Haematologica, Volume 103, Issue 4 by Haematologica

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

Ancient Greek

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

Scientific Latin

aÂma [haima] = blood a·matow [haimatos] = of blood lÒgow [logos]= reasoning

Scientific Latin

haematologicus (adjective) = related to blood

Modern English

haematologica (adjective, plural and neuter, used as a noun) = hematological subjects The oldest hematology journal, publishing the newest research results. 2016 JCR impact factor = 7.702

Haematologica, as the journal of the European Hematology Association (EHA), aims not only to serve the scientific community, but also to promote European cultural identify.

haematologica Journal of the European Hematology Association Published by the Ferrata Storti Foundation Editor-in-Chief Luca Malcovati (Pavia)

Managing Director Antonio Majocchi (Pavia)

Associate Editors Omar I. Abdel-Wahab (New York), Hélène Cavé (Paris), Simon Mendez-Ferrer (Cambridge), Pavan Reddy (Ann Arbor), Andreas Rosenwald (Wuerzburg), Monika Engelhardt (Freiburg), Davide Rossi (Bellinzona), Jacob Rowe (Haifa, Jerusalem), Wyndham Wilson (Bethesda), Paul Kyrle (Vienna), Swee Lay Thein (Bethesda), Pieter Sonneveld (Rotterdam)

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

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

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

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

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

Information for readers, authors and subscribers Haematologica (print edition, pISSN 0390-6078, eISSN 1592-8721) publishes peer-reviewed papers on all areas of experimental and clinical hematology. The journal is owned by a non-profit organization, the Ferrata Storti Foundation, and serves the scientific community following the recommendations of the World Association of Medical Editors (www.wame.org) and the International Committee of Medical Journal Editors (www.icmje.org). Haematologica publishes editorials, research articles, review articles, guideline articles and letters. Manuscripts should be prepared according to our guidelines (www.haematologica.org/information-for-authors), and the Uniform Requirements for Manuscripts Submitted to Biomedical Journals, prepared by the International Committee of Medical Journal Editors (www.icmje.org). Manuscripts should be submitted online at http://www.haematologica.org/. Conflict of interests. According to the International Committee of Medical Journal Editors (http://www.icmje.org/#conflicts), “Public trust in the peer review process and the credibility of published articles depend in part on how well conflict of interest is handled during writing, peer review, and editorial decision making”. The ad hoc journal’s policy is reported in detail online (www.haematologica.org/content/policies). Transfer of Copyright and Permission to Reproduce Parts of Published Papers. Authors will grant copyright of their articles to the Ferrata Storti Foundation. No formal permission will be required to reproduce parts (tables or illustrations) of published papers, provided the source is quoted appropriately and reproduction has no commercial intent. Reproductions with commercial intent will require written permission and payment of royalties. Detailed information about subscriptions is available online at www.haematologica.org. Haematologica is an open access journal. Access to the online journal is free. Use of the Haematologica App (available on the App Store and on Google Play) is free. For subscriptions to the printed issue of the journal, please contact: Haematologica Office, via Giuseppe Belli 4, 27100 Pavia, Italy (phone +39.0382.27129, fax +39.0382.394705, E-mail: info@haematologica.org). Rates of the International edition for the year 2018 are as following: Print edition

Institutional Euro 600

Personal Euro 150

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

haematologica calendar of events

Journal of the European Hematology Association Published by the Ferrata Storti Foundation Emirates Society of Haematology (ESH) Chairs: AS Al Olama, KM Belhoul, RM Seliem, I Mirza April 12-14, 2018 Abu Dhabi, United Arab Emirates EHA-SWG Scientific Meeting on New Molecular Insights and Innovative Management Approaches for Acute Lymphoblastic Leukemia Chair: N Gökbuget April 12-14, 2018 Barcelona, Spain 1st European Myeloma Network Meeting Società Italiana di Ematologia (SIE) Chair: M Boccadoro April 19-21, 2018 Torino, Italy EHA-TSH Hematology Tutorial on Acute Leukemias April 28-29, 2018 Istanbul, Turkey The 4th World Congress on Controversies in Multiple Myeloma Chairs: M Mohty, A Nagler, T Facon May 3-5, 2018 Paris, France

EHA Hematology Tutorial on Thalassemia May 10-11, 2018 Shiraz, Iran 23rd Congress of EHA June 14-17, 2018 Stockholm, Sweden EHA-SAH Hematology Tutorial on lymphoid Malignancies and Plasma Cell Dyscrasias September 14-15, 2018 Buenos Aires, Argentina EHA-SWG Scientific Meeting on Aging and Hematology Chair: D Bron October 12-14, 2018 Location TBC Location TBC

Calendar of Events updated on March 12, 2018

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

Table of Contents Volume 103, Issue 4: April 2018 Cover Figure

Hypercellular bone marrow from a patient with multiple myeloma showing a massive infiltrate of large cells with low nucleus-cytoplasmic ratio, central or eccentric nucleus with loose chromatin and prominent nucleoli, deeply basophilic cytoplasm. Courtesy of Prof. Rosangela Invernizzi.

Editorials 559

Does being overweight contribute to longer survival rates in myelodysplastic syndrome? Eric Solary and Michaela Fontenay

561

CD83 in Hodgkin lymphoma Ralf KĂźppers

563

Bendamustine plus rituximab in chronic lymphocytic leukemia: is there life in the old dog yet? Clemens-Martin Wendtner

Review Article 565

The many faces of IKZF1 in B-cell precursor acute lymphoblastic leukemia RenĂŠ Marke et al.

Articles Rec Cell Biology & its Disorders

575

New pathogenic mechanisms induced by germline erythropoietin receptor mutations in primary erythrocytosis Florence Pasquier et al.

Bone Marrow Failure

587

Involvement of hepcidin in iron metabolism dysregulation in Gaucher disease Thibaud Lefebvre et al.

Myelodysplastic Syndrome

597

Leptin-deficient obesity prolongs survival in a murine model of myelodysplastic syndrome Michael J. Kraakman et al.

Myeloproliferative Disorders

607

Non-adherence to treatment with cytoreductive and/or antithrombotic drugs is frequent and associated with an increased risk of complications in patients with polycythemia vera or essential thrombocythemia (OUEST study) Ronan Le Calloch et al.

Acute Myeloid Leukemia

614

Gfi1b: a key player in the genesis and maintenance of acute myeloid leukemia and myelodysplastic syndrome Aniththa Thivakaran et al.

626

Association of mutations with morphological dysplasia in de novo acute myeloid leukemia without 2016 WHO Classification-defined cytogenetic abnormalities Olga K. Weinberg et al.

Acute Lymphoblastic Leukemia

634

Dynamic clonal progression in xenografts of acute lymphoblastic leukemia with intrachromosomal amplification of chromosome 21 Paul. B. Sinclair et al.

Haematologica 2018; vol. 103 no. 4 - April 2018 http://www.haematologica.org/

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

645

Prevalence and characteristics of metabolic syndrome in adults from the French childhood leukemia survivorsâ&#x20AC;&#x2122; cohort: a comparison with controls from the French population Claire Oudin et al.

Hodgkin Lymphoma

655

CD83 is a new potential biomarker and therapeutic target for Hodgkin lymphoma Ziduo Li et al.

Non-Hodgkin Lymphoma

666

Mixed-species RNAseq analysis of human lymphoma cells adhering to mouse stromal cells identifies a core gene set that is also differentially expressed in the lymph node microenvironment of mantle cell lymphoma and chronic lymphocytic leukemia patients Gustav Arvidsson et al.

679

Histone modifier gene mutations in peripheral T-cell lymphoma not otherwise specified Meng-Meng Ji et al.

Chronic Lymphocytic Leukemia

688

Tumor necrosis factor receptor signaling is a driver of chronic lymphocytic leukemia that can be therapeutically targeted by the flavonoid wogonin Claudia DĂźrr et al.

698

Rituximab plus bendamustine or chlorambucil for chronic lymphocytic leukemia: primary analysis of the randomized, open-label MABLE study Anne-Sophie Michallet et al.

Plasma Cell Disorders

707

Modeling multiple myeloma-bone marrow interactions and response to drugs in a 3D surrogate microenvironment Daniela Belloni et al.

Stem Cell Transplantation

717

Tocilizumab, tacrolimus and methotrexate for the prevention of acute graft-versus-host disease: low incidence of lower gastrointestinal tract disease William R. Drobyski et al.

Hemostasis

728

Macrophage scavenger receptor SR-AI contributes to the clearance of von Willebrand factor Nikolett Wohner et al.

Coagulation & its Disorders

738

A three-year prospective study of the presentation and clinical outcomes of major bleeding episodes associated with oral anticoagulant use in the UK (ORANGE study) Laura Green et al.

Letters to the Editor Letters are available online only at www.haematologica.org/content/103/4.toc

e141

Levels of the erythropoietin-responsive hormone erythroferrone in mice and humans with chronic kidney disease Mark R. Hanudel et al. http://www.haematologica.org/content/103/4/e141

Haematologica 2018; vol. 103 no. 4 - April 2018 http://www.haematologica.org/

haematologica Journal of the European Hematology Association Published by the Ferrata Storti Foundation e143

Emergence and evolution of TP53 mutations are key features of disease progression in myelodysplastic patients with lower-risk del(5q) treated with lenalidomide Laurence LodĂŠ et al. http://www.haematologica.org/content/103/4/e143

e147

Trametinib inhibits RAS-mutant MLL-rearranged acute lymphoblastic leukemia at specific niche sites and reduces ERK phosphorylation in vivo Mark Kerstjens et al. http://www.haematologica.org/content/103/4/e147

e151

Impact of histological grading on survival in the SWOG S0016 follicular lymphoma cohort Lisa M. Rimsza, et al. http://www.haematologica.org/content/103/4/e151

e154

The prohibitin-binding compound fluorizoline induces apoptosis in chronic lymphocytic leukemia cells ex vivo but fails to prevent leukemia development in a murine model Marina Wierz et al. http://www.haematologica.org/content/103/4/e154

e158

No improvement in long-term survival over time for chronic lymphocytic leukemia patients in stereotyped subsets #1 and #2 treated with chemo(immuno)therapy Panagiotis Baliakas et al. http://www.haematologica.org/content/103/4/e158

e162

Cytogenetic aberrations in multiple myeloma are associated with shifts in serum immunoglobulin isotypes distribution and levels Pankaj Yadav et al. http://www.haematologica.org/content/103/4/e162

e165

Rapid hematologic responses improve outcomes in patients with very advanced (stage IIIb) cardiac immunoglobulin light chain amyloidosis Richa Manwani et al. http://www.haematologica.org/content/103/4/e165

e169

The use of romiplostim in treating chemotherapy-induced thrombocytopenia in patients with solid tumors Hanny Al-Samkari et al. http://www.haematologica.org/content/103/4/e169

Case Reports Case reports are available online only at www.haematologica.org/content/103/4.toc

e173

Lethal neonatal bone marrow failure syndrome with multiple congenital abnormalities, including limb defects, due to a constitutional deletion of 3â&#x20AC;&#x2122; MECOM Lars T. van der Veken et al. http://www.haematologica.org/content/103/4/e173

e177

Single-agent dabrafenib for BRAFV600E-mutated histiocytosis Ankush Bhatia et al. http://www.haematologica.org/content/103/4/e177

e181

Use of thrombin generation assay to personalize treatment of breakthrough bleeds in a patient with hemophilia and inhibitors receiving prophylaxis with emicizumab Yesim Dargaud et al. http://www.haematologica.org/content/103/4/e181

Haematologica 2018; vol. 103 no. 4 - April 2018 http://www.haematologica.org/

EDITORIALS Does being overweight contribute to longer survival rates in myelodysplastic syndrome? Eric Solary1,2 and Michaela Fontenay3,4 1

Université Paris-Sud, Faculté de Médecine, Le Kremlin-Bicêtre; 2INSERM U1170, Gustave Roussy, Villejuif; 3Université Paris-Descartes, Faculté de Médecine, Paris and 4Institut Cochin, INSERM U1016, CNRS UMR 8104, Paris, France E-mail: eric.solary@gustaveroussy.fr doi:10.3324/haematol.2018.188854

eing overweight or obese, defined as having an abnormal or excessive accumulation of body fat, have been associated with an increased risk of cancer development in the bone marrow and other tissues.1 Through its metabolic, endocrine and inflammatory consequences, obesity could have either an inductive or a selective effect on a malignant clone.2 While the effect of a patient being overweight or obese on cancer prevalence appears to be clear, the impact of excess body fat on patient survival at the time of cancer diagnosis is less well-defined. In this issue of Haematologica, Kraakman et al. demonstrate that in a mouse model of myelodysplastic syndrome (MDS) obesity improves survival in the absence of treatment, and propose some biological explanations to this survival advantage.3 Based on their previous results,4 Kraakman and colleagues began their study by applying the hypothesis that, through generating an inflammatory setting that includes overproduction of interleukin-1 (IL)-1β, obesity may promote the progression of MDS to acute myeloid leukemia (AML) and decrease survival. They tested this hypothesis in Ob/Ob mice in which the Lep gene was mutated.5 Leptin is an adipocyte-released cytokine/adipokine that regulates food intake, and its mutation leads to rapid weight gain that, importantly, persists following bone marrow transplantation procedures.3 Ob/Ob animals were engrafted with marrow from NHD13 transgenic mice expressing the NUX98HOXD3 fusion protein, which induces an MDS phenotype that can subsequently evolve into acute leukemia.6 Seven months post-transplantation of NHD13 bone marrow cells, Ob/Ob mice and their lean counterparts all remained alive and demonstrated a defective bone marrow hematopoiesis. These animals subsequently showed macrocytic anemia, severe lymphocytopenia, decreased platelet count, and splenomegaly. In addition to these background abnormalities, obesity was associated with a stronger increase in the fraction of circulating monocytes and a specific shift of hematopoiesis from the bone marrow to the spleen. Unexpectedly, follow up of these animals showed that Ob/Ob mice lived, on average, 100 days longer with MDS than lean animals with the same disease, despite the fact that a complete bone marrow failure had been observed in both genetic settings. The prevalence of secondary AML was similar in Ob/Ob and lean animals, however, the exacerbated monocytosis which developed in the obese animals mimicked chronic myelomonocytic leukemia, whereas the lean MDS animals developed more T-cell acute lymphoblastic leukemias. A striking difference between Ob/Ob and lean animals was the lower loss of body fat in Ob/Ob mice developing MDS. In addition, morphological analysis of visceral adipose tissue detected a massive infiltration of activated CD11b+ myeloid cells and CD11c+ pro-inflammatory macrophages haematologica | 2018; 103(4)

surrounding remodeled adipocytes in obese MDS mice (Figure 1). Conversely, infiltration of the spleen and the liver by myeloid cells was significantly higher in lean MDS animals. The proposed hypothesis is that the expanded adipose tissue acts as a sink for myeloid cells, which spares other organs, such as the liver, from myeloid cell infiltration and functional degradation. This model stresses the complexity of the relationship between obesity and cancer. In 2014, an estimated 640 million adults worldwide were obese, and the obesity-related cancer burden was estimated as being between 3% and 4% of the entire cancer burden.1 The risk is generally higher in women compared with men, reaching 9% among women in countries, including North America, Europe, and the Middle East, where the body mass index (BMI) is the highest.7 Regarding hematological malignancies, the report from the International Agency for Research on Cancer (IARC) Working Group1 emphasized a positive association between increased BMI and the risk of multiple myeloma, with relative risks ranging from 1.2 for overweight to 1.5 for severe obesity. This report did not comment on other hematopoietic malignancies, including that of MDS, but a significant association between increased BMI and the risk of MDS had been suggested by previous independent studies.8-10 Obesity is a chronic and complex pathological state that affects bone marrow homeostasis. Increased fat mass changes the composition of the bone marrow niche, either

Figure 1. Ob/Ob animals and their lean littermates engrafted with marrow from transgenic mice expressing the NUX98-HOXD3 fusion protein develop an MDS phenotype that can secondarily evolve into acute leukemia. Increased survival of Ob/Ob animals correlates with the retention of CD11b+ myeloid cells in the adipose tissue. According to the hypothesis proposed Kraakman et al., this retention could protect other organs, such as the liver, from myeloid cell infiltration and functional degradation.

559

Editorials

directly, by modifying the composition of local adipose tissue,11,12 or indirectly, through diet-induced modification of the gut microbiota.13 In turn, these alterations of the niche disrupt the hematopoietic stem cell compartment, e.g., through deregulating the transcription factor Gfi1,14 promoting myeloid skewing and overproduction of monocytes and neutrophils.4 A link has recently been established between saturated fatty acids that accumulate in the serum of obese people - the fatty acid binding protein FABP4 to which they bind, which is highly expressed in leukemic cells - and a cellular pathway that leads to DNA hypermethylation and fuels AML cell growth, suggesting innovative therapeutic strategies in this disease.15 A pertinent question is whether and how overweight and obesity affect the progression of established disease. The murine model described in this issue of Haematologica3 suggests some effect of obesity on the natural evolution of the disease. Seven months after the engraftment of NHD3 bone marrow cells, Ob/Ob mice share a largely common phenotype with their lean counterparts, including decreased mature blood cell counts and bone marrow progenitors; Ob/Ob mice also demonstrate an increase in circulating monocytes and splenic hematopoiesis, which may reflect the signals induced by the obese inflammatory state as suggested by the accumulation of Ly6Chigh monocyte subsets.4 This deregulated myelopoiesis, which occurs at an intermediate time point in disease evolution, contrasts with the dramatic increase in the overall survival of obese animals. At the time of animal sacrifice, the adipose tissue of Ob/Ob mice engrafted with NHD3 cells had recruited more myeloid cells, including CD11b+ myeloid cells and macrophages, that, by contrast, were less present in the spleen and the liver. However, it remains unclear whether, and how, this distinct repartition of myeloid cells affects disease evolution and promotes monocyte accumulation rather than acute leukemia evolution. The authors did not evaluate the impact of obesity on disease response to treatment. As surprising as it may sound, the retrospective analysis of 1,974 AML patients enrolled in SWOG studies had identified an increased response rate to a chemotherapeutic regimen in overweight and obese patients.16 Howbeit, the opposite observation was made in a cohort of childhood AML,17 which could be related to the demonstrated ability of adipocytes to sequester chemotherapeutic drugs.18 The influence of BMI on MDS and AML therapeutic response therefore deserves further investigation. As acknowledged by Kraakman et al.,3 a limitation of their study is the use of Ob/Ob mice in which the Lep gene is disrupted. Leptin levels are elevated in overweight individuals in which this pro-inflammatory adipokine was shown to affect the behavior of tumor cells and their microenvironment. A proliferative and anti-apoptotic effect of leptin has also been depicted on AML blast cells.19 Therefore, the absence of leptin in the tested model may alter the natural history of the disease in an overweight setting, which demands validation in another model of obesity in which leptin secretion is maintained. The demonstration that improved survival in MDS animals is related to the absence of leptin would foster the

560

therapeutic development of leptin antagonists, including leptin analogs and antibodies targeting leptin or its transmembrane receptor.20 The manuscript by Kraakman et al. points to a counterintuitive, and thus thrilling hypothesis of a protective effect of obesity on the progression of installed MDS, and suggests a series of future investigations in order to validate this premise, explore the cellular and molecular mechanisms involved, and determine if this protective effect also applies to the therapeutic response.

References 1. Lauby-Secretan B, Scoccianti C, Loomis D, Grosse Y, Bianchini F, Straif K; International Agency for Research on Cancer Handbook Working Group. Body fatness and cancer - viewpoint of the IARC Working Group. N Engl J Med. 2016;375(8):794-798. 2. Lichtman MA. Obesity and neoplasms of lymphohematopoietic cells. Blood Adv. 2016;1(1):101-103. 3. Kraakman MJ, Kammoun HL, Dragoljevic D, et al. Leptin-deficient obesity prolongs survival in a murine model of myelodysplastic syndrome. Haematologica. 2018;103(4)597-606. 4. Nagareddy PR, Kraakman M, Masters SL, et al. Adipose tissue macrophages promote myelopoiesis and monocytosis in obesity. Cell Metab. 2014;19(5):821-835. 5. Pelleymounter MA, Cullen MJ, Baker MB, et al. Effects of the obese gene product on body weight regulation in ob/ob mice. Science. 1995;269(5223):540-543. 6. Lin YW, Slape C, Zhang Z, Aplan PD. NUP98-HOXD13 transgenic mice develop a highly penetrant, severe myelodysplastic syndrome that progresses to acute leukemia. Blood. 2005;106(1):287-295. 7. Arnold M, Leitzmann M, Freisling H, et al. Obesity and cancer: an update of the global impact. Cancer Epidemiol. 2016;41:8-15. 8. Ma X, Lim U, Park Y, et al. Obesity, lifestyle factors, and risk of myelodysplastic syndromes in a large US cohort. Am J Epidemiol. 2009;169(12):1492-1499. 9. Murphy F, Kroll ME, Pirie K, Reeves G, Green J, Beral V. Body size in relation to incidence of subtypes of haematological malignancy in the prospective Million Women Study. Br J Cancer. 2013;108(11):2390-2398. 10. Poynter JN, Richardson M, Blair CK, et al. Obesity over the life course and risk of acute myeloid leukemia and myelodysplastic syndromes. Cancer Epidemiol. 2016;40:134-140. 11. Adler BJ, Kaushansky K, Rubin CT.Obesity-driven disruption of haematopoiesis and the bone marrow niche. Nat Rev Endocrinol. 2014 ;10(12):737-748. 12. Naveiras O, Nardi V, Wenzel PL, Hauschka PV, Fahey F, Daley GQ. Bone-marrow adipocytes as negative regulators of the haematopoietic microenvironment. Nature. 2009;460(7252):259-263. 13. Luo Y, Chen GL, Hannemann N, et al. Microbiota from obese mice regulate hematopoietic stem cell differentiation by altering the bone niche. Cell Metab. 2015;22(5):886-894. 14. Lee JM, Govindarajah V, Goddard B, et al. Obesity alters the longterm fitness of the hematopoietic stem cell compartment through modulation of Gfi1 expression. J Exp Med. 2018;215(2) :627-644. 15. Yan F, Shen N, Pang JX, et al. A vicious loop of fatty acid-binding protein 4 and DNA methyltransferase 1 promotes acute myeloid leukemia and acts as a therapeutic target. Leukemia. 2017 Oct 10. [Epub ahead of print] 16. Medeiros BC, Othus M, Estey EH, Fang M, Appelbaum FR. Impact of body-mass index on the outcome of adult patients with acute myeloid leukemia. Haematologica. 2012;97(9):1401-1404. 17. Orgel E, Tucci J, Alhushki W, et al. Obesity is associated with residual leukemia following induction therapy for childhood B-precursor acute lymphoblastic leukemia. Blood. 2014;124(26):3932-3938. 18. Sheng X, Parmentier JH, Tucci J, et al. Adipocytes sequester and metabolize the chemotherapeutic daunorubicin. Mol Cancer Res. 2017;15(12):1704-1713. 19. Konopleva M, Mikhail A, Estrov Z, et al. Expression and function of leptin receptor isoforms in myeloid leukemia and myelodysplastic syndromes: proliferative and anti-apoptotic activities. Blood. 1999;93(5):1668-1676. 20. Ray A, Cleary MP. The potential role of leptin in tumor invasion and metastasis. Cytokine Growth Factor Rev. 2017;38:80-97.

haematologica | 2018; 103(4)

Editorials

CD83 in Hodgkin lymphoma Ralf KĂźppers Institute of Cell Biology (Cancer Research), University of Duisburg-Essen, Medical Faculty, Germany E-mail: ralf.kueppers@uk-essen.de doi:10.3324/haematol.2018.188870

he development of combined chemotherapy with or without radiotherapy for classical Hodgkin lymphoma (HL) can be considered as a major success story in onocology. With current treatment protocols, a long-term cure is obtained in about 80-90% of patients.1 However, these therapies come with considerable toxicity and a risk for the development of secondary cancers, which is particularly problematic not only for non-fit elderly patients, but also for young adult HL patients. Hence, there is currently much concentrated effort being put into the development of a more targeted and less toxic therapy. One very promising approach is the use of a toxin-coupled anti-CD30 antibody, brentuximab vedotin, which directly targets the Hodgkin and Reed-Sternberg (HRS) tumor cells in HL, as they consistently express high levels of CD30.1 A second targeted therapy with exciting results from clinical studies involves antibodies against programmed cell death1 (PD-1) or programmed cell death ligand 1 (PD-L1).1 PDL1 is expressed by HRS cells and inhibits PD-1-expressing activated T cells as a means of immune evasion.2 In this issue of Haematologica, Li and colleagues focus on CD83 as a further potentially attractive candidate, both as a biomarker and target for therapy.3 CD83 is a membrane glycoprotein belonging to the immunoglobulin superfamily. It is frequently used as a general marker for dendritic cells, but it is also expressed by some other cell types, including a fraction of B cells and T cells.4 CD83 is also released from cells, and the sol-

uble form (sCD83) is even detectable at a low concentration in the serum of healthy individuals. This release seems to be predominantly mediated by proteolytic cleavage from membrane-anchored CD83, but may also involve differential splicing to produce a secreted form. Until recently there was no indication for the CD83 ligand(s).5 However, a number of studies have since revealed numerous immunosuppressive functions of sCD83.4-6 The expression of CD83 by HRS cells was already described more than 20 years ago by Hart and colleagues.7 A more recent study confirmed the frequent expression of CD83 by HRS cells, and showed that this can serve as a valuable marker to distinguish classical HL from anaplastic lymphoma kinase (ALK)-negative anaplastic large cell lymphoma, which can be a difficult differential diagnosis.8 As CD83 was initially considered to be a dendritic cell marker, its expression by HRS cells was originally interpreted as a hint for a dendritic cell origin of HRS cells,7 but we now know that CD83 is also specifically expressed by centrocytes, the non-proliferating subset of germinal center B cells.9 Hence, although HRS cells, which are derived from germinal center B cells,10 have largely lost their B cell typical gene expression pattern,11,12 expression of CD83 by HRS cells in the majority of cases of HL may reflect their germinal center B-cell origin. The retained expression of this marker by HRS cells may indicate that it is of selective advantage for HRS cells to keep it expressed and not to downregulate it;

Figure 1. Features of CD83 in HL and potential clinical applications involving CD83. CD83 is expressed on HRS cells in most cases of classical HL, which can be used for differential diagnosis. Soluble CD83 (sCD83) is also released from HRS cells. sCD83 levels in serum may serve as a biomarker for disease load. sCD83 has immunosuppressive functions when binding to its still poorly characterized ligands on target cells. CD83 can also be transfered to other cells in the HL microenvironment by trogocytosis, a process in which membrane fragments are transfered from one cell to another. This process may also be immunosuppressive, as it causes upregulation of PD-1 on T cells, likely rendering them more responsive to inhibiting signals from PD-L1 on HRS cells. Finally, a toxin-coupled monoclonal antibody against human CD83 has been developed, which efficiently kills HL cell line cells. This antibody needs to be further tested for its suitability for targeted therapy. HRS: Hodgkin and Reed-Sternberg; PD-1: programmed cell death-1.

haematologica | 2018; 103(4)

561

Editorials

most other B-cell markers are downregulated by HRS cells. In their study, Li and collagues studied several aspects of CD83 in HL (Figure 1).3 First, they validated that over 80% of cases of classical HL show CD83 expression by HRS cells, and they report that in the positive cases the fraction of CD83+ HRS cells varies from 10% to more than 90%. Second, they show that in vitro, CD83 can be transfered from HL cell line cells to T cells by trogocytosis. Trogocytosis is a process in which fragments of cell membranes are transfered from one cell to another.13 Interestingly, the trogocytosed CD4+ T cells that acquired CD83 though this process upregulated PD-1 expression, and consequently may become further suppressed in their activity against HRS cells. However, it remains unclear whether this is due to CD83 or other consequences of acquisition of membrane fragments from HRS cells. Third, extending upon an earlier observation by Hart and colleagues that a HL cell line releases sCD83,14 the group now shows that this is also a feature of other HL cell lines, and that sCD83 contributes to the inhibition of the proliferation of stimulated T cells. This became evident from the observation that the inhibitory effect of supernatants from HL cell line cultures on T-cell proliferation was partially abolished when sCD83 in the supernatants was captured by an anti-CD83 antibody. Fourth, serum levels of sCD83, measured by the enzyme-linked immunosorbent assay (ELISA), correlated with clinical response; this, however, requires more detailed analyses, as only six patients were studied in the work by Li and colleagues. Fifth, a human anti-human anti-CD83 antibody (3C12C) was tested for its suitability to target HRS cells. Whereas the unconjugated antibody had variable cytotoxic effetcs when tested on three HL cell lines, toxicity became more pronounced and consistent when the antibody was coupled to monomethyl auristatin E (MMAE). The unconjugated form showed no general toxicity when applied to non-human primates, baboons. However, B-cell numbers where reduced in the animals, a fact from which the authors conclude that 3C12C has a targeted effect, as a fraction of B cells in baboons expresses CD83. Thus, 3C12C-MMAE should be further modified and tested as a novel targeted treatment option for HL patients. The multifaceted study by Derek Hart and his team addresses a multitude of aspects about the biology of CD83 in HL and its clinical implications. This will, hopefully, stimulate more investigative work on this interesting topic. Regarding the application of the anti-CD83 antibody for the treatment of HL patients, a number of critical questions need to be addressed in future studies. The presence of sCD83 in HL patients may pose a major restriction for a successful therapy by capturing antiCD83 antibodies, which may cause difficulty in obtaining high enough concentrations of free antibody in the lymph nodes to attack the HRS cells. It is a possibility that the capturing of sCD83 may be part of an efficient therapy, as

562

sCD83 has immunosuppressive features, therefore reducing its concentration may work synergistically with the direct killing of HRS cells. A further caveat is that 3C12CMMAE may also eliminate mature dendritic cells, which are CD83+, and thereby impair normal immune responses in patients. Neverheless, positive therapeutic effects may predominate also in this regard if dendritic cells in the HL microenvironment, which are considered to contribute to the complex immune evasion strategies in HL,2,15 are efficiently eliminated by the antibody-drug conjugate treatment. Moreover, although administration of 3C12C had no toxic effects on baboons, it remains to be clarified what the off-target toxicity of the toxin-coupled form of the antibody is, this being that which one would like to use in therapy. Finally, as most cases of HL express CD83 at levels detectable by immunohistochemistry only on a fraction of HRS cells (whereas CD30, the target of brentuximab vedotin, is expressed on virtually all HRS cells), it remains to be clarified how efficiently the HRS cell clone is eliminated when exposed to the anti-CD83 antibody toxin conjugate.

References 1. Eichenauer DA, Engert A. The evolving role of targeted drugs in the treatment of Hodgkin lymphoma. Expert Rev Hematol. 2017;10(9):775-782. 2. Wein F, Küppers R. The role of T cells in the microenvironment of Hodgkin lymphoma. J Leukoc Biol. 2016;99(1):45-50. 3. Li Z, Ju X, Lee K, et al. CD83 is a new potential biomarker and therapeutic target for Hodgkin lymphoma. Haematologica. 2018;103(4):655-665. 4. Fujimoto Y, Tedder TF. CD83: a regulatory molecule of the immune system with great potential for therapeutic application. J Med Dent Sci. 2006;53(2):85-91. 5. Horvatinovich JM, Grogan EW, Norris M, et al. Soluble CD83 inhibits T cell activation by binding to the TLR4/MD-2 complex on CD14+ monocytes. J Immunol. 2017;198(6):2286-2301. 6. Breloer M, Fleischer B. CD83 regulates lymphocyte maturation, activation and homeostasis. Trends Immunol. 2008;29(4):186-194. 7. Sorg UR, Morse TM, Patton WN, et al. Hodgkin's cells express CD83, a dendritic cell lineage associated antigen. Pathology. 1997;29(3):294-299. 8. Döring C, Hansmann ML, Agostinelli C, et al. A novel immunohistochemical classifier to distinguish Hodgkin lymphoma from ALK anaplastic large cell lymphoma. Mod Pathol. 2014;27(10):1345-1354. 9. Victora GD, Dominguez-Sola D, Holmes AB, Deroubaix S, DallaFavera R, Nussenzweig MC. Identification of human germinal center light and dark zone cells and their relationship to human B-cell lymphomas. Blood. 2012;120(11):2240-2248. 10. Küppers R, Engert A, Hansmann M-L. Hodgkin lymphoma. J Clin Invest. 2012;122(10):3439-3447. 11. Schwering I, Bräuninger A, Klein U, et al. Loss of the B-lineage-specific gene expression program in Hodgkin and Reed-Sternberg cells of Hodgkin lymphoma. Blood. 2003;101(4):1505-1512. 12. Tiacci E, Döring C, Brune V, et al. Analyzing primary Hodgkin and Reed-Sternberg cells to capture the molecular and cellular pathogenesis of classical Hodgkin lymphoma. Blood. 2012;120(23):4609-4620. 13. Dhainaut M, Moser M. Regulation of immune reactivity by intercellular transfer. Front Immunol. 2014;5:112. 14. Hock BD, Kato M, McKenzie JL, Hart DN. A soluble form of CD83 is released from activated dendritic cells and B lymphocytes, and is detectable in normal human sera. Int Immunol. 2001;13(7):959-967. 15. Wein F, Weniger MA, Hoing B, et al. Complex immune evasion strategies in classical Hodgkin lymphoma. Cancer Immunol Res. 2017;5(12):1122-1132.

haematologica | 2018; 103(4)

Editorials

Bendamustine plus rituximab in chronic lymphocytic leukemia: is there life in the old dog yet? Clemens-Martin Wendtner University of Cologne and Klinikum Schwabing, Academic Teaching Hospital of the University of Munich, Germany E-mail: clemens.wendtner@uni-koeln.de doi:10.3324/haematol.2018.188862

he results of the multicenter international MABLE trial published in this issue of Haematologica represent the first randomized phase III data comparing two frequently used chemoimmunotherapies, chlorambucil plus rituximab (Clb-R) versus bendamustine plus rituximab (BR), in patients with chronic lymphocytic leukemia (CLL) and concomitant comorbidities.1 Michallet and colleagues report a significant benefit for BR with respect to complete response (CR) rate, progression-free survival (PFS) and minimal residual disease (MRD) negativity rate, while the safety profiles of both chemoimmunotherapies were quite similar. A dominant inclusion criterion was that patients had to be ineligible for treatment with fludarabine due to comorbidities. Altogether, 357 patients were randomized while 241 of them were treatment-naive. In this frontline population the primary endpoint of the trial was met, i.e., higher CR rates after 6 cycles of treatment in favor of BR (24% vs. 9%, respectively). As secondary endpoints, PFS (30 months vs. 20 months, respectively) and the MRD negativity rate (66% vs. 36%, respectively) were also significantly superior in the BR arm, whereas overall response rate (ORR) and overall survival (OS) did not differ between arms. While the differences in efficacy endpoints between both treatment arms are quite impressive, some critical points need further reflection. The entire MABLE trial contains a quite heterogeneous and not well-defined group of CLL patients, i.e., first-line and second-line patients, while the aforementioned analysis carried out by Michallet et al. focuses only on the subgroup of treatment-naive CLL patients. The recruitment of second-line patients had been stopped by an amendment of the trial, caused by slow accrual. Furthermore, comorbidity is not well-characterized as an inclusion criterion for this trial, and seems rather subjective when given the discriminator “fludarabine-ineligibility”. It would have been desirable if a comorbidity scoring, based, for example, on the cumulative illness rating scale (CIRS), as has been used in similar trials (e.g., COMPLEMENT-1, CLL11, etc.), had also been applied for the MABLE trial population.2,3 In addition, the dose of bendamustine that was chosen for firstline use (90mg/m2) attests to a reasonably fit patient population (70 mg/m2 is the typical standard for unfit patients) and makes a comparison with the COMPLEMENT-1 trial (Clb vs. Clb plus ofatumumab) or the CLL11 trial (Clb vs. Clb-R vs. Clb plus obinutuzumab [Obi]) difficult. In the initially reported first-line data for BR at a dose of 90 mg/m2 based on a phase II trial (CLL2M trial) of the German CLL Study Group (GCLLSG), only a minority of patients was comorbid and/or over 70 years of age.4 In a phase III trial comparing BR against fludarabine, cyclophosphamide, and rituximab (FCR) in fitter patients, it was shown that the 90 haematologica | 2018; 103(4)

mg/m2 bendamustine dose was quite toxic in patients over the age of 65, inducing severe infections of grade III and IV in more than 20% of patients.5 Therefore, an international consensus panel recommended a lower dose of bendamustine (70 mg/m2) in elderly patients.6 Excepting the question of the adequately-dosed chemotherapy backbone, there is a need to discuss whether the antiCD20 monoclonal antibody (mAb) rituximab should still be the standard-of-care for unfit CLL patients nowadays. Within the CLL11 trial of the GCLLSG a combination of chlorambucil plus the type II mAb obinutuzumab has been shown to be superior to the doublet of Clb-R, at least with respect to PFS. Besides the problematic comparison of MABLE data to other chemoimmunotherapy trials focusing on a less fit CLL population, there must be a critical discussion as to whether the question regarding which chemoimmunotherapy is superior as a frontline approach in CLL patients is still relevant nowadays, given the fact that many other therapeutic options have become available over the last few years. We have learned from the RESONATE-II trial that the Bruton’s tyrosine kinase (BTK) inhibitor, ibrutinib, is very effective as a first-line therapy in CLL patients, regardless of the risk factors, e.g., an unmutated IGHV status.7 Although a direct comparison of ibrutinib monotherapy to a chemoimmunotherapy standard, such as Clb-R or BR, is lacking thus far, it is difficult to imagine that one of the main players of the MABLE trial could beat ibrutinib, which induces durable responses in the frontline setting (median PFS not reached after a median observation time of 18.4 months). However, data from the ILLUMINATE trial performed by the UK CLL Study Group, which is comparing chlorambucil plus obinutuzumab versus ibrutinib plus obinutuzumab, are pending. In addition, the U.S. based trial A041202 (Alliance) will answer the question of whether BR still has a role compared to ibrutinib or ibrutinib plus

Table 1. Efficacy of different first-line treatment options in less fit/elderly CLL patients.

Clb-R (MABLE)1 BR (MABLE)1 Clb-Obi (CLL11)3 Ibrutinib (RESONATE-II)7 Venetoclax-Obi (CLL14)8

Median age (yrs)

ORR (%)

CR (%)

MRD negativity PFS (%) (months)

72 72 74

75 74 77

9 24 22

13 41 38

30 40 27

73 75

86 100

4 58

0 92

NR NR

ORR: overall response rate; CR: complete response; MRD: minimal residual disease; PFS: progression-free survival; Clb: chlorambucil; R: rituximab; BR: bendamustine plus rituximab; Obi: obinutuzumab; NR: not reached.

563

Editorials

rituximab. In addition to the potential role of ibrutinib as a broad future first-line standard in CLL, we have to be aware that the BCL2 inhibitor, venetoclax, will further challenge any chemoimmunotherapy standard, both alone and in combination with antibodies, or even as a doublet including ibrutinib. The initial data regarding venetoclax in combination with the type II anti-CD20 mAb, obinutuzumab, seem to be very promising, with a MRD negativity rate of 100% in peripheral blood, based on a small run-in cohort as part of a large randomized phase III trial of the GCLLSG (CLL14 trial), which is correlating this combination with that of chlorambucil plus obinutuzumab (Table 1).8 Taken together, and in spite of some limitations in the trial design and trial performance, the data from MABLE are important as they represent the only available phase III data of the BR combination regimen compared to a chlorambucil/anti-CD20 comparator arm. The data from MABLE provide important cognizance with regard to the frontline treatment portfolio, keeping in mind that in many countries worldwide, upcoming treatment options based on B-cell receptor inhibitors, like ibrutinib, or BCL2 inhibitors such as venetoclax, will not be an available or affordable option in the near future.9 On account of the MABLE data, treating physicians will now have a rationale to use BR as an alternative treatment option compared to chlorambucil-based regimens in patients who are not eligible for more aggressive, fludarabine-based therapies. In other words, with respect to BR, and for the time being, there seems to be life in the old dog yet.

564

References 1. Michallet A-S, Aktan M, Hiddemann W, et al. Rituximab plus bendamustine or chlorambucil for chronic lymphocytic leukemia: primary analysis of the randomized, open-label MABLE study. Haematologica 2018;103(4):698-706. 2. Hillmen P, Robak T, Janssens A, et al. Chlorambucil plus ofatumumab versus chlorambucil alone in previously untreated patients with chronic lymphocytic leukaemia (COMPLEMENT 1): a randomised, multicentre, open-label phase 3 trial. Lancet. 2015;385(9980):18731883. 3. 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):1101-1110. 4. Fischer K, Cramer P, Busch R, et al. Bendamustine in combination with rituximab for previously untreated patients with chronic lymphocytic leukemia: a multicenter phase II trial of the German Chronic Lymphocytic Leukemia Study Group. J Clin Oncol. 2012;30(26):3209-3216. 5. Eichhorst B, Fink AM, Busch R, et al. Frontline chemoimmunotherapy with fludarabine (F), cyclophosphamide (C), and rituximab (R) (FCR) shows superior efficacy in comparison to bendamustine (B) and rituximab (BR) in previously untreated and physically fit patients (pts) with advanced chronic lymphocytic leukemia (CLL): final analysis of an international, randomized study of the German CLL Study Group (GCLLSG) (CLL10 Study). Blood. 2014;124(21):19. 6. Cheson BD, Brugger W, Damaj G, et al. Optimal use of bendamustine in hematologic disorders: Treatment recommendations from an international consensus panel - an update. Leuk Lymphoma. 2016;57(4):766-82. 7. Burger JA, Tedeschi A, Barr PM, et al. Ibrutinib as initial therapy for patients with chronic lymphocytic leukemia. N Engl J Med. 2015;373(25):2425-2437. 8. Fischer K, Al-Sawaf O, Fink AM, et al. Venetoclax and obinutuzumab in chronic lymphocytic leukemia. Blood. 2017 May 11;129(19):2702-2705. 9. Chen Q, Jain N, Ayer T, et al. Economic burden of chronic lymphocytic leukemia in the era of oral targeted therapies in the United States. J Clin Oncol. 2017;35(2):166-174.

haematologica | 2018; 103(4)

REVIEW ARTICLE

The many faces of IKZF1 in B-cell precursor acute lymphoblastic leukemia

Ferrata Storti Foundation

RenĂŠ Marke,1 Frank N. van Leeuwen1 and Blanca Scheijen1,2

1 Laboratory of Pediatric Oncology, Radboud University Medical Center, and 2Department of Pathology, Radboud University Medical Center; Radboud Institute for Molecular Life Sciences (RIMLS), Nijmegen, the Netherlands

ABSTRACT

ranscription factor IKZF1 (IKAROS) acts as a critical regulator of lymphoid differentiation and is frequently deleted or mutated in Bcell precursor acute lymphoblastic leukemia. IKZF1 gene defects are associated with inferior treatment outcome in both childhood and adult B-cell precursor acute lymphoblastic leukemia and occur in more than 70% of BCR-ABL1-positive and BCR-ABL1-like cases of acute lymphoblastic leukemia. Over the past few years, much has been learned about the tumor suppressive function of IKZF1 during leukemia development and the molecular pathways that relate to its impact on treatment outcome. In this review, we provide a concise overview on the role of IKZF1 during normal lymphopoiesis and the pathways that contribute to leukemia pathogenesis as a consequence of altered IKZF1 function. Furthermore, we discuss different mechanisms by which IKZF1 alterations impose therapy resistance on leukemic cells, including enhanced cell adhesion and modulation of glucocorticoid response.

Introduction B-cell precursor acute lymphoblastic leukemia (BCP-ALL) is the most common malignancy in children and involves uncontrolled expansion of B-lymphoid progenitors in the bone marrow. The disease is frequently initiated by a chromosomal translocation but becomes manifest only when leukemic progenitors in the bone marrow have accumulated a number of additional gene deletions and mutations that drive disease progression. With current treatment protocols long-term survival approaches 90%;1 however, relapses still pose a significant clinical challenge due to resistance to chemotherapy of the recurrent disease.1 Both in pediatric and adult BCP-ALL, specific genetic subtypes with distinct prognostic outcomes can be identified2 Some of these subtypes, such as hyperdiploid ALL and ETV6-RUNX1rearranged ALL are associated with a favorable outcome, while other genetic hallmarks, such as MLL gene rearrangements, hypodiploidy, intrachromosomal translocation of chromosome 21 (iAMP21), or the presence of the t(9;22) BCR-ABL1 translocation predict poor outcome. Moreover, the presence of a gene expression profile similar to that of BCR-ABL1-positive ALL, which frequently involves genetic alterations that deregulate cytokine receptor and/or tyrosine kinase signaling, is similarly associated with poor outcome.2 In addition to these gross chromosomal rearrangements, deletions or mutations affecting the B-cell transcription factor IKZF1, are a strong and independent predictor of poor outcome in BCP-ALL.3,4 Together with its role as a critical regulator of B-cell development and a leukemia tumor suppressor, there is mounting evidence that IKZF1 loss also affects signaling pathways that modulate therapy response. Here, we provide an overview of the complex role of transcription factor IKZF1 during normal lymphopoiesis and consequences of IKZF1 loss for the pathogenesis of BCP-ALL. Finally, we discuss some of the molecular mechanisms by which IKZF1 gene alterations may contribute to therapy resistance.

Transcription regulation by IKAROS zinc-finger protein 1

Haematologica 2018 Volume 103(4):565-574

Correspondence: blanca.scheijen@radboudumc.nl

Received: November 29, 2017. Accepted: February 12, 2018. Pre-published: March 8, 2018. doi:10.3324/haematol.2017.185603 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/565 ÂŠ2018 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.

The IKAROS family of transcription factors consists of five different IKAROS zinc-finger proteins (IKZF1-IKZF5) that are able to bind DNA directly at the core haematologica | 2018; 103(4)

565

R. Marke et al.

motif A/GGGAA through their N-terminal zinc-finger domain.5,6 Furthermore, all IKAROS family members harbor two additional C-terminal zinc-fingers required for homo- and heterodimerization between the different IKZF proteins (Figure 1A). The formation of homo- or heterodimers between IKAROS zinc-finger proteins with a functional DNA binding domain strongly enhances their DNA affinity and transcriptional activity. However, a common feature of IKZF1 and related family members is the presence of shorter variants due to alternative splicing. These variants often lack DNA binding activity but retain the ability to interact with full-length IKZF1-IKZF5, thereby creating dominant-negative isoforms. A well-known splice variant of both the mouse and human IKZF1 gene is the IK6 isoform, which lacks exons 4 to 7 that encode the four N-terminal zinc-fingers representing the DNA binding domain (Figure 1B). IKZF1 mainly regulates gene expression through association with the nucleosome remodeling and deacetylase complex,7-10 which includes histone deacetylases HDAC1, HDAC2 and the ATP-dependent chromatin remodeling proteins CHD3 and CHD4. The nucleosome remodeling and deacetylase complex is involved in both transcriptional repression as well as gene activation by IKZF1.11,12 Gene silencing by IKZF1 is also facilitated through interaction

with Polycomb repressive complex 2, which promotes histone H3 lysine 27 trimethylation to maintain genes in an inactive state.13,14 Other transcriptional co-factors that can associate with IKZF1 and mediate gene regulation include CtBP, CtIP and SWI/SNF-related complex.15-17 On the other hand, IKZF1 may itself participate in transcription initiation through direct interactions with the general transcription factors TFIIB and TBP.16 IKZF1 also controls transcription elongation via association with protein phosphatase 1Îą and cyclin-dependent kinase 9 (CDK9), the enzymatic component of the positive transcription elongation factor b.18-20 IKZF1-mediated transfer of protein phosphatase 1Îą to CDK9 promotes activation of positive transcription elongation factor b and recruitment to gene regulatory regions, thereby facilitating transcription elongation of IKZF1-target genes in hematopoietic cells.18 Distinct post-translational modifications are able to modify the function of IKZF1. Phosphorylation of IKZF1 at multiple serine and threonine residues by casein kinase II impairs its function as a transcription factor.20-22 Conversely, casein kinase II inhibition enhances the transcriptional repressor function of IKZF1.23 On the other hand, dual-specificity kinases BTK and SYK both phosphorylate IKZF1 on specific serine residues in close proximity of the DNA binding domain to augment its nuclear

Figure 1. Overview of the human family of IKAROS zinc-finger (IKZF) transcription factors and IKZF1 isoforms. (A) Schematic representation of the five IKZF proteins (IKZF1-IKZF5), including the N-terminal zinc-fingers that define the DNA-binding domain and the two C-terminal zinc-fingers representing the dimerization domain. The colored boxes indicate the individual regions within the protein that are encoded by distinct exons. (B) The common IKZF1 splice variants (IK1-IK8) are shown, including the shorter isoforms that are generated by alternative splicing. The splice variants lacking exons 4 and 5 (IK6-IK8) represent dominant-negative isoforms of IKZF1.

566

haematologica | 2018; 103(4)

IKZF1 in leukemia and therapy response

localization and DNA binding activity.24,25 Sumoylation of IKZF1 on lysine residues occurs within the nucleus and seems to interfere with transcriptional repression.26,27 It was previously shown that IKZF1 is also subject to ubiquitination,20 and there is now renewed interest in this pathway, since both IKZF1 and IKZF3 are targets of the immunomodulatory drugs thalidomide, lenalidomide, pomalidomide and CC-122.28 These immunomodulatory drugs promote proteosomal degradation of IKZF1 and IKZF3 by redirecting the substrate specificity of the CRL4CRBN ubiquitin ligase complex.29,30 Immunomodulatory drugs show therapeutic effects in a broad range of hematologic malignancies through their ability to target the malignant cells and modulate the immune system and its microenvironment.

IKZF1 is essential for normal lymphopoiesis Studies performed in both constitutive and conditional Ikzf1 knockout mouse models have demonstrated that IKZF1 function is not only required at different stages of lymphopoiesis,12,31,32 but also for normal myeloid, megakaryocyte and erythroid differentiation.33-36 Ikzf1deficient mice (Ikzf1null/null) lack all B cells, natural killer cells, plasmacytoid dendritic cells and fetal T cells31,37 (Figure 2). Nonetheless, post-natal Ikzf1-null mice harbor early T lineage progenitors within the thymus and export mature T

cells to the periphery.38 Mice homozygous mutant for a hypomorphic allele of Ikzf1 (Ikzf1L/L) show reduced B-cell progenitors in the bone marrow compartment, but still generate normal counts of mature B2 cells.39 These splenic B cells display alterations in isotype selection during immunoglobulin class switch recombination and a hyperproliferation phenotype upon antigenic stimulation.40,41 Although spontaneous progression to B-cell ALL is not observed in Ikzf1L/L mice, haplodeficient Ikzf1L/+ animals demonstrate an accelerated onset of B-cell leukemia in combination with a BCR-ABL1 transgene.42 Moreover, all Ikzf1L/L mice develop thymic lymphoma within a period of 10 months through activation of the Notch pathway.43 Ikzf1 mutant mice expressing dominant-negative isoforms of IKZF1 (Ikzf1DN/DN and Ikzf1Plstc/Plstc) demonstrate a widespread failure of hematopoiesis,44,45 highlighting the importance of IKAROS transcription factors in hemato-lymphoid differentiation. Notably, heterozygous Ikzf1 mutant mice develop T-cell malignancies with very high penetrance and short latency in the case of the dominant-negative isoforms,46,47 while this phenotype is less obvious in Ikzf1+/- mice.48 Detailed gene expression profiling has revealed that IKZF1 is essential for the generation of common lymphoid progenitors by priming lymphoid lineage-specific signatures in hematopoietic stem cells and lymphoid-primed

Figure 2. Summary of the observed phenotypes in the different constitutive Ikzf1 knockout mouse models. The knockout allele shows a schematic representation at which position the deletion or mutation is present in the mouse Ikzf1 gene. DN: dominant negative; Plstc: ENU-induced dominant-negative point mutation, called Plastic; Neo: neomycin gene; Î˛Geo: fusion between LacZ and neomycin gene; ZF: zinc-finger; HSC: hematopoietic stem cell; NK: natural killer; pDCs: plasmacytoid dendritic cells; LT-HSC: long-term hematopoietic stem cell; GMPs: granulocyte-macrophage precursors; mo: months.

haematologica | 2018; 103(4)

567

R. Marke et al.

multipotent progenitors.49 At different stages of T-lineage differentiation and development, IKZF1 is engaged by setting thresholds for (pre-)T-cell receptor-controlled checkpoints as well as T-cell activation downstream of interleukin-2 receptor signaling.50,51 In B-cell progenitors, Ikzf1 is required to induce Rag1 and Rag2 expression, and mediates chromatin accessibility during immunoglobulin gene rearrangement and allelic exclusion at the Igk locus.12,32,52 During pre-B-cell differentiation, IKZF1 regulates the transcription of genes implicated in pre-B-cell receptor signaling, cell survival, stromal-cell adhesion and B-cell commitment, such as Pax5, Foxo1 and Ebf1.12,32,53 Many of those regulatory activities during B-lineage differentiation are navigated by super-enhancer networks controlled by IKZF1 and other B-cell master transcription factors.54 Besides regulating expression of B-lymphoid genes, IKZF1 is actively involved in repression of a lineage-inappropriate transcriptional program normally prevalent in epithelial and mesenchymal precursors.54 To further delineate the function of the individual zincfingers within the DNA-binding domain of IKZF1 in Blymphopoiesis, Ikzf1 mouse mutants have been generated with targeted deletion of exon 4, which encodes zinc-finger 1 (Ikzf1ΔF1/ΔF1), or exon 6 encoding zinc-finger 4 (Ikzf1ΔF4/ΔF4).37 Germline deletion of either exon 4 or 6 results in decreased B-cell precursors with a stronger developmental block in Ikzf1ΔF1/ΔF1 mice, especially at the pre-B-cell stage.37 In contrast, the fraction of large pre-B cells is strongly increased in Ikzf1ΔF4/ΔF4 mice as compared to wild-type control animals. Interestingly, deletion of zinc-finger 4, but not zinc-finger 1, accelerates the onset of BCR-ABL1-mediated B-cell leukemia.37,55 Conditional deletion of exon 5 (Ikzf1E5Δ/Δ), which encodes zinc-fingers 2 and 3, at the stage of common lymphoid progenitors also results in an expansion of large pre-B cells within the bone marrow compartment, which is followed by a subsequent block in the transition to small pre-B cells.56 These findings indicate that Nterminal zinc-fingers 2, 3 and 4 of IKZF1 limit cell proliferation and survival at the time of active pre-B-cell receptor signaling, while zinc-fingers 1, 2 and 3 are absolutely required for the transition to the pre-B-cell stage.

IKZF1 gene lesions drive leukemia development and relapse In the past decade, complementary genome-wide approaches have been employed to identify the genetic drivers implicated in the pathogenesis of ALL. Those studies revealed that the IKZF1 gene, which is located on chromosome band 7p12.2, is recurrently affected by different types of genetic alterations in BCP-ALL. Analysis of copy number alterations has demonstrated that IKZF1 gene deletions are present in about 15% of cases of childhood BCP-ALL and 40%-50% of adult patients with BCPALL.57-60 These deletions frequently involve the whole gene (DEL1-8) that results in loss of expression of wildtype IKZF1, as well as focal deletions that alter the function of IKZF1, such as the dominant-negative isoform IK6 (DEL4-7). Other common variants include deletions affecting exons 2-3, exons 2-7 and exons 4-8.61 In most cases these are monoallelic IKZF1 deletions where one functional copy of IKZF1 is retained, although biallelic deletions are also observed in a fraction of BCP-ALL cases.62,63 In addition, IKZF1 function is compromised by insertions, frameshift and missense mutations, which represent ~7% of IKZF1 alterations in BCP-ALL.63 568

Furthermore, rare in-frame gene fusions involving IKZF1 have been identified by RNA sequencing in BCP-ALL, including IKZF1-NUTM1, IKZF1-SETD5 and the reciprocal SETD5-IKZF1.64 However, it remains to be established whether these IKZF1 gene fusions are pathogenic and contribute to leukemia development. An interesting feature is the strongly increased prevalence of IKZF1 deletions and mutations in high-risk BCPALL cases with an activated tyrosine kinase profile, particularly BCR-ABL1-positive ALL (~85%),65 and BCR-ABL1like ALL (~70%), which is characterized by a range of genetic alterations driving cytokine receptor and kinase signaling.3,66-68 Similarly, IKZF1 deletions and mutations are highly abundant in chronic myeloid leukemia that has progressed to lymphoid blast crises, but IKZF1 alterations are virtually absent in chronic-phase and myeloid blast crisis chronic myeloid leukemia.65,69,70 IKZF1 deletions are also rarely detected in ETV6-RUNX1-positive BCP-ALL (3%), TCF3-rearranged (~3%) and MLL-rearranged (~5%) B-cell ALL.58,71,72 The distribution of IKZF1 deletions among the remaining subtypes, including hyperdiploid and Bother leukemia, ranges from 15%-20%.72 IKZF1 acts as a critical tumor suppressor in mouse Tlymphoid malignancies,43,46,47 but IKZF1 gene lesions are not very prevalent in T-ALL. Copy number alterations and mutations affecting the IKZF1 gene can be detected in ~4% of T-ALL.58,65,71,73 Notably, IKZF1 alterations occur in ~13% of early T-cell precursor ALL, a high-risk subtype of T-ALL characterized by recurrent mutations activating tyrosine kinases (FLT3, JAK1, JAK3) and cytokine signaling (IL7R).74 IKZF1 alterations have also been reported in myeloproliferative neoplasms,75 and both pediatric and adult acute myeloid leukemia harbor IKZF1 deletions that affect its function.76,77 Thus, the tumor suppressive activity of IKZF1 is not uniquely restricted towards the lymphoid lineage and extends to a broader range of hematologic malignancies. Besides its critical role in the pathogenesis of leukemia, IKZF1 alterations are also associated with adverse prognosis in BCP-ALL.3,4,78 even within the high-risk group of BCR-ABL1-positive ALL.79,80 Notably, the occurrence and prognostic impact of IKZF1 alterations is not restricted to high-risk cases, but is also observed in standard-risk B-ALL subtypes,72 including high hyperploidy.81 Indeed, IKZF1 deletion represents one of the strongest independent predictors of poor treatment outcome in childhood BCPALL.71,72,82 Similar data have been reported in adult BCPALL, where loss-of-function gene deletions of IKZF1 predict poor treatment outcome in BCR-ABL1-negative cases.83-86 Interestingly, the presence of other co-occurring gene lesions may either enhance or negate the prognostic value of IKZF1 deletions. For instance, focal deletions affecting both transcriptional regulator BTG1 and IKZF1 represent a high-risk group with a worse outcome than those with IKZF1 alterations alone.48 On the other hand, the BCP-ALL subtype characterized by deregulation of transcription factors ERG and DUX4 has a favorable outcome, despite the presence of IKZF1 deletions in approximately 40% of these patients.64,87-90 An explanation for this latter observation remains elusive.

Genetic alterations that cooperate with IKZF1 deletions in B-cell precursor acute lymphoblastic leukemia There is accumulating evidence that recurrent chromosomal aberrations present in BCP-ALL, such as BCR-ABL1 haematologica | 2018; 103(4)

IKZF1 in leukemia and therapy response

translocations or CRLF2 rearrangements, act as driver lesions and represent early events in leukemia development. Genome-wide analysis has established that several other genetic alterations cooperate before B-cell leukemia becomes manifest. Gene lesions that inactivate the lymphoid transcription factor IKZF1 are frequently observed in BCR-ABL1-positive and CRLF2-rearranged BCPALL.65,69,91,92 The latter group is associated with concomitant JAK1 and JAK2 activating mutations.91 Similarly, IKZF1 alterations are highly prevalent in tyrosine kinaseactivating lesions that define BCR-ABL1-like ALL.67,68 These include rearrangements involving ABL1/ABL2, CSF1R, EPOR, JAK2 and PDGFRB, or sequence mutations affecting FLT3, IL7R or SH2B3. Indeed, loss of IKZF1 may permit more effective STAT5 target gene regulation downstream of these pathways.93 Collectively, these findings argue that loss of IKZF1 function strongly cooperates with activated tyrosine kinase signaling pathways linked to enhanced progenitor B-cell proliferation and immortalization (Figure 3). The predilection for IKZF1 gene alterations in BCRABL1-mediated lymphoid versus myeloid malignancies has been further corroborated in mouse studies. In a bone

marrow transplantation model using lineage-negative hematopoietic progenitor cells, it was shown that expression of IK6 skews BCR-ABL1-mediated leukemia from an exclusive myeloproliferative disease towards a combined myeloid and B-lymphoid disease.63 Introducing p19Arf-deficiency further strengthens this trend towards uniformly induced B-cell ALL. This is in agreement with the finding that BCR-ABL1-positive BCP-ALL is characterized by the co-occurrence of IKZF1 and CDKN2A gene deletions.65 Another group of genetic changes that frequently cooccur with IKZF1 alterations in BCP-ALL include gene deletions affecting lymphoid transcription factors, such as EBF1 and PAX5, and the transcriptional co-factor BTG148,58 (Figure 3). BTG1 belongs to the BTG/TOB antiproliferative (APRO) family of proteins,94 which control gene transcription by their ability to interact with specific transcription factors, such as nuclear receptors and homeobox proteins.95,96 the CCR4-NOT transcriptional regulatory complex,97 or through recruitment of protein arginine methyl transferase PRMT1.98 In addition, BTG1 through interaction with the CCR4-NOT, may also regulate mRNA deadenylation and consequently mRNA decay.99,100 Mice deficient for Btg1 show a partial block in B-cell development,

Figure 3. Pathways cooperating with IKZF1 alterations in leukemia pathogenesis. Pathways involving cytokine receptor signaling and B-cell differentiation by lymphoid transcriptional regulators in normal progenitor B cells are schematically indicated on the left. Alterations of these pathways co-occur frequently with IKZF1 deletions and mutations in B-cell progenitor acute lymphoblastic leukemia (BCP-ALL) as indicated on the right. These include, activating mutations in FLT3, IL7R, JAK2 (*), upregulation of CRLF2 (+), C-terminal truncations or upregulation of EPOR (^), chromosomal translocations generating fusion proteins with PDGFR or CSF1R (-), and BCR-ABL1, which collectively results in activated cytokine receptor and tyrosine kinase signaling leading to STAT activation. In addition, IKZF1 alterations cooccur with gene deletions affecting the activity of B-lymphoid transcriptional regulators EBF1, PAX5 and BTG1, which results in a block of B-cell differentiation. FLT3: FMS related tyrosine kinase 3; IL7R: interleukin 7 receptor; CRLF2: cytokine receptor like factor 2; C-KIT: mast/stem cell growth factor receptor Kit; JAK, Janus kinase; STAT: signal transducer and activator of transcription; BTG1: B-cell translocation gene 1; EBF1: early B-cell factor 1; PAX5: paired box 5; IKZF1: IKAROS family zinc finger 1; CSF1R: colony-stimulating factor 1 receptor; EPOR: erythropoietin receptor; PDGFR: platelet-derived growth factor receptor.

haematologica | 2018; 103(4)

569

R. Marke et al.

which is even more evident in Btg1-/-;Btg2-/- mice.101 These studies have demonstrated that BTG1, together with BTG2, is required to suppress a T-lineage inappropriate expression program in progenitor B cells. Thus, monoallelic gene deletions of IKZF1 in combination with EBF1, PAX5 or BTG1 may contribute to a more prominent block in B-cell development and increased proliferative expansion of precursor B cells. Indeed, intercrossing haplodeficient Ikzf1 animals with heterozygous Ebf1 or Pax5 knockout mice promotes the onset of ALL, giving rise to both BALL and T-ALL.102 On the other hand, Btg1-deficiency specifically accelerates the development of T-ALL in Ikzf1+/- mice, which suggests that B-lineage-restricted mouse models will be required to establish their synergistic action in the pathogenesis of B-ALL.

Effector pathways downstream of IKZF1 involved in leukemia pathogenesis Since lymphoid transcription factors are commonly deleted in BCP-ALL, the tumor suppressive functions of IKZF1 and other B-cell master regulators, such as EBF1 and PAX5, have been mostly linked to the suppression of their B-cell differentiation programs in these leukemic cells. However, this would not fully explain the predilection of IKZF1 alterations in BCR-ABL1-positive and BCR-ABL1like leukemia, suggesting that IKZF1 also regulates other molecular pathways. Furthermore, loss of IKZF1 function probably affects different target genes in human leukemic cells as compared to mouse progenitor B cells, which could even be distinct from those deregulated by expression of dominant-negative isoforms, such as IK6. Nonetheless, mouse studies performed over the past 5 years have been very instrumental in deciphering the transcriptional networks downstream of IKZF1. Thus, gene expression profiling in different Ikzf1 knockout mouse models combined with genome-wide chromatin immunoprecipitation studies has uncovered IKZF1-specific targets that are not only linked to lymphoid lineage commitment and B-cell differentiation, but also to leukemia development. A large group of those Ikzf1-target genes can be classified as signal transducers, some of which drive early lymphoid differentiation, such as c-Kit, Flt3 and Il7r.12,32,37,53 Adult ALL samples harboring IKZF1 deletions display increased expression of IL7R together with reduced expression of SH2B3, which represents a defined subset of high-risk B-ALL.103 Other genes differentially expressed in Ikzf1-mutant mice are important for pre-B-cell receptor signaling, and several of these IKZF1 targets appear to be deregulated in BCR-ABL1-positive B-ALL, including IGLL1, SYK, and SLP65.104,105 Indeed, defective pre-B-cell receptor function is a hallmark of BCR-ABL1-positive ALL, and loss of IKZF1 function enhances SRC phosphorylation at the expense of the SYK/SLP65 pathway activation, which is required for pre-B-cell differentiation.104 Besides transcriptional regulation of signal transducers, Ikzf1 controls the expression of cell surface receptors, such as CD34 and CD43, and these molecules confer a leukemic growth advantage to IKZF1-mutated BCR-ABL1-positive B-ALL cells.55 Another group of IKZF1 target genes identified in mouse progenitor B cells seems to converge on a cellular network coupling cell surface protein expression with intracellular Wnt and Rho signaling as well as catenin-driven gene regulation inside the nucleus.55,106 A critical target 570

gene within this subgroup includes Ctnnd1 encoding p120catenin. This is a multifunctional protein that regulates cadherin stability at the cell membrane, activation of the Rho family of GTPases in the cytoplasm and Wnt/βcatenin target genes within the nucleus by interacting with Kaiso.107 Activation of CTNND1 expression is observed in samples from patients with IKZF1 deletions,108 and inactivation of p120-catenin reduces the proliferative capacity of BCR-ABL1-positive leukemic cells.55,106 A related downstream effector pathway of IKZF1 that plays an eminent role during mouse B-cell development is integrindependent survival signaling, which involves activation of focal adhesion kinase (FAK).12,56 In mouse models of BCRABL1-positive B-ALL, perturbation of Ikzf1, including lossof-function deletions and expression of IK6, leads to activation of an adhesive phenotype, which correlates with overexpression of FAK.63,109 FAK pathway upregulation is also observed in BCR-ABL1-positive BCP-ALL, especially in the context of IK6 expression.109 Moreover, FAK inhibition potentiates the responsiveness to the ABL inhibitor dasatinib in a xenograft model system and improves survival.109 Recently, it has been proposed that the B-lymphoid transcriptional program regulated by IKZF1, as well as PAX5, acts as a metabolic barrier against malignant transformation of B-cell precursor cells.110 Inducible reconstitution of functional IKZF1 in patient-derived IKZF1-deleted B-ALL cells results in activation of the LKB1-AMPK energy-stress-sensor pathway, and decreased protein levels of the insulin receptor, the glucose transporters GLUT1, GLUT3 and GLUT6, as well as the effectors of glucose metabolism, such as HK2, HK3, and G6PD. On the other hand, the expression of glucose-transport inhibitors, such as TXNIP and CNR2, are strongly induced by IKZF1. Consequently, these IKZF1-reconstituted B-ALL cells transit into a state of chronic energy deficit. Thus, this ‘metabolic gatekeeper’ function of IKZF1 may force silent preleukemic clones that carry potentially oncogenic lesions to remain in a latent state. Besides imposing a change on pre-B-cell receptor signaling, cell adhesion and metabolic state, IKZF1 alterations in combination with BCR-ABL1 expression also result in acquisition of stem cell-like features and enhanced selfrenewal of progenitor B cells63,105 (Figure 4). Activation of THY1 expression has been linked to enhanced self-renewal,63 and Ikzf1 has been shown to regulate expression of multiple genes involved in cell cycle regulation, including Cdkn1a, Cdkn2a, and Cdk6.53,55 In mouse progenitor B cells and human B-ALL, BCL6 and MYC have been identified as IKZF1 targets,32,111-113 and probably both contribute to enhanced cell proliferation of IKZF1-deleted B-ALL. However, it remains to be established whether targeting these pathways has therapeutic potential in high-risk BALL patients.

IKZF1 alterations mediate therapy resistance The presence of IKZF1 gene lesions in BCR-ABL1-positive B-ALL results in inferior treatment outcome and mouse xenograft models suggest that IKZF1 loss contributes to resistance to tyrosine kinase inhibitor-based therapy.63,80 Reactivation of cell adhesion pathways by perturbation of IKZF1 function leads to elevation of key adhesion molecules, such as integrins (ITGA5) and CD90, and adhesion regulators, such as FAK, as well as increased phosphorylation of FAK itself, which permits relocalization of leukemic haematologica | 2018; 103(4)

IKZF1 in leukemia and therapy response

Figure 4. Effector pathways downstream of IKZF1 involved in leukemia pathogenesis. Loss of IKZF1 function due to IKZF1 gene deletions and mutations affects multiple pathways, including pre-B-cell receptor signaling, cell adhesion and proliferation, metabolic pathways and signal transducers and cell surface receptors. IKZF1 affects the expression of defined key molecules within each of these pathways, as indicated in the boxes. Green boxes define targets that are upregulated upon loss of IKZF1 function, while red boxes represent repressed targets. SRC: sarcoma proto-oncogene tyrosine kinase; SYK: spleen tyrosine kinase; IGLL1: immunoglobulin lambda-like polypeptide 1; SLP65: B-cell linker; RHO: RHO family of GTPases ; CTNND1: catenin delta 1/p120 catenin; FAK: focal adhesion kinase; ITGA5: integrin subunit alpha 5; THY1: thymus cell antigen 1; CDK6: cyclin dependent kinase 6; CDKN1A: cyclin dependent kinase inhibitor 1A; BCL6: B-cell lymphoma 6; CDKN2A: cyclin dependent kinase inhibitor 2A; c-MYC: cellular myelocytomatosis oncogene; GLUT1/3/6: glucose transporter 1/3/6; INSR: insuline receptor; HK2: hexokinase 2; HK3: hexokinase 3; AMPK: AMP-activated protein kinase; LKB1: liver kinase B1; G6PD: glucose-6-phosphate dehydrogenase; TXNIP: thioredoxin interacting protein; NR3C1: nuclear receptor subfamily 3, group C, member 1 (glucocorticoid receptor); CNR2: cannabinoid receptor 2; FLT3: FMS related tyrosine kinase 3; CD34: hematopoietic progenitor cell antigen; c-KIT: KIT receptor tyrosine kinase; CD43: sialophorin; IL7R: interleukin 7 receptor.

cells to the bone marrow niche. Indeed, FAK inhibition resensitizes BCR-ABL1 leukemic cells to tyrosine kinase inhibitor therapy.109 Similar results are observed after treatment with retinoids, specifically retinoid X receptor agonists, which induce expression of wild-type IKZF1, but not IK6, thereby abrogating expression of stem cell and adhesion molecules.63 Although these studies have provided important clues about how IKZF1 deletions alter treatment response especially in the context of BCR-ABL1-positive ALL, alternative mechanisms of therapy resistance may exist besides protection through cell interactions within the bone marrow microenvironment. Synthetic glucocorticoids, such as prednisolone, constitute essential drugs in the treatment of ALL patients and glucocorticoid resistance remains a substantial problem in the treatment of BCP-ALL. There is accumulating evidence that IKZF1 deletions mediate prednisolone resistance in vivo,114,115 but different mechanisms have been proposed. IKZF1 actively represses genes of the phosphatidylinositol-3 kinase pathway, including PIK3CD and PIK3C2B.23 Disruption of IKZF1 function, and subsequent activation of the PI3K/AKT/mTOR pathway can promote glucocorticoid resistance.116,117 IKZF1 controls expression of several genes involved in glucose and energy supply.110 This metabolic program may alter the threshold for responses to glucocorticoids in BCP-ALL. Specifically, the glucocorticoid receptor NR3C1 was reported to be a target of IKZF1 in pre-B ALL cells, and downregulation of NR3C1 protein levels could be observed upon expression of IK6.110 However, studies performed in murine Ikzf1+/- B cells and human BCP-ALL cell lines with short hairpinmediated IKZF1 knockdown have demonstrated that loss of IKZF1 function induces glucocorticoid resistance indehaematologica | 2018; 103(4)

pendently of altered NR3C1 mRNA and protein expression.114 Indeed, IKZF1 itself appears to regulate NR3C1dependent gene transcription.114 The transcriptional regulator BTG1 has been identified as a modifier of IKZF1mediated resistance to glucocorticoid therapy and the combined loss of BTG1 and IKZF1 leads to an even stronger inhibition of glucocorticoid-induced cell death.48 Finally, IKZF1 target gene EMP1,106 which itself represents a poor prognostic factor in pediatric ALL, was shown to regulate the response to prednisolone, but also, on the other hand, to affect normal leukemic cell viability and proliferation.118 Collectively, these findings demonstrate that IKZF1, through modulation of different signaling pathways and acting directly on glucocorticoid target genes, alters treatment response, thereby mediating therapy resistance in BCP-ALL (Figure 5).

Conclusions and perspectives From this review it becomes clear that loss of IKZF1 function affects a broad variety of biological pathways which may all contribute to leukemia development. Moreover, the recently established roles for IKZF1 in cell adhesion, metabolism and glucocorticoid-dependent target gene regulation seem to be important determinants of therapy resistance. Preclinical studies are helping with the identification of molecular pathways that can be exploited for targeted therapy of IKZF1-deleted BCP-ALL. Over the past decade, a large series of studies conducted in both childhood and adult ALL have provided clear evidence that IKZF1 alterations predict adverse outcome in BCP-ALL, both in BCR-ABL1-positive and -negative B571

R. Marke et al.

Figure 5. IKZF1 alterations mediate therapy resistance. Overview of IKZF1-affected pathways contributing to tyrosine kinase inhibitor (TKI) resistance and glucocorticoid (GC) resistance. Enhanced cell adhesion due to loss of IKZF1 function has been shown to contribute to both TKI and GC resistance. Deregulation of metabolic pathways, such as LKB1/AMPK signaling and glucose metabolism, attenuated glucocorticoid receptor (GR) target gene regulation and upregulation of epithelial membrane protein 1 (EMP1) have been implicated in mediating GC resistance of IKZF1-deleted BCP-ALL. Green boxes indicate activated targets or pathways, while red boxes define attenuated pathways. Targets within the metabolic pathway can either promote or inhibit GC resistance.

ALL. However, more recently the role of IKZF1 deletions as an independent prognostic marker has been challenged,119 as has the specific contributions of whole gene versus intragenic dominant-negative IKZF1 deletions.86 One potential explanation for such disparities may relate to differences in scheduling and dosing of specific therapeutic agents between different treatment protocols. It will, therefore, be important to study these protocol-dependent differences in order to define what is currently the most efficient treatment for IKZF1-deleted ALL. Certain adjustments, such as the addition of vincristine-

References 1. Hunger SP, Mullighan CG. Acute lymphoblastic Leukemia in children. N Engl J Med. 2015;373(16):1541-1552. 2. Iacobucci I, Mullighan CG. Genetic Basis of Acute Lymphoblastic Leukemia. J Clin Oncol. 2017;35(9):975-983. 3. 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. 4. 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. 5. John LB, Ward AC. The Ikaros gene family: transcriptional regulators of hematopoiesis and immunity. Mol Immunol. 2011;48(910):1272-1278. 6. Yoshida T, Georgopoulos K. Ikaros fingers on lymphocyte differentiation. Int J Hematol. 2014;100(3):220-229. 7. Kim J, Sif S, Jones B, et al. Ikaros DNA-binding proteins direct formation of chromatin remodeling complexes in lymphocytes. Immunity. 1999;10(3):345-355. 8. Koipally J, Renold A, Kim J, Georgopoulos K. Repression by Ikaros and Aiolos is mediated through histone deacetylase complexes. EMBO J. 1999;18(11):3090-3100.

572

steroid pulses during maintenance therapy,81 may already prevent relapses. For the near future, more systematic screens aimed at determining specific vulnerabilities of IKZF1-deleted ALL may lead to the identification of targeted therapies that can re-sensitize this high-risk ALL subgroup to curative treatment. Acknowledgments This work was supported by Stichting Kinderen Kankervrij (KiKa; grant numbers KiKa 2009-55 and KiKa 2010-77) and Stichting KOC Nijmegen.

9. Sridharan R, Smale ST. Predominant interaction of both Ikaros and Helios with the NuRD complex in immature thymocytes. J Biol Chem. 2007;282(41):30227-30238. 10. Liang Z, Brown KE, Carroll T, et al. A highresolution map of transcriptional repression. Elife. 2017;6. 11. Zhang J, Jackson AF, Naito T, et al. Harnessing of the nucleosome-remodelingdeacetylase complex controls lymphocyte development and prevents leukemogenesis. Nat Immunol. 2011;13(1):86-94. 12. Schwickert TA, Tagoh H, Gultekin S, et al. Stage-specific control of early B cell development by the transcription factor Ikaros. Nat Immunol. 2014;15(3):283-293. 13. Ross J, Mavoungou L, Bresnick EH, Milot E. GATA-1 utilizes Ikaros and polycomb repressive complex 2 to suppress Hes1 and to promote erythropoiesis. Mol Cell Biol. 2012;32(18):3624-3638. 14. Oravecz A, Apostolov A, Polak K, et al. Ikaros mediates gene silencing in T cells through Polycomb repressive complex 2. Nat Commun. 2015;6:8823. 15. Koipally J, Georgopoulos K. Ikaros interactions with CtBP reveal a repression mechanism that is independent of histone deacetylase activity. J Biol Chem. 2000;275(26):19594-19602. 16. Koipally J, Georgopoulos K. Ikaros-CtIP interactions do not require C-terminal bind-

17.

18.

19.

20.

21.

22.

23.

ing protein and participate in a deacetylaseindependent mode of repression. J Biol Chem. 2002;277(26):23143-23149. O'Neill DW, Schoetz SS, Lopez RA, et al. An ikaros-containing chromatin-remodeling complex in adult-type erythroid cells. Mol Cell Biol. 2000;20(20):7572-7582. Bottardi S, Mavoungou L, Pak H, et al. The IKAROS interaction with a complex including chromatin remodeling and transcription elongation activities is required for hematopoiesis. PLoS Genet. 2014;10(12): e1004827. Bottardi S, Zmiri FA, Bourgoin V, et al. Ikaros interacts with P-TEFb and cooperates with GATA-1 to enhance transcription elongation. Nucleic Acids Res. 2011;39(9):35053519. Popescu M, Gurel Z, Ronni T, et al. Ikaros stability and pericentromeric localization are regulated by protein phosphatase 1. J Biol Chem. 2009;284(20):13869-13880. Gowda C, Song C, Kapadia M, et al. Regulation of cellular proliferation in acute lymphoblastic leukemia by Casein Kinase II (CK2) and Ikaros. Adv Biol Regul. 2017;63:71-80. Gomez-del Arco P, Maki K, Georgopoulos K. Phosphorylation controls Ikaros's ability to negatively regulate the G(1)-S transition. Mol Cell Biol. 2004;24(7):2797-2807. Song C, Gowda C, Pan X, et al. Targeting

haematologica | 2018; 103(4)

IKZF1 in leukemia and therapy response

24. 25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

casein kinase II restores Ikaros tumor suppressor activity and demonstrates therapeutic efficacy in high-risk leukemia. Blood. 2015;126(15):1813-1822. Ma H, Qazi S, Ozer Z, et al. Regulatory phosphorylation of Ikaros by Bruton's tyrosine kinase. PloS one. 2013;8(8):e71302. Uckun FM, Ma H, Zhang J, et al. Serine phosphorylation by SYK is critical for nuclear localization and transcription factor function of Ikaros. Proc Natl Acad Sci USA. 2012;109(44):18072-18077. Gomez-del Arco P, Koipally J, Georgopoulos K. Ikaros SUMOylation: switching out of repression. Mol Cell Biol. 2005;25(7):26882697. Apostolov A, Litim-Mecheri I, Oravecz A, et al. Sumoylation inhibits the growth suppressive properties of Ikaros. PloS one. 2016;11(6):e0157767. Lindner S, Kronke J. The molecular mechanism of thalidomide analogs in hematologic malignancies. J Mol Med (Berl). 2016;94(12):1327-1334. Kronke J, Udeshi ND, Narla A, et al. Lenalidomide causes selective degradation of IKZF1 and IKZF3 in multiple myeloma cells. Science. 2014;343(6168):301-305. Lu G, Middleton RE, Sun H, et al. The myeloma drug lenalidomide promotes the cereblon-dependent destruction of Ikaros proteins. Science. 2014;343(6168):305-309. Wang JH, Nichogiannopoulou A, Wu L, et al. Selective defects in the development of the fetal and adult lymphoid system in mice with an Ikaros null mutation. Immunity. 1996;5(6):537-549. Heizmann B, Kastner P, Chan S. Ikaros is absolutely required for pre-B cell differentiation by attenuating IL-7 signals. J Exp Med. 2013;210(13):2823-2832. Yoshida T, Ng SY, Zuniga-Pflucker JC, Georgopoulos K. Early hematopoietic lineage restrictions directed by Ikaros. Nat Immunol. 2006;7(4):382-391. Lopez RA, Schoetz S, DeAngelis K, O'Neill D, Bank A. Multiple hematopoietic defects and delayed globin switching in Ikaros null mice. Proc Natl Acad Sci USA. 2002;99(2): 602-607. Rao KN, Smuda C, Gregory GD, Min B, Brown MA. Ikaros limits basophil development by suppressing C/EBP-alpha expression. Blood. 2013;122(15):2572-2581. Malinge S, Thiollier C, Chlon TM, et al. Ikaros inhibits megakaryopoiesis through functional interaction with GATA-1 and NOTCH signaling. Blood. 2013;121(13): 2440-2451. Schjerven H, McLaughlin J, Arenzana TL, et al. Selective regulation of lymphopoiesis and leukemogenesis by individual zinc fingers of Ikaros. Nat Immunol. 2013;14(10):10731083. Allman D, Sambandam A, Kim S, et al. Thymopoiesis independent of common lymphoid progenitors. Nat Immunol. 2003;4(2):168-174. Kirstetter P, Thomas M, Dierich A, Kastner P, Chan S. Ikaros is critical for B cell differentiation and function. Eur J Immunol. 2002;32(3):720-730. Sellars M, Reina-San-Martin B, Kastner P, Chan S. Ikaros controls isotype selection during immunoglobulin class switch recombination. J Exp Med. 2009;206(5): 10731087. Heizmann B, Sellars M, Macias-Garcia A, Chan S, Kastner P. Ikaros limits follicular B cell activation by regulating B cell receptor signaling pathways. Biochem Biophys Res Commun. 2016;470(3):714-720.

haematologica | 2018; 103(4)

42. Virely C, Moulin S, Cobaleda C, et al. Haploinsufficiency of the IKZF1 (IKAROS) tumor suppressor gene cooperates with BCR-ABL in a transgenic model of acute lymphoblastic leukemia. Leukemia. 2010;24(6):1200-1204. 43. Dumortier A, Jeannet R, Kirstetter P, et al. Notch activation is an early and critical event during T-Cell leukemogenesis in Ikaros-deficient mice. Mol Cell Biol. 2006;26(1):209-220. 44. Georgopoulos K, Bigby M, Wang JH, et al. The Ikaros gene is required for the development of all lymphoid lineages. Cell. 1994;79(1):143-156. 45. Papathanasiou P, Perkins AC, Cobb BS, et al. Widespread failure of hematolymphoid differentiation caused by a recessive niche-filling allele of the Ikaros transcription factor. Immunity. 2003;19(1):131-144. 46. Winandy S, Wu P, Georgopoulos K. A dominant mutation in the Ikaros gene leads to rapid development of leukemia and lymphoma. Cell. 1995;83(2):289-299. 47. Mantha S, Ward M, McCafferty J, et al. Activating Notch1 mutations are an early event in T-cell malignancy of Ikaros point mutant Plastic/+ mice. Leuk Res. 2007;31(3):321-327. 48. Scheijen B, Boer JM, Marke R, et al. Tumor suppressors BTG1 and IKZF1 cooperate during mouse leukemia development and increase relapse risk in B-cell precursor acute lymphoblastic leukemia patients. Haematologica. 2017;102(3):541-551. 49. Ng SY, Yoshida T, Zhang J, Georgopoulos K. Genome-wide lineage-specific transcriptional networks underscore Ikaros-dependent lymphoid priming in hematopoietic stem cells. Immunity. 2009;30(4):493-507. 50. Avitahl N, Winandy S, Friedrich C, et al. Ikaros sets thresholds for T cell activation and regulates chromosome propagation. Immunity. 1999;10(3):333-343. 51. Winandy S, Wu L, Wang JH, Georgopoulos K. Pre-T cell receptor (TCR) and TCR-controlled checkpoints in T cell differentiation are set by Ikaros. J Exp Med. 1999;190(8): 1039-1048. 52. Reynaud D, Demarco IA, Reddy KL, et al. Regulation of B cell fate commitment and immunoglobulin heavy-chain gene rearrangements by Ikaros. Nat Immunol. 2008;9(8):927-936. 53. Ferreiros-Vidal I, Carroll T, Taylor B, et al. Genome-wide identification of Ikaros targets elucidates its contribution to mouse Bcell lineage specification and pre-B-cell differentiation. Blood. 2013;121(10):1769-1782. 54. Hu Y, Zhang Z, Kashiwagi M, et al. Superenhancer reprogramming drives a Bcell-epithelial transition and high-risk leukemia. Genes Dev. 2016;30(17):19711990. 55. Schjerven H, Ayongaba EF, Aghajanirefah A, et al. Genetic analysis of Ikaros target genes and tumor suppressor function in BCRABL1+ pre-B ALL. J Exp Med. 2017;214(3): 793-814. 56. Joshi I, Yoshida T, Jena N, et al. Loss of Ikaros DNA-binding function confers integrin-dependent survival on pre-B cells and progression to acute lymphoblastic leukemia. Nat Immunol. 2014;15(3):294304. 57. Kuiper RP, Schoenmakers EF, van Reijmersdal SV, et al. High-resolution genomic profiling of childhood ALL reveals novel recurrent genetic lesions affecting pathways involved in lymphocyte differentiation and cell cycle progression. Leukemia. 2007;21(6):1258-1266.

58. Mullighan CG, Goorha S, Radtke I, et al. Genome-wide analysis of genetic alterations in acute lymphoblastic leukaemia. Nature. 2007;446(7137):758-764. 59. Paulsson K, Cazier JB, Macdougall F, et al. Microdeletions are a general feature of adult and adolescent acute lymphoblastic leukemia: Unexpected similarities with pediatric disease. Proc Natl Acad Sci USA. 2008;105(18):6708-6713. 60. Tokunaga K, Yamaguchi S, Iwanaga E, et al. High frequency of IKZF1 genetic alterations in adult patients with B-cell acute lymphoblastic leukemia. Eur J Haematol. 2013;91(3):201-208. 61. 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):3238. 62. Dupuis A, Gaub MP, Legrain M, et al. Biclonal and biallelic deletions occur in 20% of B-ALL cases with IKZF1 mutations. Leukemia. 2013;27(2):503-507. 63. 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. 64. Lilljebjorn H, Henningsson R, HyreniusWittsten A, et al. Identification of ETV6RUNX1-like and DUX4-rearranged subtypes in paediatric B-cell precursor acute lymphoblastic leukaemia. Nat Commun. 2016;7:11790. 65. Mullighan CG, Miller CB, Radtke I, et al. BCR-ABL1 lymphoblastic leukaemia is characterized by the deletion of Ikaros. Nature. 2008;453(7191):110-114. 66. Den Boer ML, van Slegtenhorst M, De Menezes RX, et al. A subtype of childhood acute lymphoblastic leukaemia with poor treatment outcome: a genome-wide classification study. Lancet Oncol. 2009;10(2): 125134. 67. Roberts KG, Morin RD, Zhang J, et al. Genetic alterations activating kinase and cytokine receptor signaling in high-risk acute lymphoblastic leukemia. Cancer Cell. 2012;22(2):153-166. 68. Roberts KG, Li Y, Payne-Turner D, et al. Targetable kinase-activating lesions in Phlike acute lymphoblastic leukemia. N Engl J Med. 2014;371(11):1005-1015. 69. Iacobucci I, Storlazzi CT, Cilloni D, et al. Identification and molecular characterization of recurrent genomic deletions on 7p12 in the IKZF1 gene in a large cohort of BCRABL1-positive acute lymphoblastic leukemia patients: on behalf of Gruppo Italiano Malattie Ematologiche dell'Adulto Acute Leukemia Working Party (GIMEMA AL WP). Blood. 2009;114(10):2159-2167. 70. Grossmann V, Kohlmann A, Zenger M, et al. A deep-sequencing study of chronic myeloid leukemia patients in blast crisis (BC-CML) detects mutations in 76.9% of cases. Leukemia. 2011;25(3):557-560. 71. Dorge 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. 72. 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. 73. Liu Y, Easton J, Shao Y, et al. The genomic landscape of pediatric and young adult T-lin-

573

R. Marke et al.

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

574

eage acute lymphoblastic leukemia. Nat Genet. 2017;49(8):1211-1218. Zhang J, Ding L, Holmfeldt L, et al. The genetic basis of early T-cell precursor acute lymphoblastic leukaemia. Nature. 2012;481(7380):157-163. Jager R, Gisslinger H, Passamonti F, et al. Deletions of the transcription factor Ikaros in myeloproliferative neoplasms. Leukemia. 2010;24(7):1290-1298. Crescenzi B, La Starza R, Romoli S, et al. Submicroscopic deletions in 5q- associated malignancies. Haematologica. 2004;89(3): 281-285. de Rooij JD, Beuling E, van den HeuvelEibrink MM, et al. Recurrent deletions of IKZF1 in pediatric acute myeloid leukemia. Haematologica. 2015;100(9):1151-1159. 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. Martinelli G, Iacobucci I, Storlazzi CT, et al. IKZF1 (Ikaros) deletions in BCR-ABL1-positive acute lymphoblastic leukemia are associated with short disease-free survival and high rate of cumulative incidence of relapse: a GIMEMA AL WP report. J Clin Oncol. 2009;27(31):5202-5207. 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. 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. Olsson L, Ivanov Ofverholm I, NorenNystrom U, et al. The clinical impact of IKZF1 deletions in paediatric B-cell precursor acute lymphoblastic leukaemia is independent of minimal residual disease stratification in Nordic Society for Paediatric Haematology and Oncology treatment protocols used between 1992 and 2013. Br J Haematol. 2015;170(6):847-858. Ribera J, Morgades M, Zamora L, et al. Prognostic significance of copy number alterations in adolescent and adult patients with precursor B acute lymphoblastic leukemia enrolled in PETHEMA protocols. Cancer. 2015;121(21):3809-3817. Yao QM, Liu KY, Gale RP, et al. Prognostic impact of IKZF1 deletion in adults with common B-cell acute lymphoblastic leukemia. BMC Cancer. 2016;16:269. Zhang W, Kuang P, Li H, Wang F, Wang Y. Prognostic significance of IKZF1 deletion in adult B cell acute lymphoblastic leukemia: a meta-analysis. Ann Hematol. 2017;96(2): 215-225. Kobitzsch B, Gokbuget N, Schwartz S, et al. Loss-of-function but not dominant-negative intragenic IKZF1 deletions are associated with an adverse prognosis in adult BCRABL-negative acute lymphoblastic leukemia. Haematologica. 2017;102(10): 1739-1747. Yasuda T, Tsuzuki S, Kawazu M, et al. Recurrent DUX4 fusions in B cell acute lymphoblastic leukemia of adolescents and young adults. Nat Genet. 2016;48(5):569574. 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. 89. Zaliova M, Zimmermannova O, Dorge P, et al. ERG deletion is associated with CD2 and attenuates the negative impact of IKZF1 deletion in childhood acute lymphoblastic leukemia. Leukemia. 2014;28(1):182-185. 90. Zhang J, McCastlain K, Yoshihara H, et al. Deregulation of DUX4 and ERG in acute lymphoblastic leukemia. Nat Genet. 2016;48(12):1481-1489. 91. Harvey RC, Mullighan CG, Chen IM, et al. Rearrangement of CRLF2 is associated with mutation of JAK kinases, alteration of IKZF1, Hispanic/Latino ethnicity, and a poor outcome in pediatric B-progenitor acute lymphoblastic leukemia. Blood. 2010;115 (26):5312-5321. 92. Russell LJ, Jones L, Enshaei A, et al. Characterisation of the genomic landscape of CRLF2-rearranged acute lymphoblastic leukemia. Genes Chromosomes Cancer. 2017;56(5):363-372. 93. Katerndahl CDS, Heltemes-Harris LM, Willette MJL, et al. Antagonism of B cell enhancer networks by STAT5 drives leukemia and poor patient survival. Nat Immunol. 2017;18(6):694-704. 94. Matsuda S, Rouault J, Magaud J, Berthet C. In search of a function for the TIS21/PC3/BTG1/TOB family. FEBS Lett. 2001;497(2-3):67-72. 95. Busson M, Carazo A, Seyer P, et al. Coactivation of nuclear receptors and myogenic factors induces the major BTG1 influence on muscle differentiation. Oncogene. 2005;24(10):1698-1710. 96. van Galen JC, Kuiper RP, van Emst L, et al. BTG1 regulates glucocorticoid receptor autoinduction in acute lymphoblastic leukemia. Blood. 2010;115(23):4810-4819. 97. Prevot D, Morel AP, Voeltzel T, et al. Relationships of the antiproliferative proteins BTG1 and BTG2 with CAF1, the human homolog of a component of the yeast CCR4 transcriptional complex: involvement in estrogen receptor alpha signaling pathway. J Biol Chem. 2001;276(13):9640-9648. 98. Lin WJ, Gary JD, Yang MC, Clarke S, Herschman HR. The mammalian immediate-early TIS21 protein and the leukemiaassociated BTG1 protein interact with a protein-arginine N-methyltransferase. J Biol Chem. 1996;271(25):15034-15044. 99. Bogdan JA, Adams-Burton C, Pedicord DL, et al. Human carbon catabolite repressor protein (CCR4)-associative factor 1: cloning, expression and characterization of its interaction with the B-cell translocation protein BTG1. Biochem J. 1998;336( Pt 2):471-481. 100. Rouault JP, Prevot D, Berthet C, et al. Interaction of BTG1 and p53-regulated BTG2 gene products with mCaf1, the murine homolog of a component of the yeast CCR4 transcriptional regulatory complex. J Biol Chem. 1998;273(35):2256322569. 101. Tijchon E, van Emst L, Yuniati L, et al. Tumor suppressors BTG1 and BTG2 regulate early mouse B-cell development. Haematologica. 2016;101(7):e272-276. 102. Farrar M, Harris LH, Kornblau S, et al. B cell transcription factors define a novel tumor suppressor gene network in acute lymphoblastic leukemia. J Immunol. 2013;190(1 Suppl):52-55. 103. Ge Z, Gu Y, Xiao L, et al. Co-existence of IL7R high and SH2B3 low expression distin-

guishes a novel high-risk acute lymphoblastic leukemia with Ikaros dysfunction. Oncotarget. 2016;7(29):46014-46027. 104. Trageser D, Iacobucci I, Nahar R, et al. Pre-B cell receptor-mediated cell cycle arrest in Philadelphia chromosome-positive acute lymphoblastic leukemia requires IKAROS function. J Exp Med. 2009;206(8):1739-1753. 105. Iacobucci I, Iraci N, Messina M, et al. IKAROS deletions dictate a unique gene expression signature in patients with adult B-cell acute lymphoblastic leukemia. PloS One. 2012;7(7):e40934. 106. Witkowski MT, Hu Y, Roberts KG, et al. Conserved IKAROS-regulated genes associated with B-progenitor acute lymphoblastic leukemia outcome. J Exp Med. 2017;214(3): 773-791. 107. Kourtidis A, Ngok SP, Anastasiadis PZ. p120 catenin: an essential regulator of cadherin stability, adhesion-induced signaling, and cancer progression. Prog Mol Biol Transl Sci. 2013;116:409-432. 108. Vitanza NA, Zaky W, Blum R, et al. Ikaros deletions in BCR-ABL-negative childhood acute lymphoblastic leukemia are associated with a distinct gene expression signature but do not result in intrinsic chemoresistance. Pediatr Blood Cancer. 2014;61(10):17791785. 109. Churchman ML, Evans K, Richmond J, et al. Synergism of FAK and tyrosine kinase inhibition in Ph+ B-ALL. JCI Insight. 2016;1(4). 110. Chan LN, Chen Z, Braas D, et al. Metabolic gatekeeper function of B-lymphoid transcription factors. Nature. 2017;542(7642): 479-483. 111. Ge Z, Guo X, Li J, et al. Clinical significance of high c-MYC and low MYCBP2 expression and their association with Ikaros dysfunction in adult acute lymphoblastic leukemia. Oncotarget. 2015;6(39):4230042311. 112. Ge Z, Zhou X, Gu Y, et al. Ikaros regulation of the BCL6/BACH2 axis and its clinical relevance in acute lymphoblastic leukemia. Oncotarget. 2017;8(5):8022-8034. 113. Ma S, Pathak S, Mandal M, et al. Ikaros and Aiolos inhibit pre-B-cell proliferation by directly suppressing c-Myc expression. Mol Cell Biol. 2010;30(17):4149-4158. 114. 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. 115. 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. 116. Piovan E, Yu J, Tosello V, et al. Direct reversal of glucocorticoid resistance by AKT inhibition in acute lymphoblastic leukemia. Cancer Cell. 2013;24(6):766-776. 117. Evangelisti C, Cappellini A, Oliveira M, et al. Phosphatidylinositol 3-kinase inhibition potentiates glucocorticoid response in B-cell acute lymphoblastic leukemia. J Cell Physiol. 2018;233(3):1796-1811. 118. Aries IM, Jerchel IS, van den Dungen RE, et al. EMP1, a novel poor prognostic factor in pediatric leukemia regulates prednisolone resistance, cell proliferation, migration and adhesion. Leukemia. 2014;28(9):1828-1837. 119. Palmi C, Valsecchi MG, Longinotti G, et al. What is the relevance of Ikaros gene deletions as a prognostic marker in pediatric Philadelphia-negative B-cell precursor acute lymphoblastic leukemia? Haematologica. 2013;98(8):1226-1231.

haematologica | 2018; 103(4)

ARTICLE

Red Cell Biology & Its Disorders

New pathogenic mechanisms induced by germline erythropoietin receptor mutations in primary erythrocytosis

Ferrata Storti Foundation

Florence Pasquier,1,2,3,4 Caroline Marty,1,2,4 Thomas Balligand,5 Frédérique Verdier,4,6 Sarah Grosjean,1,2,4 Vitalina Gryshkova,5 Hana Raslova,1,2,4 Stefan N. Constantinescu,5 Nicole Casadevall,1,7 William Vainchenker,1,2,4 Christine Bellanné-Chantelot1,8* and Isabelle Plo1,2,4*

INSERM, UMR 1170, Gustave Roussy, Laboratoire d’Excellence GR-Ex, Villejuif, France; Université Paris-Sud, UMR 1170, Gustave Roussy, Villejuif, France; 3Service d’Hématologie, Département d’Oncologie Médicale, Gustave Roussy, Villejuif, France; 4Laboratoire d’Excellence GR-Ex, Paris, France; 5Ludwig Institute for Cancer Research, and Université Catholique de Louvain, de Duve Institute, Brussels, Belgium; 6INSERM U1016, Institut Cochin, CNRS UMR8104, Université Paris Descartes, France; 7Laboratoire d'Hématologie, Hôpital Saint Antoine, Assistance Publique Hôpitaux de Paris, France and 8Département de Génétique, Hôpital Universitaire Pitié-Salpêtrière, Assistance Publique Hôpitaux de Paris, France 1 2

* CB-C and IP contributed equally to this work.

Haematologica 2018 Volume 103(4):575-586

ABSTRACT

rimary familial and congenital polycythemia is characterized by erythropoietin hypersensitivity of erythroid progenitors due to germline nonsense or frameshift mutations in the erythropoietin receptor gene. All mutations so far described lead to the truncation of the C-terminal receptor sequence that contains negative regulatory domains. Their removal is presented as sufficient to cause the erythropoietin hypersensitivity phenotype. Here we provide evidence for a new mechanism whereby the presence of novel sequences generated by frameshift mutations is required for the phenotype rather than just extensive truncation resulting from nonsense mutations. We show that the erythropoietin hypersensitivity induced by a new erythropoietin receptor mutant, p.Gln434Profs*11, could not be explained by the loss of negative signaling and of the internalization domains, but rather by the appearance of a new C-terminal tail. The latter, by increasing erythropoietin receptor dimerization, stability and cell-surface localization, causes pre-activation of erythropoietin receptor and JAK2, constitutive signaling and hypersensitivity to erythropoietin. Similar results were obtained with another mutant, p.Pro438Metfs*6, which shares the same last five amino acid residues (MDTVP) with erythropoietin receptor p.Gln434Profs*11, confirming the involvement of the new peptide sequence in the erythropoietin hypersensitivity phenotype. These results suggest a new mechanism that might be common to erythropoietin receptor frameshift mutations. In summary, we show that primary familial and congenital polycythemia is more complex than expected since distinct mechanisms are involved in the erythropoietin hypersensitivity phenotype, according to the type of erythropoietin receptor mutation.

Introduction Primary erythrocytosis (also known as primary familial and congenital polycythemia, PFCP) is a pathology of erythroid progenitors, which display hypersensitivity to erythropoietin.1-14 This rare inherited entity is usually passed down with an autosomal dominant pattern with complete penetrance.2-6,8-14 PFCP is characterized by an isolated primary polycythemia in which an increased red cell mass is associated with subnormal serum erythropoietin levels. It can therefore mimic the clinical presentation of polycythemia vera. However, the hematopoiesis is polyhaematologica | 2018; 103(4)

Correspondence: isabelle.plo@gustaveroussy.fr or christine.bellanne-chantelot@aphp.fr Received: July 12, 2017. Accepted: December 21, 2017. Pre-published: December 21, 2017.

doi:10.3324/haematol.2017.176370 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/575 ©2018 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.

575

F. Pasquier et al.

clonal and PFCP patients do not show any risk of their disease transforming into myelofibrosis or leukemia. Moreover, somatic JAK2V617F15 and JAK2 exon1216 mutations, hallmarks of polycythemia vera, are not found in PFCP,12,13 which is, however, characterized by the presence of germline mutations in the erythropoietin receptor gene (EPOR). Erythropoietin receptor (EPOR) is a type I cytokine receptor,17 essentially expressed by erythroid progenitors in hematopoietic tissue. The binding of erythropoietin to its receptor at the cell surface leads to the transient activation of preformed EPOR-JAK2 complexes18,19 and downstream signaling pathways, including signal transducer and activators of transcription (STAT),20 phosphatidylinositol 3 kinase/Akt21 and mitogen-activated protein kinase pathways. Erythropoietin-induced signaling is crucial for the proliferation and the survival of erythroid progenitors as well as for terminal erythroid differentiation.22 Around 20 germline heterozygous nonsense and frameshift mutations located in exon 8 of EPOR have been described so far in PFCP, all leading to the truncation of the C-terminal part of the receptor.3-11,13,14,23-28 Interestingly, similar EPOR truncations have been described in BCR-ABL1like acute lymphoblastic leukemia, due to acquired rearrangements of EPOR with immunoglobulin chain loci.29 The C-terminal part of the receptor includes several conserved tyrosine residues that are docking sites for positive and negative regulators of EPOR signaling. The erythropoietin hypersensitivity of PFCP progenitors is, therefore, usually explained by the disappearance of negative regulatory domains located in the truncated part of the receptor.30,31 A few missense EPOR mutations have also been described, but their involvement in the PFCP phenotype is not yet clear.4,7,13,26-28,32 Previous studies have suggested that truncated EPOR mutations might not be equivalent in term of underlying mechanisms leading to EPOR activation,14,27 but relatively few functional studies have been carried out. We, therefore, investigated the mechanism of some EPOR mutations in PFCP. We identified and extensively studied a new germline frameshift EPOR mutation, c.1300dup (p.Gln434Profs*11), responsible for marked erythropoietin hypersensitivity as in JAK2V617F-positive polycythemia vera. We modeled EPOR p.Gln434Profs*11 and several other EPOR mutants already described in the Ba/F3 and UT-7 cell line and demonstrated that different mechanisms are involved in the erythropoietin hypersensitivity phenotype, according to the type of EPOR mutation, highlighting that PFCP is a more complex pathology than usually considered.

Methods

double-positive magnetic cell sorting system (AutoMACS; Miltenyi Biotec, Paris, France). CD34+ cells were amplified in erythroid conditions for 7-10 days in Iscove modified Dulbecco medium with penicillin/streptomycin/glutamine, alpha-thioglycerol, bovine serum albumin, a mixture of sonicated lipids and insulintransferrin in the presence of recombinant human cytokines (25 ng/mL stem cell factor, 100 U/mL interleukin-3, 1 U/mL erythropoietin).

EPOR sequencing Genomic DNA was extracted from blood, nails and hair using standard procedures. Exon 8 of EPOR was amplified and sequenced in both directions. Mutations are numbered as recommended by the Human Genome Variation Society (http://www.hgvs.org/) using the reference sequence NM_000121.3. Participants gave written informed consent to the genetic study.

