Haematologica, Volume 104, Issue 9 by Haematologica

haematologica Journal of 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 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 2019 are as following: Print edition

Institutional Euro 700

Personal Euro 170

Advertisements. Contact the Advertising Manager, Haematologica Office, via Giuseppe Belli 4, 27100 Pavia, Italy (phone +39.0382.27129, fax +39.0382.394705, e-mail: marketing@haematologica.org). Disclaimer. Whilst every effort is made by the publishers and the editorial board to see that no inaccurate or misleading data, opinion or statement appears in this journal, they wish to make it clear that the data and opinions appearing in the articles or advertisements herein are the responsibility of the contributor or advisor concerned. Accordingly, the publisher, the editorial board and their respective employees, officers and agents accept no liability whatsoever for the consequences of any inaccurate or misleading data, opinion or statement. Whilst all due care is taken to ensure that drug doses and other quantities are presented accurately, readers are advised that new methods and techniques involving drug usage, and described within this journal, should only be followed in conjunction with the drug manufacturer’s own published literature. Direttore responsabile: Prof. 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 Journal of the Ferrata Storti Foundation

Table of Contents Volume 104, Issue 9: September 2019 Cover Figure Bone marrow smear showing atypical mast cells in a patient with systemic mastocytosis, mast cell leukemia variant. Courtesy of Prof. Rosangela Invernizzi.

Editorials 1689

Stem cell factor: the bridge between bone marrow adipocytes and hematopoietic cells Ziru Li and Ormond A. MacDougald

1691

New potential players in hepcidin regulation Maxwell Chappell and Stefano Rivella

1694

Ubiquitination is not omnipresent in myeloid leukemia Ramesh C. Nayak and Jose A. Cancelas

1696

Six-packed antibodies punch better Christoph Rader and Adrian Wiestner

1699

The secret afterlife of platelets Nicholas A. Arce and Renhao Li

Perspective Article 1702

Hemophilia A and B: molecular and clinical similarities and differences Giancarlo Castaman and Davide Matino

Review Articles 1710

Emerging disease-modifying therapies for sickle cell disease Marcus A. Carden and Jane Little

1720

Targeting sickle cell disease root-cause pathophysiology with small molecules Yogen Saunthararajah

Articles Hematopoiesis

1731

Bone marrow adipose tissue-derived stem cell factor mediates metabolic regulation of hematopoiesis Zengdi Zhang et al.

1744

MicroRNA-127-3p controls murine hematopoietic stem cell maintenance by limiting differentiation Laura Crisafulli et al.

Iron Metabolism & its Disorders

1756

Dimeric ferrochelatase bridges ABCB7 and ABCB10 homodimers in an architecturally defined molecular complex required for heme biosynthesis Nunziata Maio et al.

Iron Metabolism & its Disorders

1768

New thiazolidinones reduce iron overload in mouse models of hereditary hemochromatosis and Î˛-thalassemia Jing Liu et al.

Myeloproliferative Neoplasms

1782

Long-term outcome after allogeneic hematopoietic cell transplantation for myelofibrosis Marie Robin et al.

Haematologica 2019; vol. 104 no. 9 - September 2019 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation

Chronic Myeloid Leukemia

1789

De novo UBE2A mutations are recurrently acquired during chronic myeloid leukemia progression and interfere with myeloid differentiation pathways Vera Magistroni et al.

Acute Myeloid Leukemia

1798

Sequential therapy for patients with primary refractory acute myeloid leukemia: a historical prospective analysis of the German and Israeli experience Ron Ram et al.

Acute Lymphoblastic Leukemia

1804

Glucocorticoids and selumetinib are highly synergistic in RAS pathway-mutated childhood acute lymphoblastic leukemia through upregulation of BIM Elizabeth C. Matheson et al.

1812

Asparagine levels in the cerebrospinal fluid of children with acute lymphoblastic leukemia treated with pegylated-asparaginase in the induction phase of the AIEOP-BFM ALL 2009 study Carmelo Rizzari et al.

Non-Hodgkin Lymphoma

1822

Burkitt-like lymphoma with 11q aberration: a germinal center-derived lymphoma genetically unrelated to Burkitt lymphoma Blanca Gonzalez-Farre et al.

Chronic Lymphocytic Leukemia

1830

Energy metabolism is co-determined by genetic variants in chronic lymphocytic leukemia and influences drug sensitivity Junyan Lu et al.

Cell Therapy & Immunotherapy

1841

CD20 and CD37 antibodies synergize to activate complement by Fc-mediated clustering Simone C. Oostindie et al.

Platelet Biology & its Disorders

1853

CD45 expression discriminates waves of embryonic megakaryocytes in the mouse Isabel Cortegano et al.

1866

Fatal dysfunction and disintegration of thrombin-stimulated platelets Oleg V. Kim et al.

1879

A novel combinatorial technique for simultaneous quantification of oxygen radicals and aggregation reveals unexpected redox patterns in the activation of platelets by different physiopathological stimuli Dina Vara et al.

Hemostasis

1892

Tspan18 is a novel regulator of the Ca2+ channel Orai1 and von Willebrand factor release in endothelial cells Peter J. Noy et al.

Letters to the Editor Letters are available online only at www.haematologica.org/content/104/9.toc

e385

Children with sickle cell anemia and APOL1 genetic variants develop albuminuria early in life Rima S. Zahr et al. http://www.haematologica.org/content/104/9/e385

Haematologica 2019; vol. 104 no. 9 - September 2019 http://www.haematologica.org/

haematologica Journal of the Ferrata Storti Foundation e388

Myelodysplastic syndrome-associated spliceosome gene mutations enhance innate immune signaling Daniel A. Pollyea et al. http://www.haematologica.org/content/104/9/e388

e393

Adult patients with de novo acute myeloid leukemia show a functional deregulation of redox balance at diagnosis which is correlated with molecular subtypes and overall survival Julie Mondet et al. http://www.haematologica.org/content/104/9/e393

e398

Sorafenib improves survival of FLT3-mutated acute myeloid leukemia in relapse after allogeneic stem cell transplantation: a report of the EBMT Acute Leukemia Working Party Ali Bazarbachi et al. http://www.haematologica.org/content/104/9/e398

e402

TP53, ETV6 and RUNX1 germline variants in a case series of patients developing secondary neoplasms after treatment for childhood acute lymphoblastic leukemia Stefanie V. Junk et al. http://www.haematologica.org/content/104/9/e402

e406

A high definition picture of key genes and pathways mutated in pediatric follicular lymphoma Federica Lovisa et al. http://www.haematologica.org/content/104/9/e406

e410

A B-cell receptor-related gene signature predicts response to ibrutinib treatment in mantle cell lymphoma cell lines Tiziana D'Agaro et al. http://www.haematologica.org/content/104/9/e410

e415

A metabolic switch in proteasome inhibitor-resistant multiple myeloma ensures higher mitochondrial metabolism, protein folding and sphingomyelin synthesis Lenka Besse et al. http://www.haematologica.org/content/104/9/e415

e420

Quantitative dynamic 18F-fluorodeoxyglucose positron emission tomography/computed tomography before autologous stem cell transplantation predicts survival in multiple myeloma Christos Sachpekidis et al. http://www.haematologica.org/content/104/9/e420

e424

The diagnostic performance of renal function-adjusted D-dimer testing in individuals suspected of having venous thromboembolism Vincent ten Cate et al. http://www.haematologica.org/content/104/9/e424

Case Reports Case Reports are available online only at www.haematologica.org/content/104/9.toc

e428

Compound heterozygosity in PKLR gene for a previously unrecognized intronic polymorphism and a rare missense mutation as a novel cause of severe pyruvate kinase deficiency Shruti Bagla et al. http://www.haematologica.org/content/104/9/e428

e432

Daratumumab for treatment of blastic plasmacytoid dendritic cell neoplasm. A single-case report Katrine F. Iversen et al. http://www.haematologica.org/content/104/9/e432

e434

Venetoclax resistance and acquired BCL2 mutations in chronic lymphocytic leukemia Eugen Tausch et al. http://www.haematologica.org/content/104/9/e434

Haematologica 2019; vol. 104 no. 9 - September 2019 http://www.haematologica.org/