Quantification of clonogenic progenitors in semi-solid cultures Colony assays were performed with 500 purified CD34+ progenitors per culture dish in duplicate in H4100 Methocult media (StemCell Technologies, Grenoble, France) supplemented with 12% bovine serum albumin, 30% or no fetal bovine serum, 2-βmercaptoethanol (1 mM), 1% L-glutamine, stem cell factor (25 ng/mL), interleukin-3 (100 U/mL) and in the absence or presence of increasing concentrations of erythropoietin (0.001; 0.01; 0.1 and 1 U/mL). Burst-forming units-erythroid (BFU-E)-derived colonies were counted on day 14.

DNA manipulation and retrovirus production EPOR mutations were introduced into the pMX-HA-human EPOR WT-IRES-GFP plasmid by the QuikChange site-directed mutagenesis method using the PfuUltra high-fidelity DNA polymerase (Agilent Technologies, Stratagene, Les Ulis, France): c.1300dup (p.Gln434Profs*11, EPOR FS); c.1330G>T (p.Gln444*, EPOR STOP); c.1303_1304delinsGC (p.Leu435Ala, EPOR WTdiL or EPOR STOPdiL); c.1195G>T (p.Glu399*); c.1273G>T (p.Glu425*); c.1311_1312del (p.Pro438Metfs*6); c.1327_1329delinsTAA (p.Pro443*); c.[1300dup; 1311G>C] (p.[Gln434Profs*11; Ala437Arg]) (Table 1). Full-length EPOR mutant cDNA was verified by sequencing.

Cell lines The murine Ba/F3 and human UT-7 cells were grown in Dulbecco modified Eagle medium supplemented with 10% fetal bovine serum (StemCell Technologies, Grenoble, France) and 5% WEHI-conditioned medium as a source of murine interleukin-3 or granulocyte-macrophage colony-stimulating factor (10 ng/mL). Cells were transduced with pMX-IRES-GFP retrovirus to stably express the human wild-type EPOR or EPOR mutants, carrying a N-terminal HA-tag. GFP+ cells were subsequently sorted by flow cytometry (Influx, Beckton-Dickinson, Le Pont-de-claix, France).

Materials Human recombinant erythropoietin and interleukin-3 were generous gifts from Amgen (Neuilly, France). Stem cell factor was purchased from Biovitrum AB (Stockholm, Sweden).

Patients, cell purification and erythroblast cultures Peripheral blood samples from the patient or healthy donors were collected by leukapheresis. Written informed consent was obtained from the patient in accordance with the Declaration of Helsinki and the study was approved by the ethics committee of La Pitié-Salpétrière Hospital. Mononuclear cells were separated over a Ficoll density gradient and CD34+ cells were purified by a 576

Proliferation assays The premixed WST-1 cell proliferation assay was carried out according to the manufacturer’s instructions (Takara Bio Europe, Clontech, Saint-Germain-en-Laye, France). Experiments were performed in triplicate. Dose-response curves to erythropoietin were expressed as percentages of viability of the maximal response.

Western blot analysis For signaling studies, cells were starved for 5 h and then incubated with increasing concentrations of erythropoietin for 15 min or stimulated with 1 U/mL erythropoietin for 15 min and washed haematologica | 2018; 103(4)

Erythropoietin receptor mutations in PFCP

three times in phosphate-buffered saline 1X to remove the cytokine. Signaling studies were performed using polyclonal antibodies against the phosphorylated forms of JAK2 (Tyr 1007/1008), STAT5 (Tyr 694), ERK1/2 (Thr 202/Tyr 204) and Akt (Ser 473) and against the pan proteins (Cell Signaling Technology, Ozyme, Montigny le-Bretonneux, France). β-actin (Sigma, Saint Quentin-Fallaviers) was used as a loading control. For glycosidase digestion, cell lysates were incubated with either endoglycosidase H (Endo H) or peptide:N-glycosidase F (PNGase F) (1,000 U) at 37°C for 16 h according to the recommendations of the supplier (New England Biolabs, Evry, France).

Dimerization of human erythropoietin receptor monomers assessed by split Gaussia luciferase assay. A split Gaussia subunit, Gluc1 or Gluc2, was fused to the C-terminus of each EPOR construct.33 When an EPOR monomer with a Gluc1 fusion subunit dimerizes with another EPOR monomer with a Gluc2 fusion subunit, the two Gluc subunit proteins recombine into a catalytically active luciferase that is able to degrade coelenterazine, thus emitting light. Both EPOR constructs fused to a C-terminal Gluc1 and a Gluc2 subunit were transiently transfected at a 50:50 ratio into HEK cells using Transit-LT1 (Mirus, Euromedex, Souffelweyersheim, France) as a transfecting agent. The pGL3-control vector (Promega, Charbonnières-les-Bains, France) that constitutively expresses the firefly luciferase was cotransfected in each condition as a transfection control reporter. After 48 h, luciferase signals were read by a GloMax discovery system (Promega) after addition of coelenterazine and firefly luciferin to each well. A 530LP filter was used to discriminate the luminescence of the firefly luciferase from that of the Gaussia luciferase.

Erythropoietin receptor stability assay Ba/F3 cells were incubated for different periods of time with 50 ng/mL cycloheximide (Sigma). After cell lysis, western blotting was performed with anti-HA antibody. HA-tagged EPOR was quantified with Image J software, using β-actin as a loading control. The half-life of the receptor was calculated using GraphPad PRISM software.

Cell-surface localization of HA-tagged erythropoietin receptor by flow cytometry Cells were labeled with monoclonal mouse anti-HA antibody conjugated to phycoerythrin (Miltenyi Biotecs) and processed by flow cytometry (FACSCanto, Beckton-Dickinson, Le Pont-declaix, France).

Erythropoietin labeling and binding and erythropoietin receptor internalization studies Erythropoietin labeling using IODO-GEN (Pierce, Rockford, IL, USA), erythropoietin binding and EPOR internalization studies were performed as previously described.34-36 Nonspecific binding was determined using a 250-fold excess of unlabeled erythropoietin and was less than 5% in each case. All reported data represent specific binding. The number of cell-surface receptors was determined by comparing the radioactivity of Ba/F3-EPOR cells to the radioactivity of the reference UT-7 cell line.34 For internalization experiments, after incubation with 125Ierythropoietin, 5 x 106 cells per condition were washed twice at 4°C to remove unbound ligand. An acidic wash was then performed to separate cell surface–bound from internalized erythropoietin. Cells were incubated in 0.5 mL acidic buffer (150 mM NaCl, 50 mM sodium acetate, pH 3.5) for 3 min at 4°C. The pH was then adjusted to 7.4 using 1 M Tris-HCl, pH 9 and the cell suspension was centrifuged. The radioactivity of the supernatant (cell surface–bound erythropoietin) and of the cell pellet (internalized erythropoietin) was determined. When 125Ierythropoietin was bound to the cells at 4°C to inhibit erythropoietin internalization, more than 95% of cell-bound 125I-erythropoietin was recovered in the acidic wash supernatant using this method. Each experiment was performed three times with similar results.

Results Identification of a new germline EPOR mutation responsible for marked erythropoietin hypersensitivity Primary polycythemia was diagnosed in a 28-year old woman without a history of thrombosis. She had high hemoglobin concentration (21 g/dL) and hematocrit (60%) and an increased red cell mass (65%) with a low erythropoietin level (1.2 mU/mL; laboratory standard: 5-25 mU/mL). She had no splenomegaly at physical examination. Leukocyte and platelet counts in the peripheral blood as well as bone marrow aspiration and biopsy were strictly normal. Later no JAK2V617F or JAK2 exon12 mutation was found, rendering the diagnosis of polycythemia vera unlikely. The search for abnormal hemoglobin affinity and for VHL, PHD1/2, SH2B3 (LNK) pathological mutations was negative. EPOR sequencing identified a new germline heterozygous frameshift mutation, c.1300dup (p.Gln434Profs*11), which generates a new ten-amino acid C-terminal tail and a stop codon at position 444, lead-

Table 1. EPOR mutations investigated in this study and their functional consequences.

Mutation (DNA)

Mutation (protein)

Truncation (number of AA lost)

Number of remaining tyrosines

Reference

c.1195G>T c.1273G>T c.1311_1312del c.1327_1329delinsTAA c.1300dup

p.Glu399* p.Glu425* p.Pro438Metfs*6 p.Pro443* p.Gln434Profs*11 (EPOR FS) p.Gln444*

109 83 65 65 64

1 1 2 2 2

Arcasoy et al.11 Kralovics et al.10 Bento et al.25 This study This study

This study

c.1330G>T (EPOR STOP)

EPOR: erythropoietin receptor gene (NM_000121.3), AA: amino acids.

haematologica | 2018; 103(4)

577

F. Pasquier et al.

ing to the truncation of 64 amino acids of the wild-type receptor and the loss of the six last conserved tyrosine residues (Table 1, Figure 1A,B). The mutation was found in neutrophils and CD3+ T lymphocytes as well as in nonhematopoietic tissues such as nails and hair allowing the diagnosis of PFCP to be made. There was no familial his-

tory of polycythemia. The blood counts of the mother and two children were strictly normal. The fatherâ&#x20AC;&#x2122;s blood count was not available. Erythroid progenitors displayed autonomous growth when cultured in the presence of serum (Figure 1C), an effect probably explained by residual erythropoietin in the

Figure 1. Study of the EPOR c.1300dup (p.Gln434Profs*11) in primary cells. (A) Electropherograms showing the EPOR c.1300dup mutation in hematopoietic cells (CD3+ T-lymphocytes and neutrophils) and in non-hematopoietic tissues (nails, hair and buccal swab). (B) Scheme of EPOR wild-type and of the new frameshift EPOR mutants (EPOR FS). Negative signaling regulators (on the left) and internalization/degradation sites (on the right) are indicated. The tyrosine (Y) number of the mature EPOR is also indicated in parentheses. (C, D) Effect of erythropoietin (EPO) concentration on erythroid colony formation in the presence (C) or absence (D) of fetal bovine serum. BFU-E colonies were counted at day 14. Each experiment was performed twice in duplicate. The results are expressed in percentages of the number of colonies at 1 U/mL of EPO. (E) Effect of EPO concentration on EPOR signaling in erythroblasts. After 7 to 10 days in culture with stem cell factor (SCF), interleukin-3 (IL3) and EPO, CD34+ cells from the patient or healthy donors were cytokine-starved for 5 h then stimulated for 15 min with increasing concentrations of EPO. JAK2 and STAT5 phosphorylations were examined by western blotting. One of three independent experiments is presented and fold activation is indicated below. (F) Persistence of JAK2 and STAT5 phosphorylation in erythroblasts. After 7 to 10 days in culture with SCF, IL3 and EPO, CD34+ cells from the patient or healthy donors were cytokine-starved for 5 h prior to 15 min of stimulation with EPO 1 U/mL. Cells were then washed to remove the EPO and cultured in the absence of cytokine or serum. JAK2 and STAT5 phosphorylations were examined by western blotting in a time-dependent manner: after 5 h of starvation (-), after the EPO stimulation (+EPO) and at different times after EPO removal (10 min, 30 min, 1 h and 4 h). One out of two independent experiments is presented and fold activation is indicated below.

578

haematologica | 2018; 103(4)

Erythropoietin receptor mutations in PFCP

serum, as in the absence of serum, no spontaneous growth was observed, but there was greater erythropoietin hypersensitivity (a nearly 5-fold increase) compared to that of control cells from a healthy donor (Figure 1D). We performed signaling experiments with primary CD34+ cells from the PFCP patient which were cultured in

vitro for 7-10 days in the presence of interleukin-3, stem cell factor and erythropoietin. Unlike control cells, PFCP erythroid progenitors exhibited constitutive and persistent phosphorylation of JAK2 as well as constitutive activation of STAT5 with a hypersensitive erythropoietin doseresponse (Figure 1E,F). This erythropoietin-induced JAK2

Figure 2. Functional study of EPOR c.1300dup (p.Gln434Profs*11) in the Ba/F3 cell line. (A) Ba/F3 cells were transduced with pMX-HA-huEPOR-IRES-GFP retrovirus to stably express the wild-type receptor (EPOR WT), a truncated mutant at position 444 (p.Gln444*, EPOR STOP) or the frameshift mutant EPOR c.1300dup (p.Gln434Profs*11, EPOR FS). (B) Proliferation was assessed 48 h after culturing Ba/F3-EPOR cells in the absence or presence of increasing doses of erythropoietin (EPO) (0.01, 0.02, 0.03, 0.05, 0.1, 0.3, and 1 U/mL) by a WST-1 proliferation assay. Dose-response curves are means expressed in percentages of maximum growth value ± SEM (n = 3 in triplicate). Two-tailed t-test, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. (C) Effect of EPO concentration on EPOR signaling. Ba/F3 cells expressing different EPOR constructs were examined by western blotting for the presence and phosphorylation status of various signaling molecules. Cells were serum- and cytokine-starved for 5 h prior to stimulation for 15 min with increasing doses of EPO (0, 0.001, 0.01, 0.1 and 1 U/mL). Expression of β-actin was used as a loading control. One of three independent experiments is presented. (D) Phosho-STAT5/actin spontaneous phosphorylation was quantified compared to WT using Image J. Results represent the mean ± SEM (n = 3). Two-tailed t-test, *P<0.05, ***P<0.001, (E) Persistence of STAT5 phosphorylation. Ba/F3-EPOR cells were serum- and cytokine-starved for 5 h prior to 15 min of stimulation with EPO 1 U/mL. Cells were then washed to remove the EPO and cultured in the absence of cytokine or serum. STAT5 phosphorylation was examined by western blotting in a time-dependent manner: after 5 h of starvation (-), after EPO stimulation (+EPO) and at different times after EPO removal (5 min, 30 min, 1 h, 2 h and 4 h). One out of two independent experiments is presented.

haematologica | 2018; 103(4)

579

F. Pasquier et al.

and STAT5 activation also persisted for 4 h after removal of erythropoietin from the PFCP cells, whereas it was much more transient in control cells (Figure 1F). Here we show that EPOR c.1300dup (p.Gln434Profs*11) is a strong gain-of-function mutation, which induces major erythropoietin hypersensitivity in primary erythroid progenitors, similar to that observed in patients with polycythemia vera, as well as constitutive and persistent activation of JAK2 and STAT5.

Functional analysis of EPOR c.1300dup (p.Gln434Profs*11) in the Ba/F3 cell line In order to study the functional impact of the new frameshift mutant, Ba/F3 cells were transduced to express different HA-tagged human EPOR: the wild-type receptor (EPOR WT), EPOR p.Gln434Profs*11 (EPOR FS), identical to the patient’s mutation or EPOR p.Gln444* (EPOR STOP) generating a truncated receptor at position 444 (Table 1, Figure 2A). EPOR FS and EPOR STOP differed by the nature of the C-terminal ten amino acid residues, namely the new sequence PALASMDTVP in EPOR FS and the natural sequence QLLRPWTLCP in EPOR STOP. These cells expressed quite similar levels of exogenous EPOR, as detected with anti-HA antibody (Figure 2C). Interestingly, EPOR FS migrated slightly above EPOR STOP, suggesting differences in post-translational modification. However, the glycosylation states of EPOR WT, EPOR STOP and EPOR FS were similar after using Endo H and PGNase F (Online Supplementary Figure S1). MTT-like assays were performed to investigate the potential effects of EPOR STOP and EPOR FS on cell proliferation. None of these mutants was able to induce cytokine-independent cell growth. However, EPOR FS conferred a 4- to 5-fold greater erythropoietin hypersensitivity to Ba/F3 cells compared to EPOR WT. Interestingly the erythropoietin-induced growth of EPOR STOP Ba/F3 cells was similar to that of EPOR WT cells (Figure 2B). To confirm these results in a human setting, we transduced EPOR WT, STOP and FS in the human UT-7 cell line, which expresses endogenous EPOR and found similar results as in the Ba/F3 cell line (Online Supplementary Figure S2A). We checked the signaling pathways and observed that EPOR FS induced constitutive phosphorylation of STAT5, AKT and ERK compared to EPOR WT and to a higher extent than EPOR STOP in Ba/F3 cells (Figure 2C) and UT7 cells (Online Supplementary Figure S2B). Semi-quantitative analysis of spontaneous STAT5 phosphorylation showed a significant increase in EPOR FS compared to EPOR WT and EPOR STOP (Figure 2D). Moreover, we observed a similar persistent STAT5 activation in both EPOR FS and EPOR STOP Ba/F3 cells compared to EPOR WT cells (4 h, 2 h and 30 min, respectively, after erythropoietin removal) (Figure 2E). These results show that EPOR FS confers a similar erythropoietin hypersensitivity to Ba/F3 and UT-7 cells as that observed in PFCP erythroid progenitors, while the truncated mutant EPOR STOP did not confer such erythropoietin hypersensitivity to these cells. The greater erythropoietin hypersensitivity induced by EPOR p.Gln434Profs*11 cannot, therefore, be explained by the receptor truncation itself and the loss of the two SHP-1 and SOCS3 binding sites which are responsible for a persistent activation, but rather by the appearance of a new C-terminal tail that confers spontaneous signaling. 580

Effects of the c.1300dup (p.Gln434Profs*11) mutation on erythropoietin receptor stability, cell surface expression and dimerization To further characterize this new EPOR mutant, the cell surface expression of the wild-type receptor and both mutants was studied by flow cytometry. At similar levels of GFP (transduction of cells with IRES-GFP retroviruses), EPOR FS was significantly more abundant at the cell surface (more than 2-fold, P=0.0002) than EPOR WT and EPOR STOP in both Ba/F3 and UT-7 cell lines (Figure 3A and Online Supplementary Figure S3). These results were further confirmed in Ba/F3 cells using radiolabeled 125I-erythropoietin (2,412±494 receptors for EPOR FS, P=0.015, 1,018±284 receptors for EPOR STOP and 1,201±304 receptors for EPOR WT) (Figure 3b). We next investigated the stability of the receptors using treatment with the protein synthesis inhibitor cycloheximide. EPOR FS was more stable than both EPOR WT and EPOR STOP (halflife of 2 h and 1 h, respectively) (Figure 3C). We also studied the dimerization of human EPOR monomers by split Gaussia luciferase assay in steady-state conditions in HEK cells, in the absence of erythropoietin (Figure 3D). Close proximity between the C-terminal cytosolic domains of EPOR was increased by 4-fold with EPOR FS compared to control and EPOR STOP, as assessed by reconstitution of the split Gaussia luciferase activity (Figure 3E). The increased stability and cell surface localization of EPOR FS may also result in a defect in the receptor internalization pattern due to the loss of specific domains located in the C-terminal part of the wild-type receptor. However, EPOR WT, STOP and FS displayed the same internalization pattern as assessed by 125I-erythropoietin labeling experiments (Figure 4A,B). Furthermore, we noted that a dileucine motif, known to be a potential clathrin-dependent endocytosis signal,37 was lost in the new C-terminal tail of the EPOR FS. We assumed that this particular modification could be involved in the increased cell surface expression of EPOR FS. Thus, based on EPOR WT and EPOR STOP constructs, we generated two other mutants, EPOR WT/diL and EPOR STOP/diL, where the dileucine motif was removed (Figure 4C). However its abrogation did not modify either the erythropoietin sensitivity of EPOR WT or STOP Ba/F3 cells (Figure 4D) or the localization of the receptors at the cell surface (Figure 4E). Likewise, EPOR WT/diL and EPOR STOP/diL Ba/F3 cells displayed a similar signaling pattern to that of EPOR WT and EPOR STOP cells, respectively (Figure 4F). Collectively these results show that p.Gln434Profs*11 increases EPOR stability, dimerization and localization at the cell surface without modifying internalization of the receptor.

Different mechanisms are involved in the erythropoietin hypersensitivity phenotype in primary erythrocytosis according to the type of EPOR mutation We wondered whether this model could be extended to other EPOR mutations already described in PFCP. We therefore investigated the impact of the frameshift EPOR c.1311_1312del (p.Pro438Metfs*6) mutation25 and its designed nonsense mutant counterpart, EPOR c.1327_1329delinsTAA (p.Pro443*) (Figure 5A) on the proliferation rate of Ba/F3 cells. These mutants lack the last 65 amino acids of the receptor, retaining two of the eight conserved tyrosine residues, Tyr-368 and Tyr-426 (Table 1, Figure 6A,C,D). Interestingly, EPOR p.Pro438Metfs*6 and haematologica | 2018; 103(4)

Erythropoietin receptor mutations in PFCP

EPOR p.Gln434Profs*11 (EPOR FS) share the same fiveamino acid terminal sequence (MDTVP). MTT-like assays showed that erythropoietin hypersensitivity was induced by EPOR p.Pro438Metfs*6, but not by EPOR p.Pro443* (Figure 5B), similarly to the results obtained with EPOR p.Gln434Profs*11 (EPOR FS) and EPOR p.Gln444* (EPOR STOP), respectively. We also studied two proximal nonsense mutations, EPOR c.1195G>T (p.Glu399*)11 and c.1273G>T (p.Glu425*)10 (Figure 5A), which are responsible for more extensive truncations (109 and 83 amino acids, respectively) and the loss of seven of the eight cytoplasmic tyrosine residues, retaining only Tyr-368 (Table 1, Figure 6A,B). Interestingly, these nonsense mutants, unlike EPOR p.Pro434* and p.Gln444*, were able to confer erythropoietin hypersensitivity to Ba/F3 cells (Figure 5C). These results show that extensive truncations, unlike shorter ones, are sufficient per se to induce the erythropoietin hypersensitivity phenotype. In comparison, erythropoietin hypersensitivity induced by frameshift EPOR

mutations is due to a distinct common mechanism based on the appearance of new amino acid sequences.

Discussion In this study, we identified a new germline heterozygous EPOR mutation, c.1300dup (p.Gln434Profs*11) in a patient suffering from PFCP. Like the other mutations already described in this pathology, c.1300dup is located in exon 8 of EPOR. The frameshift generates a new cytoplasmic tail of ten amino acids and a premature stop codon at position 444 leading to the loss of 64 amino acids in the C-terminal part of the receptor. Two other mutations, c.1281dup (p.Ile428Tyrfs*17)7 and c.1288dup (p.Asp430Glyfs*15),4 leading to a similar truncation at position 444 had been previously reported but, as for most other EPOR mutants, no extensive functional study had been carried out. Here we show that EPOR p.Gln434Profs*11 is a strong gain-of-

haematologica | 2018; 103(4)

Figure 3. Effects of c.1300dup (p.Gln434Profs*11) mutation on EPOR stability, dimerization and cell surface expression. (A) Cell-surface expression of the different EPOR was assessed by flow cytometry using PE fluorescent labeling of the extracellular HA-tag. The histogram shows the ratio of mean fluorescence intensiy (MFI) of PE-labeled cell-surface erythropoietin (EPO) on the respective MFI of GFP. Results are the mean ± SEM of seven independent experiments. (B) Cell-surface expression of the different EPOR was assessed with radiolabeled 125IEPO. The results are expressed in cpm normalized to EPOR WT. The number of cell-surface receptors was determined by comparison between the radioactivity of transduced Ba/F3 cells and parental UT-7 cells that express 7000 receptors. (C) EPOR stability. Ba/F3-EPOR cells were incubated with cycloheximide for different times (0 min, 15 min, 30 min, 1 h, 2 h, 4 h, 6 h) and HA expression was studied by western blotting. HA-EPOR and β-actin were quantified using Image J software. The curves represent the HA/β-actin ratios. Three independent experiments were done. (D) Schematic representation of split Gaussia princeps luciferase complementation assay used to test EPOR dimerization in HEK293-derived BOSC cells. (E) Dimerization of human EPOR monomers was assessed by split Gaussia luciferase assay in steady-state conditions, in the absence of EPO in HEK cells. Close proximity between the C-terminal cytosolic domains of EPOR is significantly promoted by EPOR FS. The results represent the mean ± SEM from three independent experiments, each performed with eight biological replicates per condition. For each experiment, raw values were normalized to the average of the EPOR WT condition before pooling the data of all experiments together. Unpaired two-tailed t-test with Welch correction, ****P<0.0001.

581

F. Pasquier et al.

Figure 4. Role of diLeucine motif (diL) loss in the EPOR c.1300dup (p.Gln434Profs*11, EPOR FS) mechanism. (A, B) EPOR internalization was performed with radiolabeled 125I-erythropoietin (EPO). Cells were incubated with 125I-EPO and washed at 4°C to remove unbound ligand. An acidic wash was then performed to separate cell surface–bound from internalized EPO. The radioactivity levels of the (A) supernatant (cell surface–bound EPO) and of (B) the cell pellet (internalized EPO) were determined. Each experiment was performed three times with similar results. (C) To study the potential role of the dileucine motif loss in the EPO hypersensitivity phenotype induced by EPOR FS, Ba/F3 cells were transduced with pMX-HA-huEPOR-IRES-GFP retrovirus to stably express EPOR WT or EPOR STOP receptor with abrogation of the dileucine motif EPOR WT/diL and EPOR STOP/diL respectively. (D) Proliferation was assessed 48 h after culturing Ba/F3-EPOR cells in the absence or presence of increasing doses of EPO (0.01, 0.02, 0.03, 0.05, 0.1, 0.3, and 1 U/mL) by a WST-1 proliferation assay. Dose-response curves are means expressed in percentages of maximum growth value ± SEM (n = 3 in triplicate). Two-tailed t-test, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. (E) Cell-surface expression of the different EPOR was assessed by flow cytometry using PE fluorescence labeling of the extracellular HA-tag. The histogram shows the ratio of mean fluorescence intensity (MFI) of PE-labeled cell-surface EPOR on the respective MFI of GFP. Results are the mean ± SEM of three independent experiments. (F) Effect of EPO concentration on EPOR signaling. Ba/F3 cells expressing different EPOR constructs were examined by western blotting for the presence and phosphorylation status of various signaling molecules. Cells were serum- and cytokine-starved for 5 h prior to stimulation for 15 min with increasing doses of EPO (0, 0.001, 0.01, 0.1 and 1 U/mL). Expression of β-actin was used as a loading control. One of three independent experiments is presented and fold activation is indicated below.

582

haematologica | 2018; 103(4)

Erythropoietin receptor mutations in PFCP

function mutant that induces major erythropoietin hypersensitivity in primary cells as well as in transduced Ba/F3 cells, without spontaneous growth, in accordance with the diagnosis of PCFP. Previous reports assumed that the erythropoietin hypersensitivity phenotype observed in PFCP is due to the loss of negative regulatory domains located in the C-terminal part of the receptor, especially some conserved tyrosine residues.30,31 Both the magnitude and duration of EPOR signaling are indeed crucial for erythropoiesis and are therefore tightly regulated by several mechanisms induced as soon as erythropoietin binds to its cognate receptor. A classic negative feedback loop is achieved by signaling

inhibitors such as the tyrosine phosphatase SHP-138 and the suppressor of cytokine signaling (SOCS) proteins SOCS3 and CIS,39 which bind the conserved Tyr-426, -454 and -456 on the cytoplasmic tail of EPOR (Figure 6A). Rapid internalization and degradation of the receptor induced by erythropoietin binding34 also contribute to the negative regulation of EPOR signaling. Several possible cooperative and partially redundant mechanisms are involved in this process, but their relative contributions remain unclear. The C-terminal part of EPOR is degraded at the cell surface by the proteasome.34,35 This step requires binding of the E3-ligase Î˛Trcp to a conserved motif Asp461-Ser462-Gly463 of EPOR (Figure 6A).36 The ery-

Figure 5. Different mechanisms are involved in the erythropoietin hypersensitivity phenotype of primary familial and congenital polycythemia depending on the type of EPOR mutation. (A) Ba/F3 cells were transduced with pMX-HA-huEPOR-IRES-GFP retrovirus to stably express different kind of EPOR mutations already described: the frameshift EPOR c.1311_1312del mutant (p.Pro438Metfs*6) or its nonsense designed counterpart EPOR c.1327_1329delinsTAA (p.Pro443*), more proximal truncations due to the nonsense mutants EPOR c.1195G>T (p.Glu399*) and c.1273G>T (p.Glu425*). (B, C) Proliferation was assessed 48 h after culturing EPOR p.Pro438Metfs*6 and p.Pro443* Ba/F3 cells (B) or EPOR p.Glu399* and p.Glu425* Ba/F3 cells (C) in the absence or presence of increasing doses of EPO (0.01, 0.02, 0.03, 0.05, 0.1, 0.3, and 1 U/mL) and compared to EPOR WT and EPOR FS growth by a WST-1 proliferation assay. Dose-response curves are means expressed in percentages of maximum growth value Âą SEM (n = 3 in triplicate). Two-tailed t-test, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

haematologica | 2018; 103(4)

583

F. Pasquier et al.

thropoietin-EPOR complex is then internalized in a process dependent on Tyr-454, -456 and -504 and the p85 subunit of phosphatidylinositol 3 kinase, and targeted to the lysosomes for degradation (Figure 6A).30,31,35 The loss of these negative signaling regulatory and degradation domains due to EPOR truncation is usually thought to explain erythropoietin hypersensitivity in PFCP.6,7,14,27,2931,36,40 In the present study, we studied two extensive truncations, p.Glu399* and p.Glu425*, which lack seven of the eight conserved tyrosine residues (Table 1, Figure 6B) and all sequences that constitute the EPOR negative regulatory domains. These mutants conferred greater erythropoietin hypersensitivity to Ba/F3 cells compared to EPOR WT. Thus, confirming previous reports, we showed that extensive truncations lacking all EPOR neg-

ative regulatory sites are sufficient in themselves to induce the PFCP phenotype. Our results highlight that a different mechanism underlies erythropoietin hypersensitivity due to frameshift EPOR mutations. The study of EPOR p.Gln434Profs*11 (EPOR FS) and EPOR p.Pro438Metfs*6 and their designed nonsense counterparts, EPOR p.Gln444* (EPOR STOP) and EPOR p.Pro434*, respectively, allowed us to discriminate between the effects due to the truncations themselves leading to the loss of SHP-1 and SOCS3 binding sites and those due to the appearance of a new cytoplasmic tail. In accordance with the loss of negative regulatory sites and previous reports,9,14,27,29 EPOR FS and EPOR STOP Ba/F3 cells displayed persistent phosphorylation of STAT5 after erythropoietin removal, but only EPOR FS was able

Figure 6. Regulation of EPOR signaling and positions of the EPOR truncations. (A) Negative regulators of EPOR signaling and their binding sites in the C-terminal part of the receptor: negative signaling regulators (on the left) and internalization/degradation sites (on the right). The tyrosine (Y) number of the mature EPOR is also indicated in parentheses. (B) Proximal truncations due to EPOR p.Glu399* and p.Glu425* lack all negative regulation sites of the receptor. (C and D) More distal truncations due to frameshift EPOR mutants and their designed counterparts retain the Tyr-426.

584

haematologica | 2018; 103(4)

Erythropoietin receptor mutations in PFCP

to induce constitutive STAT5 phosphorylation and erythropoietin hypersensitivity. In a similar way to EPOR p.Gln434Profs*11 (EPOR FS), Ba/F3 cells expressing EPOR p.Pro438Metfs*6 displayed erythropoietin hypersensitivity, unlike EPOR p.Pro443*, suggesting a common mechanism for the frameshift EPOR mutations in PFCP due to the presence of a new, partially common cytoplasmic tail (MDTVP). Of note EPOR p.Pro443* and p.Gln444* have never been identified in PFCP patients. Furthermore, a previously described murine mutant that lacks the last 40 amino acids and has a new cytoplasmic tail, which does not include the MDTVP motif, is also responsible for erythropoietin hypersensitivity. This suggests that other sequences of amino acids could be involved in erythropoietin hypersensitivity.41 Furthermore, another mutant, EPOR p.Trp439*, was found in a large family with erythrocytosis, with a mild phenotype probably associated with incomplete penetrance of the mutation.3 This finding suggests that our in vitro cellular model may be insufficiently sensitive to unmask weak functional effects or that the wild-type WTLCP motif may act as a negative regulator of erythropoietin sensitivity. We also demonstrated that the increased stability and the greater localization at the cell surface of EPOR FS compared to EPOR WT and EPOR STOP were not due to the loss of internalization sites located in the cytoplasmic tail of the protein as these receptors displayed similar internalization patterns. These results suggest better addressing to the cell surface, increased stability or greater recycling of EPOR FS. Moreover, we found that EPOR FS increased the basal dimerization of the receptor, correlating with the spontaneous activation of JAK2 and downstream signaling. Two recent studies have emphasized the role of EPOR dimerization and conformation in signal transduction. Indeed, while the EPO R150Q mutant was unable to mediate full signaling due to decreased EPOR dimerization and JAK2 activation,42 alteration of EPOR conformation by diabodies was able to finely tune the receptor signaling.43 The conformation of EPOR is indeed crucial for its optimal activation. In the absence of ligand, inactive EPOR dimers are pre-formed at the cell surface44 through their transmembrane domains.45 Erythropoietin binding to its receptor induces conformational changes, such as the formation of a 120° angle between the D1 domains of the two EPOR molecules46 and the reorientation of the continuous rigid alpha helix formed by the transmembrane domain and the cytosolic juxtamembrane region that contains a rigid

References 1. Yoshimura A, Longmore G, Lodish HF. Point mutation in the exoplasmic domain of the erythropoietin receptor resulting in hormone-independent activation and tumorigenicity. Nature. 1990;348(6302): 647-649. 2. Juvonen E, Ikkala E, Fyhrquist F, Ruutu T. Autosomal dominant erythrocytosis caused by increased sensitivity to erythropoietin. Blood. 1991;78(11):3066-3069. 3. de la Chapelle A, Traskelin AL, Juvonen E. Truncated erythropoietin receptor causes dominantly inherited benign human ery-

haematologica | 2018; 103(4)

hydrophobic motif, composed of residues Leu278, Ile282 and Trp283.47-49 It has been shown that the orientation of this conserved motif is crucial for JAK2 activation and JAK2-induced EPOR phosphorylation.47 Other active EPOR conformations have been described but this particular dimer orientation is the optimal one for signaling.50 As EPOR is not permissive for self-activation in terms of conformation, the appearance of a new cytoplasmic tail due to frameshift mutations might induce reorientation of EPOR transmembrane and/or juxtamembrane domains, leading to pre-activation of the EPOR-JAK2 complex at the cell surface. To our knowledge this is the first extensive functional study of EPOR mutations in PFCP. We demonstrated here that different mechanisms, depending on the type and location of EPOR mutations, contribute to the erythropoietin hypersensitivity phenotype, as suggested previously. Extensive truncations lacking all negative regulatory and degradation domains are sufficient by themselves to confer erythropoietin hypersensitivity, whereas more distal truncations induced by frameshift mutants confer erythropoietin hypersensitivity that depends on the appearance of a new C-terminal tail. The latter, by increasing EPOR dimerization and stability at the cell surface, cause preactivation of EPOR and JAK2, constitutive signaling and hypersensitivity to erythropoietin similar to that occurring in JAK2V617F-positive polycythemia vera. Acknowledgments We are deeply grateful to the patient involved in the study. We thank the Imaging and Cytometry Platform (PFIC) of Gustave Roussy, especially Philippe Rameau and Yann Lecluse. We also thank Gwendoline Leroy for the sequencing. This work was supported by grants from l’Agence Nationale de la Recherche (ANR-13-JVSV1-GERMPN-01), the Laurette Fugain foundation, the GIS-Institute for rare diseases for high throughput sequencing (AO9102LS), the Association de Recherche sur la Moelle Osseuse (ARMO), the Association pour la Recherche contre le Cancer (ARC) (Fondation ARC libre 2012) and the regional PHRC AOR07014. The “Investissements d’avenir” program is funding the Labex GR-Ex (IP, WV and FV). FP was supported by Ph.D. grants from the ARC. CM was supported by ANR-Blanc 2013 GERMPN. WV is a recipient of a research fellowship from IGR INSERM (contrats d’interface). TB is a Télévie PhD fellow and SNC is supported by the Ludwig Institute for Cancer, Fondation Contre le Cancer, Fondation Salus Sanguinis, FNRS-FRSM-PDR, Project ARC10/15-027 and PAI Belgian Medical Genetics Initiative Project.

throcytosis. Proc Natl Acad Sci USA. 1993;90(10):4495-4499. 4. Sokol L, Luhovy M, Guan Y, Prchal JF, Semenza GL, Prchal JT. Primary familial polycythemia: a frameshift mutation in the erythropoietin receptor gene and increased sensitivity of erythroid progenitors to erythropoietin. Blood. 1995;86(1):15-22. 5. Arcasoy MO, Degar BA, Harris KW, Forget BG. Familial erythrocytosis associated with a short deletion in the erythropoietin receptor gene. Blood. 1997;89(12):4628-4635. 6. Furukawa T, Narita M, Sakaue M, et al. Primary familial polycythaemia associated with a novel point mutation in the erythro-

poietin receptor. Br J Haematol. 1997;99(1): 222-227. 7. Kralovics R, Indrak K, Stopka T, Berman BW, Prchal JF, Prchal JT. Two new EPO receptor mutations: truncated EPO receptors are most frequently associated with primary familial and congenital polycythemias. Blood. 1997;90(5):2057-2061. 8. Kralovics R, Sokol L, Prchal JT. Absence of polycythemia in a child with a unique erythropoietin receptor mutation in a family with autosomal dominant primary polycythemia. J Clin Invest. 1998;102(1):124-129. 9. Watowich SS, Xie X, Klingmuller U, et al. Erythropoietin receptor mutations associat-

585

F. Pasquier et al.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

586

ed with familial erythrocytosis cause hypersensitivity to erythropoietin in the heterozygous state. Blood. 1999;94(7): 2530-2532. Kralovics R, Prchal JT. Genetic heterogeneity of primary familial and congenital polycythemia. Am J Hematol. 2001;68(2):115121. Arcasoy MO, Karayal AF, Segal HM, Sinning JG, Forget BG. A novel mutation in the erythropoietin receptor gene is associated with familial erythrocytosis. Blood. 2002;99(8):3066-3069. Rives S, Pahl HL, Florensa L, et al. Molecular genetic analyses in familial and sporadic congenital primary erythrocytosis. Haematologica. 2007;92(5):674-677. Al-Sheikh M, Mazurier E, Gardie B, et al. A study of 36 unrelated cases with pure erythrocytosis revealed three new mutations in the erythropoietin receptor gene. Haematologica. 2008;93(7):1072-1075. Perrotta S, Cucciolla V, Ferraro M, et al. EPO receptor gain-of-function causes hereditary polycythemia, alters CD34 cell differentiation and increases circulating endothelial precursors. PLoS One. 2010;5(8):e12015. James C, Ugo V, Le Couedic JP, et al. A unique clonal JAK2 mutation leading to constitutive signalling causes polycythaemia vera. Nature. 2005;434(7037): 1144-1148. Scott LM, Tong W, Levine RL, et al. JAK2 exon 12 mutations in polycythemia vera and idiopathic erythrocytosis. N Engl J Med. 2007;356(5):459-468. D'Andrea AD, Fasman GD, Lodish HF. Erythropoietin receptor and interleukin-2 receptor beta chain: a new receptor family. Cell. 1989;58(6):1023-1024. Huang LJ, Constantinescu SN, Lodish HF. The N-terminal domain of Janus kinase 2 is required for Golgi processing and cell surface expression of erythropoietin receptor. Mol Cell. 2001;8(6):1327-1338. Witthuhn BA, Quelle FW, Silvennoinen O, et al. JAK2 associates with the erythropoietin receptor and is tyrosine phosphorylated and activated following stimulation with erythropoietin. Cell. 1993;74(2):227236. Socolovsky M, Nam H, Fleming MD, Haase VH, Brugnara C, Lodish HF. Ineffective erythropoiesis in Stat5a(-/-)5b(-/-) mice due to decreased survival of early erythroblasts. Blood. 2001;98(12):3261-3273. Bouscary D, Pene F, Claessens YE, et al. Critical role for PI 3-kinase in the control of erythropoietin-induced erythroid progenitor proliferation. Blood. 2003;101(9):34363443. Wu H, Liu X, Jaenisch R, Lodish HF. Generation of committed erythroid BFU-E and CFU-E progenitors does not require erythropoietin or the erythropoietin receptor. Cell. 1995;83(1):59-67. Petersen KB, Hokland P, Petersen GB, Nyvold CG. Erythropoietin receptor defect: a cause of primary polycythaemia. Br J

Haematol. 2004;125(4):537-538. 24. O'Rourke K, Fairbairn DJ, Jackson KA, Morris KL, Tey SK, Kennedy GA. A novel mutation of the erythropoietin receptor gene associated with primary familial and congenital polycythaemia. Int J Hematol. 2011;93(4):542-544. 25. Bento C, Almeida H, Maia TM, et al. Molecular study of congenital erythrocytosis in 70 unrelated patients revealed a potential causal mutation in less than half of the cases (where is/are the missing gene(s)?). Eur J Haematol. 2013;91(4):361368. 26. Bento C, Percy MJ, Gardie B, et al. Genetic basis of congenital erythrocytosis: mutation update and online databases. Hum Mutat. 2014;35(1):15-26. 27. Gross M, Ben-Califa N, McMullin MF, et al. Polycythaemia-inducing mutations in the erythropoietin receptor (EPOR): mechanism and function as elucidated by epidermal growth factor receptor-EPOR chimeras. Br J Haematol. 2014;165(4):519528. 28. Chauveau A, Luque Paz D, Lecucq L, et al. A new point mutation in EPOR inducing a short deletion in congenital erythrocytosis. Br J Haematol. 2016;172(3):475-477. 29. Iacobucci I, Li Y, Roberts KG, et al. Truncating erythropoietin receptor rearrangements in acute lymphoblastic leukemia. Cancer Cell. 2016;29(2):186-200. 30. Sulahian R, Cleaver O, Huang LJ. Ligandinduced EpoR internalization is mediated by JAK2 and p85 and is impaired by mutations responsible for primary familial and congenital polycythemia. Blood. 2009;113(21):5287-5297. 31. Bulut GB, Sulahian R, Yao H, Huang LJ. Cbl ubiquitination of p85 is essential for Epoinduced EpoR endocytosis. Blood. 2013;122(24):3964-3972. 32. Le Couedic JP, Mitjavila MT, Villeval JL, et al. Missense mutation of the erythropoietin receptor is a rare event in human erythroid malignancies. Blood. 1996;87(4):1502-1511. 33. Remy I, Michnick SW. A highly sensitive protein-protein interaction assay based on Gaussia luciferase. Nat Methods. 2006;3(12):977-979. 34. Verdier F, Walrafen P, Hubert N, et al. Proteasomes regulate the duration of erythropoietin receptor activation by controlling down-regulation of cell surface receptors. J Biol Chem. 2000;275(24):1837518381. 35. Walrafen P, Verdier F, Kadri Z, Chretien S, Lacombe C, Mayeux P. Both proteasomes and lysosomes degrade the activated erythropoietin receptor. Blood. 2005;105(2): 600-608. 36. Meyer L, Deau B, Forejtnikova H, et al. beta-Trcp mediates ubiquitination and degradation of the erythropoietin receptor and controls cell proliferation. Blood. 2007;109(12):5215-5222. 37. Letourneur F, Klausner RD. A novel dileucine motif and a tyrosine-based motif independently mediate lysosomal targeting

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

and endocytosis of CD3 chains. Cell. 1992;69(7):1143-1157. Jiao H, Berrada K, Yang W, Tabrizi M, Platanias LC, Yi T. Direct association with and dephosphorylation of Jak2 kinase by the SH2-domain-containing protein tyrosine phosphatase SHP-1. Mol Cell Biol. 1996;16(12):6985-6992. Sasaki A, Yasukawa H, Shouda T, Kitamura T, Dikic I, Yoshimura A. CIS3/SOCS-3 suppresses erythropoietin (EPO) signaling by binding the EPO receptor and JAK2. J Biol Chem. 2000;275(38):29338-29347. Arcasoy MO, Harris KW, Forget BG. A human erythropoietin receptor gene mutant causing familial erythrocytosis is associated with deregulation of the rates of Jak2 and Stat5 inactivation. Exp Hematol. 1999;27(1):63-74. D'Andrea AD, Yoshimura A, Youssoufian H, Zon LI, Koo JW, Lodish HF. The cytoplasmic region of the erythropoietin receptor contains nonoverlapping positive and negative growth-regulatory domains. Mol Cell Biol. 1991;11(4):1980-1987. Kim AR, Ulirsch JC, Wilmes S, et al. Functional selectivity in cytokine signaling revealed through a pathogenic EPO mutation. Cell. 2017;168(6):1053-1064 e1015. Moraga I, Wernig G, Wilmes S, et al. Tuning cytokine receptor signaling by reorienting dimer geometry with surrogate ligands. Cell. 2015;160(6):1196-1208. Livnah O, Stura EA, Middleton SA, Johnson DL, Jolliffe LK, Wilson IA. Crystallographic evidence for preformed dimers of erythropoietin receptor before ligand activation. Science. 1999;283 (5404):987-990. Constantinescu SN, Keren T, Socolovsky M, Nam H, Henis YI, Lodish HF. Ligandindependent oligomerization of cell-surface erythropoietin receptor is mediated by the transmembrane domain. Proc Natl Acad Sci USA. 2001;98(8):4379-4384. Syed RS, Reid SW, Li C, et al. Efficiency of signalling through cytokine receptors depends critically on receptor orientation. Nature. 1998;395(6701):511-516. Constantinescu SN, Huang LJ, Nam H, Lodish HF. The erythropoietin receptor cytosolic juxtamembrane domain contains an essential, precisely oriented, hydrophobic motif. Mol Cell. 2001;7(2):377-385. Lu X, Gross AW, Lodish HF. Active conformation of the erythropoietin receptor: random and cysteine-scanning mutagenesis of the extracellular juxtamembrane and transmembrane domains. J Biol Chem. 2006;281 (11):7002-7011. Li Q, Wong YL, Huang Q, Kang C. Structural insight into the transmembrane domain and the juxtamembrane region of the erythropoietin receptor in micelles. Biophys J. 2014;107(10):2325-2336. Seubert N, Royer Y, Staerk J, et al. Active and inactive orientations of the transmembrane and cytosolic domains of the erythropoietin receptor dimer. Mol Cell. 2003;12(5):1239-1250.

haematologica | 2018; 103(4)

ARTICLE

Bone Marrow Failure

Involvement of hepcidin in iron metabolism dysregulation in Gaucher disease

Ferrata Storti Foundation

Thibaud Lefebvre,1,2* Niloofar Reihani,3* Raed Daher,1 Thierry Billette de Villemeur,4 Nadia Belmatoug,5 Christian Rose,6 Yves Colin-Aronovicz,3 Hervé Puy,1,2 Caroline Le Van Kim,3 Mélanie Franco3** and Zoubida Karim1** University Sorbonne Paris Cité, Paris Diderot University, Inserm U1149 / ERL 8252, Inflammation Research Center (CRI), Laboratory of Excellence GR-Ex, Paris; 2 AP-HP, Centre Français des Porphyries, Hôpital Louis Mourier, Colombes; 3 University Sorbonne Paris Cité, Paris Diderot University, Inserm, INTS, “Biologie Intégrée du Globule Rouge” Department, Laboratory of Excellence GR-Ex, Paris; 4 Sorbonne Universités, UPMC, GRC ConCer-LD and AP-HP, Hôpital Trousseau, Service de Neuropédiatrie, Centre de Référence des Maladies Lysosomales, Paris; 5 Hôpitaux Universitaires Paris Nord Val de Seine, Assistance Publique-Hôpitaux de Paris, Hôpital Beaujon, Service de Médecine Interne, Centre de Référence des Maladies Lysosomales, Clichy and 6Université Catholique de Lille, Hôpital Saint Vincent de Paul, Service d'Hématologie, France 1

*TL and NR contributed equally to this work. **MF and ZK contributed equally to this work.

Haematologica 2018 Volume 103(4):587-596

ABSTRACT

aucher disease (GD) is an inherited deficiency of glucocerebrosidase leading to accumulation of glucosylceramide in tissues such as the spleen, liver, and bone marrow. The resulting lipid-laden macrophages lead to the appearance of “Gaucher cells”. Anemia associated with an unexplained hyperferritinemia is a frequent finding in GD, but whether this pathogenesis is related to an iron metabolism disorder has remained unclear. To investigate this issue, we explored the iron status of a large cohort of 90 type I GD patients, including 66 patients treated with enzyme replacement therapy. Ten of the patients treated with enzyme replacement were followed up before and during treatment. Serum levels of hepcidin, the iron regulatory peptide, remained within the physiological range, while the transferrin saturation was slightly decreased in children. Inflammation-independent hyperferritinemia was found in 65% of the patients, and Perl’s staining of the spleen and marrow smear revealed iron accumulation in Gaucher cells. Treated patients exhibited reduced hyperferritinemia, increased transferrin saturation and transiently increased systemic hepcidin. In addition, the hepcidin and ferritin correlation was markedly improved, and, in most patients, the hemoglobin level was normalized. To further explore eventual iron sequestration in macrophages, we produce a Gaucher cells model by treating the J774 macrophage cell line with a glucocerebrosidase inhibitor and showed induced local hepcidin and membrane retrieval of the iron exporter, ferroportin. These data reveal the involvement of Gaucher cells in abnormal iron sequestration, which may explain the mechanism of hyperferritinemia in GD patients. Local hepcidin-ferroportin interaction was involved in this pathogenesis.

Correspondence: zoubida.karim@inserm.fr

Received: August 1, 2017. Accepted: January 3, 2018. Pre-published: January 5, 2018. doi:10.3324/haematol.2017.177816 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/587 ©2018 Ferrata Storti Foundation

Introduction Gaucher Disease (GD) is caused by an inherited deficiency of the lysosomal enzyme glucocerebrosidase, leading to the accumulation of glycosphingolipids in various organ systems, particularly in myeloid mononuclear cells.1 GD is mostly characterized by the presence of lipid-laden macrophages that turn into "Gaucher cells" with a striated appearance and off-center nuclei.2 GD is classified into three clinical types according to the absence (type 1) or presence (types 2 and 3) of central neurological impairment. Type I GD (GD1) remains the most prevalent form of GD.3 In all forms, the clinical presentation of GD and the age of diagnosis are highly variable. The accumulation of Gaucher cells in the spleen, liver and bone marrow leads to significant organomegaly, cytopenia and bone disorders.3 In addition, haematologica | 2018; 103(4)

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.

587

T. Lefebvre et al. 36% of Gaucher patients are anemic at diagnosis,3–5 but there is no direct correlation between anemia and the degree of splenomegaly or cytopenia, suggesting that other mechanisms may be involved. Accordingly, we have recently shown that red blood cells (RBCs) exhibit abnormal deformability and adhesion properties that may trigger ischemic events and phagocytosis in GD.6 Using erythroid progenitor cultures derived from peripheral CD34+ cells, we also demonstrated that GD1 patients exhibit ineffective erythropoiesis with increased plasma levels of erythropoietin (EPO) and GDF15 factor.7 The treatment of GD is based on enzyme replacement therapy (ERT)8 using either imiglucerase (Cerezyme®, Genzyme, SANOFI company, Cambridge®, MA, USA) or velaglucerase alfa (VPRIV®, Shire HGT, Lexington, MA, USA). Substrate reduction therapy (SRT) using miglustat (Zavesca®, Actelion Pharmaceuticals, Allschwil, Switzerland) and the new molecule eliglustat (Cerdelga®, Sanofi-Genzyme) may also be used. However, ERT remains the current standard of care for GD treatment, and once started, ERT is generally administered for life. Several reports have documented the appearance of clinical hyperferritinemia in GD patients in whom the level of serum ferritin may exceed 2-3 times the normal level, suggesting an altered iron homeostasis pathogenesis in GD.4,9-11 Ferritin is an intracellular cytosolic protein that stores iron in almost all tissues, including the liver and spleen, but small amounts are secreted into the serum. The amount of this secreted fraction significantly increases when ferritin synthesis is exceeded due to iron accumulation in target tissues. However, serum ferritin may also increase in the case of an inflammatory stimulus.12 Iron homeostasis is mainly based on iron storage in the liver and macrophages of the spleen, which together contain the largest pool of iron (derived from the phagocytosis of senescent red cells and catabolism of heme) and control its release according to the body’s needs (20-30 mg of iron daily). The intestine is the second compartment that provides 1-2 mg of iron, corresponding to the amount lost daily by the body. In macrophages, as well as in enterocytes, iron is exported to the bloodstream through the iron exporter ferroportin (FPN). Hepcidin, secreted by hepatocytes and acting as a hyposideremic factor, is the main regulator of these iron fluxes.13,14 Hepcidin was first shown in HEK-293 cells transfected with GFP-tagged ferroportin to regulate iron efflux through a direct interaction with ferroportin, leading to the internalization and lysosomal degradation of the exporter.15 This result was confirmed in primary cultured macrophages expressing endogenous ferroportin.16 In addition, hepcidin acts on the iron importer DMT1 (divalent iron transporter 1) at the apical side of the enterocytes, possibly leading to efficient inhibition of iron intestinal absorption.17,18 Hepcidin synthesis is regulated by iron through a complex of integral hemochromatosis proteins, i.e. HFE, HJV (hemojuvelin) and TfR2 (Transferrin Receptor 2). They tightly co-ordinate signaling through the BMP6/HJV/SMAD pathway and increase hepcidin gene (named Hamp) expression when serum or tissue iron levels increase. Inflammation with an accompanying production of cytokines such as IL-6 can also contribute to the elevation of hepcidin synthesis through the JAK/STAT3 signaling pathway.19 Thus, inflammatory conditions can raise the levels of hepcidin, leading to the anemia of inflammation characterized by a high plasma ferritin and low transferrin saturation (TS).20 Interestingly, besides 588

being predominantly produced by the liver, hepcidin was found to be expressed in other organs, although to a lesser extent (e.g. the kidney, gut, and retina).21-28 In addition, hepcidin is expressed in macrophages, where it is supposed to act through an autocrine/paracrine pathway to decrease iron release.29 Two previous studies have reported contradictory results concerning the serum hepcidin concentration in untreated patients with GD1. Thus, the question of the serum hepcidin concentration and its role in iron metabolism in GD requires more investigation.10,11 In this study, we sought to decipher the origin of hyperferritinemia in GD. We investigated iron metabolism parameters and measured the serum hepcidin levels in a large cohort of patients with GD1. In vitro studies, as well as the longitudinal follow up of 10 ERT-treated patients, allowed us to highlight iron retention in Gaucher cells and the improved redistribution of this iron after ERT in patients.

Methods Patients We conducted a retrospective observational study on a cohort of 90 patients with type 1 Gaucher Disease (GD1). Three centers participated in the study (Beaujon and Trousseau Hospital, Assistance Publique-Hôpitaux de Paris, and Saint Vincent Hospital, Lille, France). The patients were registered from 2009 to 2015 (for patients' characteristics see Online Supplementary Table S1). Patients with iron overload, defined by hyperferritinemia associated with a transferrin saturation exceeding 45%, and splenectomized patients were excluded from the study. Most of the patients (n=66) were treated with ERT using either imiglucerase or velaglucerase alfa. Twenty-four patients were untreated at the time of the study, and 10 of them had a longitudinal biological follow up with blood samples taken every 6-10 months. Only patients who received at least six months of ERT were considered as treated patients, and the longest time that a patient had been treated was 23 years. The cohort was explored according to sex and age. Patients aged 16 years or younger were considered as children (pediatric population). Hyperferritinemia was considered when the serum ferritin values were above 300 µg/L for men and above 150 µg/L for children and women. Cases with hyperferritinemia of known origin were excluded, including chronic or acute inflammatory syndrome, history of transfusion or iron supplementation, cellular lysis, dysmetabolic syndrome, and excessive alcohol consumption. Anemia was determined by hemoglobin levels below 13 g/dL for men, 12 g/dL for women and 11.5 g/dL for children. The study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the local ethics committee.

Hematologic and biochemical analysis Serum iron, ferritin, transferrin, C-reactive protein (CRP) and soluble transferrin receptor (sTfR) were quantified using the Dimension RXL and Vista 1500 system (Siemens Healthcare, Saint-Denis, France). Transferrin saturation (TS) was calculated as the percent of [serum iron (mmol/L) / serum transferrin (g/L) × 25]. The hemoglobin level was measured on an automated counter (Sysmex, Roissy, France), and quantification of IL-6 was performed using a cytometric bead array (BD Bioscience, Le Pont de Cliax, France). Serum hepcidin was quantified by the previously published method of LC-MSMS.30 Biological markers for each subgroup were expressed in mean±Standard Error of Mean (SEM). haematologica | 2018; 103(4)

Hepcidin and hyperferritinemia in Gaucher disease Table 1. Iron-related biological parameters of the untreated and treated Type 1 Gaucher disease (GD1) cohorts.

Group

ERT

Age (years)

Ferritin (mg/L)

Fer (mmol/L)

Transferrin saturation (%)

Hepcidin (ng/mL)

Hemoglobin (g/dL)

MCV (fL)

Men

20-65

Women

19-63

Children

3-13

Men

Yes

17-80

Women

Yes

17-64

Children

Yes

9-14

563±148 [53-1443] 518±162 [27-2141] 287 ±62 [97-779] 372±52 [56-1341] 278±88 [33-2572] 104*±19 [52-175] 30-300 15-150 15-150

16.0±2.2 [8-22] 13.9±1.4 [6-20] 10.5±0.6 [9-13] 17.5±0.9 [8-28] 16.9±1.2 [5-34] 16.6±1.0 [1-34] 12-26 10-26 10-26

23.5±3.3 [10.0-35.6] 21.7±2.3 [10.0-33.6] 14.9±1.5 [10.8-21.0] 28.7±1.7 [11.0-48.2] 29.8±2.8 [8.0-73.0] 19.9±2.7 [11.0-28.0] 20-40 15-35 15-35

12.9±3.6 [2.3-35.1] 6.4±2.1 [0-20.5 6.6±1.2 [1.0-15.0] 24.1±5.1 [1.1-127.7] 10.2±2.5 [0-46.1] 6.4±2.4 [0-17]

14.1±0.3 [12.9-16.2] 11.9±0.4 [6.4-13.7] 11.3±0.3 [9.5-12.5] 15.5***±0.2 [13.6-17.3] 13.6***±0.2 [12.2-16.5] 13.2**±0.4 [12.1-14.4] 13-18 12-17 11,5-16

84.8±1.8 [75.0-91.9] 88.2±1.8 [76.0-100.0] 74.1±1.5 [67.0-80.0] 85.1±0.9 [75.0-96.0] 89.3±0.9 [81.0-98.0] 79.7*±1.7 [70.0-83.0]

Normal values

Men Women Children

1-20

80-100

Data are expressed as the means ± Standard Error of Mean, and the range is shown in brackets. The two-tailed Mann-Whitney test was used to compare each subgroup between the untreated and treated patients (*P≤0,05; **P≤0,01; ***P≤0,001). N: number; ERT: enzyme replacement therapy. MCV: mean corpuscular volume; Fer: serum iron.

Cell culture The J774 cell line derived from mouse macrophages was purchased from the American Type Culture Collection (ATCC) and was cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with a 10% low-endotoxin fetal bovine serum (FBS), 4 mM glutamine and 1.5 g/L sodium bicarbonate at 37°C with 5% CO2 in the atmosphere. Cells were plated and maintained overnight in the above medium in a 6-well plate. Confluent cells were first treated with FeNTA (100 mM Fe/400 mM NTA) to drive the expression of FPN at the cell membrane. Twenty-four hours (h) later, 1 mM conduritol B epoxide (CBE) was added for 96 h.

Immunofluorescence and confocal microscopy Treated cells were washed 3 times with PBS, fixed with 3% paraformaldehyde for 10 min at room temperature, washed again 3 times with PBS, incubated with 20 mmol/L glycine for 10 min, and subsequently permeabilized for 30 min with PBS containing 0.1% saponin. FPN staining was performed using an anti-FPN antibody (Alpha Diagnostic International, San Antonio, USA) (1/50, for 1 h at room temperature). Actin staining was performed using Alexa Fluor 635 phalloidin (1/200, for 30 min at room temperature). The coverslips were mounted using Dako glycerol medium supplemented with 2.5% 1,4 diazabicyclo-(2.2.2)octane (DABCO) (Sigma Aldrich, St Louis, MO, USA) as the anti-fading reagent. Confocal images were taken with a high-resolution confocal microscopy LSM 780 (ZEISS, Marly-Le-Roi, France) equipped with a 63X oil immersion objective.

Statistical analysis Statistical analysis was performed using Prism 7 software (GraphPad Software, La Jolla, CA, USA). The unpaired MannWhitney test and the paired Wilcoxon test were used to compare biological values. The χ2 test was used to compare patient groups. Spearman’s test was performed to qualify the correlations. P<0.05 was considered statistically significant. Other methods are described in the Online Supplementary Appendix. haematologica | 2018; 103(4)

Results Biological features of the untreated patients The main biological parameters are shown in Table 1. As expected, high ferritinemia was found in untreated patients, with mean values of 563, 518 and 287 g/L in the subgroups of men, women and children, respectively. Twenty-two (65%) out of these 34 patients showed high serum ferritin levels (67% of men, 57% of women and 73% of children) (Online Supplementary Table S2). Despite a wide range of values, we found that the serum ferritin level in women was particularly high, with an average value exceeding three times the accepted normal level. Men and women (adult patients) exhibited a low normal level of TS (means of 22%). However, young patients were iron deficient, with a mean TS of 14.9% and decreased serum iron (Table 1). We also measured serum hepcidin levels by liquid chromatographic-tandem mass spectrometry (LC-MSMS). In all groups, the level of serum hepcidin was within the normal range (mean, 12.9±4 ng/mL for men, 6.4±2 ng/mL for women and 6.6±1 ng/mL for children; the normal range is 1-20 ng/mL) (Table 1). As expected, anemia was observed in 33% of our patients (40% of children, 43% of women and one man, 11%), although the hemoglobin level was only slightly reduced (Online Supplementary Table S3). Anemia was normocytic with respect to a mean corpuscular volume (MCV) of 74.1±1.5 fL in children (Table 1). Soluble transferrin receptor (sTfR) was also slightly increased with a mean of 2.40±0.24 mg/L (normal values ≤1.76 mg/L).

Hyperferritinemia is independent of the global iron status and systemic inflammation Because ferritin is stimulated during inflammation, we investigated the inflammatory status of our cohort of GD patients. All individuals included in the study were negative for CRP (<5 mg/L). We also measured the serum level 589

T. Lefebvre et al. A

Figure 1. Hemoglobin levels, serum ferritin and serum hepcidin values in untreated Gaucher disease (GD) patients. Correlation of the hemoglobin concentration (Hb) with serum ferritin (A-C) and serum hepcidin (E-F) in men (A and D), women (B and E) and children (C and F), respectively. According to the Spearman test, no significant correlation was found (r<0.3; P>0.05) between hemoglobin and ferritin but correlations were positive between hemoglobin and hepcidin in women and children. The dotted lines represent the high normal value of ferritin on the x-axis and low normal level of hemoglobin on the y-axis.

of the pro-inflammatory interleukin 6 (IL-6) in 22 patients (mix of treated and untreated patients) and found that this cytokine was not induced (data not shown), suggesting an absence of systemic inflammation. We conclude that hyperferritinemia is independent of the inflammatory state in our GD cohort. Furthermore, there was no correlation between serum ferritin and hemoglobin in untreated patients (Figure 1). There was even a positive correlation between hemoglobin and serum hepcidin in women and children, who were more affected by anemia in our cohort, excluding the eventual presence of an anemia of inflammation. Because some untreated GD patients have normal ferritinemia (NF), we compared their levels of serum hepcidin and TS with those of untreated patients with hyperferritinemia (HF) (Figure 2). For both parameters, no significant differences were found between NF and HF groups, and correlation between serum hepcidin and ferritin was not significant in any group of the cohort (Online Supplementary Figure S1). Thus, the ferritin level in GD patients is increased independently of systemic inflammation, the iron pool and the serum hepcidin level.

Hyperferritinemia reflects local iron sequestration in Gaucher cells Serum ferritin is a primary marker of tissue iron overload; thus, we investigated whether hyperferritinemia might reflect local iron sequestration in Gaucher cells. Perlâ&#x20AC;&#x2122;s staining on the myelogram of one patient and on histological sections of the spleen from 2 patient biopsies showed high iron accumulation in Gaucher cells in both 590

organs (Figure 3A and B, respectively). We noticed a high variation in the amount of accumulated iron within these Gaucher cells, particularly in the spleen. To investigate the underlying mechanism of this iron sequestration, we decided to mimic the Gaucher phenotype in vitro using the J774 macrophagic cell line incubated with the GCerase inhibitor CBE. The inhibition of GCerase activity in the presence of CBE was confirmed by flow cytometry showing a decreased fluorescence of its PFB-FDGlu substrate (Online Supplementary Figure S2). We hypothesized that the inhibition of GCerase activity could affect the protein expression of the iron exporter FPN (Figure 4). In the absence of CBE, FPN protein staining was strongly observed both at the plasma membrane localized with actin and in the subapical compartment, reminiscent of the known intracellular diffuse distribution of the FPN protein.16 However, when added to the culture media at a final concentration of 1 mM for 96 h, CBE dramatically reduced the level of FPN membrane expression (Figure 4A). We also explored the expression pattern of CD11b, a macrophage cell surface protein, which was found not to be affected by CBE treatment (Online Supplementary Figure S3), confirming specificity in FPN regulation by CBE. The decrease in the FPN protein level was not associated with decreased FPN mRNA, the level of which was slightly but not significantly increased (Figure 4B), indicating a posttranslational downregulation of FPN expression due to GCerase activity inhibition. Furthermore, FPN downregulation was associated with increased level of cellular ferritin in CBE-treated cells, reminiscent of an iron sequestration in Gaucher cells (Figure 4D). Because the post-transhaematologica | 2018; 103(4)

Hepcidin and hyperferritinemia in Gaucher disease

lational downregulation of FPN may involve hepcidin and macrophages can produce hepcidin locally, we quantified the expression level of the Hamp gene encoding hepcidin in J774 cells under the same conditions. We found that inhibition of GCerase activity significantly increased the mRNA level of hepcidin, which may lead, at least in part, to a reduction in FPN protein via an autocrine-paracrine interaction and resulting in iron retention in these cells (Figure 4C). Since hepcidin in liver was found to be induced by inflammatory cytokines, likely by IL-6 or IL1β, we also explored the J774 cells inflammatory profile by quantifying secreted cytokines and chemokines in their supernatant. under CBE treatment, the levels of IL-1β, MCP-1 and CCL-5 levels were increased, the levels of IL-

10 and TNF-α levels were not affected, and INF-g, IL-6 and IL-12 remained under limit of detection (Figure 5). Thus, one can postulate that increased hepcidin in CBE-treated J774 macrophages is dependent on a local inflammatory environment.

The improvement of biological abnormalities by enzyme replacement therapy confirms the local iron sequestration hypothesis Serum ferritin was significantly reduced in children (104 vs. 287 mg/L; P=0.008). A trend toward reduced values was observed in adults (Table 1). Hyperferritinemia was observed in 42% of treated rather than 65% of untreated patients (P=0.02) (Online Supplementary Table S2). This

B Figure 2. Serum hepcidin level and transferrin saturation (TS) are independent of the ferritin status. Serum hepcidin (A) and TS (B) levels were measured in the normal ferritin (NF) and hyperferritin (HF) untreated patient groups. The means are represented by the horizontal line. According to a two-tailed Mann-Whitney test, no significant differences were observed.

B Figure 3. Iron sequestration in Gaucher cells. Tissue iron content was determined by Perl’s staining of medullar smear (A) and spleen sections (B). The large cells with a laminated aspect were Gaucher cells. The representative images show iron deposition mostly in Gaucher cells. The classical description of Gaucher cells (black arrows) is limited to cells 20-100 μm in diameter with eccentrically placed nuclei and cytoplasm with characteristic crinkles and striations. Images were taken at 20X magnification and a higher magnification (40X).

haematologica | 2018; 103(4)

591

T. Lefebvre et al.

amelioration was more pronounced in children, with hyperferritinemia observed in 14% of treated versus 73% of untreated patients (P=0.008). ERT also tended to improve TS, with a mean value of 28.7% in men, 29.8% in women and 19.9% in children (Table 1). Regarding the hepcidin response, we found that ERT led to slightly increased serum hepcidin levels in men. However, in women and children, the serum hepcidin levels remained unchanged, most likely because hemoglobin in these two groups was lower and may maintain hepcidin levels within the minimal range for erythropoiesis iron availability. ERT markedly increased the hemoglobin levels, and none of the treated patients remained anemic (Online Supplementary Table S3). Indeed, the average hemoglobin levels increased by 1.4 g/dL in men, 1.7 g/dL in women and 1.9 g/dL in the pediatric group (Table 1). Moreover, the longitudinal monitoring of the biological parameters of patients before/after ERT analyzed by a paired comparison significantly highlighted the tendencies observed in the global cohort study. This patient follow up demon-

strated reduced ferritin levels, increased TS and improved hemoglobin levels, suggesting the involvement of iron release and underlying the need for its bioavailability for erythropoiesis recovery (Figure 6A-C). We also quantified the serum hepcidin level at different time points during ERT. Interestingly, we observed that the kinetics of systemic hepcidin secretion reflected a positive and transient adaptive response against iron release in circulation. Compared to untreated patients, the hepcidin level increased significantly between six months and five years of treatment and then returned to the baseline value five years post ERT (Figure 6D). In addition, we also observed significantly lower serum soluble transferrin receptor levels in all treated patients than in untreated patients (Figure 6E). We monitored the time course of different biological parameters in 2 representative patients: one who began the ERT during adolescence and the other who began treatment later in adulthood. The 2 patients were followed up to 25 months post ERT, and the ERT efficacy

Merge

Figure 4. Impact of glucocerebrosidase activity inhibition on ferroportin (FPN) and hepcidin expression in the J774 cell line. J774 cells were incubated with 1 mM CBE (+CBE) or with vehicle (-CBE) for 96 hours (h). (A) Staining of FPN and actin were performed as described in the Methods section. FPN is labeled in green and actin in red. In the absence of CBE, FPN is stained mostly at the plasma membrane with some intracellular extent. In the presence of CBE, FPN membrane staining was reduced, and its localization was mostly intracellular. The cross-section images demonstrated a significant overlap of FPN and actin staining in the -CBE condition but not in the +CBE condition. Images were taken by confocal microscopy (60X). (B and C) Quantification of the mRNA expression levels of FPN (B) and hepcidin (C) from treated and untreated J774 cells. Data were normalized by the housekeeping transcript Hprt1 and expressed as the percentage of the -CBE group meanÂąStandard Error of Mean. Mann-Whitney test was used to compare RNA levels. (D) Quantification of ferritin in cellular extracts from treated and untreated J774 cells. Data were expressed as percentage of the control mean. Medians are represented by the horizontal line. According to a two-tailed Mann-Whitney test, the level of cellular ferritin was significantly higher in the CBE treated cells; P=0.021.

592

haematologica | 2018; 103(4)

Hepcidin and hyperferritinemia in Gaucher disease

Discussion

regarding iron parameters was compared between them (Figure 7A and B). In Patient 1, an 18-year-old woman, the ferritin level decreased dramatically after only five months of ERT, and this decline was maintained continuously over time (Figure 7A). Consequently, the TS began to increase, reaching a maximal level (approximately 40%) at 10 months post ERT, and then returned to a steady state value of approximately 30%. The kinetics of systemic hepcidin secretion was again positive and transient with a maximal level at 18 months post ERT, allowing the TS to reach normal values. This controlled iron release was beneficial for restoring hemoglobin synthesis because its serum level requires ten months post ERT to stabilize. Similar kinetic patterns were confirmed in another young treated patient (Online Supplementary Figure S4), although the TS values and intermediate time data were missing. Patient 2 was a woman who received ERT for the first time at 52 years of age. Interestingly, the decrease in the ferritin level was not evident as late as 15 months post ERT. The kinetic patterns of the other parameters were similar to those of Patient 1 (Figure 7B). Finally, we analyzed the correlation between the hepcidin and ferritin values in untreated and treated GD1 patients. Compared with that in untreated patients, the correlation between both parameters was significantly improved under ERT with a higher slope (Figure 7C). Indeed, the hepcidin response levels were more adapted toward the ferritin values, confirming the restored systemic hepcidin control on iron metabolism. Furthermore, this recovery of controlled iron metabolism was associated with a reduction in the soluble transferrin receptor level in treated patients (Figure 6E).

This study demonstrated for the first time that hyperferritinemia observed in GD1 patients is related to abnormal iron metabolism and the sequestration of iron in Gaucher cells. ERT allows for the release of iron, improves the iron status, and corrects the anemia. In addition to spleen dysfunction, bone marrow infiltration by Gaucher cells and ineffective erythropoiesis, part of the anemia can likely be explained by this abnormal iron metabolism. From a large cohort of 90 GD1 patients, we showed that most of them (65%) exhibited high levels of serum ferritin, with one-third presenting moderate anemia. These results are in line with previous data.9-11 Serum iron and TS were in a normal range as previously shown in adult GD1 patients.4,9 However, the pediatric population exhibited decreased iron release into plasma as shown by a low TS and significant anemia. Indeed, children were the most affected patients, most likely because the iron requirements are very high during childhood due to the high energy intake and active erythropoiesis, limiting the body pool of iron.31 In our cohort of GD patients, hyperferritinemia was not related to an obvious systemic inflammation because neither CRP nor IL-6 were detected. Furthermore, there was no difference in hepcidin level, which is also induced by inflammation,32,33 between the groups of GD patients with and without hyperferritinemia. A normal serum hepcidin concentration despite hyperferritinemia has already been reported by Lorenz et al. in a limited number of GD1 patients (n=11).11 Although MedranoEngay et al. reported increased hepcidin levels in 3 untreated GD1 patients,10 our data indicate that the ferritin levels

Figure 5. Impact of CBE treatment on J774 inflammatory profile. IL-1Î˛, IL-10, MCP-1, CCL-5 and TNF-Îą were quantified in supernatant of J774 cells after incubation with (+CBE) and without (-CBE) CBE for 96 hours (h). Mann-Whitney test was used to compare cytokine levels.

haematologica | 2018; 103(4)

593

T. Lefebvre et al. A

Figure 6. Impact of ERT on iron metabolism. Paired comparison of the levels of ferritin (A), TS (B) and hemoglobin (C) in patients before (pre-ERT) and after initiation of enzyme replacement treatment (ERT) (from 6 months at least). (D) The hepcidin level was measured in untreated (NT) and treated patients during different periods of time. Each group was compared with the NT group. The hepcidin level was transiently increased during the first five years of ERT (<5). (E) Comparison of the levels of serum soluble transferrin receptor in untreated (n=24) versus treated (ERT; n=22) patients. (A-C) A paired Wilcoxon test was used; (D and E) an unpaired Mann-Whitney test was used.

in GD patients are increased independently of the hepcidin levels and inflammation. Indeed, our measures of hepcidin were analyzed in a larger cohort of untreated GD1 patients (n=34) and interpreted in the context of other iron metabolism parameters. Previous reports have documented that GD may trigger pro-inflammatory and anti-inflammatory pathways including IL-6 coming from monocytes/macrophages lineages, and pathways due to glucocerebroside accumulation.34-36 These pathways could lead to high ferritin levels and increased hepcidin transcription. However, the pathological manifestations of GD are highly variable, leading to different levels of accumulated glucocerebroside among patients and individually within organs. Thus, one may speculate that the management of our patients was performed before the critical threshold of glucosylceramide storage was reached, leading to the absence of systemic inflammation in these patients. Moreover, the appearance of local inflammation around Gaucher cells cannot be excluded. Indeed, Gaucher cells have been shown to exhibit a phenotype resembling activated M2 macrophages not expressing pro-inflammatory cytokines; however, they are surrounded by macrophages with an M1 signature.37 Under steady-state conditions, the serum ferritin concentration is well correlated with total body iron stores.38 The sequestration of iron in Gaucher cells may then account for the increased production and secretion of ferritin. Of interest, iron accumulation in the macrophage compartment has already been described in GD.39-41 Herein, using Perlâ&#x20AC;&#x2122;s staining of a spleen biopsy and medullar smear from one GD patient, we confirmed a significant iron overload in 594

Gaucher cells. The origin of this iron accumulation is not fully elucidated, but evidence for increased erythrophagocytic activity in Gaucher cells with the ingestion of mainly mature erythrocytes has already been provided.42,43 Our data suggested that iron efflux is also affected in these cells. Indeed, the expression of FPN, the sole iron exporter, was significantly diminished upon glucocerebrosidase inhibition and accumulation of glucocerebroside into J774 macrophage cells. Because hepcidin transcripts were also enhanced, we postulate that one of the mechanisms involved in this FPN downregulation may be the induction of local hepcidin in this in vitro system. Macrophages have already been shown to be able to express endogenous hepcidin.29 Our data confirmed these reports and also suggested that, in Gaucher cells, the hepcidin/ferroportin axis may be active via an autocrine/paracrine pathway. Whether hepcidin is induced via a local inflammatory state or directly via lipid-mediated pathways remains to be determined. However, one cannot exclude the possibility that FPN may also be directly affected by lipid-mediated pathways in addition to local hepcidin effect. Enzyme replacement therapy was associated with a notable reversal of hyperferritinemia that appeared to correlate with iron release from Gaucher cells. Indeed, longitudinal monitoring of 10 patients under ERT showed the opposite time course of ferritin and rates of TS. This improvement in iron availability occurred under the control of systemic hepcidin as evidenced by the moderate elevation in TS compared to the strong decrease in ferritinemia. This is also reinforced by the improvement in the correlation between serum hepcidin and ferritinemia post ERT. We postulate that the recovery of hemoglobin haematologica | 2018; 103(4)

Hepcidin and hyperferritinemia in Gaucher disease

Figure 7. Recovery of hepcidin control on iron metabolism in treated Gaucher disease (GD) patients. (A and B) Time course of iron-related parameters in 2 treated patients from enzyme replacement therapy (ERT) initiation until 18-25 months after ERT initiation. Patient 1: 18-year-old woman; Patient 2: 52-year-old woman. (C) According to the Spearman test, hepcidin/ferritin correlation in the untreated and treated patients r and P were shown for each condition and slope of linear regression.

levels under ERT may also be partly due to this iron availability. This was confirmed by the lower level of soluble transferrin receptor in treated patients. Similarly, an increased number of reticulocytes in treated patients would have helped to highlight this mechanism, but unfortunately these data were not available. However, iron restriction may not be the only cause of GD anemia, as we have already shown that GD patients exhibit an ineffective erythropoiesis,7 as well as altered morphological properties of RBCs that could accelerate their erythrophagocytosis.43,44 In this study, the efficacy of ERT was particularly demonstrated in the pediatric population in which the iron requirement is high. Likewise, the low ferritin responsiveness in Patient 2 under ERT may imply a low efficacy of this treatment in older patients, but this hypothesis requires additional studies on a larger number of patients. Interestingly, the level of serum hepcidin was increased after ERT. This is probably to fight against iron excess due to the improvement of iron release from Gaucher cells, and to regain a balanced iron status. In conclusion, our study revealed for the first time that hyperferritinemia in GD is related to the sequestration of iron in Gaucher cells due to the downregulation of the

References 1. Brady RO, Kanfer JN, Shapiro D. Metabolism of glucocerebrosides. II. Evidence of an enzymatic deficiency in Gaucher’s disease. Biochem Biophys Res Commun. 1965;18:221-225.

haematologica | 2018; 103(4)

iron exporter ferroportin. Hepcidin from macrophages, rather than the systemic peptide, seems to be responsible for this FPN repression. ERT restored the iron availability and improved the anemia in these patients. Overall, these results provide new perspectives for better management of GD patients. Indeed, the measurement of liver iron content by magnetic resonance imaging and the analysis of iron status must be systematically evaluated to prevent iron deficiency in patients. Acknowledgments We are very grateful to the patients who kindly contributed to this study. Inserm and Paris Diderot University, the Laboratory of excellence, GR-Ex, Paris, France, supported this work. Funding The labex GR-Ex, reference ANR-11-LABX-0051 is funded by the program “Investissements d’Avenir” of the French National Research Agency, reference ANR-11-IDEX-0005-02. We thank Nathalie Dessendier, Nicolas Ducrot and Olivier Thibaudeau for their technical support, Samira Benadda for the confocal microscope support, and Karima Yousfi for providing assistance in collecting the patient data.

2. Grabowski GA. Phenotype, diagnosis, and treatment of Gaucher’s disease. Lancet Lond Engl. 2008;372(9645):1263-1271. 3. Charrow J, Andersson HC, Kaplan P, et al. The Gaucher registry: demographics and disease characteristics of 1698 patients with Gaucher disease. Arch Intern Med. 2000;160(18):2835-2843.

4. Stein P, Yu H, Jain D, Mistry PK. Hyperferritinemia and iron overload in type 1 Gaucher disease. Am J Hematol. 2010;85(7):472–476. 5. Regenboog M, van Kuilenburg ABP, Verheij J, Swinkels DW, Hollak CEM. Hyperferritinemia and iron metabolism in Gaucher disease: Potential pathophysio-

595

T. Lefebvre et al.

10.

11.

12. 13.

14.

15.

16.

17.

596

logical implications. Blood Rev. 2016; 30(6):431-437 Franco M, Collec E, Connes P, et al. Abnormal properties of red blood cells suggest a role in the pathophysiology of Gaucher disease. Blood. 2013; 121(3):546555. Reihani N, Arlet J-B, Dussiot M, et al. Unexpected macrophage-independent dyserythropoiesis in Gaucher disease. Haematologica. 2016;101(12):1489-1498. Barton NW, Brady RO, Dambrosia JM, et al. Replacement therapy for inherited enzyme deficiency--macrophage-targeted glucocerebrosidase for Gaucher’s disease. N Engl J Med. 1991;324(21):1464-1470. Mekinian A, Stirnemann J, Belmatoug N, et al. Ferritinemia during type 1 Gaucher disease: mechanisms and progression under treatment. Blood Cells Mol Dis. 2012;49(1):53-57. Medrano-Engay B, Irun P, Gervas-Arruga J, et al. Iron homeostasis and infIammatory biomarker analysis in patients with type 1 Gaucher disease. Blood Cells Mol Dis. 2014;53(4):171-175. Lorenz F, Pawłowicz E, Klimkowska M, et al. Ferritinemia and serum inflammatory cytokines in Swedish adults with Gaucher disease type 1. Blood Cells Mol Dis. 2018;68:35-42. Finch CA, Bellotti V, Stray S, et al. Plasma ferritin determination as a diagnostic tool. West J Med. 1986;145(5):657-663. Nicolas G, Bennoun M, Devaux I, et al. Lack of hepcidin gene expression and severe tissue iron overload in upstream stimulatory factor 2 (USF2) knockout mice. Proc Natl Acad Sci USA. 2001; 98(15):8780-8785. Pigeon C, Ilyin G, Courselaud B, et al. A new mouse liver-specific gene, encoding a protein homologous to human antimicrobial peptide hepcidin, is overexpressed during iron overload. J Biol Chem. 2001; 276(11):7811-7819. Nemeth E, Tuttle MS, Powelson J, et al. Hepcidin regulates cellular iron efflux by binding to ferroportin and inducing its internalization. Science. 2004; 306(5704): 2090-2093. Delaby C, Pilard N, Gonçalves AS, Beaumont C, Canonne-Hergaux F. Presence of the iron exporter ferroportin at the plasma membrane of macrophages is enhanced by iron loading and down-regulated by hepcidin. Blood. 2005;106(12):3979-3984. Brasse-Lagnel C, Karim Z, Letteron P, et al. Intestinal DMT1 cotransporter is downregulated by hepcidin via proteasome internalization and degradation. Gastroenterology. 2011;140(4):1261-1271.

18. Mena NP, Esparza A, Tapia V, Valdés P, Núñez MT. Hepcidin inhibits apical iron uptake in intestinal cells. Am J Physiol Gastrointest Liver Physiol. 2008; 294(1):G192-198. 19. Nemeth E, Rivera S, Gabayan V, et al. IL-6 mediates hypoferremia of inflammation by inducing the synthesis of the iron regulatory hormone hepcidin. J Clin Invest. 2004; 113(9):1271-1276. 20. Weiss G, Goodnough LT. Anemia of chronic disease. N Engl J Med. 2005; 352(10):1011-1023. 21. Kulaksiz H, Theilig F, Bachmann S, et al. The iron-regulatory peptide hormone hepcidin: expression and cellular localization in the mammalian kidney. J Endocrinol. 2005;184(2):361–370. 22. Moulouel B, Houamel D, Delaby C, et al. Hepcidin regulates intrarenal iron handling at the distal nephron. Kidney Int. 2013; 84(4):756-766. 23. Gnana-Prakasam JP, Martin PM, Mysona BA, et al. Hepcidin expression in mouse retina and its regulation via lipopolysaccharide/Toll-like receptor-4 pathway independent of Hfe. Biochem J. 2008; 411(1):79-88. 24. Schwarz P, Kübler JAM, Strnad P, et al. Hepcidin is localised in gastric parietal cells, regulates acid secretion and is induced by Helicobacter pylori infection. Gut. 2012; 61(2):193-201. 25. Kulaksiz H, Fein E, Redecker P, et al. Pancreatic beta-cells express hepcidin, an iron-uptake regulatory peptide. J Endocrinol. 2008;197(2):241-249. 26. Merle U, Fein E, Gehrke SG, Stremmel W, Kulaksiz H. The iron regulatory peptide hepcidin is expressed in the heart and regulated by hypoxia and inflammation. Endocrinology. 2007;148(6):2663-2668. 27. Frazier MD, Mamo LB, Ghio AJ, Turi JL. Hepcidin expression in human airway epithelial cells is regulated by interferon-g. Respir Res. 2011;12:100. 28. Bekri S, Gual P, Anty R, et al. Increased Adipose Tissue Expression of Hepcidin in Severe Obesity Is Independent From Diabetes and NASH. Gastroenterology. 2006;131(3):788-796. 29. Peyssonnaux C, Zinkernagel AS, Datta V, et al. TLR4-dependent hepcidin expression by myeloid cells in response to bacterial pathogens. Blood. 2006;107(9):3727-3732. 30. Lefebvre T, Dessendier N, Houamel D, et al. LC-MS/MS method for hepcidin-25 measurement in human and mouse serum: clinical and research implications in iron disorders. Clin Chem Lab Med. 2015; 53(10):1557-1567. 31. Domellöf M, Braegger C, Campoy C, et al.

32.

33.

34.

35. 36.

37.

38. 39.

40.

41. 42. 43.

44.

Iron requirements of infants and toddlers. J. Pediatr. Gastroenterol. Nutr. 2014; 58(1):119-129. Nicolas G, Chauvet C, Viatte L, et al. The gene encoding the iron regulatory peptide hepcidin is regulated by anemia, hypoxia, and inflammation. J Clin Invest. 2002; 110(7):1037-1044. Nemeth E, Rivera S, Gabayan V, et al. IL-6 mediates hypoferremia of inflammation by inducing the synthesis of the iron regulatory hormone hepcidin. J Clin Invest. 2004; 113(9):1271-1276. Allen MJ, Myer BJ, Khokher AM, Rushton N, Cox TM. Pro-inflammatory cytokines and the pathogenesis of Gaucher’s disease: increased release of interleukin-6 and interleukin-10. QJM Mon. J Assoc Physicians. 1997;90(1):19-25. Barak V, Acker M, Nisman B, et al. Cytokines in Gaucher’s disease. Eur Cytokine Netw. 1999;10(2):205-210. Pandey MK, Burrow TA, Rani R, et al. Complement drives glucosylceramide accumulation and tissue inflammation in Gaucher disease. Nature. 2017; 543(7643):108-112. Boven LA, van Meurs M, Boot RG, et al. Gaucher cells demonstrate a distinct macrophage phenotype and resemble alternatively activated macrophages. Am J Clin. Pathol. 2004;122(3):359-369. Hunt JR. How important is dietary iron bioavailability? Am J Clin Nutr. 2001; 73(1):3-4. Bogoeva B, Petrusevska G. Immunohistochemical and ultrastructural features of Gaucher’s cells--five case reports. Acta Medica Croat. Cas. Hravatske Akad Med Znan. 2001;55(3):131-134. Lee RE, Balcerzak SP, Westerman MP. Gaucher’s disease. A morphologic study and measurements of iron metabolism. Am J Med. 1967;42(6):891-898. Lorber M. Adult-type Gaucher’s disease: a secondary disorder of iron metabolism. Mt Sinai J Med N Y. 1970;37(4):404-417. Bitton A, Etzell J, Grenert JP, Wang E. Erythrophagocytosis in Gaucher cells. Arch Pathol Lab Med. 2004;128(10):1191-1192. Bratosin D, Tissier J-P, Lapillonne H, et al. A cytometric study of the red blood cells in Gaucher disease reveals their abnormal shape that may be involved in increased erythrophagocytosis. Cytometry B Clin Cytom. 2011;80(1):28-37. Franco M, Collec E, Connes P, et al. Abnormal properties of red blood cells suggest a role in the pathophysiology of Gaucher disease. Blood. 2013; 121(3):546555.

haematologica | 2018; 103(4)

ARTICLE

Myelodysplastic Syndrome

Leptin-deficient obesity prolongs survival in a murine model of myelodysplastic syndrome

Ferrata Storti Foundation

Michael J. Kraakman,1,2* Helene L. Kammoun,1,3* Dragana Dragoljevic,1,3 Annas Al-Sharea,1,3 Man K.S. Lee,1,3 Michelle C. Flynn,1,3 Christian J. Stolz,1,3 Andrew A. Guirguis,4 Graeme I. Lancaster,1,3 Jaye Chin-Dusting,3 David J. Curtis4 and Andrew J. Murphy1,3 Baker Heart and Diabetes Institute, Melbourne, VIC, Australia; 2Naomi Berrie Diabetes Center and Department of Medicine, Columbia University, NY, USA; 3Monash University, Melbourne, VIC, Australia and 4Australian Centre for Blood Diseases, Monash University, Melbourne, VIC, Australia

*MJK AND HLK contributed equally to this work.

Haematologica 2018 Volume 103(4):597-606

ABSTRACT

besity enhances the risk of developing myelodysplastic syndromes. However, the effect of obesity on survival is unclear. Obese people present with monocytosis due to inflammatory signals emanating from obese adipose tissue. We hypothesized that obesity-induced myelopoiesis would promote the transition of myelodysplastic syndrome to acute myeloid leukemia and accelerate mortality in obesity. Obese Ob/Ob mice or their lean littermate controls received a bone marrow transplant from NUP98-HOXD13 transgenic mice, a model of myelodysplastic syndrome. The metabolic parameters of the mice were examined throughout the course of the study, as were blood leukocytes. Myeloid cells were analyzed in the bone, spleen, liver and adipose tissue by flow cytometry halfway through the disease progression and at the endpoint. Survival curves were also calculated. Contrary to our hypothesis, transplantation of NUP98-HOXD13 bone marrow into obese recipient mice significantly increased survival time compared with lean recipient controls. While monocyte skewing was exacerbated in obese mice receiving NUP98-HOXD13 bone marrow, transformation to acute myeloid leukemia was not enhanced. Increased survival of obese mice was associated with a preservation of fat mass as well as increased myeloid cell deposition within the adipose tissue, and a concomitant reduction in detrimental myeloid cell accumulation within other organs. The study herein revealed that obesity increases survival in animals with myelodysplastic syndrome. This may be due to the greater fat mass of Ob/Ob mice, which acts as a sink for myeloid cells, preventing their accumulation in other key organs, such as the liver.

Correspondence: andrew.murphy@baker.edu.au

Received: October 3, 2017. Accepted: January 19, 2018. Pre-published: January 25, 2018. doi:10.3324/haematol.2017.181958

Introduction Obesity represents a major health risk and is independently associated with the development of a cluster of disorders commonly referred to as metabolic diseases, including type 2 diabetes mellitus (T2DM), cardiovascular disease, stroke, neurodegeneration, and liver diseases. Furthermore, causal associations between body mass index (BMI) and many cancers are becoming increasingly apparent.1-4 In the last decade, obesity has been associated not only with most forms of tumor-based cancer, but also with hematological malignancies.5 Obese patients have an increased risk of leukemia,6 and, in children, excess fat mass is linked both to enhanced incidence and lower overall survival for leukemia.7 In addition, obesity has been increasingly associated with an enhanced risk for the incidence of myelodysplastic syndromes (MDS).8,9 MDS encompass a group of bone marrow disorders characterized by defective hematopoiesis.10 The risk of progressing to leukemia is high in individuals with MDS, with approximately 30% proceeding to develop acute myeloid leukemia (AML), an aggressive hematopoietic malignancy with a low 5-year survival prognosis. While the environmental factors for developing MDS and its progression to AML remain poorly understood, recent epidemiological studies have revealed haematologica | 2018; 103(4)

Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/597 ÂŠ2018 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.

597

M.J. Kraakman et al.

obesity to be associated with a small but significant increase in the risk of developing AML.11,12 While the exact biological mechanisms underpinning the increased leukemia risk in obesity are likely to be complex and multifactorial, the presence of chronic low-grade inflammation induced by obesity is thought to contribute to the increased cancer risk.13 Chronic low-grade inflammation is broadly characterized by alterations in circulating immuno-modulatory cytokines and leukocytes within the adipose tissue as well as the activation of stress pathways in metabolically important tissues.14 We and others have recently reported that obesity causes prominent monocytosis due to enhanced myelopoiesis and altered hematopoiesis driven by interleukin-1β (IL-1β).15,16 Interestingly, IL-1β, along with other myeloid promoting cytokines (i.e., IL-3, granulocyte-macrophage colonystimulating factor [GM-CSF]) has also been shown to promote the progression of AML.17 Thus, we sought to investigate the influence of obesity on the transition of MDS to AML and survival. We hypothesized that obesityinduced inflammation would promote the progression of MDS to AML through heightened myelopoiesis and hematopoietic stress.

Methods Detailed methods are available in Online Supplementary Methods 10-week-old male Ob/Ob mice, along with littermate lean controls, purchased from the Jackson Laboratory (USA), underwent a bone marrow transplant (BMT), receiving marrow from 6/8-week-old male wild-type (WT) C57bl/6 mice (Jackson Laboratory) or male NHD13 mice sourced from colonies maintained within the Alfred Medical Research Education Precinct (AMREP) Animal Centre. All animal experiments were approved by the AMREP Animal Ethics Committee and conducted in accordance with the National Health and Medical Research Council of Australia Guidelines for Animal Experimentation (Ethics E/1444/2014/B).

Results Obese mice exhibit a hematopoietic phenotype after seven months of myelodysplastic syndrome To determine the impact of obesity on the development of MDS, we performed bone marrow (BM) transplantation studies using NHD13 transgenic donor mice or WT littermate BM as a control into lean (Ob/+) and obese (Ob/Ob) recipient mice (Figure 1A). We considered the diet-induced obesity (DIO) model, but explicitly chose to conduct this experiment in Ob/Ob mice for the following reasons: i) Ob/Ob mice are guaranteed to maintain and gain weight following a BM transplant, which is in contrast to our prior experience whereby WT mice had limited weight gain post-BMT on a high fat diet, ii) the loss of leptin means these mice will feed consistently, and removes a variable of changes in feed patterns as the myelodysplasia progresses, iii) the diets are matched, ruling out effects of altered nutritional composition of standard chow and high fat diets, and iv) we have previously shown that Ob/Ob and DIO mice display enhanced myelopoiesis through the same mechanism (i.e., increased IL-1β production emanating from the adipose 598

tissue).15 We acknowledge that leptin plays a role in hematopoiesis, but this is mainly restricted to the lymphoid system,18 which is suppressed in MDS and not likely to be a confounding factor in our experiment. Moreover, we have shown that supplementing Ob/Ob mice with leptin, while causing weight reduction, had no impact on myelopoiesis.15 The MDS model we chose to employ was the NHD13 transgenic mouse, which overexpresses a NUX98HOXD13 fusion protein that has been associated with human MDS.19 These mice have an MDS phenotype that can develop into AML with a penetrance of ∼30% within 14 months.19 Accordingly, we initially assessed the impact of obesity on hematopoiesis, and specifically myelopoiesis, seven months post-transplantation with BM from NHD13 transgenic donor mice, a time point at which all mice remained alive (i.e., the predicted halfway survival point of the model). Total blood cell counts revealed the expected MDS features of decreased white blood cells (WBC) and red blood cells (RBC), effects that were observed in both lean and obese MDS mice (Figure 1B). As expected, the decrease in WBC counts was primarily due to lymphopenia (Figure 1C). Platelets were also significantly decreased in both lean and obese MDS mice. These changes occurred despite obese mice presenting with increased platelet numbers in the healthy state (Figure 1D). Both lean and obese MDS mice failed to compensate this thrombocytopenia with enhanced platelet production, as suggested by the increased percentage of newly formed reticulated platelets (Figure 1D). Next, we analyzed numbers of circulating myeloid cells to determine whether obesity influenced myelopoiesis. Consistent with our previous findings, obesity induced a significant increase in the proportion of monocytes when compared with lean controls, and this effect was strongly amplified in the presence of MDS (Figure 1E). Both Ly6Chi and Ly6-Clo subsets contributed to this increase in the proportion of monocytes (Figure 1E). Interestingly, while the proportion of neutrophils was elevated in the lean NHD13/WT mice, this effect was absent in obese MDS mice (Figure 1E). Consistent with the known phenotype of MDS mice, defective hematopoiesis in BM progenitor cells was evident, with a decrease in the abundance and proliferation of all of the hematopoietic stem and progenitor cells occurring in both lean and obese MDS mice (Figure 1F,G). The obesity-induced increase in circulating monocytes could be explained by extramedullary myelopoiesis in the MDS mice, made apparent by increased hematopoietic stem and progenitor cells (HSPCs, also referred to as LSKs), common myeloid progenitors (CMPs), granulocyte-macrophage progenitors (GMPs), and megakaryocyte-erythroid progenitors (MEPs) (Figure 1H). Consequently, a significant accumulation of CD11b+ myeloid cells and F4/80+ macrophages was also observed in the spleen (Figure 1I). Of note, the MDS pathology did not appear to significantly affect the metabolic function of lean or obese mice relative to their littermate controls (Online Supplementary Figure S1). Taken together, these data, at the half-way point in the pathogenesis of this model, demonstrated an enhanced monocyte response in obese MDS mice compared with lean MDS mice. Further, failing BM hematopoiesis appeared to promote a shift of this process to the spleen in the mice suffering from MDS. haematologica | 2018; 103(4)

Obesity and myelodysplasia

Figure 1. Ob/Ob mice display a similar hematopoietic phenotype to wild-type mice when challenged with myelodysplastic syndrome despite pre-existing monocytosis. (A) Experimental overview: Ob/Ob mice and WT littermate controls transplanted with either WT or NHD13 bone marrow (BM) were followed until the development of MDS symptoms required euthanasia. Seven months after the bone marrow transplant, mice were bled for analysis (B-E) and a subset was culled for tissue analysis (F-I). (B) Blood counts obtained by CBC, (C) flow cytometry analysis of lymphocytes and (D) CBC platelet counts and flow cytometry analysis of reticulated platelets (% platelets). (E) myeloid cells analysis by flow cytometry on lysed blood. (F) Flow cytometry analysis of BM cells including long term stem cells and myeloid progenitors, and (G) cell cycle analysis (DAPI). (H-I) Flow cytometry analysis of spleen immune cells including (H) long term stem cells and myeloid progenitors and (I) myeloid populations. (A-E); n=12-16; (F-I); n=3. All data expressed as mean Âą SEM. *P<0.05, for obesity effect; #P<0.05, for MDS effect as analyzed by 2-way ANOVA. WT: wild-type; WBC: white blood cell; RBC: red blood cell; HSPC: hematopoietic stem and progenitor cell; CMP: common myeloid progenitor; GMP: granulocyte-macrophage progenitor; MEP: megakaryocyte-erythroid progenitor.

haematologica | 2018; 103(4)

599

M.J. Kraakman et al.

Obese mice display prolonged survival when challenged with myelodysplastic syndrome despite increased myelopoiesis As the obese mice presented with increased myelopoiesis with increased splenic macrophages in the setting of MDS, we hypothesized that these mice would have reduced survival compared with the lean MDS mice. However, to our surprise, obese MDS mice had a significantly prolonged survival compared with their lean MDS counterparts (Figure 2A). Obesity appeared to influence the cause of death, by promoting the development of a chronic myelomonocytic leukemia (CMML) at the expense of T-cell acute lymphoblastic leukemia (T-ALL) and cytopenias (Figure 2B). Overall, the obese MDS mice lived an average of 100 days longer than the lean mice, a survival advantage that occurred despite an increase in the proportion of circulating monocytes, driven primarily by the inflammatory Ly6-Chi subset, this being consistently higher compared with those in the lean MDS counterparts (Figure 2C-E). Interestingly, while the proportion of neutrophils were higher in the lean NHD13 transplanted

600

mice, neutrophils in obese MDS mice were comparable to mice without MDS (Figure 2F). It is unlikely that differences in circulating T cells or B cells could account for the improvement in longevity, as obese MDS mice were not protected from a sustained lymphopenia induced by the NHD13 BM compared with lean mice (Figure 2G,H). A similar situation was observed when looking at the total numbers of these circulating cells, with obese MDS mice having a trend to higher blood monocytes, driven by the Ly6-Chi subset (Online Supplementary Figure S2A-D). We also examined the abundance of ckit+ cells in the blood, which were unchanged, suggesting that refractory anemia with excess blasts was not occurring (Online Supplementary Figure S2E,F).

Endpoint blood and spleen characteristics To explore why prolonged survival was observed in obese MDS mice relative to their lean counterparts, we first analyzed the key features of end-stage disease for MDS. Once the mice began to show the characteristic signs of the terminal stage of the disease, euthanasia was

Figure 2. Ob/Ob mice display prolonged survival when challenged with myelodysplastic syndrome despite preexisting monocytosis. Ob/Ob mice and WT littermate controls transplanted with either WT or NHD13 BM were followed for (A) Kaplan-Meier survival curve. (B) Proportion of disease contributing to death. Circulating myeloid cell populations, including (C) total monocytes, (D) Ly-6Chi monocytes, (E) Ly-6Clo monocytes and (F) neutrophils analyzed by flow cytometry on lysed blood. Circulating lymphoid populations, including (G) B cells and (H) T cells analyzed by flow cytometry on lysed blood. (A,B); n=16; (C-H); n=12-16. All data expressed as mean Âą SEM. WT: wild-type; MDS: myelodysplastic syndrome; T-ALL: T-cell acute lymphoblastic leukemia; CMML: chronic myelomonocytic leukemia; AML: acute myeloid leukemia; BMT: bone marrow transplant.

haematologica | 2018; 103(4)

Obesity and myelodysplasia

performed alongside healthy lean and obese controls for comparative analysis. Consistent with the data obtained throughout the course of the disease, monocytosis was observed in lean MDS mice, and this effect was exacerbated in the obese MDS mice (Figure 3A). The inefficient hematopoiesis which is characteristic of MDS occurred at similar levels in lean and obese MDS mice, both of which presented with severe lymphopenia, reduced circulating progenitor cells, and anemia (Figure 3B-F). Anemia was

accompanied by macrocytosis, as demonstrated by the significantly increased mean corpuscular volume, indicative of RBC volume (Figure 3G). Platelet counts were also reduced with MDS, irrespective of obesity status (Figure 3H). Finally, prominent splenomegaly occurred in MDS mice, consistent with the monocytosis and aberrant extramedullary myelopoiesis observed at the sevenmonth time point (Figure 3I,J; Figure 1E,H). Consistent with MDS progression, and irrespective of body weight,

Figure 3. Ob/Ob mice exhibit a similar disease phenotype at MDS/AML endpoint despite aggravated monocytosis. Ob/Ob mice and WT littermate controls transplanted with either WT or NHD13 bone marrow (BM) were followed until the development of MDS symptoms required euthanasia. Blood flow cytometry analysis of (A) myeloid cells, (B) lymphocytes and (C) progenitor cells. CBC analysis of (D) red blood cells, (E) hemoglobin, (F) hematocrit, (G) mean corpuscular volume and (H) platelets. (I) Spleen weights and representative images, scale 0.5cm. (J) Flow cytometry analysis of HSPCs and myeloid progenitors in the BM. (K) Representative images of BM from lean and Ob/Ob mice, arrows indicate dilated blood vessels, A indicates adipocytes. (L) Flow cytometry analysis of HSPCs in the spleen. (A-I); n=11-16; (J-L); n=4-8. All data expressed as mean Âą SEM. *P<0.05, for obesity effect; #P<0.05, for MDS effect as analyzed by 2-way ANOVA. WT: wild-type; RBC: red blood cell; HSPC: hematopoietic stem and progenitor cell; CMP: common myeloid progenitor; GMP: granulocyte-macrophage progenitor; MEP: megakaryocyte-erythroid progenitor; MCV: mean corpuscular volume.

haematologica | 2018; 103(4)

601

M.J. Kraakman et al.

we observed reduced stem cells in the BM with fewer HSPCs and GMPs (Figure 3K). Interestingly, obese mice transplanted with NHD13 BM showed a trend towards restored CMP and MEP populations, despite terminal stage of the disease (Figure 3L). Next, we compared the gross morphology of the BM between the groups. While processing the bones in the obese NHD13 mice we observed two distinct phenotypes, which we traced back to mice dying due to MDS or transformation to AML. Thus, in the NHD13 mice we took the opportunity to explore the phenotype amongst mice that had died of either MDS or AML (Figure 3L). As expected, the BM from lean control mice appeared normal. When we explored the marrow of lean NHD13 mice with AML, we noted characteristic crowding of the marrow and dilated vessels, with an abundant number of cells that appeared to be leaving the marrow, compared to the NHD13 mice with MDS which presented with more disperse marrow. Again, as expected, adipocytes were more abundant in the BM of obese control mice compared with lean control mice, along with more megakaryocytes, as we have previously described.20 However, the most notable change in the overall morphology was observed in the NHD13 obese mice. Those that died from AML had fewer adipocytes and more cellular marrow compared to the obese control mice, suggesting that the leukemic cells had used the lipid stored within these adipocytes for energy,

similar to the interaction between HSPCs and adipocytes post-BMT.21 Interestingly, NHD13 obese mice that died from MDS had almost a complete lack of cells in the BM, which was loaded with adipocytes, suggesting that hematopoiesis was likely being supported by another organ, and at this end-stage of disease showed complete BM failure. In addition, stem cells and progenitor cells that were increased in the spleen at the seven-month time point appeared to have become exhausted and were rapidly maturing into myeloid cells (i.e., more GMPs; Figure 3L). Together, these data demonstrate a similar profile of hematological changes in obese and lean MDS mice at disease endpoint, with a clear difference being the adiposity observed in the marrow of the obese NHD13 mice that died due to MDS.

Obese mice present with remodeled adipose tissue in response to MDS Given that the blood profile did not provide an explanation for the prolonged survival of the obese mice with MDS, we sought an alternate explanation. Interestingly, when analyzing tissues after sacrifice, we noted that lean MDS mice had lost almost all of their body fat (Figure 4A). In contrast, the obese mice, which started with substantially more fat mass, retained the majority of their adiposity when facing MDS. Indeed, when expressed as a percentage relative to mice without MDS, lean MDS mice

Figure 4. Adipose tissue parameters in Ob/Ob mice subjected to MDS. Ob/Ob mice and WT littermate controls transplanted with either WT or NHD13 BM were followed until the development of MDS symptoms required euthanasia. After death, adipose tissues were dissected and analyzed. (A) Fat and lean mass measured by EchoMRI. (B) Percentage weight loss relative to respective control groups. (C) Representative H&E stained section of visceral adipose tissue (VAT). (D) Flow cytometry analysis of VAT stromal vascular fraction, including percentages of activated myeloid cells (CD11b+), macrophages (F4/80+) and pro-inflammatory macrophages (CD11c+). (E) Adipocyte number analysis and (F) adipocyte size analysis using Image Pro J. (G) PicroSirius red staining of VAT for collagen visualization in polarized light. (A-B); n=9-16; (C-G); n=4-8. All data expressed as mean Âą SEM. *P<0.05, for obesity effect; #P<0.05, for MDS effect as analyzed by 2-way ANOVA. WT: wildtype.

602

haematologica | 2018; 103(4)

Obesity and myelodysplasia

had lost almost 75% of their fat mass at the time of sacrifice, while obese mice had only lost 20% of their fat mass overall, thus maintaining a stable percentage of body fat mass (Figure 4B, Online Supplementary Figure S3). Of note, examining endpoint lean mass revealed that obesity did not protect against muscle cachexia, as lean and obese

MDS mice lost comparable amounts of lean mass (Figure 4A,B). However, at the 37-week time point, when all the lean NHD13 mice had died, the obese NHD13 still had lean mass equal to control WT mice, suggesting that their lean mass had not fallen to dire levels (Online Supplementary Figure S3B). Exploring the morphological

Figure 5. Spleen and liver parameters in response to MDS/AML in Ob/Ob mice. Spleen analysis by flow cytometry including (A) monocytes and neutrophils, (B) macrophages (F4/80+) and (C) activated myeloid cells (CD11b+). (D) Histological H&E staining of the spleen. Liver analysis by (E) histological H&E analysis, (F) flow cytometry analysis of hepatic immune cells, including activated myeloid cells (CD11b+) and macrophages (F4/80+) and (G) PicroSirius red staining of liver for collagen visualization. Plasma analysis of (H) liver enzymes alanine amino transferase (ALT) and aspartate amino transferase (AST) and (I) bilirubin. (A-G); n=4-8; (H-I); n=57. All data expressed as mean Âą SEM. *P<0.05, for obesity effect; #P<0.05, for MDS effect as analyzed by 2-way ANOVA. WT: wild-type.

haematologica | 2018; 103(4)

603

M.J. Kraakman et al.

phenotype of the epididymal fat pads, there was no major differences in gross morphology between the lean control and lean MDS mice. However, a massive infiltration of small nucleated cells surrounded most adipocytes in obese mice with MDS compared to control obese mice (Figure 4C). Using flow cytometry to identify this cell infiltrate, we observed significantly more macrophages, particularly CD11c+ pro-inflammatory macrophages, in the adipose tissue stromal vascular fraction of obese mice, but this was not impacted by MDS status (Figure 4D). However, the obese NHD13 transplanted mice had a significant and selective increase in CD11b+ cells, suggesting the accumulation of activated myeloid cells (Figure 4D). To reaffirm that the obese mice were unlikely to be prone to extramedular forms of leukemia (i.e., we saw no increase in circulating blood ckit+ cells; Online Supplementary Figure 2E,F), we performed an ex vivo migration assay whereby isolated ckit+ cells from WT or NHD13 mice were allowed to migrate to conditioned media from lean or obese mice. This revealed that independent of genotype, there was a suppressed migratory response to obese fat compared to lean, suggesting that ckit+ cells were not being encouraged to migrate to the obese adipose tissue and evolve into leukemic cells (Online Supplementary Figure 3D). The recruitment of immune cells into the adipose has been associated with fat tissue remodeling, thus we set out to analyze adipocyte characteristics.22 Quantification of adipocyte size revealed that the visceral adipose tissue (VAT) from obese MDS mice contained more smaller adipocytes and fewer large adipocytes, signifying extensive remodeling of the fat tissue in response to MDS, with a similar trend in the lean mice (Figure 4E,F). Consistent with increased immune cell recruitment and adipose tissue remodeling, we observed increased VAT fibrosis in the

obese MDS animals when looking at collagen staining (Figure 4C-G). Given this data, we hypothesize that one mechanism by which obesity prolongs survival in mice with MDS is through a preservation of fat mass.

Impact of MDS on spleen and liver immune cell populations in WT and Ob/Ob mice Given the increased recruitment of activated myeloid cells in the VAT of obese MDS mice and their prolonged survival, we hypothesized that the increased recruitment of myeloid cells to VAT may spare the recruitment of these cells to other organs. In the spleen, consistent with the splenomegaly and similar to the disease profile at seven months, we still observed a striking increase in Ly6Chi and Ly6-Clo monocyte populations in both lean and obese MDS mice, consistent with MDS characteristics (Figure 3I and Figure 5A). Interestingly, the strong neutrophil accumulation we noted was restricted to the lean MDS mice, as obese MDS mice showed no increase in splenic neutrophils (Figure 5A). As observed in the VAT, and consistent with the obese phenotype, F4/80+ macrophages were increased in the spleens of obese mice, and there was a trend for MDS to potentiate this profile (Figure 5B). Strikingly, contrary to the VAT profile, there was a 7-fold increase in the abundance of splenic CD11b+ cells observed in lean MDS conditions compared to only a doubling in the obese animals (Figure 5C). This enhanced myeloid cell infiltration, particularly in the lean MDS mice, was supported by the gross morphology of the spleen (Figure 5D). Overall, it appeared that obese mice were partly protected from splenic myeloid cell accumulation. This prompted us to explore immune cell recruitment in the liver, where macrophages tend to home in on a con-

Figure 6. Schematic overview of the proposed mechanism of enhanced survival in the obese mice transplanted with NHD13 bone marrow. MDS: myelodysplastic syndrome; AML: acute myeloid leukemia.

604

haematologica | 2018; 103(4)

Obesity and myelodysplasia

text of obesity. Exploring the liver, initially at a macroscopic level, it appeared that MDS resulted in the preferential accumulation of immune cells in lean mice, and to a lesser extent in the obese MDS mice (Figure 5E). Quantitative flow cytometry data confirmed this observation, where MDS resulted in similar F4/80+ macrophages in lean and obese hepatic tissue, but significantly more hepatic CD11b+ myeloid cells in the lean animals (Figure 5F). These findings were correlated with the observed liver fibrosis, which appeared to be more prevalent in the livers of lean MDS mice (Figure 5G). Interestingly, the fatty liver phenotype observed in obese mice (Figure 5E) was associated with an increase in circulating liver enzymes aspartate transaminase (AST)/ alanine transaminase (ALT) (Figure 5H). Although MDS did not influence ALT levels, lean MDS mice presented with increased AST levels when obese MDS mice showed no difference with their obese healthy counterparts (Figure 5H). Interestingly, in support of impaired liver function, lean MDS mice had higher levels of bilirubin, indicative of impaired uptake and degradation by the liver of these mice (Figure 5I), effects that were not apparent in the obese MDS mice. Finally, we assessed the basal levels of circulating creatine kinase as a proxy measure of muscle damage. This tended to be higher in the lean MDS mice, suggesting a possible impact of MDS on muscle integrity which did not appear to occur in their obese counterparts (Online Supplementary Figure S4). Overall, these data support the idea that the obese VAT preferentially attracts specific myeloid cell populations in MDS mice which results in a concomitant decrease in the accumulation of these cells in other tissues, such as the spleen and liver.

Discussion Obesity has become increasingly associated with cancer and is now recognized as a risk factor for many malignant pathologies.5,23 In addition to solid tumors, obesity has also been linked with different forms of leukemia.11 Obesity is associated with an increased prevalence of MDS, however the data is less clear regarding survival outcome in obese patients presenting with MDS. We have previously demonstrated that the obese adipose tissue interacts with the BM, and promotes monocytosis through the stimulation of the myeloid pathway via IL-1β. In this context, we hypothesized that this increased basal rate of myelopoiesis would contribute to MDS and the progression to AML, thereby decreasing survival. Surprisingly, while obesity-induced myelopoiesis was observed in the setting of MDS, this was associated with significantly improved survival. We hypothesize that the expanded adipose tissue in obese mice acts as a sink for the increased myeloid cells, sparing other vital organs from myeloid cell burden and subsequent dysfunction. Furthermore, we propose that the preservation of fat mass observed in obese MDS mice likely contributes to their survival advantage relative to their lean counterparts (Figure 6). Our paradoxical findings highlight that obesity might not always be associated with enhanced mortality risk in people suffering from hematological disorders. One major difference between our pre-clinical data and the clinical course of MDS resides in the fact that patients with this haematologica | 2018; 103(4)

disease may be treated with transfusions, chemotherapy, including epigenetic modifiers, and occasionally BMT. In general, studies supporting an obesity paradox in cancer tend to describe a ‘U-shape’ correlation between BMI and overall survival.24 This could indicate that increased energy stores in overweight and moderately obese patients could allow for longer survival, but severely obese patients with comorbidities would be at risk of decreased survival rate. Consistent with this obesity paradox, there are accumulating reports demonstrating that adipose tissue can be critical for patient’s health and survival. In particular, subcutaneous adipose tissue has been associated with increased survival in several conditions, including the hematological disorder multiple myeloma.25-27 Adipose tissue is the main energy store in the human body, hence, conserving enough fat stores could lead to better survival outcomes by allowing slower rates of muscle proteolysis, another important source of energy. In addition, the adipose tissue has now been fully recognized as an endocrine organ impacting various bodily functions. Indeed, depletion of white adipose tissue has been associated with increased inflammatory signalling and disrupted circadian regulation.28 Low levels of adiponectin, one of the main hormones secreted by the VAT, has been associated with an increased risk of cancer and poor diagnosis.29 Thus, the maintenance of fat stores in the Ob/Ob mice transplanted with NHD13 may play a crucial role in their prolonged survival compared to the lean MDS that lose the majority of their VAT. In the obese mice with MDS we found cells homed in to the VAT in significant numbers. In the absence of treatment, it appeared that the preferential homing of cells to the adipose tissue could partially protect other organs from infiltration, providing an explanation for the prolonged survival of obese mice when confronted with NHD13-induced MDS. Transformation from MDS to AML takes place in the stem and progenitor cells, not mature myeloid cells. Interestingly, transformation could have occurred in the obese VAT and at the same time prevented outgrowth of the leukemia, keeping these cells somewhat dormant. Prescience comes from the discovery that leukemic stem cells can reside within the VAT in a quiescent nature, conferring their protection from chemotherapy.30,31 One striking observation was the different cellular makeup of the BM between WT and obese mice that died of MDS. While the marrow of the WT mice that died of MDS displayed classical signs of dysplasia, with fewer hematopoietic cells, the marrow from the obese MDS mice was full of adipocytes, with few hematopoietic cells evident. Whether marrow adiposity alters the course of the disease requires further investigation. Of note, the bones of lean and obese mice that died from AML looked similar, thus there is a possibility that the leukemic cells, upon transition from MDS to AML, could have used this stored lipid in the early stages of proliferation as a source of energy, which in the lean mice would have come from peripheral organs (i.e., adipose tissue). The role of BM adipocytes is not well described, but was recently shown to play an important role in hematopoietic regeneration following ablation via chemotherapy or irradiation.21 Given the blood profile of the obese MDS mouse, where there was an increase in the Ly6-Chi monocytes, an early transition to CMML could be occurring; however, this hypothesis requires further investigation. Moreover, the 605

M.J. Kraakman et al.

mice that were deemed to have died from cytopenias could have potentially developed aplastic anemia, as the morphology of the bone marrow and numbers of circulating cells is similar to that described in this disease. This study has a number of limitations which should be taken into account when interpreting our findings. Firstly, we used a genetic model of obesity, where leptin is deficient, thus causing hyperphagia. While this model is key in maintaining adiposity, it does not reflect a scenario where a change in diet drives obesity, in which, for example, changes in lipids may influence the transformation of the disease. Additionally, a role for leptin in the evolution of MDS cannot be excluded. Further, we opted to perform BMTs, as opposed to crossing the NHD13 mice with the Ob/Ob mice. Whether the hematopoietic stress associated with transplantation and engraftment altered the course of the disease is probably unlikely, but should be kept in mind. Finally, we only tested one model of MDS; whether this holds true in other models is yet to be determined. In the study herein, we have demonstrated that obesity confers a survival advantage when mice are confronted with NHD13-induced MDS. It appears that the increased adiposity allows for dampening of the systemic leukemic

References 1. Khandekah MJ, Cohen P, Spiegelman BM. Molecular mechanisms of cancer development in obesity. Nat Rev Cancer. 2011; 11(12):886-895. 2. Kahn SE, Hull RL, Utzschneider KM. Mechanisms linking obesity to insulin resistance and type 2 diabetes. Nature. 2006;444(7121):840-846. 3. Hildreth KL, Van Pelt RE, Schwartz RS. Obesity, insulin resistance, and Alzheimer's disease. Obesity (Silver Spring). 2012; 20(8):1549-1557. 4. Masuoka HC, Chalasani N. Nonalcoholic fatty liver disease: an emerging threat to obese and diabetic individuals. Ann N Y Acad Sci. 2013;1281:106-122. 5. Calle EE, Rodriguez C, Walker-Thurmond K, Thun MJ. Overweight, obesity, and mortality from cancer in a prospectively studied cohort of U.S. adults. N Engl J Med. 2003;348(17):1625-1638. 6. Larsson SC, Wolk A. Overweight and obesity and incidence of leukemia: a meta-analysis of cohort studies. Int J Cancer. 2008;122(6):1418-1421. 7. Orgel E, Genkinger JM, Aggarwal D, Sung L, Nieder M, Ladas EJ. Association of body mass index and survival in pediatric leukemia: a meta-analysis. Am J Clin Nutr. 2016;103(3):808-817. 8. Ma X, Lim U, Park Y, et al. Obesity, lifestyle factors, and risk of myelodysplastic syndromes in a large US cohort. Am J Epidemiol. 2009;169(12):1492-1499. 9. Murphy F, Kroll ME, Pirie K, Reeves G, Green J, Beral V. Body size in relation to incidence of subtypes of haematological malignancy in the prospective Million Women Study. Br J Cancer. 2013;108(11):2390-2398. 10. Tefferi A, Vardiman JW. Myelodysplastic syndromes. N Engl J Med. 2009; 361(19):1872-1885. 11. Poynter JN, Richardson M, Blair CK, et al.

606

12.

13.

14. 15.

16.

17.

18.

19.

20.

21.

insult, as leukemic cells preferentially home to the adipose tissue, protecting vital organs such as the liver and spleen. Additionally, maintaining adequate fat stores per se may also contribute to the improved survival observed in obese MDS mice. Our findings support the growing literature suggesting that despite increased incidences of MDS and AML in obese patients, their overall survival may not be different to lean patients and, in some instances, may even be prolonged. Thus, taken together with the recent findings of Carey et al.,17 it may also be important to understand the cytokine profile of the patients before treatment, to ensure more effective eradication of the leukemia and to promote the restoration of normal hematopoiesis (i.e., inhibiting IL-1Î˛ in obese patients). Funding This work was supported by NHMRC grants (APP1083138, APP1106154 and APP1142938) to AJM. MJK is a Russell Berrie Foundation Scholar in Diabetes Research from the Naomi Berrie Diabetes Centre. AJM is supported by a Career Development Fellowship from the NHMRC (APP1085752), a Future Leader Fellowship from the National Heart Foundation (100440) and a Centenary Award from CSL.

Obesity over the life course and risk of acute myeloid leukemia and myelodysplastic syndromes. Cancer Epidemiol. 2016; 40:134140. Castillo JJ, Reagan JL, Ingham RR, et al. Obesity but not overweight increases the incidence and mortality of leukemia in adults: a meta-analysis of prospective cohort studies. Leuk Res. 2012;36(7):868-875. Khandekar MJ, Cohen P, Spiegelman BM. Molecular mechanisms of cancer development in obesity. Nat Rev Cancer. 2011;11(12):886-895. Hotamisligil GS. Inflammation and metabolic disorders. Nature. 2006; 444(7121):860867. Nagareddy PR, Kraakman M, Masters SL, et al. Adipose tissue macrophages promote myelopoiesis and monocytosis in obesity. Cell Metab. 2014;19(5):821-835. Singer K, DelProposto J, Morris DL, et al. Diet-induced obesity promotes myelopoiesis in hematopoietic stem cells. Mol Metab. 2014;3(6):664-675. Carey A, Edwards DKt, Eide CA, et al. Identification of Interleukin-1 by functional screening as a key mediator of cellular expansion and disease progression in acute Myeloid leukemia. Cell Rep. 2017; 18(13):3204-3218. Bennett BD, Solar GP, Yuan JQ, Mathias J, Thomas GR, Matthews W. A role for leptin and its cognate receptor in hematopoiesis. Curr Biol. 1996;6(9):1170-1180. Lin YW, Slape C, Zhang Z, Aplan PD. NUP98-HOXD13 transgenic mice develop a highly penetrant, severe myelodysplastic syndrome that progresses to acute leukemia. Blood. 2005;106(1):287-295. Kraakman MJ, Lee MK, Al-Sharea A, et al. Neutrophil-derived S100 calcium-binding proteins A8/A9 promote reticulated thrombocytosis and atherogenesis in diabetes. J Clin Invest. 2017;127(6):2133-2147. Zhou BO, Yu H, Yue R, et al. Bone marrow adipocytes promote the regeneration of

22.

23.

24. 25.

26.

27.

28.

29.

30.

31.

stem cells and haematopoiesis by secreting SCF. Nat Cell Biol. 2017;19(8):891-903. Crewe C, An YA, Scherer PE. The ominous triad of adipose tissue dysfunction: inflammation, fibrosis, and impaired angiogenesis. J Clin Invest. 2017;127(1):74-82. Donohoe CL, Lysaght J, O'Sullivan J, Reynolds JV. Emerging concepts linking obesity with the hallmarks of cancer. Trends Endocrinol Metab. 2017;28(1):46-62. Lennon H, Sperrin M, Badrick E, Renehan AG. The obesity paradox in cancer: a review. Curr Oncol Rep. 2016;18(9):56. Lindauer E, Dupuis L, Muller HP, Neumann H, Ludolph AC, Kassubek J. Adipose tissue distribution predicts survival in amyotrophic lateral sclerosis. PLoS One. 2013; 8(6):e67783. Takeoka Y, Sakatoku K, Miura A, et al. Prognostic effect of low subcutaneous adipose tissue on survival outcome in patients with Multiple Myeloma. Clin Lymphoma Myeloma Leuk. 2016; 16(8):434-441. Antoun S, Bayar A, Ileana E, et al. High subcutaneous adipose tissue predicts the prognosis in metastatic castration-resistant prostate cancer patients in post chemotherapy setting. Eur J Cancer. 2015;51(17):25702577. Tsoli M, Schweiger M, Vanniasinghe AS, et al. Depletion of white adipose tissue in cancer cachexia syndrome is associated with inflammatory signaling and disrupted circadian regulation. PLoS One. 2014;9(3): e92966. Katira A, Tan PH. Evolving role of adiponectin in cancer-controversies and update. Cancer Biol Med. 2016;13(1):101119. Ye H, Adane B, Khan N, et al. Adipose tissue functions as a reservoir for leukemia stem cells and confers chemo-resistance. Blood. 2015;126(23):845-845. Ye H, Adane B, Khan N, et al. Leukemic stem cells evade chemotherapy by metabolic adaptation to an adipose tissue niche. Cell Stem Cell. 2016;19(1):23-37.

haematologica | 2018; 103(4)

ARTICLE

Myeloproliferative Disorders

Ronan Le Calloch,1,2 Karine Lacut,3,4,5 Christelle Le Gall-Ianotto,6 Emmanuel Nowak,3 Morgane Abiven,3 Adrian Tempescul,2,7 Florence Dalbies,2,7 Jean-Richard Eveillard,2,7 Valérie Ugo,8 Stéphane Giraudier,9 Gaëlle Guillerm,2,7 Eric Lippert,10 Christian Berthou2,7 and Jean-Christophe Ianotto2,4,7

Ferrata Storti Foundation

Haematologica 2018 Volume 103(4):607-613

Service de Médecine Interne-Maladies du Sang-Maladies Infectieuses (MIIS), CHIC de Quimper; 2Fédération Inter Hospitalière d’Immuno–Hématologie de Bretagne Occidentale (FIHBO); 3CIC 1412, INSERM, Brest; 4EA3878 G.E.T.B.O, Université de Bretagne Occidentale, Brest; 5Département de Médecine Interne et Pneumologie, CHRU de Brest; 6 Laboratoire Interactions Epitheliums-Neurones, EA 4685, Université de Bretagne Occidentale; 7Service d’Hématologie Clinique, Institut de Cancérologie et Hématologie, CHRU de Brest; 8Laboratoire d’Hématologie, CHU d’Angers; 9Laboratoire d'Hématologie, Hôpital St-Louis, AP-HP, Paris and 10Laboratoire d’Hématologie, CHRU de Brest and Equipe ECLA, INSERM U1078, Université de Bretagne Occidentale, Brest, France 1

ABSTRACT

he purpose of this study was to identify the incidence, causes and impact of non-adherence to oral and subcutaneous chronic treatments for patients with polycythemia vera or essential thrombocythemia. Patients receiving cytoreductive drugs for polycythemia vera or essential thrombocythemia were recruited at our institution (Observatoire Brestois des Néoplasies Myéloprolifératives registry). They completed a one-shot questionnaire designed by investigators (Etude de l’Observance Thérapeutique et des Effets Secondaires des Traitements study). Data about complications (thrombosis, transformation and death) at any time in the patient’s life (before diagnosis, up until consultation and after the completion of the questionnaire) were collected. Sixty-five (22.7%) of 286 patients reported poor adherence (<90%) to their treatment with cytoreductive drugs and 46/255 /18%) also declared non-adherence to antithrombotic drugs. In total, 85/286 patients (29.7%) declared they did not adhere to their treatment. Missing an intake was rare and was mostly due to forgetfulness especially during occupational travel and holidays. Patients who did not adhere to their treatment were characterized by younger age, living alone, having few medications but a high numbers of pills and determining their own schedule of drug intake. Having experienced thrombosis or hematologic evolution did not influence the adherence rate. Non-adherence to oral therapy was associated with a higher risk of phenotypic evolution (7.3 versus 1.8%, P=0.05). For patients treated for polycythemia vera or essential thrombocythemia, non-adherence to cytoreductive and/or antithrombotic therapies is frequent and is influenced by age, habitus and concomitant treatments, but not by disease history or treatment side effects. Phenotypic evolution seems to be more frequent in the non-adherent group. (ClinicalTrials.gov #NCT02893410, #NCT02897297).

haematologica | 2018; 103(4)

Correspondence: jean-christophe.ianotto@chu-brest.fr

Received: September 11, 2017. Accepted: December 15, 2017. Pre-published: December 15, 2017.

doi:10.3324/haematol.2017.180448 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/607 ©2018 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.

607

R. Le Calloch et al.

Introduction Polycythemia vera (PV) and essential thrombocythemia (ET) are myeloproliferative neoplasms (MPN) arising from the clonal expansion of a multipotent hematopoietic stem cell, causing deregulated proliferation of myeloid lineages. Therapeutic management of these chronic pathologies has two objectives: reduction of the thrombotic risk induced by blood hyperviscosity (short-term) and reduction of the risk of transformation into myelofibrosis or acute myeloid leukemia (long-term).1 Available treatments are administered orally (hydroxycarbamide, pipobroman, anagrelide or ruxolitinib) or subcutaneously (pegylated interferon), usually in combination with antithrombotic drugs. The problem of lack of adherence to treatment is likely as old as the practice of medicine, as indicated by Hippocrates’ statement that “Patients often lie when they say that they take their medication”. The World Health Organization reported that “improving patient adherence to chronic treatment should be more beneficial than any biomedical discovery”.2 Because anticancer drugs can significantly increase patients’ survival, some cancers have become chronic diseases. Most drugs are costly, induce side effects and their efficacy frequently depends on the dose. For these reasons, adherence to cancer therapy is critical for an optimal benefit-risk ratio. Adherence to treatment has been extensively studied in asthma and diabetes, but few studies have approached this issue in patients with malignant diseases.3 The impact of non-adherence on the achievement of sustained remission was observed in patients with chronic myeloid leukemia in whom poor adherence to imatinib therapy may be the predominant reason for not reaching an optimal molecular response.4-8 No “gold standard” exists to measure adherence, but a minimum of 90% drug intake was described as a good cut-off to discriminate treatment-adherent versus non-adherent patients.7,9-11 In a meta-analysis, Noens et al. showed that the rate of treatment non-compliant patients with chronic myeloid leukemia was variable (from 3% to 56%), depending on the evaluation method. Adherence to hydroxycarbamide therapy has been studied in patients with sickle cell disease and appears suboptimal in most cases; better adherence was associated with improved clinical and economic outcomes.12 To the best of our knowledge, adherence to treatment has not been studied in PV or ET patients. We conducted a prospective clinical study to analyze adherence rates, reasons for non-adherence including the impact of previous complications and the influence of non-adherence on the clinical outcome of PV and ET patients.

Commission Nationale Informatique et Libertés (CNIL) (N. 13809*03). We excluded patients treated for other MPN, patients who did not receive any treatment and those with physical or mental disabilities who were unable to consent and complete the questionnaire. These patients were identified by their inclusion in the “OBENE” observational registry for patients diagnosed with and treated for Philadelphia-negative MPN at our hospital (NCT02897297).

Questionnaire and data collected This single-center prospective study was based on a closed questionnaire (with simple and multiple-choice questions) given to the patient at the end of a consultation or sent by e-mail. The questionnaire varied according to the route of administration of the cytoreductive drugs (oral or subcutaneous) (Online Supplementary Figure S1). The questionnaires were validated by both the Scientific and Ethical Committees of our hospital. A complete blood count was also performed at the time the questionnaire was administered. The questionnaire was filled in by patients, the results of complete blood counts were collected and sent directly to the data analyzers. Consultants were not allowed to know which patients were or were not adherent to treatment. Non-adherence to drug prescription was defined by at least three omissions of medication during the preceding month (representing ≥10% of the dose) for the group treated orally (group 1) and omission of at least one injection during the two preceding months for the subcutaneously treated group (group 2). These definitions were chosen in accordance with the cut-offs identified by Marin et al.6 The patients were followed prospectively and new events (thrombosis, hematologic evolution and death) were recorded at the end of the study on February 1, 2017. At that point, the patients’ identities were revealed to the consultants and the global analyses were performed.

Statistical methods The responses were analyzed using conventional descriptive parameters. The response items were described in terms of frequency for qualitative responses and as the median ± the extreme values for quantitative answers. Statistical analyses were performed by the Clinical Investigation Center of Brest Hospital (INSERM CIC 1412) using SAS software (SAS, Brie Comte Robert, France). The data were compared using the chi-square test for qualitative parameters and non-parametric tests for quantitative parameters. A P value <0.05 was considered statistically significant. The risks of thrombosis and transformation were analyzed by calculating the hazard ratios between treatment-adherent and non-adherent patients.

Results Description of the population study

Methods Recruitment of patients Between December 2014 and December 2015, adult patients followed for PV or ET at the Institut de Cancéro-Hématologie (CHRU of Brest, France) and treated with oral (group 1) or subcutaneous (group 2) cytoreductive therapies for more than 6 months were enrolled in the OUEST study (NCT02893410). All patients signed informed consent to participation in the study, which was approved by the regional authorities of the Ethics Committee “CPP Ouest V” dated 04/09/2014, pursuant to Article L.1121-1 of the Code of Public Health, and has been declared to the 608

We included 286 patients in the study: 136 (47.6%) with PV and 150 (52.4%) with ET as their initial diagnoses. The sex ratio was 0.74 in the whole cohort (164 males, 122 females). Of the 286 patients, 233 (81.5%) received their treatment orally (group 1) and 53 (18.5%) received it subcutaneously (group 2). All the patients’ characteristics are summarized in Table 1. Before completion of the questionnaire, most patients had experienced a complication related to their MPN: thrombotic events before or at diagnosis of MPN in 31.8% (74/233) and 18.9% (10/53) of patients in group 1 and 2, respectively, and between diagnosis and inclusion in the haematologica | 2018; 103(4)

MPN patients’ non-adherence to treatment

study in 20.6% (48/233) and 28.3% (15/53); or phenotypic evolution in 12% (28/233) and 17% (9/53), respectively. At the time of being administered the questionnaire, the patients’ median age was 69.8 years old (range, 26-98.4) and the median follow-up since diagnosis had been 8.3 years (range, 0.5-36.9). In group 1, 163/233 patients (70%) took hydroxycarbamide with a median number of 10.7 pills per week whereas all patients but one in group 2 were receiving injections of pegylated interferon-α2a, in most cases every 3 weeks. Ongoing treatment was the first-line therapy in 72.3% and 22.6% of patients in groups 1 and 2, respectively. Antithrombotic drugs were administered to 217/233 patients (93.1%) of group 1 (78.3% low-dose aspirin) and 48/53 patients (90.6%) in group 2 (68.8% low-dose aspirin).

Characteristics of treatment non-adherent patients Incidence of non-adherence

Table 1. Characteristics of the population studied.

Characteristics

Cohort

Number of patients 286 Age at the time of consultation (y) 69.8 Sex ratio 1.34 Pathologies (ET/PV) 150/136 On-going treatment (n/%) Hydroxycarbamide 163 (57) Anagrelide 37 (12.9) Pipobroman 24 (8.4) Ruxolitinib 9 (3.1) Pegulated interferon α2a 52 (18.2) Pegulated interferon α2b 1 (0.4) Low dose aspirin 194 (72.9) Vitamin-K antagonists 43 (16.2) Clopidogrel 18 (6.8) Associations of antithrombotic drugs 9 (3.4) History of thrombotic events 84 (29.4) before diagnosis (n./%) Cardiovascular risk factors (n/%) High blood pressure 131 (45.8) Hypercholesterolemia 61 (21.3) Tobacco use 26 (9.1) Diabetes 22 (7.7) Median follow-up before consultation (y) 8.3 Patients with complications from diagnosis to consultation (n/%) Thromboses 63 (22) Hematologic evolutions 37 (12.9) Median follow-up from consultation (y) 1.8 Patients with complications after 35 (12.3) completion of questionnaire(n/%) Thromboses 18 (6.3) Hematologic evolutions 8 (2.8) Death 17 (5.9) Non-adherence analyses (n%) Cytoreductive drugs 65 (22.7) Antithrombotic drugs 46 (18) Both 27 (9.4) Total 85 (29.7)

Group 1 Oral drugs

Group 2 SC drugs

233 72.4 1.4 128/105

53 61.2 1.12 22/31

163 (70) 37 (16) 24 (10) 9 (4) na na 163 (74.8) 32 (14.7) 16 (7.3) 7 (3.2) 74 (31.8)

na na na na 52 (98,1) 1 (1,9) 31 (64.6) 11 (22.9) 2 (4.2) 2 (4.2) 10 (18.9)

112 (48) 53 (22.8) 20 (8.6) 20 (8.6) 7.6

19 (35.8) 8 (15.1) 6 (11.3) 2 (3.7) 11.9

48 (20.6) 28 (12) 1.8 32 (13.7)

15 (28.3) 9 (17) 1.8 3 (5.7)

16 (6.9) 7 (3) 17 (7.3)

2 (3.8) 1 (1.9) 0

55 (23.6) 33 (15.2) 22 (9.4) 67 (28.8)

10 (18.9) 13 (27.1) 5 (9.4) 18 (34)

ET: essential thrombocythemia; na: non-applicable; n: number; PV: polycythemia vera; SC: subcutaneous; y: years; %: percent.

haematologica | 2018; 103(4)

Using the criteria of non-adherence defined in the Methods section (missing at least 3 doses in the preceding month for oral drugs or 1 injection in the 2 preceding months for subcutaneous drugs), 65/286 patients (22.7%) were considered non-adherent to their cytoreductive drug therapy. Non-adherence was more frequent in the orally treated group (group 1) than in the subcutaneously treated group (group 2): 55/233 (23.6%) and 10/53 (18.9%) in groups 1 and 2, respectively, P=0.46). All the characteristics are showed in Table 2. Regarding compliance with antithrombotic drug therapy, 46/286 patients (16.1%) in the whole cohort declared that they did not adhere fully to their treatment. The patients in group 2 declared a higher rate of non-adherence to their antithrombotic drugs compared to the patients in group 1 [13/53 (27.1%) versus 33/233 (15.2%), respectively, P=0.055]. In both groups, patients who were non-adherent to their cytoreductive drug treatment were also less adherent to their antithrombotic therapy compared to patients who were adherent to their cytoreductive drug treatment [26/65 (40%) versus 20/221 (9%), P<107 ; ORR=6.64, 95% CI: 3.22-13.94]. In total, 85/286 patients (29.7%) were non-adherent to either cytoreduction or antithrombotic drugs and 27/286 to both treatments (9.4% of the total cohort or 31.8% of the non-adherent patients). In both groups, the number of treatment omissions was close to the threshold defining non-adherence (around 10% of intake omissions) for most patients (96%) while the remaining 4% had very poor adherence to treatment (≥20%).

Analysis of the cohort To tease out the characteristics of treatment non-adherent patients, we then analyzed the responses of these patients to cytoreductive drugs (n=65). In the whole cohort, non-adherent patients were younger (68.1 versus 70.7 years, P=0.007), more frequently male (1.1 versus 0.66, P=0.07), taller (172 versus 165 cm, P=0.004) and heavier (74 versus 69 kg, P=0.01). The differences in height and weight were probably not only due to the gender bias because treatment non-adherent patients remained significantly taller even when only male patients were analyzed, (176 versus 173, P=0.002). This was not true for female patients. Furthermore, in the treatment non-adherent group there was a higher proportion of patients choosing their own drug intake schedule (55.4 versus 34.8%, P=0.003), a lower proportion of patients following a fixed intake schedule (64.6 versus 83.7%, P=0.0008) and fewer polymedicated patients (67.7 versus 79.2%, P=0.05). Interestingly, diabetic patients were significantly more adherent to their cytoreductive treatment [21/221 (9%) versus 1/65 (1.53), P=0.03; ORR=6.69, 95% CI: 1.03-281.86]. Since groups 1 and 2 had different rates of compliance, we analyzed the characteristics of the treatment nonadherent patients for each group. In group 1, treatment non-adherent patients were younger than adherent patients (66.6 versus 73.4 years old; P=0.0013), more frequently determined the pill intake schedule themselves (52.8% versus 33.7%; P=0.029), dispersed their pill intake through the day instead of grouping the pills together (30.9% versus 10.8%; P=0.0003) and had fewer drugs to take (70.9% versus 82.5%; P=0.06). For group 2 patients, the only significant difference concerned 609

R. Le Calloch et al.

the dose of interferon: treatment non-adherent patients had higher doses (81.5 versus 46.5 mg/week, P=0.05). Organizing pills according to intake schedules (drug diary) or having someone to remind the patients when to take their drug did not improve adherence. For both groups, having experienced thrombosis or phenotypic evolution did not modify the adherence to treatment. In the subpopulation of patients treated for ET, no significant association was found between non-adherence and thrombotic events, whether for cytoreductive therapy (ORR=1.25; 95% CI: 0.55â&#x20AC;&#x201C;2.83) or for antithrombotic drugs (ORR=1.3; 95% CI: 0.45-3.72). This was also true for patients with PV (ORR=1.36; 95% CI: 0.65-2.82 and ORR=1.72; CI: 0.6-4.94, respectively) (Online Supplementary Figure S2).

ness and difficulties in managing their treatment during holidays and travel (62% and 55%, respectively). Interestingly, the former was more frequently claimed by group 1 (48%) while the latter was the most frequent reason in group 2 (40%). In group 2, professional schedule constraints were also frequently brought up (26%). Some patients expressed more than one reason. It is interesting to note that 12% of patients reported voluntarily omitting some treatment and that no patients mentioned side effects as a cause of non-adherence. All the reasons are presented in Figure 1. It is also interesting that 20/55 (40.8%) of the patients who did not adhere fully to their treatment believed that forgetting their treatment on occasions had no influence on its efficacy, and only four of them increased the dose following a missed intake.

Reasons for treatment non-adherence To gain further insight into the causes of non-adherence to treatment, patients were asked to identify the most important reasons why they had skipped doses. The two most frequent reasons were simply forgetful-

Incidence of thrombotic events and hematologic evolutions after completion of the questionnaire To determine whether non-adherence to treatment had an impact on the evolution of the MPN, as has been

Table 2. Analyses of treatment non-adherence in the studied population.

Non-adherence (n/%) Cytoreductive drugs Antithrombotic drugs Both Total Age at the time of consultation (y) Sex ratio Pathology (n/%) ET PV Style of life (n/%) Living alone City resident History of thrombosis (n/%) History of evolution (n/%) Treatment (n/%) Hydroxycarbamide Doses (pills/wk) Duration (>1 y) Second-line therapy Pegylated interferon Doses(injections/wk) Duration (>1 y) Second-line therapy Drug diary Help to remember Same timing of intake Own timing for intake Other medications Full blood counts at inclusion CHR (n/%) Hemoglobin (g/dL) Platelets (x109/L) Leukocytes (x109/L) Neutrophils (x109/L)

Non-adherent pts

Whole cohort Adherent pts

65 (22.7) 46 (16.1) 26 (9.1) 85 (29.7)

221 (77.3) 240 (83.9) 260 (90.9) 201 (70.3)

68.1 1.1

70.7 0.66

32 (49.2) 33 (50.8)

Group 1 Non-adherent pts Adherent pts 55 (23.6) 33 (15.2) 21 (9) 67 (28.8)

178 (76.4) 200 (84.8) 212 (91) 166 (71.2)

0.007 0.07

66.6 0.96

73.4 0.65

118 (53.4) 103 (46.6)

0.55

27 (49.1) 28 (50.9)

25 (38.5) 33 (50.8) 18 (27.7) 7 (10.8)

64 (29) 121 (54.8) 45 (20.4) 30 (13.6)

0.15 0.12 0.21 0.55

41 (74.6) 12.3 50 (90.9) 21 (38.2) 42 (79.2) 81.5 (53.1) 9 (90) 6 (60) 13 (24.1) 9 (16.4) 42 (64.6) 36 (55.4) 44 (67.7)

121 (68.6) 10.1 158 (89.3) 67 (37.8) 10 (18.9) 46.5 (26.1) 40 (93) 35 (81.4) 54 (30.5) 19 (10.7) 185 (83.7) 77 (34.8) 175 (79.2)

33 (50.8) 13.6 355 6.15 4

144 (65.2) 13.3 319 6.4 4.2

Group 2 Non-adherent pts Adherent pts 10 (18.9) 13 (27.1) 5 (9.4) 18 (34)

43 (81.1) 40 (72.9) 48 (90.6) 35 (66)

0.0013 0.2

61.1 2.33

56.8 0.72

0.31 0.16

101 (56.7) 77 (43.3)

0.52

5 (50) 5 (50)

17 (39.5) 26 (60.5)

0.5

23 (41.8) 28 (50.9) 14 (25.5) 4 (7.3)

53 (29.8) 100 (56.2) 34 (19.2) 24 (13.6)

0.1 0.35 0.31 0.34

2 (20) 5 (50) 4 (40) 3 (30)

11 (25.6) 21 (48.8) 11 (25.6) 6 (13.9)

1 1 0.4 0.35

0.31 0.11 0.73 0.97 1 0.05 1 0.2 0.36 0.26 0.0008 0.003 0.05

41 (74.6) 12.3 50 (90.9) 21 (38.2) na na na na 13 (24.1) 9 (16.4) 38 (69.1) 28 (52.8) 39 (70.9)

121 (68.6) 10.1 158 (89.3) 67 (37.8) na na na na 54 (30.5) 19 (10.7) 157 (89.2) 59 (33.7) 146 (82.5)

0.31 0.11 0.73 0.97

0.36 0.26 0.0003 0.03 0.06

na na na na 42 (79.2) 81.5 (53.1) 9 (90) 6 (60) na na 4 (40) 8 (80) 5 (50)

na na na na 10 (18.9) 46.5 (26.1) 40 (93) 35 (81.4) na na 28 (65.1) 28 (66.6) 29 (67.4)

0.03 0.21 0.11 0.26 0.17

28 (50.9) 12.4 354 7.81 7.28

110 (61.8) 12.1 339 7.12 5.11

0.3 0.18 0.51 0.51 0.45

5 (50) 13.9 379 5.7 3.5

34 (79.1) 13.5 252 5.7 3.1

CHR: complete hematological response; ET: essential thrombocythemia; n: number; PV: polycythemia vera; pts: patients; wk: week; y: years; %: percent.

610

haematologica | 2018; 103(4)

1 0.05 1 0.2 0.17 0.84 0.46 0.33 0.98 0.08 1 0.74

MPN patients’ non-adherence to treatment

shown in other pathologies, the cohort was prospectively followed for an additional median time of 1.8 years (range, 1-2.4). During the follow-up period, we recorded new events in 35/286 patients (12.2%), among whom 32/35 (91%) were in group 1. The recorded events were thrombosis in 18 cases (6.3%; 12 arterial, 6 venous), phenotypic evolution in 7 (2.4%; 1 case of post-ET PV, 3 cases of secondary myelofibrosis and 3 cases if secondary acute myeloid leukemias) and death in 17 cases (5.9%), all occurring in group 1 (P=0.05) (Table 3). In the whole cohort, non-adherence to cytoreductive therapy was associated with a significant reduction in the complete hematologic remission rate compared to that in the group adhering to treatment: 50.8 versus 65.2% (P=0.03) (ORR=1.85, 95% CI: 1.01-3.36). This difference was lost when analyzing groups 1 and 2 separately (Table 2). No significant association was found between nonadherence and thrombosis or death. In group 2, nonadherence was not significantly associated with the outcome, but there were only a few events in this group. However, non-adherence to cytoreductive therapy was associated with an increased risk of hematologic transformation both in the whole cohort [4/65 (6.1%) versus 3/221 (1.3%), P=0.05; ORR=4.73, 95% CI: 0.78-33.14] and in group 1 [4/55 (7.3%) versus 3/178 (1.8%), P=0.05; ORR=4.54, 95% CI: 1-31.98]. Furthermore, these evolutions also occurred sooner in the treatment non-adherent group (P=0.05) (Figure 2).

Discussion The importance of treatment compliance has now been clearly established in many pathological conditions, and especially in hematologic malignancies.9,13-16 These studies typically demonstrate that poor adherence has a negative impact on clinical evolution. However, to the best of our knowledge, no such data were previously available regarding patients with Philadelphia-negative MPN. Yet, these chronic disorders have very variable clinical evolution and are prone to complications. We, therefore, decided to assess MPN patients’ compliance with cytoreductive and antithrombotic treatments. Many ways of assessing patients’ adherence to treat-

ment have been described, including pill counts, drug plasma levels, various microelectronic monitoring systems and dispensation by a third party. All methods have their pros and cons. We chose to assess patients’ adherence using a single questionnaire. The self-evaluation method using a questionnaire is easier and less expensive to implement, even though patients’ reluctance to admit omitting drug intake could theoretically bias the results. Because of the blind process of this study, there was no influence from the consultant or staff on completion of the questionnaires. We cannot, however, exclude some degree of under-declaration of non-adherence. Despite this fact, the proportion of patients not adherent with treatment in this study was equivalent to that reported by Marin et al. who used a microelectronic monitoring system, suggesting that

Table 3. New events observed after completion of the questionnaires.

Non-adherent Adherent patients patients Whole cohort N. of patients Events (n/%) Total Thrombosis Evolution Death Group 1 N. of patients Events (n/%) Total Thrombosis Evolution Death Group 2 N. of patients Events (n/%) Total Thrombosis Evolution Death

286

221

35 (12.2) 18 (6.3) 7 (2.4) 17 (5.9)

8 (12.3) 4 (6.1) 4 (6.1) 2 (3.1)

27 (12.2) 14 (6.3) 3 (1.4) 15 (6.8)

233

178

32 (13.7) 16 (6.9) 7 (3) 17 (7.3)

7 (12.7) 3 (5.6) 4 (7.3) 2 (3.7)

25 (14) 13 (7.3) 3 (1.8) 15 (8.4)

3 (5.7) 2 (3.8) 0 0

1 (10) 1 (10) 0 0

2 (4.7) 1 (2.3) 0 0

0.98 1 0.05 0.37

0.8 0.77 0.05 0.37

0.47 0.34 na na

na: non-applicable; n: number; pts: patients; %: percent.

the questionnaire does not grossly underestimate non-

Figure 1. Reasons for non-adherence. Gray represents the answers of patients from group 1 (oral intake) and black represents the answers of patients from group 2 (sub-cutaneous injection). The results are expressed as percentage of answers. Patients could state more than one reason for non-adherence. FBC: full blood count.

haematologica | 2018; 103(4)

611

R. Le Calloch et al.

adherence. This study by Marin et al. in patients with chronic myeloid leukemia found that 26% of patients were non-adherent with their treatment, while in the present study, 22% were non-adherent to cytoreduction and 29% to either cytoreduction or antithrombotic drugs. Furthermore, our questionnaire provided insight into the reasons why the patients missed taking their drugs. The causes for non-adherence were also approached in chronic myeloid leukemia by Marin et al. who reported younger age as a major factor.6 Likewise, in acute lymphoblastic leukemia, Bhatia et al. found that adherence to maintenance therapy was suboptimal in teenagers among whom the non-adherence rate was 20.5%.9 This study also pointed to socio-economic conditions as a major determinant of adherence. In our study, factors significantly associated with non-adherence were found mostly in patients who took their treatment orally, and included younger age, choosing the pill intake schedule themselves, dispersing their intake through the day and a small number of different drugs to take (Table 2). This indicates that, in addition to personal traits (age, ethnic, socio-economic background), the way of managing patientsâ&#x20AC;&#x2122; drug intake (route of administration, time of the day, number of different treatments) has an impact on adherence. Interestingly, patients receiving subcutaneously injected cytoreduction showed poorer adherence to oral antithrombotic drugs than patients receiving oral cytoreduction. Together with the fact that treatment adherent patients were more likely to be taking several drugs and the fact that diabetic patients showed a higher rate of adherence to treatment, this suggests that having a higher number of oral drugs to take makes it less likely to miss one intake. Physicians should therefore probably put even more effort into helping patients being treated with subcutaneous cytoreduction ensure good adherence to their antithrombotic drug treatment. Unexpectedly, the occurrence of adverse effects to the drugs was not reported by patients as a determining factor in their non-adherence. We also observed that having suffered a complication and/or phenotypic evolution of the disease did not increase adherence to treatment. This is coherent with the fact that most non-adherent patients reported not believing that non-adherence may affect the clinical outcome of their disease. These ele-

ments suggest that good comprehension of the disease and treatment should improve adherence, as has been shown in other chronic diseases such as diabetes. To assess the risk of transformation associated with non-adherence, we analyzed the events (thrombosis, phenotypic evolution) that occurred before compilation of the questionnaire (retrospective study) and followed up the cohort prospectively. The frequencies of events before compilation of the questionnaire (median time of observation of 11.7 years) were similar in treatment adherent and non-adherent patients. Thrombotic events occurred in 19.2% of treatment adherent and 25.5% of treatment non-adherent patients. These frequencies are slightly higher than those reported for patients with PV (26.4% versus 9.3-19%) and in accordance with the scientific literature for patients with ET (18.2% versus 7.6-22%).17 Regarding the events that occurred during the prospective follow-up, the median time was shorter (1.8 years). However, events were recorded in 35 (12.2%) patients. No differences were noted between the treatment adherent and non-adherent groups regarding thrombotic events or death, but phenotypic evolution was more frequent in the treatment non-adherent patients, especially in group 1. Although this result must be interpreted with caution given the small number of affected patients (n=7), the impact of adherence on phenotypic evolution of MPN is reminiscent of data reported for chronic myeloid leukemia or acute lymphoblastic leukemia, confirming the importance of constant therapeutic pressure for the control of malignant clones. Unexpectedly, the impact on thrombosis was less obvious. This may be related to the fact that thrombosis is a more acute event, depending on the immediate hemostatic status at the time of thrombus constitution, whereas the phenotypic evolution of chronic hematologic malignancies may be more the result of long-term evolution of the clone, reflecting its exposure to therapeutic pressure. This is coherent with the observation that treatment non-adherent patients were less likely to achieve a complete hematologic response. Only a few events were observed in group 2, suggesting that interferon may ensure better long-term control of MPN clones as has been suggested in recent publications.18 Further evaluation of the long-term impact of non-adher-

Figure 2. Kaplan-Meier evolution-free survival curves for treatment adherent or non-adherent patients.

612

haematologica | 2018; 103(4)

MPN patients’ non-adherence to treatment

ence would be necessary to confirm these observations. This evaluation will be not possible with our cohort, since after the blinding had been removed, all treatment non-adherent patients were managed to improve their adherence. Larger multicenter studies could confirm the “non-adherent profile” which sometimes pointed to unexpected findings, such as taller height in male patients not adhering fully with treatment. To our knowledge, OUEST is the first study on the incidence, determinants and impact of treatment non-adherence on the outcome of patients with Philadelphia negative MPN. The occurrence of non-adherence is relatively common, with an incidence of 28%, but is generally moderate. Younger age and the route and schedule of drug

References 1. Barbui T, Barosi G, Birgegard G, et al. Philadelphia-negative classical myeloproliferative neoplasms: critical concepts and management recommendations from European LeukemiaNet. J Clin Oncol. 2011;29(6):761-770. 2. Sabaté E, World Health Organization, editors. Adherence to long-term therapies: evidence for action. Geneva: World Health Organization; 2003,p.198. 3. Salmeron S, Liard R, Elkharrat D, Muir J, Neukirch F, Ellrodt A. Asthma severity and adequacy of management in accident and emergency departments in France: a prospective study. Lancet. 2001;358 (9282):629-635. 4. Söderlund S, Dahlén T, Sandin F, et al. Advanced phase chronic myeloid leukaemia (CML) in the tyrosine kinase inhibitor era - a report from the Swedish CML register. Eur J Haematol. 2017;98(1):57-66. 5. Santoleri F, Lasala R, Ranucci E, et al. Medication adherence to tyrosine kinase inhibitors: 2-year analysis of medication adherence to imatinib treatment for chronic myeloid leukemia and correlation with the depth of molecular response. Acta Haematol. 2016;136(1):45-51. 6. Marin D, Bazeos A, Mahon F-X, et al. Adherence is the critical factor for achiev-

haematologica | 2018; 103(4)

10.

11. 12.

administration seem to be the major determinants of poor treatment adherence. Phenotypic evolution seems to be more frequent in the group not adherent to treatment, suggesting that cytoreductive drug pressure could help to reduce the risk of evolution. Major efforts should be invested into improving treatment adherence. Acknowledgments The authors would like to thank all the patients who took the time to complete the questionnaire as a mark of interest in this study. The authors would also like to thank the France Intergroup of Myeloproliferative neoplasms (FIM) for its help with this study: RLC, VU, SG, EL and JCI are members of this cooperative group.

ing molecular responses in patients with chronic myeloid leukemia who achieve complete cytogenetic responses on imatinib. J Clin Oncol. 2010;28(14):2381-2388. Noens L, van Lierde M-A, De Bock R, et al. Prevalence, determinants, and outcomes of nonadherence to imatinib therapy in patients with chronic myeloid leukemia: the ADAGIO study. Blood. 2009;113(22): 5401-5411. Noens L, Hensen M, Kucmin-Bemelmans I, Lofgren C, Gilloteau I, Vrijens B. Measurement of adherence to BCR-ABL inhibitor therapy in chronic myeloid leukemia: current situation and future challenges. Haematologica. 2014;99(3): 437-447. Bhatia S, Landier W, Hageman L, et al. 6MP adherence in a multiracial cohort of children with acute lymphoblastic leukemia: a Children’s Oncology Group study. Blood. 2014;124(15):2345-2353. Cortes JE, Egorin MJ, Guilhot F, Molimard M, Mahon FX. Pharmacokinetic/ pharmacodynamic correlation and blood-level testing in imatinib therapy for chronic myeloid leukemia. Leukemia. 2009;23(9):1537-1544. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med. 2005;353(5):487497. Candrilli SD, O’Brien SH, Ware RE, Nahata MC, Seiber EE, Balkrishnan R. Hydroxyurea adherence and associated outcomes among Medicaid enrollees with

13.

14.

15.

16.

17.

18.

sickle cell disease. Am J Hematol. 2011;86 (3):273–277. Amitai I, Leader A, Raanani P. Adherence to tyrosine kinase inhibitors in chronic myeloid leukemia: the challenge that lies ahead. Acta Haematol. 2016;136(1):43-44. Cuisset T, Quilici J, Fugon L, et al. Nonadherence to aspirin in patients undergoing coronary stenting: negative impact of comorbid conditions and implications for clinical management. Arch Cardiovasc Dis. 2011;104(5):306-312. Schwartz KA, Schwartz DE, Ghosheh K, Reeves MJ, Barber K, DeFranco A. Compliance as a critical consideration in patients who appear to be resistant to aspirin after healing of myocardial infarction. Am J Cardiol. 2005;95(8):973-975. Simpson SH, Eurich DT, Majumdar SR, et al. A meta-analysis of the association between adherence to drug therapy and mortality. BMJ. 2006;333(7557):15. Casini A, Fontana P, Lecompte TP. Thrombotic complications of myeloproliferative neoplasms: risk assessment and risk-guided management. J Thromb Haemost. 2013;11(7):1215-1227. Masarova L, Patel KP, Newberry KJ, et al. Pegylated interferon alfa-2a in patients with essential thrombocythaemia or polycythaemia vera: a post-hoc, median 83 months follow-up of an open-label, phase 2 trial. Lancet Haematol. 2017;4(4):e165e175.

613

ARTICLE

Acute Myeloid Leukemia

Ferrata Storti Foundation

Gfi1b: a key player in the genesis and maintenance of acute myeloid leukemia and myelodysplastic syndrome

Aniththa Thivakaran,1 Lacramioara Botezatu,1 Judith M. Hönes,1,2 Judith Schütte,1 Lothar Vassen,1 Yahya S. Al-Matary,1 Pradeep Patnana,1 Amos Zeller,1 Michael Heuser,3 Felicitas Thol,3 Razif Gabdoulline,3 Nadine Olberding,1 Daria Frank,1 Marina Suslo,1 Renata Köster,1 Klaus Lennartz,4 Andre Görgens,5,6 Bernd Giebel,5 Bertram Opalka,1 Ulrich Dührsen1 and Cyrus Khandanpour1,7

Department of Haematology, West German Cancer Center, University Hospital Essen, University of Duisburg-Essen, Essen, Germany; 2Department of Endocrinology, Diabetes and Metabolism, University Hospital Essen, University of Duisburg-Essen, Essen, Germany; 3Department of Haematology, Haemostaseology, Oncology, and Stem Cell Transplantation, Medical University of Hannover, Germany; 4Institute for Cell Biology, University Hospital Essen, University of Duisburg-Essen, Essen, Germany; 5Institute for Transfusion Medicine, University Hospital Essen, University of Duisburg-Essen, Essen, Germany; 6Department of Laboratory Medicine, Karolinska Institutet, Stockholm, Sweden and 7Department of Medicine A, Hematology, Oncology and Pneumology, University Hospital Münster, Germany 1

Haematologica 2018 Volume 103(4):614-625

ABSTRACT

Correspondence: cyrus.khandanpour@uk-essen.de

Received: June 29, 2017. Accepted: January 5, 2018. Pre-published: January 11, 2018.

ifferentiation of hematopoietic stem cells is regulated by a concert of different transcription factors. Disturbed transcription factor function can be the basis of (pre)malignancies such as myelodysplastic syndrome (MDS) or acute myeloid leukemia (AML). Growth factor independence 1b (Gfi1b) is a repressing transcription factor regulating quiescence of hematopoietic stem cells and differentiation of erythrocytes and platelets. Here, we show that low expression of Gfi1b in blast cells is associated with an inferior prognosis of MDS and AML patients. Using different models of human MDS or AML, we demonstrate that AML development was accelerated with heterozygous loss of Gfi1b, and latency was further decreased when Gfi1b was conditionally deleted. Loss of Gfi1b significantly increased the number of leukemic stem cells with upregulation of genes involved in leukemia development. On a molecular level, we found that loss of Gfi1b led to epigenetic changes, increased levels of reactive oxygen species, as well as alteration in the p38/Akt/FoXO pathways. These results demonstrate that Gfi1b functions as an oncosuppressor in MDS and AML development.

doi:10.3324/haematol.2017.167288

Introduction Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/614 ©2018 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.

614

Myelodysplastic syndrome (MDS) is characterized by disturbed function of the myeloid compartment of the bone marrow (BM),1 leading in some cases to acute myeloid leukemia (AML).2 AML is characterized by an accumulation of immature myeloid blasts in the BM.2 Hematopoietic development, among other functions, is regulated by transcription factors (TFs).3 Functional dysregulation of several TFs4,5 can induce malignant transformation. The hematopoieticTF Growth factor independence 1b (Gfi1b) regulates dormancy and proliferation6 of hematopoietic stem cells (HSCs), the development of erythroid and megakaryocytic cells,7-10 as well as B and T cells.1113 Constitutive deletion of Gfi1b in mice is embryonically lethal at day E15 due to bleeding and anemia.9 Conditional loss of Gfi1b leads to a significant expansion of functional HSCs in the BM and peripheral blood.6 In human primary hematopoietic progenitors, forced expression of GFI1B results in expansion of immature erythroblasts and repression of myeloid differentiation.14 Gfi1b exerts its function by recruiting histone modifying enzymes, such as CoREST, G9a, LSD1 or HDACs, to induce deacetylation of H3K9, demethylation of H3K4 and/or methylation of H3K9.15-18 We report that a reduced level or absence of GFI1B negatively influences the prognosis of MDS/AML patients. Moreover, we present evidence that loss/reduced expression of Gfi1b promotes AML development in different murine models of human AML. haematologica | 2018; 103(4)

Gfi1b in AML and MDS

Furthermore, reduced expression of Gfi1b in murine models of human leukemia leads to a higher number of leukemic stem cells (LSCs). On a molecular level, aberrant regulation of the ROS/p38/Akt/FoXO pathway as a consequence of reduced Gfi1b level might contribute to these phenotypic changes.

Methods Study samples Characteristics of different patient cohorts have been described previously.19-25

Boundaries of GFI1B expression To set boundaries for GFI1B expression levels in AML and MDS patients, we correlated expression levels with the survival outcome of patients.

Mice Gfi1bfl/fl and Gfi1bEGFP/WT, MxCre, NUP98/HOXD13 and Kras mice have been described previously.6,26-28 Mice were housed in specific pathogen-free conditions in the animal facility of University Hospital Essen. All mouse experiments were performed with the approval of the local ethics committee for animal use (authorization n. G1196/11).

Poly(I:C) treatment MxCretg mice harboring the poly(I:C) inducible Cre recombinase gene under the control of the Mx1 promoter were crossed to Gfi1bfl/fl mice. To conditionally delete the Gfi1b alleles in the NUP98/HOXD13 MDS mouse model, Gfi1bfl/flMxCretg NUP98/HOXD13tg mice were injected intraperitoneally (i.p.), as shown previously.6 For Gfi1bfl/flMxCretgKras+/fl mice, two poly(I:C) injections were sufficient to activate the Kras oncogene and delete the Gfi1b alleles. As a control, Gfi1bfl/fl or Gfi1bwt/wt mice not carrying the MxCretg were injected with poly(I:C). Three weeks after transplantation of MLL-AF9-transduced lineage negative (Lin-) BM cells from Gfi1bfl/flMxCretg or Gfi1bfl/flMxCrewt mice, primary recipient mice were injected with poly(I:C) 4 times every second day.

Isolation, retroviral transduction, and transplantation of murine hematopoietic progenitor cells Mouse leukemia was induced by transplanting Lin- BM cells that were retrovirally transduced with the MLL-AF9 oncofusion gene as well as the GFP-encoding gene, as previously described.4,27 For the limiting dilution assay, different numbers of leukemic cells were retransplanted into sublethally irradiated (3 Gy) secondary recipient mice (3-4 mice/group). The frequency of functional LSCs was determined using ELDA software.29

ChIP and ChIPâ&#x20AC;&#x201C;Seq analyses Chromatin Immunoprecipitation (ChIP) and ChIP-Seq analyses were performed as previously described.4,27 Data are available from: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE88934

Online Supplementary Appendix Details on the experimental procedures and figures are available in the Online Supplementary Appendix.

Results Low level of GFI1B are indicative of an inferior prognosis of MDS and AML patients To obtain a first insight into the role of GFI1B in AML prognosis, we analyzed two well-annotated published haematologica | 2018; 103(4)

data sets.19-21,25 In these sets, CD34+ leukemic cells and CD34+ control HSCs were used. CD34+ leukemic cells represent a fraction in which LSCs are enriched, whereas CD34+ cells from healthy donors represent a fraction of cells in which HSCs are enriched.21,30 GFI1B showed lower expression in CD34+ AML blasts compared to CD34+ control HSCs (Figure 1A). MDS can progress to AML, and therefore, we wanted to elucidate how GFI1B expression changes during the progression of MDS to AML. Again, GFI1B showed a lower expression in AML blasts compared to GFI1B expression in CD34+ cells from the BM of patients with MDS (Figure 1B). We also analyzed an independent data set, which provided whole genome expression data for LSCs in different types of AML as well as different human hematopoietic progenitor cells.20,25 GFI1B showed a lower expression in human LSCs of different AML subtypes compared to its expression in normal human myeloid progenitors (GMPs) or HSCs (Figure 1C).20 GMPs and HSCs are two fractions from which LSCs arise in mice and humans.31 We analyzed whether GFI1B level might also be informative regarding the prognosis of MDS and AML patients. Based on available expression data of GFI1B and the associated survival data, we could distinguish two distinct populations with regard to GFI1B expression (Figure 2A). A low level of GFI1B (see Methods section and Online Supplementary Appendix for details) in leukemic blast cells was associated with inferior outcome with regard to overall survival (OS) of all AML patients (Figure 2B) as well as OS and event-free survival (EFS) in the group of patients with no overt cytogenetic aberrations (Figure 2C and D). We also performed a multivariate analysis, including additional factors such as age, sex and cytogenetic status, as well as mutational status of certain genes. There was a tendency for a very low GFI1B level to be an independent prognostic marker (P=0.12), but this did not reach a level of significance (data not shown). Low GFI1B expression might be associated with an inferior prognosis, but other confounding factors contribute to this association. Finally, we examined whether low GFI1B expression (the lowest 5% compared with the highest 20% of expression levels) was associated with a certain gene expression signature to obtain a first insight into how GFI1B might influence prognosis. We performed Signaling Pathway Enrichment using Experimental Datasets (SPEED) analysis (see Online Supplementary Appendix) on two separate studies,21,22 for which expression data of the full length GFI1B and associated clinical data were available. Low level of GFI1B expression was associated with a reactive oxygen species (ROS)-mediated signature pathway as well as activation of mitogen-activated protein kinase (MAPK), JAK, TGFB and TLR signaling pathways (Figure 2E). We also examined whether GFI1B expression level influences survival and disease progression from MDS to AML using a separate set of data.23,27 Again, we could distinguish two different populations with regard to GFI1B expression (low and high) (Figure 2F): low expression of GFI1B correlated with poor EFS (Figure 2G). Anguita et al.32 observed a positive correlation between the expression of a mutated form of GFI1B, which acts in a dominant-negative manner, and the expression of MLLT3 and a negative correlation with regard to SPI1. In addition, Chowdhury et al. described a negative correlation between GFI1B expression and MEIS1.33 In our patient cohorts, we also found an inverse correlation 615

A. Thivakaran et al.

between GFI1B expression level and SPI1 expression as well as MEIS1 and a positive correlation with MLLT3 (Online Supplementary Figure S1).

Reduced expression level or loss of Gfi1b promotes progression of MDS to AML in a murine MDS model To investigate a connection between Gfi1b level and AML, we used different mouse strains and models of human leukemia. We used one strain in which both Gfi1b alleles can be conditionally deleted in the hematopoietic system upon poly(I:C) administration, resembling absence of Gfi1b expression (Gfi1bfl/flMxCretg).6 In a second mouse model, one coding allele of Gfi1b is replaced by enhanced green fluorescence protein (EGFP) cDNA (Gfi1bEGFP/wt),13 which leads to a lower expression level of Gfi1b (see below). Finally, wild-type mice were used to model normal/high Gfi1b expression. To study whether reduced Gfi1b expression accelerates MDS to AML progression, we crossed the above-mentioned mouse strains with NUP98/HOXD13tg mice, which represent a model for human MDS/AML.34 We first used the Gfi1b:EGFP knock-in reporter mouse strain and crossed these mice with NUP98/HOXD13tg mice (Figure 3A). Loss of one allele of Gfi1b shortened the latency period of AML development (Figure 3B). In BM cells derived from heterozygous leukemic mice, the expression of Gfi1b mRNA and Gfi1b protein levels were reduced to approximately 50% compared to BM cells from Gfi1bwt/wt leukemic mice (Online Supplementary Figure S2A and B). Furthermore, we found that the EGFP expression level and hence Gfi1b expression level was significantly lower in the myeloid blasts when the disease onset was within the first 250 days compared to Gfi1b expression in blasts from mice that developed overt leukemia more than 250 days after birth (Figure 3C). The leukemic cells from Gfi1bwt/wt or Gfi1bEGFP/wt animals showed no significant differences with regard to surface marker expression, spleen size, white blood cell and platelet counts, or cytological appearance, but showed significant differences with regard to hemoglobin and red blood cells (Figure 3D, Online Supplementary Figure S2C-F and data not shown), which might be due to a potential dose-dependent role of Gfi1b in erythropoiesis.6,9

We next examined how complete absence of Gfi1b influences MDS to AML progression. We used the Gfi1b conditional knockout mouse model (Gfi1bfl/flMxCretg), whereby the expression of Gfi1b can be conditionally abrogated in the hematopoietic system upon poly(I:C) administration6 (Figure 3E). The absence of Gfi1b resulted in a substantially earlier onset of AML with a median survival time of approximately 50 days (P<0.0001) (Figure 3F). Cre-mediated excision was verified to be efficient in leukemic Gfi1bfl/flMxCretgNUP98/HOXD13tg mice after poly(I:C) administration with non-excised Gfi1b alleles below detection levels (Figure 3G), and this was associated with practically no expression of Gfi1b mRNA and protein (Online Supplementary Figure S2A and B).6 Leukemic cells from Gfi1bfl/flMxCretg and Gfi1bfl/flMxCretg animals showed no significant differences in spleen size, white blood cells or cytological appearance but significant differences in hemoglobin, red blood cells and platelet counts (Figure 3H and Online Supplementary Figure S2G-J and data not shown), which might be due to a dose-dependent role of Gfi1b in erythropoiesis.6,9 The absence of Gfi1b led to a reduced frequency of myeloid cells (Figure 3I, left, middle, and Online Supplementary Figure S3A-C). CD117 (c-Kit) was uniformly higher expressed on all Gfi1b-deficient blast cells (derived from Gfi1bfl/flMxCretg) mice compared to Gfi1b expressing blasts (Gfi1bfl/flMxCrewt) (Figure 3I, right). Finally, there was no difference with regard to apoptosis in NUP98/HOXD13tg mice (Online Supplementary Figure S3D). In our murine model of MDS/AML development, we did not observe a positive correlation between Gfi1b and Mllt3 expression nor a negative correlation between Gfi1b and Spi1 expression, which might be disease context-dependent and thus not reproducible in all types of AML (Online Supplementary Figure S3A and B). We also analyzed the expression level of Meis1, since Chowdhury et al. observed a negative correlation between GFI1B and MEIS1.33 We were able to confirm this finding for this model of AML (Online Supplementary Figure S4C).

Loss of Gfi1b promotes the progression of myeloproliferative disorder in a conditional Kras mouse model To validate the results above in a second model, we used mice conditionally expressing a mutated form of

Figure 1. Correlation between different GFI1B expression levels and myelodysplastic syndrome (MDS) or acute myeloid leukemia (AML) prognosis. (A) Expression of GFI1B in CD34+ AML cells (n=269) compared to CD34+ control cells (n=8) based on the patient cohort published by Valk et al.;21 Pâ&#x2030;¤0.0001. (B) Expression of GFI1B in CD34+ MDS cells (n=23) compared to CD34+ AML cells (n=501) based on the patient cohort published by Wouters et al.;19 Pâ&#x2030;¤0.0001. (C) Expression of GFI1B in leukemic stem cells (LSCs)20 of different AML subtypes compared to normal hematopoietic stem cells (HSCs) or common myeloid progenitor cells (CMPs) in published gene expression arrays.31

616

haematologica | 2018; 103(4)

Gfi1b in AML and MDS

Kras. RAS mutations are found in 5-10% of AML patients.2 These mice harbor a transcriptional stop codon flanked by loxP sites upstream of a mutated Kras allele and, after removal of the stop codon, develop myeloproliferative disorder.35 We crossed these mice with Gfi1bfl/flMxCretg or Gfi1bwt/wtMxCretg mice, and after poly(I:C) administration, we observed mice for the emergence of disease (Figure 4A). While Gfi1bwt/wtMxCretgKras+/fl mice developed a lethal myeloproliferative disorder with a median survival of approximately 25 days, loss of Gfi1b significantly shortened the latency period of the disease to a median survival of approximately seven days (Figure 4B). There was no difference with regard to cytological appearance, number of myeloid cells or level of apoptosis (Figure 4C and D and Online Supplementary Figure S5A-D). We also did not observe any significant difference with regard to white blood counts, platelet counts or spleen size but a significant difference in hemoglobin and red blood cells between Gfi1bfl/flMxCretg and Gfi1bwt/wtMxCretg animals (Online Supplementary Figure S5E-H), which might be due to the role of Gfi1b in erythropoiesis.6,9

haematologica | 2018; 103(4)

Loss of Gfi1b promotes the progression of AML initiated by retroviral MLL-AF9 expression The Mixed Lineage Leukemia (MLL) gene is a common target for chromosomal translocations.2 MLL-AF9 is a fusion protein frequently occurring in a subset of AML patients,2 and its expression in hematopoietic progenitors has been linked to the induction of AML in mice.36 As a third AML mouse model, we thus used mice that developed AML through the induction of MLL-AF9 expression, the product of the t(9;11)(q22;p23) translocation. Lin- BM cells derived from Gfi1bwt/wtMxCretg or Gfi1bfl/flMxCretg mice were transduced with a retrovirus expressing MLL-AF9 and transplanted into lethally irradiated C57BL/6J mice. For Cre-mediated excision of Gfi1b in the transplanted cells, mice were injected with poly(I:C) three weeks after transplantation (Figure 4E). Poly(I:C)-injected mice with MLL-AF9-transduced Gfi1bfl/flMxCretg (Gfi1b-deficient) cells succumbed faster to leukemia than mice injected with poly(I:C) and transplanted with MLL-AF9-transduced Gfi1bwt/wtMxCretg (Gfi1b-expressing) cells (Figure 4F). However, there were no major qualitative differences concerning cytological findings and or blood parameters

Figure 2. Different Gfi1B levels are indicative of prognosis of myelodysplastic syndrome (MDS) or acute myeloid leukemia (AML) patients. (A) High expression of GFI1B in human AML cells (n=116) compared to lower expression of GFI1B (0-15%) in AML cells (n=394) based on the patient cohort published by Verhaak et al.22 (B) Overall survival (OS) from patients described in Verhaak et al.22 with regard to GFI1B expression (P=0.0443). (C) Overall survival (OS) from patients described in Verhaak et al.22 (restricted to cytogenetically normal patients) with regard to GFI1B expression; P=0.0407. (D) Same as in (C) but with regard to event-free survival (EFS); P=0.0350. (E) Analysis of signaling pathways with low GFI1B expression (the lowest 5% compared with the highest 20% of expression levels) in a bigger dataset from Verhaak et al.22 Analysis was performed by Signaling Pathway Enrichment using Experimental Datasets (SPEED) analysis. Pathways such as reactive oxygen species (ROS; H2O2), MAPK, JAK, TGFB and TLR are highly significant. (F) High expression of GFI1B (31-100%) in human MDS patients (n=32) compared to lower expression of GFI1B (0-30%) in MDS patients (n=85) based on the patient cohort published by Papaemmanuil et al.23 (G) EFS of patients described in Papaemmanuil et al.23 with regard to GFI1B expression; P=0.032.

617

A. Thivakaran et al. A

Figure 3. Low level or absence of Gfi1b accelerates the progression of myelodysplastic syndrome (MDS) to acute myeloid leukemia (AML) in the NUP98/HOXD13 MDS mouse model. (A) Crossing of the Gfi1bwt/wt and Gfi1bEGFP/wt mouse strains with the NUP98/HOXD13 mouse model. (B) Survival of Gfi1bwt/wt and Gfi1bEGFP/wtNUP98/HOXD13 transgenic mice; P=0.0039. Number of mice succumbing to AML is indicated. (C) Mean fluorescence intensity (MFI) of the GFP expression level (and hence Gfi1b promoter activity) in Gfi1bEGFP/wt mice that died of AML before 250 days (n=7) or after 250 days (n=5); P=0.0272. (D) Wright-Giemsa staining of bone marrow (BM) cytospins from representative Gfi1bEGFP/wt and Gfi1bwt/wt NUP98/HOXD13 leukemic mice (bar=20 mm). (E) Crossing of the Gfi1bfl/flMxCrewt and Gfi1bfl/flMxCretg mouse strains with the NUP98/HOXD13 mouse model. After Cre-mediated deletion of the Gfi1b gene upon poly(I:C) administration, the mice were monitored for signs of leukemia. (F) Survival of Gfi1bfl/flMxCrewt (Gfi1bwt/wt) and Gfi1bfl/flMxCretg (Gfi1bKO/KO) mice transgenically expressing NUP98/HOXD13 after poly(I:C) administration; P<0.0001. Number of mice succumbing to AML is indicated. (G) Polymerase chain reaction genotyping of DNA from BM cells of poly(I:C)-injected Gfi1bfl/flMxCrewt and Gfi1bfl/flMxCretg NUP98/HOXD13 leukemic mice. (H) Wright-Giemsa staining of BM cytospins from representative poly(I:C)-injected Gfi1bfl/flMxCrewt and Gfi1bfl/flMxCretgNUP98/HOXD13 leukemic mice (bar=20 mm). (I) The frequency of monocytes (Mac-1hiGr-1int) (left panel, ****P<0.0001), granulocytes (Mac1hiGr-1hi) (middle panel, *P=0.0206) and CD117+ (c-Kit) cells (right panel, ****P<0.0001) in the BM of mice described in (F) (n=15 for Gfi1bfl/flMxCrewt mice; n=13 for Gfi1bfl/flMxCretgNUP98/HOXD13 mice).

618

haematologica | 2018; 103(4)

Gfi1b in AML and MDS

(Figure 4G and data not shown). We did not observe a significant change in the number of overall myeloid cells or apoptosis level in the different settings (Figure 4H and Online Supplementary Figure S6A-D).

Loss of Gfi1b increases the number of LSCs Loss of Gfi1b leads to an expansion in the number of functional HSCs;6 therefore we investigated whether the same applies to LSCs. We performed a limiting dilution assay by transplanting MLL-AF9 leukemic BM cells derived from poly(I:C)-treated Gfi1bfl/flMxCrewt or Gfi1bfl/flMxCretg leukemic mice into sublethally irradiated congenic mice (Figure 5A). Gfi1b-deficient MLL-AF9 BM cells had a LSC frequency of 1:3500 compared with an LSC frequency of 1:63000 cells in Gfi1b-expressing leukemic cells (Figure 5B). The increased number of func-

tional LSCs in Gfi1b-deficient leukemic cells could explain why loss of Gfi1b accelerated disease progression, as it has already been shown that a higher number of LSCs is associated with a poor prognosis of leukemia patients.37 However, this hypothesis needs to be confirmed in independent experiments.

Loss of Gfi1b induces gene expression changes supporting AML development To further study the molecular function of Gfi1b in AML, we performed whole genome gene expression analysis using Gfi1b-expressing and Gfi1b-deficient NUP98/HOXD13tg leukemic mice (Figure 6A). This model was used since the difference between Gfi1b-deficient and Gfi1b-expressing leukemic cells was most striking in the NUP98/HOXD13tg mouse model. Using gene set enrich-

haematologica | 2018; 103(4)

Figure 4. Absence of Gfi1b accelerates the progression of myeloproliferative disorder and acute myeloid leukemia (AML). (A) Crossing of the Gfi1bwt/wtMxCretg and Gfi1bfl/flMxCretg mouse strains with the Kras+/fl mouse model. (B) Survival of Gfi1bwt/wtMxCretg and Gfi1bfl/flMxCretg mice transgenically expressing Kras after poly (I:C) administration; ****P<0.0015. Numbers indicate the number of mice succumbing to AML. (C) WrightGiemsa staining of bone marrow (BM) cytospins from representative Gfi1bwt/wtMxCretg and Gfi1bfl/flMxCrewt leukemic mice transgenically expressing Kras after poly(I:C) administration (bar=20 mm). (D) Flow cytometric analysis of the BM from the leukemic mice shown in (B) with regard to Gr-1 and Mac-1 expression. (E) Isolation, transduction and transplantation of lineage-negative (Lin-) cells from Gfi1bfl/flMxCrewt and Gfi1bfl/flMxCretg mice with MLL-AF9-expressing retrovirus. After Cre-mediated deletion of the Gfi1b gene upon poly(I:C) administration, the mice were monitored for signs of leukemia. (F) Survival of the mice transplanted with Gfi1bfl/flMxCrewt and Gfi1bfl/flMxCretg MLL-AF9 transduced cells; ***P=0.0002. Number of mice succumbing to AML is indicated. (G) Wright-Giemsa staining of BM cytospins from the leukemic mice described in (F) (bar=20 mm). (H) Flow cytometric analysis of the BM from the leukemic mice described in (F) with regard to Gr-1 and Mac-1 expression.

619

A. Thivakaran et al. A

Figure 5. Loss of Gfi1b increases the stemness of LSCs. (A) In vivo limiting dilution assay for determination of functional LSCs. The indicated numbers of poly (I:C)-treated Gfi1bfl/flMxCrewt and Gfi1bfl/flMxCretg MLL-AF9 transduced cells were retransplanted into sublethally irradiated mice. (B) Survival of the mice transplanted with different numbers of poly(I:C)-treated Gfi1bfl/flMxCrewt MLL-AF9 and Gfi1bfl/flMxCretg MLL-AF9 transduced cells. (C) Determination of functional LSCs by limiting dilution assay.

ment analysis (GSEA), loss of Gfi1b was associated with a signature showing enrichment of genes involved in AML development as well as regulation of stemness (Figure 6B). This is of interest since we observed an increase in the number of LSCs upon deletion of Gfi1b. Gfi1b recruits different histone-modifying enzymes, among them HDACs,16 to its target genes. This in turn leads to deacetylation of H3K9, which leads to epigenetic silencing of the particular Gfi1b target genes.16 We, therefore, analyzed the genome-wide H3K9 acetylation level of leukemic blasts from Gfi1b-expressing and Gfi1b-deficient NUP98/HOXD13tg leukemic mice. Loss of Gfi1b leads to a genome-wide increase in H3K9 acetylation level (Figure 6C). In a subsequent step, we analyzed those genes, which showed an elevated level of H3K9 acetylation in Gfi1b-deficient leukemic cells compared to the H3K9 acetylation level of the same genes found in Gfi1b-expressing leukemic cells. Using GSEA, we found a significant enrichment of gene sets associated with the regulation of cell growth and MAPK signaling (Figure 6D). We also performed a Kyoto encyclopedia of genes and 620

genomes (KEGG) pathway analysis of those genes, which exhibited differentially H3K9 acetylated promoter areas in Gfi1b-expressing and Gfi1b-deficient leukemic cells. We found a number of processes involved in erythroid regulation (Online Supplementary Figure S7A and B), which is a main function of Gfi1b and hence underscores the validity of our results.6,9,16 Finally, we analyzed the differentially acetylated genes in Gfi1b-deficient and Gfi1b-expressing leukemic cells and compared these gene sets based on the gene sets provided by the Molecular Signatures Database (MSigDB). Using this approach, we repeatedly found signatures associated with p38 (Figure 6E). We observed increased H3K9 acetylation of the promoter area of genes involved in stem cell function in Gfi1b-deficient leukemic cells, and these epigenetic changes correlated with the gene-expression changes described above (Figure 6B). As described, gene expression arrays revealed an enrichment of a stem cell/leukemic stem cell gene signature in Gfi1b-deficient leukemic cells. To validate these results, we selected 14 different genes haematologica | 2018; 103(4)

Gfi1b in AML and MDS

Figure 6. Loss of Gfi1b induces gene expression changes supporting acute myeloid leukemia (AML) development. (A) Microarray analysis was performed with Gfi1bfl/flMxCrewtNUP98/HOXD13 and Gfi1bfl/flMxCretgNUP98/HOXD13 leukemic cells. (B) Based on the results of the micro-array analysis, a Gene set enrichment analysis (GSEA) of leukemic cells from Gfi1bfl/flMxCrewtNUP98/HOXD13tg an d Gfi1bfl/flMxCretgNUP98/HOXD13tg mice w as performed. As a result, Gfi1b deficient leukemic cells showed an enrichment of the gene set of VALK AML cluster 8 with a normalized enrichment score (NES) of 2.1 and false discovery rate (FDR) of qval=0.00191213. Gfi1b deficient cells showed also an enrichment for RAMALHO STEMNESS with an NES=2.41 and an FDR q-val=0.(C) ChIP and ChIP-Seq analysis was performed with and Gfi1bfl/flMxCrewtNUP98/HOXD13 fl/fl tg Gfi1b MxCre NUP98/HOXD13 leukemic cells. ChIP-Seq analysis for differences in the frequency of H3K9 acetylation of Gfi1b-deficient (Gfi1bfl/flMxCretg) leukemic blasts from NUP98/HOXD13tg mice compared to leukemic cells with normal Gfi1b expression (Gfi1bfl/flMxCrewt). (D) GSEA of genes with an elevated acetylation level in Gfi1b deficient mice are associated with regulation of cell growth NES=1.78 and FDR q-val=0.055857178 and GNF2_MAP2K3 NES=2.02 and FDR qval=1.7715618E-4. (E) Upon analyzing the differentially acetylated genes and using the MSigDB Pathway approach, we found significant enrichment of different pathways, among them the p38 pathway. P-value was used to rank the enrichment.

(Abcg2, Gata3, Itga2, Thy1, Cd24a, Pecam1, Prom1, Plaur, Klf4, Mycn, Ptch1, Pecam1, Sav1, and Notch1), which were differentially expressed by more than 2-fold between Gfi1b-expressing and Gfi1b-deficient leukemic mice in the gene expression arrays. We selected these genes based on their diverse role in regulating stem cell function. We then examined these genes and confirmed that these genes were also differentially expressed using RT-PCR (Online Supplementary Figure S8A).

Loss of Gfi1b leads to increased ROS levels and decreased levels of activated p38 To obtain further insight into the molecular mechanism behind our observation we compared the whole genome haematologica | 2018; 103(4)

gene expression pattern in murine Gfi1b-expressing and Gfi1b-deficient leukemic cells based on an AltAnalyze approach (see Online Supplementary Appendix). Using the same algorithm, we compared the gene expression pattern found in AML blasts with low GFI1B expression and high GFI1B expression (data obtained from published studies from Valk et al.21 and Verhaak et al.22). For the analysis of the human dataset, we analyzed the 10% of patients with the lowest and the 20% of patients with the highest GFI1B expression level in order to have enough observations from which to draw any conclusions. Then we compared which pathways were similarly deregulated in the human and murine leukemia sets. ROS and MAPK signaling were among the pathways differentially expressed between both murine and human 621

A. Thivakaran et al.

Gfi1b/GFI1B-deficient/low and Gfi1b/GFI1B highexpressing leukemic cells (Figure 7A). As ROS plays an important role in the pathogenesis of AML,38 we examined the level of ROS. For non-malignant HSCs, it was shown that HSCs with low ROS had a high self-renewal capacity.39 In contrast, HSCs with elevated ROS were mostly located in the vascular niche, had a reduced selfrenewal capacity, and were more restricted with regard to their differentiation potential.39 Based on this and our previous report that loss of Gfi1b leads to higher level of ROS in HSCs,6 we examined whether ROS level might differ between Gfi1b-deficient and Gfi1b expressing LSCs. Due to the difficulty of defining a distinct LSC population in each set of AML samples, we used CD117 (c-Kit) expression as a surrogate marker to define a population which is enriched for LSCs. CD117 expression has been used to identify a fraction that is enriched for LSCs.40 We identified two distinct populations in the CD117+ blast cells that differ with regard to their ROS expression (a population with low ROS expression and a population with high ROS expression). Loss of Gfi1b led to an increased level of ROS (defined as the mean fluorescent intensity, MFI) in both ROS-low and ROS-high populations (Figure 7B-D and Online Supplementary Figure S8B).

Altered activity of the AKT pathway in Gfi1b-deficient AML Elevated levels of ROS promote AML development, but ROS also activates various redox-sensitive signaling transduction cascades,41 including the MAPK pathway, which limits the stemness of the affected cells, at least in a non-malignant setting.42 In the presence of ROS, the MAPK pathway component p38 is activated, which subsequently results in an exhaustion of the HSC population.43 It has also been shown that activation of p38 limits oncogenic transformation.44 Despite a higher level of ROS, in our models, Gfi1b-deficient leukemic cells have escaped p38 activation, indicated by a decreased level of phosphorylated p38 compared to Gfi1bfl/flMxCrewtNUP98/HOXD13tg (Figure 7E). The fact that Gfi1b might directly or indirectly regulate p38 is also supported by the analysis of differentially acetylated genes in Gfi1b-deficient and Gfi1b-expressing leukemic cells. Because decreased p38 levels are associated with higher oncogenic potential,44 this could partially explain the higher number of functional LSCs we observed in the Gfi1b-deficient leukemic cells. Activation of p38 leads to an increased level of AktSer473.45 AktSer473 activity inversely correlates with the number of LSCs in AML.46 We thus examined the connection between loss of Gfi1b and AktSer473 and found that the level of phosphorylated AktSer473 is reduced in Gfi1b-deficient leukemia (Figure 7F). Akt represses the function of FoXO3, and since FoXO3 acts as an oncogene in AML,46 we determined the FoXO3 protein level. Gfi1b-deficient leukemic cells showed an increased expression of FoXO3 protein in the nucleotide (NER) and cytoplasmatic (CER) cell fraction compared to the expression level of FoXO3 in Gfi1b-expressing leukemic cells (Figure 7G). To obtain an insight into whether this increased level of FoXO3 also has any functional consequences, we re-examined the whole genome expression datasets in Gfi1b-expressing and Gfi1b-deficient leukemic cells and found an enrichment of FoXO3 binding sites among the promoter areas of those genes, which were differentially expressed between cells with 622

absence of Gfi1b expression and cells with intact expression of Gfi1b (Online Supplementary Figure S9), showing that altered level of FoXO3 might be one additional explanation for our observations (Figure 7H).

Discussion In the datasets analyzed by us, GFI1B was expressed at a lower level in LSCs compared to the control. Low GFI1B was also indicative of an inferior prognosis for MDS and AML patients, with the caveat that these statements are based on retrospective studies. Larger prospective studies would be required to make such a claim on a solid basis. We previously reported that low GFI1 expression level in AML blasts was associated with poor outcome and here we report that low GFI1B expression levels were associated with poor survival. This might appear surprising since GFI1 and GFI1B repress each other, but in our cohorts we observed that low GFI1B expression level can also be associated with low GFI1 expression (data not shown), therefore, the reciprocal regulation between GFI1 and GFI1B might be different in leukemic cells. We postulate that GFI1B plays a dose-dependent role in human/murine AML pathogenesis. Anguita et al. showed that a mutated dominant-negative form of GFI1B contributes to AML development. These reports show how altering the function of GFI1B can influence normal and malignant development. Recent studies have highlighted the role of different isoforms of GFI1B in the course of erythroid and megakaryocytic development.6,47-49 It remains to be elucidated whether altering the expression of these isoforms might also contribute to AML development. Loss of Gfi1b in our murine models increased the number of LSCs on a functional level. These data are in line with our previous reports that loss of Gfi1b leads to an expansion of functional HSCs.6 On a molecular level, loss of Gfi1b resulted in an increased level of H3K9ac among its target genes, which is in line with other reports regarding the epigenetic function of Gfi1b.11,16 Among these target genes are a number of genes involved in the regulation of leukemogenesis and stem cell regulation, indicating that the absence of Gfi1b leads to a gene expression signature that directly or indirectly contributes to an increased number of LSCs. Both murine and human data also indicated a possible connection between Gfi1b and ROS/p38/Akt signaling. P38 and AktSer473 activation limit oncogenic and stemness potential.43,44,46 Conceivably, lower expression of these proteins would increase the oncogenic potential. P38 and Akt were down-regulated in Gfi1b-deficient leukemic cells in vivo. In line with this, Saleque et al. demonstrated that Gfi1b is involved in the regulation of p38 and that reduced Gfi1b levels are associated with lower p38 signaling.50 In addition to the ROS/p38/Akt/FoXO3 signaling cascade, other pathways were altered. It remains to be elucidated which role these pathways might play in the pathogenesis of human and murine AML. In addition, how Gfi1b/GFI1B regulates ROS, p38, Akt and FoXO3 levels remains to be analyzed. In summary, epigenetic changes and alteration of the ROS/p38/Akt/FoXO signaling cascade might facilitate the progression of normal hematopoietic cells to LSCs. In the future, testing will be needed to determine whether alterhaematologica | 2018; 103(4)

Gfi1b in AML and MDS

Figure 7. Loss of Gfi1b deregulates acute myeloid leukemia (AML) signaling pathway. (A) Analysis to see which pathways were enriched in GFI1B low-expressing blast cells in two independent sets of myelodysplastic syndrome (MDS) or AML patients compared to the expression pattern found in GFI1B high-expressing MDS/AML blast cells. The same approach was repeated for Gfi1b-expressing and Gfi1b-non-expressing leukemic cells from our mice experiments. As an overlap, enrichment was observed in pathways of JAK-STAT-, MAPK- and ROS-related signaling. (B) Representative flow cytometric analysis of bone marrow (BM) from Gfi1bfl/flMxCretgNUP98/HOXD13tg mice compared to Gfi1bfl/flMxCrewtNUP98/HOXD13tg mice showing the gating strategy for determining ROS low and ROS high levels. (C) Mean fluorescence intensity (MFI) for ROS in the ROS-low population of c-Kit+ blast cells derived from Gfi1bfl/flMxCrewtNUP98/HOXD13tg (n=6) and Gfi1bfl/flMxCretgNUP98-HOXD13tg (n=5); *P=0.0488. (D) Mean fluorescence intensity (MFI) for ROS in the ROS-high population of c-Kit+ blast cells derived from Gfi1bfl/flMxCrewtNUP98-HOXD13tg (n=6) and Gfi1bfl/flMxCretg NUP98/HOXD13tg (n=5); *P=0.0191. (E) Flow cytometric analysis of p38 MAPK (pT180/pY182) in CD117+ blast cells derived from Gfi1bfl/flMxCrewtNUP98/HOXD13tg (n=5) and Gfi1bfl/flMxCretg NUP98/HOXD13tg (n=6); *P=0.0144. (F) Flow cytometric analysis of Akt (pS473) in c-Kit+ blast cells derived from Gfi1bfl/flMxCrewtNUP98/HOXD13tg (n=4) and Gfi1bfl/flMxCretg NUP98/HOXD13tg (n=5); **P=0.0040. (G) FoXO3 protein level was detected in nuclear extraction (NER)- and cytoplasmic extraction (CER)-derived BM cells from Gfi1bfl/flMxCretgNUP98/HOXD13tg and Gfi1bfl/flMxCrewtNUP98/HOXD13tg. (H) Working model hypothesis: normal levels of Gfi1b lead to reduced ROS levels, resulting in normal maturation and differentiation of progenitor cells. Loss of Gfi1b in leukemic cells is associated with higher ROS levels, which have been shown to promote AML development. However, through a still undefined mechanism, this results in lower levels of p38 MAPK and pAkt and higher levels of unphosphorylated FoXO3, which might explain the increased number of functional leukemic stem cells in the Gfi1b-deficient AML population. LSC: leukemic stem cells.

haematologica | 2018; 103(4)

623

A. Thivakaran et al.

ation of the ROS pathway could be a targeted therapeutic approach to treat AML patients with low GFI1B expression.

Funding This work was supported by the Deutsche Forschungsgemeinschaft and the IFORES program of University Hospital Essen.

Acknowledgments We thank the animal facility of University Hospital Essen.

References 1. Tefferi A, Vardiman JW. Myelodysplastic syndromes. N Engl J Med. 2009; 361(19):1872-1885. 2. Dohner 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. 3. Calero-Nieto FJ, Ng FS, Wilson NK, et al. Key regulators control distinct transcriptional programmes in blood progenitor and mast cells. EMBO J. 2014;33(11):12121226. 4. Hones JM, Botezatu L, Helness A, et al. GFI1 as a novel prognostic and therapeutic factor for AML/MDS. Leukemia. 2016; 30(6):1237-1245. 5. Gentner B, Pochert N, Rouhi A, et al. MicroRNA-223 dose levels fine tune proliferation and differentiation in human cord blood progenitors and acute myeloid leukemia. Exp Hematol. 2015;43(10):858868. 6. Khandanpour C, Sharif-Askari E, Vassen L, et al. Evidence that growth factor independence 1b regulates dormancy and peripheral blood mobilization of hematopoietic stem cells. Blood. 2010;116(24):5149-5161. 7. Laurent B, Randrianarison-Huetz V, Marechal V, Mayeux P, Dusanter-Fourt I, Dumenil D. High-mobility group protein HMGB2 regulates human erythroid differentiation through trans-activation of GFI1B transcription. Blood. 2010;115(3):687-695. 8. Foudi A, Kramer DJ, Qin J, et al. Distinct, strict requirements for Gfi-1b in adult bone marrow red cell and platelet generation. J Exp Med. 2014;211(5):909-927. 9. Saleque S, Cameron S, Orkin SH. The zincfinger proto-oncogene Gfi-1b is essential for development of the erythroid and megakaryocytic lineages. Genes Dev. 2002; 16(3):301-306. 10. Vassen L, Beauchemin H, Lemsaddek W, Krongold J, Trudel M, Moroy T. Growth factor independence 1b (gfi1b) is important for the maturation of erythroid cells and the regulation of embryonic globin expression. PLoS One. 2014;9(5):e96636. 11. van der Meer LT, Jansen JH, van der Reijden BA. Gfi1 and Gfi1b: key regulators of hematopoiesis. Leukemia. 2010; 24(11):1834-1843. 12. Schulz D, Vassen L, Chow KT, et al. Gfi1b negatively regulates Rag expression directly and via the repression of FoxO1. J Exp Med. 2012;209(1):187-199. 13. Vassen L, Okayama T, Moroy T. Gfi1b:green fluorescent protein knock-in mice reveal a dynamic expression pattern of Gfi1b during hematopoiesis that is largely complementary to Gfi1. Blood. 2007;109(6):2356-2364.

624

14. Osawa M, Yamaguchi T, Nakamura Y, et al. Erythroid expansion mediated by the Gfi-1B zinc finger protein: role in normal hematopoiesis. Blood. 2002;100(8):27692777. 15. Vassen L, Fiolka K, Moroy T. Gfi1b alters histone methylation at target gene promoters and sites of gamma-satellite containing heterochromatin. EMBO J. 2006; 25(11):2409-2419. 16. Saleque S, Kim J, Rooke HM, Orkin SH. Epigenetic regulation of hematopoietic differentiation by Gfi-1 and Gfi-1b is mediated by the cofactors CoREST and LSD1. Mol Cell. 2007;27(4):562-572. 17. McGhee L, Bryan J, Elliott L, et al. Gfi-1 attaches to the nuclear matrix, associates with ETO (MTG8) and histone deacetylase proteins, and represses transcription using a TSA-sensitive mechanism. J Cell Biochem. 2003;89(5):1005-1018. 18. Duan Z, Zarebski A, Montoya-Durango D, Grimes HL, Horwitz M. Gfi1 coordinates epigenetic repression of p21Cip/WAF1 by recruitment of histone lysine methyltransferase G9a and histone deacetylase 1. Mol Cell Biol. 2005;25(23):10338-10351. 19. Wouters BJ, Lowenberg B, ErpelinckVerschueren CA, van Putten WL, Valk PJ, Delwel R. Double CEBPA mutations, but not single CEBPA mutations, define a subgroup of acute myeloid leukemia with a distinctive gene expression profile that is uniquely associated with a favorable outcome. Blood. 2009;113(13):3088-3091. 20. Rapin N, Bagger FO, Jendholm J, et al. Comparing cancer vs normal gene expression profiles identifies new disease entities and common transcriptional programs in AML patients. Blood. 2014;123(6):894-904. 21. Valk PJ, Verhaak RG, Beijen MA, et al. Prognostically useful gene-expression profiles in acute myeloid leukemia. N Engl J Med. 2004;350(16):1617-1628. 22. Verhaak RG, Wouters BJ, Erpelinck CA, et al. Prediction of molecular subtypes in acute myeloid leukemia based on gene expression profiling. Haematologica. 2009;94(1):131-134. 23. Papaemmanuil E, Gerstung M, Malcovati L, et al. Clinical and biological implications of driver mutations in myelodysplastic syndromes. Blood. 2013;122(22):3616-3627. 24. Gerstung M, Pellagatti A, Malcovati L, et al. Combining gene mutation with gene expression data improves outcome prediction in myelodysplastic syndromes. Nat Commun. 2015;6:5901. 25. Bagger FO, Sasivarevic D, Sohi SH, et al. BloodSpot: a database of gene expression profiles and transcriptional programs for healthy and malignant haematopoiesis. Nucleic Acids Res. 2016;44(D1):D917-924. 26. Al-Matary YS, Botezatu L, Opalka B, et al. Acute myeloid leukemia cells polarize macrophages towards a leukemia support-

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38. 39.

40.

ing state in a Growth factor independence 1 dependent manner. Haematologica. 2016; 101(10):1216-1227. Botezatu L, Michel LC, Helness A, et al. Epigenetic therapy as a novel approach for GFI136N-associated murine/human AML. Exp Hematol. 2016;44(8):713-726 e714. Khandanpour C, Krongold J, Schuette J, et al. The human GFI136N variant induces epigenetic changes at the Hoxa9 locus and accelerates K-RAS driven myeloproliferative disorder in mice. Blood. 2012; 120(19):4006-4017. Hu Y, Smyth GK. ELDA: Extreme limiting dilution analysis for comparing depleted and enriched populations in stem cell and other assays. J Immunol Methods. 2009; 347(1â&#x20AC;&#x201C;2):70-78. Bonnet D, Dick JE. Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat Med. 1997;3(7):730-737. 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):138152. Anguita E, Gupta R, Olariu V, et al. A somatic mutation of GFI1B identified in leukemia alters cell fate via a SPI1 (PU.1) centered genetic regulatory network. Dev Biol. 2016;411(2):277-286. Chowdhury AH, Ramroop JR, Upadhyay G, Sengupta A, Andrzejczyk A, Saleque S. Differential transcriptional regulation of meis1 by Gfi1b and its co-factors LSD1 and CoREST. PLoS One. 2013;8(1):e53666. Lin YW, Slape C, Zhang Z, Aplan PD. NUP98-HOXD13 transgenic mice develop a highly penetrant, severe myelodysplastic syndrome that progresses to acute leukemia. Blood. 2005;106(1):287-295. Chan IT, Kutok JL, Williams IR, et al. Conditional expression of oncogenic K-ras from its endogenous promoter induces a myeloproliferative disease. J Clin Invest. 2004;113(4):528-538. Krivtsov AV, Twomey D, Feng Z, et al. Transformation from committed progenitor to leukaemia stem cell initiated by MLL-AF9. Nature. 2006;442(7104):818-822. Meyer LH, Eckhoff SM, Queudeville M, et al. Early relapse in all is identified by time to leukemia in NOD/SCID mice and is characterized by a gene signature involving survival pathways. Cancer Cell. 2011; 19(2):206-217. Hole PS, Darley RL, Tonks A. Do reactive oxygen species play a role in myeloid leukemias? Blood. 2011;117(22):5816-5826. Jang YY, Sharkis SJ. A low level of reactive oxygen species selects for primitive hematopoietic stem cells that may reside in the low-oxygenic niche. Blood. 2007;110(8):3056-3063. Wang Y, Krivtsov AV, Sinha AU, et al. The

haematologica | 2018; 103(4)

Gfi1b in AML and MDS

Wnt/beta-catenin pathway is required for the development of leukemia stem cells in AML. Science. 2010;327(5973):1650-1653. 41. Hole PS, Zabkiewicz J, Munje C, et al. Overproduction of NOX-derived ROS in AML promotes proliferation and is associated with defective oxidative stress signaling. Blood. 2013;122(19):3322-3330. 42. Essers MA, Offner S, Blanco-Bose WE, et al. IFNalpha activates dormant haematopoietic stem cells in vivo. Nature. 2009;458(7240):904-908. 43. Ito K, Hirao A, Arai F, et al. Reactive oxygen species act through p38 MAPK to limit the lifespan of hematopoietic stem cells. Nat Med. 2006;12(4):446-451.

haematologica | 2018; 103(4)

44. Bulavin DV, Fornace AJ Jr. p38 MAP kinase's emerging role as a tumor suppressor. Adv Cancer Res. 2004;92:95-118. 45. Park S, Chapuis N, Tamburini J, et al. Role of the PI3K/AKT and mTOR signaling pathways in acute myeloid leukemia. Haematologica. 2010;95(5):819-828. 46. Sykes SM, Lane SW, Bullinger L, et al. AKT/FOXO signaling enforces reversible differentiation blockade in myeloid leukemias. Cell. 2011;146(5):697-708. 47. Polfus LM, Khajuria RK, Schick UM, et al. Whole-Exome Sequencing Identifies Loci Associated with Blood Cell Traits and Reveals a Role for Alternative GFI1B Splice Variants in Human Hematopoiesis. Am J

Hum Genet. 2016;99(3):785. 48. Schulze H, Schlagenhauf A, Manukjan G, et al. Recessive grey platelet-like syndrome with unaffected erythropoiesis in the absence of the splice isoform GFI1B-p37. Haematologica. 2017;102(9):e375-e378. 49. Vassen L, Khandanpour C, Ebeling P, et al. Growth factor independent 1b (Gfi1b) and a new splice variant of Gfi1b are highly expressed in patients with acute and chronic leukemia. Int J Hematol. 2009;89 (4):422-430. 50. Sengupta A, Upadhyay G, Sen S, Saleque S. Reciprocal regulation of alternative lineages by Rgs18 and its transcriptional repressor Gfi1b. J Cell Sci. 2016;129(1):145-154.

625

ARTICLE

Acute Myeloid Leukemia

Ferrata Storti Foundation

Haematologica 2018 Volume 103(4):626-633

Association of mutations with morphological dysplasia in de novo acute myeloid leukemia without 2016 WHO Classification-defined cytogenetic abnormalities Olga K. Weinberg,1 Christopher J. Gibson,2 Traci M. Blonquist,3 Donna Neuberg,3 Olga Pozdnyakova,4 Frank Kuo,4 Benjamin L. Ebert5 and Robert P. Hasserjian5

1 Department of Pathology, Boston Children’s Hospital; 2Division of Hematology, Brigham and Women’s Hospital, Dana Farber Cancer Institute; 3Department of Biostatistics and Computational Biology, Dana Farber Cancer Institute; 4Department of Pathology, Brigham and Women’s Hospital and 5Department of Pathology, Massachusetts General Hospital, Boston, MA, USA

ABSTRACT

Correspondence: olga.weinberg@childrens.harvard.edu

Received: September 29, 2017. Accepted: January 4, 2018. Pre-published: January 11, 2018. doi:10.3324/haematol.2017.181842 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/626 ©2018 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.