EDITORIALS Stem cell factor: the bridge between bone marrow adipocytes and hematopoietic cells Ziru Li and Ormond A. MacDougald Department of Molecular & Integrative Physiology, University of Michigan Medical School, Ann Arbor, MI, USA. E-mail: ZIRU LI - liziru@umich.edu doi:10.3324/haematol.2019.224188

hite adipocytes serve as an energy reservoir to store excessive calories in the form of lipid droplets and protect other tissues or organs from ectopic lipid accumulation. Brown adipocytes express uncoupling protein 1 and are integral to adaptive thermogenesis. Whereas the functions of adipocytes in either white or brown adipose tissues are well documented, our knowledge of bone marrow adipocytes (BMA) remains in its infancy. Bone marrow adipose tissue (BMAT) occupies approximately 50-70% of the bone marrow volume in human adults.1 It is a dynamic tissue and responds to multiple metabolic conditions. For example, BMAT increases with obesity, aging, diabetes, caloric restriction, and irradiation.2 Although the significance of BMAT expansion under these conditions is still largely unknown, BMA interact locally with hematopoietic and bone cells, and contribute to global metabolism through secretion of adiponectin, leptin, stem cell factor (SCF), and other functional factors. For example, A-ZIP/F1 mice, which lack adipose tissues throughout the body, including BMAT, have delayed hematopoietic regeneration in long bones after irradiation.3 Our latest work also observed that depletion of BMA by bariatric surgery is associated with a decrease in bone marrow erythroid cells and anemia.4 The importance of BMA and the derived factors on hematopoiesis is further enhanced by a study in this issue of the Journal, in which Zhang et al.5 demonstrate that BMATderived SCF mediates metabolic regulation of hematopoiesis. Stem cell factor, also known as Kit ligand (Kitl), is a hematopoietic cytokine expressed in fibroblasts and endothelial cells, as well as in BMA.3 Together with its receptor, c-Kit, SCF plays important roles in the maintenance of hematopoietic stem cells (HSC) and hematopoiesis. Blockade of the interaction between c-Kit and SCF with antic-Kit antibody promotes the clearance of HSC, which indicates the importance of Kitl/c-Kit signaling in HSC selfrenewal.6 Loss-of-function mutations in c-Kit cause macrocytic anemia, or even embryonic lethality under some severe mutations.7 Inversely, mice with c-Kit gain-of-function mutations developed erythrocytosis compatible with myeloproliferative disorders.8 Analyses of multiple cell populations isolated from bone marrow and adipose tissue have demonstrated that BMA and LepR-positive (+) stromal cells are the primary sources of SCF, which is required for the regeneration of HSC and hematopoiesis after irradiation.3 Zhang et al. report that BMA-derived SCF is important for hematopoietic homeostasis under basal (Figure 1), obese and aging conditions, and in response to Î˛3-adrenergic agonists.5 Knockout of SCF in adipocytes with an adiponectin driver does not influence circulating SCF concentrations or phenotypes of the peripheral adipose depots, which is perhaps due to compensatory expression of SCF from other sources, such as endothelial cells, fibroblasts and stromal cells. Interestingly, Zhang et al. observed a significant loss of SCF haematologica | 2019; 104(9)

in the bone marrow supernatant, which indicates that BMAT is a primary source of SCF in bone marrow.5 Deficiency of SCF in BMAT reduces the bone marrow cellularity, hematopoietic stem and progenitor cells (HSPC), common myeloid progenitors (CMP), megakaryocyte-erythrocyte progenitor (MEP) and granulocyte-monocyte progenitors (GMP) under steady-state condition. Consistent with these changes in the progenitor cells of bone marrow, mice deficient for adipocyte SCF develop macrocytic anemia and reduction of neutrophils, monocytes and lymphocytes in circulation. In contrast to results in this study, Zhou et al. reported that the conditional deficiency of SCF in adipocytes driven by adiponectin-Cre/ER had no effect on hematopoiesis under basal conditions.3 Although further investigation is necessary, the discrepancy between these two studies might be due to the time-frame of SCF deletion, tamoxifen injection and/or animal lines. Of note, the deletion of SCF has no effect on the proliferation of HSPC evidenced by colony-forming assays, which suggests that defects in BMAT-derived SCF influences the bone marrow microenvironment rather than the intrinsic function of HSPC. Since adiponectin-Cre is expressed in both peripheral adipocytes and BMA, it is possible that there might be effects on hematopoiesis that are independent of BMAT. To more specifically study effects of BMA on the bone marrow niche and hematopoiesis, Zhang et al. also deleted the Kitl using osterix promoter, which traces BMA but not the other adipocytes. Again, knockout of Kitl from the osterix-positive (+) cells reduced bone marrow cellularity, hematopoietic progenitor populations and mature blood cells including red blood cells (RBC), neutrophils and monocytes, which is consistent with the phenotypes from mice lacking adipocytic Kitl. Of note, in addition to BMA, osterix+ progenitors also trace to osteoblasts.9,10 Mesenchymal and osteoblast lineage cells are involved in the maintenance and regulation of the supportive microenvironments necessary for quiescence, self-renewal and differentiation of HSC.11,12 However, the SCF from osteoblasts is not required for HSC maintenance in adult bone marrow under steady-state conditions.13 Although the possible effects of SCF derived from osterix+ progenitors on hematopoiesis could not be excluded and the bone phenotypes were not explored in this mouse model, it should be appreciated that authors used both adiponectinand osterix-driven Cre enzyme to confirm the phenotypes of SCF-deficiency on hematopoiesis. These results strongly point to BMA as an important source of SCF since the common cell type traced by adiponectin and osterix drivers is the BMA; however, development of BMA-specific transgenic mouse tools will be required to truly confirm these observations of BMA and the roles of SCF in the bone marrow niche homeostasis and hematopoiesis. The authors also investigated whether BMA-derived SCF is required for hematopoietic adaptation to aging or high fat 1689

Editorials

Figure 1. Bone marrow adipocytes influence the maintenance of hematopoietic stem cell (HSC) and hematopoiesis. Bone marrow cellularity is complex, but is mainly composed of hematopoietic cells and bone marrow adipocytes (BMA), which appear after birth and accumulate with age, obesity and irradiation. BMA originate from osterix-positive (+) progenitor cells and secret adiponectin, stem cell factor (SCF) and other functional factors. In this study, Zhang et al.5 have demonstrated that BMAT-derived SCF plays important roles in HSC maintenance and hematopoietic differentiation under baseline, aging and obese conditions. Deficiency of SCF in BMAT hinders the self-renewal of HSC by influencing the bone marrow microenvironment and hematopoiesis through unknown mechanisms. RBC: red blood cell; MPP: multipotent progenitor; CMP: common myeloid progenitor; MEP: megakaryocyte-erythrocyte progenitor; GMP: granulocyte-monocyte progenitor; CLP: common lymphoid progenitor.

diet (HFD)-induced obesity. Whereas HFD, per se, did not increase the SCF concentrations in bone marrow supernatant, this treatment increased bone marrow cellularity, HSPC, and mature blood cells, including granulocytes, monocytes and lymphocytes, the effects of which were eliminated by SCF deficiency in adipocytes. Aging causes similar increases in the HSPC, especially in the myeloid lineage populations, and most of these effects required adipocyte-derived SCF. Further, these investigators explored a potential role for SCF in mediating effects of a β3-adrenergic receptor agonist. Activation of these receptors induces the lipolysis of white adipocytes, and while although BMAT lipolysis is relatively resistant to β-adrenergic signaling,14 Zhang et al. observed that after administration of a β3-adrenoceptor agonist, CL316, 243, SCF expression was increased in bone marrow without significant changes in the BMA numbers.5 Consistent with the elevated SCF in bone marrow, the numbers of HSPC, including Lin-Sca1+c-Kit+ (LSK) cell, multipotent progenitor (MPP), MEP, GMP and CLP were increased by CL316, 243 injection, the effects of which were compromised by adipocyte-specific deficiency of SCF. Based on the animal 1690

models described above, it should be noted that alterations of BMAT, SCF and hematopoiesis were not tightly associated under these conditions, which suggests that hematopoietic metabolism is regulated by factors beyond BMAT and its derived SCF. The global effects of obesity, aging and β3-adrenoceptor activation cannot be excluded from this scenario. In addition, other secreted factors from BMAT may also play significant roles in hematopoiesis under these conditions. Unfortunately, the secretome of BMAT remains largely unexplored. In summary, Zhang et al.5 have extended our understanding of the roles of BMAT in the bone marrow niche and the interaction between BMA and hematopoietic cells. They thoroughly addressed their hypotheses using a variety of animal models and complete profiling of hematopoietic changes. However, due to the complexity of whole-body metabolism and the lack of BMA-specific transgenic tools, further work will be required to determine whether BMA-derived SCF regulates hematopoiesis directly through Kitl/c-Kit signaling in hematopoietic cells or indirectly by changing the microenvironment of the bone marrow niche. haematologica | 2019; 104(9)

Editorials

Acknowledgments This work was supported by grants from the NIH to OAM (R24 DK092759; R01 DK62876), and from the American Diabetes Association to ZL (1-18-PDF-087).