626

espite improvements in our understanding of the molecular basis of acute myeloid leukemia (AML), the association between genetic mutations with morphological dysplasia remains unclear. In this study, we evaluated and scored dysplasia in bone marrow (BM) specimens from 168 patients with de novo AML; none of these patients had cytogenetic abnormalities according to the 2016 World Health Organization Classification. We then performed targeted sequencing of diagnostic BM aspirates for recurrent mutations associated with myeloid malignancies. We found that cohesin pathway mutations [q (FDRadjusted P)=0.046] were associated with a higher degree of megakaryocytic dysplasia and STAG2 mutations were marginally associated with greater myeloid lineage dysplasia (q=0.052). Frequent megakaryocytes with separated nuclear lobes were more commonly seen among cases with cohesin pathway mutations (q=0.010) and specifically in those with STAG2 mutations (q=0.010), as well as NPM1 mutations (q=0.022 when considering the presence of any vs. no megakaryocytes with separated nuclear lobes). RAS pathway mutations (q=0.006) and FLT3-ITD (q=0.006) were significantly more frequent in cases without evaluable erythroid cells. In univariate analysis of the 153 patients treated with induction chemotherapy, NPM1 mutations were associated with longer event-free survival (EFS) (P=0.042), while RUNX1 (P=0.042), NF1 (P=0.040), frequent micromegakaryocytes (P=0.018) and presence of a subclone (P=0.002) were associated with shorter EFS. In multivariable modeling, NPM1 was associated with longer EFS, while presence of a subclone and frequent micromegakaryocytes remained significantly associated with shorter EFS. Introduction Acute myeloid leukemia (AML) is a complex and dynamic disease, characterized by multiple somatically acquired driver mutations, co-existing competing clones, and disease evolution over time.1-3 A major challenge in the application of mutation information to clinical management of AML patients is how to integrate it into existing AML risk stratification and classification schemes. The 2016 WHO Classification incorporates cytogenetics, clinical ontogeny, dysplastic morphology in background hematopoietic cells, and the status of certain mutations (NPM1, RUNX1, and CEBPA).4 One main distinction between the WHO Classification and other AML risk stratification schemes, such as the European LeukemiaNet5 and National Comprehensive Cancer Network (NCCN),6 is its use of dysplastic morphology to categorize cases in an adverse risk group, AML with myelodysplasiarelated changes (AML-MRC). Although there is general agreement as to the adverse effect of MDS-related cytogenetics on the outcome of de novo AML, the clinical and biological significance of morphological dysplasia in this setting is controversial.7 haematologica | 2018; 103(4)

Correlation of mutations with morphological findings in AML

Studies have shown that, although dysplasia is frequently seen in NPM1-mutated AML, it does not confer a worse clinical outcome.8,9 Another study found that in AML patients with wild-type NPM1, those with multilineage dysplasia showed an inferior response to induction and, in younger patients, a lower 5-year survival, suggesting that the prognostic relevance of multilineage dysplasia in AML might depend on NPM1 mutation status.10 In a recent study of patients with de novo AML lacking specific cytogenetic findings, we reported that WHOdefined multilineage dysplasia had no impact on outcome, but the presence of certain specific dysplastic features (micromegakaryocytes and hypogranulated myeloid cells) was associated with adverse EFS.11 However, this study did not evaluate the impact of gene mutations other than FLT3 and NPM1. Moreover, despite the recently expanded knowledge of the mutations affecting multiple functional pathways in AML, an association of mutations with specific dysplastic features has not been closely evaluated. In myelodysplastic syndromes (MDS), Della Porta et al. found an association between granulocytic dysplasia and mutations in ASXL1, RUNX1, TP53 and SRSF2.12 Devillier et al. evaluated 94 patients with AML-MRC and found that ASXL1 mutations were associated with a higher degree of dysgranulopoiesis, but not dyserythropoiesis or dysmegakaryopoiesis.13 However, this study did not evaluate individual dysplastic features, and included patients with both MDS-related cytogenetics and a prior history of MDS, who are known to have a poor prognosis and frequent ASXL1 mutations.14,15 In AML cases without a history of MDS or MDS-associated karyotype findings, it is uncertain if the presence of background morphological dysplasia indicates a true relationship to MDS. The goal of the current study is to analyze the association between dysplastic findings and somatic mutations in de novo AML. We investigated the associations between specific dysplastic features with individual mutations, gene pathway alterations, and clonal architecture, and explored the effects of these parameters on patients' outcome.

Methods Patients Cases of newly diagnosed de novo AML were identified from the pathology archives of Brigham and Women’s Hospital/DanaFarber Cancer Institute and Massachusetts General Hospital between 2009-2016. This study has been approved by an institutional review board (IRB) (IRB Protocol n. 2009P001369). All cases had bone marrow (BM) aspirate smear and biopsy slides available for review that were diagnosed as AML prior to any therapy being administered. Only cases with adequate karyotype and clinical follow-up information were included. Patients who had received any prior cytotoxic therapy, had a prior diagnosis of any myeloid neoplasm, or had defining cytogenetic abnormalities of AMLMRC or AML with recurrent genetic abnormalities according to the 2016 WHO Classification4 were excluded.

Morphology assessment Bone marrow aspirate and biopsy smears from each case were viewed in a blinded fashion by 3 hematopathologists (OW, RH and OP) who scored dysplasia in each lineage in increments of 10%, as previously described;11 the median score for all 3 observers was used for all analyses. A minimum of 10 megakaryhaematologica | 2018; 103(4)

ocytes (on biopsies and/or aspirate smears) and 20 erythroids and 20 myeloid elements (in aspirate smears) were required, otherwise a lineage was designated as “not evaluable”. Specific dysplastic features in each lineage were also scored on a semi-quantitative scale (<10% cells showing the dysplastic feature = 0, 10-25%=1, 26-50%=2, 51-75%=3, >75%=4). Specific dysplastic features scored in the erythroid lineage were: 1) megaloblastoid change; 2) multinucleation; 3) nuclear irregularities; and 4) pyknosis. Dysgranulopoiesis features scored were: 1) abnormal nuclear shape (including pseudo Pelger-Huet anomaly); and 2) hypogranulation. Dysmegakaryopoiesis features scored were: 1) micromegakaryocytes; 2) presence of two or multiple separated nuclear lobes; and 3) megakaryocytes with hypolobated or monolobated nuclei.

Clinical data The complete blood count (CBC) and white blood cell (WBC) count differential results at the time of AML diagnosis were recorded. Type of treatment, including date of any allogeneic stem cell transplant (SCT), date of relapse or disease refractoriness to two induction regimens, and status at last follow up were recorded for each patient. Complete remission (CR) was determined as defined by clinical standards13 and follow-up information (relapse, death) was recorded.

Targeted sequencing We performed targeted sequencing on BM aspirates obtained at the time of diagnosis for all 168 patients, as previously described.16-18 We enriched target regions of 87 genes (Baylor Custom SureSelect hybrid capture system, Agilent Technologies, Santa Clara, CA, USA) in 105 patients or 95 genes [Rapid Heme Panel kit, Illumina Truseq Custom Amplicon (TSCA), San Diego, CA, USA] in 66 patients, which were selected on the basis of pathogenic involvement in myeloid malignancies, on either banked DNA samples or diagnostic BM aspirate smears. We classified variants as pathogenic driver mutations based on mutation type and position, and on frequency in publicly available single nucleotide polymorphism databases. The median coverage was 1200X across all the genes. Average coverage across the entire hybrid capture run was 500X. The minimal coverage was 50X, or at least 5 alternative reads required to call a variant. Variants with a variant allele fraction of less than 0.05 were excluded to ensure consistency across both sequencing platforms. FLT3-ITD was identified by filtering aligned BAM files for which one side maps to exon 13, 14 or 15 of FLT3 and the other side is unmapped. These one-sided mapped reads were then scanned for the presence of duplications of at least 10 bp. Reads that contained duplications were realigned among themselves to make a final determination of FLT3-ITD status.18 Mutations were grouped into 7 pathways as follows: DNA methylation: DNMT3A, IDH1, IDH2, and/or TET2 Epigenetic regulators: ASXL1, EZH2, BCOR, SETBP1, BCORL1, SH2B3, SETD2, and/or CREBBP Transcription factors: CEBPA (single or double), RUNX1, ETV6, WT1, and/or PHF6 Cohesin complex: STAG2, PDS5B, RAD21, and/or SMC3 RAS pathway: KRAS, NRAS, ITD, FLT3, KIT, CBL, RIT1, PTPN11, and/or NF1 Spliceosome pathway: U2AF1, ZRSR2, PRPF40b, SRSF2, and/or SF3B1 We separately explored the association between morphology and genetics based on AML ontogeny as specified by Lindsley et al.,14 in which the presence of SRSF2, SF3B1, U2AF1, ZRSR2, ASXL1, EZH2, BCOR, or STAG2 mutations defined secondary type AML, the presence of an NPM1 mutation [unless any second627

O.K. Weinberg et al.

Figure 1. Examples of typical morphological dysplastic features found in de novo acute myeloid leukemia (AML). Megakaryocytes often show separated lobes (A) and small size micromegakaryocytes (B). Dysplastic changes in myeloid cells, including hypogranular cytoplasm and abnormal nuclear lobation (C). Dysplastic erythroid cells are shown with irregular nuclear contors (D).

ary-type mutation was present at >5% variant allele frequency (VAF)] implied de novo AML, and all other cases were designated as having a pan-AML mutation pattern.14

Subclonal analysis The presence of a leukemia subclone was defined as having at least one mutation at a frequency of at least 10% less than the mutation with highest VAF, and with the sum of the highest mutation and subclonal mutation VAF over 55%.19

Statistical analysis Fisher exact test and the Wilcoxon rank sum test were used to compare categorical and continuous variables, respectively. In the event that a continuous variable was compared across 3 or more groups, the Kruskall-Wallis test was used. The Kendall coefficient of concordance, W, with a correction for ties, was used to compare the 3 observersâ&#x20AC;&#x2122; score of dysplasia percentages and individual dysplastic feature scores; Kendall W ranges from 0 (no agreement) to 1 (complete agreement). The association between each degree of dysplasia in all three lineages was assessed by both continuous and categorical analyses. Individual dysplastic findings assessed by categorical analysis explored different cut offs (>0, â&#x2030;Ľ2, and â&#x2030;Ľ3) for each evaluable dysplastic feature. We explored the associations between age, WBC, peripheral blasts (%), BM cellularity (%), BM blasts (%), platelet count, and hemoglobin level, considering individual mutations present more than 5 patients (>3%), mutation pathways, molecu628

lar ontogeny, and subclone status. There were 29 planned analyses based on mutation data for each of the morphological measures and patients' clinical characteristics; therefore, a false discovery rate adjustment (q-value) was made for the planned analyses within the respective outcome. All tests were two-sided and a q-value (FDR-adjusted P-value) <0.05 was considered significant Event-free survival (EFS), defined as the time in months from the date of diagnosis to the date of disease progression or death, was explored in the subset of patients that received induction chemotherapy. Nominal P-values are presented for univariate analyses. Cox proportional hazards models with transplant as a time varying covariate were considered to evaluate the association of mutation and morphological findings with EFS. Mutation data and morphological findings underwent univariate analysis with transplant-adjusted Cox models. Multivariable Cox models were considered with mutations and morphological findings with a univariate P-value <0.2 and the backwards elimination method was used to select a final model retaining transplant and variables with a P-value <0.1.

Results Patients A total of 168 cases of de novo AML without WHOdefined cytogenetic abnormalities were identified with a median age of 60.6 years (range 19.9-86.7). A total of 137 haematologica | 2018; 103(4)

Correlation of mutations with morphological findings in AML

Table 1. Transplant-adjusted univariate analysis of factors affecting event-free survival (EFS).

Pathway mutation DNA methylation Epigenetic regulators Transcription factors Cohesin pathway Ras pathway Spliceosome pathway Individual mutation significantly associated with EFS RUNX1 NPM1 NF1 Subclone Presence of subclone Lindsley AML Ontogeny Grouping* Pan AML Secondary AML

Coefficient

Exp (coefficient) [Hazard Ratio]

SE (coefficient)

-0.213 0.372 0.163 0.34 0.027 0.446

0.808 1.451 1.177 1.404 1.027 1.562

0.224 0.217 0.225 0.283 0.219 0.233

0.342 0.086 0.469 0.23 0.903 0.055

0.537 -0.44 0.887

1.71 0.644 2.429

0.264 0.216 0.431

0.042 0.042 0.04

0.686

1.987

0.222

0.002

0.378 0.62

1.46 1.858

0.276 0.244

0.17 0.011

EXP: expansion; SE: Standard Error; AML: acute myeloid leukemia. *Reference group is de novo AML.

cases (81%) had a normal karyotype and 31 (18%) had an abnormal karyotype. The most common cytogenetic abnormalities were +8 (8 cases), +11 (4 cases), and +13 (2 cases); other trisomies were seen in 8 cases, chromosome losses in 3 cases, ring chromosome in 1 case, and balanced or unbalanced translocations in 5 cases. The individual mutations identified in more than 5 patients (>3% of the study cohort) were: NPM1 (n=72), DNMT3A (n=68), FLT3-ITD (n=44), TET2 (n=40), FLT3 (n=26), IDH1 (n=31), ASXL1 (n=30), NRAS (n=30), SRSF2 (n=28), IDH2 (n=27), RUNX1 (n=27), PTPN11 (n=20), WT1 (n=15), STAG2 (n=14), BCOR (n=10), BCORL1 (n=6), CEBPA (n=11, including 3 with double CEBPA mutations), KRAS (n=8), RIT1 (n=8), NF1 (n=7) and CBL (n=6). Mutations were grouped into pathways as follows: DNA methylation (n=120), RAS pathway (n=97), NPM1 (n=72), epigenetic regulators (n=50), transcription factors (n=51) and spliceosome (n=39). The Lindsley et al. AML ontogeny groupings were as follows: secondary type (n=60), de novo type (n=64), and pan-AML type (n=42); 2 cases could not be assigned to the AML ontogeny group because of missing information for one or more group-defining mutations.14 The Pappaemanuil et al. AML classification groupings were as follows: chromatin-spliceosome (n=60), NPM1 (n=69), CEBPA (n=3), other driver mutations (n=23), no mutations (n=5), IDH2R172 (n=6).26 Comparing the two groupings, the chromatin/spliceosome group was composed of 54 secondary and 6 pan-AML cases; the NPM1 group 64 de novo, 4 secondary, and 1 pan-AML; the other driver group 22 pan-AML and 1 secondary AML; the no mutation and IDH2R172 groups were all pan-AML. A total of 92 cases (55%) had a leukemia subclone.20 The 2016 WHO Classification categories (AML-MRC being defined as â&#x2030;Ľ50% dysplastic cells present in at least 2 lineages, in the absence of NPM1, double CEBPA, or RUNX1 mutations) included the following: AML-MRC (n=19), AML-NOS (n=50), AML-NPM1 (n=71), AML-RUNX1 (n=25), and AML-CEBPA (n=3). haematologica | 2018; 103(4)

Mutation association with morphology Specific dysplastic features were scored on a scale of 04 ( Figure 1). A summary of the specific dysplastic features in each lineage and the Kendall coefficient of concordance for each lineage is shown in Online Supplementary Table S1. The median score for all 3 observers was used for all analyses. The distribution of dysplasia scores by pathway mutation and individual mutation is presented in Figure 2. Distribution of subclones by mutations is shown in Figure 3. When evaluating the overall degree of dysplasia in each lineage according to mutational pathways, we found that cohesin mutations (q=0.046) were associated with greater overall megakaryocytic dysplasia, while RAS pathway mutations were marginally associated with greater megakaryocytic dysplasia. Evaluating mutations individually indicated that STAG2 and RIT1 were marginally associated with greater overall megakaryocyte dysplasia (q=0.064 and q=0.056, respectively) and STAG2 also marginally associated with greater overall myeloid lineage dysplasia (q=0.052). No association was detected between Lindsley et al. molecular ontogeny groupings or the presence of a subclone with the overall dysplasia in any lineages. RAS pathway mutations (q=0.006) and specifically an FLT3-ITD (q=0.006) were significantly associated with non-evaluable erythroid lineage due to a lack of significant erythropoiesis in the leukemic BM. Next, the associations between evaluable specific dysplastic features in each lineage, with individual mutations, pathways, AML genetic ontogeny grouping and subclones were explored. The presence of any (score >0) megakaryocytes with separated nuclear lobes was associated with RAS pathway mutations (q=0.0425) and NPM1 mutations (q=0.022). Frequent megakaryocytes with separated nuclear lobes (score â&#x2030;Ľ3) were associated with cohesin pathway (q=0.010), and STAG2 mutations in particular (q=0.010). No significant associations were detected when considering a score cut off of 2 or over for megakaryocytes 629

O.K. Weinberg et al. A

Figure 2. The distribution of dysplasia scores by mutational pathway, subclone status and individual mutation. (A) Erythoid lineage dysplasia scores with mutation pathways and individual mutations. (B) Myeloid dysplasia scores. (C) Megakaryocyte dysplasia scores. (D) Dysplasia scores based on Lindsley et al. molecular ontogeny group.14

with separated lobes. A lack (0 score) of erythroid nuclear irregularities was associated with epigenetic regulator mutations (q=0.035).

Mutation association with patientsâ&#x20AC;&#x2122; characteristics and other morphology The age distribution differed by AML ontogeny groupings (q=0.002), where the median ages in the AML genetic ontogeny groups were 47.1 years [Interquartile Range (IQR 34.0 - 64.2)] for patients with pan-AML mutations only, 64.8 years (IQR 55.5-70.2) for patients with secondary AML type mutations, and 60.5 years (IQR) 51.5-67.3 for patients with de novo AML type mutations. Older age was also associated with spliceosome pathway mutations (q<0.0001), specifically SRSF2 mutation (q<0.0001). Patients with secondary-type AML mutations had the lowest median WBC (3.4x109/L; IQR 2.0-35.1) followed by pan-AML (16.2x109/L; IQR 3.4-43.7) and de novo AML had the highest median WBC (60.5x109/L; IQR 51.5-67.3, q=0.003). At presentation, a lower median WBC was found in AML with epigenetic regulator mutations (q=0.003), spliceosome mutations (q=0.015), ASXL1 (q=0.015) and BCOR (q=0.0029) mutations. A higher median WBC was found in AML with RAS pathway mutations (q<0.0001) and AML with NPM1 mutations (q=0.004). Evaluation of BM showed that higher BM cellularity was associated with RAS pathway mutations (q=0.029), 630

specifically FLT3-ITD (q=0.029), and NPM1 mutations (q=0.029). No significant differences in blast percentages in blood or BM were found in any of the mutations or mutation groupings.

Influence of mutations, including subclone analysis, and morphological findings on event-free survival An analysis evaluating the association between EFS adjusted for receipt of allogeneic stem cell transplant (alloSCT) and individual mutations, mutation pathways, presence or absence of subclones, and morphological dysplastic findings was performed in the 153 patients who were treated with standard induction chemotherapy. In this group, 75 patients (49%) received allo-SCT in first CR and 26 (17%) additional patients received allo-SCT after disease relapse. In transplant-adjusted univariate analysis, NPM1 mutation was associated with longer EFS (nominal P=0.042), while RUNX1 (P=0.042) and NF1 (P=0.04) mutations were associated with shorter EFS. Patients with secondary AML type mutations had shorter EFS (P=0.011) compared to patients with de novo AML type mutations (P=0.011). The presence of a subclone (see Figure 3) was also associated with shorter EFS (P=0.002). Among morphological features, the presence of frequent micromegakaryocytes (score â&#x2030;Ľ3) was associated with shorter EFS (P=0.018). Regarding the effect of the WHO Classification-defined AML-MRC category, there was no significant difference in haematologica | 2018; 103(4)

Correlation of mutations with morphological findings in AML

Table 2. Final transplant-adjusted multivariable model including mutational and morphological features

Stem cell transplant Micromegakaryocytes (score â&#x2030;Ľ3) Subclone present NPM1 mutation NF1 mutation

Coefficient

Exp ( coefficient) Hazard Ratio

SE coefficient

z value

-0.348 0.764 0.799 -0.696 0.756

0.706 2.146 2.223 0.499 2.13

0.271 0.319 0.248 0.248 0.447

-1.283 2.391 3.22 -2.81 1.69

0.199 0.017 0.001 0.005 0.091

EXP: expansion.

EFS between AML-MRC versus all the other AML cases (P=0.941), nor was there any significant difference in EFS between AML-MRC versus AML-RUNX1 (P=0.253), AML-MRC versus combined AML-NPM1/AML-CEBPA (P=0.407), and AML-MRC versus AML-NOS (P=0.704). The final multivariable model indicated that micromegakaryocytes (score â&#x2030;Ľ3) and presence of a subclone were associated with shorter EFS while NPM1 mutation was associated with longer EFS; mutation in the RAS pathway gene NF1 was marginally associated with shorter EFS (Table 2).

Discussion To our knowledge, this is the first study to explore the relationship of specific dysplastic findings with mutation patterns in de novo AML. We found that among the broad spectrum of dysplastic morphologies evaluated in our study, only megakaryocyte morphology showed significant associations with mutation pattern or outcome. Interestingly, we found that one specific morphological feature, megakaryocytes with separated lobes, was correlated with cohesin pathway and NPM1 mutations but did not impact outcome, while another morphological feature, micromegakaryocytes, was associated with poor outcome, but did not correlate with any mutations. The biological significance of morphological dysplasia in de novo AML is controversial, despite the fact that multilineage dysplasia (at least 50% dysplastic cells in at least 2 hematopoietic lineages) defines a subset of cases of AML with myelodysplasia changes that lack a defining cytogenetic abnormality or antecedent MDS.4 In our cohort of de novo AML cases, we found that cohesin pathway (and specifically STAG2) mutations were associated with greater overall megakaryocytic dysplasia and the specific finding of megakaryocytes with separated nuclear lobes. Mutations in cohesin pathway genes were present in 11% of all patients in our cohort; a higher rate than that reported by Thol et al. (5.9%) or in The Cancer Genome Atlas (2.5-3.5%). However, this rate might be influenced by the fact that our study included only AML cases without a history of a prior myeloid neoplasm and without MDS-associated cytogenetic abnormalities.2,17 Interestingly, Thol et al. and The Cancer Genome Atlas research network found a strong association between NPM1 mutations and mutations in cohesin genes,2,17 and in our study 6 (31%) of 19 cases with cohesin mutations also had NPM1 mutations, and the latter were also associated with frequent megakaryocytes with separated nuclear lobes. RAS pathhaematologica | 2018; 103(4)

way mutations were frequently seen in our cohort, being present in 58% of the cases and, similar to cohesin pathway and NPM1 mutations, were associated with megakaryocytes with separated nuclear lobes. The underlying reasons for the association of cohesin pathway mutations with dysmegakaryopoiesis, and specifically separated megakaryocyte nuclei, are uncertain. Cohesin knockdown mice display a skewing in their stem cell compartment in the BM including myeloid hyperplasia, decrease in erythroid and megakaryocytic progenitors, and increase in nuclear size.21 Studies have also found that ASXL1 loss leads to MDS-like disease in mice and increased frequencies of dysplastic myeloid cells in BM.22 Li et al. suggested that ASXL1 binds to the cohesin complex and plays an essential role in maintaining normal chromatin separation during cell division, suggesting an overlapping molecular mechanism that underlies the pathogenesis of the myeloid disorders.22 Devillier et al. evaluated patients with AML-MRC and found that ASXL1 mutations were associated with a higher amount of dysgranulopoiesis, but not dyserythropoiesis or dysmegakaryopoiesis.13 Cho et al. found that spliceosome mutations were significantly found in AML-MRC, especially in cases with a preceding MDS or dysplasia.23 We did not find any associations between ASXL1 mutations and dysplasia, which could reflect our exclusion of cases with history of MDS or MDS-type cytogenetic abnormalities. In multivariable analysis assessing the effects of individual mutations on outcome in our patient cohort, only NPM1 mutation was significantly associated with EFS. This finding validates the revision in the 2016 WHO Classification in which an NPM1 mutation supercedes the presence of multilineage dysplasia; cases with NPM1 mutation and multlineage dysplasia have been removed from the category of AML-MRC and are now retained in the prognostically favorable category of AML with mutated NPM1.4 We also found that the presence of a subclone (using criteria from Mossner et al.19) adversely affected outcome in the multivariable analysis. A subclone was most commonly seen in cases with DNA methylation (73 cases, 79%), RAS pathway (63 cases, 68%) and NPM1 (44 cases, 48%) mutations. Interestingly, although a subclone was associated with poor outcome in a multivariable analysis, it was not significantly associated with any specific mutation pattern, clinical features, or morphological features. Eisfield et al. investigated chronology of mutations during clonal evolution stratified by functional groups and showed that mutations in tumor suppressor genes, the cohesin complex or the spliceosome are commonly first 631

O.K. Weinberg et al.

Figure 3. Presence of mutations in de novo acute myeloid leukemia (AML). Analyzed by presence of a subclone.

mutational events.3 In the Papaemmanuil et al. study, mutations in DNMT3A, ASXL1, IDH1/2 and TET2 appeared to be acquired the earliest and were almost never found in isolation, while NPM1 mutations were usually secondary events.24 The only morphological features we identified to be associated with outcome was the presence of frequent (score ≥3) micromegakaryocytes; a finding that was independent of any mutations or mutation pattern. In MDS, Della Porta et al. also found that severity of megakaryocytic dysplasia correlated with out-

References 1. Patel JP, Gönen M, Figueroa ME, et al. Prognostic relevance of integrated genetic profiling in acute myeloid leukemia. N Engl J Med. 2012;366(12):1079-1089. 2. Cancer Genome Atlas Research Network, Ley TJ, Miller C, Ding L, et al. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med. 2013;368(22):2059-2074. 3. Eisfeld AK, Mrózek K, Kohlschmidt J, et al. The mutational oncoprint of recurrent cytogenetic abnormalities in adult patients with de novo acute myeloid leukemia. Leukemia. 2017;31(10):2211-2218. 4. Arber DA, Orazi A, Hasserjian R, et al. The 2016 revision to the World Health Organization classification of myeloid neoplasms and acute leukemia. Blood. 2016;127(20):2391-2405. 5. O'Donnell MR, Tallman MS, Abboud CN, et al. Acute Myeloid Leukemia, Version 3.2017, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw. 2017;15(7):926-957. 6. Dohner H, Estey EH, Amadori S, et al. Diagnosis and management of acute myeloid leukemia in adults: Recommendations from an international expert panel, on behalf of the European

632

10.

come.12 Feng et al. found that micromegakaryocytes were independent prognostic factors and improved predictive accuracy of the International Prognostic Scoring Systemrevised (IPSS-R) in their study of 422 MDS patients.25 Our results suggest that micromegakaryocytes in de novo AML could represent a better marker for aggressive disease than the traditional WHO definition of multilineage dysplasia that takes into account dyserythropoiesis and dysgranulopoiesis, which did not affect outcome in our study. Our findings suggest that some cases currently diagnosed as AML-MRC on the basis of dyserythropoiesis and dysgranulopoiesis may not be truly biologically secondary disease and may not be appropriately classified with other AMLMRC cases defined by cytogenetics or history of MDS. Conversely, dysmegakaryopoiesis showed significant associations with specific mutations (when manifesting as forms with separated nuclear lobes) as well as shortened survival (when manifesting as micromegakaryocytes). Our findings highlight the continued relevance of evaluating the background maturing hematopoiesis, specifically megakaryocytes, in de novo AML, even in the current era of detailed mutational analysis. The independent impact of micromegakaryocytes on outcome may reflect factors influencing megakaryocyte morphology that are not directly related to the leukemia mutation pattern, such as epigenetics, the BM microenvironment, or individual patients' characteristics. The finding that a leukemia subclone adversely and independently impacted outcome implies that, not only the number and types of mutations, but also the clonal leukemia architecture should be taken into account in future AML risk stratification schemes. Acknowledgments The authors would like to thank Susan Wong for her editorial assistance with this work.

LeukemiaNet. Blood. 2010;115(3):453474. Miesner M, Haferlach C, Bacher U, et al. Multilineage dysplasia (MLD) in acute myeloid leukemia (AML) correlates with MDS-related cytogenetic abnormalities and a prior history of MDS or MDS/MPN but has no independent prognostic relevance: a comparison of 408 cases classified as "AML not otherwise specified" (AMLNOS) or "AML with myelodysplasia-related changes" (AML-MRC). Blood. 2010;116(15):2742-2751 Falini B1, Macijewski K, Weiss T, Bacher U, Schnittger S, Kern W, et al. Multilineage dysplasia has no impact on biologic, clinicopathologic, and prognostic features of AML with mutated nucleophosmin (NPM1). Blood. 2010;115(18):3776-3786. Díaz-Beyá M, Rozman M, Pratcorona M, et al. The prognostic value of multilineage dysplasia in de novo acute myeloid leukemia patients with intermediate-risk cytogenetics is dependent on NPM1 mutational status. Blood. 2010;116(26):61476148. Rozman M1, Navarro JT, Arenillas L, et al Multilineage dysplasia is associated with a poorer prognosis in patients with de novo acute myeloid leukemia with intermediaterisk cytogenetics and wild-type NPM1.

Ann Hematol. 2014;93(10):1695-1703. 11. Weinberg OK, Pozdnyakova O, Campigotto F, et al. Reproducibility and prognostic significance of morphologic dysplasia in de novo acute myeloid leukemia. Mod Pathol. 2015;28(7):965976. 12. Della Porta MG, Travaglino E, Boveri E, et al. Rete Ematologica Lombarda (REL) Clinical Network. Minimal morphological criteria for defining bone marrow dysplasia: a basis for clinical implementation of WHO classification of myelodysplastic syndromes. Leukemia. 2015;29(1):66-75. 13. Devillier R, Mansat-De Mas V, Gelsi-Boyer V, et al. Role of ASXL1 and TP53 mutations in the molecular classification and prognosis of acute myeloid leukemias with myelodysplasia-related changes. Oncotarget. 2015;6(10):8388-8396. 14. Lindsley RC, Mar BG, Mazzola E, et al. Acute myeloid leukemia ontogeny is defined by distinct somatic mutations. Blood. 2015;125(9):1367-1376. 15. Alvarez Argote J, Dasanu CA. ASXL1 mutations in myeloid neoplasms: pathogenetic considerations, impact on clinical outcomes and survival. Curr Med Res Opin. 2017:1-7. 16. Gibson CJ, Lindsley RC, Tchekmedyian V, et al. Clonal Hematopoiesis Associated

haematologica | 2018; 103(4)

Correlation of mutations with morphological findings in AML With Adverse Outcomes After Autologous Stem-Cell Transplantation for Lymphoma. J Clin Oncol. 2017;35(14):1598-1605. 17. Weinberg OK, Gibson CJ, Blonquist TM, et al. NPM1 mutation but not RUNX1 mutation or multilineage dysplasia defines a prognostic subgroup within de novo acute myeloid leukemia lacking recurrent cytogenetic abnormalities in the revised 2016 WHO classification. Am J Hematol. 2017;92(7):E123-E124. 18. Kluk MJ, Lindsley RC, Aster JC, et al. Validation and Implementation of a Custom Next-Generation Sequencing Clinical Assay for Hematologic Malignancies. J Mol Diagn. 2016;18(4):507515.

haematologica | 2018; 103(4)

19. Mossner M, Jann JC, Wittig J, et al. Mutational hierarchies in myelodysplastic syndromes dynamically adapt and evolve upon therapy response and failure. Blood. 2016;128(9):1246-1259. 20. Thol F, Bollin R, Gehlhaar M, et al. Mutations in the cohesin complex in acute myeloid leukemia: clinical and prognostic implications. Blood. 2014;123(6):914-920. 21. Mullenders J, Aranda-Orgilles B, Lhoumaud P, et al. Cohesin loss alters adult hematopoietic stem cell homeostasis, leading to myeloproliferative neoplasms. J Exp Med. 2015;212(11):1833-1850. 22. Li Z, Zhang P, Yan A, et al. ASXL1 interacts with the cohesin complex to maintain chromatid separation and gene expression

for normal hematopoiesis. Sci Adv. 2017;3(1):e1601602. 23. Cho YU, Jang S, Seo EJ, et al. Preferential occurrence of spliceosome mutations in acute myeloid leukemia with preceding myelodysplastic syndrome and/or myelodysplasia morphology. Leuk Lymphoma. 2015;56(8):2301-2308. 24. 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. 25. Feng G, Gale RP, Cui W, et al. A systematic classification of megakaryocytic dysplasia and its impact on prognosis for patients with myelodysplastic syndromes. Exp Hematol Oncol 2016;5:12.

633

ARTICLE

Acute Lymphoblastic Leukemia

Ferrata Storti Foundation

Haematologica 2018 Volume 103(4):634-644

Dynamic clonal progression in xenografts of acute lymphoblastic leukemia with intrachromosomal amplification of chromosome 21

Paul. B. Sinclair,1 Helen H. Blair,1 Sarra L. Ryan,1 Lars Buechler,1 Joanna Cheng,1 Jake Clayton,1 Rebecca Hanna,1 Shaun Hollern,1 Zoe Hawking,1 Matthew Bashton,1 Claire J. Schwab,1 Lisa Jones,1 Lisa J. Russell,1 Helen Marr,1 Peter Carey,2 Christina Halsey,3 Olaf Heidenreich,1 Anthony V. Moorman1 and Christine J. Harrison1 Wolfson Childhood Cancer Research Centre, Northern Institute for Cancer Research, Newcastle University, Newcastle-upon-Tyne; 2Department of Clinical Haematology, Royal Victoria Infirmary, Newcastle-upon-Tyne and 3Wolfson Wohl Cancer Research Centre, Institute of Cancer Sciences, University of Glasgow, UK 1

ABSTRACT

Correspondence: christine.harrison@newcastle.ac.uk

Received: May 15, 2017. Accepted: February 8, 2018. Pre-published: February 15, 2018.

doi:10.3324/haematol.2017.172304 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/634 ÂŠ2018 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.

634

ntrachromosomal amplification of chromosome 21 is a heterogeneous chromosomal rearrangement occurring in 2% of cases of childhood precursor B-cell acute lymphoblastic leukemia. These abnormalities are too complex to engineer faithfully in animal models and are unrepresented in leukemia cell lines. As a resource for future functional and preclinical studies, we have created xenografts from the leukemic blasts of patients with intrachromosomal amplification of chromosome 21 and characterized them by in-vivo and ex-vivo luminescent imaging, flow immunophenotyping, and histological and ultrastructural analyses of bone marrow and the central nervous system. Investigation of up to three generations of xenografts revealed phenotypic evolution, branching genomic architecture and, compared with other B-cell acute lymphoblastic leukemia genetic subtypes, greater clonal diversity of leukemia-initiating cells. In support of intrachromosomal amplification of chromosome 21 as a primary genetic abnormality, it was always retained through generations of xenografts, although we also observed the first example of structural evolution of this rearrangement. Clonal segregation in xenografts revealed convergent evolution of different secondary genomic abnormalities implicating several known tumor suppressor genes and a region, containing the B-cell adaptor, PIK3AP1, and nuclear receptor corepressor, LCOR, in the progression of B-cell acute lymphoblastic leukemia. Tracking of mutations in patients and derived xenografts provided evidence for co-operation between abnormalities activating the RAS pathway in B-cell acute lymphoblastic leukemia and for their aggressive clonal expansion in the xeno-environment. Bi-allelic loss of the CDKN2A/B locus was recurrently maintained or emergent in xenografts and also strongly selected as RNA sequencing demonstrated a complete absence of reads for genes associated with the deletions.

Introduction Xenograft models of leukemia have been used to address a number of important fundamental and translational research questions relating to: the nature of leukemia stem cells, clonal evolution and experimental therapies.1-8 Limiting dilution studies have demonstrated that leukemia-initiating cells are common, while fluorescence in situ hybridization (FISH), genomic arrays, analysis of immunoglobulin/T-cell receptor rearrangements, immunophenotype and drug response have suggested that the haematologica | 2018; 103(4)

iAMP21 xenografts

disease can be propagated through multiple generations of mice with high fidelity. Intrachromosomal amplification of chromosome 21 (iAMP21) is an intriguing cytogenetic abnormality, defining a specific subgroup of approximately 2% of cases of childhood precursor B-cell acute lymphoblastic leukemia (B-ALL). Chromosome 21 genomic profiles, although highly variable, always involve amplifications, flanked by regions of normal copy number or deletion.9,10 We have shown that the oncogenic potential of chromosome 21 is optimized through a combination of catastrophic sequence reorganization, driven by chromothripsis, deletion and amplification, resulting from dicentric chromosome formation, breakage-fusion-bridge cycles and whole chromosome arm duplications.11 This mechanism has the potential to produce a near infinite number of alternative derivative chromosomes 21. The structure of the iAMP21 chromosome is stabilized by telomere acquisition or duplication, while a combination of protected amplified genes are postulated to confer an overall growth advantage, leading to the development of ALL. Several lines of evidence indicate that iAMP21 is a stable, primary genetic change: (i) among 530 patients, iAMP21 was reported as a sub-clonal abnormality in only a single case;12 (ii) the iAMP21 chromosome morphology remains consistent between cells in the same patient; and (iii) the same chromosome structure is observed at diagnosis and relapse.9 A range of specific secondary genetic abnormalities: CRLF2 activating rearrangements, X chromosome gain, deletions of RB1, ETV6, the long arm of chromosome 7 (7q) and 11q, and mutations of the RAS pathway frequently co-occur with iAMP21.9,12,13 This distinct iAMP21-ALL subgroup is clinically defined by older age (median 9 years), low white blood cell counts and a high risk of relapse on standard therapy.14-16 Intensive therapy significantly reduces the risk of relapse17,18 but associated morbidity highlights an urgent need for less toxic regimens. Development of rational targeted therapies requires understanding of the mechanism by which these rearrangements initiate leukemia. However the requisite tools for functional studies are lacking because no iAMP21-ALL cell lines exist and the complex nature of the abnormalities exclude their reproduction in engineered animal models. To address this shortfall, we transplanted primary leukemia cells from five iAMP21-ALL patients into NOD/LtSz-scid IL2Rg null (NSG) mice. In-vivo luminescent imaging, to track the physical development of ALL, was used to assess these xenografts as potential models for functional and pre-clinical studies. In addition, we characterized the disease morphology at the microscopic and ultrastructural levels and, as a first application, have performed extensive genomic analysis to investigate clonal heterogeneity of iAMP21-ALL, from which some intriguing findings have emerged.

Xenografts and isolation of leukemia cells Primografts were created by intrafemoral injection of patients’ cells into NSG mice, as previously described.1 Between 2x104 and 2x106 primograft bone marrow or spleen cells from each mouse were used in the same procedure to create secondary and tertiary xenografts (Online Supplementary Table S4). Xenografts were culled at end stage-disease as defined in the Online Supplementary Methods. Bone marrow cells flushed from femora and disaggregated spleens were passed through 40 µm cell strainers. Leukemic cells used for all experimental work were purified from spleen preparations by separation over Ficoll-Paque [density 1.077 g/mL] (G.E. Health Care, Buckinghamshire, UK). Proportions of human and mouse cells and immunophenotypes of the human leukemia fractions were determined by flow cytometry as described in the Online Supplementary Methods.

Lentiviral transduction, in-vitro culture and in-vitro, in-vivo and ex-vivo imaging of xenograft cells Detailed procedures are provided in the Online Supplementary Methods.

Histopathology and transmission electron microscopy Detailed procedures are provided in the Online Supplementary Methods.

Single nucleotide polymorphism arrays DNA extraction and SNP6.0 array hybridization and analysis were performed, as previously described.13 To define regions of chromosome 21 copy number evolution, single nucleotide polymorphism copy number values were subtracted between secondary xenografts 2°1B and 2°1A. Copy number abnormalities (CNA) in immunoglobulin or T-cell receptor regions, those not involving coding gene regions, present in patients’ remission samples or listed in the Toronto Database of Genomic Variants, are not reported. Genomic positions are those in the Hg19 database.

Fluorescence in situ hybridization and multiplex ligation-dependent probe amplification Dual color FISH on 100-200 interphase cells was performed using fluorescently labeled BAC probes hybridizing to the RUNX1 (RP11-773I18) and APP (RP11-66H5 and RP11-15D13) genomic regions or commercially available probes to the CDKN2A/B genomic region and chromosome 9 centromere (CytoCell, Cambridge, UK), as previously described.20 Multiplex ligation-dependent probe amplification (MLPA) was performed using the SALSA MLPA 335 kit (MRC-Holland, the Netherlands), as previously described.21

Analysis of RAS pathway mutations and RNA sequencing Detailed procedures are provided in the Online Supplementary Methods.

Results Methods Patients

Development of xenografts and characterization by in-vivo and ex-vivo imaging

Viable cells from children diagnosed with iAMP21-ALL, as previously defined,19 were provided by the Bloodwise Childhood Leukaemia Cell Bank. Ethical approval was obtained for all patients and informed consent was granted in accordance with the Declaration of Helsinki. Karyotypes and demographic details of the patients used to generate xenografts or as controls for histological analysis are presented in Online Supplementary Tables S1 - S3.

Of six primary and two relapsed cases of iAMP21-ALL transplanted into femora of NSG mice, five, including one relapsed case, developed ALL derived from the human cells, in one or more animals (Online Supplementary Table S1). The mean time to development of end-stage disease in primografts was 30 weeks and splenomegaly was seen in all engrafted animals. Secondary and tertiary xenografts were established in three and one cases, respectively, and

haematologica | 2018; 103(4)

635

P.B. Sinclair et al.

all were assigned unique identifiers indicating passage number and patient of origin, for example 2Â°3e was one of several secondary xenografts derived from patient 3. Xenograft leukemia cells constituted between 40-92% of bone marrow and 23-53% and 79-99% of crude and purified spleen samples, respectively (Online Supplementary Table S4). Essentially all human cells isolated from xenografts expressed the B-cell markers CD19 and CD10 but analysis of CD34 and CD38 demonstrated considerable phenotypic divergence between mice (Online Supplementary Table S5 and Online Supplementary Figure S1). To investigate their potential for use in in-vivo and invitro functional studies, we transduced xenograft stocks from four iAMP21-ALL patients with the pSLIEW lentivirus vector that expresses luciferase and enhanced green fluorescent protein (EGFP).22 Three days after transduction, a total of 3x106 cells from each patient were transplanted by intra-femoral injection into two NSG

mice each, here identified by the patientsâ&#x20AC;&#x2122; number followed by aSLIEW or bSLIEW. Less than 1% of transduced cells were EGFP-positive by fluorescence-activated cell sorting analysis at this time point (Online Supplementary Figure S2) or by fluorescence microscopy after 1 week of culture on MS-5 feeder layers (data not shown); nevertheless, by 2-4 weeks following transplantation, luminescent signals, clearly localized to the injected femora, were seen on whole body imaging of all mice. Leukemia spread to other bones and organs with noticeable variation in the strength of signal at some sites (Online Supplementary Figure S3). This variation was highlighted by measurement of luminescent signals from organs post-mortem and by analysis of the relationship between signal development at different sites over time (Figure 1A,B and Table 1). Signal variation in the spleen was shown to relate to the proportion of infiltrating blast cells that expressed EGFP rather than to overall tumor load (Figure 1C, Table

636

Figure 1. In xivo and ex-vivo imaging of xenografts. (A) Serial quantification of the luminescent signal from injected and contralateral femora and the whole body for each xenograft suggesting variations in the rate at which cells migrate from the site of injection and the degree to which different sites are infiltrated by transduced cells. For example, in 3a/bSLIEW the strength of signal from the contralateral femur lagged substantially behind that of the injected femur until weeks 10-13, while in 1a/bSLIEW and 2bSLIEW the two femora showed similar levels from week 2. (B) Example of ex-vivo imaging of dissected organs showing total luminescent readings for spleen, liver and kidney. Images are representative of six animals analyzed. (C) Examples of FACS analysis of cells isolated from the spleens of 3aSLIEW and 4aSLIEW, demonstrate marked contrast in the proportion of CD19-positive cells that are EGFP-positive. (D) A single example of serial threedimensional reconstructions of luminescent signals in 2aSLIEW, 11 and 15 weeks after transplant. Arrows point to regions of the skull, spleen and a third site showing strong signal increase between these two time points. The skeleton projected for orientation is not derived from this animal.

haematologica | 2018; 103(4)

iAMP21 xenografts

1 and Online Supplementary Table S4). Serial three-dimensional reconstructions of one xenograft showed dramatic increases in signals from the spleen and head between weeks 11-15 (Figure 1D).

Morphology and ultrastructure of bone marrow and central nervous system reveals patient-specific heterogeneity, including evidence for systemic bone marrow niche destruction To investigate bone marrow morphology and central nervous system (CNS) involvement in the iAMP21-ALL models, we examined sections through tibiae, sterna and heads of the mice engrafted with SLIEW-transduced cells. Control NSG bone marrow stained with hematoxylin and eosin resembled that of a wild-type mouse23 and was negative for anti-human CD19, CD45 and Ki-67 staining (Figure 2A and Online Supplementary Figure S4). In xenografts, two types of morphology were seen, both of which differed from controls; type A, which closely resembled that of iAMP21-ALL patientsâ&#x20AC;&#x2122; trephines (Figure 2B,G, Online Supplementary Figures S4 and S5 and Online Supplementary Table S2) and type B, which although abnormal, clearly differed from the patientsâ&#x20AC;&#x2122; trephines (Figure 2C,G and Online Supplementary Figure S4). Cells in type A but not type B sections were actively cycling and of human origin as indicated by staining with anti-human Ki67, CD19 and CD45 antibodies (Figure 2D and Online Supplementary Figure S4). Individual sections presented with either type A or B morphology only with the exception of one sternal segment in which both types co-existed (Figure 2D). There was relatively sharp demarcation between the A and B type areas, suggesting that the iAMP21 ALL cells were organized into massive clumps that did not diffuse easily within the lumen. We performed transmission electron microscopy of decalcified sections of different tibiae or forelimb bones from each mouse. Xenograft ultrastructure always differed from that of wild-type controls and, as with bright field microscopy, two distinct categories could be identified (Figure 2E and Online Supplementary Figure S6). The first was termed viable leukemia (VL), as mitotic figures were present and cells appeared normal, although homogeneous by comparison with controls. There were more connections and less extracellular space between cells in preparations from 4a/bSLIEW compared with controls and 1a/bSLIEW.

The second category, equivalent to histological type B, was termed apoptotic (AP), as no mitotic figures were present and cells were depleted in number with classical signs of apoptosis, in the form of condensed chromatin localized to the periphery of the nucleus.24 We also examined sections through the skulls and CNS of seven of the eight xenografts (Figure 2F and Online Supplementary Figure S7), revealing calvaria in all cases, packed with homogeneously stained cells resembling the A type morphology of tibial sections. CNS involvement ranged from small foci of leukemia cells to heavy meningeal infiltration, extending into the choroid plexus in one case. Comparison of CNS histological grades for each xenograft with bone marrow histopathology and transmission electron microscopy data (Table 2) showed heavy CNS involvement only in 1a/bSLIEW and 4a/bSLIEW, correlating with tibia marrow histological type A and transmission electron microscopy type VL. We infer the proportion of transduced cells infiltrating the CNS varied between mice because luminescent signals from the head failed to correlate with histological grade (Table 2). To investigate the relative incidence of the morphological types we examined bone marrow sections, stained with hematoxylin and eosin and anti-CD19, from 13 additional xenografts derived from seven B-ALL patients (Online Supplementary Tables S3 and S6). Among these cases, type B morphology was seen in two primografts, one derived from a relapsed iAMP21-ALL patient and the second from a case with high hyperdiploid ALL; areas of A type morphology were also seen in both (Online Supplementary Figure S8). Other xenografts displayed A type morphology either exclusively (Online Supplementary Figure S9) or infiltrating apparently normal mouse bone marrow (Online Supplementary Figure S10). Interestingly these latter cases supported our initial observation that the ALL cells may grow in clumps because CD19-positive cells formed distinct clusters.

Segregation of copy number abnormalities in xenografts implicates known and novel genes in the progression of acute lymphoblastic leukemia with intrachromosomal amplification of chromosome 21 We used SNP6.0 array profiles to evaluate the genomic stability of iAMP21-ALL in 21 xenografts from five patients. Presentation and remission samples were avail-

Table 1. In-vivo and ex-vivo luminescent imaging data and spleen weights for xenografts transplanted with iAMP21-ALL cells transduced with pSLIEW.

Xenograft

4SLIEWa 4SLIEWb 3SLIEWa 3SLIEWb 1SLIEWa 1SLIEWb 2SLIEWa 2SLIEWb

Whole Body Luminescence (photons/second) Peak whole body Peak injected Ratio (PWB) femur (PIF) PWB/PIF 6.87E+08 2.97E+08 1.64E+09 1.81E+09 3.79E+08 1.57E+08 3.30E+09 3.65E+09

2.53E+08 5.09E+07 1.90E+08 4.00E+08 2.44E+07 1.43E+07 1.06E+08 3.64E+08

2.72 5.84 8.63 4.53 15.53 10.83 31.13 10.03

Dissected Organ Luminescence (photons/second) Kidney Spleen Liver (mean) ND ND 5.69E+08 3.85E+08 1.12E+07 2.60E+07 1.04E+09 3.10E+09

ND ND 3.77E+07 1.99E+07 4.10E+06 4.27E+06 9.84E+07 1.83E+08

ND ND 1.49E+05 3.02E+05 2.50E+05 1.84E+05 3.67E+05 6.65E+05

Spleen weight/ % GFP +ve blasts* Spleen radiance/g ND ND 2.47E+09 1.43E+09 2.11E+07 2.95E+07 2.60E+09 5.64E+09

0.83g / 0.2 0.81g / 0.0 0.23g / 44 0.27g / 68% 0.53g / 1.6% 0.88g / 0.8% 0.40g / 24% 0.55g / 26%

*blasts are CD19-positive cells isolated from spleen.

haematologica | 2018; 103(4)

637

P.B. Sinclair et al.

able for all except patient 4. A core of three to 16 concordant CNA, involving coding gene regions, were clonal at presentation and retained in all xenografts (Online Supplementary Table S7). The existence of competing subclones and branching genomic evolution was demonstrated by discordant CNA, which occurred at a rate of

between four and 12 (Figure 3 and Online Supplementary Table S7). Clonal trisomies or copy number neutral loss of heterozygosity, present in each patients’ sample, were lost after transplantation, while deletions and amplifications were typically sub-clonal and either lost or increased in level or newly emergent as sub-clones in xenografts.

Figure 2. Histological sections of bone marrow from controls and xenografts transplanted with SLIEW transduced cells. (A) Control NSG mouse: hematoxylin and eosin (H&E) stained femur showing heterogeneous cell types, abundant megakaryocytes and vascular structures. (B) H&E stained femur from 4aSLIEW showing tightly packed homogeneously stained cells and an absence of megakaryocytes and vascular structures (image representative of six animals showing only morphology type A). (C) H&E stained femur from 3bSLIEW showing heterogeneous cell types but in comparison with controls, loss of cellularity and organization, absence of vascular structures, reduced numbers of megakaryocytes and presence of small darkly stained cells or cellular fragments (image representative of two animals showing morphology type B). (D) The only example seen of co-existence of A and B type morphology in a sternal segment of 3bSLIEW: first left panel; H&E stained whole section with box marking the region shown in the second left panel. Second left panel; H&E stained detail of a single sternum segment displaying both A and B type morphologies. Middle panel; whole sections stained with anti-human CD19 and Ki-67 antibodies, arrowheads indicate three regions corresponding to high resolution images in the right hand panels. Right hand panels; high resolution images of anti-CD19 and Ki-67 staining from regions 1, 2 and 3. Anti-CD19 and Ki67 stained human leukemia cells remain tightly packed (region 2) with little diffusion to adjacent areas of acellular marrow (region 3). Anti-Ki67 and CD19 staining demonstrating that the proportion of cycling human cells are reduced in region 2 compared with region 1 suggesting a microenvironment less favorable for leukemia cell growth. (E) Examples of TEM images of xenograft tibia sections. In 4aSLIEW and 1aSLIEW cells appear homogeneous compared with controls (Online Supplementary Figure S6) and have a high nuclear to cytoplasmic ratio (VL morphology). 3aSLIEW and 2aSLIEW displaying evidence of cell death and characteristics of apoptosis, such as chromatin clumping and nuclear fragmentation (AP morphology). Images are representative of four animals each showing VL and AP morphology. (F) Examples of skull and brain sections from 3bSLIEW (left panel) and 1bSLIEW (right panel) showing heavy infiltration of the calvaria in both cases (arrow 1), light (3bSLIEW) and heavy (1bSLIEW ) infiltration of the meninges, respectively (arrow 2). Images are representative of seven animals. (G) Two examples of H&E stained patients’ trephines. The left hand panel is from patient 1 for comparison with the leukemia cells from the same patient in a mouse (1aSLIEW) tibia section shown in (B). Images are representative of seven iAMP21-ALL patients’ trephines analyzed. Scale bars are; (A, B, C) two right hand panels of (D and G) - 50 mm, far left hand and two middle panels of (D) – 1 mm, second from left panel of (D and F) 200 mm and (E) - 10 mm.

638

haematologica | 2018; 103(4)

iAMP21 xenografts

Table 2. Summary of histological and ultrastructural (EM) data for bone marrow and CNS of mice transplanted with iAMP21-ALL cells.

Xenograft

4aSLIEW 4bSLIEW 3aSLIEW 3bSLIEW 1aSLIEW 1bSLIEW 2aSLIEW 2bSLIEW

H&E

CD19

Tibia/sternum CD45

Ki67

H&E

Calvaria CD19

(grade)

CNS Peak head luminescence*

A A A A/B A A A B

+ve +ve +ve +/-ve +ve +ve +ve -ve

+ve +ve -ve -ve +ve +ve -ve -ve

+ve +ve +ve +/-ve +ve +ve +ve -ve

VL VL AP AP VL VL AP AP

N/A A A A A A A A

+ve +ve +ve +ve +ve +ve +ve +ve

N/A 5 1-2 1-2 3-4 + CP 4 1 2

8.3x107 3.4x107 3.4x108 2.3x108 3.7x107 6.4x106 3.7x108 6.3x108

CNS: central nervous system; H&E: hematoxylin and eosin; EM: electron microscopy. *Units of luminescence are total flux (photons / second).

Exceptionally in xenografts of patient 1, a sharp transition in genomic architecture, involving clonal gain of CNA of three chromosomes, occurred. Chromosome 21 profiles usually remained unchanged across samples (Online Supplementary Figure S11), but interestingly in cells from patient 1, we observed structural evolution of the iAMP21 chromosome, involving a small region of copy number gain and nine distinct regions of loss of one or two copies (Figure 3A and Online Supplementary Table S8). Importantly the additional deletions did not affect two regions predicted to contain critical oncogenes11 but did re-define the proximal boundary of the region of highest level of amplification9 from 21:32,813,553-37 to 21:33,949,423. By FISH, we confirmed that the RUNX1 and APP gene regions were maintained at the same level of amplification and reduced in copy number from three to one, respectively. Additional rearrangements included bi-allelic deletion of the short arm of chromosome 9 (9p), resulting in homozygous loss of CDKN2A/B, as confirmed by MLPA (Online Supplementary Table S9), and mono-allelic deletion of 3p, involving the CMTM genes 6-8. SNP6.0 array and FISH provided no evidence of these CNA prior to their emergence in 2°1a. However, as previously reported, two reads in whole genome sequencing data were consistent with the presence of a minor clone carrying the chromosome 3 deletion in the patient’s cells at presentation.11 Suggestive of convergent clonal evolution and highlighting the relevance of specific chromosomal regions to disease progression, several were targeted by different abnormalities segregated in xenografts from the same patient. Consistent with an oncogenic role for genes on Xp, patient 2 carried competing sub-clones marked by gain of a whole X chromosome or isochromosome Xp (Figure 3B). Whether emergence of a focal deletion of Xp, involving the zinc finger genes, ZNF157 and ZNF41, was related to the presence of the larger scale CNA remains unclear but they were unlikely to have been driven by CRLF2 overexpression, as genetic analysis ruled out rearrangement of this locus in the patient’s sample.25 In patient 3 large overlapping deletions of 12p, both involving ETV6, were segregated (Figure 3C). Although no patient’s material was available, differences in CNA involving the long arm of chromosome 10 were identified in xenografts from a relapse sample of patient 4 (Figure 3D). Strongly indicative of convergent evolution and hence a role in leukemia progression, the same focal bi-allelic deletion, involving PIK3AP1 (BCAP) and LCOR (C10orf21), was nested within two dishaematologica | 2018; 103(4)

tinctly different, large, mono-allelic 10q deletions. Evidently of independent origin, as it was detected only in a single xenograft, one of the large deletions also harbored a second likely co-operating focal bi-allelic deletion that resulted in loss of BLNK. Lastly, passage of patient 5 cells in a primary xenograft resulted in concomitant loss of copy number neutral loss of heterozygosity of 12q, with progression from sub-clonal to clonal deletion of a region of 12q containing SH2B3 (Figure 3E). Comparison between relapse and the xenograft showed no overlap in progression of specific CNA, although interestingly the EBF1 gene was targeted by different deletions in the two samples.

Mutations affecting the RAS pathway drive clonal expansion To investigate progression of mutations activating the RAS pathway previously identified in patient 113 we performed whole exome sequencing of selected and Sanger sequencing of all derived xenografts (Figure 4A). Interestingly, while an NF1 mutation remained clonal in all samples, two different mutations affecting NRAS and one of KRAS marked a dramatic clonal evolution. Remarkably the KRAS mutation, present as a dominant clone in both primary and one secondary xenograft was undetectable at a read depth of over 6000 in the presentation sample and also undetected in other xenografts which instead carried a dominant NRAS mutation detected in only 1% of reads at presentation. NRAS and FLIT3 mutations detected at presentation in patients 3 and 5, respectively, were clonal in xenografts while the FLIT3 mutation was lost at relapse (Figure 4B,C).

Transcriptional environment associated with deletions Anticipating that xenograft preparations, in contrast to patients’ samples, would be free of non-leukemic human cells, we used them to investigate the transcriptional environment associated with bi-allelic deletions (Table 3). By comparison with non-deleted samples, and confirming clonal dominance of the chromosome 10 deletions, we saw marked reductions in the levels of transcription of PI3KAP1, LCOR and BLNK. Loss of the BLNK genomic region also resulted in silencing of DNTT and OPALIN. Three of the four sequenced xenografts carried bi-allelic 9p deletions affecting CDKN1A/B. Indicating strong pressure for clonal selection of bi-allelic loss, read counts within this region were reduced to zero in deleted cases. Interestingly two focal bi-allelic deletions were associated with silencing 639

P.B. Sinclair et al.

Figure 3. Analysis of evolution of copy number abnormalities in xenograft models. Panels in blue boxes show heat maps of copy number from SNP6.0 arrays for chromosomal regions showing evidence of clonal genomic evolution in xenografts, P (presentation), R (remission), 1°,2° and 3° (primary, secondary and tertiary xenografts). Box-flow diagrams illustrate loss, gain or change in level of genomic abnormalities. White boxes; sample analyzed, gray boxes; sample not analyzed, red type; CN gain of one, purple; CN gain of two, light red; sub-clonal CN gain, black; CN loss of one, blue; CN loss of two, gray; subclonal loss, green CN-LOH. * indicates deletion was detected by whole genome sequencing but not by SNP6.0 array. (A) Patient 1 – including heat maps of 9p demonstrating gain of bi-allelic deletion of a region containing CDKN2A/B and focal mono-allelic deletion of a region of 3p containing the CMTM genes 6-8. Chromosome 21 heat maps illustrate a complex pattern of copy number gain and loss characteristic of iAMP21 and also demonstrate additional CN changes in one 2° and all 3° xenografts. The color-coded bar depicts regions of chromosome loss or gain in the derivative iAMP21, [der(iAMP21)]. Green box; chromosome 21 CN profiles before and after iAMP21 evolution and showing the position of probes used for FISH analysis. Upper black box; FISH image of metaphase and interphase cells showing multiple copies of the RUNX1 region and three copies of the APP region at presentation. Lower black box; FISH images showing three or one copy of the APP region in cells from the patient, a 1° and two 2° xenografts. (B) CNA discordant between patient 2 and xenografts including; loss of whole chromosome (WC) X, gain of an iso(Xp) and gain of a focal deletion of Xp involving the genes ZNF157 and ZNF41. (C) CNA discordant between patient 3 and xenografts including different deletions involving ETV6 one of which potentially resulted in a novel ETV6-BICD1 fusion gene. (D) Differences in copy number profiles between three 1° xenografts from patient 4. These included two different large mono-allelic deletions of chromosome 10 (1 and 2) both of which contained a focal bi-allelic deletion (3), present in all xenografts and involving PIK3AP1 and LCOR. A second bi-allelic deletion (4) was present in two xenografts only and involved BLNK. Other discordant CNA included complex rearrangements of 17q and 14q, focal deletions of 4q34.1, 16p13 and 22q13.1 and a focal amplification of 4q13.2. E. Sub-clonal deletions at 2p25.1, 3q13.3 and 12q24.1 present in patient 5 became dominant clones in the xenograft while a sub-clonal deletion of 3p21.3 was lost and sub-clonal deletions of 1q42.3 and 5q33.3 (marked by red arrow 1) were gained. In the patients’ relapse sample the 3p21.3 deletion became dominant while other sub-clonal abnormalities detected at presentation were lost. As with the xenograft a focal deletion of 5q33.1 (marked by red arrow 2) emerged at relapse. The two 5q deletions both resulted in loss of coding exons of EBF1. Genomic positions of breakpoints derived from SNP6.0 analysis and genes contained within focal CNA, for all patients and xenografts, are annotated in Online Supplementary Table S7.

640

haematologica | 2018; 103(4)

iAMP21 xenografts

Figure 4. Analysis of mutations affecting the RAS pathway. The key to the analytical methods used are shown top right. Blue boxes show non-synonymous RAS pathway mutations identified. White boxes summarize the estimated variant allele (VA) frequencies and methods of analysis used for patients’ and xenograft samples. (A) In patient 1 an NF1 mutation was clonal at presentation and in all xenografts. By contrast mutations of NRAS and KRAS showed marked fluctuations in VA frequency; NRAS (1), present in the patient as a sub-clone, was detected in both 1° xenografts but not in blasts from 2° and 3° animals. NRAS (2), identified by high depth targeted sequencing in 1% of patients’ sample reads, was undetected by whole exome sequencing in primografts but emerged as a dominant clone in 2°2a and all derivative 3° xenografts. A mutation of KRAS, although undetected in more than 6000 reads by targeted sequencing of the patient’s sample, marked the dominant clone present in primografts and 2°2b. Sanger sequence traces illustrate the relationship between the NF1, NRAS (2) and KRAS mutations in the two 2° xenografts, traces shown for 2°1a are also representative of 1°1a and 1°1b, traces shown for 2°1b are also representative of 3°1a-g. (B) In patient 3 an NRAS mutation identified in the patient remained clonal in all 1° and 2° xenografts. The Sanger sequencing trace shown for 1°3a is representative of all xenografts. (C) In patient 5 a FLT3 itd, detected as a minor sub-clone by exome sequencing of the presentation sample, became dominant in the 1° xenograft as demonstrated by the generation of two distinct exon 14 polymerase chain reaction (PCR) amplicons of equal intensity (first lane bottom right). In contrast only a single PCR product was amplified from the relapse sample of this patient (second lane bottom right). Mutations detected in the patients’ presentation samples have been previously published.13

not only of the physically deleted genes but also of TUSC1, positioned more than 3 Mb away from the deletion boundaries.

Discussion Compared with genetically engineered animal models, xenografts bring several advantages to the study of ALL, not least, their potential to fully recapitulate the spectrum of genomic abnormalities that occur within individual patients of a given genetic sub-type. This is particularly relevant to iAMP21-ALL, in which the primary abnormality is structurally complex, unique to each patient and impossible to reproduce in engineered animal models. As there are no cell lines carrying iAMP21, the xenografts presented here represent an important resource for future functional and pre-clinical studies. Highlighting the potential of lentiviral constructs integrated into xenograft cells, we demonstrated their in vivo expression. However we observed considerable temporal and spatial variation in signal development that, as demonhaematologica | 2018; 103(4)

strated by analysis of spleen and CNS, was apparently related to heavy skewing of the ratios of transduced to non-transduced cells at specific sites. It seems likely that this skewing was caused largely by clonal expansion of small founder populations, particularly as tracking of specific genomic abnormalities demonstrated aggressive expansion of minor sub-clones in xenografts. As a consequence, accurate analysis of disease burden by in-vivo imaging, will in future require enrichment for EGFPexpressing cells prior to engraftment. Unexpectedly, light microscopy and transmission electron microscopy together provided strong evidence that transplantation of NSG mice with iAMP21-ALL cells from two patients led to destruction of the bone marrow niche. As we saw no examples of similar morphology among iAMP21-ALL patients’ samples, it seemed likely that this phenomenon was xenograft-specific and a consequence of initiating a heavy leukemic burden at one site. Cells populating the affected areas, although damaged, were heterogeneous and showed little if any staining with human CD19 and CD45 suggesting they were of host origin. Histology was therefore consistent with destruction hav641

P.B. Sinclair et al. Table 3. The expression of genes in xenograft-derived iAMP21-ALL cells within and neighboring regions of bi-allelic deletion of chromosomes 10 and 9.

Chromosome 10

Xenograft 3°1e

Gene ZNF518A BLNK DNTT OPALIN TLL2 TM9SF3 PIK3AP1 LCOR SLIT1 ARHGAP19

1°4b

2°2e

2°3d

Genomic position

RCM

96,129,715-96,205,288 96,191,702-96,271,587 96,304,396-96,338,564 96,343,216-96,359,365 96,364,606-96,513,918 96,518,109-96,587,452 96,593,312-96,720,514 96,832,260-96,981,043 96,998,038-97,185,920 97,222,173-97,292,673

3513 18573 166450 458 235 14506 35985 8942 0 2486

2 2 2 2 2 2 2 2 2 2

2746 21 344 2 1 8325 4 21 0 4151

1 0 1 1 1 1 0 0 1 1

2474 11315 170269 698 152 12641 20616 7568 0 3164

2 2 2 2 2 2 2 2 2 2

3316 6251 215495 471 111 19522 28820 4218 0 3524

2 2 2 2 2 2 2 2 2 2

1 0 0 0 0 0 0 0 1

0 0 0 0 0 0 0 0 0

918 0 0 0 0 0 0 686 1411

1 0 0 0 1 1 1 1 1

2 0 0 0 0 0 0 576 1230

0 0 0 0 1 1 1 1 1

1846 883 62 82 0 0 415 1307 1939

2 2 2 2 2 2 2 2 2

Chromosome 9 MTAP CDKN2A CDKN2B CDKN2B-AS1 DMRTA1 ELAVL2 TUSC1 CAAP1 PLAA

21,802,543-21,937,651 21,967,753-21,995,301 22,002,903-22,009,363 21,994,778-22,121,097 22,446,841-22,455,740 23,690,104-23,826,337 25,676,389-25,678,440 26,840,685-26,892,804 26,904,083-26,947,463

Gene expression units are read counts / million (RCM). Regions of complete / near complete loss of expression and copy number (CN) of 0 are highlighted.

ing occurred in areas of normal bone marrow before malignant cell infiltration suggesting systemic suppression of normal hematopoiesis, possibly through a mechanism such as cytokine scavenging, as recently reported to account for cytopenia in acute myeloid leukemia.26 The effect did not correlate with late stage disease, as the mean times between transplantation and culling were almost identical and spleen weights were greater and CNS infiltration heavier for mice with no evidence of niche destruction. Further analysis demonstrated that niche destruction is not restricted to the iAMP21-ALL subtype but is probably less common than suggested by our initial data. Global analysis of genomes from the iAMP21-ALL patients and xenografts revealed a dynamic branching of genomic architecture, similar to that reported previously for B-ALL.4,6,7,27,28 However the rate of newly emergent CNA and their diversity in iAMP21-ALL xenografts suggested a leukemia-initiating cell compartment characterized by greater genetic heterogeneity compared with other B-ALL sub-types. Genomic arrays revealed an average of five CNA per transplanted iAMP21-ALL sample, while similar analysis defined only a single change among seven KMT2A-rearranged infant ALL samples engrafted into multiple mice.28 Additionally, among 12 BCR-ABL1-positive ALL samples, half showed no CNA discordance in xenografts.27 The iAMP21-ALL primografts also developed disease with a relatively long latency. Together with the older age of patients at diagnosis of iAMP21-ALL,12 these data suggest that the primary iAMP21 rearrangement confers only a moderate growth advantage, producing an 642

indolent disease course over which diverse genetic subclones are sampled. As genetic diversity has been linked to clinical aggressiveness,29 this clonal heterogeneity of iAMP21-ALL may underlie the affected patients’ poor response to standard therapy.17 Although each iAMP21 chromosome is unique with respect to the balance of regions amplified and deleted, within clinical trials patients are treated homogeneously.17 Our data support this approach, as they further confirm iAMP21 to be a primary abnormality, because the region identified as consistently amplified and spared from chromothripsis,11 was always retained. However in xenografts from one patient, we observed segregation of a structurally evolved iAMP21 chromosome which, together with other CNA, marked a clone that appeared to confer an exceptionally strong growth advantage. Structural evolution of iAMP21 has not been reported previously, although only few presentation/relapse pairs have been analyzed at the whole genome level and FISH is usually targeted only to the RUNX1 gene. This case demonstrates that even after stabilization of the iAMP21 chromosome evident at the time of diagnosis, these rearrangements can undergo further evolution, potentially influencing clinical features and treatment response. However this iAMP21 chromosome may be atypical, as it was reported to be a ring chromosome, which are known to be inherently unstable structures.30 It may be that other iAMP21 ring chromosomes have a tendency to further evolution, but this case was the only one included in this study. Whether this iAMP21 chromosomal evolution acted as a driver of leukemia prohaematologica | 2018; 103(4)

iAMP21 xenografts

gression remains uncertain, as it was co-selected with other abnormalities, including an NRAS mutation and biallelic loss of CDKN2A/B. Among the four other cases transplanted, three were affected by concordant or discordant CDKN2A deletions, two bi-allelic and one mono-allelic, detected by MLPA only and without apparent involvement of CDKN2B. Further suggesting strong selective pressure for loss of CDKN2A/B in the xenografts, as evidenced by RNA sequencing data, the bi-allelic deletions were all highly clonal. As deletion of this locus only occurs in about 12% of iAMP21-ALL patients,21 these observations support previous reports that CDKN2A/B loss is associated with rapid disease manifestation27 and is selected for in B-ALL xenografts,4 and are also in keeping with a xenograft-specific expression signature enriched for cell cycle genes.31 Alternative mutations of NRAS and KRAS were also strongly selected and both apparently cooperated with an NF1 mutation in xenografts. To our knowledge NF1 and RAS mutations have always been reported as mutually exclusive in B-ALL patients, although their co-occurrence in juvenile myelomoncytic leukemia has been associated with aggressive disease. In mouse models, a combination of NF1 deficiency and KRAS activating mutation reduced the latency of myeloid malignancy compared with either abnormality alone.13,32,33 Other chromosomal regions were strongly implicated in the progression of ALL, as targets of overlapping abnormalities segregated in different clones of xenografts. These included genes known to be involved in B-ALL; ETV6, SH2B3 and BLNK (SLP-65),34-36 as well as novel candidate tumor suppressor genes. Two distinct large deletions, selected in different xenografts, resulted in conversion to homozygosity of a micro-deletion involving PIK3AP1 and LCOR. PIK3AP1 encodes an adaptor protein linking the Bcell receptor and CD19 to activation of PI3K/Akt.37-39 A similar function in the transduction of pre-B-cell receptor signaling is likely and, although not previously implicated in childhood leukemia, focal deletions of PIK3AP1 have been

References 1. Rehe K, Wilson K, Bomken S, et al. Acute B lymphoblastic leukaemia-propagating cells are present at high frequency in diverse lymphoblast populations. EMBO Mol Med. 2013;5(1):38-51. 2. Barrett DM, Seif AE, Carpenito C, et al. Noninvasive bioluminescent imaging of primary patient acute lymphoblastic leukemia: a strategy for preclinical modeling. Blood. 2011;118(15):e112-117. 3. Morisot S, Wayne AS, Bohana-Kashtan O, et al. High frequencies of leukemia stem cells in poor-outcome childhood precursorB acute lymphoblastic leukemias. Leukemia. 2010;24(11):1859-1866. 4. Schmitz M, Breithaupt P, Scheidegger N, et al. Xenografts of highly resistant leukemia recapitulate the clonal composition of the leukemogenic compartment. Blood. 2011;118 (7):1854-1864. 5. Terziyska N, Castro Alves C, Groiss V, et al. In vivo imaging enables high resolution preclinical trials on patients' leukemia cells growing in mice. PLoS One. 2012;7(12): e52798. 6. Anderson K, Lutz C, van Delft FW, et al.

haematologica | 2018; 103(4)

10.

11.

12.

reported in three cases of adult B-ALL.40,41 Combined with our evidence of strong selective pressure for its conversion from mono- to bi-allelic deletion, PIK3AP1 is an interesting candidate, possibly worthy of addition to the growing list of pre-B-cell receptor-related genes disrupted in childhood B-ALL.42 However a role for LCOR, which functions as a co-repressor of several nuclear hormone receptors,43 and has been reported to interact with methyltransferase complexes including polycomb repressive complex 2,44 should not be discounted. In conclusion, we present the first successful xenografts of iAMP21-ALL and demonstrate their potential as experimental models for functional investigation of this poorly understood genetic subtype. These xenografts could also serve as models in pre-clinical trials or for personalized medicine, with the caveat that systemic niche destruction occurred in some cases. Tracking of CNA, to investigate genomic evolution in xenografts, revealed a surprisingly high rate of instability and examples of marked divergence in CN status of known leukemia driver genes. Further genomic screening of iAMP21-ALL xenografts is likely to reveal many more clonal abnormalities undetected in patientsâ&#x20AC;&#x2122; samples, augmenting data generated from clinical trial cohorts, as well as potentially guiding treatment in individual cases. Acknowledgments The authors would like to thank the Department of Cellular Medicine, Royal Victoria Infirmary, Newcastle-upon-Tyne NHS Foundation Trust for fixing and staining histological samples. Tracey Davey and Kath White, Electron Microscopy Research Services, Newcastle University for EM preparations and images and Clare Orange and Lynn Stevenson, University of Glasgow, for the brain histology and imaging. The brain histology slides were scanned by Glasgow University slide scanning and image analysis service at the QEUH. We would also like to thank the Bloodwise Childhood Leukaemia Cell Bank for providing primary childhood leukemia samples.

Genetic variegation of clonal architecture and propagating cells in leukaemia. Nature. 2011;469(7330):356-361. Patel B, Dey A, Castleton AZ, et al. Mouse xenograft modeling of human adult acute lymphoblastic leukemia provides mechanistic insights into adult LIC biology. Blood. 2014;124(1):96-105. Townsend EC, Murakami MA, Christodoulou A, et al. The public repository of xenografts enables discovery and randomized phase II-like trials in mice. Cancer Cell. 2016;29(4):574-586. Rand V, Parker H, Russell LJ, et al. Genomic characterization implicates iAMP21 as a likely primary genetic event in childhood B-cell precursor acute lymphoblastic leukemia. Blood. 2011;117(25):6848-6855. Strefford JC, van Delft FW, Robinson HM, et al. Complex genomic alterations and gene expression in acute lymphoblastic leukemia with intrachromosomal amplification of chromosome 21. Proc Natl Acad Sci USA. 2006;103(21):8167-8172. Li Y, Schwab C, Ryan SL, et al. Constitutional and somatic rearrangement of chromosome 21 in acute lymphoblastic leukaemia. Nature. 2014;508(7494):98-102. Harrison CJ, Moorman AV, Schwab C, et al.

13. 14.

15.

16.

17.

An international study of intrachromosomal amplification of chromosome 21 (iAMP21): cytogenetic characterization and outcome. Leukemia. 2014;28(5):1015-1021. Ryan SL, Matheson E, Grossmann V, et al. The role of the RAS pathway in iAMP21ALL. Leukemia. 2016;30(9):1824-1831. Robinson HM, Broadfield ZJ, Cheung KL, et al. Amplification of AML1 in acute lymphoblastic leukemia is associated with a poor outcome. Leukemia. 2003;17(11): 2249-2250. Moorman AV, Richards SM, Robinson HM, et al. Prognosis of children with acute lymphoblastic leukemia (ALL) and intrachromosomal amplification of chromosome 21 (iAMP21). Blood. 2007;109(6):2327-2330. Attarbaschi A, Mann G, Panzer-Grumayer R, et al. Minimal residual disease values discriminate between low and high relapse risk in children with B-cell precursor acute lymphoblastic leukemia and an intrachromosomal amplification of chromosome 21: the Austrian and German acute lymphoblastic leukemia Berlin-FrankfurtMunster (ALL-BFM) trials. J Clin Oncol. 2008;26(18):3046-3050. Moorman AV, Robinson H, Schwab C, et al. Risk-directed treatment intensification

643

P.B. Sinclair et al.

18.

19.

20. 21.

22.

23. 24. 25.

644

significantly reduces the risk of relapse among children and adolescents with acute lymphoblastic leukemia and intrachromosomal amplification of chromosome 21: a comparison of the MRC ALL97/99 and UKALL2003 trials. J Clin Oncol. 2013;31(27):3389-3396. Heerema NA, Carroll AJ, Devidas M, et al. Intrachromosomal amplification of chromosome 21 is associated with inferior outcomes in children with acute lymphoblastic leukemia treated in contemporary standardrisk Children's Oncology Group studies: a report from the Children's Oncology Group. J Clin Oncol. 2013;31(27):3397-3402. Harrison CJ, Haas O, Harbott J, et al. Detection of prognostically relevant genetic abnormalities in childhood B-cell precursor acute lymphoblastic leukaemia: recommendations from the Biology and Diagnosis Committee of the International Berlin-Frankfurt-Munster study group. Br J Haematol. 2010;151(2):132-142. Schwab C, Harrison CJ. Acute lymphoblastic leukaemia. Methods Mol Biol. 2011;730:99-117. Schwab CJ, Chilton L, Morrison H, et al. Genes commonly deleted in childhood Bcell precursor acute lymphoblastic leukemia: association with cytogenetics and clinical features. Haematologica. 2013;98(7):1081-1088. Bomken S, Buechler L, Rehe K, et al. Lentiviral marking of patient-derived acute lymphoblastic leukaemic cells allows in vivo tracking of disease progression. Leukemia. 2013;27(3):718-721. Travlos GS. Normal structure, function, and histology of the bone marrow. Toxicol Pathol. 2006;34(5):548-565. Elmore S. Apoptosis: a review of programmed cell death. Toxicol Pathol. 2007;35(4):495-516. Russell LJ, Capasso M, Vater I, et al. Deregulated expression of cytokine receptor gene, CRLF2, is involved in lymphoid transformation in B-cell precursor acute

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

lymphoblastic leukemia. Blood. 2009;114(13):2688-2698. Rauch PJ, Ellegast JM, Widmer CC, et al. MPL expression on AML blasts predicts peripheral blood neutropenia and thrombocytopenia. Blood. 2016;128(18):2253-2257. Notta F, Mullighan CG, Wang JC, et al. Evolution of human BCR-ABL1 lymphoblastic leukaemia-initiating cells. Nature. 2011;469(7330):362-367. Bardini M, Woll PS, Corral L, et al. Clonal variegation and dynamic competition of leukemia-initiating cells in infant acute lymphoblastic leukemia with MLL rearrangement. Leukemia. 2015;29(1):38-50. Park SY, Gonen M, Kim HJ, Michor F, Polyak K. Cellular and genetic diversity in the progression of in situ human breast carcinomas to an invasive phenotype. J Clin Invest. 2010;120(2):636-644. Guilherme RS, Meloni VF, Kim CA, et al. Mechanisms of ring chromosome formation, ring instability and clinical consequences. BMC Med Genet. 2011;12:171. Clappier E, Gerby B, Sigaux F, et al. Clonal selection in xenografted human T cell acute lymphoblastic leukemia recapitulates gain of malignancy at relapse. J Exp Med. 2011;208(4):653-661. Caye A, Strullu M, Guidez F, et al. Juvenile myelomonocytic leukemia displays mutations in components of the RAS pathway and the PRC2 network. Nat Genet. 2015;47(11):1334-1340. Cutts BA, Sjogren AK, Andersson KM, et al. Nf1 deficiency cooperates with oncogenic K-RAS to induce acute myeloid leukemia in mice. Blood. 2009;114(17):3629-3632. Moriyama T, Metzger ML, Wu G, et al. Germline genetic variation in ETV6 and risk of childhood acute lymphoblastic leukaemia: a systematic genetic study. Lancet Oncol. 2015;16(16):1659-1666. Perez-Garcia A, Ambesi-Impiombato A, Hadler M, et al. Genetic loss of SH2B3 in acute lymphoblastic leukemia. Blood. 2013;122(14):2425-2432.

36. Jumaa H, Bossaller L, Portugal K, et al. Deficiency of the adaptor SLP-65 in pre-Bcell acute lymphoblastic leukaemia. Nature. 2003;423(6938):452-456. 37. Inabe K, Kurosaki T. Tyrosine phosphorylation of B-cell adaptor for phosphoinositide 3-kinase is required for Akt activation in response to CD19 engagement. Blood. 2002;99(2):584-589. 38. Okada T, Maeda A, Iwamatsu A, Gotoh K, Kurosaki T. BCAP: the tyrosine kinase substrate that connects B cell receptor to phosphoinositide 3-kinase activation. Immunity. 2000;13(6):817-827. 39. Castello A, Gaya M, Tucholski J, et al. Nckmediated recruitment of BCAP to the BCR regulates the PI(3)K-Akt pathway in B cells. Nat Immunol. 2013;14(9):966-975. 40. Okamoto R, Ogawa S, Nowak D, et al. Genomic profiling of adult acute lymphoblastic leukemia by single nucleotide polymorphism oligonucleotide microarray and comparison to pediatric acute lymphoblastic leukemia. Haematologica. 2010;95(9):1481-1488. 41. Safavi S, Hansson M, Karlsson K, Biloglav A, Johansson B, Paulsson K. Novel gene targets detected by genomic profiling in a consecutive series of 126 adults with acute lymphoblastic leukemia. Haematologica. 2015;100(1):55-61. 42. Eswaran J, Sinclair P, Heidenreich O, et al. The pre-B-cell receptor checkpoint in acute lymphoblastic leukaemia. Leukemia. 2015;29(8):1623-1631. 43. Fernandes I, Bastien Y, Wai T, et al. Liganddependent nuclear receptor corepressor LCoR functions by histone deacetylasedependent and -independent mechanisms. Mol Cell. 2003;11(1):139-150. 44. Alekseyenko AA, Gorchakov AA, Kharchenko PV, Kuroda MI. Reciprocal interactions of human C10orf12 and C17orf96 with PRC2 revealed by BioTAPXL cross-linking and affinity purification. Proc Natl Acad Sci USA. 2014;111(7):24882493.

haematologica | 2018; 103(4)

ARTICLE

Complications in Hematology

Prevalence and characteristics of metabolic syndrome in adults from the French childhood leukemia survivors’ cohort: a comparison with controls from the French population Claire Oudin,1,2 Julie Berbis,2 Yves Bertrand,3 Camille Vercasson,2 Frédérique Thomas,4 Pascal Chastagner,5 Stéphane Ducassou,6 Justyna Kanold,7 Marie-Dominique Tabone,8 Catherine Paillard,9 Marilyne Poirée,10 Dominique Plantaz,11 Jean-Hugues Dalle,12 Virginie Gandemer,13 Sandrine Thouvenin,14 Nicolas Sirvent,15 Paul Saultier,1 Sophie Béliard,16 Guy Leverger,8 André Baruchel,12 Pascal Auquier,² Bruno Pannier4 and Gérard Michel1,2

Ferrata Storti Foundation

Haematologica 2018 Volume 103(4):645-654

Department of Pediatric Hematology and Oncology, Timone Enfants Hospital and Aix-Marseille University, Marseille; ²Research Unit EA 3279 and Department of Public Health, Aix-Marseille University and Timone Hospital, Marseille; 3Department of Pediatric Hematology and Oncology, University Hospital of Lyon; 4Preventive and Clinical Investigation Centre, Paris; 5Department of Pediatric Onco-Haematology, Children’s Hospital of Brabois, Vandoeuvre Les Nancy; 6Department of Pediatric Hematology and Oncology, University Hospital of Bordeaux; 7Department of Pediatric Hematology and Oncology, CIC Inserm 501, University Hospital of Clermont-Ferrand; 8Pediatric Hematology Department, Trousseau Hospital, Paris; 9Department of Pediatric Hematology-Oncology, University Hospital, Strasbourg; 10Pediatric Hematology and Oncology Department, University Hospital L’Archet, Nice; 11Department of Pediatric Hematology-Oncology, University Hospital of Grenoble; 12Pediatric Hematology Department, Robert Debré Hospital, Paris; 13Department of Pediatric Hematology and Oncology, University Hospital of Rennes; 14Pediatric Hematology, University Hospital, Saint Etienne; 15Pediatric Hematology and Oncology Department, University Hospital, Montpellier and 16Department of Endocrinology and Nutrition, Timone Hospital, Marseille, France 1

ABSTRACT

he prevalence of the metabolic syndrome among adults from the French LEA childhood acute leukemia survivors’ cohort was prospectively evaluated considering the type of anti-leukemic treatment received, and compared with that of controls. The metabolic profile of these patients was compared with that of controls. A total of 3203 patients from a French volunteer cohort were age- and sexmatched 3:1 to 1025 leukemia survivors (in both cohorts, mean age: 24.4 years; females: 51%). Metabolic syndrome was defined according to the National Cholesterol Education Program’s Adult Treatment Panel III criteria. Metabolic syndrome was found in 10.3% of patients (mean follow-up duration: 16.3±0.2 years) and 4.5% of controls, (OR=2.49; P<0.001). Patients transplanted with total body irradiation presented the highest risk (OR=6.26; P<0.001); the other treatment groups also showed a higher risk than controls, including patients treated with chemotherapy only. Odd Ratios were 1.68 (P=0.005) after chemotherapy only, 2.32 (P=0.002) after chemotherapy and cranial irradiation, and 2.18 (P=0.057) in patients transplanted without irradiation. Total body irradiation recipients with metabolic syndrome displayed a unique profile compared with controls: smaller waist circumference (91 vs. 99.6 cm; P=0.01), and increased triglyceride levels (3.99 vs. 1.5 mmol/L; P<0.001), fasting glucose levels (6.2 vs. 5.6 mmol/L; P=0.049), and systolic blood pressure (137.9 vs. 132.8 mmHg; P=0.005). By contrast, cranial irradiation recipients with metabolic syndrome had a larger waist circumference (109 vs. 99.6 cm; P=0.007) than controls. Regardless of the antileukemic treatment, metabolic syndrome risk was higher among childhood leukemia survivors. Its presentation differed depending on the treatment type, thus suggesting a divergent pathophysiology. This study is registered at clinicaltrials.gov identifier: 01756599. haematologica | 2018; 103(4)

Correspondence: gmichel@ap-hm.fr

Received: July 10, 2017. Accepted: January 17, 2018. Pre-published: January 19, 2018. doi:10.3324/haematol.2017.176123 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/645 ©2018 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.

645

C. Oudin et al.

Introduction

Methods

The number of long-term survivors of childhood acute leukemia (AL) has dramatically increased over the past decades. Improvements in treatment and supportive care have resulted in 5-year survival rates exceeding 85% in children with acute lymphoblastic leukemia (ALL)1-3 and approximately 55-60% among children with acute myeloid leukemia (AML).4-6 Therefore, the long-term chronic health condition of such patients has become a major public health concern.7 Several authors have shown that long-term survivors of childhood cancer are at risk of metabolic syndrome, a well-known marker of cardiovascular morbidity and mortality.8-12 Overall, previous studies in this field showed that longterm survivors of childhood AL are at risk of developing a metabolic syndrome, as compared with controls, but several issues need to be addressed. Firstly, our team and others have previously shown that hematopoietic stem cell transplantation (HSCT) and cranial irradiation11,13-18 are clearly associated with a higher risk of metabolic syndrome, while the risk of metabolic syndrome for patients who received chemotherapy only has not been specifically assessed in large comparative studies. Secondly, prevalence of the metabolic syndrome varies greatly from one country to another.19-21 Many metabolic syndrome studies were carried out in the USA,11,17,22 where metabolic syndrome occurs more frequently among the general population than in France19,23-25 or other European countries. Consequently, extrapolating US-based results to understand metabolic syndrome in a non-US country may not be appropriate. Furthermore, the incidence of metabolic syndrome increases with age and is influenced by sex, country of origin, socio-economic status24 and academic level.26 Thus, the real risk of developing metabolic syndrome among survivors of childhood AL cannot be accurately evaluated without considering all these parameters. This study is based on data from the LEA (French acronym for “leukemia in children and adolescents”) cohort, a French prospective multicenter cohort designed to evaluate the long-term health status of childhood AL survivors. The primary objective of the present study was to evaluate the prevalence of metabolic syndrome and its components in adults from the LEA cohort, and to compare it with that found in controls from a French volunteer cohort. More than 1000 patients from the LEA cohort (60% of them treated with chemotherapy only), were compared with age- and sex-matched controls from the Investigation and Clinical Prevention (IPC) cohort, in which volunteer patients were recruited from one of the largest preventive healthcare centers in France. We also aimed to determine the potential effects of different therapeutic modalities used during leukemia treatment on metabolic syndrome occurrence, and to assess the risk of metabolic syndrome among patients who received only chemotherapy [without HSCT and without central nervous system (CNS) irradiation]. Lastly, we aimed to investigate whether the metabolic syndrome profile (i.e. the cluster of components that constitute metabolic syndrome) varied between the IPC cohort and the LEA cohort, which would suggest divergent mechanisms concerning metabolic syndrome development between these two groups.

This study is based on a comparison between the LEA cohort and a control group from the IPC cohort.

646

The LEA group The LEA program was implemented in 2004 to prospectively evaluate the long-term health status, Quality of Life and socioeconomic status of childhood AL survivors enrolled in treatment programs from 1980 to the present, in 16 cancer centers in France. The details of the program have been previously described.27,28 Data regarding different long-term complications were collected during specific medical visits at pre-defined dates (Online Supplementary Appendix). Since 2007, assessment of metabolic syndrome has been systematically proposed to all adults participating in the LEA program. The study was approved by the French National Program for Clinical Research, the National Cancer Institute, and by the review boards of the institutions involved. All patients provided written informed consent for participation in the study. The inclusion criteria for the current study were: 1) participation in the LEA program between 2007 and 2014; 2) older than 18 years of age at last LEA evaluation; and 3) at least one complete evaluation for metabolic syndrome.

Comparison cohort: IPC group The IPC centers are dedicated to the evaluation of the general health status of French patients living in the Paris area (France). These medical centers, funded by the French National Social Security, offer a free medical examination every five years to working and retired employees and their families. During each examination, patients benefit from a medical check-up including medical examination and biological tests. A self-administered questionnaire provides information concerning higher education, medical history, current health status and medication. During these medical check-ups all patients were screened for metabolic syndrome. Selected controls were age- (2-year categories) and sexmatched 3:1 to the LEA patients.

Outcome measurements Metabolic syndrome was defined according to the National Cholesterol Education Program-Adult Treatment Panel III (NCEPATPIII) revised in 200523 as the combination of at least three of the following criteria: 1) increased waist circumference (≥102 cm in men and ≥88 cm in women); 2) increased blood pressure (systolic blood pressure ≥130 mmHg and/or diastolic blood pressure ≥85 mmHg) or treatment for hypertension; 3) reduced HDL-cholesterol [<40 mg/dL (1.03 mmol/L) in men, <50 mg/dl (1.3 mmol/l) in women]; 4) elevated fasting glucose levels (≥5.5 mmol/L) or treatment for hyperglycemia; and 5) increased triglycerides (≥1.7 mmol/L) or treatment for hypertriglyceridemia. For further details, see the Online Supplementary Appendix.

Statistical analysis Statistical analysis was performed using SPSS 20.0 (SPSS Inc., Chicago, IL, USA) and Intercooled Stata 9.0 for Windows. Qualitative data are expressed as percentages. Quantitative data are shown as mean±Standard Error of Mean (SEM). χ2 and Fischer exact tests were used to compare qualitative variables. Quantitative variables were compared using the Student test or the Mann-Whitney test. Logistical regression, adjusted for sex and age, was used to evaluate the probability of developing metabolic syndrome in the LEA subgroups and the IPC group. Odds Ratios (ORs) were estimated with 95% confidence interval (CI). P<0.05 was considered significant. haematologica | 2018; 103(4)

Metabolic syndrome in childhood leukemia survivors

Table 1. Leukemia subtypes and treatment regimen in patients in the Leukemia in Children and Adolescents (LEA) cohort. Comparison between included and eligible but non-included patients.

Leukemia subtype ALL AML Mean age at leukemia diagnosis, years Mean follow-up duration (between leukemia diagnosis and last LEA evaluation), years Total dose of steroids, mg/m² Treatment modalities Chemotherapy alone Chemotherapy + central nervous system irradiation Hematopoietic stem cell transplantation after TBI conditioning Hematopoietic stem cell transplantation after Bu-based conditioning regimen Hematopoietic stem cell transplantation Autologous stem cell transplantation Allogeneic stem cell transplantation Donor type (n, % of allo-transplant recipients) Matched sibling donor Mismatched related donor Matched unrelated donor Cord blood unit Central nervous system irradiation 18 Grays (n, % of irradiation recipients) 24 Grays (n, % of irradiation recipients) Other Irradiation subtype Cranial irradiation (n, % of irradiation recipients) Cranio-spinal irradiation (n, % of irradiation recipients)

Included LEA patients (n=1025) n(%) or mean ± SEM

Eligible nonincluded (n=437) n(%) or mean ± SEM

867 (84.6) 158 (15.4) 8.37 ± 0.15 16.32 ± 0.21

379 (86.7) 58 (13.3) 8.50 ± 0.23 15.64 ± 0.33

0.29 0.63 0.08

4688.28 ± 148.35

4880.52 ± 195.31

0.43

637 (62.2) 143 (13.9) 168 (16.4) 77 (7.5) 245 (23.9) 65 (26.5) 180 (73.5)

261 (59.7) 70 (16.0) 83 (19.0) 23 (5.3) 106 (24.3) 24 (22.6) 82 (77.4)

0.20

105 (62.1) 9 (5.3) 32 (18.9) 23 (13.6) 168 (16.4) 128 (76.2) 28 (16.7) 8 (4.8)

47 (61.8) 5 (6.6) 15 (19.7) 9 (11.8) 77 (17.7) 54 (70.1) 13 (16.9) 10 (130)

122 (72.6) 44 (26.2)

61 (79.2) 16 (20.8)

0.88 0.65

0.96

0.53 0.08

0.32

n: number; SEM: Standard Error of Mean; ALL: acute lymphoblastic leukemia; AML: acute myeloid leukemia; TBI: total body irradiation; Bu-based: busulfan-based.

Results Characteristics of the LEA cohort Among the 3188 participants in the LEA cohort between 2007 and 2014, 1462 were eligible (aged >18 years at last evaluation) of whom 1025 had a complete evaluation of the metabolic syndrome; all 1025 were included in the present study (for details, see the flow chart in Online Supplementary Figure S1). Characteristics of the LEA patients are summarized in Table 1, which also includes a comparison between included and eligible but not included patients. No significant difference was noted between included and eligible but not included patients in terms of AL subtype, age at diagnosis, and AL treatment modalities. Among the 1025 included patients, 524 (51.1%) were females; 867 patients (84.6%) had had ALL and 15.4% AML. The mean follow-up duration from leukemia diagnosis to last metabolic syndrome evaluation was 16.32±0.21 years. Patients were treated according to the protocols in use at the time of AL diagnosis, depending on leukemia subtype (AML or ALL) (i.e. FRALLE, EORTC, LAME or ELAM). Most of the included patients received haematologica | 2018; 103(4)

chemotherapy only (n=637, 62.2%), while 143 patients (13.9%) were treated with chemotherapy and CNS irradiation. Overall, 245 patients (23.9%) received HSCT after a conditioning regimen with (n=168, 68.6%) or without (n=77, 31.4%) TBI. Finally, 180 (73.5%) of the 245 transplanted patients received allogeneic stem cell transplantation.

Characteristics of the IPC group The control group included 3203 patients who were age- and sex-matched 3:1 to the LEA patients. Table 2 provides demographic characteristics and socio-economic data concerning the LEA and IPC patients. As expected, no difference in age at evaluation for metabolic syndrome was noted between the LEA and IPC groups (24.4±0.2 vs. 24.4± 0.1 years, respectively). Similarly, the sex ratio was the same in both groups. Notably, the LEA patients had an overall higher socio-economic status and higher education level compared to controls; for example, LEA patients were more frequently currently employed than controls (79.5% vs. 47.3%, respectively; P<0.001) and were more likely to have a higher level of education (57.8% vs. 647

C. Oudin et al.

40.2%, respectively; P<0.001). Furthermore, controls (IPC group) had a higher mean body mass index (BMI) than LEA patients (23.7±4.71 and 23.3±4.46 for IPC and LEA patients, respectively; P=0.01).

Prevalence of metabolic syndrome: comparison between the LEA and the IPC groups The metabolic syndrome prevalence was 10.3% (n=106/1025) in the LEA group and 4.5% (n=145/3203) in the IPC group (P<0.001), with an OR of 2.49 (95%CI: 1.91-3.25) (Table 3). The metabolic syndrome occurred in 9.7% of female LEA patients, whereas the syndrome occurred in 4% of control females (OR: 2.56, 95%CI: 1.753.74; P<0.001). Metabolic syndrome was observed in 11%

of male LEA patients and 5% of male controls (OR: 2.33, 95%CI: 1.63-3.34; P<0.001). Since blood pressure is the variable the most affected by fluctuations and might be overestimated in some cases, we analyzed the metabolic syndrome occurrence excluding the hypertension criteria: 46 LEA patients (4.5%) had a metabolic syndrome versus 66 controls (2.1%). The difference was still statistically significant (P<0.001). We analyzed the cumulative incidence of the metabolic syndrome among LEA patients over time: 7.86% (95%CI: 5.99-10.29) at 25 years, and 14.42% (95%CI: 11.22-18.43) at 30 years (Online Supplementary Appendix). Prevalence of the metabolic syndrome components

648

Figure 1. Biological markers. Biological markers of metabolic syndrome (triglycerides, HDL-cholesterol and fasting glucose levels) among Leukemia in Childhood and Adolescents (LEA) cohort patients displaying a metabolic syndrome (n=106) according to treatment modality: hematopoietic stem cell transplantation (HSCT) with total body irradiation (TBI): n=39; HSCT without TBI: n=7; no HSCT with central nervous system (CNS) irradiation: n=18; no HSCT/no CNS irradiation: n=42. LEA patients were compared with Investigation and Clinical Prevention (IPC) group patients (controls) with metabolic syndrome (n=145), adjusted according to sex and age. Results are expressed as mean±Standard Error of Mean (SEM). (A) Triglyceride levels. (B) HDL cholesterol levels. (C) Fasting glucose levels.

haematologica | 2018; 103(4)

Metabolic syndrome in childhood leukemia survivors

(considered separately) also differed between the LEA and IPC patients: LEA patients were more likely than IPC subjects to have a larger waist circumference [17% vs. 10.6%, respectively, OR=1.79 (95%CI: 1.43-2.23); P<0.001], elevated triglycerides [14% vs. 4.5%, respectively, OR=3.59 (95%CI: 2.8-4.6); P<0.001] and high blood pressure [33.9% vs. 22.8%, respectively, OR=1.81 (95%CI: 1.532.14); P<0.001]. By contrast, the prevalence of elevated fasting glucose and low HDL-cholesterol levels were not elevated in the LEA group.

The highest prevalence of metabolic syndrome was found in patients who received HSCT (prevalence: 18.8%, OR: 4.87, 95%CI: 3.4-6.99; P<0.001). TBI before HSCT was associated with the highest prevalence of metabolic syndrome (23.2%) as well as the highest relative risk of developing metabolic syndrome (OR=6.26, 95%CI: 4.179.36; P<0.001). Notably, women who received HSCT after TBI were at particularly high risk of developing metabolic syndrome compared with females from the control group [OR=9.25 (95%CI: 5.33-16.1); P<0.001]. Male patients

haematologica | 2018; 103(4)

Figure 2. Clinical markers. Clinical markers (blood pressure and waist circumference) of metabolic syndrome among Leukemia in Childhood and Adolescents (LEA) cohort patients who show a metabolic syndrome (n=106) according to treatment modality: hematopoietic stem cell transplantation (HSCT) with total body irradiation (TBI): n=39; HSCT without TBI: n=7; no HSCT with central nervous system (CNS) irradiation: n=18; no HSCT/no CNS irradiation: n=42. LEA patients with metabolic syndrome were compared with Investigation and Clinical Prevention (IPC) group patients (controls) with metabolic syndrome (n=145). Results are expressed as meanÂąStandard Error of Mean (SEM). (A) Waist circumference. (B) Systolic blood pressure. (C) Diastolic blood pressure. *Significant difference,

649

C. Oudin et al.

receiving TBI, compared with males from the IPC cohort, were also at higher risk of metabolic syndrome [OR=4.13 (95%CI: 2.26-7.56); P<0.001]. Transplanted patients without TBI had also a higher risk of metabolic syndrome, even if it did not reach the significance threshold (OR=2.18, 95%CI: 0.97-4.86; P=0.057). We also found that patients who underwent CNS irradiation were more likely to develop metabolic syndrome compared with controls [OR= 2.32 (95%CI: 1.36-3.97); P=0.002]. Interestingly, patients who received chemotherapy only were also more likely to develop metabolic syndrome

than controls [OR= 1.68 (95%CI: 1.17-2.41); P=0.005] (Table 3). Age at diagnosis or at transplantation did not impact the metabolic syndrome risk among the LEA patients.

Metabolic syndrome profile: impact of treatment modalities We also aimed to determine whether the metabolic profile among patients who had a metabolic syndrome was different between LEA patients and controls. This is the

Table 2. Characteristics of the Leukemia in Children and Adolescents (LEA) and the Investigation and Clinical Prevention) (IPC) groups. Comparison of age, sex, socio-economic status and education level.

Mean age, years Sex Male Female Socio-economic status Marital status Married/living with a partner or with family Yes No Missing data House-owner Yes No Missing data Supplementary health insurance Yes No Missing data Complementary free medical care Education level Ongoing education Education completed Missing data Higher education No higher education Occupation Currently employed Seeking employment Other (without employment, not seeking employment) Other cardio vascular risk factors Body Mass Index Smoking habits Smoker Non smoker Former smoker Missing data

IPC n=3203 n (%) or mean ± SEM

LEA n= 1025 n (%) or mean ± SEM

24.4 ± 0.1

24.4 ± 0.2

1573 (49) 1630 (51)

501 (49) 524 (51)

2016 (64) 1134 (36) 53

731 (82.0) 161 (18.0) 11

<0.001

234 (7.4%) 2930 (92.6%) 39

117 (18.7%) 510 (81.3%) 276

<0.001

2123 (67.1%) 1039 (32.9%) 41 446 (16.2%)

794 (92.0%) 69 (8.0%) 40 39 (4.5%)

<0.001

1095 (34.4%) 2087 (65.6%) 21 930 (40.2%) 1385 (59.8%)

368 (41.1%) 52 (58.9%) 8 424 (57.8%) 310 (42.2%)

<0.001

987 (47.3%) 1099 (52.7%) 0 (0.0%)

408 (79.5%) 92 (17.9%) 13 (2.6%)

<0.001

23.7 (±4.71)

23.3 (±4.46)

0.01

854 (26.7) 1708 (53.3) 242 (7.6) 398 (12.4)

185 (18) 493 (48.1) 9 (0.9) 338 (33)

<0.001

<0.01

n: number; SEM: Standard Error of Mean; NS: not significant. Significant values are in bold. Supplementary health insurance is typically purchased by individuals with a good economic status, whereas complementary free medical care is dedicated only to the poorest patients.

650

haematologica | 2018; 103(4)

Metabolic syndrome in childhood leukemia survivors

reason why in this study we selected only patients with a metabolic syndrome from the LEA and IPC cohorts (Figures 1 and 2). When assessing the metabolic syndrome patients in both groups (n=106 in the LEA group, n=145 in the control group), metabolic profile differed between the two groups of patients. In the LEA group with metabolic syndrome, patients had significantly higher levels of triglycerides [mean triglyceride level: 2.74±0.4 mmol/L (LEA patients) vs. 1.5±0.1 mmol/L (IPC patients); P=0.001] and elevated systolic and diastolic blood pressure (mean systolic blood pressure: 137.1±1.3 (LEA) and 132.8±1.1 mmHg (IPC); P=0.005; mean diastolic blood pressure: 81.7±1.1 (LEA) vs. 78.5±0.8 mmHg (IPC); P=0.01 (Figures 1 and 2). Patients with metabolic syndrome from the LEA group who received TBI had a particular metabolic profile: in spite of a smaller mean abdominal circumference [91±2 (LEA) vs. 99.6±1.5 cm (IPC); P=0.01], these patients displayed higher mean triglyceride levels (3.99±1.06 vs. 1.5±0.07 mmol/L, respectively; P<0.001), higher mean blood pressure (systolic blood pressure: 137.9±2.2 mmHg vs. 132.8±1.1, respectively; P=0.005) and higher mean fasting glucose levels (6.2±0.3 vs. 5.6±0.1 mmol/L, respectively; P=0.049) than patients with metabolic syndrome from the control group. By contrast, metabolic syndrome patients from the LEA group who received CNS irradiation without TBI also displayed higher triglyceride levels (mean: 2.17±0.25 mmol/L vs. 1.5±0.07, respectively; P=0.002) but had a larger waist circumference (109±4.5 cm vs. 99.6±1.5 cm, respectively; P=0.007) compared with metabolic syndrome patients in the control group. Metabolic syndrome patients who received HSCT without TBI in the LEA cohort showed higher systolic blood pressure levels (mean: 140.6±8.1 mmHg vs. 132.8±1.1 mmHg, respectively; P=0.039) but lower fasting glucose levels (mean: 4.4±0.6 mmol/L vs. 5.6±0.1 mmol/L; P=0.049) compared with metabolic syndrome controls. Lastly, LEA patients with metabolic syndrome who were treated with chemotherapy only displayed higher triglyceride levels (1.94±0.17 vs. 1.5±0.07 mmol/L, respectively; P=0.008) and higher systolic blood pressure (138.6±1.5 vs. 132.8±1.1; P=0.02) compared with metabolic syndrome controls.