References 1. Cawthorn WP, Scheller EL, Learman BS, et al. Bone marrow adipose tissue is an endocrine organ that contributes to increased circulating adiponectin during caloric restriction. Cell Metab. 2014;20(2):368375. 2. Li Z, Hardij J, Bagchi DP, Scheller EL, MacDougald OA. Development, regulation, metabolism and function of bone marrow adipose tissues. Bone. 2018;110:134-140. 3. Zhou BO, Yu H, Yue R, et al. Bone marrow adipocytes promote the regeneration of stem cells and haematopoiesis by secreting SCF. Nat Cell Biol. 2017;19(8):891-903. 4. Li Z, Hardij J, Evers SS, et al. G-CSF partially mediates effects of sleeve gastrectomy on the bone marrow niche. J Clin Invest. 2019;130:2404-2416. 5. Zhang Z, Huang Z, Ong B, Sahu C, Zeng H, Ruan HB. Bone marrow adipose tissue-derived stem cell factor mediates metabolic regulation of hematopoiesis. Haematologica. 2019;104(9):1731-1743. 6. Czechowicz A, Kraft D, Weissman IL, Bhattacharya D. Efficient transplantation via antibody-based clearance of hematopoietic stem

cell niches. Science. 2007;318(5854):1296-1299. 7. Nocka K, Majumder S, Chabot B, et al. Expression of c-kit gene products in known cellular targets of W mutations in normal and W mutant mice--evidence for an impaired c-kit kinase in mutant mice. Genes Dev. 1989;3(6):816-826. 8. Bosbach B, Deshpande S, Rossi F, et al. Imatinib resistance and microcytic erythrocytosis in a KitV558Delta;T669I/+ gatekeepermutant mouse model of gastrointestinal stromal tumor. Proc Natl Acad Sci U S A. 2012;109(34):E2276-E2283. 9. Song L, Liu M, Ono N, Bringhurst FR, Kronenberg HM, Guo J. Loss of wnt/beta-catenin signaling causes cell fate shift of preosteoblasts from osteoblasts to adipocytes. J Bone Miner Res. 2012;27(11):23442358. 10. Mizoguchi T, Pinho S, Ahmed J, et al. Osterix marks distinct waves of primitive and definitive stromal progenitors during bone marrow development. Dev Cell. 2014;29(3):340-349. 11. Calvi LM, Adams GB, Weibrecht KW, et al. Osteoblastic cells regulate the haematopoietic stem cell niche. Nature. 2003;425(6960):841846. 12. Visnjic D, Kalajzic Z, Rowe DW, Katavic V, Lorenzo J, Aguila HL. Hematopoiesis is severely altered in mice with an induced osteoblast deficiency. Blood. 2004;103(9):3258-3264. 13. Ding L, Saunders TL, Enikolopov G, Morrison SJ. Endothelial and perivascular cells maintain haematopoietic stem cells. Nature. 2012;481(7382):457-462. 14. Scheller EL, Khandaker S, Learman BS, et al. Bone marrow adipocytes resist lipolysis and remodeling in response to beta-adrenergic stimulation. Bone. 2019;118:32-41.

New potential players in hepcidin regulation Maxwell Chappell and Stefano Rivella Division of Hematology, Department of Pediatrics, Children’s Hospital of Philadelphia, Cell and Molecular Biology Graduate Group, University of Pennsylvania, Abramson Research Center, Philadelphia, PA, USA E-mail: STEFANO RIVELLA - rivellas@email.chop.edu doi:10.3324/haematol.2019.224311

he manuscript by Liu and colleagues, published in this issue of Haematologica, reports the identification of novel compounds able to increase hepcidin expression in normal mice as well as in animals affected by hemochromatosis and β-thalassemia intermedia (or non-transfusion-dependent thalassemia) (Figure 1A).1 Hepcidin is the master regulator of iron secreted from the liver and acts on ferroportin, a transmembrane protein that functions as an iron exporter.2,3 Once hepcidin binds ferroportin, the complex is rapidly degraded, preventing iron egress.2,3 Ferroportin is expressed in many types of cells, including enterocytes and macrophages.2,3 Therefore, the relative abundance of hepcidin in the circulation and ferroportin on cell membranes control iron absorption (from enterocytes) and iron recycling (from macrophages).2,3 Hepcidin expression is regulated by iron, inflammation and erythropoiesis.2,3 With regard to iron-mediated control of hepcidin, this is achieved through at least two mechanisms. The first senses the amount of intracellular iron in liver sinusoidal endothelial cells and responds by synthesizing BMP6, and other similar ligands, belonging to the TGFβ-like family.2-4 Increased intracellular concentration of iron leads to secretion of BMP6 from these cells.2-4 As a consequence, BMP6 binds and activates receptors that trigger phosphorylation of a SMAD complex and stimulate hepcidin expression in hepatic cells.2-4 The second mechanism senses the iron in circulation by recognizing iron-loaded transferrin molecules.3 haematologica | 2019; 104(9)

Molecules such as HFE, transferrin receptor-2, and others communicate intracellularly when the transferrin saturation levels increase.3 It has been hypothesized that this sensing complex potentiates the SMAD complex activated by BMP6.5 Alternatively, or in addition, it has been suggested that this complex acts upon hepcidin expression by decreasing the ERK1/2 pathway.10 Under conditions that require enhanced red cell production (as a consequence of a transient or chronic anemia), hepcidin synthesis is normally suppressed.2 A few factors have been identified that could play a role in this mechanism, such as erythroferrone and platelet-derived growth factor BB.7,8 In particular, erythroferrone is secreted by erythroid cells and acts as a trap ligand, limiting the activity of BMP6 and other similar molecules.9 Another player in the regulation of hepcidin is the molecule matriptase-2 (or TMPRSS6).2,3 This molecule prevents hepcidin overexpression, which could lead to hypoferremia and anemia.3,10 Although it is unclear which pathways and molecules control TMPRSS6, it has been shown that TMPRSS6 is required for erythropoietinmediated hepcidin suppression in mice.11,12 In primary forms of hemochromatosis, patients show excessive iron absorption and suffer from iron overload (Figure 1A).7,8 This happens when hepcidin, or other genes that control its expression, are mutated.2,3 In secondary forms of hemochromatosis (as in β-thalassemia), the anemia triggers increased iron absorption, likely by increased expression of erythroferrone and other hypox1691

Editorials

Figure 1. Role of hepcidin and novel thiazolidinone compounds in the treatment of hemochromatosis and Î˛-thalassemia. (A) Relationship between disease, genetic mutations, hepcidin levels, drug administration and phenotype in primary and secondary forms of hemochromatosis. (B) Potential mechanisms of action and therapeutic effects of new thiazolidinone compounds in hemochromatosis and Î˛-thalassemia.

ia-related molecules that converge on suppressing hepcidin synthesis or increasing ferroportin expression (Figure 1A).2,3,13 Conversely, mutations in the TMPRSS6 gene lead to overexpression of hepcidin. In this case, patients suffer from a condition indicated as iron-refractory iron deficiency anemia or IRIDA.14 These individuals suffer from a form of anemia that typically does not improve with oral iron treatment, but requires parenteral iron administration.14 The elucidation of these pathways and their association with disease led to the development of pharmacological compounds that increase hepcidin expression, mimic its activity, or decrease ferroportin activity, which can decrease iron absorption and improve iron overload in 1692

primary and secondary forms of hemochromatosis (Figure 1A).15,16 Intriguingly, the same drugs also showed beneficial effects on anemia in animal models of nontransfusion-dependent thalassemia (Figure 1A).13,17,18 In this case, it was observed that these drugs not only decreased iron absorption, but also erythroid iron intake.13,17,18 In thalassemic erythroid progenitor cells, this can reduce the detrimental effects of oxidative stress triggered by the excess of iron and heme not included in normal hemoglobin molecules.13,17,18 This improves the quality and lifespan of red blood cells, and increases hemoglobin levels.13,17,18 The overall effect is to improve ineffective erythropoiesis and the associated iron overload (Figure 1A). So far, most of the compounds identified as leading to haematologica | 2019; 104(9)