Discussion Here we report on one of the largest comparative studies on metabolic syndrome prevalence among adults treat-

ed for AL during childhood or adolescence. We found that patients from the LEA cohort were at greater risk of developing metabolic syndrome (OR=2.49, 95%CI: 1.91-3.25) compared to controls, regardless of the treatment they received. Furthermore, the risk was significantly higher for patients treated exclusively with chemotherapy compared to controls (OR= 1.68, 95%CI: 1.17-2.41; P=0.005), a fact which has never been shown before, even though it had been suspected. The results presented in previous metaanalyses or comparative studies have so far remained unconfirmed, probably due to the relatively low number of patients receiving chemotherapy only11,29-31 or the lack of proper controls.32,33 The Saint Jude study published by Nottage et al.11 reported a higher risk of metabolic syndrome in 784 AL survivor patients as compared to controls, but 64.6% of them had received cranial irradiation. In the subgroup of patients treated exclusively with chemotherapy, no significant increase in the risk of developing metabolic syndrome could be shown, probably due to the small number of patients who received chemotherapy only (n=277). By contrast, the LEA patients included in this study were mainly treated with chemotherapy only (n=637, 62.2%). So, we demonstrate here that, even in the case of treatment with chemotherapy only, the metabolic syndrome risk is higher among long-term AL survivors than in the control population. This is important since nowadays, the majority of children treated for an AL will receive only chemotherapy, without any irradiation or HSCT. Several mechanisms have been discussed to explain how metabolic syndrome develops after chemotherapy treatment. Alkylating agents are known to induce mitochondrial dysfunction and endothelial cytotoxicity, which can lead to insulin resistance, steatosis and hypertension. Anthracyclins and antimetabolites can also cause mitochondrial and endothelial dysfunction. Steroids can induce hyperglycemia and dyslipidemia. Vinca alkaloids induce endothelial toxicity and can cause hyperglycemia by inhibition of GLUT2/4 vesicle translocation.34 Lastly, iron overload, a frequent complication after multiple transfusions in patients treated for acute leukemia, could also increase the risk for metabolic syndrome through hepatic toxicity. As a consequence, the notable metabolic syndrome risk in patients treated exclusively with chemotherapy, who represent the majority of future AL survivors, should be taken into careful consideration throughout their long-term follow up. This high risk among LEA patients is striking given that the LEA patients were found to benefit from more favor-

Table 3. Metabolic syndrome prevalence among Leukemia in Children and Adolescents (LEA) patients according to treatment modality. Comparison with the Investigation and Clinical Prevention) (IPC) group (sex- and age-matched controls).

IPC patients LEA patients Chemotherapy alone, no CNS irradiation Chemotherapy+CNS irradiation HSCT without TBI HSCT after TBI conditioning

Metabolic syndrome n (%)

3203 1025 637 / 1025 (62.2%) 143 / 1025 (13.9%) 77 / 1025 (7.5%) 168 / 1025 (16.4%)

145 (4.5%) 106 (10.3%) 42 (6.6%) 18 (12.6%) 7 (9.1%) 39 (23.2%)

OR (95% CI)

2.49 1.68 2.32 2.18 6.26

(1.91 – 3.25) (1.17 – 2.41) (1.36 – 3.97) (0.97 – 4.86) (4.17 – 9.36)

P <10-3 0.005 0.002 0.057 <10-3

N/n: number; OR: Odds Ratio; CI: Confidence Interval; CNS: central nervous system; HSCT: hematopoietic stem cell transplantation; TBI: total body irradiation. Significant values are in bold.

haematologica | 2018; 103(4)

651

C. Oudin et al.

able socio-economic conditions (e.g. higher level of education, higher rates of employment) than IPC patients. Indeed, it is well known that lower socio-economic status is associated with a higher risk of metabolic syndrome.24,35 As the control group (IPC population) was characterized by particularly unfavorable social and economic conditions (unemployed: 52.7%, no higher education: 59.8%), the relative risk of metabolic syndrome in the LEA group may be underestimated. Due to these unfavorable socioeconomic conditions, the IPC group is probably at higher risk of metabolic syndrome than the general French population, which makes the risk described in the LEA patients even higher. Moreover, LEA patients had a lower mean BMI, and were less frequently smokers than controls, which could contribute to their relative risk of developing a metabolic syndrome being underestimated. Few studies on metabolic syndrome in AL survivors consider socio-economic factors,36 whereas studies about metabolic syndrome in the general population have shown this to be very important. Given the impact of socio-economic status on the development of the metabolic syndrome, further studies are needed to investigate a potential correlation between lifestyle (e.g. sedentary behavior), eating habits, and the occurrence of metabolic syndrome in AL survivors. Our findings concerning metabolic syndrome prevalence in the LEA cohort (10.3%) is difficult to compare with studies from other countries, as metabolic syndrome prevalence in the general population varies from one country to another. Furthermore, metabolic syndrome prevalence is lower in France than in many other industrialized countries.25,37 Previous reports of metabolic syndrome prevalence among AL survivors range from 4.2% to 49%,11,16,32,38-40 with prevalence depending mainly on the treatment type (chemotherapy, HSCT or CNS irradiation), follow-up duration and metabolic syndrome definition. In several studies, HSCT was associated with a high prevalence of metabolic syndrome, ranging from 31% to 49%.16,39,40 The study by Nottage et al. reported a higher metabolic syndrome prevalence among non-transplanted patients (33.6%) than that found in our study, and this can partly be explained by the number of patients who received CNS irradiation (507 of 784 patients) and the older age of the patients.11 The risk of developing metabolic syndrome was even higher among patients who received CNS irradiation (OR: 2.32, 95%CI: 1.36-3.97; P=0.002), which confirms the results of previous studies.15,17,38 Interestingly, those patients had a much larger waist circumference compared with the IPC group. This highlights the impact of CNS irradiation on the development of obesity, an observation which has been previously described11,41,42 and debated.43,44 The highest risk of metabolic syndrome was observed among transplanted patients (prevalence: 18.8%, OR: 4.87, 95%CI: 3.4-6.99; P<0.001), especially those who received TBI (OR: 6.26, 95%CI: 4.17-9.36; P<0.001). The deleterious impact of TBI on the development of the metabolic syndrome has previously been reported by our group13 and others;18,32 here, these data are confirmed with a larger number of patients. Apart from TBI, so far additional risk factors for metabolic syndrome in HSCT patients have not been clearly documented: some authors have suggested that graft-versus-host disease would increase this risk,18 but this remains controversial. Several studies, including ours, have also shown an 652

association between metabolic syndrome and growth hormone deficiency,14,18,33,45 particularly in patients who received CNS irradiation or TBI. However, the precise mechanism by which growth hormone deficiency could induce a metabolic syndrome remains unclear. We also aimed to determine whether the metabolic profile among patients who had a metabolic syndrome was different between LEA patients and controls. This is why we only selected patients with a metabolic syndrome from among LEA and IPC patients in order to study this metabolic profile (Figures 1 and 2). Interestingly, we found that, compared with metabolic syndrome controls, patients with metabolic syndrome who received HSCT after TBI had a specific metabolic profile: they had more elevated triglycerides and fasting glucose levels, as well as higher blood pressure. This suggests that these patients develop a metabolic syndrome with more severe features than that of controls, which could lead to higher rates of cardiovascular morbidity, as suggested for the general population.46 Patients who developed metabolic syndrome after TBI had a smaller waist circumference than IPC patients with metabolic syndrome. Altogether, these results suggest that obesity is not a key factor after TBI, in contrast to the general population. Different hypotheses can be made concerning the pathophysiology in those patients. Irradiation of the pancreas during TBI can induce diabetes47 and therefore metabolic syndrome. Furthermore, some authors have found that, in patients treated with TBI, insulin resistance was not associated with obesity but rather with abnormal fat mass repartition.48,49 Modification of adipose tissue metabolism has been recognized as a fundamental mechanism behind metabolic syndrome development.50 Therefore, TBI exposure may induce adipose tissue abnormalities, as suggested by animal models,51 and contribute to the development of metabolic syndrome. Our previous studies indicated that high-dose corticosteroids do not have an impact on metabolic syndrome,11,15 and, therefore, this hypothesis remains controversial.11,15 Patients with metabolic syndrome who received only chemotherapy displayed higher systolic blood pressure and increased triglyceride levels compared with metabolic syndrome controls, thus suggesting a more severe form of metabolic syndrome. The waist circumference of patients who received CNS irradiation was markedly larger than that of the controls. Obesity caused by CNS irradiation, as previously described,41 is probably linked to the metabolic syndrome development. As the hypothalamus exerts central neuroendocrine functions that control hunger and satiety, hypothalamic irradiation could lead to modifications in food intake and energy balance. One limitation of our study involved the fact that some important metabolic syndrome factors such as physical activity, eating habits and lifestyle have not been documented in our population. However, those factors are known to be very important in the metabolic syndrome genesis. A reduction in physical activity in LEA patients, as well as unhealthy eating habits could worsen the risk of developing metabolic syndrome. Another bias is linked to the control population, which includes only people from the Ile de France region, whereas LEA patients are recruited from throughout France. Some regional differences in the metabolic syndrome prevalence could be found, which might make it difficult to extrapolate the results haematologica | 2018; 103(4)

Metabolic syndrome in childhood leukemia survivors

from the IPC population to the general French population. In summary, our study suggests that metabolic syndrome may develop through different mechanisms, depending on the treatments received. In the present study, we used age-matched controls, which enabled us to accurately evaluate metabolic syndrome risk in this young AL survivor population (mean age: 24 years), regardless of the fact that metabolic syndrome prevalence among individuals in this young age range is poorly described. This prevalence will probably increase with age, as is the case for the general population.23 In conclusion, this study reveals an increased risk of metabolic syndrome among adult survivors of AL, regardless of the treatment they received. Moreover, metabolic syndrome seems to be more severe in the LEA patients than in the control group. The highest risk is observed in patients who received TBI, a group that displays a specific metabolic profile, but all patients treated for childhood AL should be considered at risk of metabolic syndrome, regardless of the treatment they

References 1. Silverman LB, Stevenson KE, O'Brien JE, et al. Long-term results of Dana-Farber Cancer Institute ALL Consortium protocols for children with newly diagnosed acute lymphoblastic leukemia (1985-2000). Leukemia. 2010;24(2):320-334. 2. 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. 3. Hunger SP, Mullighan CG. Acute Lymphoblastic Leukemia in Children. N Engl J Med. 2015;373(16):1541-1552. 4. Zwaan CM, Kolb EA, Reinhardt D, et al. Collaborative Efforts Driving Progress in Pediatric Acute Myeloid Leukemia. J Clin Oncol. 2015;33(27):2949-2962. 5. Creutzig U, Zimmermann M, Bourquin JP, et al. Randomized trial comparing liposomal daunorubicin with idarubicin as induction for pediatric acute myeloid leukemia: results from Study AML-BFM 2004. Blood. 2013;122(1):37-43. 6. Gamis AS, Alonzo TA, Meshinchi S, et al. Gemtuzumab ozogamicin in children and adolescents with de novo acute myeloid leukemia improves event-free survival by reducing relapse risk: results from the randomized phase III Children's Oncology Group trial AAML0531. J Clin Oncol. 2014;32(27):3021-3032. 7. Bhatia S, Armenian SH, Armstrong GT, et al. Collaborative Research in Childhood Cancer Survivorship: The Current Landscape. J Clin Oncol. 2015;33(27):30553064. 8. Mottillo S, Filion KB, Genest J, et al. The metabolic syndrome and cardiovascular risk a systematic review and meta-analysis. J Am Coll Cardiol. 2010;56(14):1113-1132. 9. Ford ES. Risks for all-cause mortality, cardiovascular disease, and diabetes associated with the metabolic syndrome: a summary of the evidence. Diabetes Care. 2005;28(7):1769-1778.

haematologica | 2018; 103(4)

received, even in the case of chemotherapy only. We hypothesize that if early detection of metabolic syndrome is followed by changes in lifestyle (e.g. improved eating habits, more physical activity), it will help to prevent cardiovascular events in this at-risk population.36 We are currently planning controlled intervention studies in order to explore such an approach. Funding The study was funded by the French National Clinical Research Program, the French National Cancer Institute (InCA), the “Laurette Fugain” association, the French National Research Agency (ANR), the “Ligue Contre le Cancer” association, Cancéropôle PACA, the Regional Council PACA and the French Institute for Public Health Research (IRESP). Acknowledgments The authors would like to thank the LEA study group (Supplemental Data), for data collection, as well as the patients and their family.

10. Gami AS, Witt BJ, Howard DE, et al. Metabolic syndrome and risk of incident cardiovascular events and death: a systematic review and meta-analysis of longitudinal studies. J Am Coll Cardiol. 2007;49(4):403-414. 11. Nottage KA, Ness KK, Li C, Srivastava D, Robison LL, Hudson MM. Metabolic syndrome and cardiovascular risk among longterm survivors of acute lymphoblastic leukaemia - From the St. Jude Lifetime Cohort. Br J Haematol. 2014;165(3):364374. 12. Eschwege E. The dysmetabolic syndrome, insulin resistance and increased cardiovascular (CV) morbidity and mortality in type 2 diabetes: aetiological factors in the development of CV complications. Diabetes Metab. 2003;29(4 Pt 2):6S19-27. 13. Oudin C, Simeoni MC, Sirvent N, et al. Prevalence and risk factors of the metabolic syndrome in adult survivors of childhood leukemia. Blood. 2011;117(17):4442-4448. 14. Oudin C, Auquier P, Bertrand Y, et al. Metabolic syndrome in adults who received hematopoietic stem cell transplantation for acute childhood leukemia: an LEA study. Bone Marrow Transplant. 2015;50(11):1438-1444. 15. Saultier P, Auquier P, Bertrand Y, et al. Metabolic syndrome in long-term survivors of childhood acute leukemia treated without hematopoietic stem cell transplantation: an L.E.A. study. Haematologica. 2016;101(12):1603-1610. 16. Majhail NS, Flowers ME, Ness KK, et al. High prevalence of metabolic syndrome after allogeneic hematopoietic cell transplantation. Bone Marrow Transplant. 2009;43(1):49-54. 17. Oeffinger KC, Adams-Huet B, Victor RG, et al. Insulin resistance and risk factors for cardiovascular disease in young adult survivors of childhood acute lymphoblastic leukemia. J Clin Oncol. 2009;27(22):36983704. 18. Friedman DN, Hilden P, Moskowitz CS, et al. Cardiovascular Risk Factors in Survivors of Childhood Hematopoietic Cell Transplantation Treated with Total Body

19.

20.

21.

22.

23.

24.

25.

26.

27.

Irradiation: A Longitudinal Analysis. Biol Blood Marrow Transplant. 2017;23(3):475482. Ervin RB. Prevalence of metabolic syndrome among adults 20 years of age and over, by sex, age, race and ethnicity, and body mass index: United States, 20032006. Natl Health Stat Report. 2009;(13):17. Procopiou M, Philippe J. The metabolic syndrome and type 2 diabetes: epidemiological figures and country specificities. Cerebrovasc Dis. 2005;20(Suppl 1):2-8. Cameron AJ, Shaw JE, Zimmet PZ. The metabolic syndrome: prevalence in worldwide populations. Endocrinol Metab Clin North Am. 2004;33(2):351-375. Oeffinger KC. Are survivors of acute lymphoblastic leukemia (ALL) at increased risk of cardiovascular disease? Pediatr Blood Cancer. 2008;50(2 Suppl):462-467. Pannier B, Thomas F, Eschwege E, et al. Cardiovascular risk markers associated with the metabolic syndrome in a large French population: the "SYMFONIE" study. Diabetes Metab. 2006;32(5 Pt 1):467-474. Vernay M, Salanave B, de Peretti C, et al. Metabolic syndrome and socioeconomic status in France: The French Nutrition and Health Survey (ENNS, 2006-2007). Int J Public Health. 2013;58(6):855-864. Alberti KG, Eckel RH, Grundy SM, et al. Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity. Circulation. 2009;120(16):16401645. Scuteri A, Vuga M, Najjar SS, et al. Education eclipses ethnicity in predicting the development of the metabolic syndrome in different ethnic groups in midlife: the Study of Women's Health Across the Nation (SWAN). Diabet Med. 2008;25(12):1390-1399. Berbis J, Michel G, Baruchel A, et al.

653

C. Oudin et al.

28.

29.

30.

31.

32.

33.

34.

654

Cohort Profile: the French childhood cancer survivor study for leukaemia (LEA Cohort). Int J Epidemiol. 2015;44(1):49-57. Michel G, Bordigoni P, Simeoni MC, et al. Health status and quality of life in longterm survivors of childhood leukaemia: the impact of haematopoietic stem cell transplantation. Bone Marrow Transplant. 2007;40(9):897-904. Karakurt H, Sarper N, Kilic SC, Gelen SA, Zengin E. Screening survivors of childhood acute lymphoblastic leukemia for obesity, metabolic syndrome, and insulin resistance. Pediatr Hematol Oncol. 2012;29(6):551-561. Follin C, Thilen U, Ahren B, Erfurth EM. Improvement in cardiac systolic function and reduced prevalence of metabolic syndrome after two years of growth hormone (GH) treatment in GH-deficient adult survivors of childhood-onset acute lymphoblastic leukemia. J Clin Endocrinol Metab. 2006;91(5):1872-1875. Blijdorp K, van Waas M, van der Lely AJ, Pieters R, van den Heuvel-Eibrink M, Neggers S. Endocrine sequelae and metabolic syndrome in adult long-term survivors of childhood acute myeloid leukemia. Leuk Res. 2013;37(4):367-371. Chow EJ, Simmons JH, Roth CL, et al. Increased cardiometabolic traits in pediatric survivors of acute lymphoblastic leukemia treated with total body irradiation. Biol Blood Marrow Transplant. 2010; 16(12):1674-1681. Gurney JG, Ness KK, Sibley SD, et al. Metabolic syndrome and growth hormone deficiency in adult survivors of childhood acute lymphoblastic leukemia. Cancer. 2006;107(6):1303-1312. Rosen GP, Nguyen HT, Shaibi GQ. Metabolic syndrome in pediatric cancer survivors: a mechanistic review. Pediatr Blood Cancer. 2013;60(12):1922-1928.

35. Vernay M, Malon A, Oleko A, et al. Association of socioeconomic status with overall overweight and central obesity in men and women: the French Nutrition and Health Survey 2006. BMC Public Health. 2009;9:215. 36. Smith WA, Li C, Nottage KA, et al. Lifestyle and metabolic syndrome in adult survivors of childhood cancer: a report from the St. Jude Lifetime Cohort Study. Cancer. 2014;120(17):2742-2750. 37. Gavrila D, Salmeron D, Egea-Caparros JM, et al. Prevalence of metabolic syndrome in Murcia Region, a southern European Mediterranean area with low cardiovascular risk and high obesity. BMC Public Health. 2011;11:562. 38. Faienza MF, Delvecchio M, Giordano P, et al. Metabolic syndrome in childhood leukemia survivors: a meta-analysis. Endocrine. 2015;49(2):353-360. 39. Annaloro C, Usardi P, Airaghi L, et al. Prevalence of metabolic syndrome in longterm survivors of hematopoietic stem cell transplantation. Bone Marrow Transplant. 2008;41(9):797-804. 40. Paris C, Yates L, Lama P, Zepeda AJ, Gutierrez D, Palma J. Evaluation of metabolic syndrome after hematopoietic stem cell transplantation in children and adolescents. Pediatr Blood Cancer. 2012; 59(2):306-310. 41. Garmey EG, Liu Q, Sklar CA, et al. Longitudinal changes in obesity and body mass index among adult survivors of childhood acute lymphoblastic leukemia: a report from the Childhood Cancer Survivor Study. J Clin Oncol. 2008;26(28):4639-4645. 42. Oeffinger KC, Mertens AC, Sklar CA, et al. Obesity in adult survivors of childhood acute lymphoblastic leukemia: a report from the Childhood Cancer Survivor Study. J Clin Oncol. 2003;21(7):1359-1365. 43. Chow EJ, Pihoker C, Hunt K, Wilkinson K,

44.

45.

46.

47.

48.

49.

50.

51.

Friedman DL. Obesity and hypertension among children after treatment for acute lymphoblastic leukemia. Cancer. 2007; 110(10):2313-2320. Razzouk BI, Rose SR, Hongeng S, et al. Obesity in survivors of childhood acute lymphoblastic leukemia and lymphoma. J Clin Oncol. 2007;25(10):1183-1189. Link K, Moell C, Garwicz S, et al. Growth hormone deficiency predicts cardiovascular risk in young adults treated for acute lymphoblastic leukemia in childhood. J Clin Endocrinol Metab. 2004;89(10):5003-5012. DeBoer MD, Gurka MJ, Woo JG, Morrison JA. Severity of the metabolic syndrome as a predictor of type 2 diabetes between childhood and adulthood: the Princeton Lipid Research Cohort Study. Diabetologia. 2015;58(12):2745-2752. de Vathaire F, El-Fayech C, Ben Ayed FF, et al. Radiation dose to the pancreas and risk of diabetes mellitus in childhood cancer survivors: a retrospective cohort study. Lancet Oncol. 2012;13(10):1002-1010. Frisk P, Rossner SM, Norgren S, Arvidson J, Gustafsson J. Glucose metabolism and body composition in young adults treated with TBI during childhood. Bone Marrow Transplant. 2011;46(10):1303-1308. Wei C, Thyagiarajan MS, Hunt LP, Shield JP, Stevens MC, Crowne EC. Reduced insulin sensitivity in childhood survivors of haematopoietic stem cell transplantation is associated with lipodystropic and sarcopenic phenotypes. Pediatr Blood Cancer. 2015;62(11):1992-1999. Virtue S, Vidal-Puig A. Adipose tissue expandability, lipotoxicity and the Metabolic Syndrome--an allostatic perspective. Biochim Biophys Acta. 2010;1801(3):338-349. Poglio S, Galvani S, Bour S, et al. Adipose tissue sensitivity to radiation exposure. Am J Pathol. 2009;174(1):44-53.

haematologica | 2018; 103(4)

ARTICLE

Hodgkin Lymphoma

CD83 is a new potential biomarker and therapeutic target for Hodgkin lymphoma

Ferrata Storti Foundation

Ziduo Li,1,2 Xinsheng Ju,1,2 Kenneth Lee,2,3 Candice Clarke,3 Jennifer L. Hsu,1,2 Edward Abadir,1,2 Christian E. Bryant,1,4 Suzanne Pears,5 Neroli Sunderland,5 Scott Heffernan,5 Annemarie Hennessy,5 Tsun-Ho Lo,1,2 Geoffrey A. Pietersz,6,7 Fiona Kupresanin,1 Phillip D. Fromm,1,2 Pablo A. Silveira,1,2 Con Tsonis,1 Wendy A. Cooper,2,8,9 Ilona Cunningham,10 Christina Brown,2,4 Georgina J. Clark1,2 and Derek N.J. Hart,1,2

Dendritic Cell Research, ANZAC Research Institute, Sydney; 2Sydney Medical School, University of Sydney; 3Department of Anatomical Pathology, Concord Repatriation General Hospital, Sydney; 4Institute of Haematology, Royal Prince Alfred Hospital, Sydney; 5Animal Facility, Royal Prince Alfred Hospital, Sydney; 6Burnet Institute, Melbourne; 7Baker Heart and Diabetes Institute, Melbourne; 8Tissue Pathology and Diagnostic Oncology, Royal Prince Alfred Hospital, Sydney; 9School of Medicine, University of Western Sydney and 10Department of Haematology, Concord Repatriation General Hospital, Sydney, Australia 1

Haematologica 2018 Volume 103(4):655-665

ABSTRACT

hemotherapy and hematopoietic stem cell transplantation are effective treatments for most Hodgkin lymphoma patients, however there remains a need for better tumor-specific target therapy in Hodgkin lymphoma patients with refractory or relapsed disease. Herein, we demonstrate that membrane CD83 is a diagnostic and therapeutic target, highly expressed in Hodgkin lymphoma cell lines and Hodgkin and Reed-Sternberg cells in 29/35 (82.9%) Hodgkin lymphoma patient lymph node biopsies. CD83 from Hodgkin lymphoma tumor cells was able to trogocytose to surrounding T cells and, interestingly, the trogocytosing CD83+T cells expressed significantly more programmed death-1 compared to CD83â&#x20AC;&#x201C; T cells. Hodgkin lymphoma tumor cells secreted soluble CD83 that inhibited T-cell proliferation, and anti-CD83 antibody partially reversed the inhibitory effect. High levels of soluble CD83 were detected in Hodgkin lymphoma patient sera, which returned to normal in patients who had good clinical responses to chemotherapy confirmed by positron emission tomography scans. We generated a human anti-human CD83 antibody, 3C12C, and its toxin monomethyl auristatin E conjugate, that killed CD83 positive Hodgkin lymphoma cells but not CD83 negative cells. The 3C12C antibody was tested in dose escalation studies in non-human primates. No toxicity was observed, but there was evidence of CD83 positive target cell depletion. These data establish CD83 as a potential biomarker and therapeutic target in Hodgkin lymphoma.

Introduction Hodgkin lymphoma (HL) is a B-cell neoplasm that is defined by the presence of Hodgkin Reed-Sternberg cells (HRS). During recent decades, the long-term survival of HL patients has increased, and most patients can be cured through multi-agent chemotherapy, radiotherapy and/or hematopoietic stem cell transplantation.1 Despite this, 25-30% of patients experience either disease relapse or are refractory to chemotherapy and their survival is substantially reduced, especially for elderly patients who do not tolerate intensive therapy.2,3 New targeted therapies for HL are warranted, especially for refractory/relapsed patients and elderly patients where limiting treatment toxicity is essential. Recent studies have focused on the development of therapeutic agents that target HL-specific antigens or regulate the natural immune response in patients. Antibodies targeting HL surface antigens such as CD25 (daclizumab),4 CD20 (rituximab, tositumomab)5,6 or CD30 (brentuximab)7-10 have shown promising results. The programmed death-1(PD-1)/PD-ligand 1 (PDL1) checkpoint inhibitors (nivolumab, pembrolizumab), that reverse the suppreshaematologica | 2018; 103(4)

Correspondence: xinsheng.ju@sydney.edu.au

Received: August 22, 2017. Accepted: January 10, 2018. Pre-published: January 19, 2018. doi:10.3324/haematol.2017.178384 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/655 ÂŠ2018 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.

655

Z. Li et al.

sive communication between the tumor and immune system in tumor microenvironments have also been effective in HL patients.11-13 To date, the main utility of identifying membranebound CD83 has been to define activated dendritic cells (DC), but CD83 is also expressed on the surface of some activated B cells, T cells, macrophages and neutrophils.1418 In addition to a membrane-bound form, there is a membrane cleaved soluble (s) form of CD83. We reported that lymphoma tumor cells (HL and non-Hodgkin lymphoma [NHL]) expressed CD83 and released sCD83 into serum.19,20 Recombinant sCD83 protein has immune inhibitory function in mice and humans.21,22 Recently, CD83 was identified as one of the four classifiers to distinguish HL with anaplastic lymphoma kinase (ALK)anaplastic large cell lymphoma.23 Despite its potential as a relatively specific target, CD83 has not been investigated as a therapeutic target on either HL or NHL. We generated a human anti-human CD83 antibody, 3C12C, which prevents graft-versus-host disease (GvHD) but preserves anti-tumor T-cell function in mice after transplantation with human peripheral blood mononuclear cell (PBMC).24,25 The availability of this potential therapeutic anti-CD83 antibody prompted us to investigate CD83 biology in HL. We show herein that an antibody that detects CD83 in paraffin sections stains Hodgkin and Reed-Sternberg (HRS) cells in most HL lymph node biopsy samples, that HL tumor cells secrete sCD83, and the serum sCD83 level in HL patients correlates with the clinical response. The 3C12C antibody, and its toxin conjugate, killed HL lines and 3C12C depleted CD83 target cells in non-human primate studies without any evidence of toxicity.

Methods HL tissue section and serum samples Lymph node biopsies and serum of HL patients were collected after approval by the Sydney Local Health District (SLHD) Human Research Ethics Committee, consistent with the Declaration of Helsinki. Archival paraffin embedded lymph node biopsies were obtained from 35 HL patients at initial diagnosis (Table 1), while serum samples were collected from six HL patients at diagnosis and during chemotherapy.

Table 1. Characteristics of 35 Hodgkin lymphoma patients.

Age at enrolment (mean, range)

35 (17-71)

Sex Male; n (%) Female; n (%)

18 (51.4%) 17 (48.6%)

Histologic subtype cHL-Nodular sclerosis (NS) cHL-Mixed cellularity (MC) cHL-Lymphocyte rich (LR) cHL-unspecified (CHL-U) Nodular lymphocyte predominant (NLP) Stage at onset I II III IV

21 (60.0%) 7 (20.0%) 1 (2.9%) 2 (5.7%) 4 (11.4%) 3 (8.6%) 19 (54.3%) 5 (14.3%) 8 (22.9%)

cHL: classic Hodgkin lymphoma.

change. Human sCD83 was analyzed by a sCD83 ELISA kit (Sino Biological Inc.).

Antibody Dependent Cell Cytotoxicity (ADCC) Assays Target HL cells labeled with 25mM Calcein-AM (Life Technologies) were co-cultured at a ratio of 1:25 with human PBMC of a healthy donor used as effector cells. Supernatants were collected after three hours to measure released calcein using an enzyme-linked immunosorbent assay (ELISA) Reader (Perkin Elmer). The percentage of specific cytolysis was calculated as described.25

3C12C conjugation with monomethyl auristatin E (3C12C-MMAE) and cytotoxicity on CD83+ cell lines 3C12C is a human immunoglobulin G1 (IgG1) anti-human CD83 mAb selected from a phage display library26 and further engineered to improve affinity.25,27 To produce 3C12C-MMAE, a lysosomal cathepsin B-cleavable, self-emolative dipeptide (ValCit) maleimide linker was prepared from MMAE for conjugation to partially reduced 3C12C using a similar method to brentuximab vedotin.28 The cytotoxic activity was assessed by 7-amino-actinomycin D (7AAD, Thermo Fisher Scientific) staining using flow cytometry.

3C12C trials in non-human primates

Immunohistochemical staining was performed on 3mm sections of formalin fixed paraffin embedded lymph node biopsies of HL patients or non-human primates. The primary antibodies used were mouse anti-human CD20 (L26, Dako), CD83 monoclonal antibodies (mAb; F5, Santa Cruz Biotechnology), CD30 (Ber-H2, Dako), and staining was performed on a Leica Bond III Autostainer (Leica Biosystems) using a Bond Polymer Refine Detection kit for visualization with 3, 3’-diaminobenzidine (DAB). Images were taken with an Olympus BX51 microscopy with an Olympus PP71 camera using Olympus labSens software.

The SLHD Animal Research Ethics Committee approved the study of five non-human primates (Papio Hamadryas baboon), which received intravenous human-IgG (Intragam, CSL) (10 mg/kg) or 3C12C mAb (1, 5, 10, 10 mg/kg) at days 0, 7, 14 and 21. PBMC were analyzed for immune cell populations including Dendritic cells (DC), T cells and B cells on a Fortessa X20 flow cytometer (BD Biosciences). Liver and kidney function were assessed by measuring alkaline phosphatase (ALP), aspartate transaminase (AST) and creatinine in serum samples using the Cobas 8000 (Roche). Lymph nodes were taken from 3C12C (10mg/kg) or human IgG (10mg/kg) treated animals at day 28 for immunohistological staining.

sCD83 analysis

Statistical Analysis

For the analysis of sCD83 levels, KM-H2, L428 and HDLM2 cells were cultured at concentrations of 106 cells/ml in complete roswell park memorial institute (RPMI) medium containing 10% fetal calf serum, 2mM glutaMAX™, 100U/ml penicillin, 100µg/ml streptomycin (Thermo Fisher Scientific) at 37°C, in 5% CO2. Cell culture supernatant were collected 24 hours after fresh medium

Mean values with standard error of mean (SEM) bars are shown in graphs. Statistical analyses were performed using Prism 6.0 (GraphPad Software). A Mann-Whitney or one-way analysis of variance (ANOVA) test with Greenhouse-Geisser correction for multiple comparisons were used. Differences with P<0.05 were considered significant.

Immunohistochemistry

656

haematologica | 2018; 103(4)

CD83 and Hodgkin lymphoma

Results CD83 is expressed on HL cell lines and HRS cells in lymph node biopsies of HL patients Expression of CD83 was analyzed using the mouse antihuman antibodies HB15a, HB15e and potential therapeutic human anti-human CD83 antibody 3C12C.25 KM-H2 cells expressed the most expressive CD38 cell surface, stained as it was with all three anti-CD83 antibodies, whilst the L428 and HDLM2 lines expressed less CD83. All three lines expressed CD30 (Figure 1A). This data was confirmed by confocal CD83 staining on KM-H2 cells (Figure 1B), detection of CD83 messenger ribonucleic acid (mRNA) transcripts by reverse transcription polymerase chain reaction (RT-PCR) and intracellular CD83 expression in the three HL lines (Online Supplementary Figure S1).

Next, CD83 expression was analyzed on the paraffinembedded lymph node biopsies of 35 HL patients (Table 1). The HRS cells were identified as CD30+ (Figure 2A). Of note, 8/35 (22.9%) biopsies expressed high levels of CD83 on the HRS cells (>90% positive), 21/35 (60%) expressed middle levels (10-90% positive), and 6/35 (17.1%) expressed low levels of CD83 (<10% positive) (Figure 2B,C). Of the 29 biopsies with high or middle expression, 21 (72.4%) had strong or moderate intensity, while 8/29 (27.6%) showed weak intensity of CD83 on HRS cells (Online Supplementary Table S1). The subtype analysis showed that 16/21 (79.2%) of HRS cells in nodular sclerosis (NS) HL were CD83 high or middle, and 85.7% were CD83 high or middle in mixed cellularity (MC) HL. Most (20/22, 90.9%) of stage I-II HL were CD83 high or middle, and 9/13 (69.2%) HL in stage III-IV were CD83 high or

Figure 1. CD83 is expressed on Hodgkin lymphoma cell lines. (A) Expression of CD83 was analyzed by flow cytometry on KM-H2, L428 and HDLM2 cell lines, which were stained with HB15a-fluorescein isothiocyanate (FITC), HB15e-FITC or 3C12C-FITC anti-CD83 mAbs, respectively. Gray histograms represent isotype control, while open histograms represent anti-CD83 antibodies. CD30 staining was used as a positive control. These data are representative of three independent experiments with comparable results. (B) CD83 expression (red) on KM-H2 cells with HB15a, HB15e or 3C12C mAb were imaged by confocal microscopy. Nuclei were stained with 4',6-diamidino-2-phenylindole (DAPI; blue). Human IgG1 was used as control for 3C12C mAb. Scale bar: 5Âľm.

haematologica | 2018; 103(4)

657

Z. Li et al.

middle. Interestingly, strong CD83 expression on HRS cells was found in two out of three relapsed HL (Online Supplementary Table S1). Epstein-Barr virus (EBV) infection is associated with an increasing risk of developing EBVpositive HL. A number of viral products, including EBV nuclear antigens (EBNA), EBV latent membrane proteins (LMP1 and LMP2) and EBV encoding small ribonucleic acids (RNA; EBER) have been implicated. LMP1 induced CD83 in EBV-infected human B cells by activation of NFkB.29 CD83/LMP1 has been reported to be correlated in MC HL, but not for NS HL.23 By in situ staining of EBER of 35 HL samples, we found that seven HL were EBER posi-

tive, including 2/7(28.6%) MC and 3/22 (13.6%) NS HL (Figure 2D). On six out of seven EBER positive HL samples, CD83 staining of HRS were strong or moderate (Online Supplementary Table S1).

CD83 is trogocytosed from HL cells to T cells We found previously that CD83 was able to transfer from the membrane of DC to T cells via trogocytosis.15 Similar trogocytosis was observed to occur between HL cell lines and T cells. When these two cell types were cocultured for four hours, CD83 surface expression was detected on 5-15% of T cells (Figure 3A,B), whereas no

Figure 2. CD83 is expressed on Hodgkin and Reed-Sternberg (HRS) cells in Hodgkin lymphoma patients. (A) CD83 and CD30 expression (brown) on paraffin-embedded lymph node biopsy samples of HL was imaged by microscopy with ×200 magnification. One representative sample of 35 biopsies shown. (B) Pie chart analysis of CD83 expression level in HRS cells of HL patients (n=35). High: CD83 positive in >90% HRS cells; middle: 10-90% CD83+in HRS cells; low: 10% CD83+ in HRS cells. One representative sample of each group is shown in (C), upper panel: original magnification ×40, lower panel shown with high amplification (×200). Arrows indicate HRS cells expressing CD83. (D) Epstein-Barr virus encoding small ribonucleic acids (RNA; EBER) in 35 HL biopsies were detected by in situ hybridization; one of the seven EBER positive samples is shown.

658

haematologica | 2018; 103(4)

CD83 and Hodgkin lymphoma

CD83 was detected on T cells in the absence of KM-H2 cells. Furthermore, separating the T and KM-H2 cells during culture by a 0.4mm transwell filter prevented trogocytosis (Online Supplementary Figure S2). To confirm the trogocytosis involved membrane transfer, KM-H2 cells were labeled with fluorescent dye (CellVue Claret) and co-cultured with CD3+ T cells. Cell membrane transfer from KM-H2 cells to T cells was confirmed by flow cytometry and confocal microscopy (Figure 3C,D, and Online Supplementary Figure S2). No differences were observed in the CD4+ and CD8+ T-cell ratio during the co-culture of

KM-H2 and T cells within four hours (data not shown). However, the CD83+ T cells expressed significantly higher levels of PD-1 than CD83– T cells (P=0.048) and T cells cultured without KM-H2 (P=0.005) (Figure 3E). The increase in PD-1 was significantly higher on the trogocytosed CD83+CD4+ T cells than non-trogocytosed CD83– T cells (P=0.049). In contrast, no difference in PD-1 expression was seen between the CD83+ and CD83– + C D 8 T cells, (P=0.185) although both KM-H2 co-cultured CD4+ and CD8+ T cells had higher PD-1 expression than

haematologica | 2018; 103(4)

Figure 3. Trogocytosis of CD83 molecule from Hodgkin and ReedSternberg cells to T cells. (A) T cells from healthy donor PBMCs were co-cultured with KM-H2 cells for four hours at a ratio of 1:5. CD83 and CD30 expression on CD3+T cells was analyzed by flow cytometry, data were from one of seven experiments and summarized data (mean± SEM and Pvalue) are shown in (B). (C) KM-H2 cells were labeled with CellVue Claret (red) and co-cultured with purified T cells (green) at a ratio of 5:1 for four hours. CellVue Claret and CD83 expression on T cells was analyzed by flow cytometry. (D) Confocal microscopy image of Claret labeled KMH2 cells co-cultured with T cells that stained with biotinylated mouse anti-human CD3 mAb and Stepdavidin-AF488. Nuclei were stained with DAPI. Scale bar: 5mm. Upper insert: trogocytosed T cells, lower insert: non-trogocytosed T cells. Data representative of three experiments. (E) PD-1 expression on CD83+ trogocytosed T cells co-cultured with KM-H2 cells for four hours was determined by flow cytometry (n=4). P-value of one-way ANOVA analysis shown. (F) PD-1 expression on trogocytosed CD4+T or CD8+ T cells after co-culture with KM-H2 cells for four hours was analyzed (n=4). P-value of one-way ANOVA analysis shown. A representative experiment shown in (G). FSC: forward scatter; SSC: side scatter; PD-1: programmed death1; FMO: fluorescence minus one.

659

Z. Li et al.

T cells cultured alone (Figure 3F,G). The CD83+CD4+ T cells had the same proportion of regulatory T cells (Treg) as non-trogocytosed CD4+ T cells (Online Supplementary Figure S2).

Supernatant from HL cell lines inhibits T-cell proliferation Surface CD83 can be cleaved into sCD83.15,20 We detected it in the supernatants of activated DC and B lympho-

cytes,20 as well as the serum of NHL and chronic lymphocytic leukemia patients.30 High levels of sCD83 were found in the supernatant of KM-H2 (460.6±11.8 pg/ml) and L428 (200.8±53.2 pg/ml), but low in HDLM2 (21.67±1.45 pg/ml) (Figure 4A). HL patients had significantly higher serum sCD83 (360.5±54.82 pg/ml, n=10) at diagnosis than healthy donors (52.6±9.5 pg/ml. Figure 4A). We then tested the effect of KM-H2 cell supernatant on T-cell function. KM-H2 supernatant containing sCD83

Figure 4. Soluble CD83 (sCD83) from Hodgkin lymphoma (HL) cell lines inhibits T-cell proliferation which is abolished by binding to 3C12C. (A) sCD83 was detected in the supernatant of KM-H2, L428, and HDLM2 lines that were cultured for 24 hours at 1x106/ml after changing fresh complete Roswell Park Memorial Institute (RPMI) medium and diagnostic sera of HL patients by ELISA. The P-value of A Mann-Whitney test is shown. (B) Carboxyfluorescein N-hydroxysuccinimidyl ester (CFSE) labeled purified T cells were stimulated with CD2/CD3/CD28 beads (3:1) in the presence of 25% supernatant (SN) of KM-H2 or plus 3C12C (5 mg/ml) for five days. Cells were analyzed by flow cytometry and the proliferation index (PI), that is defined as the total number of divisions divided by the number of cells that went into division, were calculated for total CD3+, CD4+ and CD8+ T cells using Flow Jo (n=6). The P-value of one-way ANOVA analysis is shown. (C) Different volumes (v/v) of KM-H2 supernatant were added to CD2/CD3/CD28 microbead-stimulated CFSE-labeled human T cells. T cells were collected and CFSE was analyzed by flow cytometry at day five. The PI and division index (DI), that is the average number of cell divisions that a cell in the original population has undergone, were calculated as indicators for proliferation. Representative data from one of three similar experiments shown. (D) CFSE-labeled T cells were stimulated with CD2/CD3/CD28 microbeads. T cells were then cultured in 25% (v/v) KM-H2 SN with 3C12C (5 and 10 mg/ml). T-cell proliferation was analyzed on day five. (E) The effect of different concentrations of 3C12C on proliferation of CFSE-labeled T cells was determined after CD2/CD3/CD28 microbead stimulation.

660

haematologica | 2018; 103(4)

CD83 and Hodgkin lymphoma

inhibited CD2/CD3/CD28 bead stimulated T-cell proliferation (Figure 4B) in a dose-dependent manner (Figure 4C). Only proliferation of CD8+ T cells seemed inhibited by KM-H2 supernatant (P=0.09), and not CD4+ T-cell proliferation (P=0.732). Administration of the anti-CD83 antibody, 3C12C, partially abolished the inhibitory effect of KM-H2 supernatant (Figure 4D). 3C12C alone had no effect on T-cell proliferation (Figure 4E).

HL patient serum sCD83 declined to normal levels correlated with a complete or partial response by PET-CT scan We monitored changes in circulating sCD83 in six HL patients during sequential chemotherapy. All assessments of response were made by positron emission tomography â&#x20AC;&#x201C; computed tomography (PET-CT) scan using the Lugano classification system. All patients received 3-6 cycles of chemotherapy; five achieved a complete response (CR) and one patient a partial response (PR) by PET-CT scan (Figure 5, Online Supplementary Table S2). Serum sCD83

decreased, returning to normal levels when the patients had a CR to chemotherapy, as documented by PET-CT scan in patients #1 and #2. In patients #3 and #6, the serum sCD83 level was still elevated when the PET-CT scan showed CR but normalized after one further cycle of chemotherapy. Patient #4 showed a PR prior to cycle 5 by PET-CT-scan, however the serum sCD83 level only started to decrease during cycle #5 reaching a normal range in cycle 6, coinciding with CR. PET-CT scans in patient #5 showed progressive disease (PD) after cycle 2, but a PR after another two cycles of chemotherapy, when the corresponding sCD83 reduced to normal level.

3C12C and 3C12C-MMAE kills HL cell lines The ADCC activity of the anti-CD83 mAb, 3C12C, was tested on the three HL lines: KM-H2, L428 and HDLM2. Whilst 3C12C killed KM-H2 and L428 efficiently, HDLM2 was relatively resistant to it (Figure 6A). To elucidate this, the stability of 3C12C binding on the HL cell surface were tested. HL lines were cultured in saturating concentration

Figure 5. Time course of soluble CD83 (sCD83) in Hodgkin lymphoma patients during chemotherapy. The sCD83 level in the sera of six HL patients during different cycles of chemotherapy was examined by ELISA. Arrows indicate when PETCT scans were performed and the results of complete response (CR), partial response (PR) or progressive disease (PD) are noted.

haematologica | 2018; 103(4)

661

Z. Li et al. of 3C12C (10 mg/ml) on ice followed by washing off unbound antibody. Cells were then cultured without 3C12C for up to two hours. The remaining 3C12C bound on the cell surface were detected by a secondary antihuman antibody. Though L428 and HDLM2 have a lower level of surface CD83 expression compared to KM-H2, our analysis showed that the 3C12C level on the surface of HDLM2 reduced much faster than on L428, while the 3C12C bound to L428 were far more stable (Figure 6B). This suggested 3C12C was rapidly internalized in KM-H2 and HDLM2, while 3C12C was internalized slower in L428. To investigate further potential therapeutic applications, we generated a 3C12C toxin-conjugate (3C12CMMAE). In vitro, 3C12C-MMAE killed CD83+ KM-H2 cells most efficiently, followed by HDLM2 and L428, while CD83– HL-60 cells were the least sensitive to 3C12C-MMAE (Figure 6C and Online Supplementary Figure S3). In addition, the intracellular CD83 level in HDLM2 was much higher than L428, lending to more sensitivity of the HDLM2 to the killing of 3C12C-MMAE (Online Supplementary Figure S1).

different staining pattern to CD30. Thus, CD83 is potentially another diagnostic marker of HL. More importantly, this work suggests that the majority of HL patients might be suitable for a therapeutic mAb targeting CD83. An anti-CD83 mAb may also work synergistically with chemotherapy, which is similar to the treatment of stage III or IV HL with anti-CD30 antibody drug conjugate

Administration of 3C12C is safe in non-human primates To “de-risk” the antibodies before advancing 3C12C into a clinical trial, we performed pre-clinical dose-escalation studies of 3C12C in non-human primates. Five baboons were injected intravenously with 3C12C (1, 5, 10 mg/kg on d0, 7, 14, and 21). No adverse clinical events were recorded during follow up for 84 days. We assessed blood counts and biochemistry weekly, and monitored different immune cell populations by flow cytometry or immune histology. Administration of 3C12C did not affect blood cell counts (white blood cells [WBC], red blood cells [RBC], and platelets), liver (ALP and AST) or kidney (creatinine) function (Online Supplementary Figure S4). The total T-cell number, and ratio of CD4+/CD8+ T cells all remained normal up to day 84 (data not shown). However, there was evidence of 3C12C efficacy in that CD1c+DC counts were reduced. We found that baboon blood B cells expressed CD83 as human B cells (data not shown), and reductions in blood B cells were noted by flow cytometry (Figure 7A). In addition, B-cell areas in lymph nodes were reduced in the 3C12C-treated animals (10mg/kg) compared to the control animals (human IgG 10mg/kg) (Figure 7B).

Discussion HL is driven by the malignant HRS cell, which are of B lineage origin.31,32 A significant number of patients experience relapsed/refractory disease following first-line chemotherapy.33 Less toxic treatments for relapsed/refractory HL would be highly desirable, as exemplified by the introduction of the anti-CD30 antibody drug conjugate (brentuximab).10,34 In the study herein, we were able to identify sCD83 as a new potential biomarker for HL, and CD83 as a target for a therapeutic mAb and derivatives. CD83 was first described on activated B cells and we originally detected CD83 on HL using frozen sections.19 Our ability to stain paraffin embedded lymph node biopsy samples of HL patients encouraged this study and allows for the assessment of CD83 expression in routine clinical practice. CD83 was highly expressed on HRS cells with a 662

Figure 6. 3C12C and 3C12C conjugation with monomethyl auristatin E (3C12C-MMAE) kill Hodgkin lymphoma (HL) cell lines in vitro. (A) Target cells KM-H2, L428 or HDLM2, labeled with Calcein-AM were co-cultured with effector cells (human PBMC) at effector: target ratio of 25:1 with increasing 3C12C concentration from 0 mg/ml to 1 mg/ml at 37°C for three hours. Supernatant was collected for fluorescence reading (excitation 485nm, emission 538nm) of released Calcein. Antibody (Ab)-dependent cell cytotoxicity was calculated (n=3). (B) HL cells were cultured in 3C12C saturation concentration (10 mg/ml) on ice followed by intensive washing and culture without 3C12C from 0-2 hours. The remaining levels of 3C12C bound on the cell surface were detected by a secondary anti-human antibody with flow cytometry. The remaining surface level of 3C12C on KM-H2, L428 and HDLM2 was normalized to the level of time 0. (n=3). (C) CD83+ KM-H2, L428, HDLM2 or CD83–HL-60 cells were cultured with different concentrations of 3C12C-MMAE for three days before determining viable cells by 7-amino-actinomycin D (7AAD) staining with flow cytometry. The half maximal inhibitory concentration (IC50) is shown. Data was from one of four representative experiments.

haematologica | 2018; 103(4)

CD83 and Hodgkin lymphoma

(brentuximab).10 As CD83 is inducible, it is possible that either drug induced or inflammatory activation would induce greater CD83 expression on HRS cells. We also identified serum sCD83 as a potential disease marker. Its immunosuppressive effect was reversed by anti-CD83 mAb at levels readily obtained in vivo. We predict that 3C12C would target HL cells directly through ADCC, but it has the additional therapeutic effect of reversing the inherent immunosuppressive effect of CD83. Such a synergistic response has the potential to have a significant clinical effect with limited toxicity. Recent studies revealed the impact of tumor microenvironment on tumor progression and therapy. HL is a leading example. The low frequency malignant HRS cells secrete several factors and generate a surrounding infiltrate of immune cells that contribute to the pathogenesis of the disease.35-37 CD83 appears to be involved in this process. We previously demonstrated that the transfer of membrane proteins on myeloma cells to T cells disrupted the immune response and was associated with poor prog-

nosis.38 We found that HL tumor cells express CD83 and can transfer surface CD83 molecules by trogocytosis. CD83 transfer from KM-H2 to T cells in vitro was consistent with the finding that some lymphocytes in the lymph node biopsy samples, especially in CD83 high expression patients, expressed CD83. The proportion of Treg in the trogocytosed CD83+CD4+ T cells was not increased, but CD83+ T cells, especially CD4+ T cells, expressed a higher level of PD-1 than CD83â&#x20AC;&#x201C; T cells. PD-1 and PD-1L interaction contributes to the immunosuppressive microenvironment of HL.39 Such PD-1high CD83+ T cells might become unresponsive in the tumor microenvironment.40 A CD83 target therapy might be combined with brentuximab and PD-1 blockage to enhance the clinical response. The serum of some hematopoietic malignancies have increased levels of sCD83.20,30 The supernatant of HL cells inhibited T-cell proliferation, but this inhibitory effect was not related to Treg induction (data not shown). The antiCD83 mAb, 3C12C, partially abolished the inhibition by KM-H2 supernatant. Thus, sCD83 from the supernatant

haematologica | 2018; 103(4)

Figure 7. 3C12C reduced B cells in non-human primates. Five nonhuman primates were injected with 3C12C (1, 5, 10, 10 mg/kg, n=4) or human Immunoglobulin G (IgG; 10mg/kg, n=1) at days 0, 7, 14 and 21. Blood and serum samples were collected for cell counts (red cells, white cells and platelets), liver function (ALP and AST levels) and kidney function (creatinine level) analysis. (A) CD19+ B cells were enumerated from PBMC of five animals by flow cytometry. Dashed lines indicate the base cell number at day 0. *indicates one time point when WBC was extremely high on that animal. (B) A lymph node biopsy was taken at day 28 from 3C12C (10mg/kg) and control-treated animals. B cells stained with antihuman CD20 mAb on paraffinembedded lymph node biopsy samples are shown. One of the two similar results for the two animals receiving 10 mg/kg 3C12C showing reduced B-cell areas compared to the human IgG control animal.

663

Z. Li et al.

plays a key inhibitory role on T cells. We also found that HL cell line supernatant inhibited normal B-cell proliferation (data not shown). Other cytokines or soluble factors from the KM-H2 supernatant may also contribute to the inhibitory effect, e.g., sCD30. An 85kDa soluble form of the CD30 molecule (sCD30) has been shown to be released by CD30+ cells in vitro and in vivo. sCD30 was elevated in the serum of HL41,42 and other CD30-expressing tumors, as well as inflammatory conditions with strong Tor B-cell activation. The CD30-Fc fusion protein inhibits T-cell proliferation,43 whilst sCD30 is also involved in the pathogenesis of renal, islet transplant rejection.44,45 The effect of brentuximab on sCD30 has not been investigated. The thymus and activation related chemokine (TARC; CCL17) is expressed by HRS cells.46 TARC is confirmed as a biomarker of HL, since elevated serum TARC reflected the disease activity and correlated with clinical response.47,48 Herein, as well as confirming elevation of serum sCD83 in active HL, we monitored sCD83 on six HL patients who underwent sequential cycles of chemotherapy. A complete response shown by PET-CT scan correlated with the decreased sCD83 levels. Thus sCD83 may be another biomarker candidate for monitoring the potential clinical response. A much larger cohort of HL patients will be monitored prospectively in order to explore this further. Additional investigations regarding the effect of sCD83 on HL biology may well assist therapeutic development in HL. Natural sCD83 has proved difficult to obtain for functional studies, suggesting the sCD83 structure and/or function is sensitive to in vitro manipulation.49 We have developed a human anti-human CD83 mAb, 3C12C, to investigate in clinical trials. It kills HL cells through ADCC, but in order to enhance its activity, we developed a 3C12C toxin conjugate (3C12C-MMAE).

References 1. Ansell SM. Hodgkin lymphoma: 2016 update on diagnosis, risk-stratification, and management. Am J Hematol. 2016;91(4):434-442. 2. Castagna L, Carlo-Stella C, Mazza R, Santoro A. Current role of autologous and allogeneic stem cell transplantation for relapsed and refractory hodgkin lymphoma. Mediterr J Hematol Infect Dis. 2015;7(1):e2015015. 3. Reddy NM, Perales MA. Stem cell transplantation in Hodgkin lymphoma. Hematol Oncol Clin North Am. 2014;28(6):10971112. 4. Janik JE, Morris JC, O'Mahony D, et al. 90Ydaclizumab, an anti-CD25 monoclonal antibody, provided responses in 50% of patients with relapsed Hodgkin's lymphoma. Proc natl Acad Sci USA. 2015; 112(42):1304513050. 5. Rehwald U, Schulz H, Reiser M, et al. Treatment of relapsed CD20+ Hodgkin lymphoma with the monoclonal antibody rituximab is effective and well tolerated: results of a phase 2 trial of the German Hodgkin Lymphoma Study Group. Blood. 2003;101(2):420-424. 6. Jacene H, Crandall J, Kasamon YL, et al.

664

8. 9.

10.

11.

12.

The unconjugated parent anti-CD30 antibody SGN-30 had no effect on HL in a study of 38 patients.50 Brentuximab vedotin SGN35, which is a drug conjugate of SGN-30 with MMAE, has, however, proven to be a highly promising drug with CD30+ lymphoma.10 We found 3C12C-MMAE kills CD83+ HL cell lines KM-H2 and HDLM2 very efficiently. Although HDLM2 cells express less surface CD83 and are resistant to ADCC killing, the high killing efficiency of HDLM2 with 3C12C-MMAE is likely related to the rapid antibody internalization and high intracellular CD83 turnover. 3C12C binding on L428 is relatively stable, rendering it sensitive to killing by ADCC, but less likely to be killed with 3C12C-MMAE, which is mediated via antibody internalization. Further improvements in the 3C12CMMAE conjugate preparation are planned. Finally, we tested the safety of 3C12C in non-human primates. We saw no clinical toxicity, abnormalities of blood count, liver or renal function or a decrease in the target CD1c+DC population (data not shown). By monitoring B cells, we saw depletion in the blood and lymph nodes. The depletion of activated CD83+ B cells led to a significant reduction in the B-cell area of lymph nodes. This early evidence of CD83 target cell depletion in non-human primates is most encouraging, suggesting an ADCC effect that should translate readily to the clinic. Taken together, these data demonstrate that CD83 is a new potential diagnostic HL marker and serum sCD83 levels are likely to reflect HL disease load. CD83 is a target for therapeutic mAb development as well as CD83 target derivatives. The potential therapeutic human anti-CD83 antibody, 3C12C, kills HL cells efficiently in vitro. It is safe in non-human primates, and depletes CD83+ target cells. Further development of 3C12C in human studies merits serious consideration.

Initial experience with tositumomab and I131-labeled tositumomab for treatment of relapsed/refractory Hodgkin lymphoma. Mol Imaging Biol. 2017;19(3):429-436. Younes A, Bartlett NL, Leonard JP, et al. Brentuximab vedotin (SGN-35) for relapsed CD30-positive lymphomas. N Engl J Med. 2010;363(19):1812-1821. Deng C, Pan B, O'Connor OA. Brentuximab vedotin. Clin Cancer Res. 2013;19(1):22-27. Gravanis I, Tzogani K, van Hennik P, et al. The European Medicines Agency review of brentuximab vedotin (Adcetris) for the treatment of adult patients with relapsed or fefractory CD30+ Hodgkin lymphoma or systemic anaplastic large cell lymphoma: summary of the scientific assessment of the Committee for Medicinal Products for Human Use. Oncologist 2016;21(1):102-109. Connors JM, Jurczak W, Straus DJ, et al. ' Brentuximab vedotin with chemotherapy for stage III or IV Hodgkin's lymphoma.. N Engl J Med. 2017 Dec 12. [Epub ahead of print] Villasboas JC, Ansell SM. Nivolumab for the treatment of classical Hodgkin lymphoma after failure of autologous stem cell transplant and brentuximab. Expert Rev Anticancer Ther. 2016;16(1):5-12. Ansell SM, Lesokhin AM, Borrello I, et al. PD-1 blockade with nivolumab in relapsed

13.

14.

15.

16.

17.

18.

or refractory Hodgkin's lymphoma. N Engl J Med. 2015;372(4):311-319. Kasamon YL, de Claro RA, Wang Y, Shen YL, Farrell AT, Pazdur R. FDA approval summary: nivolumab for the treatment of relapsed or progressive classical Hodgkin lymphoma. Oncologist. 2017; 22(5):585591. Zhou LJ, Tedder TF. Human blood dendritic cells selectively express CD83, a member of the immunoglobulin superfamily. J Immunol. 1995;154(8):3821-3835. Ju X, Silveira PA, Hsu WH, et al. The analysis of CD83 expression on human immune cells identifies a unique CD83+-activated T cell population. J Immunol. 2016; 197(12):4613-4625. Kretschmer B, Luthje K, Guse AH, et al. CD83 modulates B cell function in vitro: increased IL-10 and reduced Ig secretion by CD83Tg B cells. PloS One. 2007;2(8):e755. Cao W, Lee SH, Lu J. CD83 is preformed inside monocytes, macrophages and dendritic cells, but it is only stably expressed on activated dendritic cells. Biochem J. 2005;385(Pt 1):85-93. Yamashiro S, Wang JM, Yang D, Gong WH, Kamohara H, Yoshimura T. Expression of CCR6 and CD83 by cytokine-activated human neutrophils. Blood. 2000; 96(12):3958-3963.

haematologica | 2018; 103(4)

CD83 and Hodgkin lymphoma

19. Sorg UR, Morse TM, Patton WN, et al. Hodgkin's cells express CD83, a dendritic cell lineage associated antigen. Pathology. 1997;29(3):294-299. 20. Hock BD, Kato M, McKenzie JL, Hart DN. A soluble form of CD83 is released from activated dendritic cells and B lymphocytes, and is detectable in normal human sera. Int Immunol. 2001;13(7):959-967. 21. Dudziak D, Nimmerjahn F, Bornkamm GW, Laux G. Alternative splicing generates putative soluble CD83 proteins that inhibit T cell proliferation. J Immunol. 2005; 174(11):6672-6676. 22. Horvatinovich JM, Grogan EW, Norris M, et al. Soluble CD83 inhibits T cell activation by binding to the TLR4/MD-2 complex on CD14+ monocytes. J Immunol. 2017;198(6):2286-2301. 23. Doring C, Hansmann ML, Agostinelli C, et al. A novel immunohistochemical classifier to distinguish Hodgkin lymphoma from ALK anaplastic large cell lymphoma. Mod Pathol. 2014;27(10):1345-1354. 24. Wilson J, Cullup H, Lourie R, et al. Antibody to the dendritic cell surface activation antigen CD83 prevents acute graft-versus-host disease. J Exp Med. 2009;206(2):387-398. 25. Seldon TA, Pryor R, Palkova A, et al. Immunosuppressive human anti-CD83 monoclonal antibody depletion of activated dendritic cells in transplantation. Leukemia. 2016;30(3):692-700. 26. Sheets MD, Amersdorfer P, Finnern R, et al. Efficient construction of a large nonimmune phage antibody library: the production of high-affinity human single-chain antibodies to protein antigens. Proc Natl Acad Sci U S A. 1998;95(11):6157-6162. 27. Jones ML, Seldon T, Smede M, et al. A method for rapid, ligation-independent reformatting of recombinant monoclonal antibodies. J Immunol Methods. 2010; 354(1-2):85-90. 28. Francisco JA, Cerveny CG, Meyer DL, et al. cAC10-vcMMAE, an anti-CD30monomethyl auristatin E conjugate with potent and selective antitumor activity. Blood. 2003;102(4):1458-1465. 29. Dudziak D, Kieser A, Dirmeier U, et al. Latent membrane protein 1 of Epstein-Barr

haematologica | 2018; 103(4)

30.

31. 32. 33.

34.

35.

36.

37.

38.

39.

40.

virus induces CD83 by the NF-kappaB signaling pathway. J Virol. 2003;77(15):82908298. Hock BD, Haring LF, Steinkasserer A, Taylor KG, Patton WN, McKenzie JL. The soluble form of CD83 is present at elevated levels in a number of hematological malignancies. Leuk Res. 2004;28(3):237-241. Kuppers R. The biology of Hodgkin's lymphoma. NaR Rev Cancer. 2009;9(1):15-27. Scott DW, Gascoyne RD. The tumour microenvironment in B cell lymphomas. Nat Rev Cancer. 2014;14(8):517-534. Adams HJ, Nievelstein RA, Kwee TC. Systematic review and meta-analysis on the prognostic value of complete remission status at FDG-PET in Hodgkin lymphoma after completion of first-line therapy. Ann Hematol. 2016;95(1):1-9. Forero-Torres A, Holkova B, Goldschmidt J, et al. Phase 2 study of frontline brentuximab vedotin monotherapy in Hodgkin lymphoma patients aged 60 years and older. Blood. 2015;126(26):2798-2804. Aldinucci D, Gloghini A, Pinto A, De Filippi R, Carbone A. The classical Hodgkin's lymphoma microenvironment and its role in promoting tumour growth and immune escape. J Pathol. 2010;221(3):248-263. Ma Y, Visser L, Roelofsen H, et al. Proteomics analysis of Hodgkin lymphoma: identification of new players involved in the cross-talk between HRS cells and infiltrating lymphocytes. Blood. 2008;111(4):2339-2346. Vardhana S, Younes A. The immune microenvironment in Hodgkin lymphoma: T cells, B cells, and immune checkpoints. Haematologica. 2016;101(7):794-802. Brown R, Kabani K, Favaloro J, et al. CD86+ or HLA-G+ can be transferred via trogocytosis from myeloma cells to T cells and are associated with poor prognosis. Blood. 2012;120(10):2055-2063. Yamamoto R, Nishikori M, Kitawaki T, et al. PD-1-PD-1 ligand interaction contributes to immunosuppressive microenvironment of Hodgkin lymphoma. Blood. 2008;111(6):3220-3224. Jiang Y, Li Y, Zhu B. T-cell exhaustion in the tumor microenvironment. Cell Death Dis. 2015;6:e1792.

41. Gause A, Pohl C, Tschiersch A, et al. Clinical significance of soluble CD30 antigen in the sera of patients with untreated Hodgkin's disease. Blood. 1991;77(9):1983-1988. 42. Nadali G, Vinante F, Ambrosetti A, et al. Serum levels of soluble CD30 are elevated in the majority of untreated patients with Hodgkin's disease and correlate with clinical features and prognosis. J Clin Oncol. 1994;12(4):793-797. 43. Su CC, Chiu HH, Chang CC, Chen JC, Hsu SM. CD30 is involved in inhibition of T-cell proliferation by Hodgkin's Reed-Sternberg cells. Cancer Res. 2004;64(6):2148-2152. 44. Billing H, Sander A, Susal C, et al. Soluble CD30 and ELISA-detected human leukocyte antigen antibodies for the prediction of acute rejection in pediatric renal transplant recipients. Transpl Int. 2013; 26(3):331-338. 45. Saini D, Ramachandran S, Nataraju A, et al. Activated effector and memory T cells contribute to circulating sCD30: potential marker for islet allograft rejection. Am J Transplant. 2008;8(9):1798-1808. 46. van den Berg A, Visser L, Poppema S. High expression of the CC chemokine TARC in Reed-Sternberg cells. A possible explanation for the characteristic T-cell infiltratein Hodgkin's lymphoma. Am J Pathol. 1999; 154(6):1685-1691. 47. Weihrauch MR, Manzke O, Beyer M, et al. Elevated serum levels of CC thymus and activation-related chemokine (TARC) in primary Hodgkin's disease: potential for a prognostic factor. Cancer Res. 2005; 65(13):5516-5519. 48. Plattel WJ, van den Berg A, Visser L, et al. Plasma thymus and activation-regulated chemokine as an early response marker in classical Hodgkin's lymphoma. Haematologica. 2012;97(3):410-415. 49. Hock BD, Fernyhough LJ, Gough SM, Steinkasserer A, Cox AG, McKenzie JL. Release and clinical significance of soluble CD83 in chronic lymphocytic leukemia. Leuk Res. 2009;33(8):1089-1095. 50. Forero-Torres A, Leonard JP, Younes A, et al. A Phase II study of SGN-30 (anti-CD30 mAb) in Hodgkin lymphoma or systemic anaplastic large cell lymphoma. Br J Haematol. 2009;146(2):171-179.

665

ARTICLE

Non-Hodgkin Lymphoma

Ferrata Storti Foundation

Haematologica 2018 Volume 103(4):666-678

Gustav Arvidsson,1 Johan Henriksson,2 Birgitta Sander3 and Anthony P. Wright4 Department of Laboratory Medicine, Clinical Research Center, Karolinska Institutet; Department of Biosciences and Nutrition, Karolinska Institutet; 3Department of Laboratory Medicine, Division of Pathology, Karolinska Institutet and Karolinska University Hospital and 4Department of Laboratory Medicine, Clinical Research Center, Karolinska Institutet Stockholm, Sweden

1 2

ABSTRACT

Correspondence: anthony.wright@ki.se

Received: October 10, 2017. Accepted: February 9, 2018. Pre-published: February 15, 2018. doi:10.3324/haematol.2017.182048 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/666

subset of hematologic cancer patients is refractory to treatment or suffers relapse, due in part to minimal residual disease, whereby some cancer cells survive treatment. Cell-adhesion-mediated drug resistance is an important mechanism, whereby cancer cells receive survival signals via interaction with e.g. stromal cells. No genome-wide studies of in vitro systems have yet been performed to compare gene expression in different cell subsets within a co-culture and cells grown separately. Using RNA sequencing and species-specific read mapping, we compared transcript levels in human Jeko-1 mantle cell lymphoma cells stably adhered to mouse MS-5 stromal cells or in suspension within a co-culture or cultured separately as well as in stromal cells in co-culture or in separate culture. From 1050 differentially expressed transcripts in adherent mantle cell lymphoma cells, we identified 24 functional categories that together represent four main functional themes, anti-apoptosis, B-cell signaling, cell adhesion/migration and early mitosis. A comparison with previous mantle cell lymphoma and chronic lymphocytic leukemia studies, of gene expression differences between lymph node and blood, identified 116 genes that are differentially expressed in all three studies. From these genes, we suggest a core set of genes (CCL3, CCL4, DUSP4, ETV5, ICAM1, IL15RA, IL21R, IL4I1, MFSD2A, NFKB1, NFKBIE, SEMA7A, TMEM2) characteristic of cells undergoing cell-adhesion-mediated microenvironment signaling in mantle cell lymphoma/chronic lymphocytic leukemia. The model system developed and characterized here together with the core gene set will be useful for future studies of pathways that mediate increased cancer cell survival and drug resistance mechanisms. Introduction

ÂŠ2018 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.

666

Novel therapy regimes have improved the prognosis for many types of lymphoma and leukemia.1 However, unsolved problems remain for many patients who are refractory to treatment or experience disease relapse. For example, mantle cell lymphoma (MCL) is an aggressive and generally incurable B-cell neoplasm, comprising about 8% of non-Hodgkin lymphomas (NHL).2 MCL is characterized by a high relapse rate and frequent resistance to therapy, which together with high median age at diagnosis makes MCL difficult to cure. While median overall survival time has doubled in recent years,3 the prognosis for MCL patients is poor with fewer than 15% long-term survivors.4 Signaling pathway defects intrinsic to MCL cells, due to genetic aberrations, provide only a partial explanation for its aggressiveness and frequent relapse. These include the molecular hallmark translocation t(11;14)(q13;q32) CCND1/IGH,

haematologica | 2018; 103(4)

Global mRNA changes in stroma-bound lymphoma cells

which leads to cyclin D1 overexpression and cell cycle deregulation. Other frequent genetic aberrations are mutations in crucial DNA damage response genes, such as ATM and TP53. Relevant for the present work, however, recent studies show an important role for extrinsic, soluble and adhesion-mediated signals from the tumor microenvironment in supporting MCL trafficking, homing and susceptibility to therapy.5,6 Consistent with the important role of microenvironments, primary MCL and chronic lymphocytic leukemia (CLL) cells can only be cultured in vitro for a few days before they undergo spontaneous apoptosis.7,8 If co-cultured with mesenchymal stromal cells on the other hand, the in vitro cultures can be sustained for weeks.8,9 Furthermore, stromal cells of both human and murine origin can protect MCL and CLL cells from spontaneous and drug-induced apoptosis.8,10-12 While soluble molecules secreted by stromal cells such as BAFF8,13 and CXCL1214 have been shown to increase survival in malignant B cells, the protective effect is more prominent for lymphoma cells that physically adhere to stromal cells,6,8 and direct interactions between lymphoma cells and stromal cells can induce cell cycle arrest in MCL and diffuse large B-cell lymphoma (DLBCL).15 These mechanisms, involving soluble and adhesion-mediated signaling, may specifically confer survival advantages to lymphoma cells that home to protective microenvironmental niches through the activation of anti-apoptotic programs and downregulation of genes involved in proliferation.16 Targeted cell-culture studies have elucidated effects of microenvironment interactions in MCL and CLL. Increased levels of immunomodulatory cytokines, such as CCL3, CCL4, CCL22, IL-10 and TNF, with the capacity to alter microenvironment cellular composition have been reported in co-cultures of MCL or CLL cells with stromal cells or under other conditions that mimic microenvironment interactions.17-20 The adhesive properties of non-Hodgkin lymphoma (NHL) cells have been shown to increase upon treatment with anti-IgM, CXCL12 or CXCL13.17 The CXCR4 cytokine receptor protein, central to normal B-cell migration and homing, is down-regulated in adherent CLL cells.14,21 In co-culture and analogous studies, increased expression of anti-apoptotic proteins, such as BCL-XL and MCL-1, have been reported.11,22,23 Co-cultivation of MCL cells with stromal cells has also been reported to increase protein levels of the cell cycle inhibitors p21Cip1 and p27Kip1, along with an increased ratio of G0/G1 cells relative to Sphase cells.15 Many of these effects may be associated with an adhesion-related induction of both the canonical and non-canonical NF-κB pathways.8 While important signaling mechanisms relevant for cell adhesion-mediated survival of lymphoma cells have been revealed by targeted studies, the present work is the first systematic study of global changes in gene expression in a defined model system that allows discrimination of gene expression changes in the different cell types in the co-culture as well as their relationship to the same cells grown in isolation.

Methods Cell culture Cells were cultivated in a humidified incubator at 37°C and 5% CO2 in media supplemented with 100 U/mL penicillin and 100 haematologica | 2018; 103(4)

mg/mL streptomycin. The mouse stromal cell line MS-5 and the MCL cell line Jeko-1 were purchased from DSMZ and maintained in αMEM-glutamax (Gibco) supplemented with 10% heat-inactivated fetal bovine serum (H.I. FBS; Gibco) and 2 mM sodium pyruvate or RPMI-glutamax (Gibco) supplemented with 10% HI FBS, respectively. Co-cultures of Jeko-1 with MS-5 at a 10:1 ratio were maintained under the same conditions as for MS-5 cells alone.

Cell-cell binding assay Unlabeled Jeko-1 suspension cells were added to established MS-5 monolayers. After 24 h, unlabeled Jeko-1 cells in suspension were removed and replaced with an equivalent number of CFDASE labeled Jeko-1 cells. Adhered unlabeled/labeled Jeko-1 cells were counted at 24 h and 48 h. The order of addition of labeled/unlabeled Jeko-1 cells was subsequently reversed.

RNA extraction, library preparation and sequencing Total RNA was extracted using RNeasy with QIAshredders (Qiagen). Libraries were prepared using TruSeq sample prep kit v.2.0 and included a poly-A enrichment step. Samples were 16-plexed on an Illumina HighSeq 2500 instrument generating 230,700,000 2x101bp short reads (Online Supplementary Table S1).

Species-based read separation and mapping to reference genomes Reference genomes hg19 and mm10 were obtained from UCSC. Raw reads were separated based on species origin using Xenome (v.1.0.1).24 Separated reads of human and murine origin were aligned to reference genomes hg19 and mm10, respectively, using Tophat2 (2.0.11)25 using default options. Fragments per feature were counted using summarizeOverlaps from the Bioconductor package GenomicAlignments (1.0.6) with counting mode set to “Union”. Gene annotations and coordinates were from the Bioconductor packages TxDb.Hsapiens.UCSC.hg19.knownGene (v.2.14.0) and TxDb.Mmusculus.UCSC.mm10.knownGene (v.2.14.0).

Differential gene expression analysis and GSEA Differential expression was determined by DESeq (1.16.0).26 Normalized count tables were used for Gene Set Enrichment Analysis (GSEA) (Broad Institute, 2.2.0)27 using canonical pathways (c2.cp.v5.0.symbols.gmt) and GO processes (c5.bp.v5.0.symbols.gmt), from the MsigDB.27 Functional clusters from the leading edge analysis were identified as ngenes≥10.

Protein interaction network analysis A unified list with unique gene identifiers based on differentially expressed genes (FDR q-value≤0.05) and leading edge genes from the GSEA (n=1458) was uploaded to the STRING interaction database (v.10.0) with the confidence level set to 0.4.28 Node genes were defined as those having 8 or more interactions.

Microarray analysis Microarray datasets were downloaded from GEO:29 GSE2102930 and GSE70910.31 Overlaps between these datasets and adhesionrelated genes from the present study were interrogated by twosided Fisher’s exact tests.

Data availability Raw RNA-seq data are available via the gene expression omnibus (GEO) repository29 by accession number: GSE99501. A detailed account of materials and methods used is available in the Online Supplementary Methods. 667

A. Arvidsson et al.

Results

incubation for a further 24 h did not significantly displace the unlabeled bound cells (Figure 1A). Interestingly, the labeled MCL cells were able to bind to stromal cells independently of the previously bound MCL cells. Similar results were obtained when the order of addition of labeled and unlabeled cells was reversed (Figure 1A), indicating that the CFDA-SE label does not significantly affect the adherence characteristics of Jeko-1 cells in this assay. Thus, the co-cultured MCL cells could be divided into two relatively stable subsets (adherent and suspension), and therefore it was of interest to characterize differences in

MCL cells adhere stably to stromal cells Approximately 10% of MCL cells (Jeko-1), which normally grow in suspension, adhered to a mono-layer of the bone marrow-derived, adherent stromal cell line (MS-5) upon co-culture and remained in place when MCL cells remaining in suspension were poured away after 24 h (Figure 1A). The interaction with stromal cells was stable because addition of a 10-fold excess of CFDA-SE labeled MCL cells to the stromal cells with adhered MCL cells and

668

Figure 1. Identification of differentially regulated genes in adherent co-culture cells. (A) A fraction of co-cultured Jeko-1 cells adhere to an MS-5 cell mono-layer and remain stably bound in the presence of excess suspension cells. The number of unlabeled (light gray) or CFDA-SE labeled (dark gray) Jeko1 cells adhered to MS-5 stromal cells after 24 hours (h) was not significantly reduced after a further 24 h of incubation (48 h) with excess CFDA-SE (P=0.841) or unlabeled (P=0.391) Jeko-1 cells, respectively. Mean±Standard Deviation are shown for cultures (n=3). (B) Experimental design showing the three Jeko-1 and the two MS-5 stromal cell fractions for which RNAseq data were acquired. (C) Heatmap representation of relative transcript levels for the 3697 genes with significantly changed mRNA levels (FDR q-value ≤ 0.05) in at least one Jeko-1 cell fraction after hierarchical clustering. (D) Venn diagram illustrating number of differentially expressed genes in pair-wise comparisons between the three Jeko-1 cell fractions (FDR q-value ≤ 0.05, total number of changed genes: 3697). The three comparisons in the Venn diagram are: adherent Jeko-1 cells in co-culture (ADH) compared with suspension Jeko-1 cells in co-culture (SUSP), n=1050 (ADH/SUSP), ADH compared to mono-cultured Jeko-1 cells (SEP), n=3453 (ADH/SEP) SUSP compared to SEP n=1471 (SUSP/SEP). (E) Ranked fold changes of 1050 genes with changed transcript levels in ADH relative to SUSP Jeko-1 cells in co-culture.

haematologica | 2018; 103(4)

Global mRNA changes in stroma-bound lymphoma cells

Table 1. Differential transcript levels in adherent compared to suspension mantle cell lymphoma (MCL) cells in co-culture.

Symbol TPM SEP TPM SUSP TPM ADH FC Higher transcript levels in adherent Jeko-1 cells:

FDR q

TNF EGR1 FOS EGR3 CCL3 MIR146A EGR2 CCL4 MYCN NR4A2 CSF1 PHLDA1 LTA PLK2 MAFF IL10 SLAMF1 RIN2 CCL4L2 CD69 SEMA7A TNFRSF9 DUSP2 GPR3 STX11 FOSL1 HES1 IL4I1 ETV5 NR4A3 IER3 XIRP1 TNFAIP3 PPP1R15A RGS3 DUSP4 GADD45B NFKBIA CD83 ICAM1 HBEGF KCNK5 BCL2A1 NR4A1 ZC3H12C IL21R LRRC32 KLF10 SPRY2 GEM

7.6E-46 4.1E-23 7.8E-13 3.6E-20 1.4E-26 9.9E-03 5.0E-13 1.2E-33 1.3E-03 4.5E-16 1.1E-34 1.5E-11 2.9E-09 2.5E-05 1.1E-03 3.8E-11 3.9E-02 1.2E-03 1.7E-08 1.4E-05 2.3E-47 5.8E-05 5.8E-50 5.6E-03 7.8E-06 3.5E-03 2.8E-08 5.9E-09 1.5E-03 8.3E-22 1.7E-13 3.9E-02 2.0E-09 4.0E-06 4.1E-20 1.1E-02 5.8E-32 9.9E-11 2.0E-15 8.2E-36 3.3E-07 1.2E-02 4.1E-17 2.1E-41 1.6E-13 7.9E-14 1.4E-26 1.8E-06 3.6E-11 1.7E-08

0.6 4.3 0.8 0.2 10.2 3.9 9.4 17.9 0.3 1.3 3.1 0.1 0.3 0.4 0.1 0.8 0.2 0.2 4.4 17 6.7 0 75.2 0.6 0.1 0.4 0.6 0.2 0.2 11.5 8 0.1 13.3 14.8 0.8 0 14.6 97.2 98.7 16.2 5 0.1 6.9 27.9 2.9 0.7 0.6 13.1 2.9 1.7

1.7 10.7 1.8 5.2 36.2 5.1 14.7 70.5 0.3 3.8 5.1 1.1 0.6 0.5 0.5 2.2 0.2 0.5 14.4 23.3 20 0.4 123.1 0.7 0.6 1.1 2.5 1.1 0.5 53.1 8.1 0.2 20.1 17.6 1.9 0.3 28.7 121 292.2 34.1 2.9 0.6 59.5 57.9 6.4 4.3 11.7 17 8.2 4.6

39.7 178 25.5 66.4 424.2 52.9 135.3 596.9 2.5 30.1 37.9 8.1 4.3 3.5 3.4 14.9 1.3 3.1 84 133.1 110.8 2.1 637.7 3.5 2.9 5.3 11.8 5.1 2.3 241.2 36.7 0.9 90.1 75.6 7.7 1.2 114.7 468.2 1128.8 128.5 10.9 2.2 216.9 207.1 22.8 15.3 41 59.5 28 15.5

23.35 16.64 14.17 12.77 11.72 10.37 9.20 8.47 8.33 7.92 7.43 7.36 7.17 7.00 6.80 6.77 6.50 6.20 5.83 5.71 5.54 5.25 5.18 5.00 4.83 4.82 4.72 4.64 4.60 4.54 4.53 4.50 4.48 4.30 4.05 4.00 4.00 3.87 3.86 3.77 3.76 3.67 3.65 3.58 3.56 3.56 3.50 3.50 3.41 3.37

Symbol TPM SEP TPM SUSP TPM ADH Lower transcript levels in adherent Jeko-1 cells: RAG1 GPR56 SMAD6 CCDC42B NEIL1 C10orf71 NYNRIN OSR2 BMP3 WDR66 PCDH9 PIF1 CXCR4 GADD45A AICDA PSRC1 FBXO32 LINC01089 AOX2P KCNA3 ABCA1 RASSF6 TP53INP1 ASPM RNF144B SORT1 SOX4 WNT5A PROX1 SMAD1 ZNF850 LOC730101 LOC440173 FOXN4 SPPL2B TRIM52 HUNK NUAK2 KIF20A MIAT MYLIP KLHL24 CDC42BPB

42.6 14.3 4.1 14.5 6.5 22.8 5.2 10.3 33.9 9.1 96 30.9 901.8 38.9 31.3 29.9 3.7 70.2 7.8 24.8 36.9 208.1 10.8 134.5 93.1 24 68.3 34.8 13.3 102.7 4.9 14.9 18.9 15.6 30.7 23.9 8.4 14.2 102.3 16.7 23.3 22.3 49.8

30.3 6.3 4.1 11.2 7.3 14 3.9 8.8 18.7 5.7 53.6 28.8 744.3 34.1 34.6 25 3.1 67.5 4.5 15.2 33 130.5 6.9 109.9 56.8 17.5 53.8 29.2 13.4 70.2 4.2 9.3 16.2 13 27.5 25.1 5.9 12.6 96.3 10.6 28.4 14.1 35.8

5.7 1.7 1.4 4.4 2.9 5.7 1.6 3.7 8 2.5 24.5 13.2 344.4 16 16.3 11.9 1.5 33 2.2 7.5 16.3 65.2 3.5 56.4 29.5 9.2 28.6 15.9 7.3 38.3 2.3 5.1 8.9 7.2 15.3 14 3.3 7.1 54.4 6 16.1 8 20.4

FDR q

-5.32 -3.71 -2.93 -2.55 -2.52 -2.46 -2.44 -2.38 -2.34 -2.28 -2.19 -2.18 -2.16 -2.13 -2.12 -2.10 -2.07 -2.05 -2.05 -2.03 -2.02 -2.00 -1.97 -1.95 -1.93 -1.90 -1.88 -1.84 -1.84 -1.83 -1.83 -1.82 -1.82 -1.81 -1.80 -1.79 -1.79 -1.77 -1.77 -1.77 -1.76 -1.76 -1.75

3.5E-74 4.7E-07 2.3E-02 1.4E-02 7.6E-03 1.3E-09 2.9E-03 8.7E-03 1.8E-12 7.3E-03 2.5E-25 4.5E-10 4.9E-72 1.6E-04 4.0E-09 2.6E-05 4.9E-02 1.2E-11 4.0E-02 2.1E-04 1.3E-24 4.1E-41 3.3E-03 1.1E-28 5.3E-10 1.1E-08 1.3E-15 7.3E-11 1.3E-03 2.0E-13 4.2E-02 1.8E-02 3.3E-02 6.1E-03 3.0E-08 2.0E-03 1.5E-02 4.2E-03 1.5E-15 8.1E-06 5.5E-05 3.0E-06 9.1E-12

The most up- and down-regulated genes with an absolute fold change â&#x2030;¥1.75 (max. 50) in the adherent MCL fraction (ADH) as compared to MCL cells in suspension (SUSP) in co-culture with stromal cells. Transcript levels are shown as transcripts per million reads (TPM) for ADH and SUSP as well as for MCL cells grown in mono-culture (SEP), where read counts per gene have been normalized to library size and gene length. Fold change (FC) for comparison of ADH relative to SUSP presented with the associated P-value adjusted for multiple testing by false discovery rate (FDR q-value).

haematologica | 2018; 103(4)

669

A. Arvidsson et al.

gene expression between these subsets and in relation to mono-cultured cells.

Adhesion to stromal cells affects gene expression in MCL cells Physical separation of adherent MCL cells from stromal cells in co-cultures was avoided by direct RNA isolation from the mixed cell population in order to minimize separation-associated artifacts in the RNAseq data. The mixed-species short RNAseq reads were separated in silico by species-specific read separation using reference genomes for human (MCL cells) and mouse (stromal cells). Read counts per sample prior to and following speciesspecific read separation are presented in Online Supplementary Table S1. The experimental design allowed for global transcript level comparisons between three distinct MCL fractions: mono-cultured MCL cells (SEP), adherent MCL cells in coculture (ADH), and suspension MCL cells in co-culture (SUSP) (Figure 1B). Pairwise comparisons of transcript levels for the three MCL fractions identified a total 3697 genes that are differentially expressed in at least one comparison after 24 h mono-/co-culture (FDR q-value â&#x2030;¤0.05). The heat map in Figure 1C shows relative transcript levels across the three cell populations. Overall, the largest difference is between the ADH fraction and the suspension cells in mono-culture (SEP) and co-culture (SUSP), which while distinct, are more similar to each other (column dendrogram, Figure 1C). The Venn diagram sets in Figure 1D show the number of genes that are differentially expressed between each pair of MCL cell populations: 3453, 1471 and 1050 genes were differentially expressed in the respective comparisons of ADH relative to SEP (ADH/SEP), SUSP relative to SEP (SUSP/SEP) and ADH relative to SUSP (ADH/SUSP). The three comparison groups of differentially regulated genes are highly overlapping with half the genes occurring in two or more groups (Figure 1D). Given the previously reported survival advantage conferred to lymphoma cells adhering to stromal cells in co-culture,6,8 and as this comparison provides the best opportunity for specifically understanding molecular aspects associated with adhesion to stromal cells, we focused the analysis on the 1050 genes with altered transcript level between the two MCL cell fractions within the co-culture system (ADH/SUSP). Altogether, 137 of these genes were changed more than 2fold at the transcript level (Figure 1E). Significantly changed genes in the ADH fraction with the highest fold change are presented in Table 1 (Complete lists of significantly changed genes are available in Online Supplementary Tables S2-S4).

Gene expression and co-culture dependent changes in stromal cells The study design did not allow for an in-depth comparison of transcript levels in stromal cells analogous to the analysis of cancer cells above. Thus a detailed analysis of differentially expressed stromal cell genes is not reported here. Nonetheless, 100 genes were significantly changed at the transcript level between mono- and cocultured stromal cells (Figure 2). These include the chemotactic molecules Ccl2 and Ccl7, both with higher transcript levels in the co-cultured MS-5 cells. These can interact with the CCR2 cytokine receptor, which is expressed in Jeko-1 MCL cells. A complete list with sig670

Figure 2. Significant transcript level changes between mono- and co-cultured stromal cells. (A) Heatmap representation of the 100 genes with significant transcript level changes between co-cultured (COCULT) and mono-cultured (SEP) and MS-5 stromal cells (FDR q-value â&#x2030;¤0.05) after hierarchical clustering. (B) Ranked fold changes for the 100 genes with altered transcript levels between COCULT and SEP MS-5 stromal cell fractions.

haematologica | 2018; 103(4)

Global mRNA changes in stroma-bound lymphoma cells

Table 2. Functional clusters of genes distinguishing stromal-cell attached MCL cells from mantle cell lymphoma (MCL) suspension cells in coculture.

C Description Functional categories Genes Functional clusters for which gene transcripts are more abundant in adherent MCL cells compared to suspension MCL cells in the same co-culture c1

CD40 and NF-κB signaling

I kappaB kinase kappaB cascade, Positive regulation of I KappaB kinase NF kappaB cascade, positive regulation of signal transduction, Protein kinase cascade, regulation of I kappaB kinase NF kappaB cascade

BST2, BUD31, CARD9, CASP1, CD40, CDKN1C, FAF1, HMOX1, IKBKE, LGALS1, LGALS9, LITAF, MAP3K6, MIER1, MKNK1, MKNK2, MUL1, NDFIP2, NEK6, NMI, NR4A3, OTUD7B, PLCE1, PRDX4, REL, RHOC, RIPK2, SIK1, SLC20A1, SLC35B2, SQSTM1, STK17B, STK38L, TAB2, TBK1, TICAM1, TLR6, TMEM9B, TRAF2, TRIP6, VAPA

c2 Toll-like receptor and Reactome activated TLR4 signaling, Reactome innate immune system, MAP kinase signaling Reactome map kinase activation in TLR cascade, Reactome MAPK targets nuclear events mediated by map kinases, Reactome MYD88 mal cascade initiated on plasma membrane, Reactome NFκB and MAP kinases activation mediated by TLR4 signaling repertoire, Reactome nuclear events kinase and transcription factor activation, Reactome toll receptor cascades, Reactome TRAF6 mediated induction of NFκB and MAP kinases upon TLR7 8 or 9 activation, Reactome TRIF mediated TLR3 signaling

APP, ATF1, CD180, CNPY3, DUSP3, DUSP7, IRAK4, JUN, MAP2K3, MAP2K4, MAP2K6, MAPK10, MAPK11, MAPK7, MAPKAPK2, MEF2C, PELI1, TLR1, TLR10, TLR2, TLR4, TLR7

Respiration

Reactome respiratory electron transport ATP synthesis by chemiosmotic coupling and heat production by uncoupling proteins, Reactome TCA cycle and respiratory electron transport

ATP5B, ATP5C1, ATP5D, ATP5F1, ATP5G1, ATP5J, BSG, COX4I1, COX5A, COX5B, COX6B1, COX6C, COX7A2L, COX7B, COX7C, COX8A, CYC1, CYCS, DLD, ETFA, ETFB, ETFDH, FH, IDH3A, IDH3B, IDH3G, L2HGDH, LDHA, LDHB, MDH2, NDUFA1, NDUFA12, NDUFA2, NDUFA3, NDUFA5, NDUFA8, NDUFA9, NDUFAB1, NDUFB10, NDUFB2, NDUFB3, NDUFB5, NDUFB7, NDUFB9, NDUFS3, NDUFS4, NDUFS5, NDUFS6, NDUFS8, NDUFV2, OGDH, PDHA1, PDHX, PDK1, PDP1, PDP2, SDHB, SDHD, SLC16A1, SLC16A3, SUCLA2, SUCLG1, UQCR11, UQCRB, UQCRC2, UQCRFS1, UQCRH, UQCRQ

Lymphocyte activation

KEGG allograft rejection, KEGG autoimmune thyroid disease, KEGG cell adhesion molecules cams, KEGG graft versus host disease, KEGG type I diabetes mellitus, KEGG viral myocarditis, PID IL12 STAT4 pathway KEGG viral myocarditis signal, PID IL12 STAT4 pathway

CAV1, CD28, CD80, CD86, HLA-DOB, HLA-DQA2, HLA-DRA, HLA-DRB1, HLA-E, MYH15, PRF1

BCL2 family Anti apoptosis, Apoptosis GO, Negative regulation of apoptosis, and anti-apoptosis Negative regulation of developmental process, Negative regulation of programmed cell death, Programmed cell death

AATF, ACVR1, AMIGO2, ANXA5, BAG1, BCL2A1, BCL2L1, BCL3, BIK, BNIP1, BRE, CASP7, CD70, CD74, CDK5R1, CDKN2D, COL4A2, DAD1, DDAH2, DUSP22, FAIM3, GADD45B, GPX1, IER3, IL24, MCL1, MRPS30, NOL3, NOTCH4, PEA15, PIM1, PLAGL1, PMAIP1, PPP1R13B, PPP1R15A, PSEN2, RUNX3, SERPINB9, SIRT2, STK17A, TNFAIP8, TNFSF9, TRIAP1, TSPO, UTP11L, XIAP

Extracellular KEGG cytokine cytokine receptor interaction, matrix - Matrisome NABA matrisome associated, NABA matrisome, NABA secreted factors

ADAM8, ANGPTL6, ANXA6, CCL4L2, CLEC17A, COL9A2, CRELD2, CRIM1, CSF1, CST7, CSTB, CTSH, CTSZ, EDA, EMILIN2, HBEGF, HYAL2, IGFBP4, INHBE, KAZALD1, P4HA1, PDGFA, PLOD3, PLXNA1, S100A11, S100A4, S100A6, SDC3, SDC4, SEMA7A, SERPINE2, SRGN, THPO, VEGFA, VWA5A, WNT10A, ZP3

B-cell receptor/ Cell activation, Immune response, Immune system process, cytokine signaling Leukocyte activation, Lymphocyte activation, Positive regulation of immune system process, Positive regulation of multicellular organismal process, Regulation of immune system process, Regulation of lymphocyte activation, Regulation of multicellular organismal process, T-cell activation

AIMP1, CD164, CD22, CD79A, CD83, CD97, DYRK3, ELF4, FTH1, GEM, GPR65, HDAC9, HELLS, ICOSLG, IL32, IL4R, LAT2, LAX1, LRMP, LY86, MS4A1, NFIL3, NLRC3, PDCD1, PRELID1, PREX1, PTGER4, SIT1, SP2, TAPBP, TGFB1, TNFRSF14, XBP1

haematologica | 2018; 103(4)

671

A. Arvidsson et al. c8

Cell adhesion and migration VAV2,VAV3 c9 Vacuolar ATPase and acidification of cell compartments

KEGG focal adhesion, KEGG leukocyte transendothelial migration, Reactome integrin cell surface interactions

ACTB, ACTG1, ARHGAP5, BCAR1, CTNNA2, MYL12A, PTK2B, RAC2, RAC3, RAP1A, RASSF5, VASP,

KEGG epithelial cell signaling in Helicobacter pylori infection, Reactome iron uptake and transport, Reactome latent infection of homo sapiens with mycobacterium tuberculosis

ATP6V0A1, ATP6V0B, ATP6V0E1, ATP6V1C1, ATP6V1C2, ATP6V1D, ATP6V1E1, FTL, HMOX2, TFRC

c10 Cell-cell signaling

Cell cell signaling

CHRNB1, GCH1, GRIK1, HOMER1, HPRT1, HTR2C, KCNMB2, KLF10, MPZ, NMB, POMC, SRI, SYPL1, ZYX

c11 Cholesterol and nuclear receptors

Reactome cholesterol biosynthesis, Reactome nuclear receptor transcription pathway

CYP51A1, DHCR7, FDFT1, HMGCR, HMGCS1, HSD17B7, IDI1, IDI2, MSMO1, MVD, MVK, NR4A2, RARA, RARG, SQLE

c12 Proteasome

Biocarta proteasome pathway, KEGG proteasome, Reactome activation of NF kappaB in B cells, Reactome antigen processing cross presentation, Reactome CDK mediated phosphorylation and removal of CDC6, Reactome cyclin E associated events during G1 S transition, Reactome destabilization of mRNA by AUF1 HNRNP D0, Reactome downstream signaling events of B-cell receptor BCR, Reactome ER phagosome pathway, Reactome host interactions of HIV factors, Reactome regulation of apoptosis, Reactome regulation of mRNA stability by proteins that bind AU rich elements, Reactome regulation of ornithine decarboxylase ODC, Reactome signaling by the B-cell receptor BCR, Reactome VIF mediated degradation of APOBEC3G

AKT3, AP2S1, ARF1, ATP6V1H, BAD, BANF1, CASP9, ELMO1, FBXW11, GRB2, HCK, IKBKB, KPNA1, MLST8, NPM1, NR4A1, NRAS, NUPL1, PACS1, PPIA, PRKCB, PSMA1, PSMA4, PSMA7, PSMA8, PSMB1, PSMB10, PSMB2, PSMB3, PSMB4, PSMB5, PSMB6, PSMB8, PSMB9, PSMC3, PSMC4, PSMC5, PSMD11, PSMD12, PSMD13, PSMD14, PSMD3, PSMD4, PSMD7, PSMD8, PSMD9, RAN, RASGRP1, RASGRP3, RELA, RPS27A, SHC1, SKP1, SLC25A5, TCEB1, TCEB2, THEM4, TRIB3

c13 Endoplasmic reticulum stress and exosome

Reactome PERK regulated gene expression, Reactome activation of genes by ATF4

ASNS, ATF3, ATF4, ATF6, DDIT3, EIF2S1, EXOSC1, EXOSC3, EXOSC4, EXOSC7, EXOSC9, HERPUD1, NFYA

c14 MHC/ antigen presentation

KEGG antigen processing and presentation

B2M, CANX, CTSB, HLA-A, HLA-C, HLA-DMA, HLA-DMB, HLA-DPA1, HLA-DRB4, HSPA4, PSME3, RFX5 ACTR3, ARHGAP6, ARPC1B, ARPC2, ARPC3, ARPC5, CFL1, DIAPH1, GSN, PIP5K1A, PIP5K1B

c15 Actin cytoskeleton Biocarta Rho pathway remodeling c16 Cytoskeleton Reactome response to elevated platelet cytosolic CA2, remodeling/ vesicle Reactome platelet activation signaling and aggregation docking

ABCC4, CD63, DGKG, DGKH, HSPA5, PFN1, PLEK, STX4, TMSB4X, TUBA4A, WDR1

c17 Cytokine signaling and nucleus cytoplasm transport

CD44, CIITA, CISH, EIF4E, EIF4G2, FYN, IFI35, IFITM1, IFITM2, IFNGR1, IFNGR2, IL2RA, IL2RG, IRF2, IRF4, JAK2, JAK3, KPNA4, LCK, LYN, MAP3K8, MX1, NUP153, NUP35, NUP37, NUP54, OAS3, PTPN6, SEH1L, STAT1, STAT3, STAT5A, STAT5B, TNIP2, TOL LIP, UBE2L6, USP18, VAV1, YES1

Reactome cytokine signaling in immune system, Reactome IL 3 5 and GM CSF signaling, Reactome interferon alpha beta signaling, Reactome interferon gamma signaling, Reactome interferon signaling, Reactome signaling by ILS

Functional clusters for which gene transcripts are less abundant in adherent MCL cells compared to suspension MCL cells in the same co-culture c18 Histones

Reactome meiosis, Reactome meiotic recombination, Reactome amyloids, Reactome RNA POL I promoter opening, Reactome deposition of new CENPA containing nucleosomes at the centromere

DIDO1, DMC1, HIST1H2AC, HIST1H2BD, HIST1H2BF, HIST1H2BG, HIST1H2BO, HIST1H3B, HIST1H3D, HIST1H3E, HIST1H3G, HIST1H3H, HIST1H4J, HIST1H4K, HIST2H2AA4, HIST2H2BE, HIST2H4A, HIST4H4, MSH5, SUN2, SYNE1, SYNE2

c19 Cell cycle M phase

M phase, Mitosis, M phase of mitotic cell cycle, PID PLK1 pathway

ANLN, ATM, AURKA, BIRC5, BRSK1, BUB1, BUB1B, CDC25B, CDC25C, CENPE, CIT, DLGAP5, EGF, ESPL1, KIF11, KNTC1, NEK2, NUMA1, PLK1, RAD52, TAF1L, TPX2, TTK, XRCC2

PID delta NP63 pathway

ADA, BDKRB2, BRCA2, CCNB2, FASN, HBP1, MRE11A, NRG1, RAB38, STXBP4, TOP2A

c20 Î&#x201D;p63 signaling pathway

672

haematologica | 2018; 103(4)

Global mRNA changes in stroma-bound lymphoma cells

c21 Chromosome Regulation of small GTPase mediated signal transduction, organization, Chromosome organization and biogenesis, KEGG Rho/ Rac family hedgehog signaling pathway GTPase signaling and WNT signaling

ACIN1, ARAP1, ARHGAP27, ARID1A, BMP7, BNIP3, BPTF, CDC42BPB, CREBBP, DDX11, DFFB, DMPK, EHMT1, ERCC4, FGD4, FGD6, GAS1, HDAC10, HDAC6, HUWE1, KAT2A, KAT6A, KAT6B, KDM4A, MAP3K12, MRE11A, NOTCH2, NSD1, PDS5B, PIF1, PRKACB, PRMT7, RALBP1, SATB1, SMARCC2, STK36, TAF6L, TEP1, TERT, TOP2A, TSC1, WNT16, WNT5A, WNT6

c22 Kinetochore Reactome mitotic prometaphase assembly/ function

CASC5, CDCA8, CENPA, CENPI, CENPT, CKAP5, CLASP1, NDC80, NUF2, RANBP2, SGOL2, SKA2, TAOK1

c23 Mitotic centrosome

Reactome recruitment of mitotic centrosome proteins and complexes

AKAP9, ALMS1, CDK5RAP2, CENPJ, CEP135, CEP164, CEP192, CEP70, CETN2, CKAP5, CLASP1, CNTRL, CSNK1E, DCTN1, DYNC1H1, DYNC1I2, OFD1, PCM1, PCNT, TUBB4A, TUBGCP6

c24 Kinesins

Reactome kinesins

KIF15, KIF18A, KIF20A, KIF23, KIF2C, KIF3C, KIF4A, KIF4B, KIF5A, KIF9, KIFC1, KLC3, KLC4

Clusters (C) of functional categories identified from Gene Set Enrichment Analysis leading edge analysis (see Figure 3): 17 clusters (c1-c17) contain 76 of 182 functional categories and 455 of 653 genes with increased transcript levels in the adherent Jeko-1 fraction (ADH) of the co-culture as compared to MCL cells in suspension in co-culture (SUSP) and 7 clusters (c18-c24) contain 16 of 20 functional categories and 148 of 166 genes with decreased transcript levels in the ADH fraction. Genes with significantly higher (n=146) or lower (n=45) transcript levels in the ADH fraction as compared to SUSP are marked in red and blue, respectively.

nificant transcript level changes between mono- and cocultured MS-5 stromal cells is available in Online Supplementary Table S5.

Adhesion dependent changes in MCL cells are associated with four main functional themes Gene set enrichment analysis was used to identify functional categories enriched in the significantly up- (182 gene sets) or down-(20 gene sets) regulated genes in MCL cells adhered to stromal cells. Leading-edge genes (n=819), which account for the level of the enrichment score for haematologica | 2018; 103(4)

Figure 3. Identification of functionally related clusters of genes by Gene Set Enrichment Analysis leading edge analysis. Heatmaps showing clustering of functional categories and leading edge genes that are up-regulated (A) or downregulated (B) in adherent Jeko-1 cells relative to suspension of Jeko-1 cells in cocultures. The numbers of leading edge genes and functional categories are shown in each panel: 24 clusters (c1c24), defined as having ≥10 genes and ≥1 functional categories, were identified (see Table 2 for more details). Four additional groups of genes that contain many functional categories but which do not fulfill the cluster criteria are denoted i-iv.

each of the functional categories, were clustered across the set of identified functional categories to produce heat maps showing clusters of genes and functional categories that together identify functional differences between adherent and suspension cells in the co-culture. Figure 3 identifies 24 functional clusters (containing at least 10 genes and ≥ 1 functional category) for genes with increased (17 clusters, Figure 3A) or decreased (7 clusters, Figure 3B) transcript levels in adherent MCL cells. The genes and functional categories defining each cluster are listed in Table 2, and full lists of leading edge genes and 673

A. Arvidsson et al.

674

Figure 4. Differentially expressed genes in co-culture adherent mantle cell lymphoma (MCL) cells overlap with microenvironment-regulated genes in MCL and chronic lymphocytic leukemia (CLL) patients. (A) Venn diagram showing overlaps between differentially expressed genes between adherent and suspension Jeko-1 cells in co-culture (ADH/SUSP) compared with microenvironment-regulated genes from microarray studies of MCL and CLL patients, where transcript levels in lymph nodes were compared to those of peripheral blood (data from GSE21029 and GSE70910, respectively). The numbers of differentially expressed genes (FDR q-value â&#x2030;¤ 0.05) for each data set are shown. (B) Comparison of the direction of regulation for the 116 genes in the overlap of all three data sets. Color-coding is red for genes up-regulated in adherent cells/lymph node and blue for down-regulated genes. Fold change values are shown for each gene and data set. (C) Scatter plot showing fold changes for 348 genes with significant transcript level changes in both adherent MCL cells relative to suspension co-culture cells (Jeko-1 ADH-SUSP FC) and MCL patient lymph node relative to peripheral blood (MCL LN-PB FC). Spearman Rho=0.843 (P=1.83x10-68) for genes with coherent directional change in the two data sets where red data points (n=135) represent up-regulated genes and blue data points (n=114) down-regulated. Gray data points (n=99) represent genes with opposite direction of regulation between the two data sets. (D) Scatter plot showing fold changes for 228 genes with significant transcript level changes in both adherent MCL cells relative to suspension co-culture cells (Jeko-1 ADH-SUSP FC) and CLL patient lymph node relative to peripheral blood (CLL LN-PB FC). Spearman Rho=0.746 (P=8.39x10-29) for genes with coherent directional change in the two data sets where red data points (n=121) represent up-regulated genes and blue data points (n=34) down-regulated. Gray data points (n=73) represent genes with opposite direction of regulation between the two data sets. (E) A core gene set of 13 cell-adhesion related microenvironmentally regulated genes in MCL and CLL (see text for details). Transcript levels of the genes in adherent (ADH, black bars) and suspension (SUSP, gray bars) MCL cells in co-cultures are plotted; mean transcripts per million reads (TPM) where the read counts are normalized to library size and feature lengthÂąStandard Deviation.

haematologica | 2018; 103(4)

Global mRNA changes in stroma-bound lymphoma cells

enriched gene sets from GSEA are presented in Online Supplementary Tables S6-S9. Most functional gene clusters can be attributed to a small number of cellular themes, including B-cell activation and immune cell signaling (c2, c4, c7, c10, c17), apoptosis and anti-apoptosis (c1, c5), cell adhesion and migration (c6, c8, c15, c16), and early mitosis (c19, c22, c23, c24). The first three themes contain up-regulated genes in adherent cells while the fourth cluster contains down-regulated genes. Genes in the B-cell activation-related clusters included B-cell receptor components, co-stimulatory surface molecules and soluble factors (e.g. CD79A, CD86, CD180, ICOSLG, PDCD1 and TGFB1). The apoptosisrelated clusters contained anti-apoptotic BCL2-family genes (e.g. BCL2A1, BCL2L1 and MCL1) while adhesionand migration-related clusters included ACTB, ACTG1 and TUBA4A. The mitosis-related genes are primarily involved in early mitosis steps such as mitotic onset (e.g. AURKA, CCNB2 and PLK1), spindle establishment (e.g. BUB1, CENPA, CENPE), and kinetochore formation (e.g. KIF20A, KIF4A, KIF5A). These themes are each compatible with survival strategies resulting from adherence dependent changes in immune cell characteristics, apoptosis pathways and cell cycle status. Other clusters (c9, c11, c13, c20, c21) represent more general functional characteristics, which could in principle contribute to one or more of the process-related themes. Finally, there are 4 clusters (c3, c12, c14, c18) for which fewer than 10% of member genes are differentially regulated even though they do form part of the GSEA leading edge. These clusters may be important but they contain only 11 significantly differentially regulated genes. For example, the c12, containing proteasome subunits, is of interest given the use of the proteasome inhibitor, bortezomib, for treatment of refractory MCL. Apart from the clustered genes there are several leadingedge genes that are not easily clustered because they occur in many functional classes (Groups i–iv, Figure 4A). Examples include NFKB1, NFKB2, ICAM1, CCL3 and CCL4 and the groups, as well as other results are detailed in the Online Supplementary Appendix. Secreted CCL3/4 levels increased in co-cultures and upon MCL-cell simulation with anti-IgM (Online Supplementary Figure S1). Altogether 199 of the 1050 genes with significantly altered transcript levels (19%) were functionally classified by GSEA. Network analysis connected an additional 495 adhesion regulated genes to the 24 GSEA clusters, thus functionally connecting approximately 65% of the 1050 differentially expressed genes. Network analysis did not identify any gene networks not connected to GSEA clusters, strongly suggesting that we have identified the main processes and pathways defined by differential expression of genes in adherent cells.