Editorials

increased hepcidin expression show a very specific activity (i.e. hepcidin mimetics or ferroportin inhibitors).15,19 In general, these drugs belong to one of four main categories: (i) hepcidin mimetics; (ii) hepcidin inducers; (iii) ferroportin inhibitors; and (iv) erythroferrone inhibitors. TMPRSS6 inhibitors can be defined as hepcidin inducers and/or BMP/SMAD pathway activators. The ideal drug should be administered orally or injected subcutaneously very infrequently, having a long lifespan and prolonged activity. The drug should also show a large spectrum of activity, so that it can limit iron absorption in disorders such as non-transfusion-dependent thalassemia and HFE-related hemochromatosis, but also in conditions in which iron absorption is further increased (e.g., Î˛-thalassemia major), or in which iron absorption needs to be further suppressed to achieve a significant benefit (as in polycythemia vera).15,18,19 The drug should also have no side effects, particularly under conditions of chronic administration. Obviously, low cost of production would also be desirable. Furthermore, and equally important, the drug should have a clear mechanism of action. The compounds described by Liu and colleagues are derivatives of thiazolidinones, a group of versatile drugs which are also being developed for numerous clinical applications, such as anti-tuberculosis, antimicrobial, anti-cancer, anti-inflammatory, and antiviral agents.20 These new compounds increase expression of hepcidin and improve several parameters (related to both iron overload and anemia) in mice affected by primary and secondary forms of hemochromatosis (Figure 1B).1 In particular, in mice affected by hemochromatosis, the compounds described by Liu and colleagues ameliorated abnormal iron parameters, improved iron overload, and induced iron redistribution from the liver to the spleen. In mice affected by non-transfusion-dependent thalassemia, these compounds also ameliorated iron overload. In addition, as ineffective erythropoiesis was also improved, red blood cell production and hemoglobin levels increased (Figure 1B). As described in their article, these novel thiazolidinone derivatives appear to act on hepcidin expression through a variety of mechanisms, such as promoting Smad1/5/8 signaling, repressing Erk1/2 phosphorylation and decreasing Tmprss6 activity (Figure 1B). Additionally, these compounds seemed to target potential erythroid regulators (such as erythroferrone), thereby further contributing to hepcidin upregulation (Figure 1B). However, the target and mechanism of action of these compounds have not been completely elucidated. Given their many effects, there is some concern that these drugs may be relatively unselective and affect additional targets and pathways. This would be even more relevant if these drugs were to become used in a chronic fashion. Future studies should, therefore, focus on determining how these drugs interact with their target and exclude unwanted effects. In summary, these novel compounds are very promis-

haematologica | 2019; 104(9)

ing and expand the armamentarium of drugs that could benefit patients affected by disorders in which increased hepcidin expression is desirable. If proven to be safe, selective, and effective, their use will increase the chance that one or more compounds will reach the clinic, while competition between different drugs will likely diminish costs.

References 1. Liu J, Liu W, Liu Y, et al. New thiazolidinones reduce iron overload in mouse models of hereditary hemochromatosis and Î˛-thalassemia. Haematologica. 2019;104(9):1768-1781. 2. Rivella S. Iron metabolism under conditions of ineffective erythropoiesis in beta-thalassemia. Blood. 2019;133(1):51-58. 3. Muckenthaler MU, Rivella S, Hentze MW, Galy B. A red carpet for ion metabolism. Cell. 2017;168(3):344-361. 4. Rausa M, Pagani A, Nai A, et al. Bmp6 expression in murine liver non parenchymal cells: a mechanism to control their high iron exporter activity and protect hepatocytes from iron overload? PLoS One. 2015;10(4):e0122696. 5. D'Alessio F, Hentze MW, Muckenthaler MU. The hemochromatosis proteins HFE, TfR2, and HJV form a membrane-associated protein complex for hepcidin regulation. J Hepatol. 2012;57(5):1052-1060. 6. Chen H, Choesang T, Li H, et al. Increased hepcidin in transferrintreated thalassemic mice correlates with increased liver BMP2 expression and decreased hepatocyte ERK activation. Haematologica. 2016;101(3):297-308. 7. Kautz L, Jung G, Valore EV, Rivella S, Nemeth E, Ganz T. Identification of erythroferrone as an erythroid regulator of iron metabolism. Nat Genet. 2014;46(7):678-684. 8. Sonnweber T, Nachbaur D, Schroll A, et al. Hypoxia induced downregulation of hepcidin is mediated by platelet derived growth factor BB. Gut. 2014;63(12):1951-1959. 9. Arezes J, Foy N, McHugh K, et al. Erythroferrone inhibits the induction of hepcidin by BMP6. Blood. 2018;132(14):1473-1477. 10. Wahedi M, Wortham AM, Kleven MD, et al. Matriptase-2 suppresses hepcidin expression by cleaving multiple components of the hepcidin induction pathway. J Biol Chem. 2017;292(44):18354-18371. 11. Nai A, Rubio A, Campanella A, et al. Limiting hepatic Bmp-Smad signaling by matriptase-2 is required for erythropoietin-mediated hepcidin suppression in mice. Blood. 2016;127(19):2327-2336. 12. Frydlova J, Rychtarcikova Z, Gurieva I, Vokurka M, Truksa J, Krijt J. Effect of erythropoietin administration on proteins participating in iron homeostasis in Tmprss6-mutated mask mice. PLoS One. 2017;12(10):e0186844. 13. Gardenghi S, Marongiu MF, Ramos P, et al. Ineffective erythropoiesis in beta-thalassemia is characterized by increased iron absorption mediated by down-regulation of hepcidin and up-regulation of ferroportin. Blood. 2007;109(11):5027-5035. 14. Finberg KE, Heeney MM, Campagna DR, et al. Mutations in TMPRSS6 cause iron-refractory iron deficiency anemia (IRIDA). Nat Genet. 2008;40(5):569-571. 15. Casu C, Nemeth E, Rivella S. Hepcidin agonists as therapeutic tools. Blood. 2018;131(16):1790-1794. 16. Guerra A, Musallam KM, Taher AT, Rivella S. Emerging therapies. Hematol Oncol Clin North Am. 2018;32(2):343-352. 17. Casu C, Aghajan M, Oikonomidou PR, Guo S, Monia BP, Rivella S. Combination of Tmprss6- ASO and the iron chelator deferiprone improves erythropoiesis and reduces iron overload in a mouse model of beta-thalassemia intermedia. Haematologica. 2016;101(1):e8-e11. 18. Casu C, Oikonomidou PR, Chen H, et al. Minihepcidin peptides as disease modifiers in mice affected by beta-thalassemia and polycythemia vera. Blood. 2016;128(2):265-276. 19. Oikonomidou PR, Casu C, Rivella S. New strategies to target iron metabolism for the treatment of beta thalassemia. Ann N Y Acad Sci. 2016;1368(1):162-168. 20. Kaur Manjal S, Kaur R, Bhatia R, et al. Synthetic and medicinal perspective of thiazolidinones: a review. Bioorg Chem. 2017;75:406-423.

1693

Editorials

Ubiquitination is not omnipresent in myeloid leukemia Ramesh C. Nayak1 and Jose A. Cancelas1,2 1

Division of Experimental Hematology and Cancer Biology, Cincinnati Children’s Hospital Medical Center and 2Hoxworth Blood Center, University of Cincinnati Academic Health Center, Cincinnati, OH, USA E-mail: JOSE A. CANCELAS - jose.cancelas@uc.edu / jose.cancelas@cchmc.org doi:10.3324/haematol.2019.224162

hronic myelogenous leukemia (CML) is a clonal biphasic hematopoietic disorder most frequently caused by the expression of the BCR-ABL fusion protein. The expression of BCR-ABL fusion protein with constitutive and elevated tyrosine kinase activity is sufficient to induce transformation of hematopoietic stem cells (HSC) and the development of CML.1 Despite the introduction of tyrosine kinase inhibitors (TKI), the disease may progress from a manageable chronic phase to a clinically challenging blast crisis phase with a poor prognosis,2 in which myeloid or lymphoid blasts fail to differentiate. Progression of BCR-ABL-positive (+) leukemia from the chronic phase to the acute blast crisis phase is accompanied by increased BCR-ABL expression and genomic instability leading to the acquisition of secondary genetic lesions including +8, +Ph, +19, mutations in P53, Runx1, ASXL1, WT1TET2, IDH1, deletion of INK4A/ARF, and/or PAX5, IKZF1, EBF1 resulting in myeloid or B-lymphoid blast crises.3,4 However, our understanding of the mechanisms of transformation in blastic crisis remains incomplete. In this issue of the Journal, Magistroni et al.5 identified the presence of mutations in three different amino acids (D144, I33, M34) impairing the expression and/or activity of one of the alleles of the ubiquitin conjugating enzyme E2A (UBE2A, also called RAD6A) in the blastic phase, but not in the chronic phase, of two out of ten CML patients. Analysis of an unmatched, larger cohort of 24 blast crisis, 41 chronic phase, 40 acute myeloid leukemia (AML), and 38 BCR-ABL-negative CML specimens confirmed the presence of these mutations in 16.7% of blastic phase CML patients but not in any of the other groups analyzed. Mechanistically, the silencing of UBE2A or overexpression of one of the UBE2A mutants (I33M) in BCRABL+ leukemic cells results in myeloid differentiation blockade in vitro with upregulation of ITGB4, RDH10 and CLEC11A, and downregulation of CFS3R/CSF3R and RAP1GAP. The fact that UBE2A mutations were exclusively found in the blast crisis CML patients, and these mutations control the process of differentiation arrest indicates that mutant UBE2A is a potential target for intervention in blastic phase CML.

tination of target proteins through their cognate E3 ubiquitin ligases belonging to three different families (RING, HERCT, RING-between-RING or RBR type E3).7 The ubiquitin conjugating enzymes including UBE2N (UBC13) and UBE2C are over-expressed in a myriad of tumors such as breast, pancreas, colon, prostate, lymphoma, and ovarian carcinomas.8 Higher expression of UBE2A is associated with poor prognosis of hepatocellular cancer.9 In leukemia, bone marrow (BM) cells from pediatric acute lymphoblastic patients show higher levels of UBE2Q2 expression in comparison to normal counterparts.10 Ubiquitin conjugating enzyme E2E1 (UBE2E1) expression is adversely correlated with AML survival.11 However, in this report, the inactivating mutation of UBE2A seems to facilitate CML progression, and therefore UBE2A seems to act as a tumor suppressor. Based on our understanding of mechanisms controlled by UBE2A, four different signaling pathways may be involved in blast crisis transformation (Figure 1).

Ubiquitin conjugating enzymes in inflammation and cancer

The transformation to blast crisis phase is associated with selection of clones with high BCR-ABL1 expression. However, the mechanism of enhanced BCR-ABL1 expression remains poorly understood. It has been shown that arseniate, a curative agent in acute promyelocytic leukemia, induced cell apoptosis and degradation of BCRABL in CML cells. The ubiquitination and degradation of BCR-ABL was mediated by c-CBL, a RING-type E3 ligase.14 Although speculative, it is possible that c-CBL acts as a cognate E3 ligase for UBE2A for the ubiquitina-

UBE2A is an E2 ubiquitin conjugating enzyme. Ubiquitination is a highly conserved post-translational modification process affecting the proteasome-mediated degradation as well as activity of target proteins. The process occurs in three sequential steps mediated by ubiquitin-activating enzyme (E1), ubiquitin conjugating enzyme (E2), and ubiquitin ligase (E3).6 In humans, nearly 40 E2 ubiquitin conjugating enzymes regulate ubiqui1694

Inflammatory myeloid differentiation is mediated by ubiquitination First, the abundance of pro-inflammatory cytokines including interleukin-1 (IL-1) and tumor necrosis factor alpha (TNF-a) in the leukemic microenvironment regulates myeloid differentiation through the activation of NFκB and MAPK signaling pathways.12 These activities are mediated by TNF-a receptor-associated factor (TRAF) family E3 ligases.13 The activation of IL-1 and TNF-a receptor induces the recruitment of MyD88, IL-1 receptor-associated kinase (IRAK4, IRAK2) to the myddosome complex resulting in the activation of TRAF6. UBE2A might act as upstream ubiquitin conjugating enzymes for TRAF6 E3 ligase in CML myeloid blasts, and loss-of-function of UBE2A may attenuate the TRAF E3 ligase-mediated activation of NFκB and MAPK signaling pathways, leading to the impaired myeloid differentiation (Figure 1, signaling path A). This is a signaling mechanism involved in myeloid differentiation with unclear significance in the context of UBE2A loss-of-function mutations.

Ubiquitination regulates BCR-ABL and MYC expression in myeloid leukemia

haematologica | 2019; 104(9)

Editorials

tion and subsequent degradation of BCR-ABL. Furthermore, expression of transcriptional factor MYC plays a critical role in the proliferation and self-renewal of leukemic stem cells. Reavie et al. demonstrated that the E3 ubiquitin ligase FBW7 is required for the survival and maintenance of BCR-ABL+ leukemia initiating cells (LIC) by modifying the expression of MYC through FBW7mediated ubiquitination and degradation.15 Deletion of Fbw7 leads to c-Myc overexpression, p53-dependent LICspecific apoptosis, and the eventual inhibition of tumor progression. A decrease in either c-Myc protein levels or attenuation of the p53 response rescues LIC activity and disease progression. UBE2A acts as E2 conjugating enzyme for FBW7, and mutations in UBE2A attenuate FBW7 activity and maintain the basal expression level of

MYC required for survival and propagation of leukemic blast (Figure 1, signaling path B).

UBE2A activity maintains genomic integrity Myeloid blastic transformation in CML requires genomic instability which may originate from imatinibrefractory CML stem cells.16 Genomic instability is mediated by loss-of-function of DNA repair process. The UBE2A described in this report is the human homolog of yeast Rad6, and has been demonstrated to play a critical role in DNA repair and genome integrity.17 UBE2A and UBE2B regulate DNA damage through post-translational modification of proliferating cell nuclear antigen (PCNA).18 The ubiquitination of PCNA at Lys 164 in response to genotoxic stress recruits DNA polymerase

Figure 1. Schematic representation of possible UBE2A-mediated mechanisms controlling myeloid blastic transformation in BCR-ABL leukemias. Possible targets of UBE2A relevant to leukemic myeloid transformation. (A and B) Cytosolic functions. UBE2A is an E2 ubiquitin ligase important in emergency myelopoiesis induced by inflammatory cytokines, abundant in the leukemic microenvironment, through the TRAF/TRIF E3 ligases, regulators of the transcriptional factor NFkB. Loss-offunction of UBE2A may associate with impaired myeloid differentiation. (B) UBE2A regulates the activity of E3 ligases c-CBL and FBW7, which are tumor suppressors with known activity to induce degradation of BCR-ABL and MYC, whose expression in turn is required for leukemic acceleration. Question marks denote that these pathways of activity of UBE2A are speculative and not supported by direct experimental designs. (C and D) Nuclear functions. (C) UBE2A is a well known regulator of DNA repair through its cognate E3 ligase RAD18, which monoubiquitinates the proliferating cell nuclear antigen (PCNA), a modification that recruits translesion DNA polymerases to stalled replication forks. (D) Active UBE2A (phosphorylated by CDK9) regulates H2Bmonoubiquitination through recruitment of the E3 ligase RNF20/40, a major step in regulation of RNA polymerase II and gene transcription.

haematologica | 2019; 104(9)

1695

Editorials

and activates translesion synthesis DNA repair pathway.19 Furthermore, cell cycle dependent kinase-9 (CDK9) regulates UBE2A activity by phosphorylating at serine 120.20 UBE2A regulates the ubiquitination of histone H2B and proliferating cell nuclear antigen (PCNA) through the cognate E3 ubiquitin ligase RNF20/40 and RAD18, respectively. In addition to its role in transcriptional elongation, histone H2B K120 monoubiquitination plays a crucial role in DNA double strand break (DSB) repairs.21 Both these processes describe the role of UBE2A in DNA repair and maintenance of genome integrity. The loss-of-function mutations of UBE2A in advanced phase CML patients may be associated with impaired ubiquitination of H2B and PCNA, and hence increased genome instability resulting in the acquisition of additional mutations (Figure 1, signaling paths C and D). The work by Magistroni et al.5 focuses on the latter signaling paths as possibly being at the root of the myeloid transformation. While the mechanisms that control the blastic transformation of CML by UBE2A mutations remain unclear, mutation studies like that of Magistroni et al. do generate hypotheses that should be tested in further studies into BCR-ABL leukemia initiation and propagation.

References 1. Daley GQ, Van Etten RA, Baltimore D. Induction of chronic myelogenous leukemia in mice by the P210bcr/abl gene of the Philadelphia chromosome. Science. 1990;247(4944):824-830. 2. Gambacorti-Passerini CB, Gunby RH, Piazza R, Galietta A, Rostagno R, Scapozza L. Molecular mechanisms of resistance to imatinib in Philadelphia-chromosome-positive leukaemias. Lancet Oncol. 2003;4(2):75-85. 3. Perrotti D, Jamieson C, Goldman J, Skorski T. Chronic myeloid leukemia: mechanisms of blastic transformation. J Clin Invest. 2010;120(7):2254-2264. 4. Mullighan CG, Goorha S, Radtke I, et al. Genome-wide analysis of genetic alterations in acute lymphoblastic leukaemia. Nature. 2007;446(7137):758-764. 5. Magistroni V, Mauri M, D'Aliberti D, et al. De novo UBE2A mutations are recurrently acquired during chronic myeloid leukemia progression and interfere with myeloid differentiation pathways.