Overlap between adhesion-regulated MCL cell genes and microenvironment-regulated genes in MCL and CLL patients Analysis of publicly available data from two independent studies of CLL and MCL30,31 identified 2090 and 4136 differentially expressed genes, respectively, between cells from lymph node and blood (FDR q-value≤0.05). Comparison of the 1050 adhesion-associated genes observed in the in vitro co-culture system with differentially expressed genes from the CLL and MCL datasets showed significant overlaps of 228 genes (22%, haematologica | 2018; 103(4)

P=2.7x10-38) and 348 genes (34%, P=8.2x10-38), respectively (Figure 4A). In all, 116 genes were differentially expressed in all three datasets (Figure 4B). Forty-eight (41%) of these genes are included in the GSEA leading edge used to define cell-adhesion-related processes and 32 (28%) of these are members of one of the 24 functional clusters (Table 2), with 29 being in clusters that primarily define the four functional themes discussed previously. Thus the 116 genes that overlap between the three studies are representative of the gene clusters defining adhesion-associated gene expression. For 65 of the genes, the direction of regulation (up or down) was the same in each of the three datasets, assuming similarity between adherent cells in vitro and lymph node cells in vivo. Similarities are also indicated by fold change correlations for genes that are differentially regulated in both the in vitro co-culture system and the MCL and CLL datasets (Figure 4C and D). Although there are differences between in vivo and in vitro studies, the co-culture system faithfully reproduces a significant subset of differential gene regulation events observed in MCL and CLL patients. Thirteen of the 65 genes with conserved direction of regulation (CCL3, CCL4, DUSP4, ETV5, ICAM1, IL15RA, IL21R, IL4I1, MFSD2A, NFKB1, NFKBIE, SEMA7A, TMEM2) had a fold-change value of 2 or more in the present study (Figure 4E). CCL3, CCL4 and NFKBIE are members of previously identified NF-κB and BCR signatures30 and are likely to represent a broader activation of NF-κB and BCR pathways (Online Supplementary Figure S2). Of the 51 genes with a different regulation direction in one of the three datasets, 43 showed a discrepancy in the in vitro data in relation to the patient studies; 38 of these genes were related to proliferation, early mitosis or mitotic spindle formation.

Discussion Mantle cell lymphoma is a B-cell lymphoma that is difficult to cure and patients experience frequent relapses, often resulting from minimal residual disease. As described above, adherence of lymphoma cells to stromal cells within microenvironmental niches is thought to be essential for their proliferation, survival and drug resistance, as well as for their immunomodulatory ability to recruit other cell types to the microenvironment.8-11,17,18 Here, we developed a co-culture model system to systematically dissect differences in gene expression that occur in MCL cells and stromal cells that adhere to each other, using genome-wide RNA sequencing. A total of 1050 adherence-specific genes in MCL cells represent 24 functionally defined gene clusters, many of which can be attributed to four main functional themes. These correspond well with the important biological and pathological characteristics that functionally define lymphoma cells in microenvironments. The identified functions may be acquired characteristics specific to cancer cells or characteristics associated with normal B cells. The 1050 differentially regulated genes in adherent MCL cells significantly overlap with genes that are differentially expressed in lymph node, compared to blood, in MCL and CLL patients. The most differentially regulated genes included B-cell receptor signature genes that were not seen among highly regulated genes in a previous study,32 perhaps either due to the different study designs used or because differ675

A. Arvidsson et al.

ent in vitro systems reflect different aspects of microenvironment interactions.

Cell adhesion, migration and homing Mantle cell lymphoma cell adhesion to stromal cells is associated with induction of genes involved in cellular adhesion and cell motility/migration. The extracellular matrix cluster (c6) consists entirely of matrisome proteins,33 where 9 encode matrisome core proteins (glycoproteins, collagens or proteoglycans) that constitute structural components of extracellular matrix. The remaining 29 genes encode matrisome-associated proteins, which modulate extracellular matrix function, including cellular adhesion. Other clusters contain structural components important for cell rigidity and motility such as ACTB, ACTG1 and TUBA4A. Additional cell-cell adhesion-related genes with significantly higher transcript levels in adherent MCL cells include ICAM1, ITGB2 and AMIGO2. These may be important for homing of MCL cells to microenvironmental niches and their retention in the niche. The CXCR4 gene encoding a homing-related cytokine receptor is down-regulated in adherent cells, consistent with previous reports on microenvironment-associated downregulation of its mRNA and protein level in MCL and CLL patients as well as cell culture systems.17,30,34-36 The CXCR4 ligand, CXCL12, is expressed by the stromal cells and can promote adhesion of MCL cells to fibronectin and VCAM-1.17 Thus, this receptor-ligand pair could facilitate homing and subsequent adhesion of MCL cells to microenvironments.

Anti-apoptosis and cell survival Anti-apoptosis clusters are up-regulated in adherent cells and include genes involved in CD40 signaling (c1), which is known to have an anti-apoptotic effect on MCL cells.37 A second example (c5) includes upregulation of BCL2-family members (eg. BCL2A1, BCL2L1, and MCL1) as well as BCL2 that is up-regulated in the co-cultured MCL cells relative to mono-cultured cells. Importantly, increased level of alternative family members causes resistance to the BCL2 inhibitor ABT-737 in CLL,38 thus indicating their significance for designing therapies targeting the BCL2-family. Results expand previous knowledge showing that stromal cell interaction protects MCL and CLL cells from spontaneous and drug-induced apoptosis.8,10, 11

Cell proliferation The observed downregulation of early mitosis genes is consistent with previous observations showing cell cycle arrest in MCL and diffuse large B-cell lymphoma (DLBCL) cells upon interaction with stromal cells in vitro15 as well as lower response rates to cytostatic drugs observed for adherent lymphoma cells compared to cells in suspension.8 Reduced proliferation of MCL cells co-cultured with stromal cells lacking CD40L relative to cells expressing CD40L37 may indicate interaction partner-specific differences in proliferative response. While adherent MCL cells up-regulate CD40, the CD40L was not detected, consistent with downregulation of cell cycle genes in this system.

Immune cell recruitment Adherent MCL cells up-regulate genes involved in B-cell receptor (BCR) downstream signaling, seen in BCR and 676

NF-ÎşB gene signatures, including genes involved in immune-modulation via recruitment of immune cells, such as CCL3 and CCL4 that are known to be induced in adherent MCL cells.17,20 CCL3 and CCL4 attract activated T cells and monocytes, and CCL4 has been shown to attract regulatory T cells.39 T-cell infiltration has prognostic relevance in MCL40 and elevated serum levels of CCL3 and CCL4 are correlated with an inferior prognosis in DLBCL and CLL.41,42 A further example is the increased level of the immunoregulatory chemokine IL-10 in adherent cells, which permits autocrine survival signaling via the IL-10 receptor (IL10RA, 2.9-fold up-regulated in co-cultured cells) and STAT3 (cluster c17).43 BCR activation can induce such an autocrine survival loop in MCL which is attenuated by the proteasome-inhibitor drug, bortezomib.44 Elevated serum levels of IL-10 are associated with poor disease outcome in DLBCL.45 Regulatory B cells commonly express IL-10 and phenotypically-related CLL cells have been ascribed similar immunosuppressive attributes.46 The importance of BCR-signaling is supported by observations that blocking BCR signaling by the BTK inhibitor ibrutinib (PCI-32765) both lowered plasma levels of CCL22, CCL4, TNF and IL-10 and caused MCL cells to leave their protective microenvironmental niches and enter the peripheral blood.17 BTK-dependent secretion of the affected cytokines was induced in vitro by co-culture with stromal cells or IgM stimulation. Reduced transcript and secreted level of IL-10 has also been observed in DLBCL cell lines treated with ibrutinib or the mTOR inhibitor, AZD2014.47 Ccl2, which could play a role in monocyte recruitment and polarization similar to that shown in follicular lymphoma,48 and Ccl7, which promotes and directs the migration of macrophages,49 were up-regulated in adherent stromal cells. Thus adherence of MCL cells to stromal cells appears to be sufficient to establish the cytokine production needed for recruitment of monocytic cells and T-cell subsets even in the absence of such cells.

Utility of the co-culture model and identification of a core set of microenvironment genes The utility of the co-culture model system is supported by the systematic identification of functional themes that correspond well with current knowledge about how lymphomas in general and MCL in particular develop and survive. Relevance was further supported by overlap of the 1050 adherence-regulated genes with microenvironment (lymph node) regulated genes in MCL and CLL patients.30,31 The overlap is similar in extent to the overlap between the MCL and CLL data, indicating that the in vitro model system reciprocates relevant aspects of microenvironment-mediated gene regulation in lymphoma cells. The direction of regulation (up or down) was not conserved between the three datasets in 51 of the 116 genes that overlap between all three datasets, and, perhaps unsurprisingly, it is most often the in vitro data (43 of 116 genes) that differ from the in vivo studies. Interestingly, most of these genes are associated with the cell cycle, consistent with previous reports of cell cycle arrest in MCL and DLBCL cells adhering to stromal cells.15 The effect of cell adherence on the cell cycle may depend on specific properties of the interacting stromal cells, as discussed above in relation to expression of CD40L.37 haematologica | 2018; 103(4)

Global mRNA changes in stroma-bound lymphoma cells

Among 116 overlapping genes, a core gene set of 13 genes (fold change >2-fold) was identified. The core gene set contains genes that relate to the four functional themes and that have previously been coupled to lymphoma pathogenesis (CCL3, CCL4, ICAM1, NFKB1 and IL21R) as well as DUSP4, ETV5, IL15RA, IL4I1 and NFKBIE that have been described in a lymphoma context. The presence of the NFKBIE NF-ÎşB inhibitor may appear contradictory, but the apoptotic/anti-apoptotic responses to the NF-ÎşB pathway activity have been shown to be pluralistic and context dependent.50 The set also includes three genes which have not previously been associated with lymphoma pathology, MFSD2A, SEMA7A and TMEM2, suggesting the existence of at least some relevant genes that remain to be characterized. In summary, this first genome-wide, systematic study shows that genes differentially expressed in stromal cell adherent MCL cells are predominantly involved in antiapoptosis, B-cell signaling, cell adhesion and early mitosis.

References 1. Perez-Galan P, Dreyling M, Wiestner A. Mantle cell lymphoma: biology, pathogenesis, and the molecular basis of treatment in the genomic era. Blood. 2011;117(1):2638. 2. Swerdlow S, Campo E, Harris N, et al. WHO classification of tumours of haematopoietic and lymphoid tissues. 4th ed. Vol. 2. IARC; 2008. 3. Herrmann A, Hoster E, Zwingers T, et al. Improvement of overall survival in advanced stage mantle cell lymphoma. J Clin Oncol. 2009;27(4):511-518. 4. Geisler CH, Kolstad A, Laurell A, et al. Long-term progression-free survival of mantle cell lymphoma after intensive frontline immunochemotherapy with in vivopurged stem cell rescue: a nonrandomized phase 2 multicenter study by the Nordic Lymphoma Group. Blood. 2008; 112(7):2687-2693. 5. Burger JA, Peled A. CXCR4 antagonists: targeting the microenvironment in leukemia and other cancers. Leukemia. 2009;23(1):43-52. 6. Kurtova AV, Tamayo AT, Ford RJ, Burger JA. Mantle cell lymphoma cells express high levels of CXCR4, CXCR5, and VLA-4 (CD49d): importance for interactions with the stromal microenvironment and specific targeting. Blood. 2009;113(19):4604-4613. 7. Collins RJ, Verschuer LA, Harmon BV, et al. Spontaneous Programmed Death (Apoptosis) of B-Chronic LymphocyticLeukemia Cells Following Their Culture Invitro. Br J Haematol. 1989;71(3):343-350. 8. Medina DJ, Goodell L, Glod J, et al. Mesenchymal stromal cells protect mantle cell lymphoma cells from spontaneous and drug-induced apoptosis through secretion of B-cell activating factor and activation of the canonical and non-canonical nuclear factor kappaB pathways. Haematologica. 2012;97(8):1255-1263. 9. Panayiotidis P, Jones D, Ganeshaguru K, Foroni L, Hoffbrand AV. Human bone marrow stromal cells prevent apoptosis and support the survival of chronic lymphocyt-

haematologica | 2018; 103(4)

10.

11.

12.

13.

14.

15.

16. 17.

18.

Overlaps with clinical MCL and CLL data sets suggest that the identified genes also play important roles within cancer microenvironments in patients. The results support the utility of this in vitro model system for dissecting microenvironmental signaling and identify a list of 13 critical genes that should be a focus for future studies. Acknowledgments We would like to thank the core facility at Novum, BEA, Bioinformatics and Expression Analysis, which is supported by the board of research at the Karolinska Institute and the research committee at the Karolinska hospital. Funding The study was supported by grants from: The Swedish Research Council (to AW), The Swedish Cancer Society (to AW and BS), The Cancer Society in Stockholm (to BS), The Stockholm County Council (to BS) and Karolinska Institutet (to AW and BS).

ic leukaemia cells in vitro. Br J Haematol. 1996;92(1):97-103. Lagneaux L, Delforge A, Bron D, De Bruyn C, Stryckmans P. Chronic lymphocytic leukemic B cells but not normal B cells are rescued from apoptosis by contact with normal bone marrow stromal cells. Blood. 1998;91(7):2387-2396. Kurtova AV, Balakrishnan K, Chen R, et al. Diverse marrow stromal cells protect CLL cells from spontaneous and drug-induced apoptosis: development of a reliable and reproducible system to assess stromal cell adhesion-mediated drug resistance. Blood. 2009;114(20):4441-4450. Weekes CD, Pirruccello SJ, Vose JM, Kuszynski C, Sharp JG. Lymphoma cells associated with bone marrow stromal cells in culture exhibit altered growth and survival. Leuk Lymphoma. 1998;31(1-2):151165. Lwin T, Crespo LA, Wu A, et al. Lymphoma cell adhesion-induced expression of B cell-activating factor of the TNF family in bone marrow stromal cells protects non-Hodgkin's B lymphoma cells from apoptosis. Leukemia. 2009;23(1):170177. Burger JA, Tsukada N, Burger M, et al. Blood-derived nurse-like cells protect chronic lymphocytic leukemia B cells from spontaneous apoptosis through stromal cell-derived factor-1. Blood. 2000; 96(8):2655-2663. Lwin T, Hazlehurst LA, Dessureault S, et al. Cell adhesion induces p27Kip1-associated cell-cycle arrest through down-regulation of the SCFSkp2 ubiquitin ligase pathway in mantle-cell and other non-Hodgkin B-cell lymphomas. Blood. 2007;110(5):1631-1638. Scott DW, Gascoyne RD. The tumour microenvironment in B cell lymphomas. Nat Rev Cancer. 2014;14(8):517-534. Chang BY, Francesco M, De Rooij MF, et al. Egress of CD19(+)CD5(+) cells into peripheral blood following treatment with the Bruton tyrosine kinase inhibitor ibrutinib in mantle cell lymphoma patients. Blood. 2013;122(14):2412-2424. Bernard S, Danglade D, Gardano L, et al.

19.

20.

21.

22.

23.

24.

25.

26. 27.

28.

Inhibitors of BCR signalling interrupt the survival signal mediated by the micro-environment in mantle cell lymphoma. Int J Cancer. 2015;136(12):2761-2774. Schulz A, Toedt G, Zenz T, et al. Inflammatory cytokines and signaling pathways are associated with survival of primary chronic lymphocytic leukemia cells in vitro: a dominant role of CCL2. Haematologica. 2011;96(3):408-416. Burger JA, Quiroga MP, Hartmann E, et al. High-level expression of the T-cell chemokines CCL3 and CCL4 by chronic lymphocytic leukemia B cells in nurselike cell cocultures and after BCR stimulation. Blood. 2009;113(13):3050-3058. Burger JA, Burger M, Kipps TJ. Chronic lymphocytic leukemia B cells express functional CXCR4 chemokine receptors that mediate spontaneous migration beneath bone marrow stromal cells. Blood. 1999;94(11):3658-3667. Chiron D, Dousset C, Brosseau C, et al. Biological rational for sequential targeting of Bruton tyrosine kinase and Bcl-2 to overcome CD40-induced ABT-199 resistance in mantle cell lymphoma. Oncotarget. 2015;6(11):8750-8759. Pedersen IM, Kitada S, Leoni LM, et al. Protection of CLL B cells by a follicular dendritic cell line is dependent on induction of Mcl-1. Blood. 2002;100(5):1795-1801. Conway T, Wazny J, Bromage A, et al. Xenome--a tool for classifying reads from xenograft samples. Bioinformatics. 2012;28(12):i172-178. Kim D, Pertea G, Trapnell C, et al. TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions. Genome Biol. 2013;14(4):R36. Anders S, Huber W. Differential expression analysis for sequence count data. Genome Biol. 2010;11(10):R106. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA. 2005;102(43):15545-15550. Szklarczyk D, Franceschini A, Wyder S, et

677

A. Arvidsson et al.

29.

30.

31.

32.

33.

34.

35.

678

al. STRING v10: protein-protein interaction networks, integrated over the tree of life. Nucleic Acids Res. 2015;43(Database issue):D447-452. Edgar R, Domrachev M, Lash AE. Gene Expression Omnibus: NCBI gene expression and hybridization array data repository. Nucleic Acids Res. 2002;30(1):207-210. Herishanu Y, Perez-Galan P, Liu D, et al. The lymph node microenvironment promotes B-cell receptor signaling, NF-kappaB activation, and tumor proliferation in chronic lymphocytic leukemia. Blood. 2011;117(2):563-574. Saba NS, Liu D, Herman SE, et al. Pathogenic role of B-cell receptor signaling and canonical NF-kappaB activation in mantle cell lymphoma. Blood. 2016;128(1):82-92. Sivina M, Hartmann E, Vasyutina E, et al. Stromal cells modulate TCL1 expression, interacting AP-1 components and TCL1targeting micro-RNAs in chronic lymphocytic leukemia. Leukemia. 2012;26(8):18121820. Naba A, Clauser KR, Hoersch S, et al. The matrisome: in silico definition and in vivo characterization by proteomics of normal and tumor extracellular matrices. Mol Cell Proteomics. 2012;11(4):M111 014647. Mittal AK, Chaturvedi NK, Rai KJ, et al. Chronic lymphocytic leukemia cells in a lymph node microenvironment depict molecular signature associated with an aggressive disease. Mol Med. 2014;20(290301. Ghobrial IM, Bone ND, Stenson MJ, et al. Expression of the chemokine receptors CXCR4 and CCR7 and disease progression in B-cell chronic lymphocytic leukemia/

36.

37.

38.

39.

40.

41.

42.

43.

small lymphocytic lymphoma. Mayo Clin Proc. 2004;79(3):318-325. Purroy N, Abrisqueta P, Carabia J, et al. Coculture of primary CLL cells with bone marrow mesenchymal cells, CD40 ligand and CpG ODN promotes proliferation of chemoresistant CLL cells phenotypically comparable to those proliferating in vivo. Oncotarget. 2015;6(10):7632-7643. Chiron D, Bellanger C, Papin A, et al. Rational targeted therapies to overcome microenvironment-dependent expansion of mantle cell lymphoma. Blood. 2016;128(24):2808-2818. Vogler M, Butterworth M, Majid A, et al. Concurrent up-regulation of BCL-XL and BCL2A1 induces approximately 1000-fold resistance to ABT-737 in chronic lymphocytic leukemia. Blood. 2009;113(18):44034413. Bystry RS, Aluvihare V, Welch KA, Kallikourdis M, Betz AG. B cells and professional APCs recruit regulatory T cells via CCL4. Nat Immunol. 2001;2(12):11261132. Nygren L, Wasik AM, BaumgartnerWennerholm S, et al. T-cell levels are prognostic in mantle cell lymphoma. Clin Cancer Res. 2014;20(23):6096-6104. Takahashi K, Sivina M, Oki Y, et al. Serum CCL3 and CCL4 Levels Function As Novel Prognostic Markers in Diffuse Large B Cell Lymphoma. Blood. 2012;120(21):2. Sivina M, Hartmann E, Kipps TJ, et al. CCL3 (MIP-1alpha) plasma levels and the risk for disease progression in chronic lymphocytic leukemia. Blood. 2011; 117(5):1662-1669. Beguelin W, Sawh S, Chambwe N, et al. IL10 receptor is a novel therapeutic target

44.

45.

46.

47.

48.

49.

50.

in DLBCLs. Leukemia. 2015;29(8):16841694. Baran-Marszak F, Boukhiar M, Harel S, et al. Constitutive and B-cell receptor-induced activation of STAT3 are important signaling pathways targeted by bortezomib in leukemic mantle cell lymphoma. Haematologica. 2010;95(11):1865-1872. Lech-Maranda E, Bienvenu J, BroussaisGuillaumot F, et al. Plasma TNF-alpha and IL-10 Level-Based Prognostic Model Predicts Outcome of Patients with Diffuse Large B-Cell Lymphoma in Different Risk Groups Defined by the International Prognostic Index. Arch Immunol Ther Exp. 2010;58(2):131-141. DiLillo DJ, Weinberg JB, Yoshizaki A, et al. Chronic lymphocytic leukemia and regulatory B cells share IL-10 competence and immunosuppressive function. Leukemia. 2013;27(1):170-182. Ezell SA, Mayo M, Bihani T, et al. Synergistic induction of apoptosis by combination of BTK and dual mTORC1/2 inhibitors in diffuse large B cell lymphoma. Oncotarget. 2014;5(13):4990-5001. Guilloton F, Caron G, Menard C, et al. Mesenchymal stromal cells orchestrate follicular lymphoma cell niche through the CCL2-dependent recruitment and polarization of monocytes. Blood. 2012; 119(11):2556-2567. Xuan W, Qu Q, Zheng B, Xiong S, Fan GH. The chemotaxis of M1 and M2 macrophages is regulated by different chemokines. J Leukoc Biol. 2015;97(1):6169. Staudt LM. Oncogenic Activation of NFkappa B. Cold Spring Harbor Perspect Biol. 2010;2(6):30.

haematologica | 2018; 103(4)

ARTICLE

Non-Hodgkin Lymphoma

Histone modifier gene mutations in peripheral T-cell lymphoma not otherwise specified

Ferrata Storti Foundation

Meng-Meng Ji,1 Yao-Hui Huang,1 Jin-Yan Huang,1 Zhao-Fu Wang,2 Di Fu,1 Han Liu,1 Feng Liu,1 Christophe Leboeuf,3,4 Li Wang,1,3 Jing Ye,3 Yi-Ming Lu,3 Anne Janin,3,4 Shu Cheng1 and Wei-Li Zhao1,3

State Key Laboratory of Medical Genomics, Shanghai Institute of Hematology; Shanghai Rui Jin Hospital, Shanghai Jiao Tong University School of Medicine, China; 2Department of Pathology, Shanghai Rui Jin Hospital; Shanghai Jiao Tong University School of Medicine, China; 3Pôle de Recherches Sino-Français en Science du Vivant et Génomique, Laboratory of Molecular Pathology, Shanghai, China and 4U1165 Inserm/Université Paris 7, Hôpital Saint Louis, Paris, France 1

*M-MJ, Y-HH, J-YH and Z-FW contributed equally to this work.

Haematologica 2018 Volume 103(4):679-687

ABSTRACT

ue to heterogeneous morphological and immunophenotypic features, approximately 50% of peripheral T-cell lymphomas are unclassifiable and categorized as peripheral T-cell lymphomas, not otherwise specified. These conditions have an aggressive course and poor clinical outcome. Identification of actionable biomarkers is urgently needed to develop better therapeutic strategies. Epigenetic alterations play a crucial role in tumor progression. Histone modifications, particularly methylation and acetylation, are generally involved in chromatin state regulation. Here we screened the core set of genes related to histone methylation (KMT2D, SETD2, KMT2A, KDM6A) and acetylation (EP300, CREBBP) and identified 59 somatic mutations in 45 of 125 (36.0%) patients with peripheral T-cell lymphomas, not otherwise specified. Histone modifier gene mutations were associated with inferior progression-free survival time of the patients, irrespective of chemotherapy regimens, but an increased response to the histone deacetylase inhibitor chidamide. In vitro, chidamide significantly inhibited the growth of EP300-mutated T-lymphoma cells and KMT2D-mutated T-lymphoma cells when combined with the hypomethylating agent decitabine. Mechanistically, decitabine acted synergistically with chidamide to enhance the interaction of KMT2D with transcription factor PU.1, regulated H3K4me-associated signaling pathways, and sensitized T-lymphoma cells to chidamide. In a xenograft KMT2D-mutated T-lymphoma model, dual treatment with chidamide and decitabine significantly retarded tumor growth and induced cell apoptosis through modulation of the KMT2D/H3K4me axis. Our work thus contributes to the understanding of aberrant histone modification in peripheral T-cell lymphomas, not otherwise specified and the stratification of a biological subset that can benefit from epigenetic treatment (Clinical trials.gov identifiers: NCT 01746992 and NCT 02533700).

Introduction Peripheral T-cell lymphomas (PTCL) represent a heterogeneous clinicopathological entity of non-Hodgkin lymphoma with an aggressive disease course and poor clinical outcome. Approximately 50% of PTCL are unclassifiable and categorized as PTCL, not otherwise specified (PTCL-NOS).1 Using gene expression profiling, PTCL-NOS lymphocytes can be distinguished from normal T lymphocytes, with deregulation of genes involved in apoptosis, proliferation, cell adhesion, and tranhaematologica | 2018; 103(4)

Correspondence: zhao.weili@yahoo.com or orenge@medmail.com.cn Received: October 13, 2017. Accepted: January 3, 2018. Pre-published: January 5, 2018.

doi:10.3324/haematol.2017.182444 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/679 ©2018 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.

679

M-M. Ji et al.

scription regulation.2 Two subgroups of PTCL-NOS have been identified, which are characterized by high expression of either GATA3 or TBX21/T-bet transcription factors and downstream target genes.3 However, actionable biomarkers closely related to the pathogenic mechanism need to be further investigated and may become potential therapeutic targets of PTCL-NOS.4,5 Epigenetic alterations play a crucial role in tumor progression.6 Next-generation sequencing technologies have led to the discovery of epigenetic modifier gene mutations in PTCL, such as the DNA methylation genes TET2, TET1 and DNMT3A,7,8 and chromatin remodeler genes ARID1B and ARID2.9,10 Meanwhile, genes of histone methylation, such as KMT2D, KMT2A, KDM6A, SETD2 and EZH2, and those of histone acetylation, including CREBBP and EP300, have also been found in PTCL and other T-lymphoid malignancies.9,11-13 To further determine their prognostic significance and correlation with clinical treatment, here we assessed the mutational pattern of the main epigenetic modifier genes in patients with PTCL-NOS.

ral particles were incubated with Jurkat cells for 72 h. The stably transduced cells were selected by EGFP or mCherry fluorescence protein after transduction.

Methods

Online supplementary methods

Patients A total of 239 patients with previously untreated PTCL-NOS were enrolled in this study. The histological diagnosis was established according to the World Health Organization (WHO) classification.14 The study was approved by the Shanghai Rui Jin Hospital Review Board with informed consent obtained in accordance with the Declaration of Helsinki.

Targeted sequencing Targeted sequencing was performed on available tumor samples of 125 patients. To determine the mutations of candidate genes, polymerase chain reaction primers were designed by iPLEX Assay Design software (Sequenom, California, USA). Multiplexed libraries of tagged amplicons from patients with PTCL-NOS were generated by the 48×48 Access Array microfluidic platform (Fluidigm, South San Francisco, USA) according to the manufacturer’s protocol. Deep sequencing was performed with established Illumina protocols on the GAIIx and MiSeq platform (Illumina, California, USA). Matched peripheral blood samples were included to exclude germline polymorphisms and the mutations were confirmed by Sanger sequencing.

Cell line and reagents The Jurkat T-leukemia cell line was obtained from the American Type Culture Collection. Cells were grown in RPMI-1640 medium, supplemented with 10% heat-inactivated fetal bovine serum in a humidified atmosphere containing 5% CO2 at 37°C. Valproic acid (VPA, V3640) was from Sigma (San Francisco, USA). Suberoylanilide hydroxamic acid (SAHA, S1047) and romidepsin (ROMI, S3020) were from Selleck (Houston, USA). Chidamide, the histone deacetylase (HDAC) inhibitor clinically available in China, was kindly provided by Chipscreen (shenzhen,China).

Lentivirus packaging and transduction Purified plasmids pGV365-KMT2D (WT), pGV365-KMT2D (V5486M), pGV365-EP300 (WT) and pGV365-EP300 (H1377R) were transfected with package vectors into HEK-293T cells using lipofectamine 2000 (Invitrogen, California, USA; 11668019) according to the manufacturer’s protocol. The supernatant fraction of HEK-293T cell cultures was then condensed to a viral concentration of approximately 3×108 transducing units/mL. The lentivi680

Statistical analysis Data were calculated as the mean ± standard deviation from three separate experiments. The Student t-test was applied to compare two normally distributed groups and the Mann-Whitney U test to compare two groups which did not conform to normal distribution. The Bonferroni adjustment was used to perform multiple comparisons. Progression-free survival was calculated from the date when treatment began to the date when the disease progression was recognized or the date of the last follow-up. Overall survival time was measured from the date of diagnosis to the date of death or the last follow-up. Univariate hazard estimates were generated with unadjusted Cox proportional hazards models. Covariates demonstrating statistical significance with P values <0.05 on univariate analysis were included in the multivariate model. All statistical procedures were performed with the SPSS version 20.0 statistical software package or GraphPad Prism 5 software. P<0.05 was considered statistically significant.

DNA preparation, western blot, immunofluorescence, immunohistochemistry, isobolographic analysis, mRNA-seq library preparation and sequencing analysis, ChIP-seq library preparation and sequencing analysis, the TUNEL assay, murine model and micro-positron emission tomography and computed tomography imaging are described in the Online Supplementary Methods.

Results Histone modifier genes were frequently mutated in peripheral T-cell lymphoma not otherwise specified A total of 91 somatic mutations of epigenetic modifier genes were identified in 60 of 125 (48.0%) patients with PTCL-NOS by targeted sequencing (Figure 1A). Most of the somatic mutations were missense mutations (n=72), followed by nonsense (n=10) and frameshift mutations (n=9) (Figure 1B). We observed a preference for C>T/G>A alterations analogous to the somatic single-nucleotide variation spectrum in other cancers (Figure 1C). No correlation was found in terms of age and gender. Mutations of histone methylation genes (category I) most frequently occurred in KMT2D (encoding H3K4 methyltransferase, 25/125 patients, 20.0%), followed by those in SETD2 (encoding H3K36 methyltransferase, 6/125 patients, 4.8%), KMT2A (encoding H3K4 methyltransferase, 3/125 patients, 2.4%) and KDM6A (encoding H3K27 demethylase, 1/125 patients, 0.8%). No EZH2 mutation was detected. Mutations of histone acetylation genes (category II) were found in EP300 (encoding H3K18 acetyltransferase, 10/125 patients, 8.0%) and in CREBBP (encoding H3K18 acetyltransferase, 5/125 patients, 4.0%). DNA methylation genes TET2, TET1 and DNMT3A (category III), as well as chromatin remodeler genes ARID1B and ARID2 (category IV), were also affected in 12.0%, 3.2%, 3.2%, 4% and 1.6% of the patients, respectively (Figure 1A and Online Supplementary Table S1). In accordance with the conceptual classification of the mutated genes, overlap mutations were seldom present among histone methylation, histone acetylation, DNA methylation or chromatin remodeler genes. In particular, histone methylation gene haematologica | 2018; 103(4)

Histone modifier gene mutations in PTCL-NOS

mutations, such as mutations of histone acetylation genes, were mutually exclusive of each other, suggesting that histone modifying genes might be involved in distinct biological processes (Figure 1D). Alterations of histone modifier genes were primarily located at well-conserved amino acid positions across distinct species (Online Supplementary Figure S1). Typically, KMT2D/KMT2A mutations affected the PHD domain (e.g. residues 134-320, 1378-1556, 5032-5138, and 14331624, 1871-1983), HMG domain (residues 2021-2072), undetermined domain (e.g. residues 2487-4658) and SET domain (e.g. residues 5397-5519 and 3825-3969). EP300/CREBBP mutations affected the HAT domain (e.g. residues 1306-1612 and 1342-1649).

Histone modifier gene mutations were associated with disease progression in peripheral T-cell lymphoma not otherwise specified One hundred and forty patients were treated with CHOP-based chemotherapy in a historical cohort of Shanghai Ruijin Hospital from 1997 to 2011, and referred to as the training cohort. The validation cohort consisted of 99 patients enrolled in two prospective studies (NCT 01746992 and NCT 02533700, randomized trials to compare CHOP-based chemotherapy with sequential chemotherapy with CEOP/IVE/GDP or CTOP/ITE/MTX). Since 2012, 49 and 50 patients have been randomized to CHOP-based or sequential chemotherapy, respectively. No obvious differences in clinical and pathological characteristics or treatment

response were observed either between the training and the validation cohort, or between the two arms within the validation cohort (Online Supplementary Table S2). Gene mutation data were available for 73 and 52 patients of the training and validation cohorts with available tissue samples, respectively (Figure 1A). In the training cohort, the median follow-up time was 29.1 months (range, 0.5-162.0 months). The 2-year progression-free and overall survival rates of the patients were 36.7% and 47.1%, respectively. In the univariate analysis, the International Prognostic Index was a significant prognostic factor for both progression-free survival and overall survival (both P<0.001), but histone modifier mutations were only prognostic for progression-free survival and not overall survival (P=0.012 and P=0.095, respectively) (Figure 2A,B). In the multivariate analysis, when the International Prognostic Index was controlled for, the presence of a histone modifier gene mutation was an independent prognostic factor for progression-free survival (P<0.001) (Table 1). The 2-year progression-free and overall survival rates were 26.4% and 56.6% for patients with histone modifier gene mutations and 49.6% and 63.3% for patients without mutations (Figure 2A,B). In the validation cohort, the median follow-up time was 19.5 months (range, 2.1-43.0 months). Histone modifier gene mutations were associated with shorter progression-free survival in multivariate analysis (P=0.049) (Table 1). The 2year progression-free and overall survival rates were 22.2% and 24.2% for patients with histone modifier gene mutations and 41.1% and 57.5% for patients without

Figure 1. Histone modifier gene mutations in peripheral T-cell lymphoma, not otherwise specified. (A) Gene mutations identified by targeted sequencing in 125 patients with peripheral T-cell lymphomas. The number of patients (N) with mutations is listed on the right. The mutations are classified into the categories indicated on the left: I, histone methylation; II, histone acetylation; III, DNA methylation; IV, chromatin remodeler. (B) Number and type of non-silent somatic mutations. (C) Number and percentage of non-silent somatic single nucleotide variants. (D) Circos diagram according to mutation categories.

haematologica | 2018; 103(4)

681

M-M. Ji et al.

mutations (P=0.045 and P=0.224) (Figure 2C,D). Overall, these data expand the prognostic role of histone modification in disease progression in PTCL-NOS.

Histone modifier gene mutations sensitized T-lymphoma cells to the histone deacetylase inhibitor chidamide and/or the hypomethylating agent decitabine A possible structure-function relationship of the mutants was addressed using the crystal structure of the proteins encoded by KMT2D (PDB4Z4P) and EP300 (PDB4PZR), the two most frequently mutated histone modifier genes. As shown in Figure 3A, KMT2D R5389W, E5444K and V5486M might destabilize the SET domain and reduce histone methylation activity, EP300 E1377R, W1466_ and E1515V might disrupt the acetyl-CoA binding pocket, destabilize the HAT domain and reduce histone acetylation activity. Next, representative missense mutants KMT2D (V5486M) and EP300 (H1377R), as well as WT KMT2D and EP300, were established and trans-

fected into Jurkat cells. Compared with WT protein, while KMT2D mutant reduced the level of H3K4me3, this reduction was restored by the HDAC inhibitors romidepsin and chidamide (Figure 3B and Online Supplementary Figure S2). On the other hand, the level of H3K18ac was reduced in EP300 mutant, but this effect was restored by the HDAC inhibitors valproic acid, suberoylanilide hydroxamic acid, or romidepsin and chidamide (Figure 3C). These results were observed by western blot and by immunofluorescence assay (Figure 3B,C). In tumor samples from PTCL-NOS patients, a significantly lower fraction of nuclear H3K4me3 positivity (+++~++++, 30%) was observed in cases with the KMT2D mutation than in those without mutations. Similarly, a lower fraction of nuclear H3K18ac positivity (+++~++++, 17%) was present in cases with EP300/CREBBP mutations than in those without mutations (Figure 3D). Interestingly, upon treatment with chidamide (administered orally at a dose of 30 mg twice per

Table 1. Multivariate analysis of predictors of progression-free survival in patients with PTCL-NOS controlled by International Prognostic Index.

Variable International Prognostic Index Low & low/intermediate vs. intermediate/high & high risk Histone modifying gene mutations positive vs. negative

Training Cohort 95% CI

P value

Validation Cohort 95% CI

P value

4.209

2.148-8.250

<0.001

2.013

1.178-4.213

0.033

3.633

1.885-7.005

<0.001

2.017

1.062-4.414

0.049

RR. relative risk; 95% CI: confidence interval.

Figure 2. Progression-free survival and overall survival curves of patients with peripheral T-cell lymphoma not otherwise specified according to histone modifier gene mutations. (A) Progression-free survival and (B) overall survival curves of the training cohort. (C) Progression-free survival and (D) overall survival curves of the validation cohort.

682

haematologica | 2018; 103(4)

Histone modifier gene mutations in PTCL-NOS

week), relapsed patients with a histone modifier gene mutation showed a remarkably increased response rate (complete or partial remission), as compared to those without mutations (Figure 3E and Online Supplementary Table S3). Thus, such mutations might alter the protein function on chromatin state regulation, sensitizing PTCLNOS patients to HDAC inhibitors. In vitro, Jurkat cells bearing the KMT2D V5486M or EP300 H1377R mutant were treated with different concentrations of chidamide and/or the hypomethylating agent decitabine for 48 h. The combination index (CI) curve yielded most of the data points to the area <1, denoting synergistic interactions in KMT2D V5486M mutated cells. Meanwhile, the inhibitory effect on EP300

H1377R mutated cells was achieved by chidamide alone (Figure 4A). Flow cytometry revealed that chidamide and decitabine synergistically induced KMT2D V5486M mutated cell apoptosis and G0/G1 arrest (Figure 4B). The in vivo anti-tumor activity of dual treatment on T-cell lymphoma was further evaluated in a murine xenograft model in which KMT2D V5486M mutated Jurkat cells subcutaneously injected into nude mice. The tumors formed in mice co-treated with chidamide and decitabine were significantly smaller than those that formed in untreated animals or those treated with the single agents, starting from 15 days of treatment (Figure 4C, left panel), as visualized by 18F-fluorodeoxyglucose small-animal positron emission tomography â&#x20AC;&#x201C; computed tomography at 21 days of treat-

Figure 3. Effect of histone deacetylase inhibitor in KMT2D-mutated and EP300-mutated T-lymphoma. (A) Structure prediction of the missense mutations. The crystal structure of the complex of KMT2D and EP300 is PDB: 4Z4P and PDB: 4PZR, respectively. SAM, S-adenosyl-L-methionine. (B and C) Western blot and immunofluoresence assay of Jurkat cells transfected with wild-type (WT), KMT2D mutants (V5486M) (B) and EP300 mutants (H1377R) (C) upon treatment with different HDAC inhibitors. Jurkat cells were treated for 48 h at IC50. Histone 3 (H3) was used as a loading control. VPA, valproic acid, 3.7 mm; SAHA, suberoylanilide hydroxamic acid, 10 mm; ROMI, romidepsin, 5 nm; CHID, chidamide, 5 mm (48 h). Bar=10 mm. (D) Immunostaining of H3K4me3 and H3K18ac in tumor samples of PTCL-NOS patients with or without KMT2D or EP300 mutations. Bar=20 mm. (E) Response rate in relapsed PTCL-NOS patients treated with CHID according to the mutation status of histone modifier genes.

haematologica | 2018; 103(4)

683

M-M. Ji et al.

ment (Figure 4C, right panel). To search for more evidence of tumor cell apoptosis, a TUNEL assay was performed on mice tumor sections. Compared with the untreated group and the groups treated with single agents, the number of apoptotic tumor cells was increased following combined treatment (Figure 4D). In accordance with in vitro data, upregulation of H3K4me3 was more significant in the combination treatment group than in the single-agent and the untreated group (Figure 4E). To determine KMT2D-H3K4me3 DNA binding targets, ChIP-seq was performed using H3K4me3 antibody in KMT2D V5486M mutated Jurkat cells treated with chidamide (5 mm) alone or in combination with decitabine (5 mm) for 48 h. Presentation of the data in a

Venn diagram identified a significant non-overlapping portion of H3K4me3 binding promoters in the combination group, excluding 663 promoters overlapping with the chidamide group and 17 with the decitabine group (Figure 5A,B). Consistent with previous studies, H3K4me3 peaks were found at gene promoters. The group of promoters, whose H3K4me3 levels were affected by combined chidamide and decitabine treatment, but not by either chidamide or decitabine treatment alone, was enriched with binding site motifs for PU.1, a transcription factor that activates gene expression during myeloid and B-cell lymphoid cell development15,16 (Figure 5C). Furthermore, RNA sequencing analysis indicated that, in comparison with the untreated group and the single-agent

Figure 4. Effect of chidamide and decitabine in KMT2D-mutated and EP300-mutated T-lymphoma. (A) Combination index (CI) curve calculated by Compusyn software in KMT2D-mutated and EP300-mutated Jurkat cells treated with chidamide (CHID, 5 mm) and/or decitabine (DECI, 5 mm) for 48 h. (B) KMT2D-mutated Jurkat cell apoptosis and cell cycle determined by flow cytometry of cells treated with CHID and/or DECI for 48 h. *P<0.05, **P<0.01 compared with the untreated cells. (C) In vivo effect of the CHID and DECI combination in a murine T-lymphoma xenograft model. Tumor volume (left panel) and standardized uptake value (SUV) intensity of micro-positron emission tomograpy-computed tomography (right panel) of xenograft nude mice injected subcutaneously with KMT2D V5486-mutated Jurkat cells treated with CHID (12.5 mg/kg, twice weekly for 3 weeks), DECI (0.5 mg/kg, twice weekly for 3 weeks), either alone or in combination. **P<0.01 compared with the untreated group that received RPMI1640. (D) Apoptotic cells detected by the TUNEL assay (Ă&#x2014;400). Bar=20 mm. (E) Immunohistochemical assay of H3K4me3 in murine tumor samples treated with CHID and/or DECI. **P<0.01 compared with the untreated group. Bar=50 mm.

684

haematologica | 2018; 103(4)

Histone modifier gene mutations in PTCL-NOS

groups, combined treatment led to significant modulation of multiple signaling pathways associated with cancer, including those of apoptosis, cell cycle progression, cell adhesion, and transcriptional regulation (Figure 5D). Particularly, PU.1 was included in both the cancer pathway and the transcriptional pathway in the combined treatment group. Pathway enrichment analysis of the overlapping genes of the RNA-Seq and ChIP-Seq in the combination group was then performed. Significant pathways relevant to T-cell biology are shown in Figure 5E. As revealed by gene set enrichment analysis, the MAPK pathway was inactivated in the combination treatment group compared with the untreated group. Accordingly, p-ERK upregulation was observed not only in tumor samples of

PTCL-NOS patients with KMT2D mutations, but also in those of xenografted T-lymphoma mice bearing KMT2D V5486M mutants, the latter being inhibited by combined treatment with chidamide and decitabine (Figure 5F,G).

Discussion First observed in B-cell lymphoma, recurrent mutations of epigenetic modifier genes have recently been identified in PTCL-NOS.9 In the present study, we performed targeted sequencing of the main epigenetic modifier genes in a large cohort of Chinese PTCL-NOS patients. The results showed that the mutational spectrum of these genes in

Figure 5. Chip-seq and RNA sequencing data of KMT2D-mutated T-lymphoma cells treated with chidamide and/or decitabine. (A) Venn diagram depicting the overlap between transcription factors bound by H3K4me3 ChIP-seq in the combination group, as compared to the chidamide (CHID)-treated group and the decitabine (DECI)treated group in KMT2D V5486-mutated Jurkat cells. (B) The top significant transcription factors bound by H3K4me3 in the combination group. (C) ChIP-seq analysis of transcription factors bound by H3K4me3. Enriched H3K4me3-binding motifs for PU.1 analyzed by KMT2D V5486-mutated Jurkat cells treated with CHID and DECI relative to genomic background (upper panel). Genomic snapshots of PU.1 peaks bound by H3K4me3 in different groups (lower panel). (D) Cellular and genetic information processing revealed by RNA-seq on the combination group in KMT2D V5486-mutated Jurkat cells. (E) Pathway analysis of the most differentially expressed genes that overlapped in both RNA-Seq and ChIP-Seq analysis in the combination group (upper panel). Gene-set enrichment analysis of the MAPK pathway (lower panel). (F) Immunohistochemical assay of p-ERK in tumor samples of PTCL-NOS patients with or without KMT2D mutations. (G) Immunostaining of p-ERK in tumor samples of xenografted murine models bearing KMT2D V5486 mutants treated with CHID and/or DECI. **P<0.01 compared with the untreated group. Bar=20 mm.

haematologica | 2018; 103(4)

685

M-M. Ji et al.

PTCL-NOS was similar to that in B-cell lymphoma, in which predominantly missense mutations were found.17,18 Importantly, our study provided clinical evidence that histone modifier gene mutations, particularly those involved in histone methylation and acetylation, are significantly associated with tumor chemoresistance and disease progression of PTCL-NOS. The adverse prognostic effect of histone modifier gene mutations was further proven in a chemotherapy-independent manner, prompting us to explore bio-therapeutic agents that can overcome chemoresistance in PTCL-NOS patients. It is well known that HDAC inhibitors are potent anticancer drugs in hematopoietic malignancies, including lymphoma.19-21 The aim of using HDAC inhibitors is to restore normal histone modification patterns through inhibition of various components of the epigenetic machinery.22,23 In B-cell lymphoma, HDAC inhibitors can rescue deficits in histone acetylation induced by EP300/CREBBP mutations,24 rendering tumor cells more sensitive to suberoylanilide hydroxamic acid.25 This can explain why chidamide also has favorable efficacy on PTCL-NOS patients bearing EP300/CREBBP mutations. Moreover, KMT2D-mutated PTCL-NOS patients responded to chidamide. Both in vitro and in vivo, the combination of decitabine and chidamide induced apoptosis of Jurkat cells bearing the KMT2D mutant. This is in accordance with previous reports that decitabine and 5azacytidine produce a marked synergistic effect in combination with suberoylanilide hydroxamic acid and romidepsin in T-lymphoma cell lines by modulating cell cycle arrest and apoptosis.26,27 As a mechanism of action, KMT2D mutations of B-lymphoma cells promote malignant outgrowth by perturbing methylation of H3K4 that affect the JAK-STAT, Toll-like receptor, or B-cell receptor pathway.28,29 Here our study indicated that dual treatment with chidamide and decitabine enhanced the interaction of KMT2D with the transcription factor PU.1, thereby inactivating the H3K4me-associated signaling pathway

References 1. Broccoli A, Zinzani PL. Peripheral T-cell lymphoma, not otherwise specified. Blood. 2017;129(9):1103-1112. 2. Pileri SA, Piccaluga PP. New molecular insights into peripheral T cell lymphomas. J Clin Invest. 2012;122(10):3448-3455. 3. Iqbal J, Wright G, Wang C, et al. Gene expression signatures delineate biological and prognostic subgroups in peripheral Tcell lymphoma. Blood. 2014;123(19):29152923. 4. Rosenquist R, Rosenwald A, Du MQ, et al. Clinical impact of recurrently mutated genes on lymphoma diagnostics: state-of-the-art and beyond. Haematologica. 2016;101(9):1002-1009. 5. Zhang Y, Xu W, Liu H, Li J. Therapeutic options in peripheral T cell lymphoma. J Hematol Oncol. 2016;9:37. 6. Roy DM, Walsh LA, Chan TA. Driver mutations of cancer epigenomes. Protein Cell. 2014;5(4):265-296. 7. Lemonnier F, Couronne L, Parrens M, et al. Recurrent TET2 mutations in peripheral T-

686

10.

11.

12.

13.

MAPK, which is constitutively activated in T-cell lymphoma.13,30,31 The transcription factor PU.1 is involved in the development of all hematopoietic lineages32 and regulates lymphoid cell growth and transformation.33 Aberrant PU.1 expression promotes acute myeloid leukemia and is related to the pathogenesis of multiple myeloma via the MAPK pathway.34,35 On the other hand, PU.1 is also shown to interact with chromatin remodeler and DNA methyltransferease to control hematopoiesis and suppress leukemia.36 Our data thus suggested that the combined action of chidamide and decitabine may interfere with the differentiation and/or viability of PTCL-NOS through a PU.1-dependent gene expression program. In conclusion, histone modifier genes indicate clinical progression of PTCL-NOS and may represent a group of actionable biomarkers of this disease subtype. Characterized as a biological subset of PTCL-NOS, patients with dysregulation of the histone modification machinery may be amenable to therapeutic intervention with HDAC inhibitors, given either alone or in combination with hypomethylating agents. Acknowledgments The authors would like to thank all the patients involved in this study and their families. Funding This study was supported, in part, by research funding from the National Natural Science Foundation of China (81325003, 81520108003 and 81670716), Chang Jiang Scholars Program, the Shanghai Commission of Science and Technology (16JC1405800), Shanghai Municipal Education Commission Gaofeng Clinical Medicine Grant Support (20152206 and 20152208), Clinical Research Plan of SHDC (16CR2017A), Multicenter Clinical Research Project by Shanghai Jiao Tong University School of Medicine (DLY201601), Collaborative Innovation Center of Systems Biomedicine and the Samuel Waxman Cancer Research Foundation.

cell lymphomas correlate with TFH-like features and adverse clinical parameters. Blood. 2012;120(7):1466-1469. Vallois D, Dobay MP, Morin RD, et al. Activating mutations in genes related to TCR signaling in angioimmunoblastic and other follicular helper T-cell-derived lymphomas. Blood. 2016;128(11):1490-1502. Schatz JH, Horwitz SM, Teruya-Feldstein J, et al. Targeted mutational profiling of peripheral T-cell lymphoma not otherwise specified highlights new mechanisms in a heterogeneous pathogenesis. Leukemia. 2015;29(1):237-241. Odejide O, Weigert O, Lane AA, et al. A targeted mutational landscape of angioimmunoblastic T-cell lymphoma. Blood. 2014;123(9):1293-1296. Zhu X, He F, Zeng H, et al. Identification of functional cooperative mutations of SETD2 in human acute leukemia. Nat Genet. 2014;46(3):287-293. da Silva Almeida AC, Abate F, Khiabanian H, et al. The mutational landscape of cutaneous T cell lymphoma and Sezary syndrome. Nat Genet. 2015;47(12):1465-1470. Jiang L, Gu ZH, Yan ZX, et al. Exome sequencing identifies somatic mutations of

14.

15.

16.

17.

18.

19.

DDX3X in natural killer/T-cell lymphoma. Nat Genet. 2015;47(9):1061-1066. Swerdlow SH, Campo E, Pileri SA, et al. The 2016 revision of the World Health Organization classification of lymphoid neoplasms. Blood. 2016;127(20):23752390. Nutt SL, Metcalf D, D'Amico A, Polli M, Wu L. Dynamic regulation of PU.1 expression in multipotent hematopoietic progenitors. J Exp Med. 2005;201(2):221-231. Sokalski KM, Li SK, Welch I, Cadieux-Pitre HA, Gruca MR, DeKoter RP. Deletion of genes encoding PU.1 and Spi-B in B cells impairs differentiation and induces pre-B cell acute lymphoblastic leukemia. Blood. 2011;118(10):2801-2808. Pasqualucci L, Trifonov V, Fabbri G, et al. Analysis of the coding genome of diffuse large B-cell lymphoma. Nat Genet. 2011;43(9):830-837. Morin RD, Mendez-Lago M, Mungall AJ, et al. Frequent mutation of histone-modifying genes in non-Hodgkin lymphoma. Nature. 2011;476(7360):298-303. Falkenberg KJ, Johnstone RW. Histone deacetylases and their inhibitors in cancer, neurological diseases and immune disorders.

haematologica | 2018; 103(4)

Histone modifier gene mutations in PTCL-NOS

Nat Rev Drug Discov. 2014;13(9):673-691. 20. Shi Y, Jia B, Xu W, et al. Chidamide in relapsed or refractory peripheral T cell lymphoma: a multicenter real-world study in China. J Hematol Oncol. 2017;10(1):69. 21. Wozniak MB, Villuendas R, Bischoff JR, et al. Vorinostat interferes with the signaling transduction pathway of T-cell receptor and synergizes with phosphoinositide-3 kinase inhibitors in cutaneous T-cell lymphoma. Haematologica. 2010;95(4):613-621. 22. Popovic R, Licht JD. Emerging epigenetic targets and therapies in cancer medicine. Cancer Discov. 2012;2(5):405-413. 23. Stamatopoulos B, Meuleman N, De Bruyn C, Delforge A, Bron D, Lagneaux L. The histone deacetylase inhibitor suberoylanilide hydroxamic acid induces apoptosis, downregulates the CXCR4 chemokine receptor and impairs migration of chronic lymphocytic leukemia cells. Haematologica. 2010;95(7):1136-1143. 24. Jiang Y, Ortega-Molina A, Geng H, et al. CREBBP inactivation promotes the development of HDAC3-dependent lymphomas. Cancer Discov. 2017;7(1):38-53. 25. Andersen CL, Asmar F, Klausen T, Hasselbalch H, Gronbaek K. Somatic mutations of the CREBBP and EP300 genes affect

haematologica | 2018; 103(4)

26.

27.

28.

29.

30.

response to histone deacetylase inhibition in malignant DLBCL clones. Leuk Res Rep. 2012;2(1):1-3. Marchi E, Zullo KM, Amengual JE, et al. The combination of hypomethylating agents and histone deacetylase inhibitors produce marked synergy in preclinical models of Tcell lymphoma. Br J Haematol. 2015. doi: 10.1111/bjh.13566. [Epub ahead of print] Kalac M, Scotto L, Marchi E, et al. HDAC inhibitors and decitabine are highly synergistic and associated with unique geneexpression and epigenetic profiles in models of DLBCL. Blood. 2011;118(20):5506-5516. Ortega-Molina A, Boss IW, Canela A, et al. The histone lysine methyltransferase KMT2D sustains a gene expression program that represses B cell lymphoma development. Nat Med. 2015;21(10):1199-1208. Zhang J, Dominguez-Sola D, Hussein S, et al. Disruption of KMT2D perturbs germinal center B cell development and promotes lymphomagenesis. Nat Med. 2015;21(10): 1190-1198. Jones CL, Gearheart CM, Fosmire S, et al. MAPK signaling cascades mediate distinct glucocorticoid resistance mechanisms in pediatric leukemia. Blood. 2015;126(19): 2202-2212.

31. Chakraborty AR, Robey RW, Luchenko VL, et al. MAPK pathway activation leads to Bim loss and histone deacetylase inhibitor resistance: rationale to combine romidepsin with an MEK inhibitor. Blood. 2013;121(20): 4115-4125. 32. Huang G, Zhang P, Hirai H, et al. PU.1 is a major downstream target of AML1 (RUNX1) in adult mouse hematopoiesis. Nat Genet. 2008;40(1):51-60. 33. Rosenbauer F, Owens BM, Yu L, et al. Lymphoid cell growth and transformation are suppressed by a key regulatory element of the gene encoding PU.1. Nat Genet. 2006;38(1):27-37. 34. Will B, Vogler TO, Narayanagari S, et al. Minimal PU.1 reduction induces a preleukemic state and promotes development of acute myeloid leukemia. Nat Med. 2015;21(10):1172-1181. 35. Iseki Y, Nakahara M, Kubo M, Obata F, Harigae H, Takahashi S. Correlation of PU.1 and signal regulatory protein alpha1 expression in PU.1 transgenic K562 cells. Int J Mol Med. 2012;29(2):319-323. 36. van Riel B, Rosenbauer F. Epigenetic control of hematopoiesis: the PU.1 chromatin connection. Biol Chem. 2014;395(11):12651274.

687

ARTICLE

Chronic Lymphocytic Leukemia

Ferrata Storti Foundation

Haematologica 2018 Volume 103(4):688-697

Tumor necrosis factor receptor signaling is a driver of chronic lymphocytic leukemia that can be therapeutically targeted by the flavonoid wogonin Claudia Dürr,1 Bola S. Hanna,1 Angela Schulz,1 Fabienne Lucas,1,2 Manuela Zucknick,3,4 Axel Benner,3 Andrew Clear,2 Sibylle Ohl,1 Selcen Öztürk,1 Thorsten Zenz,5 Stephan Stilgenbauer,6 Min Li-Weber,7 Peter H. Krammer,7 John G. Gribben,2 Peter Lichter1 and Martina Seiffert1

Division of Molecular Genetics, German Cancer Research Center (DKFZ), Heidelberg, Germany; 2Centre for Haemato-Oncology, Barts Cancer Institute, Queen Mary University of London, UK; 3Division of Biostatistics, German Cancer Research Center (DKFZ), Heidelberg, Germany; 4Oslo Center for Biostatistics and Epidemiology; Department of Biostatistics, Institute of Basic Medical Sciences, University of Oslo, Norway; 5Molecular Therapy in Haematology and Oncology, German Cancer Research Center (DKFZ) and National Center for Tumor Diseases (NCT), and Department of Medicine V, University Hospital Heidelberg, Germany; 6Internal Medicine III, University of Ulm, Germany and 7 Division of Immunogenetics, German Cancer Research Center (DKFZ), Heidelberg, Germany 1

ABSTRACT

Correspondence: m.seiffert@dkfz.de

Received: July 31, 2017. Accepted: January 11, 2018. Pre-published: January 11, 2018. doi:10.3324/haematol.2017.177808 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/688 ©2018 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.

688

hronic lymphocytic leukemia is a malignancy of mature B cells that strongly depend on microenvironmental factors, and their deprivation has been identified as a promising treatment approach for this incurable disease. Cytokine array screening of 247 chronic lymphocytic leukemia serum samples revealed elevated levels of tumor necrosis factor (TNF) receptor-1 which were associated with poor clinical outcome. We detected a microenvironment-induced expression of TNF receptor-1 in chronic lymphocytic leukemia cells in vitro, and an aberrantly high expression of this receptor in the proliferation centers of patients’ lymph nodes. Stimulation of TNF receptor-1 with TNF-α enhanced nuclear factor κ-light-chain-enhancer of activated B cells (NFκB) activity and viability of chronic lymphocytic leukemia cells, which was inhibited by wogonin. The therapeutic effects of wogonin were analyzed in mice after adoptive transfer of Em-T-cell leukemia 1 (TCL1) leukemic cells. Wogonin treatment prevented leukemia development when given early after transplantation. The treatment of fullblown leukemia resulted in the loss of the TNF receptor-1 on chronic lymphocytic leukemia cells and their mobilization to blood. Targeting TNF receptor-1 signaling is therefore proposed for the treatment of chronic lymphocytic leukemia. Introduction Chronic lymphocytic leukemia (CLL) is a B-cell malignancy that is tightly regulated by and dependent on microenvironmental stimuli provided in lymphoid tissues.1 CLL cells in this protective niche show increased resistance to spontaneous and drug-induced apoptosis which is causative for CLL progression and relapse. In vitro studies using co-cultures of CLL and non-malignant accessory cells mirrored this dependency, and identified several CLL-relevant factors and pathways.2-4 Comparative gene expression profiling of CLL cells isolated from peripheral blood (PB), bone marrow (BM) and lymph nodes (LN) further identified enhanced B-cell receptor (BCR)-mediated signaling and nuclear factor κ-light-chain-enhancer of activated B cells (NFκB) activity in the lymphoid microenvironment compared to blood.5 In accordance with this, CLL cell migratory capability and tissue homing were shown to influence disease pathogenesis and progression.6 Data from clinical trials revealed that treatment with kinase inhibitors targeting BTK, SYK or PI3K-d leads to transient lymphocytosis accompanied by LN shrinkage due to CLL cell mobilization to PB.7-9 This impairment of CLL cell homing to lymphoid tissues subhaematologica | 2018; 103(4)

TNF receptor signaling as therapeutic target in CLL

stantially contributes to the observed high efficacy of these inhibitors.10-11 Albeit, despite their clinical success, CLL remains an incurable disease due to clonal evolution of malignant cells under treatment, followed by drug resistance and relapse.12 The current challenge is to develop new strategies by targeting not only CLL cells, but also the microenvironment, with the goal being that of eradicating the malignant cells. Tumor necrosis factor (TNF)-α and its receptors (TNFR) have been identified in the sera of CLL patients in increased concentrations, and high TNF-α levels are indicative for an aggressive disease, thus suggesting a role in CLL progression.13-16 TNF-α was shown to act as an autocrine growth factor in CLL.17,18 The inhibition of TNFR signaling by etanercept, a recombinant TNFR-2 derivative, in combination with the anti-CD20 antibody rituximab, caused durable remissions in refractory patients without 17p deletion.19 However, the detailed pathomechanism of TNFR signaling in CLL development and progression remains largely unknown. TNF-α is a pro-inflammatory cytokine that exerts its pleiotropic effects via two receptors, TNFR-1 (P55) and TNFR-2 (P75).20 Only TNFR-1 is endowed with an intracellular death domain, and can thereby induce either caspase-mediated apoptosis or prosurvival signals via NFκB activation.21 TNF-α-induced NFκB activation was shown to be blocked by wogonin, a naturally occurring flavonoid, resulting in a shift of TNFR1 signaling towards apoptosis induction.22 In a multitude of in vitro and in vivo studies, wogonin has been demonstrated to exert anti-oxidant, anti-inflammatory and antitumor activities.23 To elucidate the oncogenic role of TNFR-1 in CLL and to test its potential as a therapeutic target, we analyzed TNFR-1 expression and function in primary CLL cell cocultures and Em-T-cell leukemia 1 (TCL1) mice in the presence and absence of wogonin, and used these platforms for pre-clinical evaluation of TNFR-1 as a drug target in CLL.

Methods Samples PB of CLL patients (Online Supplementary Table S1) and healthy donors (HD) was obtained after informed consent and in accordance with the Declaration of Helsinki. The study was approved by the Institutional Review Board.

Glass slides were scanned on an Agilent microarray scanner (Agilent Technologies, Santa Clara, CA, USA) and data was analyzed using GenePix Pro software (Molecular Devices, San José, CA, USA). Serum soluble (s)TNFR-1 was quantified in mice after adoptive transfer (AT) of CLL using mouse sTNFR-1 enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems).

Statistical analysis Details of statistical analysis are provided in the Online Supplementary Methods.

Gene expression analysis CLL cells and CD19-sorted B cells of HD were cultured for one day in high cell density (1x107 cells in 4 mL per well in 6-well plates) and total ribonucleic acid (RNA) was isolated before (day 0 [d0]) and after culture (d1). Microarray-based transcriptome analysis and quantitative reverse transcription- polymerase chain reaction (RT-PCR) was performed as previously described.3

Tissue microarrays and immunohistochemistry Tissue microarrays (TMAs) including BM trephines (n=20 CLL patients; n=16 HD), LN sections (n=58 CLL patients; n=14 coincidental LN taken for non-malignant pathologies with no evidence of germinal center formation), and reactive LNs (n=28) were stained with hematoxylin and eosin (H&E) and antibodies against CD20 (Dako, clone L26, Agilent), TNFR-1 (polyclonal, Abcam, Cambridge, UK), CD3 (Labvision, clone SP7, ThermoFisher), and CD68 (Dako, clone KP1, Agilent), as detailed in Online Supplementary Methods.

Animal models and treatments AT of TCL1 splenocytes was performed as previously described.25 Briefly, 6-8-week-old C57BL/6 WT females (Charles River Laboratories, UK) were transplanted with 4x107 splenocytes pooled from leukemic TCL1 mice24 (>95% CD19+CD5+ cells) via tail vein injection. In early treatment studies, animals were randomized to daily treatment with 40 mg/kg wogonin in H2O containing arginine as an adjuvant to improve the solubility of wogonin, administered by intraperitoneal (i.p.) injection from 48 hours after AT for three weeks. For late treatment studies, PB tumor load was determined 21 days post-AT and animals were randomized to treatment with phosphate buffered saline (PBS) or 40 mg/kg wogonin (BIOTREND Chemicals AG, Wangen, Switzerland) solved in dimethyl sulfoxide (DMSO) for 21 days using daily oral gavage. At the endpoints, serum, peritoneal exudate, PB and single-cell suspensions of lymphoid organs were prepared as described previously.26

Quantification of Soluble TNFR-1 Sera from 247 CLL patients from the German CLL8 study (Online Supplementary Table S2) and from 50 age- and sex-matched healthy controls were used to quantify soluble TNFR-1 by cytometric bead arrays (BD Biosciences, Heidelberg, Germany), according to the manufacturer's protocol. Capture beads were synthesized by coupling anti-TNFR-1 antibody (Duoset, R&D Systems, Minneapolis, MN, USA) to functionalized beads. A biotinylated detection antibody (Duoset, R&D Systems) and a streptavidin conjugate were used for visualization. Data was acquired on a FACSCanto II flow cytometer and analyzed with FCAP software (BD Biosciences). Sera from final stage leukemic Em-TCL1 mice24 presenting with splenomegaly and more than 90% leukemic cells in the blood, and matched wild-type (WT) controls were screened for 144 inflammatory factors using Mouse Cytokine Array G2000 (RayBiotech, Norcross, GA, USA), according to the manufacturer’s protocol. haematologica | 2018; 103(4)

Results sTNFR-1 serum level predicts overall survival and tumor-associated deaths in CLL sTNFR-1 was quantified in serum from 247 CLL patients (Online Supplementary Table S2) and 50 age- and sex-matched controls by cytometric bead arrays. The analysis revealed a significantly higher median serum concentration of sTNFR-1 in CLL patients (2.30 ng/mL, range: 0.50-7.31) compared to controls (1.35 ng/mL, range: 0.761.74; P<0.0001; Figure 1A). Further, sTNFR-1 serum concentrations moderately but significantly correlated with β2-microglobulin (R=0.582; P<0.001; Figure 1B) and thymidine kinase (R=0.263; P<0.001; Figure 1C) serum levels, suggesting the malignant cells as the source of 689

C. Dürr et al. A

Figure 1. Elevated sTNFR-1 serum levels have prognostic relevance in CLL. (A) Serum levels of sTNFR-1 in 247 CLL patients and 50 age and sex-matched healthy donors were assessed by flow cytometry using bead arrays. The lines indicate mean concentrations. Significance analysis was performed via unpaired t-test with Welch's correction (***P<0.0001). (B) Correlations of sTNFR-1 and β2-microglobulin, (C) as well as thymidine kinase concentrations in serum samples of 247 CLL cases are depicted. (D) Associations of log2 sTNFR-1 serum concentration and overall survival, (E) along with tumor-associated deaths of CLL patients were assessed by Cox hazard model with log2 sTNFR-1 concentration as a continuous variable. Results are depicted in a Stone-Beran estimator, with the Stone-Beran estimate at the highest concentration shown as a blue line, the Stone-Beran estimate at the median concentration shown as a green line, and the Stone-Beran estimate at the lowest concentration shown as a black line. sTNFR-1: soluble tumor necrosis factor receptor-1.

sTNFR-1. Of interest, high sTNFR-1 concentrations significantly correlated with shorter overall survival (OS) (Figure 1D) and a higher incidence of tumor-associated deaths (TAD) (Figure 1E) when evaluated in a univariable Cox hazard model. Moreover, multivariate analysis revealed that sTNFR-1 represents a prognostic marker for OS and TAD irrespective of age, IGHV mutational status, 11q deletion, 17p deletion, and rituximab treatment (Online Supplementary Table S3).

Microenvironment-induced expression of TNFR-1 in CLL but not in healthy donor B cells Culturing CLL cells in high cell density provides survival supportive stimuli to the leukemic cells that would otherwise die by spontaneous apoptosis. To identify survivalstimulating pathways in CLL, we performed microarraybased gene expression profiling of CLL cells or HD B cells before and after one day of cultivation in high cell densities. Comparative analyses of the data obtained resulted in a list of 236 genes that were significantly different in their regulation between CLL and normal B cells (Online Supplementary Table S4 lists genes with the highest difference between CLL and HD). Among them, TNFR-1 (TNFRSF1A) appeared as one of the top upregulated transcripts in CLL (mean log2 fold change (FC) d1 vs. d0 =2.06), but not in HD (FC=-0.18). This finding was validated by quantitative RT-PCR using CD19-sorted CLL or normal B cells, confirming the induced expression of TNFR-1 in three out of four CLL samples (mean FC=10.13; SEM ± 3.27) but not in HD B cells (Figure 2A). We further detect690

ed significantly enhanced levels of membrane-bound TNFR-1 (mTNFR-1) in CLL cells by flow cytometry, with a mean relative median fluorescence intensity (MFI) of 1.37 (SEM ± 0.06) on freshly isolated CLL cells, and 7.35 (SEM ± 1.36) after one day of cultivation in high cell density (P<0.001; Figure 2B). To further investigate a microenvironment-dependent regulation of TNFR-1 expression, we cocultured CD19sorted CLL or healthy B cells for one day with CD14-sorted monocytes that have previously been shown to support CLL cell survival,4 and observed upregulated mTNFR1 expression in CLL cells by 4.08-fold (SEM ± 0.70), while healthy B cells remained negative (P=0.001; Figure 2C). TNFR-1 expression in CLL cells further increased in cocultures over seven days with a mean relative MFI of 7.85 (SEM ± 2.62) compared to 1.37 (SEM ± 0.01) in healthy B cells (Figure 2D). Similar results were obtained by culturing CLL cells in medium containing at least 20% human serum. Previous work has suggested that CLL cells that have recently divided in lymphoid tissues, emigrate to PB as CXCR4dimCD5bright cells.27 Over time, they begin to reexpress CXCR4 and lose CD5 expression, before entering the lymphoid system again. To test whether TNFR-1 expression is higher in CLL cells that have just left the lymphoid microenvironment, we compared TNFR-1 levels on CXCR4dimCD5bright and CXCR4brightCD5dim CLL cells from fresh blood samples of four patients, but did not observe any significant differences between the two cell subsets (data not shown). These data suggest that TNFR-1 expression is induced in haematologica | 2018; 103(4)

TNF receptor signaling as therapeutic target in CLL

Figure 2. TNFR-1 expression in CLL cells is regulated by the microenvironment. (A) TNFR-1 transcript levels were analyzed by quantitative RT-PCR using CD19-sorted CLL cells (n=4) or healthy donor (HD) B cells (n=3) before (day [d]0) and after 1 day (d1) of cultivation in high cell density (2.5 x 106 cells/mL). Results were normalized to the mean expression levels of HPRT, DCTN2, and PGK. (B) A total of 4 x 105 CLL peripheral blood mononuclear cells (PBMC) were cultured for 1 day in 200 mL complete medium in 96-well plates. Cell surface expression of TNFR-1 (mTNFR-1) was quantified by flow cytometry in freshly isolated cells (d0) and after 1 day of cultivation (d1; n=14). Relative median fluorescence intensity (MFI) of TNFR-1 normalized to isotype control staining was assessed by gating on CD20+ lymphocytes. Lines indicate means and SEM. Significance was calculated by paired Student's t-test (***P=0.0005). (C+D) A total of 5 x 105 CD19-sorted CLL cells or healthy B cells were cocultured with 1 x 105 CD14-sorted monocytes for 1 or 7 days in 400 mL complete medium in 48-well plates. mTNFR-1 was quantified on CD20+ lymphocytes by flow cytometry in freshly isolated cells (d0) and after cultivation (d1 and d7). Results are depicted as ratios of relative MFI on d1 or d7 versus d0. Lines show mean and SEM. Unpaired t-test with Welch's correction was applied for significance analysis (**P=0.001). CLL: chronic lymphocytic leukemia; PB: peripheral blood; mTNFR-1: membrane-bound TNFR-1.

CLL cells by microenvironmental stimuli, and the receptor is quickly shed from the cell surface when cells enter PB, which presumably leads to abnormally high TNFR-1 serum levels.

mTNFR-1 expression is restricted to B cells within proliferation centers To verify microenvironment-regulated expression of TNFR-1 in vivo, we performed immunohistochemical (IHC) analysis of LN and BM sections of CLL patients and HD using antibodies specific for TNFR-1, CD20, Ki-67, CD3, and CD68. In so doing, we observed co-localization of TNFR-1 with CD20 (B-cell marker) and Ki-67 (proliferation marker; Figure 3A), but not with CD3 (T-cell marker) or CD68 (marker for myeloid cells), suggesting B cell-specific expression of TNFR-1. In LNs, TNFR-1 positive cells were mainly located within proliferation centers consisting of Ki-67 positive, large, round paraimmunoblasts in nodular areas (Figure 3A). The percentage of TNFR-1 positive cells significantly correlated with that of Ki-67 positive cells (R=0.39; P<0.0001; Figure 3B), and was slightly higher in CLL-derived LN sections (n=58) with 10.68% haematologica | 2018; 103(4)

(SEM ± 1.93) compared to non-CLL samples (n=14) with 7.76% (SEM ± 3.42) TNFR-1 positive B cells (Figure 3C). In both groups, the results were very heterogeneous, ranging from 0.1-51.33% in CLL and from 0-38.34% in nonCLL samples. Analysis of reactive LN sections (n=30) revealed a clear accumulation of TNFR-1 in germinal centers with 17.48% (SEM ± 0.135) positive B cells compared to 1.74% (SEM ± 0.26) in the mantle zone (MZ; P<0.0001; Figure 3D). In BM biopsies of CLL patients (n=20), 4.12% (SEM ± 0.65) of the CD20 positive cells co-expressed TNFR-1, whereas significantly fewer co-expressing cells were detected in BM biopsies of HD (n=16; 2.38% SEM ± 0.46; P=0.04; Figure 3E). Altogether, these data show that mTNFR-1 is expressed by proliferating CLL cells that are localized in the LN and BM as well as by proliferating nonmalignant B cells in germinal centers of reactive LN.

TNF-α induced NFκB activation and survival of CLL cells in vitro is reduced by wogonin To study the downstream effects of TNFR-1 signaling in CLL, we cultured peripheral blood mononuclear cells (PBMC) from CLL patients (n=3) in 50% human serum for 691

C. Dürr et al. B

Figure 3. TNFR-1 is expressed by proliferating B cells in bone marrow and lymph nodes. Immunohistochemical staining of tissue microarrays containing lymph node (LN) sections and bone marrow (BM) trephine biopsies of CLL patients and healthy donors (HD) as well as sections of reactive LN were performed using CD20, Ki67 and TNFR-1 specific antibodies. (A) Representative micrographs of Ki-67 and TNFR-1 staining of a CLL LN section. (B) Correlation of percentages of Ki-67 positive cells and TNFR-1 positive cells in 58 CLL samples (38 LN and 20 BM). (C) CD20 positive B cells that co-expressed TNFR-1 were quantified in LN sections of CLL patients (n=38) and HD (n=14); (D) in germinal centers (GC) and marginal zones (MZ) of reactive LN sections (n=30); (E) and BM trephine biopsies isolated from CLL patients (n=20) and HD (n=16). Lines indicate mean percentage of TNFR-1 positive cells and SEM. Unpaired t-test with Welch's correction was applied for significance analysis (***P<0.0001; *P=0.04). ). CLL: chronic lymphocytic leukemia; TNFR-1: tumor necrosis factor receptor-1.

one day, which induced TNFR-1 expression. Stimulation of these cells with TNF-α significantly enhanced NFκB activity, as quantified by p65 binding to immobilized NFκB consensus sequence oligonucleotides (relative mean chemiluminescence intensity (MLI): 19.25; SEM ± 1.72), and could be blocked by neutralizing TNF-α-specific antibody (MLI: 3.45; SEM ± 1.14; P=0.001; Figure 4A). Treatment of these cultures with the flavonoid wogonin, known to impair TNF-α-induced NFκB signaling,22 resulted in a reduction of relative MLI from 2.78 (SEM ± 0.60) to 0.98 (SEM ± 0.53; Figure 4B), indicating that wogonin abolished TNF-α-induced NFκB activity. Next, the effect of TNF-α in combination with wogonin on cell survival was examined. After the induction of TNFR-1 expression by 1 day of culture in high cell densities, CLL cells were treated with increasing concentrations of wogonin, 30 minutes prior to the administration of TNF-α. Wogonin treatment for 24 h resulted in a concentration-dependent reduction in cell viability, that was significantly stronger in the presence of TNF-α (Figure 4C). Treatment with 50 µM 692

wogonin reduced average cell survival from 64.15% (SEM ± 1.80) to 44.72% (SEM ± 4.16; P=0.0003), which was further reduced to 37.92% (SEM ± 5.22; P=0.002) and 31.33% (SEM ± 2.99; P=0.004) in the presence of 10 and 50 ng/mL TNF-α, respectively. At 100 mM wogonin, survival was reduced to 27.39% (SEM ± 2.95) in the absence of TNF-α, and furthermore, to 22.13% (SEM ± 2.67; P=0.01) and 21.63% (SEM ± 3.31; P=0.009) in the presence of 10 and 50 ng/mL TNF-α, respectively. These results suggest that wogonin impacts on CLL cell viability in vitro by inhibiting TNF-α-mediated survival signals.

TNFR-1 expression and serum levels are mirrored in the Eµ-TCL1 mouse model of CLL

To investigate whether the Em-TCL1 mouse line, a wellestablished animal model of CLL,24 mirrors our findings of TNFR-1 in human CLL, we first analyzed the serum of mice with end-stage leukemia (n=5), and confirmed an average of 3.2-fold higher levels of sTNFR-1 as compared to control animals (n=3; data not shown). A significant haematologica | 2018; 103(4)

TNF receptor signaling as therapeutic target in CLL

Figure 4. TNF-α-induced NFκB activity in CLL cells is inhibited by wogonin. (A) 2 x 107 CLL PBMC (n=3) were treated with 100 ng/mL TNF-α or PBS as control in 400 mL complete medium with or without TNF-α-blocking antibody (15 mg/mL). After 30 min, NFκB activity was assessed via an oligo-based chemiluminescence ELISA that detects binding of p65 to NFκB consensus binding sequence (wt), and was normalized to p65 background binding to a mutated oligo (wt – mut). Mean values and respective SEMs are depicted relative to PBS control. Paired Student's t-test was applied for significance analysis (**P=0.001). (B) NFκB activity in 1 x 107 CLLderived PBMC (n=2) upon addition of 100 ng/mL TNF-α for 30 minutes in the presence or absence of 50 mM wogonin was assessed as described in (A). Results are depicted as mean values and SEMs relative to DMSO control (1%). (C) A total of 4 x 105 CLL PBMC were cultured for 1 day in 200 mL complete medium in 96well plates. On day 2, wogonin and TNF-α were added either alone, or in combination, at the concentrations indicated. 1% DMSO was used as untreated control. Cell survival was assessed after a further 24 h via flow cytometry by gating on annexin V-PE/7-AAD-negative cells. Results are depicted as mean survival rates and SEM of 5-8 samples. Paired Student's t-test was applied for significance analysis (*P<0.05; **P<0.01). PBS: phosphate buffered saline; wt: wild-type; mut: mutated; DMSO: dimethyl sulfoxide; NFκB: nuclear factor κ-light-chain-enhancer of activated B cells; TNF-α: tumor necrosis factor-α.

increase of sTNFR-1 serum levels was further induced in young syngeneic WT animals after AT of splenocytes from leukemic Em-TCL1 mice, which resulted in reliable and homogeneous development of CLL as formerly described.25,28 Mean concentrations of sTNFR-1 42 days after transplantation were 2.79 ng/mL (SEM ± 0.23) in TCL1 AT mice (n=7) and 2.04 ng/mL (SEM ± 0.23) in WT mice (n=6; P=0.04; Figure 5A). In addition, analysis of mTNFR-1 by flow cytometry in murine CD19+CD5+ CLL cells isolated from different tissue sites revealed significantly higher mTNFR-1 expression in the spleen, with a mean MFI of 3.53 (SEM ± 0.28) compared to 1.84 (SEM 0.28) in PB (P<0.001; Figure 5B), suggesting that microenvironmental regulation of TNFR-1 expression in malignant cells is mirrored in the Em-TCL1 model.

Wogonin reduces CLL development in the TCL1 adoptive transfer model As wogonin impaired NFκB activation and survival of CLL cells in vitro, we investigated its impact on leukemia development in the TCL1 AT model. Two days after transplantation of malignant cells, mice were randomized to daily treatment with 40 mg/kg wogonin or PBS by i.p. injection for three weeks (n=10; Figure 6A). Eight out of ten mice analyzed in this study responded to wogonin treatment with significantly lower spleen weights of an average of 0.25g (SEM ± 0.06) in treated mice as compared to 0.36g (SEM ± 0.04) in the control cohort (P=0.03; Figure 6B). To assess tumor load in all affected tissues, percentages of CD19+CD5+ cells of viable CD45+ cells were analyzed in the spleen, PB and peritoneal cavity (PC) by flow cytometry. As depicted in Figure 6C, control mice exhibited a median splenic tumor load of 23.78% (SEM ± 5.37), whereas CLL cells were almost completely absent in the eight mice responding to wogonin; the mean percentage of CLL cells in all treated animals, including the two nonresponders, was 13.42% (SEM ± 7.32). Further, the CLL haematologica | 2018; 103(4)

cell content was reduced by wogonin treatment from 31.14% (SEM ± 5.63) to 18.79% (SEM ± 5.99) in PB (Figure 6D), and from 56.31% (SEM ± 6.78) to 32.55% (SEM ± 10.67) in PC (Figure 6E). The proliferation rate of CLL cells in vivo was assessed by i.p. injection of 200 mg EdU 20 h prior to euthanization of mice, which confirmed the reduced proliferative activity of CLL cells in mice responding to wogonin treatment (Figure 6F). The data was too heterogeneous to reach significance, due to two animals that did not respond to treatment. But in the majority of mice, wogonin was able to control CLL development.

Wogonin impacts on TNFR-1 expression in vivo The effect of wogonin was further investigated in mice with advanced disease where treatment was started 21 days after AT of CLL (Figure 7A). Mice were gavaged daily with either 40 mg/kg/d wogonin (n=5) or PBS (n=9) for three weeks. No significant difference in spleen weight was observed upon wogonin treatment (0.92g, SEM ± 0.13 in treated versus 1.08g, SEM ± 0.08 in control mice; Figure 7B), although there was a tendency, in the treated group, of slightly smaller spleens. Similar results were obtained for tumor load in the spleen, which was 61.10% (SEM ± 4.48) in wogonin-treated versus 68.28% (SEM ± 3.18) in control mice (Figure 7C). By contrast, the percentage of CD19+CD5+ cells in PB was significantly increased in treated mice (74.67%, SEM ± 3.82) compared to controls (58.84%, SEM ± 3.39; P=0.01; Figure 7D). This significant increase persisted when CLL cell percentages after treatment were normalized to CD19+CD5+ percentages at the start of treatment, which was three weeks after AT of CLL (P=0.04; Figure 7E). To analyze whether this effect might be linked to the interference of wogonin with TNFR-1 signaling, we analyzed mTNFR-1 expression on splenic CLL cells after treatment. Interestingly, wogonin induced an almost complete loss of mTNFR-1 expression in CLL cells (MFI 1.50 in treated versus 3.53 in control mice; P=0.02; Figure 7F). This is most likely due to receptor shedding, since 693

C. DĂźrr et al.

quantification of sTNFR-1 in the serum of these mice by ELISA revealed significantly higher levels of 4.22 ng/mL (SEM 0.34 ng/mL) in wogonin-treated mice compared to 2.79 ng/mL (SEM 0.23 ng/mL) in control mice (P=0.005; Figure 7G). This increase in sTNFR-1 levels in wogonintreated mice is not a sign of disease progression, but rather a treatment effect, including enhanced TNFR-1 shedding

and mobilization of leukemic cells from the spleen to PB. Despite the fact that wogonin failed to effectively control advanced disease in mice, it had an impact on TNFR-1 expression, and was associated with an increased accumulation of malignant cells in PB, suggesting that prolonged treatment with wogonin, or combinations with other drugs, might lead to CLL-effective responses.

Figure 5. CLL-associated TNFR-1 expression in EÂľ-TCL1 mice. (A) Syngeneic, immunocompetent C57BL/6 mice were injected (i.v.) with 4 x 107 splenocytes from fully leukemic Em-TCL1 mice (TCL1 AT; n=7). Non-transplanted C57BL/6 mice (WT; n=6) were used as controls. Serum concentrations of sTNFR-1 were quantified 42 days after CLL cell engraftment via ELISA. Lines indicate mean concentrations and SEM. Unpaired Student's t-test was applied for significance analysis (*P=0.04). (B) Single cell suspensions of peripheral blood (PB) and spleen (SP) of TCL1 AT mice (n=9) 42 days after transplantation were analyzed for mTNFR-1 expression by flow cytometry by gating on CD45+CD5+CD19+ CLL cells. Staining of relative MFI values normalized to isotype control are depicted. Values for PB (dots) and spleen (squares) samples of each mouse (1-9) are connected by dotted lines. Paired Student's t-test was applied for significance analysis (P<0.0001). MFI: median fluorescence intensity; IgG: immunoglobulin G; sTNFR-1: soluble tumor necrosis factor receptor-1: mTNFR-1: membrane-bound TNFR-1.

Figure 6. Wogonin reduces leukemia development in the TCL1 adoptive transfer model. (A) Syngeneic, immunocompenent C57BL/6 mice were injected (i.v.) with 4 x 107 splenocytes from fully leukemic Em-TCL1 mice. Two days after transplantation, mice were either treated with 40 mg/kg/d wogonin (n=10) or PBS (ctl; n=10) by i.p. injections. Mice were sacrificed after 21 days of treatment. Serum and organs were collected. (B) Spleen weight of wogonin-treated and control mice was assessed after 21 days of treatment. (C-E) Single cell suspensions were collected from the spleen (SPL), blood (PB) and peritoneal cavity (PC). Tumor load was assessed by flow cytometry staining and is indicated as CD5+ CD19+ cells out of CD45+ cells in (C) SPL, (D) PB, (E) and PC. (F) Mice were injected i.p. with 0.1 mg/g EdU 20 h before sacrificing, and EdU incorporation in spleen CD5+ CD19+ malignant cells was analyzed after Click-iT reaction by flow cytometry. Lines in all graphs indicate mean values and SEM. Paired Student's t-test was applied for significance analysis (*P=0.03). d: day; g: gram; ctl: control.

694

haematologica | 2018; 103(4)

TNF receptor signaling as therapeutic target in CLL

Discussion The relevance of microenvironmental interactions that mediate pro-survival signaling in CLL is now generally accepted. This is, however, thus far mainly based on in vitro studies, and their role in vivo still remains ill-defined. Interfering with the crosstalk of CLL cells and their microenvironment and thereby depriving malignant cells from supportive factors has become an attractive novel approach for treatment. In the study herein, we identified TNFR-1 as a pivotal player in CLL pathology. We observed significantly elevated sTNFR-1 serum levels in CLL patients, in line with results from Digel et al.16 We further showed that sTNFR-

haematologica | 2018; 103(4)

1 levels correlate with Î˛2-microglobulin and thymidine kinase serum levels, which are indicative for tumor load as well as with OS and TAD of CLL patients. Thereby, the prognostic power of sTNFR-1 was independent of established prognostic markers. These findings are in line with observed correlations of serum sTNFR-1 and disease aggressiveness in CLL, breast, colon, and pancreatic cancer.15,16,29 We further showed that CLL development in EmTCL1 mice is associated with elevated TNFR-1 serum levels. Taken together, our findings determine sTNFR-1 as a predictor for disease progression in CLL. Albeit conflicting data has been presented concerning the expression of mTNFR-1 on freshly isolated PB-derived CLL cells,15,30,31 our results clearly demonstrate that these

Figure 7. Wogonin impacts on TNFR-1 expression in vivo. (A) Adoptive transfer of TCL1 splenocytes was performed as described in Figure 6. At d21 after transplantation, mice were treated either with 40 mg/kg/d wogonin (n=5) or PBS (ctr; n = 9) by daily oral gavage. Mice were sacrificed after 21 days of treatment. Serum and organs were collected. (B) Spleen weight of wogonin-treated and control mice was assessed after 21 days of treatment (d42). (C-E) Single cell suspensions were collected from spleen (SPL) and blood (PB; d42). Tumor load was assessed by flow cytometry staining and is indicated as CD5+CD19+ cells out of CD45+ cells in (C) SPL and (D) PB (*P=0.01). (E) Percentage tumor load in blood was further normalized to percentage at start of treatment in each mouse and is presented as ratio of d42 over d21 (*P=0.04). (F) TNFR-1 expression was quantified on CD5+CD19+ cells via flow cytometry relative to isotype control antibody (mTNFR-1; *P=0.02). (G) Serum sTNFR-1 concentration was assessed by ELISA (**P=0.005). Lines in all graphs indicate mean values and SEM. Paired Student's t-test was applied for significance analysis. d: day; g: gram; ctl: control; MFI: median fluorescence intensity; IgG: immunoglobulin G; sTNFR-1: soluble tumor necrosis factor receptor-1: mTNFR-1: membrane-bound TNFR-1.

695

C. Dürr et al.

are negative for mTNFR-1, but upregulate their expression upon in vitro cultivation. Consistent with our data, TNFR1 expression was reported on malignant cells in diffuse large B-cell lymphoma (DLBCL), which correlated with significantly shorter OS and progression-free survival rates compared to patients with TNFR-1 negative lymphoma cells.32 TNFR-1 induction upon malignant transformation has also been described in colorectal adenoma and prostate cancer, underlining its role in carcinogenesis.33,34 BCR and TNFR signaling as well as canonical NFκB activity characterize the LN microenvironment in CLL.5,35 TNFR-1 is a pleiotropic receptor which induces either cellular activation via NFκB or apoptosis via activation of caspases. NFκB activation appears to be the default pathway resulting in expression of anti-apoptotic proteins, whereas specific inhibition of NFκB prior to TNF-α stimulation triggers cell death.36 Our data, procured from tissue microarray staining of human BM and LN sections, suggests that TNFR-1 signaling contributes to NFκB activity in CLL cells. Thereby TNFR-1 expression was enriched within B-cell rich proliferation centers in CLL samples and germinal centers in reactive LN, the sites of NFκB activity.37 Along the same line, malignant B cells in the blood of Em-TCL1 mice were negative for TNFR-1, but upregulated the receptor upon recirculation to the spleen, stressing the hypothesis that TNFR-1 might be involved in CLL cell activation and survival maintenance. Inflammatory pathways are central for CLL cell survival.3 In agreement with this fact, elevated TNF-α levels were identified in CLL patients and correlated with poor prognosis.13-15 Nonetheless, data on the role of TNFα in CLL pathogenesis are controversial. It has been suggested that it acts as an autocrine and paracrine growth factor which induces CLL cell proliferation in vitro.17,18,38 However, Foa et al. showed that in the majority of CLL cases, proliferation was reduced upon TNF-α treatment.39 In our study, TNF-α had no effect on CLL cell proliferation (data not shown), but rather induced canonical NFκB activity in CLL cells in vitro. Similar results were reported by Coscia et al., who demonstrated that NFκB is activated in CLL cells with unmutated IGHV genes upon TNF-α exposure.40 Clinical intervention with TNF/TNFR signaling in CLL

References 1. Ten Hacken E, Burger JA. Microenvironment interactions and B-cell receptor signaling in chronic lymphocytic leukemia: implications for disease pathogenesis and treatment. Biochim Biophys Acta. 2016;1863(3):401-413. 2. Lagneaux L, Delforge A, Bron D, De Bruyn C, Stryckmans P. Chronic lymphocytic leukemic B cells but not normal B cells are rescued from apoptosis by contact with normal bone marrow stromal cells. Blood. 1998;91(7):2387-2396. 3. Schulz A, Toedt G, Zenz T, Stilgenbauer S, Lichter P, Seiffert M. Inflammatory cytokines and signaling pathways are associated with survival of primary chronic lymphocytic leukemia cells in vitro: a dominant role of CCL2. Haematologica. 2011; 96(3):408-416. 4. Seiffert M, Schulz A, Ohl S, Dohner H,

696

6. 7.

is currently restricted to global TNF-α neutralization, using etanercept in combination with rituximab.19 However, less attention has been paid to the receptors that mediate the pathogenic effect of TNF-α. In particular, selective TNFR-1 inhibition showed promising results in the treatment of inflammatory diseases in mice.41-43 Wogonin, a naturally occurring monoflavonoid, was shown to attenuate TNF-α-conferred NFκB activity and thereby sensitize malignant T cells to apoptosis. Moreover, wogonin was reported to exert cytostatic and cytotoxic activities against several cancer cell lines in vitro and in vivo, accompanied by no or only mild side effects and low toxicity for non-malignant cells.44-46 The mechanism of action of wogonin is based on CDK9 inhibition and interference with reactive oxygen species (ROS) homeostasis.22,45 Wogonin was shown to shift the redox status of malignant T cells to a more reduced state by increasing H2O2 levels and decreasing ·O2 levels.22 This resulted in an inhibition of the redox-sensitive protein NFκB.47,48 In the study herein, wogonin reduced TNF-αmediated NFκB activity and induced apoptosis in CLL cells. Future studies need to investigate whether this effect is based on interference with ROS levels. Pre-clinical testing of wogonin after adoptive transfer of CLL in mice revealed that early treatment start resulted in a reduced tumor load in all tissues affected by disease, which might be due to the inhibition of tumor cell survival or proliferation. Accordingly, studies with several cancer cell lines showed that wogonin attenuates cyclin expression.44,49,50 When we treated animals with full-blown leukemia, starting 21 days after tumor engraftment, wogonin reduced the CLL cell content in the spleen and significantly increased tumor load in PB. This effect was accompanied by the loss of mTNFR-1 expression in CLL cells and elevated sTNFR-1 serum levels, suggesting that the shedding of mTNFR-1 from the surface of CLL cells might be causally involved in the observed accumulation of cells in the blood. Targeting CLL cells with drugs that are currently used in clinical treatment for CLL is much more effective in the blood. Therefore, combination therapy approaches with wogonin and, for example, therapeutic antibodies like rituximab might result in improved treatment responses.

Stilgenbauer S, Lichter P. Soluble CD14 is a novel monocyte-derived survival factor for chronic lymphocytic leukemia cells, which is induced by CLL cells in vitro and present at abnormally high levels in vivo. Blood. 2010;116(20):4223-4230. Herishanu Y, Perez-Galan P, Liu D, et al. The lymph node microenvironment promotes B-cell receptor signaling, NF-kappaB activation, and tumor proliferation in chronic lymphocytic leukemia. Blood. 2011;117(2):563-574. Davids MS, Burger JA. Cell trafficking in chronic lymphocytic leukemia. Open J Hematol. 2012;3(S1). Byrd JC, Furman RR, Coutre SE, et al. Targeting BTK with ibrutinib in relapsed chronic lymphocytic leukemia. N Engl J Med. 2013;369(1):32-42. Friedberg JW, Sharman J, Sweetenham J, et al. Inhibition of Syk with fostamatinib disodium has significant clinical activity in

10.

11.

12.

non-Hodgkin lymphoma and chronic lymphocytic leukemia. Blood. 2010; 115(13):2578-2585. Hoellenriegel J, Meadows SA, Sivina M, et al. The phosphoinositide 3'-kinase delta inhibitor, CAL-101, inhibits B-cell receptor signaling and chemokine networks in chronic lymphocytic leukemia. Blood. 2011;118(13):3603-3612. de Rooij MF, Kuil A, Geest CR, et al. The clinically active BTK inhibitor PCI-32765 targets B-cell receptor- and chemokine-controlled adhesion and migration in chronic lymphocytic leukemia. Blood. 2012; 119(11):2590-2594. Woyach JA, Smucker K, Smith LL, et al. Prolonged lymphocytosis during ibrutinib therapy is associated with distinct molecular characteristics and does not indicate a suboptimal response to therapy. Blood. 2014;123(12):1810-1817. Woyach JA, Furman RR, Liu T-M, et al.

haematologica | 2018; 103(4)

TNF receptor signaling as therapeutic target in CLL

13.

14.

15.

16.

17.

18.

19.

20. 21.

22.

23. 24.

25.

26.

Resistance Mechanisms for the Bruton's Tyrosine Kinase Inhibitor Ibrutinib. N Engl J Med. 2014;370(24):2286-2294. Ferrajoli A, Keating MJ, Manshouri T, et al. The clinical significance of tumor necrosis factor-alpha plasma level in patients having chronic lymphocytic leukemia. Blood. 2002;100(4):1215-1219. Adami F, Guarini A, Pini M, et al. Serum levels of tumour necrosis factor-alpha in patients with B-cell chronic lymphocytic leukaemia. Eur J Cancer. 1994;30A(9):12591263. Bojarska-Junak A, Hus I, Szczepanek EW, Dmoszynska A, Rolinski J. Peripheral blood and bone marrow TNF and TNF receptors in early and advanced stages of B-CLL in correlation with ZAP-70 protein and CD38 antigen. Leuk Res. 2008;32(2):225-233. Digel W, Porzsolt F, Schmid M, Herrmann F, Lesslauer W, Brockhaus M. High levels of circulating soluble receptors for tumor necrosis factor in hairy cell leukemia and type B chronic lymphocytic leukemia. J Clin Invest. 1992;89(5):1690-1693. Digel W, Stefanic M, Schoniger W, et al. Tumor necrosis factor induces proliferation of neoplastic B cells from chronic lymphocytic leukemia. Blood. 1989;73(5):12421246. Cordingley FT, Bianchi A, Hoffbrand AV, et al. Tumour necrosis factor as an autocrine tumour growth factor for chronic B-cell malignancies. Lancet. 1988;1(8592):969971. Woyach JA, Lin TS, Lucas MS, et al. A phase I/II study of rituximab and etanercept in patients with chronic lymphocytic leukemia and small lymphocytic lymphoma. Leukemia. 2009;23(5):912-918. Tartaglia LA, Goeddel DV. Two TNF receptors. Immunol Today. 1992;13(5):151-153. Liu ZG, Hsu H, Goeddel DV, Karin M. Dissection of TNF receptor 1 effector functions: JNK activation is not linked to apoptosis while NF-kappaB activation prevents cell death. Cell. 1996;87(3):565-576. Fas SC, Baumann S, Zhu JY, et al. Wogonin sensitizes resistant malignant cells to TNFalpha- and TRAIL-induced apoptosis. Blood. 2006;108(12):3700-3706. Li-Weber M. Targeting apoptosis pathways in cancer by Chinese medicine. Cancer Lett. 2013;332(2):304-312. Bichi R, Shinton SA, Martin ES, et al. Human chronic lymphocytic leukemia modeled in mouse by targeted TCL1 expression. Proc Natl Acad Sci USA. 2002; 99(10):6955-6960. McClanahan F, Riches JC, Miller S, et al. Mechanisms of PD-L1/PD-1-mediated CD8 T-cell dysfunction in the context of aging-related immune defects in the Emicro-TCL1 CLL mouse model. Blood. 2015;126(2):212-221. Hanna BS, McClanahan F, Yazdanparast H, et al. Depletion of CLL-associated

haematologica | 2018; 103(4)

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

patrolling monocytes and macrophages controls disease development and repairs immune dysfunction in vivo. Leukemia. 2016;30(3):570-579. Calissano C, Damle RN, Marsilio S, et al. Intraclonal complexity in chronic lymphocytic leukemia: fractions enriched in recently born/divided and older/quiescent cells. Mol Med. 2011;17(11-12):1374-1382. Hofbauer JP, Heyder C, Denk U, et al. Development of CLL in the TCL1 transgenic mouse model is associated with severe skewing of the T-cell compartment homologous to human CLL. Leukemia. 2011;25(9):1452-1458. Aderka D, Englemann H, Hornik V, et al. Increased serum levels of soluble receptors for tumor necrosis factor in cancer patients. Cancer Res. 1991;51(20):5602-5607. Waage A, Liabakk N, Lien E, Lamvik J, Espevik T. p55 and p75 tumor necrosis factor receptors in patients with chronic lymphocytic leukemia. Blood. 1992; 80(10):2577-2583. Digel W, Schoniger W, Stefanic M, et al. Receptors for tumor necrosis factor on neoplastic B cells from chronic lymphocytic leukemia are expressed in vitro but not in vivo. Blood. 1990;76(8):1607-1613. Nakayama S, Yokote T, Tsuji M, et al. TNFalpha receptor 1 expression predicts poor prognosis of diffuse large B-cell lymphoma, not otherwise specified. Am J Surg Pathol. 2014;38(8):1138-1146. Hosono K, Yamada E, Endo H, et al. Increased tumor necrosis factor receptor 1 expression in human colorectal adenomas. World J Gastroenterol. 2012;18(38):53605368. de Miguel MP, Royuela M, Bethencourt FR, Santamaria L, Fraile B, Paniagua R. Immunoexpression of tumour necrosis factor-alpha and its receptors 1 and 2 correlates with proliferation/apoptosis equilibrium in normal, hyperplasic and carcinomatous human prostate. Cytokine. 2000; 12(5):535-538. Mittal AK, Chaturvedi NK, Rai KJ, et al. Chronic lymphocytic leukemia cells in a lymph node microenvironment depict molecular signature associated with an aggressive disease. Mol Med. 2014; 20(1):290-301. Muppidi JR, Tschopp J, Siegel RM. Life and death decisions: secondary complexes and lipid rafts in TNF receptor family signal transduction. Immunity. 2004;21(4):461465. Furman RR, Asgary Z, Mascarenhas JO, Liou HC, Schattner EJ. Modulation of NFkappa B activity and apoptosis in chronic lymphocytic leukemia B cells. J Immunol. 2000;164(4):2200-2206. van Kooten C, Rensink I, Aarden L, van Oers R. Interleukin-4 inhibits both paracrine and autocrine tumor necrosis factor-alpha-induced proliferation of B chronic

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

lymphocytic leukemia cells. Blood. 1992;80 (5):1299-1306. Foa R, Massaia M, Cardona S, et al. Production of tumor necrosis factor-alpha by B-cell chronic lymphocytic leukemia cells: a possible regulatory role of TNF in the progression of the disease. Blood. 1990;76(2):393-400. Coscia M, Pantaleoni F, Riganti C, et al. IGHV unmutated CLL B cells are more prone to spontaneous apoptosis and subject to environmental prosurvival signals than mutated CLL B cells. Leukemia. 2011; 25(5):828-837. Zettlitz KA, Lorenz V, Landauer K, et al. ATROSAB, a humanized antagonistic antitumor necrosis factor receptor one-specific antibody. MAbs. 2010;2(6):639-647. Abe Y, Nomura T, Yoshioka Y, Kamada H, Tsunoda S, Tsutsumi Y. Anti-inflammatory effects of a novel TNFR1-selective antagonistic TNF mutant on established murine collagen-induced arthritis. Adv Exp Med Biol. 2011;691:493-500. Shibata H, Yoshioka Y, Ohkawa A, et al. The therapeutic effect of TNFR1-selective antagonistic mutant TNF-alpha in murine hepatitis models. Cytokine. 2008; 44(2):229-233. Chung H, Jung Y-m, Shin D-H, et al. Anticancer effects of wogonin in both estrogen receptor-positive and -negative human breast cancer cell lines in vitro and in nude mice xenografts. Int J Cancer. 2008; 122(4):816-822. Polier G, Ding J, Konkimalla BV, et al. Wogonin and related natural flavones are inhibitors of CDK9 that induce apoptosis in cancer cells by transcriptional suppression of Mcl-1. Cell Death Dis. 2011;2:e182. Wang W, Guo Q-L, You Q-D, et al. The anticancer activities of wogonin in murine sarcoma S180 both in vitro and in vivo. Biol Pharm Bull. 2006;29(6):1132-1137. Korn SH, Wouters EF, Vos N, JanssenHeininger YM. Cytokine-induced activation of nuclear factor-kappa B is inhibited by hydrogen peroxide through oxidative inactivation of IkappaB kinase. J Biol Chem. 2001;276(38):35693-35700. Baumann S, Fas SC, Giaisi M, et al. Wogonin preferentially kills malignant lymphocytes and suppresses T-cell tumor growth by inducing PLCgamma1- and Ca2+-dependent apoptosis. Blood. 2008; 111(4):2354-2363. Yang L, Zhang HW, Hu R, et al. Wogonin induces G1 phase arrest through inhibiting Cdk4 and cyclin D1 concomitant with an elevation in p21Cip1 in human cervical carcinoma HeLa cells. Biochem Cell Biol. 2009;87(6):933-942. Zhang HW, Yang Y, Zhang K, et al. Wogonin induced differentiation and G1 phase arrest of human U-937 leukemia cells via PKCdelta phosphorylation. Eur J Pharmacol. 2008;591(1-3):7-12.