Haematologica. 2019;104(9):1789-1797. 6. Swatek KN, Komander D. Ubiquitin modifications. Cell Res. 2016;26(4):399-422. 7. Deshaies RJ, Joazeiro CA. RING domain E3 ubiquitin ligases. Annu Rev Biochem. 2009;78:399-434. 8. Gallo LH, Ko J, Donoghue DJ. The importance of regulatory ubiquitination in cancer and metastasis. Cell Cycle. 2017;16(7):634-648. 9. Shen JD, Fu SZ, Ju LL, et al. High expression of ubiquitin-conjugating enzyme E2A predicts poor prognosis in hepatocellular carcinoma. Oncol Lett. 2018;15(5):7362-7368. 10. Seghatoleslam A, Monabati A, Bozorg-Ghalati F, et al. Expression of UBE2Q2, a putative member of the ubiquitin-conjugating enzyme family in pediatric acute lymphoblastic leukemia. Arch Iran Med. 2012;15(6):352-355. 11. Luo H, Qin Y, Reu F, et al. Microarray-based analysis and clinical validation identify ubiquitin-conjugating enzyme E2E1 (UBE2E1) as a prognostic factor in acute myeloid leukemia. J Hematol Oncol. 2016;9(1):125. 12. Anand M, Chodda SK, Parikh PM, Nadkarni JS. Abnormal levels of proinflammatory cytokines in serum and monocyte cultures from patients with chronic myeloid leukemia in different stages, and their role in prognosis. Hematol Oncol. 1998;16(4):143-154. 13. Barreyro L, Chlon TM, Starczynowski DT. Chronic immune response dysregulation in MDS pathogenesis. Blood. 2018;132(15): 1553-1560. 14. Mao JH, Sun XY, Liu JX, et al. As4S4 targets RING-type E3 ligase cCBL to induce degradation of BCR-ABL in chronic myelogenous leukemia. Proc Natl Acad Sci U S A. 2010;107(50):21683-21688. 15. Reavie L, Buckley SM, Loizou E, et al. Regulation of c-Myc ubiquitination controls chronic myelogenous leukemia initiation and progression. Cancer Cell. 2013;23(3):362-375. 16. Bolton-Gillespie E, Schemionek M, Klein HU, et al. Genomic instability may originate from imatinib-refractory chronic myeloid leukemia stem cells. Blood. 2013;121(20):4175-4183. 17. Shekhar MP, Lyakhovich A, Visscher DW, Heng H, Kondrat N. Rad6 overexpression induces multinucleation, centrosome amplification, abnormal mitosis, aneuploidy, and transformation. Cancer Res. 2002;62(7):2115-2124. 18. Garg P, Burgers PM. Ubiquitinated proliferating cell nuclear antigen activates translesion DNA polymerases eta and REV1. Proc Natl Acad Sci U S A. 2005;102(51):18361-18366. 19. Hoege C, Pfander B, Moldovan GL, Pyrowolakis G, Jentsch S. RAD6-dependent DNA repair is linked to modification of PCNA by ubiquitin and SUMO. Nature. 2002;419(51):135-141. 20. Shchebet A, Karpiuk O, Kremmer E, Eick D, Johnsen SA. Phosphorylation by cyclin-dependent kinase-9 controls ubiquitinconjugating enzyme-2A function. Cell Cycle. 2012;11(11):21222127. 21. Nakamura K, et al. Regulation of homologous recombination by RNF20-dependent H2B ubiquitination. Mol Cell. 2011;41(5):515528.

Six-packed antibodies punch better Christoph Rader1 and Adrian Wiestner2 1

Department of Immunology and Microbiology, The Scripps Research Institute, Jupiter, FL and 2Hematology Branch, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, MD, USA E-mail: CHRISTOPH RADER - crader@scripps.edu ADRIAN WIESTNER - wiestnera@mail.nih.gov doi:10.3324/haematol.2019.224196

n this issue of the Journal, Oostindie et al. investigate CD37-specific monoclonal antibodies (mAb) engineered to undergo hexamerization.1 Efficient hexamer formation is induced by a single amino acid substitution, E430G, in the IgG1 constant domain previously described by the same group.2 The modification potentiates complement-dependent cytotoxicity (CDC) against chronic lymphocytic leukemia (CLL) cells in vitro. Next, the authors show that combinations of hexamerizationenhanced mAb against CD20 and CD37 provide syner1696

gistic activity. Intriguingly, the CD20- and CD37-targeting mAb formed mixed hexameric complexes on the cell surface with increased anti-tumor activity. The anti-CD20 mAb rituximab is a critical component of treatment regimens for many B-cell malignancies.3 In combination with chemotherapy, rituximab has been shown to increase response rates, response duration, and overall survival. Single-agent rituximab is quite commonly used in follicular lymphoma and as maintenance therapy in several types of B-cell non-Hodgkin lymphoma (Bhaematologica | 2019; 104(9)

Editorials

NHL), including CLL. Compared to other B-NHL, CLL cells have a relatively lower expression of CD20, and single-agent rituximab has limited activity in CLL. Few studies have investigated the combination of two mAb. The combination of rituximab with the anti-CD52 targeting mAb alemtuzumab yielded a higher rate of complete responses in CLL than had historically been seen with rituximab alone.4 However, the manufacturer withdrew alemtuzumab for the treatment of CLL. Like CD20, the tetraspanin CD37 is an integral membrane protein abundantly expressed on B cells but not on plasma cells or hematopoietic stem cells.5 T cells, natural killer (NK) cells, granulocytes, and monocytes express low levels of CD37. Tetraspanins are central to membrane organization and play important roles in cell migra-

tion and adhesion.6 CD37 has been found to co-localize with integrin a4Î˛1 on B cells and to contribute to cell adhesion and the transduction of survival signals.7 Several anti-CD37 antibodies are undergoing clinical investigation in B-cell malignancies.5 Otlertuzumab (also called TRU-016), a single-chain variable fragment (scFv) against CD37 linked to the IgG1 Fc fragment, induces apoptosis in CLL cells and mediates antibody-dependent cellular cytotoxicity (ADCC) but not CDC. Otlertuzumab has been shown to have single-agent activity in CLL,8 and in combination with bendamustine increased the response rate and prolonged progressionfree survival over single-agent bendamustine.9 BI 836826, a chimeric mouse-human mAb with Fc modifications to increase affinity to FcÎłRIIIa effectively mediates ADCC

Figure 1. Hexamerization and hetero-hexamerization of CD20- and CD37-targeting mAb on the cell surface. (A) Shown are CD20- targeting (blue) and CD37-targeting (orange) mAb in IgG1 format. Monomeric in solution, they form hexamers upon cell surface antigen binding. This natural hexamerization of wild-type IgG1 is enhanced by substituting a glutamic acid residue in the IgG1 Fc fragment with a glycine residue (E430G). Mixing CD20- and CD37-targeting mAb leads to the formation of hetero-hexamers in 3:3 (shown here), 4:2, 2:4, 5:1, or 1:5 compositions. (B) Oostindie et al.1 show a gradual increase in complement-dependent cytotoxicity (CDC) mediated by wild-type CD37-targeting IgG1 (left) to wild-type CD20-targeting IgG1 (second from left) to CD37-targeting IgG1 with E430G mutation (center) to CD20-targeting IgG1 with E430G mutation (second from right) to mixed CD20- and CD37-targeting IgG1 with E430G mutation (right). This increase in CDC correlates with the density of C1q docking sites. For simplification, only hetero-hexamers in various 3:3 compositions are shown.

haematologica | 2019; 104(9)

1697

Editorials

and also induces apoptosis of CLL cells. In a phase I doseescalation study, BI 836826 was well-tolerated up to doses of 400 mg and had a similar adverse event profile as other Fc-modified antibodies.10 The objective response rate was 61.5% in patients treated at doses â&#x2030;Ľ200 mg. Two antibody-drug conjugates and a radioimmunoconjugate targeting CD37 are also undergoing clinical investigation.5 Currently, 36 antibody-based cancer therapies approved by the US Food and Drug Administration (FDA), including 4 biosimilars, are on the market. The majority of these treatments are mAb in IgG1 format that mediate tumor cell killing on their own or in combination with chemotherapy. A key challenge has been the identification of suitable targets for therapeutic mAb as tumorspecific antigens are rare, and tumor-associated antigens are often expressed on essential healthy cells, lowering the therapeutic index. By contrast, lineage-specific antigens that are expressed on non-essential healthy cells have emerged as preferred targets of therapeutic mAb. A prime example is CD20, which is expressed on healthy and malignant B cells, and targeted by rituximab (FDA approved in 1997), its biosimilar rituximab-abbs (in 2018), ofatumumab (in 2009), and obinutuzumab (in 2013) for treatment of B-cell malignancies. The same applies to other cell surface antigens, such as CD19, that are restricted to the dispensable B-cell lineage of the hematopoietic system. The mechanism of action (MOA) by which mAb eradicate tumor cells include the induction of apoptosis by interfering with receptor/ligand interactions at the cell surface or by recruiting components of the innate immune system, such as plasma proteins in CDC, NK cells in ADCC, and macrophages in antibody-dependent cellular phagocytosis (ADCP).11 All three principle mechanisms of innate immune system recruitment, collectively known as effector functions, involve the Fc fragment of IgG1, a homodimer comprising the hinge and the second (CH2) and third (CH3) constant domains of the heavy chain. To mediate CDC, ADCC, and ADCP, the Fc fragment interacts with complement protein C1q and FcÎłRIIIa and FcÎłRIIa receptors, respectively. It also mediates prolonged circulatory half-life through neonatal Fc receptor (FcRn) recycling. All of these mechanisms can be fine tuned by subjecting the Fc fragment to protein or carbohydrate engineering.12 In fact, several of the FDAapproved mAb for cancer therapy have engineered Fc fragments. Hexabodies constitute a new class of Fc fragment-engineered therapeutic antibodies.13,14 A single amino acid substitution in CH3, E430G, enhances the formation of IgG1 hexamers upon cell surface antigen binding (Figure 1A). As such, hexamerization, which was first discovered for membrane-bound wild-type IgG1,2 facilitates the docking of the hexavalent complement protein C1q initiating CDC. Indeed, previous studies revealed that CD20targeting IgG1 with the E430G mutation mediate significantly enhanced CDC compared to the parental mAb.13,14 The current study by Oostindie et al.1 makes the same case for a CD37-targeting IgG1. In addition, combining hexameric (E430G) CD37-targeting IgG1 with one of the FDA-approved CD20-targeting IgG1 (rituximab, ofatumumab, or obinutuzumab) had a synergistic effect in terms of malignant B-cell lysis by CDC in vitro. 1698

Intriguingly, the authors provide evidence that mixing CD20- and CD37-targeting IgG1 with E430G mutation leads to the formation of hetero-hexamers that are more potent in mediating CDC than the corresponding homohexamers on their own or in combination (Figure 1B). This finding is exciting as it suggests that two mAb that target two different cell surface antigens may form bispecific hetero-hexamers in the membrane, effectively leading to target clustering and an increase in the density of C1q docking sites. It also sheds a light on a possible concerted MOA of polyclonal antibodies which might form hetero-hexamers if they target different cell surface antigens or different epitopes of the same cell surface antigen. Collectively, the study makes a strong case for investigating multispecific and multiparatopic biclonal, oligoclonal, and polyclonal antibodies for enhancing CDC compared to their parental mAb. Finding co-operative target combinations, such as CD20 and CD37 in the current study, that enable hetero-hexamer formation in the presence or absence of hexamerization-enhancing mutations is a key challenge in applying this concept to other hematologic malignancies and solid tumors. In this context, hexameric monoclonal and hetero-hexameric biclonal antibodies should also be tested for enhancing other effector functions in addition to CDC. While research into bispecific antibodies has accelerated, with a huge increase in the number of related clinical trials that are now ongoing,15 polyclonal antibodies16 may well be the next wave of antibody-based cancer therapy. Hetero-hexamerization in the membrane is a possible MOA of polyclonal antibodies in IgG1 format, providing an incentive to investigate their therapeutic utility with and without hexamerization-inducing mutations. In summary, Oostindie et al.1 make a compelling case for further exploration of hexamer-forming antibodies and the combination of two, or possibly even more, targeting mAb. The advantages of antibody combinations might include not only increased cytotoxic activity, as described here, but possibly also better tumor-specific targeting and mitigation of tumor escape through antigen loss or target internalization.17 However, several aspects of this promising technology need further exploration. How effective is hexamer formation in vivo and what kind of hetero-hexamers might be formed, especially in tissue sites? The current study is limited to in vitro studies with CLL cells in suspension. It is not immediately clear how these observations will translate to in vivo settings. Furthermore, hetero-hexamers may form in different ratios, some containing equal ratios of antibodies, while in others one antibody may dominate. Will there be an optimal ratio and if so, could a desired composition be engineered into the antibody backbone? Translation of this promising approach into clinical trials may well constitute the next breakthrough in antibody therapy of B-cell malignancies. A first clinical trial with mAb engineered to facilitate hexamerization is ongoing in solid tumors (clinicaltrials.gov identifier: NCT03576131). GEN1029 (also called HexaBodyDR5/DR5) consists of a mixture of two mAb that bind to different epitopes on DR5 and activate this death receptor to induce apoptosis. Results from this and other studies of hexamerization-enhanced mAb and mAb combinations are eagerly awaited. haematologica | 2019; 104(9)

Editorials

References 1. Oostindie SC, van der Horst HJ, Lindorfer MA, et al. CD20 and CD37 antibodies synergize to activate complement by Fc-mediated clustering. Haematologica. 2019;104(9):1841-1852. 2. Diebolder CA, Beurskens FJ, de Jong RN, et al. Complement is activated by IgG hexamers assembled at the cell surface. Science. 2014;343(6176):1260-1263. 3. Salles G, Barrett M, Foa R, et al. Rituximab in B-cell hematologic malignancies: a review of 20 years of clinical experience. Adv Ther. 2017;34(10):2232-2273. 4. Zent CS, Victoria Wang X, Ketterling RP, et al. A phase II randomized trial comparing standard and low dose rituximab combined with alemtuzumab as initial treatment of progressive chronic lymphocytic leukemia in older patients: a trial of the ECOG-ACRIN cancer research group (E1908). Am J Hematol. 2016;91(3):308-312. 5. Witkowska M, Smolewski P, Robak T. Investigational therapies targeting CD37 for the treatment of B-cell lymphoid malignancies. Expert Opin Investig Drugs. 2018;27(2):171-177. 6. Yeung L, Hickey MJ, Wright MD. The many and varied roles of tetraspanins in immune cell recruitment and migration. Front Immunol. 2018;9:1644. 7. van Spriel AB, de Keijzer S, van der Schaaf A, et al. The tetraspanin CD37 orchestrates the alpha(4)beta(1) integrin-Akt signaling axis and supports long-lived plasma cell survival. Sci Signal. 2012;5(250):ra82. 8. Byrd JC, Pagel JM, Awan FT, et al. A phase 1 study evaluating the safety and tolerability of otlertuzumab, an anti-CD37 mono-specific ADAPTIR therapeutic protein in chronic lymphocytic leukemia. Blood. 2014;123(9):1302-1308.

9. Robak T, Hellmann A, Kloczko J, et al. Randomized phase 2 study of otlertuzumab and bendamustine versus bendamustine in patients with relapsed chronic lymphocytic leukaemia. Br J Haematol. 2017;176(4):618-628. 10. Stilgenbauer S, Aurran Schleinitz T, Eichhorst B, et al. Phase 1 firstin-human trial of the anti-CD37 antibody BI 836826 in relapsed/refractory chronic lymphocytic leukemia. Leukemia. 2019 May 14. [Epub ahead of print] 11. Weiner LM, Surana R, Wang S. Monoclonal antibodies: versatile platforms for cancer immunotherapy. Nat Rev Immunol. 2010;10(5):317-327. 12. Saxena A, Wu D. Advances in therapeutic Fc engineering - modulation of IgG-associated effector functions and serum half-life. Front Immunol. 2016;7:580. 13. de Jong RN, Beurskens FJ, Verploegen S, et al. A novel platform for the potentiation of therapeutic antibodies based on antigen-dependent formation of IgG hexamers at the cell surface. PLoS Biol. 2016;14(1):e1002344. 14. Cook EM, Lindorfer MA, van der Horst H, et al. Antibodies that efficiently form hexamers upon antigen binding can induce complement-dependent ctotoxicity under complement-limiting conditions. J Immunol. 2016;197(5):1762-1775. 15. Labrijn AF, Janmaat ML, Reichert JM, Parren P. Bispecific antibodies: a mechanistic review of the pipeline. Nat Rev Drug Discov. 2019 Jun 7. [Epub ahead of print] 16. Haurum JS. Recombinant polyclonal antibodies: the next generation of antibody therapeutics? Drug Discov Today. 2006;11(13-14):655660. 17. Taylor RP, Lindorfer MA. Fcgamma-receptor-mediated trogocytosis impacts mAb-based therapies: historical precedence and recent developments. Blood. 2015;125(5):762-766.