697

ARTICLE

Chronic Lymphoblastic Leukemia

Ferrata Storti Foundation

Haematologica 2018 Volume 103(4):698-706

Rituximab plus bendamustine or chlorambucil for chronic lymphocytic leukemia: primary analysis of the randomized, open-label MABLE study

Anne-Sophie Michallet,1 Melih Aktan,2 Wolfgang Hiddemann,3 Osman Ilhan,4 Peter Johansson,5 Kamel Laribi,6 Balkis Meddeb,7 Carol Moreno,8 João Raposo,9 Anna Schuh,10 Ali Ünal,11 Tom Widenius,12 Alf Bernhardt,13 Kerstin Kellershohn,14 Dimitri Messeri,13 Stuart Osborne13 and Véronique Leblond15

1 Department of Hematology, CLCC Centre Léon Bérard, Lyon, France; 2Istanbul Medical Faculty, Istanbul University, Turkey; 3Department of Internal Medicine III, LudwigMaximilians University of Munich, Germany; 4Department of Hematology, Ankara University School of Medicine, Turkey; 5Department of Hematology, Uddevalla Hospital, Sweden; 6Department of Hematology, Centre Hospitalier du Mans, Le Mans, France; 7 Hematology Department, Aziza Othmana University Hospital, Tunis, Tunisia; 8 Department of Hematology, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain; 9 Hematology Service, Hospital de Santa Maria, Lisbon, Portugal; 10Department of Oncology, Oxford University Hospitals, UK; 11Department of Hematology, Erciyes University Medical School, Kayseri, Turkey; 12Department of Internal Medicine, Peijas Hospital, Vantaa, Finland; 13Product Development Medical Affairs, F. Hoffmann-La Roche Ltd., Basel, Switzerland; 14Medical Affairs, Roche Pharma AG, Grenzach-Wyhlen, Germany and 15Clinical Hematology, AP-HP Hôpital Pitié-Salpêtrière, UPMC University, Paris, France

ABSTRACT

Correspondence: veronique.leblond@aphp.fr

Received: August 2, 2017. Accepted: January 22, 2018. Pre-published: February 1, 2018. doi:10.3324/haematol.2017.170480 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/698 ©2018 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.

698

ABLE investigated the efficacy and safety of rituximab plus bendamustine or rituximab plus chlorambucil in fludarabine-ineligible patients with chronic lymphocytic leukemia. Patients received rituximab plus bendamustine or rituximab plus chlorambucil every four weeks for six cycles. Rituximab plus chlorambucil-treated patients without a complete response after Cycle 6 received chlorambucil monotherapy for at least six additional cycles or until complete response. The primary endpoint was complete response rate (confirmed by bone marrow biopsy) after Cycle 6 in first-line patients. Secondary endpoints included progression-free survival, overall survival, minimal residual disease, and safety. Overall, 357 patients were randomized (rituximab plus bendamustine, n=178; rituximab plus chlorambucil, n=179; intent-to-treat population), including 241 first-line patients (n=121 and n=120, respectively); 355 patients received treatment (n=177 and n=178, respectively; safety population). In first-line patients, complete response rate after Cycle 6 (rituximab plus bendamustine, 24%; rituximab plus chlorambucil, 9%; P=0.002) and median progression-free survival (rituximab plus bendamustine, 40 months; rituximab plus chlorambucil, 30 months; P=0.003) were higher with rituximab plus bendamustine than rituximab plus chlorambucil. Overall response rate and overall survival were not different. In first-line patients with a complete response, minimal residual disease-negativity was higher with rituximab plus bendamustine than rituximab plus chlorambucil (66% vs. 36%). Overall adverse event incidence was similar (rituximab plus bendamustine, 98%; rituximab plus chlorambucil, 97%). Rituximab plus bendamustine may be a valuable first-line option for fludarabine-ineligible patients with chronic lymphocytic leukemia. clinicaltrials.gov identifier: 01056510

Introduction Rituximab plus fludarabine and cyclophosphamide (R-FC) is standard treatment for medically fit chronic lymphocytic leukemia (CLL) patients,1,2 with high response rates in previously untreated patients (first-line; 1L) and treated patients (secondline; 2L).3-5 However, many CLL patients are elderly and have comorbidities, making them ineligible for fludarabine-based treatment.6 Chemotherapy options for these patients include bendamustine (B) and chlorambucil (Clb). haematologica | 2018; 103(4)

Rituximab plus bendamustine or chlorambucil in CLL

In treatment-naïve CLL patients, phase II studies showed promising efficacy with rituximab plus B (R-B)7 or Clb (R-Clb),8,9 while the phase III CLL11 study demonstrated improved efficacy with R-Clb and obinutuzumab plus Clb (G-Clb) versus Clb monotherapy.10,11 G-Clb also increased progression-free survival (PFS) and complete response (CR) rates versus R-Clb,10,11 although infusionrelated reactions and neutropenia were more common; infection rates were not increased, however.10 While a phase III study demonstrated superior efficacy in terms of CR and PFS with B versus Clb in treatment-naïve CLL,12,13 the activity of R-B versus R-Clb has not been directly compared. Herein, we present results from the randomized, openlabel, multicenter, phase IIIb MABLE study, which aimed to investigate the efficacy and safety of R-B and R-Clb in fludarabine-ineligible CLL patients.

Methods Study design Patients received rituximab (intravenous 375 mg/m2 Day [D] 1, Cycle [C] 1 and 500 mg/m2 D1, C2-C6) plus B (intravenous 90 mg/m2 [1L] or 70 mg/m2 [2L] D1 and D2, C1-C6) or Clb (oral 10 mg/m2 D1-D7, C1-C6) every four weeks for six cycles. R-Clb patients without CR after C6 received Clb monotherapy for ≤6 additional cycles or until CR. After treatment completion, patients were followed every three months for one year, then every six months until data cut-off. Treatment was discontinued if the patient had progressive disease. MABLE was conducted in accordance with the Declaration of Helsinki, Good Clinical Practice guidelines, and local laws, and approved by institutional review boards and ethics committees at participating centers. All patients provided written informed consent.

Patients

Patients were aged ≥18 years, with confirmed CLL requiring treatment as per the International Workshop on CLL (iwCLL) criteria,14 Binet stage B/C disease, Eastern Cooperative Oncology Group (ECOG) performance status ≤2, and investigator assessment of ineligibility for fludarabine-based treatment (Online Supplementary Information). Exclusion criteria included transformation to aggressive B-cell malignancy and previous malignancy within five years of enrollment (unless treated with curative intent). For full inclusion and exclusion criteria see the Online Supplementary Information. During recruitment, the protocol was amended to permit inclusion of patients with progressive Binet stage A disease and to exclude 2L patients (due to slow recruitment). All 2L patients recruited before this amendment were included in the final analysis, as tumor response in this patient subpopulation was a secondary study endpoint; however, as the number of 2L patients enrolled was relatively small, there was insufficient power to show statistically significant differences between study treatments in this patient subpopulation.

Study endpoints The primary endpoint was CR rate (confirmed by bone marrow [BM] biopsy) after C6 in 1L patients. Secondary endpoints included CR rate after C6 in 2L patients, PFS, overall survival (OS), time to next leukemic treatment, minimal residual disease (MRD), and safety. Response was assessed after C3 and C6 as per iwCLL 2008 guidelines.14 Response was also assessed in the R-Clb arm at C12, haematologica | 2018; 103(4)

with treatment being discontinued for patients showing evidence of CR during C7-C12. Assessments are detailed in the Online Supplementary Information.

Safety Adverse events (AEs) were monitored throughout the study and graded according to the National Cancer Institute Common Terminology Criteria for AEs v4.0 and coded according to the Medical Dictionary for Regulatory Activities v17.0.

Statistical analysis Efficacy analyses were conducted on the intent-to-treat (ITT) population (all randomized patients). The safety population included all randomized patients who received treatment. For 1L patients, the between-arm difference in response rates was tested using a one-sided continuity-corrected χ2 test. A twosided continuity-corrected χ2 test assessed between-arm differences in overall response rates (ORRs) and molecular responses. PFS and OS were summarized by Kaplan–Meier estimates and compared via the log-rank test. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated based on the Cox proportional hazard model, with and without baseline Binet stage as a covariate.

Results Patients The study was conducted between 23 February 2010 and 31 March 2014. Of the 357 patients in the ITT population, comprising 241 1L patients (R-B, n=121; R-Clb, n=120) and 116 2L patients (R-B, n=57; R-Clb, n=59), 355 received treatment (R-B, n=177; R-Clb, n=178). Ninetyfive patients (27%) withdrew from treatment during the study (Figure 1). Overall, 92/120 (76.7%) 1L patients treated with R-B and 97/120 (80.8%) 1L patients treated with R-Clb received six cycles of rituximab; 92 (76.7%) and 57 (47.5%) received six cycles of B and Clb, respectively. Twelve (10.0%) patients treated with R-Clb received 12 cycles of Clb. The median number of R, B and Clb doses was six for each. The median (interquartile range) dose of rituximab was 4780.5 mg (4222.5-5346.5) in 1L patients treated with R-B and 5028.5 mg (4546.0-5349.0) in 1L patients treated with R-Clb. In total, 6/121 (5.0%) patients treated with R-B and 2/120 (1.7%) patients treated with RClb had a reduction or delay in their treatment schedule due to treatment-emergent toxicities. Baseline characteristics were balanced in the 1L population (Table 1; baseline characteristics for all patients are presented in Online Supplementary Table S1). Deletion of 17p was not an exclusion criterion due to a lack of efficacy data relating to 17p deletion for the treatment combinations used at the time of study design. Thus, 13 1L patients (R-B, 10; R-Clb, 3) with 17p deletion were included. Median follow-up was 23.5 months (R-B) and 23.3 months (R-Clb). Median age in 1L patients was 72 years in both treatment arms; the majority of patients were aged 65 years or more. The median number of comorbidities (active medical conditions) in 1L patients was three in both arms (Table 1); the most common comorbidities were vascular disorders and metabolism disorders affecting 49% and 37% of patients, respectively (Online Supplementary Table S2). The great majority of all patients in the study (including 2L patients) used concomitant medication during the study (R-B, 96%; R-Clb, 94%). 699

A.-S. Michallet et al. Table 1. Demographic characteristics for patients receiving 1L therapy.

1L therapy

Age (years) Median (min, max) â&#x2030;Ľ65 years, n (%) â&#x2030;Ľ75 years, n (%) Sex Male, n (%) Female, n (%) Active medical conditions, n Median (min, max) Binet stage, n (%) A B C Missing ECOG PS, n (%) 0 1 2 Missing Body surface area, m2 Mean (SD) Min, max IGVH mutational status, n (%) Mutated Unmutated Othera Not tested 11q status, n (%) Heterozygous deletion Normal Not tested 17p status, n (%) Heterozygous deletion Normal Not tested 11q/17p deletion, n (%) Heterozygous deletion Normal Not tested 13q deletion (S25 or S319 probe)b, n (%) Homozygous deletion Two clones (one homozygote, one heterozygote) Heterozygous deletion Normal Not tested Trisomy 12, n (%) Trisomy Normal Not tested

R-B (N=121)

R-Clb (N=120)

72 (41, 86) 86 (71) 45 (37)

72 (38, 91) 90 (75) 44 (37)

70 (58) 51 (42)

80 (67) 40 (33)

3 (0, 12)

3 (0, 18)

6 (5) 73 (60) 37 (31) 5 (4)

8 (7) 66 (55) 43 (36) 3 (3)

62 (51) 50 (41) 9 (7) 0

59 (49) 51 (43) 8 (7) 2 (2)

1.811 (0.2382) 1.30, 2.48

1.807 (0.1706) 1.41, 2.29

41 (34) 73 (60) 3 (3) 4 (3)

46 (38) 59 (49) 8 (7) 7 (6)

24 (20) 96 (79) 1 (1)

19 (16) 99 (83) 2 (2)

10 (8) 110 (91) 1 (1)

3 (3) 114 (95) 3 (3)

32 (26) 88 (73) 1 (1)

22 (18) 96 (80) 2 (2)

3 (3) 15 (12) 42 (35) 61 (50) 1 (1)

1 (1) 6 (5) 5 (4) 60 (50) 2 (2)

30 (25) 90 (74) 1 (1)

19 (16) 99 (83) 2 (2)

a Other includes polyclonal and oligoclonal. bDeletion status according to at least one probe (NB, in the R-B group, one patient with two clones by S319 probe and heterozygous deletion by S25 probe is counted twice). 1L: first-line; ECOG PS: Eastern Cooperative Oncology Group performance status; R-B: rituximab plus bendamustine; R-Clb: rituximab plus chlorambucil; SD: standard deviation.

700

haematologica | 2018; 103(4)

Rituximab plus bendamustine or chlorambucil in CLL

Colony-stimulating factors were taken more frequently by patients treated with R-B (111/178 [62.4%] vs. 71/179 [39.7%] for R-Clb). ECOG performance status scores were ≥1 in approximately half the study population. At baseline, 108/177 (61.0%), 56/177 (31.6%) and 4/177 (2.3%) patients treated with R-B, and 102/178 (57.3%), 60/178 (33.7%) and 3/178 (1.7%) patients treated with R-Clb had normal; abnormal, non-clinically significant; and abnormal clinically significant calculated creatinine clearance, respectively.

Efficacy In 1L patients, the CR rate after C6 was higher with RB versus R-Clb (24% [n=29/121] vs. 9% [n=11/120];

P=0.002; Table 2). Logistic regression analysis supported the R-B treatment effect after adjusting for baseline covariates (odds ratio 4.18, 95% CI 1.77-9.87; P=0.001). None of the covariates had a statistically significant impact on the CR rate. ORRs (based on the investigator’s assessment) at the end of rituximab treatment were similar for R-B and R-Clb (91% vs. 86%; P=0.304). The proportion of patients with stable disease (3% vs. 6%) and progressive disease (3% vs. 2%) at the end of treatment were also similar for R-B vs. R-Clb, respectively. A statistically significant ten-month extension in median PFS was observed with R-B versus R-Clb (39.6 vs. 29.9 months; HR [adjusted for baseline Binet stage] 0.523, 95%

Figure 1. Patient disposition. Numbers in parentheses represent the number of patients from the 1L subpopulation. In total, 118 patients (33%) discontinued the study prematurely, due to death (R-B 16%; R-Clb 19%), patient lost to follow-up (3% per arm), investigator decision (R-B 2%; R-Clb 3%), patient withdrew consent (R-B 3%; R-Clb 2%), patient non-compliance (R-B 1%; R-Clb 2%), and ‘other’ reasons (R-B 6%; R-Clb 7%). Reason for withdrawal was not available for one patient (R-Clb). AE: adverse event; C: cycle; N: number of patients; PD: progressive disease; R-B: rituximab plus bendamustine; R-Clb: rituximab plus chlorambucil.

Table 2. CR and PRs at C6 in 1L patients.

Assessment CR confirmed by BM biopsya

Analysis CR

Logistic regressionc

PR based on the investigator’s assessment

N n (%) P-valueb N OR (95% CI) P-valued N n (%)

R-B

R-Clb

121 29 (24)

120 11 (9) 0.002

113

103 4.18 (1.77-9.87) 0.001

121 60 (50)

120 79 (66)

a CRs confirmed by BM biopsy only were included. bP-value is based on a one-sided continuity corrected χ2 test. cThe following covariates were included in the logistic regression: Binet stage (A and B vs. C); IGVH mutational status (mutated vs. unmutated); 17p/11q deletion (heterozygote deletion vs. normal); ECOG PS (0 vs. ≥1). dP-value is based on the Wald test. BM: bone marrow; CI: confidence interval; CR: complete response; OR: odds ratio; PR: partial response; R-B: rituximab plus bendamustine; R-Clb: rituximab plus chlorambucil.

haematologica | 2018; 103(4)

701

A.-S. Michallet et al. Table 3. MRD negativity at the confirmation-of-response visita in 1L patients. Overall ITT population MRD-negative patients, n (%) Patients with CR MRD-negative patients, n (%) Patients with CR or PR based on the investigator’s assessmentb MRD-negative patients, n (%)

R-B

R-Clb

(n=121) 49 (41) (n=29) 19 (66)

(n=120) 16 (13) (n=11) 4 (36)

(n=89) 47 (53)

(n=90) 16 (18)

Performed 12 weeks after the end of C6 disease response assessment. bIncludes patients with CR (with or without BM confirmation) or PR by investigator assessment. Negative MRD was defined as proportion of malignant B cells to white blood cells of <10-4, as assessed by allele-specific oligonucleotide real-time quantitative polymerase chain reaction assay measured in BM aspirate (or PB if BM unavailable). MRD data were available for 45/50 patients with a CR based on the investigator’s assessment (BM, n=42; PB, n=1; unknown=2) and 182/241 patients overall (BM, n=145; PB, n=32; unknown=5). CR: complete response; ITT: intent-to-treat; MRD: minimal residual disease; PR: partial response; RB: rituximab plus bendamustine; R-Clb: rituximab plus chlorambucil.

CI 0.339-0.806; P=0.003; Figure 2A); median OS was not significantly different (43.8 months vs. not reached; HR [adjusted for baseline Binet stage] 0.975, 95% CI 0.5051.880; P=0.939; Figure 2B). During the study, 11/121 (9.1%) patients treated with R-B and 22/120 (18.3%) patients treated with R-Clb had a documented intake of any new leukemia treatment. Due to the low numbers, the median time to next leukemic treatment could not be calculated for either treatment arm (log-rank test for comparison between treatment arms: P=0.037). MRD data were available for 45/50 patients (90%) who had a CR based on the investigator’s assessment and 182/241 patients (76%) overall. BM aspirates were available for 42/45 patients (93%) and 145/182 patients (80%), respectively. MRD-negativity rates at the confirmation-ofresponse visit (ITT population) were higher for R-B than for R-Clb (41% vs. 13%; Table 3). In 1L patients with a CR after C6, MRD-negativity rates at the confirmation-ofresponse visit were higher in the R-B group than in the RClb group (66% vs. 36%). A similar pattern was seen in those with a CR or PR according to investigator’s assessment (53% vs. 18%). Efficacy results in 2L patients are presented in the Online Supplementary Information.

Safety Safety results are presented for the pooled population (1L and 2L patients). AEs were similar between arms (R-B, 98%; R-Clb, 97%; Table 4). The most common AEs by System Organ Class (SOC) were ‘blood and lymphatic system disorders’ (R-B, 75%; R-Clb, 64%); the most commonly reported AE was neutropenia (R-B, 56%; R-Clb, 49%). AEs in the SOC ‘skin and subcutaneous tissue disorders’ were more frequent in the R-B versus R-Clb arm (36% vs. 23%), driven by a higher incidence of rash (16% vs. 5%). Grade ≥3 AEs were higher with R-B (75%) than R-Clb (64%), mainly due to a higher incidence of serious AEs (SAEs) of the SOC ‘infections and infestations’. The most common grade ≥3 AEs were of the SOC ‘blood and lymphatic system disorders’ (R-B, 56%; R-Clb, 47%). SAEs were experienced by 41% (R-B) and 32% (R-Clb) of patients, and were most frequently of the SOC ‘infections and infestations’ (R-B, 19%; R-Clb, 8%; Table 4). Rituximab-related AEs were experienced by 81% (R-B) and 73% (R-Clb) of patients. B/Clb-related AEs were 702

reported for 92% (R-B) and 80% (R-Clb) of patients. Drugrelated AEs (rituximab, B, and Clb) were most commonly of the SOC ‘blood and lymphatic system disorders’. AEs leading to rituximab discontinuation were experienced by 18% (R-B) and 11% (R-Clb) of patients. AEs leading to discontinuation of B or Clb were reported for 19% and 18% of patients, respectively. In 1L patients, AEs leading to treatment discontinuation were experienced by 22 R-B patients (18%) and 14 R-Clb patients (12%). Overall, 65 patients died (R-B, n=30, 17%; R-Clb, n=35, 20%) due to CLL (R-B, n=14, 8%; R-Clb, n=20, 11%) and AEs (R-B, n=16, 9%; R-Clb, n=14, 8%). Cause of death was missing for one patient (R-Clb arm). AEs leading to death were infection (n=4 per arm), acute myeloid leukemia or myelodysplastic syndrome (n=1 per arm), other neoplasm (R-Clb, n=1), and other causes (R-B, n=11; R-Clb, n=8). Treatment-related AEs leading to death included thrombocytopenia, neutropenic sepsis and febrile neutropenia (n=1 each) in the 1L population and multi-organ failure, pneumonia, acute myeloid leukemia and sepsis (n=1 each) in the 2L population.

Discussion R-FC is the standard 1L treatment for medically-fit patients with CLL. However, this regimen can cause significant myelosuppression and high rates of early and late infections, especially in elderly patients15 who may have comorbidities and be considered ineligible for fludarabinebased treatment.6 In MABLE, the efficacy and safety of RB and R-Clb were investigated in fludarabine-ineligible CLL patients. B is an established treatment for CLL, and a previous phase III study of B versus Clb in treatment-naïve patients demonstrated improved CR rates and median PFS with B.12,13 R combined with chemotherapeutic agents prolongs OS in previously untreated, medically fit CLL patients.16,17 In unfit patients, survival was improved by the addition of G to Clb.10 Among 1L patients in MABLE, the rates of CR and of CR with MRD-negativity were higher for R-B than for RClb. ORRs were similar in the two arms. Whereas the CR rate in 1L R-Clb patients in MABLE (9%) was comparable with that in R-Clb patients in the phase III CLL11 study (7%),10 the ORR for R-Clb-treated patients was higher in MABLE than in CLL11 (86% vs. 58%). Two explanations for this might be the different Clb doses used (MABLE, 10 haematologica | 2018; 103(4)

Rituximab plus bendamustine or chlorambucil in CLL

Figure 2. Efficacy in 1L patients. (A) PFS and (B) OS. CI: confidence interval; HR: hazard ratio; NR: not reached; OS: overall survival; PFS: progression-free survival; R-B: rituximab plus bendamustine; R-Clb: rituximab plus chlorambucil.

mg/m2; CLL11, 0.5 mg/kg), which resulted in a higher median cumulative dose (MABLE, 720 mg; CLL11, 366400 mg), and differences in the study populations, with patients in MABLE having fewer active comorbidities than those in CLL11 (medians of 3 and 5, respectively) and a better performance status.10 Earlier phase II studies in elderly CLL patients treated 1L with R-Clb reported CR rates of 10-17% and ORRs of 82-84%.8,9 The CLL2M trial, a phase II study of CLL patients treated 1L with R-B, at the same B dosage as the current study, reported a CR rate of 23% and an ORR of 88%.7 In 1L patients in the study reported herein, the median PFS of 39.6 months in the R-B arm was significantly longer (by 10 months) than the value in the R-Clb arm. This result is similar to the median PFS of 43.2 months achieved in fit CLL patients treated with R-B in the CLL10 study,18 and is consistent with the findings of the CLL2M study, which enrolled fit and unfit 1L patients and reported a median event-free survival in R-B-treated patients of 33.9 months.7 The CLL2M study did not select patients by fitness, but a substantial proportion could be considered unfit based on age, Binet stage, and renal impairment.7 In MABLE, 33% of the 1L patients were Binet stage C and the median age was 72 years. The PFS result in the R-Clb arm in MABLE was almost twice as long as that in CLL11 (29.9 vs. 15.4 months),11 however, as noted above, a direct comparison of these studies is limited by differences in Clb doses and patient fitness. haematologica | 2018; 103(4)

In MABLE, 1L patients with CR had higher MRD-negativity rates with R-B versus R-Clb, indicating a greater depth of response with R-B. Of note, MRD was assessed primarily in BM from patients with CR based on the investigatorâ&#x20AC;&#x2122;s assessment, whereas in CLL11 and previous phase II studies, MRD was measured in peripheral blood ([PB] or BM) from all patients.7,8,10 One phase II study reported a MRD-negativity rate of 12.5% (2/16 patients) using BM aspirates from R-Clb-treated patients who achieved a CR/unconfirmed CR.9 Median OS was not significantly different between treatment arms in 1L patients in MABLE and was not reached in previous phase II studies of R-B and R-Clb,7-9 or in the CLL11 and COMPLEMENT-1 studies.11,19 At current follow-up reported for CLL11 and COMPLEMENT-1, no significant OS benefit was observed for G-Clb versus R-Clb or ofatumumab plus Clb (Ofa-Clb) versus Clb, respectively,11,19 whereas a significant improvement was observed for G-Clb versus Clb alone.10 Further observation is required to determine if there is an OS benefit with G-Clb versus R-Clb or Ofa-Clb versus Clb. Safety profiles (for pooled 1L and 2L patients) were similar for R-Clb and R-B, with no new or clinically relevant safety signals, and events were as expected for CLL patients receiving immunochemotherapy.7-9 Incidences of all-grade AEs, SAEs, and treatment-related AEs were similar across arms. Grade â&#x2030;Ľ3 AE incidence was slightly higher with R-B versus R-Clb, driven by a higher incidence of 703

A.-S. Michallet et al. Table 4. Summary of AEs (safety population).

Patients, n (%)

R-B (N=177)

R-Clb (N=178)

All-grade AEs Grade ≥3 AEs SAEs Most common all-grade AEsa Blood and lymphatic system disorders Neutropenia Leukopenia Anemia Thrombocytopenia Lymphopenia Gastrointestinal disorders Nausea Diarrhea Constipation General disorders and administrative site conditions Pyrexia Asthenia Skin and subcutaneous tissue disorders Rash Most common grade ≥3 AEsb Blood and lymphatic system disorders Neutropenia Leukopenia Anemia Lymphopenia Thrombocytopenia Febrile neutropenia Most common SAEsb Infections and infestations Pneumonia Blood and lymphatic system disorders Febrile neutropenia

173 (98) 132 (75) 73 (41)

173 (97) 113 (64) 56 (32)

133 (75) 99 (56) 42 (24) 41 (23) 37 (21) 30 (17) 99 (56) 53 (30) 30 (17) 28 (16) 93 (53) 37 (21) 29 (16) 63 (36) 29 (16)

113 (64) 88 (49) 31 (17) 27 (15) 44 (25) 21 (12) 90 (51) 46 (26) 22 (12) 23 (13) 87 (49) 17 (10) 34 (19) 40 (23) 9 (5)

99 (56) 76 (43) 29 (16) 18 (10) 17 (10) 17 (10) 12 (7)

84 (47) 65 (37) 15 (8) 12 (7) 10 (6) 16 (9) 7 (4)

33 (19) 8 (5) 25 (14) 11 (6)

15 (8) 2 (1) 15 (8) 7 (4)

Preferred terms with incidence of ≥15% in either study arm. bPreferred terms with incidence of ≥5% in either study arm. Non-serious AEs were reported until 28 days after the end of the last treatment cycle. SAEs unrelated to study treatment were reported until six months after the end of treatment or until the start of new anti- chronic lymphocytic leukemia treatment. Treatment-related SAEs were to be reported indefinitely. AE: adverse event; R-B: rituximab plus bendamustine; R-Clb: rituximab plus chlorambucil; SAE: serious adverse event.

infections and infestations. Few patients in either arm discontinued therapy due to AEs. Treatment withdrawal due to AEs was reported for 18% of patients receiving 1L R-B and 12% receiving R-Clb. As previously noted, the results of the CLL11 study show that, in unfit CLL patients, G-Clb was associated with higher response rates and longer PFS than R-Clb, with more frequent MRD eradication and an acceptable toxicity profile.10 In addition, current guidelines, (European Society for Medical Oncology and National Comprehensive Cancer Network), recommend Clb in combination with an anti-CD20 antibody as standard 1L therapy in unfit CLL patients.20,21 However, in MABLE and previous studies, R-B was associated with a good response rate and improved PFS compared with R-Clb. Currently, evidence to guide the choice between R-B and G-Clb in 1L unfit patients with CLL is limited, although a recent meta704

analysis of PFS and OS results in five studies showed a trend towards better efficacy for G-Clb than R-B;22 however, the difference was not significant despite a significant difference between G-Clb and other comparators such as R-Clb, Ofa-Clb, Clb, and fludarabine. A randomized trial comparing these combinations could resolve this question. Some limitations of our study should be acknowledged. Since we relied only on investigator assessments of tumor response to evaluate efficacy, and did not include assessments by an independent review committee, this might have introduced a potential bias in the efficacy results. Comparison of our results with those of other studies in CLL patients is potentially complicated because the selection of patients based on fitness was based on a judgment made by the investigator that the patients were not eligible for fludarabine, according to a set of pre-defined critehaematologica | 2018; 103(4)

Rituximab plus bendamustine or chlorambucil in CLL

ria that were based on the prescribing information for fludarabine at the time of study design; the Cumulative Illness Rating Scale scoring was not used. However, a recent randomized study comparing the efficacy of ibrutinib with Clb used an age cut-off of 65 years as the only criterion for selecting “older” patients, and did not carry out any assessment of comorbidities.23 Although the efficacy of immunochemotherapy combinations such as R-B, Ofa-Clb, and G-Clb in CLL have been shown,7,11,19 obinutuzumab plus B (G-B) needs to be evaluated further, although it is likely that future clinical studies will focus more on the combination of anti-CD20 antibodies with novel agents such as ibrutinib, idelalisib, and venetoclax. Indeed, these agents have shown promising efficacy when combined with anti-CD20 monoclonal antibodies (including rituximab) in the relapsed/refractory CLL setting,24-29 and are now being evaluated in 1L.23 Furthermore, additional safety data for 1L B-cell receptor (BCR) signaling inhibitors are required. In March 2016, the US Food and Drug Administration alerted healthcare professionals about safety concerns with idelalisib when used in combination with rituximab or R-B in CLL and follicular lymphoma. This was related to a high rate of viral and fungal infections that led the sponsor to discontinue six trials.30 However, given the economic burden associated with new agents and the fact that BCR inhibitors are not yet available in many European countries, immunochemotherapy combinations are likely to continue to be valuable treatment options for 1L patients with CLL.31 In conclusion, in fludarabine-ineligible CLL patients, 1L R-B treatment significantly improved CR rates and median PFS versus R-Clb, and increased MRD-negativity rates, with no new safety signals reported. R-B may be a valuable 1L option for fludarabine-ineligible CLL patients and this combination continues to be widely used in clinical practice in Europe, reinforcing the interest of this large randomized study. List of MABLE investigators and centers Seppo Vanhatalo, Satakunta Central Hospital, Pori, Finland; Kimmo Porkka, HUS/Medisiininen tulosyksikkö/Hematologian klinikka, Haartmaninkatu, Finland; Anne-Sophie Michallet, Centre Hospitalier Lyon Sud, Pierre-Bénite, France; Bruno Royer, Hôpital Sud, Salouel, France; Olivier Boulat, Centre Hospitalier Henri Duffaut, Avignon, France; Christian Berthou, Hôpital Morvan – CHU Brest, France; Dominique Bordessoule, CHU de Limoges, France; Jean-Claude Eisenmann, Centre Hospitalier de Mulhouse, France; Jean-Michel Karsenti, Hôpital Archet 1, Nice, France; Eric Jourdan, CHU de Nîmes, France; Véronique Leblond, Hôpital de la Pitié-Salpêtrière, Paris, France; Laurence Sanhes, Hôpital Saint Jean, Perpignan, France; Mourad Tiab, Centre Hospitalier Départemental Vendée, La Roche Sur Yon, France; Kamel Laribi, Centre Hospitalier du Mans, Le Mans, France; Régis Costello, Hôpital la Conception, Marseille, France; Lysiane Molina, Hôpital Nord – Michallon, La Tronche, France; Laurent Sutton, CH Victor Dupouy, Argenteuil, France; Abderrazak El Yamani, Centre Hospitalier de Blois, France; João Raposo, Centro Hospitalar de Lisboa Norte, Lisbon, Portugal; Emília Cortesão, Centro Hospitalar e Universitário de Coimbra, Portugal; Cristina João, Instituto Português de Oncologia de Lisboa Francisco Gentil, Lisbon, Portugal; Joaquim Andrade, Hospital de São João, Oporto, Portugal; Ângelo Martins, Instituto Português de Oncologia do Porto Francisco Gentil, haematologica | 2018; 103(4)

Oporto, Portugal; José Antonio García Marco, Hospital Universitario Puerta de Hierro, Madrid, Spain; Marcos González Díaz, Hospital Universitario de Salamanca, Spain; Francisco Javier de la Serna Torroba, Hospital Universitario 12 de Octubre, Madrid, Spain; Carolina Moreno Atanasio, Hospital de la Sant Creu i Sant Pau, Barcelona, Spain; José María Moraleda Jiménez, Hospital Universitario Virgen de la Arrixaca, Murcia, Spain; M. Pilar Giraldo Castellano, Hospital Miguel Servet, Zaragoza, Spain; Macarena Ortiz Pareja, Hospital Regional Carlos Haya, Málaga, Spain; Alicia Rodríguez Fernández, Hospital Virgen Macarena, Sevilla, Spain; José Ignacio Olalla Antolín, Hospital Sierrallana, Cantabria, Spain; Emilio Montserrat Costa, Hospital Clínic de Barcelona, Spain; Christelle Ferra Coll, Hospital Universitari Germans Trias i Pujol, Barcelona, Spain; Kristina Wallman, Falu Iasarett, Falun, Sweden; Birgitta Lauri, Sunderby Sjukhus, Luleå, Sweden; Maria Strandberg, Sundsvall County Hospital, Sweden; Peter Johansson, Uddevalla County Hospital, Sweden; Honar Cherif, The Academic Hospital, Uppsala, Sweden; Alicja Markuszewska-Kuczynska, Norrlands Universietetssjukhus, Umeå, Sweden; Lars Timberg, Medicinkliniken, Kristianstad, Sweden; Balkis Meddeb, Hôpital Aziza Othmana, Tunis, Tunisia; Ilhan Osman, Ankara University Medical Faculty, Turkey; Filiz Vural and Seckin Cagirgan, Ege University Medical Faculty, Izmir, Turkey; Yagci Munci, Gazi University Medical Faculty, Ankara, Turkey; Undar Buient, Dokuzeylul University Medical Faculty, Izmir, Turkey; Nilgun Sayinalp, Hacettepe University Medical Faculty, Ankara, Turkey; Melih Aktan, Istanbul University Medical Faculty, Turkey; Mehmet Turgut, Ondokuzmayis University Medical Faculty, Samsun, Turkey; Muzaffer Demir, Trakya University Medical Faculty, Edirne, Turkey; Ali Unal, Erciyes Universitesi Medical Faculty, Kayseri, Turkey; Zafer Gulbas, Anadolu Health Center, Kocaeli, Turkey; Marion Macheta, Blackpool Victoria Hospital, UK; Stephen Devereux, Kings College Hospital, London, UK; Anna Schuh, Oxford Radcliffe Hospitals, UK; Adrian Bloor, Christie NHS Foundation Trust, Manchester, UK; Saad Rassam, Maidstone and Tunbridge Wells NHS Trust, Maidstone, UK; Julie Blundell, Royal Cornwall Hospital, Truro, UK; Renata Walewska, The Royal Bournemouth & Christchurch Hospital NHS Foundation Trust, Bournemouth, UK; Claire Hemmaway, Queens Hospital, Romford, UK; Peter Hillmen, St. James’ University Hospital, Leeds, UK. Acknowledgments The authors would like to thank the patients and their families, and the study investigators, study coordinators, and nurses who assisted with the rituximab clinical program. We would also like to thank Mundipharma for provision of bendamustine for use in this study. AS is supported by the Oxford Partnership Comprehensive Biomedical Research Centre with funding from the Department of Health’s National Institute of Health Research (NIHR) Biomedical Research Centre funding scheme. The views expressed in this publication are those of the authors and not necessarily those of the UK Department of Health. MABLE was sponsored by F. Hoffmann-La Roche Ltd, with provision of bendamustine and financial support from Mundipharma. Thirdparty medical writing assistance, under the direction of the authors, was provided by Susan Browne, PhD, of GardinerCaldwell Communications (Macclesfield, UK) and funded by F. Hoffmann-La Roche Ltd. Funding This work was funded by F. Hoffmann-La Roche Ltd. 705

A.-S. Michallet et al.

References 1. Roche Registration Ltd [Internet]. MabThera: Summary of product characteristics; Updated 2017 Nov 6 [cited 2018 Jan 9]. Available from: http://www.ema.europa.eu/docs/en_GB/d ocument_library/EPAR__Product_Information/human/000165/WC 500025821.pdf 2. Genentech Inc [Internet]. Rituxan: Highlights of prescribing information; Updated 2016 Apr [cited 2018 Jan 9]. Available from: https://www. gene.com/download/pdf/rituxan_prescribing.pdf. 3. Keating MJ, O'Brien S, Albitar M, et al. Early results of a chemoimmunotherapy regimen of fludarabine, cyclophosphamide, and rituximab as initial therapy for chronic lymphocytic leukemia. J Clin Oncol. 2005;23(18):4079-4088. 4. Robak T, Dmoszynska A, Solal-Céligny P, et al. Rituximab plus fludarabine and cyclophosphamide prolongs progressionfree survival compared with fludarabine and cyclophosphamide alone in previously treated chronic lymphocytic leukemia. J Clin Oncol. 2010;28(10):1756-1765. 5. Badoux XC, Keating MJ, Wang X, et al. Fludarabine, cyclophosphamide, and rituximab chemoimmunotherapy is highly effective treatment for relapsed patients with CLL. Blood. 2011;117(11):3016-3024. 6. Shah N, Tam C, Seymour JF, Rule S. How applicable is fludarabine, cyclophosphamide and rituximab to the elderly? Leuk Lymphoma. 2015;56(6):1599-1610. 7. Fischer K, Cramer P, Busch R, et al. Bendamustine in combination with rituximab for previously untreated patients with chronic lymphocytic leukemia: a multicenter phase II trial of the German Chronic Lymphocytic Leukemia Study Group. J Clin Oncol. 2012;30(26):3209-3216. 8. Hillmen P, Gribben JG, Follows GA, et al. Rituximab plus chlorambucil as first-line treatment for chronic lymphocytic leukemia: final analysis of an open-label phase II study. J Clin Oncol. 2014;32(12):1236-1241. 9. Foà R, Del Giudice I, Cuneo A, et al. Chlorambucil plus rituximab with or without maintenance rituximab as first-line treatment for elderly chronic lymphocytic leukemia patients. Am J Hematol. 2014;89(5):480-486. 10. 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. 11. Goede V, Fischer K, Engelke A, et al. Obinutuzumab as frontline treatment of chronic lymphocytic leukemia: updated results of the CLL11 study. Leukemia.

706

2015;29(7):1602-1604. 12. Knauf WU, Lissichkov T, Aldaoud A, et al. Phase III randomized study of bendamustine compared with chlorambucil in previously untreated patients with chronic lymphocytic leukemia. J Clin Oncol. 2009;27(26):4378-4384. 13. Knauf WU, Lissitchkov T, Aldaoud A, et al. Bendamustine compared with chlorambucil in previously untreated patients with chronic lymphocytic leukaemia: updated results of a randomized phase III trial. Br J Haematol. 2012;159(1):67-77. 14. Hallek M, Cheson BD, Catovsky D, et al. Guidelines for the diagnosis and treatment of chronic lymphocytic leukemia: a report from the International Workshop on Chronic Lymphocytic Leukemia updating the National Cancer Institute-Working Group 1996 guidelines. Blood. 2008;111(12):5446-5456. 15. Laurenti L, Innocenti I, Autore F, et al. Bendamustine in combination with rituximab for elderly patients with previously untreated B-cell chronic lymphocytic leukemia: a retrospective analysis of reallife practice in Italian hematology departments. Leuk Res. 2015;39(10):1066-1070. 16. Tam CS, O'Brien S, Wierda W, et al. Longterm results of the fludarabine, cyclophosphamide, and rituximab regimen as initial therapy of chronic lymphocytic leukemia. Blood. 2008;112(4):975-980. 17. Hallek M, Fischer K, Fingerle-Rowson G, et al. Addition of rituximab to fludarabine and cyclophosphamide in patients with chronic lymphocytic leukaemia: a randomised, open-label, phase 3 trial. Lancet. 2010;376(9747):1164-1174. 18. Eichhorst B, Fink AM, Busch R, et al. Frontline chemoimmunotherapy with fludarabine (F), cyclophosphamide (C), and rituximab (R) (FCR) shows superior efficacy in comparison to bendamustine (B) and rituximab (BR) in previously untreated and physically fit patients (pts) with advanced chronic lymphocytic leukemia (CLL): final analysis of an international, randomized study of the German CLL Study Group (GCLLSG) (CLL10 Study). Blood. 2014;124(21):19. 19. Hillmen P, Robak T, Janssens A, et al. Chlorambucil plus ofatumumab versus chlorambucil alone in previously untreated patients with chronic lymphocytic leukaemia (COMPLEMENT 1): a randomised, multicentre, open-label phase 3 trial. Lancet. 2015;385(9980):1873-1883. 20. Eichhorst B, Robak T, Montserrat E, et al. Chronic lymphocytic leukaemia: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2015;26(suppl 5):v78-84. 21. NCCN [Internet]. NCCN Guidelines for patients: chronic lymphocytic leukemia [cited 2018 Jan 11]. Available from: https://www.nccn.org/patients/guide-

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

lines/cll/files/assets/common/downloads/fi les/cll.pdf. Städler N, Shang A, Bosch F, et al. A systematic review and network meta-analysis to evaluate the comparative efficacy of interventions for unfit patients with chronic lymphocytic leukemia. Adv Ther. 2016;33(10):1814-1830. Burger JA, Tedeschi A, Barr PM, et al. Ibrutinib as initial therapy for patients with chronic lymphocytic leukemia. N Engl J Med. 2015;373(25):2425-2437. Burger JA, Keating MJ, Wierda WG, et al. Safety and activity of ibrutinib plus rituximab for patients with high-risk chronic lymphocytic leukaemia: a single-arm, phase 2 study. Lancet Oncol. 2014;15(10):1090-1099. Chanan-Khan A, Cramer P, Demirkan F, et al. HELIOS investigators. Ibrutinib combined with bendamustine and rituximab compared with placebo, bendamustine, and rituximab for previously treated chronic lymphocytic leukaemia or small lymphocytic lymphoma (HELIOS): a randomised, double-blind, phase 3 study. Lancet Oncol. 2016;17(2):200-211. Furman RR, Sharman JP, Coutre SE, et al. Idelalisib and rituximab in relapsed chronic lymphocytic leukemia. N Engl J Med. 2014;370(11):997-1007. Roberts AW, Ma S, Brander D, et al. Venetoclax (ABT-199/GDC-0199) combined with rituximab induces deep responses in patients with relapsed/refractory chronic lymphocytic leukemia. Haematologica. 2015;100(suppl 1):154(abstr S431). Sharman JP, Coutre SE, Furman RR, et al. Second interim analysis of a phase 3 study of idelalisib (ZYDELIG®) plus rituximab (R) for relapsed chronic lymphocytic leukemia (CLL): efficacy analysis in patient subpopulations with del(17p) and other adverse prognostic factors. Blood. 2014;124(21):330. Robak T, Wach M, Jones JA, et al. Results of a phase 3 randomized controlled study evaluating the efficacy and safety of idelalisib (idela) in combination with ofatumumab (ofa) for previously treated chronic lymphocytic leukemia (CLL). Haematologica. 2015;100(suppl 1):229(abstr LB598). US Food and Drug Administration [Internet]. FDA alerts healthcare professionals about clinical trials with zydelig (idelalisib) in combination with the other cancer medicines. Drug Safety and Availability; Updated 2016 Mar 14 [cited 2017 Jan 9]. Available from: http://www.fda.gov/Drugs/Drugsafety/uc m490618.htm. Chen Q, Jain N, Ayer T, et al. Economic burden of chronic lymphocytic leukemia in the era of oral targeted therapies in the United States. J Clin Oncol. 2017;35(2):166-174.

haematologica | 2018; 103(4)

ARTICLE

Plasma Cell Disorders

Modeling multiple myeloma-bone marrow interactions and response to drugs in a 3D surrogate microenvironment

Ferrata Storti Foundation

Daniela Belloni,1 Silvia Heltai, 1 Maurilio Ponzoni, 2,3 Antonello Villa, 4 Barbara Vergani,4 Lorenza Pecciarini,2 Magda Marcatti,5 Stefania Girlanda,5 Giovanni Tonon,6 Fabio Ciceri,3,5 Federico Caligaris-Cappio,1,3,7 Marina Ferrarini1* and Elisabetta Ferrero1*

1 Division of Experimental Oncology, IRCCS San Raffaele Scientific Institute; 2Pathology Unit, IRCCS San Raffaele Scientific Institute; 3Vita-Salute San Raffaele University; 4 Consorzio MIA, University of Milano-Bicocca; 5Hematology, IRCCS San Raffaele Scientific Institute; 6Functional Genomics of Cancer Unit, Division of Experimental Oncology, San Raffaele Scientific Institute and 7AIRC, Milan, Italy

Haematologica 2017 Volume 103(4):707-716

*MF and EF contributed equally to this work

ABSTRACT

ultiple myeloma develops primarily inside the bone marrow microenvironment, that confers pro-survival signals and drug resistance. 3D cultures that reproduce multiple myeloma-bone marrow interactions are needed to fully investigate multiple myeloma pathogenesis and response to drugs. To this purpose, we exploited the 3D Rotary Cell Culture System bioreactor technology for myelomabone marrow co-cultures in gelatin scaffolds. The model was validated with myeloma cell lines that, as assessed by histochemical and electronmicroscopic analyses, engaged contacts with stromal cells and endothelial cells. Consistently, pro-survival signaling and also cell adhesionmediated drug resistance were significantly higher in 3D than in 2D parallel co-cultures. The contribution of the VLA-4/VCAM1 pathway to resistance to bortezomib was modeled by the use of VCAM1 transfectants. Soluble factor-mediated drug resistance could be also demonstrated in both 2D and 3D co-cultures. The system was then successfully applied to co-cultures of primary myeloma cells-primary myeloma bone marrow stromal cells from patients and endothelial cells, allowing the development of functional myeloma-stroma interactions and MM cell long-term survival. Significantly, genomic analysis performed in a highrisk myeloma patient demonstrated that culture in bioreactor paralleled the expansion of the clone that ultimately dominated in vivo. Finally, the impact of bortezomib on myeloma cells and on specialized functions of the microenvironment could be evaluated. Our findings indicate that 3D dynamic culture of reconstructed human multiple myeloma microenvironments in bioreactor may represent a useful platform for drug testing and for studying tumor-stroma molecular interactions.

Correspondence: ferrero.elisabetta@hsr.it or ferrarini.marina@hsr.it Received: February 23, 2017. Accepted: December 27, 2017. Pre-published: January 11, 2018. doi:10.3324/haematol.2017.167486 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/707

Introduction Tumors develop and progress within co-evolving microenvironments that affect both the fate of the tumor and its drug sensitivity. Response to drugs may be overestimated on 2-dimensional (2D) cultures, and the discrepancy between pre-clinical findings and clinical outcomes can also be attributed to the failure of conventional 2D models.1-4 3D culture models closely reproduce tumor within its microenvironment, recapitulating tumor-stroma interactions and signaling.1-4 Multiple myeloma (MM) is a paradigm of tumor-stroma inter-dependence, as it develops almost exclusively within the bone marrow (BM),5-8 where MM cells establish tight contacts with the stroma, that in turn delivers pro-survival, anti-apoptotic signals and confers drug resistance.9 Accordingly, new drugs, including proteasome inhibitors, have been developed to target both MM cells and their BM microenvironment. However, the disease remains incurable predominantly due to development of drug resistance. Along this line, 3D models of MM cells inside their microenvironhaematologica | 2018; 103(4)

707

D. Belloni et al. A

Figure 1. Generation of a 3D multiple myeloma (MM) microenvironment in bioreactor. (A) Experimental procedure: scaffold is pre-seeded in vitro with bone marrow stromal cells (BMSC)/endothelial cells (HUVEC) and transferred to bioreactor. MM cells are then added and cultured (see Methods section). (B) Scanning electron microscopy analysis of Spongostan (bar=20 mm). (C) Input (t0) (200x103/scaffold) and recovered cell number after 18 hours (h) of 3D culture. Results are mean±Standard Error of Mean of three independent experiments. (D) Immunohistochemistry showing uniform distribution of CD138+ MM cells and CD73+ stroma. Bar=100 mm. *P≤0.05. n.s.: not significant.

ment to test the impact of drugs in a relevant human context have been recently described.10-13 We have previously contributed to 3D models for MM exploiting the Rotary Cell Culture System (RCCSTM) bioreactor technology. By providing a balance between increased mass transfer and reduced shear stress, this dynamic bioreactor generates optimal conditions for longterm ex vivo maintenance of tissue explants.14-16 Specifically, we have shown that the model preserves, for extended time periods, the morphological and functional features of MM tissue components as well as their sensitivity to drugs.16 The aim of the present study was to recreate a surrogate 3D MM microenvironment able to reproduce the functional interactions of the native MM BM. We developed a robust technology, based on the integrated use of cellrepopulated scaffolds and the RCCSTM bioreactor. We demonstrate that our model simulates crucial MM features, in particular BM-MM dynamic interactions and MM survival/proliferation, thus providing a reliable tool to test the impact of drugs on MM cells inside their microenvironment.

Methods

VCAM1 (L-VCAM) and their wild-type (wt) counterpart were maintained in DMEM or RPMI 1640 plus 10% fetal bovine serum. BM aspirates from MM patients were collected after written informed consent and ethical approval from the Institutional Review Board; primary MM cells from 7 newly diagnosed patients and one relapsed, and BMSC were obtained (see Online Supplementary Methods). Endothelial cells (HUVEC) were propagated as described.17 Osteoblasts were differentiated from BMSC (see Castrén et al.18 and Online Supplementary Methods).

Scaffold preparation and culture in RCCSTM bioreactor Scaffolds were generated as in Figure 1A in bioreactor (see Ferrarini et al.16 and Online Supplementary Methods), with MM cell lines (500x103/scaffold) and HS-5 cells or L-VCAM1 or their wt counterpart (200x103/scaffold). Alternatively, primary MM cells (200x103/scaffold) were co-cultured in scaffolds with primary pooled allogeneic BMSC and HUVEC (100x103 each/scaffold), unless otherwise indicated. Scaffolds were then fixed or digested for flow cytometric (FACS) analysis. Supernatants were collected.

Immunohistochemistry Sections were stained with hematoxylin and eosin (H&E), or with monoclonal antibodies (mAbs): anti-Ki67, anti-CD138 (Ventana Systems); anti-light chains (Immunological Sciences, Italy); anti-CD73 (abcam), anti-cleaved caspase-3 (Cell Signaling Technology).16

Cell lines and primary cells Human MM1.S, U266 and RPMI.8226 MM cell lines, HS-5 BM stromal cell line and murine L-fibroblasts transfected with human 708

Flow-cytometric analysis Flow cytometric analysis (FACS) (FC500, Beckman Coulter) haematologica | 2018; 103(4)

3D reconstructed myeloma microenvironments

Figure 2. Stroma is required for multiple myeloma (MM) cell permanence inside scaffolds. β1-integrin and α4 chain expression by flow cytometric (FACS) analysis (A) and in vitro adhesion to HS-5 cells and VCAM1 transfectant (B) of MM1.S and RPMI.8226 cells. Gray histograms represent the isotype controls. (C) Number of MM cells recovered from nude or pre-seeded scaffolds after 24 hours (input number =500x103/scaffold). Data are mean±Standard Error of Mean (SEM) of three independent experiments. (D) Immunohistochemistry (IHC) showing proliferating (Ki67+) CD138+ MM cells over a layer of HS-5 cells or CD31+HUVEC. CD138 staining of MM1.S in the presence of bone-differentiated bone marrow stromal cells is also shown. Insert represents alizarin staining of the osteoblasts-coated scaffold. Bar=100 mm. (E) Scanning electron microscopy analysis shows RPMI.8226 cells without (left panel and insert, bar=2 mm) and with HS-5 cells (middle panel) or endothelial cells (HUVEC) (right panel). Bar=20 mm.

was performed using the following mAbs: anti-β1integrin/CD29 (BioLegend), followed by ALEXA-488-conjugated goat anti-mouse antibodies (Invitrogen); FITC-conjugated anti-VLA4/CD49d; PEconjugated anti-VCAM-1/CD106, PC7-conjugated anti-CD38 (both from BD-Pharmingen).

R&D Systems), monoclonal anti-pan-Akt (clone C67E7, CellSignaling Technology), anti-β1integrin mAb (abcam), anti-β-actin mAb (Santa Cruz Biotechnology), anti-STAT3/p-Stat and survivin (abcam). Proteins were quantified by ImageJ software.20

Scanning electron microscopy analysis Response to drugs Bortezomib (Velcade®, Millenium Pharmaceuticals) was used at 10 nM for 24-48 hours (h), melphalan at 1.2 nM for 72 h, dexamethasone at 20 nM for 48 h; IL-6 (R&D Systems) at 10 ng/mL. The anti-VLA-4 mAb natalizumab19 was used at 10 mg/mL. Cells were stained with anti-CD38 mAb and Annexin-V (BD-Pharmingen) for flow cytometric analysis.

Adhesion assay Multiple myeloma cell lines (200x103) were seeded over HS-5 or L-VCAM1 in 24-well plates. After 3 h, non-adherent cells were removed. Results are expressed as percentage (%) of CD38+ recovered adherent cells over input.

Scaffolds were fixed in 2% glutaraldehyde, post-fixed in 1% OsO4, dehydrated and then sputter coated with gold. Samples were examined by FEI/Philips XL-30 SEM (FEI, the Netherlands).

Determination of soluble factors and metallo-proteasic activities in supernatants

β2-microglobulin concentration was determined by immunonephelometry. Angiopoietin-2 (Ang-2), VEGF, FGF and IL6 levels were quantified by ELISA (R&D Systems). IL-1β,IL-8/CXCL-8 and TGF-β concentrations were determined by Bio-Plex Multiple-Cytokines Assays (Bio-Rad).21 MMP-2 and MMP-9 activities were assessed through Zymography.16

Western blot analysis

Fluorescence in situ hybridization

Western blot analysis was performed as described in the Online Supplementary Methods with polyclonal anti-pAkt (against S473,

Fluorescence in situ hybridization (FISH)22 was performed using probes for the detection of trisomy 12, deletions of 11q22.3

haematologica | 2018; 103(4)

709

D. Belloni et al.

(ATM), 13q14.3 (D13S319), 13q34 (LSI13q34), 17p13 (TP53) (Multi-color Probe, Abbott Molecular) and IGH gene rearrangements (DAKO). Microscope observation was performed using a Nikon Eclipse 90i (Nikon Instruments, Japan) and analyzed by Genikon software (Nikon).

Statistical analysis Statistical analysis was performed using Student t-test or oneway ANOVA. *P≤0.05; **P≤0.01; ***P≤0.001.

Results Generation of a functional MM 3D microenvironment in RCCSTM bioreactor requires supportive stroma 3D reconstruction of a surrogate MM microenvironment was based upon the proper combination of a scaffold and stromal cells, according to the procedure depicted in Figure 1A. As scaffold we used a commercially available gelatin biomaterial, which was selected according to several parameters (Online Supplementary Figure S1),

including efficiency of cell seeding and recovery, and especially its ultrastructure, resembling that of adult bone (Figure 1B). An efficient scaffold seeding with either the HS-5 cell line or HUVEC was achieved in the RCCSTM bioreactor, as indicated by the overall number of viable recovered cells upon 18 h 3D culture (Figure 1C). As a final step, MM cells were added to the bioreactor vessels (Figure 1A) resulting in the successful formation of a homogeneously populated, dense construct where both tumor cells and stroma conserved lineage specific markers (Figure 1D). We established our model with MM1.S and RPMI.8266 MM cell lines. These cell lines differ remarkably in the expression of β1-integrin and very late antigen-4 (VLA-4) (Figure 2A) which, through the interaction with its ligand VCAM1, is involved in MM adhesion to BM stroma.23 As a result, adhesion to HS-5 or L-VCAM transfectant parallels these differences (Figure 2B). Stromal cells were also required for MM cell permanence inside the scaffold, as indicated by the significantly higher number of MM cells populating the scaffold in the presence versus the absence

Figure 3. Pro-survival signals are delivered to multiple myeloma (MM) cell lines. (A) Western blot analyses of pAkt, Akt, survivin and actin in 3D versus 2D co-cultures of HS-5 and RPMI.8266 or MM1.S cells and of primary bone marrow stromal cells (BMSC) and MM1.S cells (one representative experiment out of three) and also on isolated HS-5 cells. (B) Mean±Standard Error of Mean (SEM) of pAkt/totalAkt ratios from three independent experiments. β2-microglobulin (C), IL-6, angiopoietin-2 (Ang-2), VEGF and FGF (D) levels in 3D culture supernatants. *P≤0.05.

710

haematologica | 2018; 103(4)

3D reconstructed myeloma microenvironments

(nude scaffold) of stroma; this was particularly evident with MM1.S cells (Figure 2C). Accordingly, immunohistochemistry (IHC) indicated that both MM1.S and RPMI.8266 cells entered, were homogeneously distributed and proliferated inside the scaffolds, prevalently when pre-seeded with the HS-5 stromal cell line (Figure 2D). Other cell types within the MM BM microenvironment, including endothelial cells and osteoblasts, are increasingly recognized as participating in MM pathogenesis and progression.12,24 We then exploited our system to model MM cells-HUVEC and MM cells-osteoblasts co-cultures. The latter were obtained through bone differentiation of BMSC, as reported.18 Upon culture with osteogenic differentiation medium, BMSC underwent morphological changes, increased mineralization and acquired Alizarin staining (Online Supplementary Figure S2A-C). Moreover, parallel cultures performed with differentiated BMSC in 2D and, particularly in 3D conditions, showed alkaline phosphatase release (Online Supplementary Figure S2D). Notably, scaffolds seeded with bone differentiated BMSC and with CD31+HUVEC also supported MM cells homing and permanence (Figure 2D, right panels). Development of intimate cell-cell contacts between MM cells and microenvironment was visualized at scanning electron microscopy analysis (Figure 2E, middle panel) showing that MM cells acquired a flatter morphology over the stroma, consistent with the induction of

adhesion-mediated cytoskeletal rearrangement. Conversely, MM cells exhibited a round shape with few contact points over nude scaffolds (Figure 2E, left panel). Interestingly, the entire 3D cell surface of some MM cells was embedded when HUVEC were used as stroma, suggesting that cell-cell interactions may dynamically evolve upon contacts (Figure 2E, right panel). The recapitulated physical interactions resulted in the activation of pro-survival signals, as indicated by up-regulated Akt phosphorylation in MM cell line-stromal cell co-cultures in 3D compared to 2D co-cultures (Figure 3A and B); instead, pAkt expression was negligible in isolated HS-5 cells. Similarly, the expression of survivin was also increased in 3D co-cultures (Figure 3A).25,26 Specialized functions of both MM cells and stroma were detectable in culture supernatants, including β2-microglobulin (Figure 3C) and growth factor release. The latter varied according to the stromal elements coating the scaffolds: Ang-2 was related to the presence of HUVEC,20 while IL-6, VEGF and FGF were detectable in all co-cultures (Figure 3D).

Response to anti-myeloma drugs of MM cell lines inside the surrogate microenvironment Tumor-stroma interactions affect MM cell behavior, including survival and drug resistance. The latter is induced via two overlapping pathways, i.e. soluble factor-

Figure 4. Adhesion to stroma and 3D architecture confer drug resistance to multiple myeloma (MM) cells. (A) Bortezomib (BTZ)-induced apoptosis in MM1.S (left), RPMI.8226 (middle) and U266 (right) in 2D versus 3D experiments, evaluated as percentage (%) of CD38+annexinV (AnnV)+ by flow cytometric (FACS) analysis. (B) MM1.S death induced by melphalan (Melph). MM cells were either alone or co-cultured with HS-5 cells. Data are mean±Standard Error of Mean (SEM) of three independent experiments. (C) Comparison between 2D/3D BTZ-induced apoptosis of MM1.S co-cultured with L-VCAM. Data are mean±SEM of three independent experiments. (D) The blocking effect of natalizumab (nat) on BTZ-induced apoptosis on MM1.S co-cultured with L-VCAM is shown. Results are representative of two independent experiments. (E) Dexamethasone-induced apoptosis in MM1.S cells in the presence/absence of IL-6 (10 ng/mL) was evaluated as percentage (%) of CD38+ AnnV+ by flow cytometric analysis in 2D (left) and in 3D (right) conditions. Results are expressed as mean±Standard Deviation of three independent experiments. *P≤0.05; **P≤0.01; ***P≤0.001. WT: wild-type non transfected counterpart; NT: not treated.

haematologica | 2018; 103(4)

711

D. Belloni et al.

mediated drug resistance (SFM-DR) and cell adhesionmediated drug resistance (CAM-DR).27,28 To assess the impact of 3D microenvironment on bortezomib sensitivity, we performed parallel experiments in 2D versus 3D using HS-5 cells as stroma. Co-culture of MM1.S with HS5 cells in 2D conditions conferred higher resistance to bortezomib, compared to MM.1S alone (Figure 4A, left). Significantly, resistance to bortezomib was even more evident when MM1.S were cultured with HS-5 cells in bioreactor, underscoring the role of 3D architecture per se in determining the impact of drugs (Figure 4A, left). The protective effect exerted by HS-5 cells was negligible with the RPMI.8226 cell line (Figure 4A, middle), in agreement with their reduced adhesion to stroma (Figure 2B). Protection conferred by HS-5 cells was not restricted to bortezomibtreated MM1.S since it was also observed with other cell lines (bortezomib-treated U266) (Figure 4A, right) and other drugs (melphalan) (Figure 4B), particularly in 3D conditions. HS-5 cells in principle can provide both mechanisms of drug resistance, since they release consistent amounts of

IL-6 (Figure 3D) and serve as support to MM cells. Among tumor-stroma cell-cell interactions, we addressed as a paradigm the VLA-4/VCAM1 molecular pathway, which is considered to play a central role.27,28 In experiments performed in 2D conditions, primary VCAM1+ BMSC (Online Supplementary Figure S3A) from MM patients promoted adhesion at least in part via the counter ligand VLA-4 (Online Supplementary Figure S3B), as demonstrated by inhibition experiments with the specific α4 blocking antibody natalizumab,29 and accordingly conferred to MM1.S cells higher resistance to bortezomib compared to the VCAM1 negative HS-5 cell line (Online Supplementary Figure S3C); the release in culture supernatants of β2-microglobulin and of lactate dehydrogenase (LDH), the latter bona fide expression of bortezomib cytotoxicity, paralleled this response (Online Supplementary Figure S3C). We then modeled the VLA-4/VCAM1 pair using the murine fibroblast L-VCAM transfectant (Figure 4C). The capability of VCAM1 expressed by the transfectants (Online Supplementary Figure S3A) to engage specific interactions was demonstrated by inhibition experiments

Figure 5. Survival and functions of primary multiple myeloma (MM) cells and stroma are promoted in the scaffold. (A) Primary MM cells inside scaffold retain CD138 and light chain (κ chain) expression. (B) Western blot analyses performed on parallel 2D and 3D co-cultures of primary MM cells and MM bone marrow stromal cells (BMSC); pAkt, total Akt, β1 integrin (β1) and actin are depicted in a representative experiment (left) out of three. Right panels represent mean±Standard Error of Mean (SEM) of β1 integrin/actin and pAkt/total Akt ratios in three independent experiments. (C) Representative experiment of western blot analysis of STAT3 and Akt phosphorylation in 2D versus 3D co-cultures of primary MM cells and primary BMSC. Freshly isolated primary MM cells and BMSC are used as controls. (D) Input (t0) and recovered (t7) number of primary CD38+ MM cells from 7 patients after 3D culture for seven days. Dotted lines represent input and recovered number of CD38+ MM cells from 3 patients cultured in parallel 2D experiments. (E) Specialized functions of primary MM cells and stroma are measured in culture supernatants. Data are mean±SEM of six independent experiments. (F) IL-6 release in co-cultures with BMSC (left) and Ang-2 release in co-cultures with HUVEC (right) were determined by ELISA in parallel 2D and 3D experiments. Data are mean±SEM of three independent experiments. *P≤ 0.05; **P≤0.01. H&E: hematoxylin and eosin staining. Bar=100 mm.

712

haematologica | 2018; 103(4)

3D reconstructed myeloma microenvironments

(Figure 4D). Accordingly, the protection conferred by LVCAM1 was highly significant in both 2D and 3D conditions, particularly in the latter (Figure 4C). Finally, the involvement of soluble factors in promoting drug resistance in our system was investigated using the well-described model of dexamethasone-treated MM1.S cells, where IL-6 is recognized to exert a protective effect.30 As reported,27 the cytokine specifically triggered STAT3 pathway in MM1.S cells, as indicated by western blot analysis and inhibition experiments with the anti-IL-6R monoclonal antibody tocilizumab (TCZ) (Online Supplementary Figure S4A), and also protected MM1.S cells from dexamethasone-induced death (Figure 4E), both in 2D and in 3D conditions. To further support this finding, a viability assay was performed (Online Supplementary Figure S4B), showing that IL-6 significantly reduced the inhibitory effect of dexamethasone.

3D culture supports primary MM cells survival and functions Isolated primary MM cells outside their native microenvironment do not survive. We exploited our model using

primary MM cells from patients, with the major aim of promoting their survival. We took advantage of MM BM stroma and HUVEC onto which primary isolated CD138MM cells were seeded and cultured in bioreactor for up to seven days. In the resulting constructs viable MM cells could be identified which retained the expression of lineage-specific markers as well as light chain production (Figure 5A). To address the role of cell-cell interactions, we compared the expression of β1 integrin in 2D and 3D parallel experiments. β1 integrin was up-regulated in scaffolds, and this upregulation was paralleled by downstream Akt phosphorylation (Figure 5B). STAT3 phosphorylation was also increased in 3D compared 2D co-cultures and could be attributed almost exclusively to MM cells (Figure 5C). Accordingly, when primary MM cells from 7 newly diagnosed patients were co-cultured with a pool of allogeneic BMSC plus HUVEC, the number of cells retrieved from the scaffolds at the end of culture matched the input number (Figure 5D), at variance with parallel 2D co-cultures. Moreover, both MM cells and stroma retained their specialized functions, as indicated by β2-microglobulin and soluble factors released in the supernatant from

Figure 6. Bortezomib affects primary multiple myeloma (MM) cells viability and functions within the scaffolds. (A) Primary MM cells from 6 patients were retrieved from scaffolds after 48 hours of bortezomib (BTZ) treatment; death was calculated as the percentage of CD38+/AnnV+ cells by flow cytometric analysis. (B) A representative experiment is shown. (C) Immunohistochemistry performed on scaffolds populated with primary MM cells from 2 patients reveals the presence of apoptotic MM cells upon bortezomib exposure, as indicated by both down-regulated CD138 expression and intense nuclear caspase-3 immunoreactivity. Bar=50 mm. (D) Metalloprotease (MMPs) activities are measured in supernatants from treated (BTZ) and untreated (NT) samples by zymography and densitometric analyses (left and middle panels). Results are mean±Standard Error of Mean (SEM) of 6 patients. Right panel shows a representative experiment. (E) Angiopoietin-2 (Ang-2), IL6 and VEGF levels are determined in supernatants of treated and untreated scaffolds. Data are mean±SEM of 6 patients.*P≤ 0.05; **P≤ 0.01. a.u.: arbitrary units.

haematologica | 2018; 103(4)

713

D. Belloni et al.

the co-cultures (Figure 5E), recapitulating the native BM microenvironment.31 Of note, cytokine concentrations were significantly higher in 3D than in 2D conditions (Figure 5F).

Impact of bortezomib on a 3D culture of primary MM cells and stroma We next investigated whether our reconstructed model was suitable to assess sensitivity to bortezomib of primary MM cells within their microenvironment. To this aim, we performed parallel cultures in bioreactor in the presence/absence of bortezomib, using primary MM cells obtained from 6 patients. Scaffolds were coated with primary MM BMSC and HUVEC to better approximate the native microenvironment.12 After 48 h MM cells were retrieved from scaffolds and submitted to FACS analyses. Bortezomib-induced death, which varied among patients, could be determined as percentage of CD38+AnnV+ MM cells (Figure 6A and B). Bortezomib cytotoxicity could also

be evaluated by IHC, showing down-modulated expression of CD138 antigen in caspase-3+ MM cells undergoing apoptosis32 (Figure 6C) and through the assessment of specialized functions. In particular, MMP-2 and MMP-9 activities (Figure 6D), as well as Ang-2 and IL-6 concentrations (Figure 6E), significantly decreased in supernatants from bortezomib-treated scaffolds, underlining the impact of the drug also on stroma.

3D culture supports proliferation of an in vivo expanding MM clone Multiple myeloma cells from 6 patients analyzed survived in 3D culture. In an additional case, primary MM cells not only survived but also significantly proliferated in 3D cultures, but not in parallel 2D conditions (Figure 7A). MM cells were obtained from a patient who initially achieved a very good partial response (VGPR) upon treatment with bortezomib-thalidomide-dexamethasone (VTD) but rapidly progressed to terminal plasma cell

Figure 7. Culture in bioreactor mirrors the expansion of an in vivo proliferating multiple myeloma (MM) sub-clone. (A) Primary MM cell number (left panel) and Î˛2 microglobulin release in supernatant (right panel) after 14 days of culture under parallel 2D and 3D co-cultures with HS-5 cells. (B) Schematic representation of patientâ&#x20AC;&#x2122;s clinical course and treatments. After diagnosis (t0), response to treatment was assessed and defined as Very Good Partial Response (VGPR), Progressive Disease (PD), Plasma Cell Leukemia (PCL). Serum M protein concentration and percentage of bone marrow PC (BMPC) were serially determined. Treatments were: bortezomib-thalidomide-dexamethasone (VTD), bortezomib-doxorubicin-dexamethasone (PAD), bortezomib-cyclophosphamide-lenalidomide-dexamethasone (VCRD), dexamethasone-cisplatin-adriamycin-cyclophosphamide-etoposide (D-PACE). (C) Immunohistochemistry of the proliferation marker Ki-67 inside a scaffold (lower) and in a matched bone biopsy (upper). Insert represents CD138 staining. Bar=100 mm. (D) Interphase fluorescence in situ hybridization analysis performed in purified MM cells retrieved upon culture in bioreactor (upper), showing normal pattern of ATM and p53 (upper left) and 13q14.3/13q34 deletion (upper right); in the lower panel, a spontaneous metaphase of a PC showing 13q14.3/13q34 deletion. (E) Percentage of cells carrying the 13q14.3/13q34 deletion or IGH/FGFR3 translocation ex vivo and in vivo.

714

haematologica | 2018; 103(4)

3D reconstructed myeloma microenvironments

leukemia (PCL) despite several lines of therapy (Figure 7B). At progression, a BM biopsy was performed and MM cells were cultured in bioreactor (Figure 7B). MM cell proliferation was confirmed by the high frequency of Ki67+ cells at IHC analysis of the scaffold, which mirrored that observed in the BM biopsy (Figure 7C) and was paralleled by Î˛2-microglobulin levels in supernatants (Figure 7A). Given the unique proliferative behavior of the patientâ&#x20AC;&#x2122;s MM cells, we compared genomic changes occurring over time in vivo and ex vivo. FISH analysis performed on MM cells retrieved at the end of 3D culture identified cells carrying the 13q14.3 deletion; no deletions of ATM or p53 loci could be detected (Figure 7D, upper panels), nor IGH rearrangements, indicating the absence of the t(4;14) translocation. Cells expressing the 13q14.3 deletion represented 46% of all MM cells at the beginning of culture and became 70% after 14-day 3D culture (Figure 7E), suggesting that this clone preferentially proliferated in bioreactor. Accordingly, few spontaneous metaphases, which are extremely rare in MM samples, could be observed in 3D culture and they all presented the deletion (Figure 7D, lower panel). Notably, at diagnosis, 96% of MM cells carried the 13q14.3 deletion and 46% co-expressed the t(4;14) IGH/FGFR3 translocation, consistent with the reported frequent association between the two cytogenetic lesions.33-35 After an initial tumor burden reduction in response to VTD, MM cells with the 13q14.3 deletion, but not those also carrying the t(4;14) translocation, progressively expanded in vivo, ultimately representing the whole population in PCL cells (Figure 7E).

Discussion The availability of suitable models that recapitulate the complex tumor-host interplay is central to understanding cancer biology and developing appropriate treatments. This is especially true in MM,36 where MM-BM interactions are crucial to disease progression and responsiveness to drugs.7,8 We here show that our ex vivo 3D co-culture model in bioreactor meets the requirements of recapitulated MM-BM dialogue, permanence and survival of primary MM cells for an extended time period, thereby also incorporating the temporal dimension. The model relies on the integrated use of a gelatin scaffold seeded with tumor cells and stroma and the RCCSTM bioreactor technology. Scaffolds are a key component for the reconstitution of MM microenvironment as they provide cells with mechanical support;37-39 we selected a gelatin biomaterial that allows morphological investigations and also mimics 3D ultrastructure of MM BM. Stroma is required to achieve ex vivo cell seeding efficiency and to re-create tumor-stroma contacts and signaling, both with MM cell lines and primary MM cells. Overall, the construct reproduces the tumor-stroma 3D 'dynamic reciprocity'40 which is lost in conventional 2D co-culture. Both MM cells and stroma retain the expression of lineage specific markers as well as their specialized functions, including the release of Î˛2-microglobulin, cytokines and growth factors, thus recapitulating the profile of the native BM.29,31 The model was established with MM cell lines and successfully applied to primary MM cells. Notably, primary MM cells from patients survived for up to seven days when cultured inside scaffolds. We used as stroma prihaematologica | 2018; 103(4)

mary allogeneic BMSC from MM patients which possess the repertoire of adhesion molecules and have been previously validated for the reconstruction of an MM BM niche.12 Given the high flexibility of our model, the contribution of additional cellular elements of MM microenvironment can be addressed. As an example, the engagement of intimate EC-MM contacts inside the scaffold, documented by morphological and functional analyses, indicates that the system is suitable to elicit and study dynamic EC-MM targetable interactions. Moreover, the feasibility of co-culturing MM cells and bone-differentiated BMSC underscores the potential to fulfill the unmet need for a model to study the relationship between MM progression and bone disease.41 Finally, development of repopulated scaffolds could be further exploited to address the contribution of circulating MM cells in humanized in vivo scaffold-mouse models.42 Multiple myeloma-BM functional interactions and down-stream signaling are promoted inside our surrogate microenvironment. Indeed, higher levels of pAkt, pSTAT and survivin are more appreciable in 3D than in 2D culture with both MM cell lines and primary MM cells, mimicking the activation of pro-survival signaling pathways in MM cells from patients.25,43,44 Akt pathway is crucial for MM survival and drug resistance, and has been proposed as a promising target for future molecular-based therapies.43 Tumor-stroma interactions are considered a major determinant in drug resistance in MM via the release of soluble factors and cell-to-cell adhesion. Our data on the protective effect of stroma against bortezomib-induced apoptosis are consistent with the induction of CAM-DR; in this regard, experiments conducted with the L-VCAM transfectant indicate that the system can be exploited to model and elucidate specific molecular interactions. The additional contribution by 3D culture further supports the importance of tissue architecture per se in drug resistance. The achievement of a construct where MM cells survive via the establishment of proper 3D interactions with a compliant microenvironment is a prerequisite to test the impact of drugs in a relevant human context. Accordingly, the impact of bortezomib on primary MM cells and their microenvironment could be assessed by means of FACS and IHC analyses, and also through determination of specialized functions in supernatants. In particular, the decrease of Ang-2 can monitor the cytotoxic effect of the drug on EC,45 while variations in IL-6 and MMP-2/-9 activities may result from disruption of MM-stroma interplay.7,9 The dissection of clonal dynamics during disease progression and in response to therapy is increasingly emerging as a central issue of MM investigation.34,35 In a patient with high-risk MM, sequential BM sampling allowed the evolutionary path to be defined along the clinical course. Two sub-clones co-existing at diagnosis initially responded to first-line therapy; subsequently, only one evolved over time. Significantly, the bioreactor culture could anticipate the expansion of the same clone. These data suggest that, in selected cases, the model can be exploited to monitor the dynamics of clones inside the whole MM cell population, and possibly to identify new potential targets. Altogether, our findings indicate that 3D dynamic culture of reconstructed human MM microenvironments in RCCSTM bioreactor may represent an important platform for drug testing and the study of tumor-stroma molecular interactions. 715

D. Belloni et al.

Acknowledgments We would like to thank Dr. Andrea Motta and Dr. Tania Carenzo for technical assistance, Dr. Cristina Tresoldi, San Raffaele Scientific Institute, for providing MM samples, and Prof. V. Gattei, Aviano, Italy, for providing L-VCAM.

References 1. Tlsty TD, Coussens LM. Tumor stroma and regulation of cancer development. Annu Rev Pathol. 2006;1:119-150. 2. Hu M, Polyak K. Microenvironmental regulation of cancer development. Curr Opin Genet Dev. 2008;18(1):27-34. 3. Pampaloni F, Reynaud EG, Stelzer EH. The third dimension bridges the gap between cell culture and live tissue. Nat Rev Mol Cell Biol. 2007;8(10):839-845. 4. Yamada M, Cukierman E. Modeling Tissue Morphogenesis and Cancer in 3D. Cell. 2007;130:601-610. 5. Hallek M, Bergsagel PL, Anderson KC. Multiple myeloma: increasing evidence for a multistep transformation process. Blood. 1998;91(1):3-21. 6. Hussein MA, Juturi JV, Lieberman I. Multiple myeloma: present and future. Curr Opin Oncol. 2002;14(1):31-35. 7. Hideshima T, Mitsiades C, Tonon G, Richardson PG, Anderson KC. Understanding multiple myeloma pathogenesis in the bone marrow to identify new therapeutic targets. Nat Rev Cancer. 2007;7(8):585-598. 8. Burger JA, Ghia P, Rosenwald A, CaligarisCappio F. The microenvironment in mature B-cell malignancies: a target for new treatment strategies. Blood. 2009; 114(16):3367-3375. 9. Podar K, Chauhan D, Anderson KC. Bone marrow microenvironment and the identification of new targets for myeloma therapy. Leukemia. 2009;23(1):10-24 10. Kirshner J, Thulien KJ, Martin LD, et al. A unique three-dimensional model for evaluating the impact of therapy on Multiple Myeloma. Blood. 2008; 112(7):2935-2945. 11. Calimeri T, Battista E, Conforti F, et al. A unique three-dimensional SCID-polymeric scaffold (SCID-synth-hu) model for in vivo expansion of human primary multiple myeloma cells. Leukemia. 2011;25(4):707-711. 12. Reagan MR, Mishima Y, Glavey SV, et al. Investigating osteogenic differentiation in multiple myeloma using a novel 3D bone marrow niche model. Blood. 2014;124(22): 3250-3259. 13. Jakubikova J, Cholujova D, Hideshima T, et al. A novel 3D mesenchymal stem cell model of the multiple myeloma bone marrow niche: biologic and clinical applications. Oncotarget. 2016; 7(47):7732677341. 14. Mazzoleni G, Di Lorenzo D, Steimberg N. Modelling tissues in 3D: the next future of pharmaco-toxicology and food research? Genes Nutr. 2009;4(1):13-22. 15. Cosmi F, Steimberg N, Dreossi D, Mazzoleni G. Structural analysis of rat bone explants kept in vitro in simulated microgravity conditions. J Mech Behav Biomed Mater. 2009;2(2):164-172. 16. Ferrarini M, Steimberg N, Ponzoni M, et al. Ex-vivo dynamic 3-D culture of human tissues in the RCCS™ bioreactor allows the

716

17.

18.

19.

20.

21.

22.

23.

24.

25. 26.

27.

28.

29.

30.

Funding This work was supported by AIRC-Special Program Molecular Clinical Oncology AIRC 5x1000 project n. 9965 (to FCC) and from its extension (to PG).

study of Multiple Myeloma biology and response to therapy. PLoS One. 2013; 8(8):e71613. Belloni D, Scabini S, Foglieni C, et al The vasostatin-I fragment of chromogranin A inhibits VEGF-induced endothelial cell proliferation and migration. FASEB J. 2007;21(12):3052-3062. Castrén E, Sillat T, Oja S, et al. Osteogenic differentiation of mesenchymal stromal cells in two-dimensional and three-dimensional cultures without animal serum. Stem Cell Res Ther. 2015;6:167. Yu Y, Schürpf T, Springer TA. How natalizumab binds and antagonizes α4 integrins. J Biol Chem. 2013;288(45):231432325. Belloni D, Marcatti M, Ponzoni M, et al. Angiopoietin-2 in bone marrow milieu promotes multiple myeloma-associated angiogenesis. Exp Cell Res. 2015;330(1):1-12. Dagna L, Corti A, Langheim S, et al. Tumor necrosis factor α as a master regulator of inflammation in Erdheim-Chester disease: rationale for the treatment of patients with infliximab. J Clin Oncol. 2012;30(28):286290. Gole L, Lin A, Chua C, Chang WJ. Modified cIg-FISH protocol for multiple myeloma in routine cytogenetic laboratory practice. Cancer Genet. 2014;207(1-2):3134. Noborio-Hatano K, Kikuchi J, Takatoku M, et al. Bortezomib overcomes cell-adhesionmediated drug resistance through downregulation of VLA-4 expression in multiple myeloma. Oncogene. 2009; 28(2):231-242. Mondello P, Cuzzocrea S, Navarra M, Mian M. Bone marrow micro-environment is a crucial player for myelomagenesis and disease progression. Oncotarget. 2017; 8(12):20394-20409. Hsu J, Shi Y, Krajewski S, et al. The AKT kinase is activated in multiple myeloma tumor cells. Blood. 2001;98(9):2853-2855. Younes H, Leleu X, Hatjiharissi E, et al. Targeting the phosphatidylinositol 3-kinase pathway in multiple myeloma. Clin Cancer Res. 2007;13(13):3771-3775. Furukawa Y, Kikuchi J. Epigenetic mechanisms of cell adhesion-mediated drug resistance in multiple myeloma. Int J Hematol. 2016;104(3):281-292. Nefedova Y, Landowski TH, Dalton WS. Bone marrow stromal-derived soluble factors and direct cell contact contribute to de novo drug resistance of myeloma cells by distinct mechanisms. Leukemia. 2003; 17(6):1175-1182. Podar K, Zimmerhackl A, Fulciniti M, et al. The selective adhesion molecule inhibitor Natalizumab decreases multiple myeloma cell growth in the bone marrow microenvironment: therapeutic implications. Br J Haematol. 2011;155(4):438-448. Greenstein S, Krett NL, Kurosawa Y, et al. Characterization of the MM.1 human multiple myeloma (MM) cell lines: a model system to elucidate the characteristics, behavior, and signaling of steroid-sensitive

31.

32.

33.

34.

35.

36.

37.

38.

39. 40.

41.

42.

43.

44.

45.

and -resistant MM cells. Exp Hematol. 2003;31(4):271-282. Manier S, Sacco A, Leleu X, Ghobrial IM, Roccaro AM. Bone marrow microenvironment in multiple myeloma progression. J Biomed Biotechnol. 2012;2012:157496. Nerini-Molteni S, Ferrarini M, Cozza S, Caligaris-Cappio F, Sitia R. Redox homeostasis modulates the sensitivity of myeloma cells to bortezomib. Br J Haematol. 2008;141(4):494-503. Manier S, Salem KZ, Park J, Landau DA, Getz G, Ghobrial IM. Genomic complexity of multiple myeloma and its clinical implications. Nat Rev Clin Oncol. 2017; 14(2):100-113. Bolli N, Avet-Loiseau H, Wedge DC, et al. Heterogeneity of genomic evolution and mutational profiles in multiple myeloma. Nat Commun. 2014;5:2997. Keats JJ, Chesi M, Egan JB, et al. Clonal competition with alternating dominance in multiple myeloma. Blood. 2012; 120(5):1067-1076. Ferrarini M, Mazzoleni G, Steimberg N, et al. Innovative Models to Assess Multiple Myeloma Biology and the Impact of Drugs. In: Multiple Myeloma - A Quick Reflection on the Fast Progress. R Hajek (Ed.), InTech, 2013. DOI: 10.5772/54312. Zhang M, Boughton P, Rose B, Lee CS, Hong AM. The use of porous scaffold as a tumor model. Int J Biomater. 2013; 2013:396056. De la Puente P, Azab AK. 3D tissue-engineered bone marrow: what does this mean for the treatment of multiple myeloma? Future Oncol. 2016;12(13):1545-1547. Fischbach C, Chen R, Matsumoto T, et al Engineering tumors with 3D scaffolds. Nat Methods. 2007;4(10):855-860. Schmeichel KL, Bissell MJ. Modeling tissue-specific signaling and organ function in three dimensions. J Cell Sci. 2003; 116(12):2377-88. Reagan MR, Liaw L, Rosen CJ, Ghobrial IM. Dynamic interplay between bone and multiple myeloma: emerging roles of the osteoblast. Bone. 2015;75:161-169. Zhu D, Wang Z, Zhao JJ, et al. The Cyclophilin A-CD147 complex promotes the proliferation and homing of multiple myeloma cells. Nat Med. 2015;21(6):572580. Mimura N, Hideshima T, Shimomura T, et al. Selective and potent Akt inhibition triggers anti-myeloma activities and enhances fatal endoplasmic reticulum stress induced by proteasome inhibition. Cancer Res. 2014;74(16):4458-4469. Tsubaki M, Takeda T, Ogawa N, et al. Overexpression of survivin via activation of ERK1/2, Akt, and NF-κB plays a central role in vincristine resistance in multiple myeloma cells. Leuk Res. 2015;39(4):445452. Belloni D, Veschini L, Foglieni C, et al. Bortezomib induces autophagic death in proliferating human endothelial cells. Exp Cell Res. 2010;316(6):1010-1018.

haematologica | 2018; 103(4)

ARTICLE

Stem Cell Trasplantation

Tocilizumab, tacrolimus and methotrexate for the prevention of acute graft-versus-host disease: low incidence of lower gastrointestinal tract disease

Ferrata Storti Foundation

William R. Drobyski,1 Aniko Szabo,2 Fenlu Zhu,1 Carolyn Keever-Taylor,1 Kyle M. Hebert,3 Renee Dunn,3 Sharon Yim,1 Bryon Johnson,1 Anita Dâ&#x20AC;&#x2122;Souza,1 Mary Eapen,1 Timothy S. Fenske,1 Parameswaran Hari,1 Mehdi Hamadani,1 Mary M. Horowitz,1 J. Douglas Rizzo,1 Wael Saber,1 Nirav Shah,1 Bronwen Shaw1 and Marcelo Pasquini1 The Department of Medicine; 2The Division of Biostatistics, Institute for Health and Society and 3The Center for International Blood and Marrow Transplant Research, Medical College of Wisconsin, Milwaukee, WI, USA

Haematologica 2018 Volume 103(4):717-727

ABSTRACT

e conducted a phase 2 study in which patients undergoing allogeneic hematopoietic stem cell transplantation received tocilizumab in addition to standard immune suppression with tacrolimus and methotrexate for graft-versus-host disease prophylaxis. Thirty-five patients were enrolled between January 2015 and June 2016. The median age of the cohort was 66 (range: 22-76). All patients received busulfan-based conditioning, and were transplanted with human leukocyte antigen-matched related or matched unrelated bone marrow or peripheral stem cell grafts. The cumulative incidences of grades II-IV and III-IV acute graft-versus-host disease were 14% (95% CI 5-30) and 3% (95% CI 0-11) at day 100, and 17% (95% CI 7-31) and 6% (95% CI 116) at day 180, respectively. Notably, there were no cases of graft-versushost disease of the lower gastrointestinal tract within the first 100 days. A comparison to 130 matched controls who only received tacrolimus and methotrexate demonstrated a lower cumulative incidence of grades II-IV acute graft-versus-host disease (17% versus 45%, P=0.003) and a significant increase in grades II-IV acute graft-versus-host disease-free survival at six months (69% versus 42%, P=0.001) with tocilizumab, tacrolimus and methotrexate, which was the primary endpoint of the study. Immune reconstitution was preserved in patients treated with tocilizumab, tacrolimus and methotrexate, as T-cell and B-cell subsets recovered to near normal levels by 6-12 months post-transplantation. We conclude that tocilizumab has promising activity in preventing acute graft-versus-host disease, particularly in the lower gastrointestinal tract, and warrants examination in a randomized setting. clinicaltrials.gov Identifier:02206035

Correspondence: wdrobysk@mcw.edu

Received: October 26, 2017. Accepted: January 18, 2018. Pre-published: January 19, 2018. doi:10.3324/haematol.2017.183434 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/717

Introduction Graft-versus-host disease (GvHD) is the major complication arising from allogeneic hematopoietic stem cell transplantation (HSCT). GvHD is characterized by the overproduction of proinflammatory cytokines that induce target organ damage directly, or indirectly by activating other effector cell populations.1-3 Interleukin 6 (IL-6) has emerged as an inflammatory cytokine that plays a pivotal role in the pathophysiology of GvHD and has become a potential therapeutic target.4-6 Preclinical studies have demonstrated that IL-6 levels are increased early during GvHD and are present in all target tissues.7 Moreover, blockade of the IL-6 signaling pathway using an antibody that binds to the IL-6 receptor has been shown to reduce the severity of GvHD and prolong survival in pre-clinical murine models.7,8 In particular, IL-6 appears to have an important pathophysiological role in promoting inflammation in the gastrointestinal (GI) tract,7 which is a major cause of morbidity and mortality during GvHD. haematologica | 2018; 103(4)

717

W.R. Drobyski et al.

The importance of IL-6 in human GvHD is supported by studies which have shown that patients with elevated plasma levels of IL-6,9,10 as well as those with a recipient or donor IL-6 genotype which results in increased IL-6 production,11-14 have an increased incidence and severity of this disease. The Food and Drug Administration (FDA) approval of tocilizumab (Toc; Actemra), which is a humanized anti-IL-6 receptor antibody that blocks both the membrane and soluble forms of the receptor for the treatment of severe active rheumatoid arthritis,15,16 has allowed off-label use of this agent to determine whether blockade of IL-6 signaling attenuates GvHD. To that end, it was demonstrated that Toc induced clinical responses in patients with steroid refractory acute (a)GvHD that primarily involved the lower GI tract.17 Furthermore, the addition of Toc to standard immune suppression resulted in a low incidence of aGvHD in a patient population which was comprised primarily of myeloid maligananices,18 providing evidence that inhibition of IL-6 might also be an effective approach for the prevention of GvHD. Notably, however, in the latter study myeloablative (MA) conditioning was carried out exclusively with total body irradiation (TBI) and cyclophosphamide (Cy), which is not a widely employed regimen for the treatment of myeloid malignancies. Since the intensity of the conditioning regimen is known to affect the magnitude of inflammatory cytokine production,19 the relevance of these results to patients treated with alternative, more commonly utilized conditioning regimens is not clear. Furthermore, there was no corresponding demographically-matched control population against which to assess these results. In the study herein, we sought to determine whether the addition of Toc to standard immune suppression was effective for the prevention of aGvHD in patients who received busulfan (Bu)-based conditioning regimens, with particular emphasis given to the lower GI tract given prior pre-clinical studies and the primacy of this organ in GvHD pathophysiology. To place our results in context, we interrogated the Center for International Blood and Marrow Transplant Research (CIBMTR) database to obtain a control population that was matched for relevant demographic and transplant characteristics. We also examined longitudinal immune reconstitution and inflammatory cytokine production as additional parameters by which to assess the effect of IL-6 inhibition in transplant recipients.

Methods Patient population Patients were eligible for this trial if they met the following criteria: age >18 years; a diagnosis of acute leukemia, chronic myelogenous leukemia, myeloproliferative disease, myelodysplasia with less than 5% of blasts in the bone marrow, or a diagnosis of chronic lymphocytic leukemia, non-Hodgkin lymphoma or Hodgkin lymphoma with chemosensitive disease; availability of a 10/10 matched sibling or 8/8 matched unrelated donor; ejection fraction at rest >45% for MA conditioning or >40% for reduced intensity conditioning (RIC); estimated creatinine clearance greater than 50 mL/minute; adjusted diffusing capacity for carbon monoxide (DLCO) ≥40% and forced expiratory volume in 1 second (FEV1) ≥50%; and total bilirubin < 1.5 x and alanine transaminase (ALT)/ aspartate transaminase (AST) < 2.5x the upper normal limit. Patients were excluded if they had had a prior allogeneic HSCT, Karnofsky Performance Score <70%, uncontrolled bacterial, viral or fungal infections at time of enrollment, prior intoler718

ance or allergy to Toc, use of rituximab, alemtuzumab, antithymocyte globulin (ATG) or other monoclonal antibody at time of conditioning regimen, or history of diverticulitis, Crohn’s disease, ulcerative colitis or a demyelinating disorder.

Control population The control population was derived from cases reported to the CIBMTR.20 Eligibility criteria for the control cohort consisted of having received a first allogeneic transplant at an American center, excluding The Medical College of Wisconsin (MCW), and meeting the same eligibility as the study population, with the exception of receiving only tacrolimus (Tac)/ methotrexate (MTX) as GvHD prophylaxis. Controls were selected from the years 2010-2015. This selection process resulted in the screening of 1,442 patients to define an optimally-matched control cohort.

Conditioning regimens and GvHD prophylaxis All patients received Bu as part of the preparative regimen. Patients receiving MA conditioning were treated with either Bu 3.2 mg/kg/day (days -7 to -4) and Cy 60 mg/kg/day (days -3 and 2) or Bu 3.2 mg/kg/day (days 5- to -2) and fludarabine (Flu) 30 mg/m2/day (days -5 to -2). Bu dosing was modified after the fifth dose, if necessary, to achieve a targeted level of 900±100 ng/mL. RIC was with Flu 30 mg/m2/day (days -6 to -2) and Bu 3.2 mg/kg/day (days -5 and -4). Patients received either T-cell replete bone marrow or granulocyte colony factor-stimulated peripheral blood stem cell grafts. For GvHD prevention, Tac was administered intravenously at a dose of 0.03 mg/kg/day starting on day – 3 to maintain a level of 5-15 ng/mL. MTX was given at the doses of 15 mg/m2 IV on day +1, and 10 mg/m2 IV on days +3, +6 and +11 after hematopoietic stem cell infusion. Toc was infused intravenously at a dose of 8 mg/kg (maximum dose of 800 mg) once on day-1 approximately 24 hours prior to the hematopoietic stem cell infusion, as per Kennedy and colleagues.18

Study design The primary objective of this study was to compare the probability of grades II-IV aGvHD-free survival at day 180 post-transplant between recipients of Toc, Tac and MTX and a contemporary control population who received Tac/MTX-based GvHD prophylaxis. Pre-specified secondary objectives of the study were to compare chronic (c)GvHD, transplant related mortality (TRM), disease relapse or progression, disease-free survival (DFS) and overall survival (OS) between Toc/Tac/MTX and Tac/MTX CIBMTR controls. Online Supplementary Table S1 contains the definition of the events, censorings, and competing risks for all timeto-event outcomes. Additionally, secondary objectives included description of the incidence of grades ≥3 toxicities according to Common Terminology Criteria for Adverse Event (CTACAE) v4, neutrophil and platelet engraftment, characterization of infections, proportion of donor chimerism, extent of immune reconstitution, and production of proinflammatory cytokines among patients who received Toc/Tac/MTX. A population that consisted of patients who were otherwise eligible for the trial but did not receive Toc in addition to Tac/MTX for GvHD prophylaxis was employed as a control for the cytokine analysis. These patients were consented and enrolled on a separate study. Both protocols were approved by the Institutional Review Board at the MCW.

Outcome assessments Neutrophil recovery was defined as the first of three consecutive days with an absolute neutrophil count (ANC)>500. Platelet engraftment was defined as the first day of a sustained platelet count above 20,000 without any platelet transfusions for the preceding seven days. Five patients who received reduced intensity haematologica | 2018; 103(4)

Tocilizumab for the prevention of acute GvHD

transplants never dropped their platelet count below 20,000. In these patients, the date of platelet recovery was defined as the first day that the platelet count increased after its nadir. The grade of aGvHD was determined using the Glucksberg scale.21 cGvHD was graded using the National Institutes of Health (NIH) Chronic GvHD Consensus criteria.22

Statistical analysis The probabilities of DFS, aGvHD-free survival, and OS were calculated using the Kaplan-Meier estimator. The probabilities of neutrophil and platelet engraftment, TRM, disease progression/relapse, and aGvHD and cGvHD were calculated

Table 1. Patient characteristics.

Variable N Age, median (range) Sex (M/F) Disease (n, %) AML CR1 CR2 ALL CR1 CR2 Secondary AML, CR1 Therapy-Related AML, CR1 MDS CMML Myelofibrosis CML, CP2 T-cell Lymphoma, CR2 NK/T-cell Lymphoma, CR2 Donor Type (n, %) MRD MUD Preparative Regimen (n, %) Bu/Cy Flu/Bu4 Flu/Bu2 Graft Source (n, %) Bone Marrow Peripheral Blood Disease Risk Index (n, %) Low Intermediate High CMV Serostatus (n, %) Donor−/Recipient− Donor+/Recipient− Donor+/Recipient+ Donor−/Recipient+

Value 35 66 (22-76) 22/13 14 (40) 11 3 4 (11) 3 1 3 (9) 2 (6) 3 (9) 5 (14) 1 (3) 1 (3) 1 (3) 1 (3)

Other detailed methods Serum cytokine and immune reconstitution analyses are described in Online Supplementary Methods.

Results 14 (40) 21 (60) 5 (14) 13 (34) 7 (51) 6 (17) 29 (83) 4 (11) 22 (63) 9 (26) 12 (34) 11 (31) 4 (11) 8 (23)

AML: acute myelogenous leukemia; ALL: acute lymphoblastic leukemia; MDS: myelodysplasia; CMML: chronic myelomonocytic leukemia; CML: chronic myelogenous leukemia; MRD: matched related donor; MUD: matched unrelated donor; CMV: cytomegalovirus; Bu: busulfan; Cy: cyclophosphamide; Flu: fludarabine; CR: complete remission; CP2: second chronic phase.

haematologica | 2018; 103(4)

using the cumulative incidence estimator. GvHD was calculated using disease progression/relapse or death as competing risks. Matching criteria for controls consisted of age within 5 years; performance score (≥90 vs. <90); Bu-based regimen (Flu/Bu reduced intensity, Bu/Cy MA, or Flu/Bu MA); disease, and donor type (human leukocyte antigen [HLA]-matched sibling vs. HLAmatched unrelated donor). Up to four matches per case were selected whenever possible; when more controls were available preference was given to identical stem cell source (bone marrow, peripheral blood) and closest age. With respect to stem cell source, 124 of the 135 CIBMTR control patients were also matched for this variable. The follow up of the control patients was administratively censored at 22 months, which corresponded to the longest follow up of the trial patients. With 35 patients enrolled in the trial and 140 controls, there was an 80% power to detect an improvement of 20% on day 180 aGvHD-free survival. The outcomes were compared between groups using a stratified log-rank test or Gray’s test for survival and competing risk outcomes, respectively, with matched sets defining strata. Stratified Cox or Fine-Gray regression was used to obtain hazard ratios with 95% confidence intervals. Each cytokine was analyzed separately in the combined sample, and by conditioning regimen (ablative vs. RIC). Cytokine values were shifted by half of the smallest nonzero value and then log-transformed to improve normality of the residuals. First, a separate analysis was conducted for each treatment group (Toc and Control) followed by a joint analysis. A repeated measures analysis was performed using a mixed effects model with a random subject-specific intercept to incorporate within-subject dependence. All time-points were compared to the baseline value with Dunnett-Hsu adjustment for multiple testing. The estimates were back-transformed to the original scale for reporting.

Patient characteristics From January 29, 2015 to June 30, 2016, 35 patients were enrolled in the study. The demographic data for this population is detailed in Table 1. The median age of the cohort was 66 (range: 22-76). Diseases consisted of de novo acute myeloid leukemia (AML; n=14), acute lymphoblastic leukemia (ALL; n=4), secondary or therapy-related AML (n=5), myelodysplastic syndrome (MDS; n=3), chronic myelomonocytic leukemia (CMML; n=5), myelofibrosis (n=1), T-cell lymphoma (n=1), chronic myeloid leukemia (CML; n=1), and natural killer (NK)/Tcell lymphoma (n=1). The disease status of patients with acute and chronic leukemia are further specified in Table 1. Four of the 14 patients with AML in first remission had FLT3/ITD mutations, two additional recipients had evidence of minimal residual disease at the time of transplant, and another patient had a monosomal karyotype. All three patients with ALL in complete remission 1 (CR1) were Philadelphia chromosome positive. Using the adjusted disease risk index,23 patients were classified as low (n=4), intermediate (n=22), or high (n=9) risk.

Engraftment and chimerism There were no cases of graft rejection. The median time to an ANC>500 was 18 days (range: 14-26) (Figure 1A). 719

W.R. Drobyski et al. A

Twenty-one patients received granulocyte-colony stimulating factor (G-CSF) for 1-3 days on days 14-19 posttransplantation to accelerate white blood cell (WBC) recovery. The platelet count in five patients never dropped below 20,000. Two additional patients died before achieving engraftment of platelets. In the remaining patients, the median time to platelet engraftment was 17 days (range: 10-103) (Figure 1B). Chimerism studies conducted on day 28 post-transplantation were available for 33 patients. Median donor CD3 chimerism was 88% (range: 34-100), while median CD33 chimerism was 100% (range: 97100). Day 100 studies were performed in 27 patients and revealed a median donor CD3 chimerism of 91% (range: 43-100) and CD33 chimerism of 100% (range: 91-100).

GvHD, transplant-related mortality, relapse, and survival The cumulative incidence of grades II-IV aGvHD at days 100 and 180 was 14% (95% CI 5-30%) and 17% 720

Figure 1. Engraftment, GvHD, disease-free survival, and overall survival. (A). Cumulative incidence of achieving an absolute neutrophil count >500/mm3 for three consecutive days. (B). Cumulative incidence of patients that achieved an unsupported platelet count Five > 20,000/mm3. patients never dropped their platelet count below 20,000. (C,D). Cumulative incidence of grades II-IV and grades III-IV aGvHD. (E). Cumulative incidence of grades II-IV aGvHD in patients that received myeloablative (MA) versus reduced intensity (RIC) preparative regimens. (F). Cumulative incidence of NIH-defined cGvHD. (G) Probability of disease-free survival and (H) overall survival in patients that received Toc/Tac/MTX as GvHD prophylaxis. Dashed gray lines indicate 95% confidence interval bands.

(95% CI 7-35%), respectively, (Figure 1C). The incidence of grades III-IV aGvHD at these same time points was 3% (95% CI 0-11%) and 6% (95% CI 1-16), respectively, (Figure 1D). All patients who developed â&#x2030;Ľ grade II aGvHD within the first 100 days had isolated involvement of the skin or upper GI tract. Three patients had GvHD of the skin, one of whom developed grade IV disease, which proved to be fatal. Three patients had upper GI tract involvement, which was resolved in each case with modest doses of steroids. There were no cases of aGvHD involving the liver or lower GI tract within this time interval. One patient did develop grade II-IV aGvHD involving the lower GI tract between days 100180 post-transplantation. A second patient with a prior history of overall grade III skin involvement developed upper GI tract disease in the duodenum on day 177. The cumulative incidence of grades II-IV aGvHD for each individual tissue site is shown in Online Supplementary Figure S1. There was no difference in the incidence of haematologica | 2018; 103(4)

Tocilizumab for the prevention of acute GvHD

Figure 2. Cytokine levels in control population that received tacrolimus and methotrexate for GvHD prophylaxis. Concentration of IL-6, sIL-6R, IL-2, IL-4, IL-10, IL-17, TNF-α, and IFN-g in the serum of patients (n=11) who received Tac/MTX for the prevention of aGvHD prior to the start of conditioning, and at days 7, 14 and 28. *P<0.05, ***P<0.001. IL: interleukin; sIL: soluble interleukin; TNF-α: tumor necrosis factor α; IFN-g: interferon g.

grades II-IV aGvHD at day 180 between patients who received MA (17%, 95% CI 4-37) versus reduced intensity (18%, 95% CI 4-39) conditioning regimens (Figure 1E). The median time to onset for aGvHD in recipients of MA versus reduced intensity transplants was 44 and 78 days, respectively. The incidence of cGvHD at 12 months was 38% (95% CI 21-55) in this patient population (Figure 1F). The median follow up for surviving patients was 15 months. TRM was 14% (95% CI 5-28) at 12 months and was attributable to GvHD (n=3), sepsis (n=1), aspergillus pneumonia (n=1), idiopathic pneumonia syndrome (n=1), and respiratory failure (n=1). The one-year cumulative incidence of relapse was 29% (95% CI 15-44). DFS and OS at 12 months was 57% (95% CI 42-70) and 68% (95% CI 53-79), respectively, (Figures 1G,H).

Side effects and infectious complications There were no infusion-related reactions associated with the administration of Toc. The major immediate Tocassociated side effect within the first 28 days post-transplantation was the development of grade III liver toxicity. Nine patients (26%) had ≥ grade III ALT elevations, two patients (6%) had ≥ grade III AST levels and one patient had a grade IV bilirubin elevation. Transaminase elevations typically peaked 7-10 days after infusion, and were transient in all patients, eventually returning to baseline. The marked bilirubin level in one patient was ascribed to total parenteral nutrition administration after a biopsy revealed no evidence of GvHD or any other underlying pathology. No patient developed veno-occlusive disease of the liver. A total of 23 grade III or higher infectious complications were observed in 13 patients during the first 100 days. Fifteen of these were due to bacterial infections, of which 11 were bloodstream (staphylococcus epidermidis [n=6], bacillus cereus [n=1], strep oralis [n=1], strep mitis haematologica | 2018; 103(4)

[n=1], polymicrobial sepsis [n=1], vancomycin-resistant enterococcus [n=1]), two urinary tract (staphylococcus epidermidis [n=1], vancomycin-resistant enterococcus [n=1]), one respiratory (staphylococcus epidermidis), and one attributable to clostridium difficile. Cytomegalovirus (CMV) reactivation occurred in two of 12 (17%) seropositive recipients. Other viral infections consisted of human herpes virus 6 (HHV-6) encephalitis (n=1), enterovirus (n=1), and BK virus (n=3). One patient developed invasive aspergillus pneumonia.

Inflammatory cytokine analyses To examine how the administration of Toc altered inflammatory cytokine production, we assayed serum cytokine levels (IL-2, IL-4, IL-6, IL-10, IL-17A, tumor necrosis factor α [TNF-α] and interferon g [IFN-g]) and soluble (s)IL-6R levels in the peripheral blood of patients who received Toc (n=35) as well as a control population (n=11) which had the same trial eligibility criteria, but did not receive this agent (see patient demographics in Online Supplementary Table S2). In the control population, we observed that IL-6 was the only cytokine that was increased above baseline during the first 28 days (i.e., nine-fold increase on day 14 post-transplantation) (Figure 2). Conversely, sIL-6R levels were significantly decreased on days seven and 14 before rebounding back to baseline on day 28. In the Toc cohort, IL-6 levels were increased in patients who received Toc at days seven, 14, and 28 when compared to baseline (Figure 3A). Levels were augmented above baseline in both MA and RIC recipients (Figure 3A), although IL-6 concentrations were higher in patients who received ablative compared to reduced intensity regimens on days seven and 14, but not day 28 (Online Supplementary Figure S2A). sIL-6R levels were also significantly increased beginning on day seven post-transplantation, and were still elevated by day 28 (Figure 3B). Levels 721

W.R. Drobyski et al. A

Figure 3. Effect of tocilizumab administration on interleukin 6 and soluble interleukin 6 receptor levels based on conditioning regimen. (A). Concentration of IL-6 in the serum from patients that were treated with tocilizumab and received myeloablative or reduced intensity conditioning. (B). Concentration of soluble IL-6 receptor in the serum from patients that were treated with tocilizumab and received myeloablative or reduced intensity conditioning. ***P<0.001. IL: interleukin; sIL: soluble interleukin; RIC: reduced intensity conditioning.

were augmented in recipients of reduced intensity as opposed to MA regimens on day 14, but otherwise there were no differences at other time points (Online Supplementary Figure S2B). Of the other cytokines measured in the blood, marginal increases were observed in IL2 at day seven and IL-10 at day 28 (Online Supplementary Figure S3). A direct comparison of cytokine levels between control and Toc-treated patients demonstrated a marked increase in IL-6 and sIL-6R levels in the latter group at all time points (Figure 4). There were also significant but very modest decreases in IL-2, IL-4 and IL-10 in these patients.

Immune reconstitution Patients were tested by multi-parameter flow cytometry for reconstitution of lymphocytes and major lymphocyte subsets (CD3+, CD4+, and CD8+ T cells, B cells, and NK cells) at four intervals over the first year (Figure 5A,B). Patients recovered lymphocyte subsets to, or near, healthy control levels between six months and one year posttransplant, except for NK cells, which recovered early. The percentage of both regulatory T cells (Tregs) and T helper 17 (TH17) cells were within the expected range of healthy donors, and while Treg levels gradually decreased over the post-transplant period, the absolute number of TH17 cells remained stable (Online Supplementary Figure S4). The percentages and absolute number of B cells were particularly low at the one-month and three-month assessments (Online Supplementary Figure S5), and a subset analysis revealed other imbalances. Specifically, a functionally immature/transitional CD21â&#x2C6;&#x2019; subset found to be 722

elevated in association with autoimmunity,24-26 infection,27 and a subset of patients with cGvHD27,28 was increased throughout the assessment period. However, this subset as well as most other abnormalities (including the percentage of antigen inexperienced naĂŻve B cells) recovered through the first year. There were no significant differences seen when comparing patients experiencing aGvHD or cGvHD compared to those who did not experience GvHD for any assessment (data not shown).

Comparison to a matched control population From the CIBMTR database, four controls were identified for 30 patients, three controls for three patients, and two and one control for one patient each. The baseline characteristics for the patients in the phase 2 trial and the control cohort are detailed in Table 2. Median follow up was 15 months (range: 9-20 months) for patients receiving Toc/Tac/MTX and 13 months (range; 3-72 months) for the control cohort. The incidence of grades II-IV aGvHD at day 180 was significantly lower in the Toc/Tac/MTX cohort when compared to the Tac/MTX control population (17% versus 45% at day 180, HR=0.34 [0.17-0.69], P=0.003) (Figure 6A). Furthermore, corresponding probabilities of grade II-IV aGvHD-free survival were significantly higher in patients who received Toc/Tac/MTX than the matched cohort (69% versus 42% at day 180, HR=0.37, [0.21-0.67], P=0.001) (Figure 6B). There was no difference in the incidence of cGvHD between recipients in the Toc/Tac/MTX versus the Tac/MTX groups (38% versus 45% at 12 months, HR=0.65, [0.37-1.13], P=0.13) (Figure 6C). There was also no difference in TRM, haematologica | 2018; 103(4)

Tocilizumab for the prevention of acute GvHD

Table 2. Demographics of tocilizumab trial patients and matched controls, Age, median (range) KPS â&#x2030;Ľ 90 HCT-CI 0 1-2 3+ Disease AML (de novo) AML (secondary) ALL CMML/CML MDS/MPD NHL Donor Type MRD MUD Conditioning Regimen Flu/Bu2 Bu/Cy Flu/Bu4 Graft Source BM PBSC Median Follow up Surviving Patients

Toc/Tac/MTX (n=35)

Tac/MTX (n=130)

66 (23-76) 12 (34)

64 (23-74) 48 (37)

9 (26) 11 (31) 15 (43)

20 (15) 40 (31) 70 (54)

14 (40) 5 (14) 4 (11) 6 (17) 4 (11) 2 (6)

56 (43) 20 (15) 14 (11) 23 (18) 12 (9) 5 (4)

13 (37) 22 (63)

48 (37) 82 (63)

17 (49) 5 (14) 13 (37)

63 (48) 17 (13) 50 (38)

6 (17) 29 (83) 15 months (9-20)

22 (17) 108 (83) 13 months (3-72)

AML: acute myelogenous leukemia; ALL: acute lymphoblastic leukemia; CMML: chronic myelomonocytic leukemia; CML: chronic myelogenous leukemia; MDS: myelodysplasia; MPD: myeloproliferative disorder; NHL: non-Hodgkinâ&#x20AC;&#x2122;s lymphoma; MRD: matched related donor; MUD: matched unrelated donor; BM: bone marrow; PBSC: peripheral blood stem cells; KPS: Karnofsky Performance Score; HCT-CI: Hematopoietic cell transplantation - specific comorbidity index; Bu: busulfan; Cy: cyclophosphamide; Flu: fludarabine; Toc: tocilizumab: Tac: tacrolimus; MTX: methotrexate.

relapse, or DFS at 12 months between the two groups (Figures 6D-6F).

Discussion Inflammatory cytokine production is a proximate event in the pathophysiology of aGvHD.1-3 While a number of inflammatory molecules are produced as a consequence of the conditioning regimen and the activation and expansion of alloreactive donor T cells, IL-6 has emerged as an important cytokine mediator of tissue damage.4,7,8 In the study herein, we demonstrate that inhibition of IL-6 signaling by the administration of Toc in addition to standard immune suppression resulted in a significant reduction in grades II-IV aGvHD and an increase in grades II-IV aGvHD-free survival, when compared to a matched control population. The administration of Toc was also observed to be safe when given in the setting of MA or reduced intensity Bu-based conditioning regimens. Moreover, adverse events were largely confined to transient elevations in transaminase values, and infectious complications were not dissimilar to what we have previously observed in this patient population treated with standard immune suppression only. The results of the current study extend those reported haematologica | 2018; 103(4)

by Kennedy and colleagues,18 who also examined the efficacy of Toc for the prevention of aGvHD. These investigators observed an incidence of grades II-IV and III-IV aGvHD at day 100 of 12% and 3%, respectively, which was similar to what we observed (14% and 3% for these same endpoints). There were, however, several important differences between the two studies, which suggest that the results may be more broadly generalizable to allogeneic HSCT recipients. First of all, the median age of patients in our report was substantially higher (66 versus 48), indicating that Toc administration appears to have activity in older patients who comprise an increasing percentage of the transplant population.20 Secondly, patients in the current report received Bu-based conditioning regimens, whereas those in the study by Kennedy et al. were treated with either total body irradiation and Cy (MA conditioning) or Flu and melphalan (RIC). Since the intensity of the conditioning regimen affects the degree of inflammatory cytokine production19,29 and incidence of aGvHD,30 the fact that promising results were observed in patients who received different MA and reduced intensity regimens is evidence that inhibition of IL-6 may have activity across a spectrum of preparative regimens. Finally, we were able to provide additional context to our data by demonstrating a reduced incidence of grades II-IV aGvHD as well as an increase in grades II-IV GvHD-free survival when com723

W.R. Drobyski et al.

Figure 4. Comparative analysis of serum cytokine production in tocilizumab-treated versus patients that received Tac/MTX only. Concentration of IL-6, sIL-6R, IL-2, IL-4, IL-10, IL-17, IFN-g, and TNF-α in the serum of patients that were treated with Toc/Tac/MTX (, n=35) or Tac/MTX (control) (, n=11) for the prevention of aGvHD prior to the start of conditioning and at days 7, 14 and 28. *P<0.05, **P<0.01, ***P<0.001. IL: interleukin; sIL: soluble interleukin; TNF-α: tumor necrosis factor α; IFN-g: interferon g.

pared to a matched control population that only received Tac/MTX. Despite the reduction in grades II-IV aGvHD, however, there was no difference in TRM or OS between these two groups. There are several possible explanations for this observation. First, this was not a randomized trial, and the matching process could have resulted in unperceived differences between the two cohorts that could have impacted transplant outcome. Secondly, we observed two late GvHD deaths beyond six months, suggesting that the salutary effects conferred by Toc may be temporally limited. That there was no difference in the incidence of cGvHD is compatible with this interpretation. This does not, in our view, diminish the results, but rather highlights that effective prophylactic strategies for GvHD are likely to require a multi-tiered approach of which the mitigation of aGvHD within the first six months would be one important step. A notable finding in this study was the very low incidence of aGvHD that occurred in the lower GI tract. Specifically, there were no cases within the first 100 days, and only one case which occurred by day 180. Pre-clinical studies have shown that IL-6 messenger ribonucleic acid (mRNA) levels are significantly increased in the colons of mice,7 and the blockade of the IL-6 signaling pathway is able to significantly reduce the severity of GvHD in this tissue site.7,8 IL-6 has also been identified as a plasma biomarker that predicts for severity and nonrelapse mortality in patients with GI GvHD.31 Our findings further support the premise that IL-6 plays an important role in mediating tissue damage in the lower GI tract. Given that the incidence of liver GvHD has been declining over time,32 involvement of the GI tract has emerged 724

as the primary driver of morbidity and mortality in patients with this disease. In fact, the development of lower GI tract GvHD carries significant prognostic implications for OS. Patients with lower tract GvHD are more likely to be steroid-resistant,33 which itself is associated with increased mortality.34 Furthermore, patients with higher clinical and histological grades of lower GI tract GvHD have an increase in non-relapse mortality that results in reduced OS.35,36 Thus, given the poor prognosis associated with severe GI GvHD,37,38 therapeutic strategies that are focused on preventing the development of this complication have the potential to impact the overall course of this disease and improve transplant outcome. While we observed patients with upper GI tract GvHD, recent studies have shown that disease in this tissue site is generally responsive to modest doses of steroids, and does not impact OS.39,40 We observed that IL-6 was the only measured serum cytokine that was significantly increased above baseline in a control population of patients that received Tac and MTX but not Toc. The administration of Toc resulted in much higher serum IL-6 levels, above that seen in the control population, in recipients of both MA and RIC regimens. This was likely due to decreased consumption of IL-6 when Toc binds to the IL-6R. IL-6 signaling occurs through two distinct mechanisms; IL-6 can bind to a membrane receptor that is expressed on hematopoietic cells and hepatocytes,41 and also bind to a soluble form of the IL-6 receptor, which can in turn bind to glycoprotein (gp)130, which is ubiquitously expressed on most cells.42,43 As further support for this premise, we observed that sIL6R levels were also significantly augmented in patients who received both MA and RIC regimens. Of note, the haematologica | 2018; 103(4)

Tocilizumab for the prevention of acute GvHD

Figure 5. Reconstitution of major lymphocyte subsets in patients who received tocilizumab for GvHD prophylaxis. (A,B). The absolute number of cells per mm3 (i.e., microliter) is shown in panel A, and the percentage of the gated cells is shown in panel B. Data are shown for individual patients together with the median and 25th and 75th quartiles (red bars). Gray shading represents the upper and lower range expected for healthy control subjects. Samples were obtained at one month (n=33), three months (n=29), six months (n=22), and 12 months (n=13). Lymphocytes were gated on total CD45+ white blood cells, and all other subsets were gated on lymphocytes. NK: natural killer.

immunoassay used to detect sIL-6R levels captures free sIL-6R, IL-6R bound to IL-6, and IL-6R bound to Toc.44 Therefore, we cannot distinguish the composition of the sIL-6R complex, but it is likely that a significant component is attributable to the binding of Toc to IL-6R, given that a prior study showed that Toc can be detected for up to one month in allogeneic stem cell transplant recipients.18 This would therefore also explain the high IL-6 levels in these patients, as free IL-6 may have been precluded haematologica | 2018; 103(4)

from binding to the Toc/sIL-6R complex. The fact that IL6 levels were higher in patients treated with MA versus RIC, however, suggests that the conditioning regimen itself also contributed to the increase in IL-6 levels. Notably, we did not observe meaningful increases in any of the other cytokines that we examined (i.e., IL-2, IL-4, IL-10, TNF-Îą, IFN-g and IL-17) in control or Toc-treated patients, providing evidence that dysregulation of IL-6 is an important early event post-transplantation. Toc had no 725

W.R. Drobyski et al. A

discernible adverse effect on immune reconstitution, as patients achieved near normal T-cell and B-cell subset numbers by 6-12 months post-transplantation. In summary, this study demonstrates that Toc can be safely administered in conjunction with standard immune suppression to an older aged patient cohort treated with a Bu-based conditioning for the prevention of GvHD. The administration of Toc resulted in a low incidence of aGvHD, which was particularly evident within the lower GI tract, and was significantly less than that observed in a

References 1. Ferrara JLM, Levine JE, Reddy P, Holler E. Graft-versus-host-disease. Lancet. 2009; 373(9674):1550-1561. 2. Welniak LA, Blazar BR, Murphy WJ. Immunobiology of allogeneic hematopoietic stem cell transplantation. Ann Rev Immunol. 2007;25:139-170. 3. Shlomchik WD. Graft-versus-host disease. Nat Rev Immunol. 2007;7(5):340-352. 4. Tvedt THA, Ersvaer E, Tveita AA, Bruserud O. Interleukin 6 in allogeneic stem cell transplantation: Its possible importance for immunoregulation and as a

726

Figure 6. GvHD and transplant outcomes in patients treated with tocilizumab versus a matched CIBMTR control population. (A). Cumulative incidence of grades II-IV aGvHD in patients treated with tocilizumab versus the matched control cohort. (B). Probability of grades II-IV aGvHD-free survival. (C). Cumulative incidence of cGvHD, (D) transplant-related mortality, and (E) relapse. (F) Probability of disease-free survival in patients treated with tocilizumab versus the matched control cohort. Toc: tocilizumab: Tac: tacrolimus; MTX: methotrexate.

matched control population. There was, however, no difference in the incidence of cGvHD or a reduction in TRM. We conclude that Toc has activity for the prevention of aGvHD, and warrants further examination in a randomized setting. Funding This research was supported by a grant from the Kurtis Froedtert Foundation at the Medical College of Wisconsin Cancer Center.

therapeutic target. Front Immunol. 2017;8:667. 5. Toubai T, Mathewson ND, Magenau J, Reddy P. Danger signals and graft versus host disease: Current understanding and future perspective. Front Imunol. 2016;7:539. 6. Choi SW, Reddy P. Current and emerging strategies for the prevention of graft versus host disease. Nat Rev Clin Oncol. 2014;11(9):536-547. 7. Chen X, Das R, Komorowski R, et al. Blockade of interleukin 6 signaling augments regulatory T cell reconstitution and attenuates the severity of graft versus host disease. Blood. 2009;114(4):891-900.

8. Tawara I, Koyama M, Liu C, et al. Interleukin 6 modulates graft versus host responses after experimental allogeneic bone marrow transplantation. Clin Can Res. 2011;17(1):77-88. 9. Barak V, Levi-Schaffer F, Nisman B, Nagler A. Cytokine dysregulation in chronic graft versus host disease. Leuk Lymphoma. 1995;17(1-2):169-173. 10. Imamura M, Hashino S, Kobayashi H, et al. Serum cytokine levels in bone marrow transplantation: synergistic interaction of interleukin 6, interferon gamma, and tumor necrosis factor alpha in graft versus host disease. Bone Marrow Transplant. 1994;13(6):745-751.

haematologica | 2018; 103(4)

Tocilizumab for the prevention of acute GvHD

11. Cavet J, Dickinson AM, Norden J, Taylor PRA, Jackson GH, Middleton PG. Interferon γ and interleukin 6 gene polymorphisms associate with graft versus host disease in HLA-matched sibling bone marrow transplantation. Blood. 2001;98(5): 1594-1600. 12. Socie G, Loiseau P, Tamouza R, et al. Both genetic and clinical factors predict the development of graft versus host disease after allogeneic hematopoietic stem cell transplantation. Transplantation. 2001; 72(4):699-706. 13. Alam N, Xu W, Atenafu EG, et al. Risk model incorporating donor IL-6 and IFNG genotype and gastrointestinal GVHD can discriminate patients at high risk of steroid refractory acute GVHD. Bone Marrow Transplant. 2015;50(5):734-742. 14. Chien JW, Zhang XC, Fan W, et al. Evaluation of published single nucleotide polymorphisms associated with acute GVHD. Blood. 2012;119(22):5311-5319. 15. Nishimoto N, Yoshizaki K, Miyasaka N, et al. Treatment of rheumatoid arthritis with humanized anti-interleukin 6 receptor antibody. Arthritis Rheum. 2004;50(6):17611769. 16. Yokota S, Imagawa T, Mori M, et al. Efficacy and safety of tocilizumab in patients with systemic-onset juvenile idiopathic arthritis: a randomized, double-blind, placebo-controlled, withdrawal phase III trial. Lancet. 2008;371(9617):998-1006. 17. Drobyski WR, Pasquini M, Kovatovic K, et al. Tocilizumab for the treatment of steroid refractory graft versus host disease. Biol Blood Marrow Transplant. 2011; 17(12):1862-1868. 18. Kennedy GA, Varelias A, Vuckovic S, et al. Addition of interleukin 6 inhibition with tocilizumab to standard graft versus host disease prophylaxis after allogeneic stem cell transplantation: a phase 1/2 trial. Lancet Oncol. 2014;15(13):1451-1459. 19. Xun CQ, Thompson JS, Jennings CD, Brown SA, Widmer MB. Effect of total body irradiation, busulfan-cyclophosphamide, or cyclophosphamide conditioning on inflammatory cytokine release and development of acute and chronic graft versus host disease in H-2 incompatible transplanted SCID mice. Blood. 1994; 83(8):2360-2367. 20. D'Souza A, Lee S, Zhu X, Pasquini M. Current use and trends in hematopoietic cell transplantation in the United States. Biol Blood Marrow Transplant. 2017; 23(9):1417-1421. 21. Przepiorka D, Weisdorf D, Martin P, et al. 1994 Consensus conference on acute GVHD grading. Bone Marrow

haematologica | 2018; 103(4)

Transplant. 1995;15(6):825–828. 22. Jagasia MH, Greinix HT, Arora M, et al. National Institutes of Health Consensus Development Project on criteria for clinical trials in chronic graft versus host disease: I. The 2014 diagnosis and staging working group. Biol Blood Marrow Transplant. 2015;21(3):389-401. 23. Armand P, Kim HT, Logan BR, et al. Validation and refinement of the disease risk index for allogeneic stem cell transplantation. Blood. 2014;123(23):3664-3671. 24. Wehr C, Eibel H, Masilamani M, et al. A new CD21low B cell population in the peripheral blood of patients with SLE. Clin Immunol. 2004;113(2):161-171. 25. Isnardi I, Ng YS, Menard L, et al. Complement receptor 2/CD21− human naive B cells contain mostly autoreactive unresponsive clones. Blood. 2010;115(24): 5026-5036. 26. Shirota Y, Yarboro C, Fischer R, Pham TH, Lipsky P, Illei GG. Impact of anti-interleukin-6 receptor blockade on circulating T and B cell subsets in patients with systemic lupus erythematosus. Ann Rheum Dis. 2013;72(1):118-128. 27. Greinix HT, Pohlreich D, Kouba M, et al. Elevated numbers of immature/transitional CD21− B lymphocytes and deficiency of memory CD27+ B cells identify patients with active chronic graft-versus-host disease. Biol Blood Marrow Transplant. 2008;14(2):208-219. 28. Kuzmina Z, Krenn K, Petkov V, et al. CD19(+)CD21(low) B cells and patients at risk for NIH-defined chronic graft-versushost disease with bronchiolitis obliterans syndrome. Blood. 2013;121(10):1886-1895. 29. Hill GR, Crawford JM, Cooke KR, Brinson YS, Pan L, Ferrara JL. Total body irradiation and acute graft versus host disease: the role of gastrointestinal damage and inflammatory cytokines. Blood. 1997;90(8):32043213. 30. Flowers ME, Inamoto Y, Carpenter PA, et al. Comparative analysis of risk factors for acute graft versus host disease and for chronic graft versus host disease according to National Institutes of Health consensus criteria. Blood. 2011;117(11):3214-3219. 31. McDonald GB, Tabellini L, Storer BE, Lawler RL, Martin PJ, Hansen JA. Plasma biomarkers of acute GVHD and nonrelapse mortality: predictive value of measurements before GVHD onset and treatment. Blood. 2015;126(1):113-120. 32. Gooley TA, Chien JW, Pergam SA, et al. Reduced mortality after allogeneic hematopoietic cell transplantation. New Engl J Med. 2010;363(22):2091-2101. 33. MacMillan ML, Weisdorf DJ, Wagner JE, et

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

al. Response of 443 patients to steroids as primary therapy for acute graft versus host disease: comparison of grading systems. Biol Blood Marrow Transplant. 2002;8(2): 387-394. Castilla-Llorente C, Martin PJ, McDonald GB, et al. Prognostic factors and outcomes of severe gastrointestinal GVHD after allogeneic hematopoietic cell transplantation. Bone Marrow Transplant. 2014;49(7):966971. Macmillan ML, Robin M, Harris AC, et al. A refined risk score for acute graft versus host disease that predicts response to initial therapy, survival, and transplant-related mortality. Biol Blood Marrow Transplant. 2015;21(4):761-767. Ferrara JL, Harris AC, Greenson JK, et al. Regenerating islet-derived 3-alpha is a biomarker of gastrointestinal graft versus host disease. Blood. 2011;118(25):6702-6708. Pidala J, Anasetti C. Glucocorticoid-refractory acute graft versus host disease. Biol Blood Marrow Transplant. 2010; 16(11):1504-1518. Jamani K, RussellJA, Daly A, et al. Prognosis of grade 3-4 acute GVHD continues to be dismal. Bone Marrow Transplant. 2013;48(10):1359-1361. Mielcarek M, Storer BE, Boeckh M, et al. Initial therapy of acute graft versus host disease with low dose prednisone does not compromise patient outcomes. Blood. 2009;113(13):2888-2894. Castilla C, Perez-Simon JA, Sanchez-Guijo FM, et al. Oral beclomethasone dipropionate for the treatment of gastrointestinal acute graft versus host disease (GVHD). Biol Blood Marrow Transplant. 2006;12(9):936-941. Rose-John S, Scheller J, Elson G, Jones SA. Interleukin-6 biology is coordinated by membrane-bound and soluble receptors: role in inflammation and cancer. J Leuk Biol. 2006;80(2):227-236. Mullberg J, Schooltink H, Stoyan T, et al. The soluble interleukin 6 receptor is generated by shedding. Eur J Immunol. 1993;23(2):473-480. Jones SA, Horiuchi S, Topley N, Yamamoto N, Fuller GM. The soluble interleukin-6 receptor: mechanisms of production and implications in disease. FASEB J. 2001;15(1):43-58. Nishimoto N, Terao K, Mima T, Nakahara H, Takagi N, Kakehi T. Mechanisms and pathologic significances in increase in serum interleukin 6 (IL-6) and soluble IL-6 receptor after administration of an anti-IL6-receptor antibody, tocilizumab, in patients with rheumatoid arthritis. Blood. 2008;112(10):3959-3964.

727

ARTICLE

Hemostasis

Ferrata Storti Foundation

Macrophage scavenger receptor SR-AI contributes to the clearance of von Willebrand factor

Nikolett Wohner,1* Vincent Muczynski,1* Amel Mohamadi,1 Paulette Legendre,1 Valérie Proulle,1,2 Gabriel Aymé,1 Olivier D. Christophe,1 Peter J. Lenting,1 Cécile V. Denis1 and Caterina Casari1

Haematologica 2018 Volume 103(4):728-737

Institut National de la Santé et de la Recherche Médicale, UMR_S 1176, Univ. ParisSud, Université Paris-Saclay, 94276 Le Kremlin-Bicêtre and 2Service d’Hématologie Biologique, Centre Hospitalier Universitaire Bicêtre, Assistance Publique- Hôpitaux de Paris, 94276 Le Kremlin-Bicêtre, France 1

NW and VM contributed equally to this work.

ABSTRACT

Correspondence: peter.lenting@inserm.fr

Received: June 25, 2017. Accepted: December 27, 2017. Pre-published: January 11, 2018. doi:10.3324/haematol.2017.175216 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/728 ©2018 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.

728

reviously, we found that LDL-receptor related protein-1 on macrophages mediated shear stress-dependent clearance of von Willebrand factor. In control experiments, however, we observed that von Willebrand factor also binds to macrophages independently of this receptor under static conditions, suggesting the existence of additional clearance-receptors. In search for such receptors, we focused on the macrophage-specific scavenger-receptor SR-AI. von Willebrand factor displays efficient binding to SR-AI (half-maximum binding 14±5 nM). Binding is calcium-dependent and is inhibited by 72±4% in the combined presence of antibodies against the A1- and D4-domains. Association with SR-AI was confirmed in cell-binding experiments. In addition, binding to bone marrow-derived murine SR-AI-deficient macrophages was strongly reduced compared to binding to wild-type murine macrophages. Following expression via hydrodynamic gene transfer, we determined ratios for von Willebrand factor-propeptide over von Willebrand factorantigen, a marker of von Willebrand factor clearance. Propeptide/antigen ratios were significantly reduced in SR-AI-deficient mice compared to wild-type mice (0.6±0.2 versus 1.3±0.3; P<0.0001), compatible with a slower clearance of von Willebrand factor in SR-AI-deficient mice. Interestingly, mutants associated with increased clearance (von Willebrand factor/p.R1205H and von Willebrand factor/p.S2179F) had significantly increased binding to purified SR-AI and SR-AI expressed on macrophages. Accordingly, propeptide/antigen ratios for these mutants were reduced in SR-AI-deficient mice. In conclusion, we have identified SR-AI as a novel macrophage-specific receptor for von Willebrand factor. Enhanced binding of von Willebrand factor mutants to SR-AI may contribute to the increased clearance of these mutants. Introduction Mutations in the gene encoding von Willebrand factor (VWF) may generate proteins displaying defects in biosynthesis, secretion and/or clearance. The quintessential representative of VWF clearance mutants is VWF/p.R1205H, also known as the Vicenza variant.1 This mutation is associated with VWF antigen (VWF:Ag) levels that are usually below 20%, which are most likely due to a 5- to 10-fold reduced circulatory half-life of the mutant protein.1-3 Since the description of the Vicenza variant, additional mutations in VWF have been found to provoke a reduced survival. Such mutants have been identified by analyzing VWF survival after desmopressin treatment or by determining ratios between VWF propeptide (VWFpp) and VWF:Ag, with elevated VWFpp/VWF:Ag ratios pointing to increased VWF clearance.4-8 The mechanism by which mutations provoke accelerated clearance is poorly understood. Recently, we showed that gain-of-function mutations in the VWF A1 domain induce spontaneous binding to the clearance receptor LDL-receptor related protein-1 (LRP1).9 In addition, truncation of its N-glycans accelerates LRP1-mediathaematologica | 2018; 103(4)

SR-AI contributes to VWF clearance

ed uptake by macrophages.10,11 This is in contrast to nonmodified VWF, which only binds to macrophageexpressed LRP1 when exposed to increased shear stress.12 Apart from LRP1, other receptors have also been described to contribute to VWF clearance, including asialoglycoprotein receptor, CLEC4M and Siglec-5.13-16 Despite the various receptors having been identified, their functional absence is usually associated with a mildto-modest effect on VWF clearance. We previously observed that the majority of VWF is targeted to macrophages, and that chemical depletion of macrophages results in a 2- to 3-fold increase in VWF levels.17,18 We therefore explored the hypothesis that macrophages express one or more additional receptors that contribute to VWF clearance. Here, we present data that are compatible with the macrophage-specific receptor Scavenger receptor class A member I (SR-AI) being a clearance-receptor for VWF. Moreover, we show that two clearance mutants (VWF/p.R1205H and VWF/p.S2179F) display enhanced binding to SR-AI, providing a potential explanation for their reduced half-life in the circulation.

Microscopy analyses and immunofluorescence-based quantification

Methods

A detailed description of the microscopic analysis is provided in the Online Supplementary Material. Briefly, after saturation of nonspecific binding sites, cells were exposed to primary antibodies for 2 h at room temperature, followed by incubation for 1 h with secondary antibodies. Nuclei were counterstained with 4’,6’-diamidino-2-phenylindole. Alexa-Fluor647 or Alexa-Fluor488-labeled phalloïdin was used to determine cell boundaries. For the Duolink-Proximity Ligation Assay (Duolink-PLA) to detect close proximity between different proteins, double immunostaining was performed as described above with the second antibodies replaced by PLA probes (Sigma-Aldrich). The remainder of the protocol was conducted according to the manufacturer’s recommendations and the 550 nm wavelength detection kit was used. Hybridization between the two PLA probes leading to the fluorescent signal only occurs when the distance between the two detected antigens is less than 40 nm. Images were analyzed using ImageJ software for quantification of fluorescence by measuring the total pixel intensity per cell. Duolink-PLA experiments were analyzed using BlobFinder software (Uppsala University, Sweden) to quantify the number of fluorescent spots per cell. All images were assembled using ImageJ software.

Ethics statement

Immunosorbent binding assay

Animal housing and experiments were done as recommended by French regulations and the experimental guidelines of the European Community. This project was approved by the local ethical committee CEEA 26 (# 2012-036).

Binding experiments are described in detail in the Online Supplementary Material.

Proteins A detailed description of the proteins used in this study is provided in the Online Supplementary Material. The main proteins included plasma-derived (pd)-VWF purified from VWF concentrates (Wilfactin) and recombinant full-length VWF [wild-type (wt)-VWF, VWF/p.R1205H, VWF/p.V1316M, and VWF/p.S2179F] produced in stably transfected BHK-furin cells using serum-free medium. All variants displayed a similar distribution of multimers (data not shown). Non-purified cell culture supernatants were used for protein- and cellular binding experiments. Other proteins are detailed in the Online Supplementary Material.

Cell culture Detailed information on cell cultures is provided in the Online Supplementary Material. Briefly, human macrophages were differentiated from the THP1 acute monocytic leukemia cell line using phorbol 12-myristate 13-acetate, human macrophage colonystimulating factor and human granulocyte-macrophage colonystimulating factor as described elsewhere.18,19 Non-transfected and stable HEK293 cell lines expressing human SR-AI were cultured as described presviously.19 Murine macrophages were obtained from CD115+ cells as described by Breslin et al.20

Cellular binding experiments A detailed description of cellular binding experiments is provided in the Online Supplementary Material. Briefly, cells seeded on glass coverslips were incubated with purified pd-VWF (10 mg/mL) for 1 h at 37°C. Where indicated, culture medium containing recombinant wt-VWF, VWF/p.R1205H, VWF/p.V1316M or VWF/p.S2179F was used. In other experiments, pd-VWF was preincubated with monoclonal antibodies to VWF (MAb723 and MAb540, 167 mg/mL) for 30 min at room temperature. Following incubation, cells were gently washed twice and then fixed for 15 min with 4% paraformaldehyde at 37°C. haematologica | 2018; 103(4)

Mice Wild-type C57Bl/6 mice were purchased from Janvier Labs (Le Genest-Saint-Isle, France) and SR-AI-deficient C57Bl/6 mice (B6.Cg-Msr1tm1Csk/J) were from The Jackson Laboratory (Bar Harbor, ME, USA). Wild-type mice and SR-AI-deficient mice were not true littermates. MacLRP1-positive and macLRP1-deficient mice were described previously.12 MacLRP1 mice have a C57Bl/6 background, with >12 backcross steps.

von Willebrand factor propeptide to antigen ratios cDNA encoding human wild-type VWF, mutant VWF/p.R1205H or mutant VWF/p.S2179F was cloned into the pLIVE-plasmid (Mirus Bio LLC, Madison, WI, USA) and expressed in wild-type, macLRP1 and/or SR-AI-deficient mice following hydrodynamic gene transfer, essentially as described elsewhere.9, 21-23 Four days after injection, blood samples were taken via retroorbital puncture and plasma was prepared for the analysis of VWF propeptide (VWFpp; cat# MW1939; Sanquin Blood Supply, Amsterdam, the Netherlands) and VWF:Ag (in-house enzymelinked immunosorbent assay using a pool of monoclonal murine anti-VWF antibodies). Both assays are specific for human VWFpp and VWF:Ag, and do not detect murine proteins. VWF:Ag levels varied between 300% and 800% of normal for all constructs. VWF clearance remains unsaturated by VWF levels up to 1500%. The presence of endogenous VWF in the mouse strains used in this study does, therefore, leave clearance of hepatocyte-expressed human VWF unaffected.

Results von Willebrand factor binds to macrophages in an LRP1-independent manner Previously, we demonstrated that VWF binds to LRP1 in a shear-stress-dependent manner or when it contains type 2B gain-of-function mutations.9,12 This specific binding was 729

N. Wohmer et al.

addressed in Duolink-PLA experiments using THP1derived macrophages. This assay is positive when a receptor/ligand pair is within a 40-nm radius. As expected, no red spots above background levels could be detected when wt-VWF was analyzed for binding to LRP1, whereas numerous red spots became apparent when THP1macrophages were incubated with the VWF-type 2B mutant VWF/p.V1316M (Figure 1A-C). This difference was confirmed when quantifying fluorescent signals (Figure 1D; n=5). In control experiments, we also stained THP1-macrophages in a classical manner for the presence of VWF. Interestingly, for those cells incubated with pd-

VWF, the majority of cells proved positive for the presence of VWF (Figure 1E,F). A similar binding of VWF was observed when human macrophages derived from circulating blood precursors were used (data not shown). These data indicate that VWF is able to interact with macrophages also in an LRP1-independent manner, even under static conditions. Apparently, receptors other than LRP1 are present on macrophages that mediate binding of VWF.

Macrophage-specific receptor SR-AI as a potential receptor for von Willebrand factor In search for alternative receptors for VWF, we explored

Figure 1. von Willebrand factor can bind to macrophages independently of LRP1. (A-D) THP-derived macrophages were incubated in the absence or presence of culture-medium containing wt-VWF or VWF/p.V1316M. Association with LRP1 was detected using Duolink-PLA analysis by combining anti-VWF and anti-LRP1 antibodies. Representative images are shown in panels A-C (objective 63x). Quantification of fluorescent signals is shown in panel D. Data represent meanÂąSD (n=5 microscopic fields; 2-7 cells/field). Statistical analysis involved one-way analysis of variance followed by the Tukey multiple comparison test. (E, F) Classical immunefluorescence staining of THP1-derived macrophages incubated in the (E) absence or (F) presence of purified pd-VWF. Bound VWF was probed using monoclonal anti-VWF antibodies (objective 40x). Scale bars represent 10 mm in all panels.

730

haematologica | 2018; 103(4)

SR-AI contributes to VWF clearance

the option that SR-AI could be one of these receptors. SRAI (also known as SCARA1 or CD204) is specifically expressed on macrophages and is structurally related to SCARA5, an epithelial cell-specific receptor that was identified in genome-wide association studies to be linked to VWF plasma levels.24 We first analyzed binding of VWF to the soluble extracellular domain of SR-AI (sSR-AI) in an immunosorbent-based assay. Whereas no binding of VWF to albumin-coated control wells was observed, VWF displayed saturable and dose-dependent binding to immobilized sSR-AI (half-maximal binding 3.5±1.2 mg/mL corresponding to 14±5 nM; n=5) (Figure 2A). We next assessed

the capacity of sSR-AI to interact with various immobilized VWF fragments. sSR-AI bound dose-dependently to each of the three VWF fragments tested (the recombinant D’D3 and A1-A2-A3 regions and the D4/Fc fragment) (Figure 2B), suggesting that the interaction with SR-AI involves different regions of the VWF molecule. Further experiments showed that the A1 domain mediated binding of the A1-A2-A3 fragment to sSR-AI, while both the A2 and A3 domains were incapable of associating with sSR-AI (Figure 2C). In addition, incorporation of the VWD-type 2B mutation VWF/p.V1316M left binding of the A1-domain to sSR-AI unaffected (Figure 2C). Binding

Figure 2. von Willebrand factor interacts with SR-AI via multiple interactive sites. (A) Wells coated with recombinant human sSR-AI (closed circles) or bovine serum albumin (BSA) (open circles) were incubated with various concentrations of purified pd-VWF (0-5 mg/mL). Bound VWF was probed with peroxidase-labeled polyclonal anti-VWF antibodies. (B) Wells coated with recombinant VWF-fragments (closed squares: A1-A2-A3 domain; closed circles: D4-domain fused to Fc fragment; open squares: D’-D3 domains) or BSA (control; open circles) were incubated with various concentrations of sSR-AI (0-5 mg/mL). Bound sSR-AI was probed using biotinylated polyclonal anti-SR-AI antibodies followed by peroxidase-labeled streptavidin. (C) Wells coated with sSR-AI were incubated with various concentrations (0-5 mg/mL) of recombinant VWF-fragments (closed circles: A1-Fc; gray squares: A1-Fc/p.V1316M; gray circles: A2:Fc; open squares: A3-Fc). Bound fragments were probed using peroxidase-labeled polyclonal anti-human Fc antibodies. Control represents binding of A1-Fc to BSA-coated wells (control; open circles). Other fragments gave similar background signals. (D) sSR-AI-coated wells were incubated with recombinant D4-Fc fragment in the presence of various concentrations of recombinant A1-A2-A3 fragment (open circles) or recombinant D’D3 fragment (gray circles). Alternatively, sSR-AI-coated wells were incubated with recombinant A1-Fc fragment in the presence of various concentrations of recombinant D’D3 fragment (closed squares). Bound fragments were probed using peroxidase-labeled polyclonal anti-human Fc antibodies. (E) sSR-AI-coated wells were incubated with VWF (2.5 mg/mL) in the absence or presence of EDTA (10 mM), monoclonal anti-VWF antibody MAb723 or MAb540 (25 mg/mL) or with both antibodies simultaneously (25 mg/mL each). Bound VWF was probed using peroxidase-labeled polyclonal anti-VWF antibodies. For all panels, bound antibodies were detected via TMB-hydrolysis. Data represent mean±SD (n=3-5).

haematologica | 2018; 103(4)

731

N. Wohmer et al.

of wt-VWF to SR-AI was also not altered upon incubation with ristocetin or botrocetin, indicating that binding does not require VWF to be in its platelet-binding conformation (data not shown). Interestingly, we detected little competition between fragments (Figure 2D), suggesting that each domain binds to a distinct site within SR-AI, which at best overlap with each other. SR-AI function is partially cation-dependent,25 and we therefore assessed binding of VWF in the presence of EDTA. This revealed that binding of VWF to SR-AI was reduced by 88±6% in the presence of EDTA (Figure 2E). Specificity was further investigated by testing the effect of monoclonal antibodies targeting VWF. We identified two antibodies (i.e. Mab723 and Mab540, directed against the A1 and D4 domain, respectively) that each partially inter-

fered with the binding of VWF to sSR-AI (48±2% and 47±2% inhibition by Mab723 and Mab540, respectively; n=3) (Figure 2E). A combination of both antibodies reduced binding by 72±4% (n=4; Figure 2E). Thus, purified VWF interacts in a specific manner with SR-AI and several domains of the VWF molecule contribute to this interaction.

von Willebrand factor binds to cellular SR-AI We next tested whether VWF was able to bind cell-surface exposed SR-AI. First, binding of VWF to non-transfected or SR-AI-transfected HEK293 cells was examined. Whereas no VWF staining could be detected on non-transfected HEK293 cells, clear VWF staining was present on SR-AI-expressing HEK293 cells (Figure 3A-D). High-reso-

Figure 3. Binding of van Willebrand factor to SR-AI-expressing cells. (A-D) Non-transfected (A&C) and hSR-AI-transfected HEK293-cells (B&D) were incubated with purified pd-VWF (10 mg/mL). hSR-AI and bound VWF were probed using polyclonal anti-hSR-AI (red, A&B) and anti-VWF antibodies (green, C&D). Images were obtained via widefield microscopy (objective 40x; scale bars: 10 mm). (E-F) Spinning disk microscopy images (objective 63x; scale bars: 5 mm, z-depth 0.5 mm) of hSR-AI-transfected HEK293-cells incubated with pd-VWF (10 mg/mL). Cells were probed for VWF and hSR-AI (E, green and red, respectively) and for VWF and EEA-1 (F, green and red, respectively). Arrows indicate areas of overlapping signals. (G-J) Non-transfected and SR-AI-transfected HEK293 cells were incubated in the absence or presence of pd-VWF (10 mg/mL). Association with SR-AI was detected using Duolink-PLA analysis by combining anti-VWF and anti-SR-AI antibodies. (K-L) THP1-derived macrophages were incubated in the absence or presence of pd-VWF (10 mg/mL). Association with SR-AI was detected using Duolink-PLA analysis by combining anti-VWF and anti-SR-AI antibodies. Panels G to L: objective 63x; scale bars: 10 mm.

732

haematologica | 2018; 103(4)

SR-AI contributes to VWF clearance

lution analysis revealed that VWF co-localized with SR-AI on these cells (Figure 3E). Moreover, we also observed VWF present in EEA-1-containing endosomes (Figure 3F), indicating that SR-AI binding is followed by uptake and delivery to the lysosomal degradation pathway. DuolinkPLA analysis was performed to further confirm that VWF associates with SR-AI. This analysis revealed numerous red spots when VWF was incubated with SR-AI-expressing HEK293 cells, whereas such spots were absent upon incubation with non-transfected cells or when VWF was omitted from the incubation (Figure 3G-J). A similar colocalization between VWF and SR-AI was observed when testing the binding of VWF to THP1-macrophages (Figure 3K-L). Furthermore, binding to THP1-derived

macrophages was reduced to near background levels when VWF was incubated in the presence of anti-VWF antibodies Mab723 and Mab540 (Figure 4A-C). Finally, we analyzed binding of VWF to primary bone marrowderived macrophages obtained from wt- and SR-AI-deficient mice. We observed strongly reduced VWF staining to macrophages derived from SR-AI-deficient mice compared to macrophages derived from control mice (Figure 4D-F). Duolink-PLA analysis revealed the formation of red spots, representing complexes between human VWF and murine SR-AI on wild-type macrophages but not on macrophages from SR-AI-deficient mice (Figure 4G,H). From these observations it is conceivable that SR-AI acts as a macrophage-receptor for VWF.

Figure 4. van Willebrand factor binding to macrophages is reduced by anti-VWF antibodies or SR-AI deficiency. (A, B) Representative images of THP1-derived macrophages incubated with pd-VWF (10 mg/mL) in the absence or presence or monoclonal anti-VWF antibodies Mab723 & Mab540 (167 mg/mL). (D,E) Representative images of murine CD115+ bone marrow-derived macrophages obtained from wt- or SR-AI-deficient mice that were incubated with pd-VWF (10 mg/mL). Cell-bound VWF was probed using polyclonal anti-VWF antibodies. (C and F) Quantification of immune fluorescent signals for VWF. Data represent meanÂąSEM [(n= 64-120 cells (C); n= 62-118 cells (F)]. Statistical analysis involved a one-way analysis of variance with the Tukey multiple comparison test (C) or a two-tailed MannWhitney test (F). (G, H) wt and SR-AI-deficient murine-macrophages (CD115+) were incubated with human pd-VWF (10 mg/mL). Association between murine-SR-AI and human-VWF was detected using Duolink-PLA analysis by combining monoclonal anti-human-VWF and goat anti-murine-SR-AI antibodies. All microscopy figures: objective 40x, scale bars 10 mm.

haematologica | 2018; 103(4)

733

N. Wohmer et al.

von Willebrand factor propeptide to antigen ratio is lower in SR-AI-deficient mice than in wild-type and macLRP1-deficient mice To assess the physiological relevance of SR-AI in regulating VWF clearance, we opted to express human VWF in wild-type, macLRP1- and SR-AI-deficient mice via hydrodynamic gene transfer, and to determine the ratio between VWFpp and VWF:Ag, a measure of VWF clearance. As reported previously,9 VWFpp/VWF:Ag ratios were slightly, but significantly reduced in macLRP1-deficient mice compared to wt-mice (1.3±0.1 versus 1.1±0.1 for wt- and macLRP1-deficient mice, respectively; n=8-9; P=0.0114) (Figure 5). This confirms that LRP1 contributes to a modest extent to the clearance of VWF. Interestingly, VWFpp/VWF:Ag levels were even further reduced in SRAI-deficient mice: 0.6±0.2 versus 1.3±0.3 (n=9-14; P<0.0001) (Figure 5). VWF is apparently cleared less rapidly in SR-AI-deficient mice than in macLRP1-deficient mice. This suggests that SR-AI plays a more dominant role than that of LRP1 in basal VWF clearance.

mutants interact with SR-AI at the macrophage cell surface (Figure 6B-D). Quantitative analysis revealed that fluorescence was significantly increased for both mutants compared to wt-VWF. VWF surface coverage was 3.2±0.9% for wt-VWF, 8.7±2.4% for VWF/p.R1205H and 11.6±2.1% for VWF/p.S2179F (Figure 6E). These data indicate enhanced binding of clearance mutants VWF/p.R1205H and VWF/p.S2179F to SR-AI.

Increased clearance of mutants VWF/p.R1205H and VWF/p.S2179F is partially corrected in SR-AI-deficient mice We next investigated to what extent clearance of the mutants VWF/p.R1205H and VWF/p.S2179F is SR-AIdependent. Hydrodynamic gene transfer was applied to

Clearance mutants VWF/p.R1205H and VWF/p.S2179F show enhanced binding to SR-AI Given the involvement of the D’D3 and D4 domains in SR-AI binding (Figure 2), it was of interest to investigate whether clearance mutations in these domains affect the interaction with SR-AI. We first analyzed binding of VWF/p.R1205H (the Vicenza variant with a mutation in the D3 domain) and VWF/p.S2179F (with a mutation in the D4 domain) to sSR-AI in an immunosorbent assay. Whereas the interactions of type 2B mutant VWF/p.V1316M and wt-VWF with sSR-AI were similar (half-maximal binding at 3.1±0.7 and 3.6±0.9 mg/mL, respectively), both mutants VWF/p.R1205H and VWF/p.S2179F proved more efficient in interacting with SR-AI (1.7±0.3 and 2.3±0.4 mg/mL; P=0.0124) (Figure 6A). We then visualized binding of both mutants to SR-AI expressed on THP1-macrophages using Duolink-PLA analysis. Bright red spots were observed for mutants VWF/p.R1205H and VWF/p.S2179F, indicating that both

Figure 5. Deficiency of SR-AI results in decreased VWFpp/VWF:Ag ratios. Human-wt-VWF was expressed in macLRP1+- and SR-AI-expressing mice and in macLRP1-deficient and SR-AI-deficient mice following hydrodynamic gene transfer. Four days after injection, plasma samples were prepared for the analysis of VWFpp and VWF:Ag. Assays for VWFpp and VWF:Ag quantify only human VWF expressed via hydrodynamic gene transfer, and do not cross-react with endogenous murine VWF. VWFpp/VWF:Ag ratios for each individual mice included in the study are plotted. Data from macLRP1-mice and SR-AI-mice were compared in a pairwise manner using a two-tailed Student t-test.

734

Figure 6. Enhanced binding of von Willebrand factor mutants p.R1205H and p.S2179F to SR-AI. (A) Wells coated with sSR-AI were incubated with various concentrations of non-purified recombinant VWF (0-5 mg/mL). Closed circles: wtVWF; open squares: p.V1316M; open diamonds: p.S2179F; closed diamonds: p.R1205H. Open circles represent binding of wt-VWF to bovine serum albumincoated wells. Mutants gave similar background signals. Bound VWF was probed with peroxidase-labeled polyclonal anti-VWF antibodies. All mutants reacted similarly with these polyclonal antibodies. Data represent mean±SD (n=3). (B-E) THP1-derived macrophages were incubated in the absence or presence of nonpurified recombinant wt-VWF (B) or mutants VWF/p.R1205H (C) or VWF/p.S2179F (D). Association with SR-AI was detected using Duolink-PLA analysis by combining anti-VWF and anti-SR-AI antibodies (Objective 63x; scale bars 10 mm). (E) Quantification of fluorescent signals. Data represent mean±SD (n=5 microscopic fields; 2-5 cells/field). Statistical analysis involved one-way analysis of variance followed by the Tukey multiple comparison test.

haematologica | 2018; 103(4)

SR-AI contributes to VWF clearance

express human VWF, VWF/p.R1205H and VWF/p.S2179F in control mice and SR-AI-deficient mice, and the VWFpp/VWF:Ag ratio was determined. VWFpp/VWF:Ag ratios were markedly increased for both mutants in control mice [2.9±0.2 (n=8) and 4.4±0.5 (n=7), for VWF/p.R1205H and VWF/p.S2179F, respectively; P<0.001 compared to wt-VWF] (Figure 7), confirming that both mutations induce increased clearance of VWF. When expressed in SR-AI-deficient mice, a significant reduction in VWFpp/VWF:Ag ratio was found for both mutants: 2.0±0.4 and 2.3±0.5, for VWF/p.R1205H and VWF/p.S2170F, respectively (P<0.0001) (Figure 7). These data point to mutants VWF/p.R1205H and VWF/p.S2179F being cleared less rapidly in SR-AI-deficient mice than in wt-mice, suggesting that SR-AI contributes to the clearance of these mutants.

Discussion Sinusoidal endothelial cells and macrophages have been proposed to mediate clearance of VWF, with macrophages being particularly dominant.14,17,26,27 The molecular basis by which macrophages interact with VWF is, however, unclear. Previously, it was reported that VWF is a ligand for the scavenger-receptor LRP1, which is abundantly present on macrophages.11,12,26 Nonetheless, VWF only interacts with LRP1 when exposed to increased shear stress, or is otherwise in its active conformation, e.g. following incubation with ristocetin or botrocetin, or when harboring a VWD-type 2B mutation. In addition, modulation of the glycan structures in the A2 domain also favors spontaneous binding to LRP1.10,11 By using a Duolink-PLA strategy, we could indeed confirm that VWF is unable to associate with LRP1 under static conditions (Figure 1). In contrast, when macrophages were analyzed via classical immune-fluorescent staining, the presence of VWF on THP1-derived macrophages could readily be detected (Figure 1). These data are in agreement with previous

Figure 7. SR-AI-deficiency is associated with decreased VWFpp/VWF:Ag ratios for mutants p.R1205H and p.S2179F. Mutants VWF/p.R1205H and VWF/p.S2179F were expressed in SR-AI-expressing control mice and in SR-AIdeficient mice following hydrodynamic gene transfer. Four days after injection, plasma samples were prepared for the analysis of VWFpp and VWF:Ag. VWFpp/VWF:Ag ratios for each individual mice included in the study are plotted. Data for wt-VWF are similar to those presented in Figure 5. Statistical analysis involved one-way ANOVA followed by the Tukey multiple comparison test.

haematologica | 2018; 103(4)

observations from our laboratory, in which we have observed VWF staining on primary monocyte-derived macrophages.17,28 It should be noted that Castro-Nunez and colleagues were unable to detect VWF binding to macrophages under static conditions.26 The lack of VWF detection may be related to the conditions in which the macrophages were cultured. Alternatively, their method requires perhaps higher VWF concentrations for binding to become detectable. Macrophages express a number of candidate receptors that can be involved in VWF binding, including Siglec-5 and the asialoglycoprotein receptor. Nevertheless, their relative contribution to VWF clearance remains unclear, and is probably modest at best under regular physiological conditions. In this study, we focused on SR-AI (also known as SCARA1 or CD204) as a novel candidate receptor that is specifically expressed in macrophages and dendritic cells. The interest in this receptor mainly originates from its high structural homology with SCARA5, an epithelial receptor that has been identified in genomewide association studies to be associated with VWF plasma levels.24 SR-AI and SCARA5 are both single transmembrane scavenger receptors that interact with their ligands via an ectodomain that consists of a collagenous domain and three scavenger receptor cysteine-rich domains.29 The potential of SR-AI to interact with VWF became evident in solid-phase binding experiments, in which saturable and dose-dependent binding was observed (Figure 2). It was remarkable to note that half-maximal binding was obtained at 3.5 mg/mL VWF, corresponding to 14 nM. Although our experimental approach in combination with the multimeric structure does not allow the calculation of a true affinity constant, this value suggests that VWF is able to interact with SR-AI with relatively high affinity. This value is considerably lower than the apparent affinity constants we recently identified for the interactions of SRAI with factor X and pentraxin-2 (0.7 mM and 0.2 mM, respectively), suggesting that VWF binds to SR-AI more efficiently than factor X and pentraxin-2. It is worth mentioning that, in direct competition experiments, VWF was unable to displace factor X from SR-AI (data not shown), indicating that both ligands bind to distinct interactive sites on SR-AI. This possibility fits with the notion that factor X binding is cation-independent (data not shown), whereas VWF binding is fully cation-dependent (Figure 2). Possibly, VWF binding involves similar regions within SRAI that also mediate the cation-dependent cell adhesion.25 With regard to VWF, the interaction with SR-AI appears to involve multiple regions within the VWF molecule, including at least the D’D3-region, the A1 domain and the D4 domain (Figure 2). While testing a library of >20 monoclonal anti-VWF antibodies, we identified two antibodies that were able to interfere with the interaction between VWF and SR-AI (Figure 2). One is directed against the A1 domain (MAb723) and the other against the D4 domain (MAb540), which is in agreement with the involvement of multiple VWF regions contributing to the interaction with SR-AI. It is important to mention here that preliminary studies in our laboratory revealed that the D’D3-region and the D4 domain also contain binding sites for LRP1 (data not shown). Thus, there seems to be an overlap in domains involved in binding to SR-AI and LRP1. Nevertheless, the interaction of VWF with SR-AI is most likely distinct from its interaction with LRP1. First, antibodies MAb723 and Mab540 do not affect binding of 735

N. Wohmer et al.

VWF to LRP1 (data not shown), indicating that binding sites in these domains do not overlap. Second, introduction of the VWD-type 2B mutation leaves binding of VWF to SRAI unaffected, as does the addition of ristocetin. Hence, VWF does not need to be in its active conformation to interact with SR-AI, whereas it does need to be for binding to LRP1. As for the role of glycans present on the VWF molecule in the interaction with SR-AI, this could be the subject of further studies. However, neither the A1 domain nor the D4 domain contains glycan structures, suggesting that the interaction with SR-AI is mainly glycan-independent. This does not exclude the possibility that glycans elsewhere in the protein could modulate this interaction, akin to what has previously been reported for the binding of the A1 domain to LRP1.11 Apart from binding to purified recombinant soluble SRAI, we also observed a specific binding of VWF to cellular SR-AI (Figure 3). First, we used SR-AI-transfected HEK293 cells as a model system, and both classical immunofluorescent staining and the Duolink-PLA revealed selective binding of VWF to SR-AI. Second, the association of VWF with THP1-derived macrophages (as depicted in Figure 1) is at least in part mediated by SR-AI, as illustrated by the Duolink-PLA approach (Figure 3). Moreover, immunostaining for VWF on THP1-derived macrophages was strongly reduced in the presence of antibodies Mab723 and Mab540, which interfere with SR-AI binding (Figure 4). Finally, when binding of VWF to primary bone marrow-derived murine macrophages was tested, binding was reduced to near background levels for SR-AI-deficient macrophages compared to wt-macrophages (Figure 4). Being able to interact with SR-AI expressed on the cell surface supports a role of SR-AI as a clearance receptor for VWF. We analyzed this possibility by measuring VWFpp/VWF:Ag ratios of human VWF expressed in wtand SR-AI-deficient mice. There were several reasons for choosing this approach over measuring classical VWF survival. First, in a recent study we compared the clearance of two mutants in parallel via protein survival and via measuring VWFpp/VWF:Ag ratios.9 This analysis revealed that the VWFpp/VWF:Ag approach was clearly more sensitive than the classic protein survival experiments in detecting differences in VWF clearance, due to a markedly smaller error margin between mice. Second, this smaller error margin also favors the use of fewer mice in this type of experiments. For a classical clearance experiment, approximately 15 mice are included per molecule to be tested, whereas with the VWFpp/VWF:Ag approach fewer than ten mice per molecule are needed. Thus, from an animal ethical perspective this latter approach is to be preferred. Finally, expression in hepatocytes allows more homogenous post-translational processing compared to the production of proteins in distinct stable cells lines. Indeed, the lectin binding profile of hepatic VWF is similar to that of endothelial VWF.30 One might fear interference of clearance of hepatic VWF by endogenous endothelial-derived VWF. However, VWF clearance is similar in wt- and VWFdeficient mice,3 and even at VWF levels of 1500%, clearance remains unsaturated. Compared to VWFpp/VWF:Ag ratios obtained for wtmice (ratio=1.3) and LRP1-deficient mice (ratio=1.1), these ratios were strongly reduced in SR-AI-deficient mice (ratio=0.6). This not only points to SR-AI being a clearance

736

receptor for VWF, but also to SR-AI being more dominant in VWF clearance than LRP1. We anticipated that endogenous VWF levels would be increased in SR-AI-deficient mice compared to those in wt-mice. However, analysis of VWF levels did not reveal a statistically significant difference between SR-AI-deficient and wt-mice. We believe that the lack of difference is due to the fact that the SR-AIdeficient and wt-mice were not true littermates, which complicates a direct comparison. Indeed, even among mice with a similar genetic background, the variation in VWF levels is substantial (e.g. 0.3-1.9 U/mL),12,31 which may explain the lack of difference between SR-AI-deficient and wt-mice. Of note, we observed that murine VWF efficiently interacts with murine SR-AI, indicating that the lack of difference is not because murine VWF is unable to interact with this receptor. An intriguing aspect of VWF receptor interactions is how these are modulated by mutations in VWF, in particular those mutations that are associated with increased clearance. We previously showed that VWD-type 2B mutations promote spontaneous binding to LRP1, explaining the increased clearance of these mutants. Here we examined two clearance mutants: VWF/p.R1205H and VWF/p.S2179F.1,3,5,8 Both mutants are known to be associated with increased VWFpp/VWF:Ag ratios; in humans for VWF/p.S2179F and in human and mice for VWF/p.R1205H.5,7,8 Here we show that both mutants display increased binding to SR-AI, both to purified SR-AI and SR-AI expressed on THP1-cells (Figure 6). Increased binding of the VWF/p.R1205H is in agreement with data reported by O’Donnell et al., who also observed increased binding of this mutant to macrophages. Increased binding to SR-AI may suggest that SR-AI contributes to the accelerated removal of these mutants from the circulation. Indeed, VWFpp/VWF:Ag ratios for VWF/p.R1205H and VWF/p.S2179F were significantly reduced in SR-AI-deficient mice compared to wt-mice (Figure 7). However, even in the SR-AI-deficient mice, these ratios were substantially higher compared to wt-VWF, indicating that SR-AI is not the only receptor that mediates increased clearance of these mutants. We considered the option that LRP1 could play a role in the enhanced clearance of these mutants, and preliminary experiments revealed that both mutants did indeed display enhanced binding to LRP1 (data not shown). Apparently, enhanced receptor binding due to such clearance mutations is not always restricted to a single receptor, but may involve several receptors simultaneously, thereby multiplying the clearance rate of the mutant proteins. In summary, we identify SR-AI as a macrophage-specific receptor for VWF, and this receptor may contribute to the increased clearance of certain VWF clearance mutants. Acknowledgments This study was supported by grants from the Agence Nationale de la Recherche (ANR-13-BSV1-0014; PJL), and the Fondation pour la Recherche Médicale (FRM-SPF20130526717; NW & PJL). We would like to thank Pascal Roux & Dr. Audrey Salles (Pasteur Institute, Paris, France) for their help in accessing the confocal microscope facility and interpretation of optical microscopy data and Emilie Bouvier & Alexandre Diet (Center for Breeding & Distribution of Transgenic Animals, Orléans, France) for their technical assistance.

haematologica | 2018; 103(4)

SR-AI contributes to VWF clearance

References 1. Casonato A, Pontara E, Sartorello F, et al. Reduced von Willebrand factor survival in type Vicenza von Willebrand disease. Blood. 2002;99(1):180-184. 2. Haberichter SL, Castaman G, Budde U, et al. Identification of type 1 von Willebrand disease patients with reduced von Willebrand factor survival by assay of the VWF propeptide in the European study: molecular and clinical markers for the diagnosis and management of type 1 VWD (MCMDM1VWD). Blood. 2008;111(10):4979-4985. 3. Lenting PJ, Westein E, Terraube V, et al. An experimental model to study the in vivo survival of von Willebrand factor. Basic aspects and application to the R1205H mutation. J Biol Chem. 2004;279(13):12102-12109. 4. van Schooten CJ, Tjernberg P, Westein E, et al. Cysteine-mutations in von Willebrand factor associated with increased clearance. J Thromb Haemost. 2005;3(10):2228-2237. 5. Haberichter SL, Balistreri M, Christopherson P, et al. Assay of the von Willebrand factor (VWF) propeptide to identify patients with type 1 von Willebrand disease with decreased VWF survival. Blood. 2006;108(10):3344-3351. 6. Sztukowska M, Gallinaro L, Cattini MG, et al. Von Willebrand factor propeptide makes it easy to identify the shorter von Willebrand factor survival in patients with type 1 and type Vicenza von Willebrand disease. Br J Haematol. 2008;143(1):107114. 7. Eikenboom J, Federici AB, Dirven RJ, et al. VWF propeptide and ratios between VWF, VWF propeptide, and FVIII in the characterization of type 1 von Willebrand disease. Blood. 2013;121(12):2336-2339. 8. Pruss CM, Golder M, Bryant A, et al. Pathologic mechanisms of type 1 VWD mutations R1205H and Y1584C through in vitro and in vivo mouse models. Blood. 2011;117(16):4358-4366. 9. Wohner N, Legendre P, Casari C, Christophe OD, Lenting PJ, Denis CV. Shear stress-independent binding of von Willebrand factortype 2B mutants p.R1306Q & p.V1316M to LRP1 explains their increased clearance. J Thromb Haemost. 2015;13(5):815-820.

haematologica | 2018; 103(4)

10. O'Sullivan JM, Aguila S, McRae E, et al. Nlinked glycan truncation causes enhanced clearance of plasma-derived von Willebrand factor. J Thromb Haemost. 2016;14(12): 2446-2457. 11. Chion A, O'Sullivan JM, Drakeford C, et al. N-linked glycans within the A2 domain of von Willebrand factor modulate macrophage-mediated clearance. Blood. 2016;128(15):1959-1968. 12. Rastegarlari G, Pegon JN, Casari C, et al. Macrophage LRP1 contributes to the clearance of von Willebrand factor. Blood. 2012;119(9):2126-2134. 13. Grewal PK, Uchiyama S, Ditto D, et al. The Ashwell receptor mitigates the lethal coagulopathy of sepsis. Nat Med. 2008;14(6):648655. 14. Rydz N, Swystun LL, Notley C, et al. The Ctype lectin receptor CLEC4M binds, internalizes and clears von Willebrand factor and contributes to the variation in plasma von Willebrand factor levels. Blood. 2013;121(26):5228-5237. 15. Pegon JN, Kurdi M, Casari C, et al. Factor VIII and von Willebrand factor are ligands for the carbohydrate-receptor Siglec-5. Haematologica. 2012;97(12):1855-1863. 16. Casari C, Lenting PJ, Wohner N, Christophe OD, Denis CV. Clearance of von Willebrand factor. J Thromb Haemost. 2013;11 Suppl 1:202-211. 17. van Schooten CJ, Shahbazi S, Groot E, et al. Macrophages contribute to the cellular uptake of von Willebrand factor and factor VIII in vivo. Blood. 2008;112(5):1704-1712. 18. Muczynski V, Ayme G, Regnault V, et al. Complex formation with pentraxin-2 regulates factor X plasma levels and macrophage interactions. Blood. 2017;129(17):2443-2454. 19. Muczynski V, Bazaa A, Loubiere C, et al. Macrophage receptor SR-AI is crucial to maintain normal plasma levels of coagulation factor X. Blood. 2016;127(6):778-786. 20. Breslin WL, Strohacker K, Carpenter KC, Haviland DL, McFarlin BK. Mouse blood monocytes: standardizing their identification and analysis using CD115. J Immunol Methods. 2013;390(1-2):1-8. 21. Marx I, Christophe OD, Lenting PJ, et al. Altered thrombus formation in von Willebrand factor-deficient mice expressing von Willebrand factor variants with defec-

22.

23.

24.

25.

26.

27.

28.

29. 30.

31.

tive binding to collagen or GPIIbIIIa. Blood. 2008;112(3):603-609. Marx I, Lenting PJ, Adler T, Pendu R, Christophe OD, Denis CV. Correction of bleeding symptoms in von Willebrand factor-deficient mice by liver-expressed von Willebrand factor mutants. Arterioscler Thromb Vasc Biol. 2008;28(3):419-424. Rayes J, Hollestelle MJ, Legendre P, et al. Mutation and ADAMTS13-dependent modulation of disease severity in a mouse model for von Willebrand disease type 2B. Blood. 2010;115(23):4870-4877. Smith NL, Chen MH, Dehghan A, et al. Novel associations of multiple genetic loci with plasma levels of factor VII, factor VIII, and von Willebrand factor: the CHARGE (Cohorts for Heart and Aging Research in Genome Epidemiology) Consortium. Circulation. 2010;121(12):1382-1392. Santiago-Garcia J, Kodama T, Pitas RE. The class A scavenger receptor binds to proteoglycans and mediates adhesion of macrophages to the extracellular matrix. J Biol Chem. 2003;278(9):6942-6946. Castro-Nunez L, Dienava-Verdoold I, Herczenik E, Mertens K, Meijer AB. Shear stress is required for the endocytic uptake of the factor VIII-von Willebrand factor complex by macrophages. J Thromb Haemost. 2012;10(9):1929-1937. van der Flier A, Liu Z, Tan S, et al. FcRn rescues recombinant factor VIII Fc fusion protein from a VWF independent FVIII clearance pathway in mouse hepatocytes. PLoS One. 2015;10(4):e0124930. Casari C, Du V, Wu YP, et al. Accelerated uptake of VWF/platelet complexes in macrophages contributes to VWD type 2Bassociated thrombocytopenia. Blood. 2013;122(16):2893-2902. Zani IA, Stephen SL, Mughal NA, et al. Scavenger receptor structure and function in health and disease. Cells. 2015;4(2):178-201. Badirou I, Kurdi M, Legendre P, et al. In vivo analysis of the role of O-glycosylations of von Willebrand factor. PLoS One. 2012;7(5):e37508. Lemmerhirt HL, Broman KW, Shavit JA, Ginsburg D. Genetic regulation of plasma von Willebrand factor levels: quantitative trait loci analysis in a mouse model. J Thromb Haemost. 2007;5(2):329-335.

737

ARTICLE

Coagulation & its Disorders

Ferrata Storti Foundation

Haematologica 2018 Volume 103(4):738-745

A three-year prospective study of the presentation and clinical outcomes of major bleeding episodes associated with oral anticoagulant use in the UK (ORANGE study)

Laura Green,*1,2,3 Joachim Tan,*1 Joan K Morris,1 Raza Alikhan,4 Nicola Curry,5,6 Tamara Everington,7,8 Rhona Maclean,9 Khalid Saja,10 Simon Stanworth,5,6,11 Campbell Tait12 and Peter MacCallum1,2

Barts and the London School of Medicine and Dentistry, Queen Mary University of London; 2Barts Health NHS Trust, London; 3NHS Blood and Transplant, Colindale; University Hospital of Wales, Cardiff and Vale University Health Board; 5Department of Haematology, Oxford University Hospitals NHS Foundation Trust, Oxford 6Radcliffe Department of Medicine, University of Oxford, and Oxford BRC Haematology Theme; 7 Hampshire Hospitals NHS Foundation Trust; 8Salisbury NHS Foundation Trust; 9 Sheffield Teaching Hospitals NHS Foundation Trust; 10Barking, Havering and Redbridge University Hospitals NHS Trust; 11Transfusion Medicine, NHS Blood and Transplant, Oxford and 12Glasgow Royal Infirmary, NHS Greater Glasgow and Clyde, UK 1 4

*Joint first authors

ABSTRACT

Correspondence: laura.green@bartshealth.nhs.uk

Received: October 6, 2017. Accepted: January 22, 2018. Pre-published: January 25, 2018. doi:10.3324/haematol.2017.182220 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/x738 ÂŠ2018 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.

738

he outcomes of patients developing major bleeding while on oral anticoagulants remain largely unquantified. The objectives of this study were to: (i) describe the burden of major hemorrhage associated with all available oral anticoagulants in terms of proportion of bleeds which are intracranial hemorrhages, in-hospital mortality and duration of hospitalization following major bleeding; (ii) identify risk factors for mortality; and (iii) compare the characteristics of major hemorrhage between cases treated with warfarin and direct oral anticoagulants for the subgroups of patients with atrial fibrillation or venous thromboembolism. This was a multicenter, 3-year prospective cohort study of patients aged â&#x2030;Ľ18 years on oral anticoagulants who developed major hemorrhage leading to hospitalization. The patients were followed up for 30 days or until discharge or death, whichever occurred first. In total 2,192 patients (47% female, 81% on warfarin, median age 80 years) were reported between October 2013 and August 2016 from 32 hospitals in the UK. Bleeding sites were intracranial (44%), gastrointestinal (33%), and other (24%). The in-hospital mortality was 21% (95% CI: 19%-23%) overall, and 33% (95% CI: 30%-36%) for patients with intracranial hemorrhage. Intracranial hemorrhage, advanced age, spontaneous bleeding, liver failure and cancer were risk factors for death. Compared to warfarin-treated patients, patients treated with direct oral anticoagulants were older and had lower odds of subdural/epidural, subarachnoid and intracerebral bleeding. The mortality rate due to major bleeding was not different between patients being treated with warfarin or direct oral anticoagulants. Major bleeding while on oral anticoagulant therapy leads to considerable hospital stays and short-term mortality.

Introduction Oral anticoagulants (OAC) are highly effective for stroke prevention in patients with atrial fibrillation,1,2 for the treatment and prevention of venous thromboembolism,3 and for the prevention of thrombosis related to mechanical heart valves.4,5 It is estimated that OAC therapy is required for 1.25 million people per year in the UK with approximately 70% being for those with atrial fibrillation.6 This requirement is likely to continue to rise in an aging population, given that the prevalence haematologica | 2018; 103(4)

Major bleeding associated with oral anticoagulants

of atrial fibrillation7 and the incidence of venous thromboembolism8 both increase with age. The foremost complication of OAC therapy is the development of major bleeding. In the phase III randomized clinical trials which compared warfarin and direct OAC – namely dabigatran etexilate, rivaroxaban, apixaban and edoxaban (DOAC hereafter) – in patients with atrial fibrillation and venous thromboembolism, this risk was reported to be 1-3% per year.9 In clinical practice some studies have reported similar risks of major bleeding,10-12 while others have found that it may be considerably higher.13,14 When DOAC were first introduced into clinical practice there was concern among clinicians that the lack of specific antidotes could be detrimental to patients’ outcomes in the event of major bleeding. Recently, ‘post-approval’ observational studies have reported on the safety profile of DOAC; however, these studies have primarily focused

on patients with atrial fibrillation, using patients’ clinical data obtained from national registries/databases which were designed for different purposes.15 Moreover, these studies lacked detail on the acute management of the bleeds. The burden (with respect to in-hospital mortality, morbidity and duration of hospitalization) of major bleeding associated with all OAC, for any clinical indications, remains largely unknown. This dearth of knowledge is true even in the case of warfarin, which has been the mainstay of OAC therapy for more than 60 years. Moreover, the widespread and increasing use of OAC, particularly in the frail elderly, underscores the urgency of comparative studies to assess the burden of major bleeding events associated with warfarin and DOAC. Such information should be incorporated into the clinical assessment and counseling of patients prescribed OAC, as well as the optimization of strategies for the management of OAC-associated major bleeding events.

Table 1. Baseline characteristics by type of oral anticoagulanta.

Total patients Female Age (years), median[IQR] <65 65-74 75-84 ≥85 Indicationsc Atrial fibrillation Venous thromboembolism Mechanical heart valve Other Prescribed anti-thrombotics Single antiplatelet therapy Dual antiplatelet therapy LMWH/UFH Co-morbidities (type) Congestive heart failure Hypertension Previous stroke/TIA Liver failured Cancer Diabetes Peripheral vascular disease Ischemic heart disease Chronic kidney diseased Alcohol dependence Dementia Recurrent falls Previous major bleeding Intracranial hemorrhagee Gastrointestinal bleed Other bleeds Bleed history unknown

Total number (%)

Warfarin

Dabigatran

Rivaroxaban

Apixaban

2,192 (100) 1,028 (47) 80 [72-86] 284 (13) 408 (19) 863 (39) 637 (29)

1,771 (81) 814 (46) 79 [71-85] 246 (14) 348 (20) 714 (40) 463 (26)

46 (2) 24 (52) 85 [79-88] 2 (4) 6 (13) 12 (26) 26 (57)

283 (13) 153 (54) 82 [74-88] 29 (10) 42 (15) 96 (34) 116 (41)

89 (4) 35 (39) 81 [76-86] 7 (8) 12 (13) 39 (44) 31 (35)

1,575 (72) 456 (21) 212 (10) 184 (8)

1,265 (71) 352 (20) 211 (12) 143 (8)

38 (83) 3 (7) --11 (24)

193 (68) 89 (31) 1 (<1) 18 (6)

76 (85) 11(12) --12 (13)

193 (9) 27 (1) 53 (2)

144 (8) 23 (1) 47 (3)

3 (7) --1 (2)

35 (12) 1 (<1) 5 (2)

11 (13) 3 (3) ---

342 (16) 1,181 (54) 432 (20) 35 (2) 334 (15) 494 (23) 66 (3) 570 (26) 339 (15) 48 (2) 144 (7) 128 (6)

278 (16) 949 (54) 334 (19) 28 (2) 258 (15) 402 (23) 49 (3) 456 (26) 282 (16) 39 (2) 97 (5) 94 (5)

4 (9) 28 (61) 17 (37) 1 (2) 6 (13) 9 (20) 2 (4) 15 (33) 2 (4) 1 (2) 2 (4) 4 (9)

46 (16) 154 (54) 56 (20) 6 (2) 55 (19) 62 (22) 13 (5) 69 (24) 49 (17) 7 (2) 35 (12) 24 (8)

14 (16) 50 (56) 23 (26) --14 (16) 21 (24) 2 (2) 29 (33) 6 (7) 1 (1) 8 (9) 6 (7)

81 (4) 87 (4) 38 (2) 626 (29)

64 (4) 71 (4) 29 (2) 513 (29)

2 (4) 2 (4) 3 (7) 11 (24)

11 (4) 10 (4) 6 (2) 73 (26)

4 (4) 4 (4) --28 (31)

LMWH: low molecular weight heparin; UFH: unfractionated heparin; TIA: transient ischemic attack; aEdoxaban patients not shown separately (n=3). c10% of patients had more than one indication. dSee appendix for definitions of liver and renal failure. eIncludes two with additional non-intracranial previous bleeds

haematologica | 2018; 103(4)

739

L. Green et al.

The main objectives of this study were to: (i) describe the burden of major hemorrhage associated with all available OAC in terms of proportion of bleeds which are intracranial, in-hospital case-fatality and morbidity, and duration of hospitalization; (ii) identify risk factors for fatality; (iii) compare characteristics of major hemorrhage between patients treated with warfarin and DOAC for the subgroups anticoagulated for atrial fibrillation or venous thromboembolism, the clinical conditions for which DOAC are indicated.

Methods Study design The ORANGE (ORal ANticoagulant aGEnt-associated bleeding events reporting system) study was a prospective cohort study that collected information from multiple UK hospitals on the presentation and clinical outcomes of patients who were admitted for a major bleeding episode while on OAC therapy. Ethics approval was obtained for the study from the National Health Service, Health Research Authority (East of England-

Table 2. Bleeding characteristics and outcomes.

Number (%) Total patients All intracranial hemorrhage Subdural/epidural Intracerebrala Subarachnoidb

2,192 (100) 963 (44) 386 (18) 474 (22) 103 (5)

All gastrointestinal bleeds UpperC Lower Other bleeds Visceral Genitourinary Musculoskeletal Miscellaneous (e.g. intra-ocular, puncture and surgical sites)

712 (32) 443 (20) 269 (12) 517 (24) 131 (6) 87 (4) 224 (10) 75 (4)

Provocation of bleeding Spontaneous Trauma (excluding fall) Surgery or procedure Fall Unclassifiedd

1,434 (65) 200 (9) 105 (5) 432 (20) 21 (1)

Outcomes up to 30 days Died in hospital Discharged from hospital Still inpatient at 30 days Lost to follow-up Transferred to other hospital Not submitted

Case fatality by groups Overall

Days to event, median[IQR] 446 (20) 1,413 (64) 273 (12)

3 [1-8] 7 [3-13]

47 (2) 13 (1)

2 [1-8]

Followed-up, N

Died (%)

2,132

446 (21)

Age, years <65 65-74 75-84 85+ Type of oral anticoagulant Warfarin Direct oral anticoagulant

274 400 840 618

25 (9) 62 (16) 197 (23) 162 (26)

1,719 413

358 (21) 88 (21)

Site of bleed All intracranial All gastrointestinal Other

923 704 505

302 33) 91 (13) 53 11)

Includes 32 with additional bleeds in other sites. bIncludes 17 with additional subdural bleed. cIncludes 18 with additional lower gastrointestinal bleed. dReported as "unknown" or "not available".

740

haematologica | 2018; 103(4)

Major bleeding associated with oral anticoagulants

Cambridge South Research Ethics Committee, reference: 12/EE/0431). Data on major bleeding events were submitted by multiple hospitals across England, Scotland, Wales and Northern Ireland between October 1, 2013 and August 31, 2016. Patients underwent the normal course of treatment as directed by their clinicians and hospital protocols; at no point was their care altered for the purpose of this study.

Definition of major bleeding The definition of major bleeding adopted was an augmented version of the International Society on Thrombosis and Haemostasis criteria.16 It was defined as bleeding requiring hospitalization and at least one of the following: (i) resulting in death; (ii) transfusion of ≥2 units of red blood cells or a drop in hemoglobin of ≥20 g/L; (iii) symptomatic bleeding in a critical area or organ, such as intracranial, intraspinal, intra-ocular, retroperitoneal, intra-articular or pericardial, or intramuscular with compartment syndrome; (iv) transfusion of fresh-frozen plasma; (v) administration of prothrombin complex concentrate, recombinant activated factor VII, factor VIII inhibitor bypassing activity or fibrinogen concentrate. The rationale for appending (iv) and (v) was to ensure that the routes for case identification were as comprehensive as possible.

Data collection Any patient of 18 years or over on OAC therapy at the time when they developed major bleeding was eligible for the study.

Cases were reported consecutively and identified by clinical and research staff in participating hospitals from the emergency department, transfusion laboratory, pharmacy (if they stored hemostatic agents) and hematology doctors who were called to give medical advice on the management of these patients. The study collected information on patients’ baseline characteristics; type of OAC and indication(s), as well as co-morbidities and clinical outcomes at 30 days, death, or discharge, whichever occurred first. Further information on details of data collection/verification, and justification for sample size are given in the Online Supplementary Material.

Statistical analysis The first bleeding episode of each individual patient was analyzed, with all DOAC grouped together for comparisons. Associations between categorical variables were examined with a chi-squared or Fisher exact test and Mann-Whitney U tests were used for continuous variables (all two-sided at a 5% level of significance). Multinomial logit models were used to estimate the conditional odds ratios for DOAC versus warfarin on the site of bleeding (versus a reference site), adjusting for age and bleeding provocation. Risk factors for in-hospital 30-day mortality were investigated using mixed-effects multivariable logistic regression to take into account intra-hospital correlation, controlling for age, co-morbidities, bleeding provocation and indications for OAC therapy. All analyses were performed using STATA version 14 (StataCorp LP, USA).

Table 3. Comparison of warfarin and direct oral anticoagulants, for patients with atrial fibrillation and/or venous thromboembolism (n=1958).

Warfarin

All DOAC number (%) or median[IQR]

1,557 80 [72-85]

401 82 [75-88]

227 (15) 868 (56) 398 (26) 64 (4) 3 [2-4] 2 [1-3]

57 (14) 216 (54) 107 (27) 21 (5) 3 [2-5] 2 [2-3]

Sites of bleed Lower gastrointestinal Upper gastrointestinal Subdural Subarachnoid Intracerebral Other Patients followed-up In-hospital deaths within 30 days Days in hospital before death Days in hospital for discharged patients

166 (11) 293 (19) 313 (20) 74 (5) 342 (22) 369 (24) 1512 322 (21) 3 [1-8] 7 [3-13]

75 (19) 101 (25) 42 (10) 17 (4) 92 (23) 7 4 (18) 393 84 (21) 3 [1-10] 6 [3-11]

Number of complications in hospital None 1 2 3+

1,184 (76) 229 (15) 62 (4) 37 (2)

308 (77) 60 17 (4) 8 (2)

Total patients Age, years Number of co-morbiditiesd None 1-2 3-4 5+ CHA2DS2-VASc score HAS-BLED scoree

<0.001a

0.724b

0.1a 0.1a

<0.001b

0.973b 0.948a 0.303a

0.977c

DOAC: direct oral anticoagulants. aMann Whitney test; bchi-squared test; cFisher exact test; dco-morbidities as described in Table 1 e score does not include the labile International Normalized Ratio.

haematologica | 2018; 103(4)

741

L. Green et al. Table 4. Univariable and multivariable analysis of risk factors for mortality (n=2,132).

Age (years) <65 65-74 75-84 85+

OR [95%CI]

Univariable analysis P

Adjusted ORa [95%CI]

Multivariable analysis P

1 1.84 [1.12,3.01] 3.07 [1.97,4.80] 3.59 [2.28,5.65]

0.016 <0.001 <0.001

1 1.50 [0.89,2.53] 2.77 [1.73,4.44] 3.78 [2.33,6.14]

0.12 <0.001 <0.001

Type of oral anticoagulant Warfarin Direct oral anticoagulantb Site of bleed Other Lower gastrointestinal Upper gastrointestinal Subdural/epidural Subarachnoid Intracerebral

1 0.99 [0.76, 1.31]

0.98

1 0.96 [0.71,1.28]

0.77

1 0.65 [0.38, 1.12] 1.67 [1.14, 2.46] 2.43 [1.66, 3.55] 2.72 [1.58, 4.68] 6.34 [4.51, 8.91]

0.12 0.01 <0.001 <0.001 <0.001

1 0.45 [0.25, 0.79] 1.17 [0.78, 1.75] 2.55 [1.69, 3.84] 3.23 [1.82, 5.71] 5.75 [4.01, 8.23]

0.005 0.44 <0.001 <0.001 <0.001

Provocationc Spontaneous Trauma Surgery/procedure Fall Liver failure Cancer

1 0.82 [0.55,1.21] 0.23 [0.10,0.52] 0.97 [0.74,1.28] 2.08 [1.02,4.26] 1.29 [0.97,1.70]

0.31 0.001 0.84 0.04 0.08

1 0.54 [0.35, 0.83] 0.50 [0.21, 1.19] 0.53 [0.38, 0.73] 3.86 [1.79, 8.36] 1.37 [1.01, 1.84]

0.005 0.12 0.001 0.001 0.04

OR: odds ratio; aOdds ratios adjusted for all other variables in model; bIncludes dabigatran, rivaroxaban, apixaban and edoxaban; cunclassified provocations not displayed (n=21).

Results Cases were identified and reported by 32 hospitals across the UK, with the median number (inter-quartile range, IQR) of cases reported per hospital being 44 (25 – 88), and the median (IQR) duration for completing a case report (measured from the date of bleeding) being 38 (1387) days. Table 1 shows the demographics, OAC therapy specifications and co-morbidities of 2,192 individual patients reported with major bleeding between October 2013 and August 2016; 34 (1.6%) had more than one episode during the study period (data not shown). All patients fulfilled the International Society of Thrombosis and Haemostasis criteria for major bleeding: 20% had a fatal bleed; 62% were transfused with ≥2 units of red blood cells or had a drop of ≥20g/L hemoglobin; and 86% had symptomatic bleeding in a critical organ (overall, 67% of patients met more than one criteria). At the time of the bleed, the majority of patients were on warfarin (81%) with the remainder being on a DOAC. Patients taking an OAC for atrial fibrillation and venous thromboembolism made up 72% and 21% of the cohort respectively; approximately 10% had multiple indications. Bleeding presentations and outcomes are shown in Table 2. All intracranial hemorrhages (subdural, subarachnoid and intracerebral bleeds combined), gastrointestinal bleeds (upper and lower) and other bleeds made up 44%, 32% and 24% of the total, respectively. Two-thirds of the bleeds were reported as spontaneous and one-fifth as resulting from a fall. Outcomes up to 30 days were reported for 2,132 (97%) patients. The overall in-hospital mortality rate was 20.9% (95% CI: 19.2% - 22.7%) but was 32.7% (95% CI: 29.7-35.9%) among patients with an 742

intracranial hemorrhage. Bleeding was mentioned as the cause of death in 70% of cases, while in 22% it was not, and in 8% the cause of death was unknown/ or referred to the coroner. Patients discharged within 30 days of bleeding stayed a median of 7 days in hospital. In-hospital complications included admission to an Intensive Care Unit (10%), ventilation or acute respiratory distress syndrome (5%), hemorrhagic stroke (3%), cardiac arrest (2%), sepsis or pneumonia (2%), thrombotic events (2%), and rebleeding (2%). For the management of bleeding, patients on warfarin were given any blood transfusion (31%), prothrombin complex concentrate (78%) and vitamin K (74%); patients on DOAC were given any blood transfusion (43%), prothrombin complex concentrate (39%) and tranexamic acid (28%). OAC therapy was resumed in 43% of survivors while they were in hospital. The proportions of patients (among survivors) who restarted OAC after a bleed varied by site of bleed: intracranial hemorrhage (24%), gastrointestinal (46%), other (65%), as well as indication: atrial fibrillation (36%), mechanical heart valves (86%), venous thromboembolism (50%), other (42%). One–third of patients on a DOAC were restarted on a different OAC, as were approximately one-fifth of patients with intracranial hemorrhage or venous thromboembolism. Out of 111 DOAC patients, 29 (26%) were switched to vitamin K antagonists and out of 610 patients on warfarin, 42 (7%) were switched to a DOAC. Comparison of the 1,958 (89%) patients who were on OAC for atrial fibrillation and/or venous thromboembolism (Table 3) showed that those admitted on DOAC were significantly older than those on warfarin (P<0.001), but there was no significant difference in the number of haematologica | 2018; 103(4)

Major bleeding associated with oral anticoagulants

co-morbidities between them. The type of OAC was associated with the site of bleeding (P<0.001); after adjusting for age and bleeding provocation using multinomial models, the conditional odds ratios of admission with a subdural/epidural, subarachnoid and intracerebral bleed (versus a lower gastrointestinal bleed) on DOAC compared to warfarin, were 0.28 (95% CI: 0.17-0.47; P<0.001), 0.52 (95% CI: 0.30-0.90; P=0.02) and 0.61 (95% CI: 0.41-0.92; P=0.02), respectively. Mortality rate, the number of complications following bleeding, and the duration of hospitalization were not significantly different between patients treated with warfarin or DOAC. Table 4 shows the analysis of risk factors for fatality in 2,132 patients with known outcomes. After controlling for age, co-morbidities, bleeding site and provocation, there was no evidence that compared to warfarin, DOAC were associated with different odds of death (adjusted OR=0.96; 95% CI: 0.71-1.28; P=0.77). Intracerebral, subarachnoid and subdural/epidural bleeds were associated with 5.8 (95% CI: 4.0-8.2), 3.2 (95% CI: 1.8-5.7) and 2.6 (95% CI: 1.7-3.8) times higher odds of death, respectively, compared with other bleeds. Adjusting for other covariates, advanced age, spontaneous bleeding (compared to trauma or falls), liver failure and cancer were significantly associated with increased odds of dying, while indications for OAC were not predictors of an adverse outcome. There was negligible between-hospital variation (intraclass correlation coefficient=0.003). Given the relatively small number of patients for whom outcome was missing (n=60), sensitivity analyses were performed by assuming either all survived or died, and this did not materially alter the significance of the estimates.

Discussion The main objective of this study was to investigate the aggregate burden of major hemorrhage in patients receiving OAC, across all indications and drug types, in terms of ensuing mortality, morbidity and hospitalization. Our results showed a considerable death rate, with one-fifth of all patients, and one-third of those with intracranial hemorrhage, dying in hospital within 30 days of a major bleeding episode. Of those who were discharged within a month, half had spent 7-30 days in hospital. Our mortality rates, both overall and for the subgroup with intracranial hemorrhage, are comparable with those of other studies.17,18 However, a recent Canadian study that used a similar definition of major bleeding as ours (including only patients >66 years) showed lower mortality rates than in our study (9.2% for DOAC and 15.2% for warfarin): this difference is likely to be due to the lower rate of intracranial hemorrhage (27% overall) seen in their study.19 The proportion of patients with intracranial hemorrhage (44%) was similar to that in the study by Becattini and colleagues,20 but higher than that in other studies.19,21 The higher rate may be due to: (i) one-third of our cases being from hospitals with specialist neurosurgical units (i.e. endpoints of inter-hospital referrals); (ii) our cohort including patients who were older than those in clinical trials and thus more prone to falls and the development of intracranial hemorrhage; (iii) our definition of major bleeding being more severe than those of clinical trials, as we identified only patients who developed major hemorrhage culminating in hospitalization; (iv) cases of intracranial haematologica | 2018; 103(4)

hemorrhage being more exhaustively documented (e.g. through administration of prothrombin complex concentrates), compared with gastrointestinal or other bleeds. The pivotal randomized controlled trials of patients with atrial fibrillation and venous thromboembolism found lower rates of intracranial hemorrhage with all DOAC than with warfarin, and higher rates of gastrointestinal bleeding with dabigatran and rivaroxaban than with warfarin.9,21 A systematic review of randomized clinical trials by Connolly and colleagues22 concluded that “the risks of subdural hematoma were significantly higher with vitamin K antagonists compared with factor-Xa inhibitors and direct thrombin inhibitors” [meta-analysis odds ratio (95% CI): 2.9 (2.1–4.1) and 1.8 (1.2–2.7), respectively]. Our findings appear to be a corollary of the established evidence: while our study did not directly assess the incidence of bleeding on OAC, we observed increased odds of admission with all types of intracranial hemorrhage versus gastrointestinal bleeding on warfarin compared to DOAC, with the association being strongest for subdural/epidural bleeding. Further investigation into this association between warfarin and subdural hematoma would thus be an important topic in future research. Analysis of risk factors for death showed that intracranial hemorrhage was a strong predictor of death. However, we saw no difference in the proportions of deaths between patients treated with warfarin or DOAC even though the proportion of intracranial hemorrhage was higher for the former. A likely explanation for this is that patients admitted on DOAC were significantly older than those on warfarin, and advanced age was also an independent risk factor for death. The participants in the clinical trials of warfarin versus DOAC were significantly younger than those in our study. However, our results provide reassurance for clinicians who prescribe DOAC therapy in that, where major bleeding is concerned, the outcome is no worse than that associated with warfarin despite the lack of antidotes and the advanced age of many patients in this study. Since DOAC antidotes are now becoming available it is important to perform similar studies to assess the effect that these have on outcomes, and thereby improve OAC choices in the future. A key strength of this study is that the data were prospectively collected by over 30 hospitals across the UK, with information retrieved directly from patients’ case notes and not from large databases in which data are collected for different reasons, or missed by administrative coding.23-26 This has enabled us to document for the first time in the UK the characteristics and clinical outcomes of patients who develop major bleeding associated with different OAC when prescribed for any clinical indications. Additionally, over three-quarters of cases were fully reported within 90 days of the major bleeding episode, ensuring a high degree of fidelity with actual events and minimizing the chance of recall bias. Furthermore, we were able to collect data on the acute management of major bleeding, whereas this is not usually possible from large data sets. Our analysis showed a very low level of inter-hospital variation, which suggests that our findings should be generalizable to the wider population and potentially further afield. Moreover, the design of the study meant that we did not rely on patients’ consent, a further potential source of selection bias. Our study also provides data on the outcomes of major bleeding that are more generalizable to contemporary 743

L. Green et al.

practice than those observed in the randomized clinical trials. Patients in the latter were more highly selected, and the management of major bleeding in the trials appears different from current recommended practice; for example, among those with major bleeding in the open-label warfarin arm of the RELY trial only 27% received vitamin K and 1% received a prothrombin complex concentrate27 in contrast to over 70% for both in our study. We identified patients who developed major bleeding on OAC and were admitted to hospital, but it is possible that not every case was captured, because they were not triaged to emergency departments, and thus could have been missed by the research team. The upshot is that we may have collected data on the more severe bleeding cases with consequently worse outcomes, although we expect the numbers affected to be relatively small. Furthermore, it could be argued that warfarin cases would have been easier to capture through the prescription of prothrombin complex concentrate. However, the methods of case identification were designed to capture all cases of major bleeding from hospitals, independently of the type of OAC, but we recognize that some cases could have been inadvertently missed. It is also possible that we may have missed some cases of fatal bleeding who died suddenly in the community; however, we think these numbers will likely be small because the majority would be taken to their local emergency department because of acute deterioration. Furthermore, we might have missed the outcome of patients who were discharged before 30 days; however, it seems a reasonable stipulation that patients were discharged from hospital if and when major bleeding had been successfully treated or resolved, and hence restricting follow-up to within the hospital is justifiable on the grounds that any post-discharge deaths would not be “following from the bleeding event”. In conclusion, for clinicians who are responsible for counseling patients about oral anticoagulation our findings indicate that mortality from major bleeding on OAC that requires hospitalization is high, with one in five patients likely to die within 30 days. This is of concern given the widening use of these agents in an increasingly elderly and frail population. As far as DOAC are concerned, our results provide reassurance, in that they are

References 1. Ruff CT, Giugliano RP, Braunwald E, et al. Comparison of the efficacy and safety of new oral anticoagulants with warfarin in patients with atrial fibrillation: a meta-analysis of randomised trials. Lancet. 2014;383 (9921):955-962. 2. Camm AJ, Kirchhof P, Lip GY, et al. Guidelines for the management of atrial fibrillation: the Task Force for the Management of Atrial Fibrillation of the European Society of Cardiology (ESC). Eur Heart J. 2010;31(19):2369-23429. 3. Wolberg AS, Rosendaal FR, Weitz JI, et al. Venous thrombosis. Nat Rev Dis Primers. 2015;1:15006. 4. Cannegieter SC, Rosendaal FR, Briet E. Thromboembolic and bleeding complications in patients with mechanical heart valve prostheses. Circulation. 1994;89(2):635-641.

744

not associated with worse outcomes compared to warfarin despite the lack of antidotes at the time of the study. As the DOAC antidotes emerge, it would be important for future studies to investigate their impact on patients’ outcomes compared to current practice. Funding This study was funded by the British Society for Haematology Early Stage Researcher Fellowship awarded to Dr Laura Green. The views expressed in this publication are those of the authors and not those of the funder. The funder had no role in the study design, data collection/analysis or preparation of this article. Acknowledgments We would like to acknowledge all hospitals, NHS trusts and principal investigators who reported to the study: Newham University Hospital (Dr Olivia Kreze), Whipps Cross University Hospital (Dr Peter MacCallum); The Royal London Hospital (Dr Laura Green); Barking Havering and Redbridge University Hospitals (Dr Khalid Saja); John Radcliffe Hospital (Dr Nicola Curry); Glasgow Royal Infirmary (Dr Campbell Tait); University Hospital of Wales (Dr Raza Alikhan); Sheffield Teaching Hospitals (Dr Rhona Maclean); Basingstoke & North Hampshire Hospital and Salisbury District Hospital (Dr Tamara Everington); Ulster Hospital (Dr Margaret Bowers); Wexham Park Hospital (Dr Sarah Wilson); Withybush General Hospital (Dr Sumant Kundu); Newcastle upon Tyne Hospitals (Dr John Hanley); North Middlesex University Hospital (Dr John Luckit); Basildon and Thurrock Hospitals (Dr Godwin Simon); Torbay Hospital (Dr Nichola Rymes); University College Hospital (Dr Hannah Cohen); St George's Healthcare NHS Trust (Dr James Uprichard); Birmingham Heartlands Hospital (Dr Charalampos Kartsios); Royal Bournemouth Hospital (Dr Jason Mainwaring); Hull and East Yorkshire Hospitals NHS Trust (Dr Salama Abosaad); Northern Lincolnshire and Goole NHS Foundation Trust (Dr Sanjeev Jalihal); Dumfries and Galloway Royal Infirmary (Dr Mohamed Khan); University Hospitals of North Midlands (Dr Deepak Chandra); Gloucestershire Hospitals (Dr Oliver Miles); South Tees Hospitals (Dr Jamie Maddox); Glangwili General Hospital (Dr Saran Nicholas); Royal Berkshire NHS Foundation Trust (Dr Liza Keating); University Hospital Southampton (Dr Sara Boyce); Northumbria Healthcare NHS Foundation Trust (Dr Charlotte Bomken) and North Cumbria University Hospitals (Dr Roderick Oakes).

5. van Es N, Coppens M, Schulman S, Middeldorp S, Buller HR. Direct oral anticoagulants compared with vitamin K antagonists for acute venous thromboembolism: evidence from phase 3 trials. Blood. 2014;124(12):1968-1975. 6. NICE. Assumptions used in estimating a population benchmark. 2012. (Available from http://www nice org uk/usingguidance/commissioningguides/anticoagulationtherapyservice/popbench jsp). Accessed on 10 October 2017. 7. Chugh SS, Roth GA, Gillum RF, Mensah GA. Global burden of atrial fibrillation in developed and developing nations. Glob Heart. 2014;9(1):113-119. 8. Martinez C, Cohen AT, Bamber L, Rietbrock S. Epidemiology of first and recurrent venous thromboembolism: a populationbased cohort study in patients without active cancer. Thromb Haemost.

2014;112(2):255-263. 9. Chai-Adisaksopha C, Crowther M, Isayama T, Lim W. The impact of bleeding complications in patients receiving target-specific oral anticoagulants: a systematic review and meta-analysis. Blood. 2014;124(15):24502458. 10. Hecker J, Marten S, Keller L, et al. Effectiveness and safety of rivaroxaban therapy in daily-care patients with atrial fibrillation. Results from the Dresden NOAC Registry. Thromb Haemost. 2016;115(5): 939-949. 11. Beyer-Westendorf J, Ebertz F, Forster K, et al. Effectiveness and safety of dabigatran therapy in daily-care patients with atrial fibrillation. Results from the Dresden NOAC Registry. Thromb Haemost. 2015;113(6): 1247-1257. 12. Lip GY, Keshishian A, Kamble S, et al. Realworld comparison of major bleeding risk

haematologica | 2018; 103(4)

Major bleeding associated with oral anticoagulants

13.

14.

15.

16.

among non-valvular atrial fibrillation patients initiated on apixaban, dabigatran, rivaroxaban, or warfarin. A propensity score matched analysis. Thromb Haemost. 2016;116(5):975-986. Wang SV, Franklin JM, Glynn RJ, Schneeweiss S, Eddings W, Gagne JJ. Prediction of rates of thromboembolic and major bleeding outcomes with dabigatran or warfarin among patients with atrial fibrillation: new initiator cohort study. BMJ. 2016;353:i2607. Hylek EM, Evans-Molina C, Shea C, Henault LE, Regan S. Major hemorrhage and tolerability of warfarin in the first year of therapy among elderly patients with atrial fibrillation. Circulation. 2007;115(21):26892696. Nielsen PB, Skjoth F, Sogaard M, Kjaeldgaard JN, Lip GY, Larsen TB. Effectiveness and safety of reduced dose non-vitamin K antagonist oral anticoagulants and warfarin in patients with atrial fibrillation: propensity weighted nationwide cohort study. BMJ. 2017;356:j510. Schulman S, Kearon C, Subcommittee on Control of Anticoagulation of the S, Standardization Committee of the International Society on T, Haemostasis. Definition of major bleeding in clinical investigations of antihemostatic medicinal products in non-surgical patients. J Thromb

haematologica | 2018; 103(4)

Haemost. 2005;3(4):692-694. 17. Nieto JA, Solano R, Ruiz-Ribo MD, et al. Fatal bleeding in patients receiving anticoagulant therapy for venous thromboembolism: findings from the RIETE registry. J Thromb Haemost. 2010;8(6):1216-1222. 18. Kuramatsu JB, Gerner ST, Schellinger PD, et al. Anticoagulant reversal, blood pressure levels, and anticoagulant resumption in patients with anticoagulation-related intracerebral hemorrhage. JAMA. 2015; 313(8):824-836. 19. Xu Y, Schulman S, Dowlatshahi D, et al. Direct oral anticoagulant- or warfarin-related major Bleeding: characteristics, reversal strategies, and outcomes from a multicenter observational study. Chest. 2017;152(1):8191. 20. Becattini C, Franco L, Beyer-Westendorf J, et al. Major bleeding with vitamin K antagonists or direct oral anticoagulants in real-life. Int J Cardiol. 2017;227:261-266. 21. Chatterjee S, Sardar P, Biondi-Zoccai G, Kumbhani DJ. New oral anticoagulants and the risk of intracranial hemorrhage: traditional and Bayesian meta-analysis and mixed treatment comparison of randomized trials of new oral anticoagulants in atrial fibrillation. JAMA Neurol. 2013;70(12):14861490. 22. Connolly BJ, Pearce LA, Hart RG. Vitamin K antagonists and risk of subdural hematoma:

23.

24.

25.

26.

27.

meta-analysis of randomized clinical trials. Stroke. 2014;45(6):1672-1678. Larsen TB, Skjoth F, Nielsen PB, Kjaeldgaard JN, Lip GY. Comparative effectiveness and safety of non-vitamin K antagonist oral anticoagulants and warfarin in patients with atrial fibrillation: propensity weighted nationwide cohort study. BMJ. 2016;353: i3189. Li L, Geraghty OC, Mehta Z, Rothwell PM, Oxford Vascular S. Age-specific risks, severity, time course, and outcome of bleeding on long-term antiplatelet treatment after vascular events: a populationbased cohort study. Lancet. 2017;390 (10093):490-499. Yao X, Abraham NS, Alexander GC, et al. Effect of adherence to oral anticoagulants on risk of stroke and major bleeding among patients with atrial fibrillation. J Am Heart Assoc. 2016;5(2). pii: e003074. Graham DJ, Reichman ME, Wernecke M, et al. Stroke, bleeding, and mortality risks in elderly Medicare beneficiaries treated with dabigatran or rivaroxaban for nonvalvular atrial fibrillation. JAMA Intern Med. 2016;176(11):1662-1671. Majeed A, Hwang HG, Connolly SJ, et al. Management and outcomes of major bleeding during treatment with dabigatran or warfarin. Circulation. 2013;128(21):23252332.

745