The secret afterlife of platelets Nicholas A. Arce1,2 and Renhao Li1 1

Aflac Cancer and Blood Disorders Center, Childrenâ&#x20AC;&#x2122;s Healthcare of Atlanta, Department of Pediatrics, Emory University School of Medicine and 2Graduate Program of Molecular and Systems Pharmacology, Graduate Division of Biological and Biomedical Sciences, Emory University, Atlanta, GA, USA E-mail: RENHAO LI - renhao.li@emory.edu doi:10.3324/haematol.2019.224170

latelets express a wide variety of receptors and signaling molecules that enable responses to diverse physiological and pathological stimulants. For instance, in normal hemostasis, exposure of subendothelial collagen may elicit platelet activation at the site of injury via glycoprotein (GP)VI, integrin a2Î˛1, and, through plasma von Willebrand factor, the GPIb-IX-V complex. Moreover, GPIb-IX-V in tandem with protease-activated receptors mediate thrombin-induced platelet signaling and activation. GPIba serves as a receptor for low concentrations of thrombin, transmitting a mechanosensory signal to mediate calcium-dependent 14-3-3 signaling while GPIb-IXâ&#x20AC;&#x201C; dependent Rac1/LIMK1 signaling is modulated by protease-activated receptors.1,2 Upon activation, platelets aggregate and form clots that are interwoven with fibrin strands. Over the last several decades, much of the research effort has been focused on how platelets are rapidly activated by various agonists via their respective receptors and how activating, and sometimes inhibitory, signals amplify and propagate in the platelet. In most of these studies, the investigation ends at the cessation of blood flow, the formation of the clot, and/or the appearance of molecular signs that are well associated with platelet activation. A few minutes following platelet activation and aggregation, the blood clot contracts. In studies of clot contraction, the investigation often ends at the shrinkage of the platelet clot.3 However, little is known about the platelets in the clot haematologica | 2019; 104(9)

following the contraction of the platelet/fibrin clot. In other words, after the formation of a stable blood clot, where do platelets go? A study by Kim et al., published in this issue of Haematologica, demonstrates that after activation and contraction, thrombin-stimulated platelets break up into membrane particles, in a process termed platelet fragmentation.4 Thrombin is a major nexus between coagulation and platelet activation, as it generates fibrin to form a crosslinked fibrin plug and concurrently activates aforementioned receptors on the platelet surface.5 Platelet vesiculation and/or microparticle formation has been previously observed in response to thrombin and thrombin receptor activating peptide.6-8 The role that these platelet fragments play in hemostasis or platelet clearance has yet to be elucidated. In this new study, interestingly, Kim et al. observed a bimodal distribution of platelet fragments, the size of which can be attributed to the origin of the fragment. Filopodia as well as the main platelet body are two sources of platelet fragmentation, as smaller fragments were generated by filopodia, and larger fragments were generated from the cell body. Thus, it appears that platelet breakdown in response to thrombin stimulation is a regulated process of drastic morphological changes, platelet fragmentation, loss of function, and metabolic exhaustion. Platelet fragmentation may be a relatively newly discovered platelet behavior, adding to the ever-growing list of what 1699

Editorials

Figure 1. An illustration of thrombin-induced effects on platelets over time. Thrombin produced through coagulation binds to and activates platelet receptors, leading to platelet activation, degranulation, and filopodia formation. At a certain time after the hemostatic plug has contracted, platelets break down into membranebound fragments of defined sizes through a distinct mechanism.

happens to platelets after activation, and how they may play a role in hemostasis or clearance. This paradigm shift may help to elucidate novel mechanisms of platelet behavior and clarify the functional roles of such fragments of previously activated platelets. Using transmission and scanning electron microscopy, Kim et al. observed that after 15 min of thrombin treatment, platelets broke up into separate membrane-bound particulates that contained granules, mitochondria, and vacuoles. Remarkably, this fragmentation was not seen after exposure to other platelet agonists such as collagen or ADP, although the platelets exhibited morphological changes associated with activation such as filopodia formation and spreading. This finding suggests that platelet activation creates agonist-specific behavior, and only when thrombin is being generated can platelets start to break up into fragments. As actin is necessary for cytoskeletal rearrangement and filopodia extension,9 the authors investigated localization of actin after exposure to thrombin. After fragmentation, actin was retained in the particulates, and gradually disappeared as the fragments became smaller. It is also worthy of note that intracellular levels of calcium correlated with fragmentation, such that calcium levels dropped as the platelets disintegrated. Mitochondrial function also decreased as the platelets fragmented. Mitochondria appeared to translocate to the periphery of the cell, or even escape to the extracellular space. Clot contraction force plateaued and the generation of reactive oxygen species coincided with the initiation of platelet fragmentation, suggesting that clot contraction is stopped by the loss of cellular energy in the form of ATP due to mitochondrial dysfunction. Actin-myosin is an ATP-driven motor, and these results support the previously seen loss of the actin-regulated cytoskeleton. The authors noted that platelet fragmentation is different from typical necrosis, as fragmentation seemed to be a regulated process that did not entail cellular rupture as the membranes remained intact. To assess whether these platelet fragments were due to apoptotic signaling, caspase activity was assessed. Surprisingly, platelets exposed to up 1700

to 5 U/mL thrombin did not appear to activate caspases. If this is not apoptosis, the question is what is responsible for platelet fragmentation? The authors identified calpain, a cysteine protease that is believed to recognize tertiary structure as a cleavage site instead of sequence-specific activity,10 as one enzyme responsible for these processes. Interestingly, maximal calpain activity coincided with the initiation of fragmentation and functional mitochondrial loss. ALLN, a calpain inhibitor, was able to delay thrombininduced fragmentation and inhibit mitochondrial loss. However, ALLN was unable to inhibit calpain cleavage products completely. Also, it is clear that inhibition of calpains alone is not enough to prevent platelet fragmentation but the observations suggest that proteases are vital for the fragmentation process to occur. While Kim et al. provided an elegant in vitro characterization and outlined the mechanism of platelet fragmentation, the biological significance of this process awaits further elucidation. For instance, do these platelet fragments play a role in the breakdown of platelet-rich clots, and can aberrant fragmentation play a role in thrombosis? Moreover, it is worth noting the time delay in fragmentation upon treatment with thrombin and after platelet contraction, which may be significant to its function. While the GPIb-IX-V receptor complex does not typically lead to a fast and strong intracellular signal, protease-activated receptors, like most G protein-coupled receptors, can rapidly induce full activation of platelets. Thus, what is the mechanism in the platelet that causes fragmentation to proceed only after platelet activation and contraction events have run their courses? The authors demonstrated that force and time are likely important factors in the process of fragmentation (Figure 1). It would be extremely interesting to understand how signaling and cytoskeletal proteins in the platelet respond to these forces and temporal factors. Furthermore, perhaps there is a balance between traditional apoptotic pathways and fragmentation during exposure to a combination of agonists. When would a platelet undergo apoptosis rather than fragmentation in response to multiple agonists? Finally, it remains to be addressed how platelet fraghaematologica | 2019; 104(9)

Editorials

mentation relates to platelet clearance, or what percentage of activated platelets undergo fragmentation in vivo. It is widely accepted that activated platelets are quickly cleared from the body, but the actual molecular mechanism has been elusive. Recent work on platelet clearance has focused on investigating how platelets expose a â&#x20AC;&#x2DC;clear-meâ&#x20AC;&#x2122; signal, perhaps through desialylation of platelet surface proteins.11 The occurrence of platelet fragmentation following activation has raised the possibility that in some cases the reduction in platelet counts, which has been uniformly used as the indicator of platelet clearance, may be attributed to some extent to platelet fragmentation. It remains to be seen whether certain receptors responsible for platelet clearance can also recognize and clear platelet fragments. If this fragmentation can be observed and tracked in vivo, perhaps the question of where fragmented platelets go after activation can be answered. A recent publication by Tomaiuolo et al.12 included high resolution images of hemostatic plugs in response to a puncture in the jugular vein. What can be gleaned from these images is the notable presence of small platelet fragments in both the intravascular and extravascular boundaries of the injury site. Determining the roles that these fragments play in hemostatic plug formation and/or thrombus formation would be crucial to a complete understanding of in vivo platelet plug formation. Potentially, this mechanism could also be a pharmacological target to reduce thrombus formation or aid in thrombolysis in pathological conditions. This exciting finding may point to a novel mechanism of platelet behavior and has major implications for thrombus dissolution and platelet clearance in general.

haematologica | 2019; 104(9)

Acknowledgments This work was supported in part by National Institutes of Health grant HL082808.

References 1. Dormann D, Clemetson Kj, Kehrel Be. The GPIb thrombin-binding site is essential for thrombin-induced platelet procoagulant activity. Blood. 2000;96(7):2469-2478. 2. Estevez B, Kim K, Delaney Mk, et al. Signaling-mediated cooperativity between glycoprotein Ib-IX and protease-activated receptors in thrombin-induced platelet activation. Blood. 2016;127(5):626-636. 3. Tutwiler V, Litvinov Ri, Lozhkin Ap, et al. Kinetics and mechanics of clot contraction are governed by the molecular and cellular composition of the blood. Blood. 2016;127(1):149-159. 4. Kim OV, Nevzorova TA, Mordakhanova ER, et al. Fatal dysfunction and disintegration of thrombin-stimulated platelets. Haematologica. 2019;104(9):1866-1878. 5. Posma JJ, Posthuma JJ, Spronk HM. Coagulation and non-coagulation effects of thrombin. J Thromb Haemost. 2016;14(10):1908-1916. 6. Bode AP, Orton SM, Frye MJ, Udis BJ. Vesiculation of platelets during in vitro aging. Blood. 1991;77(4):887-895. 7. De Paoli SH, Tegegn TZ, Elhelu OK, et al. Dissecting the biochemical architecture and morphological release pathways of the human platelet extracellular vesiculome. Cell Mol Life Sci. 2018;75(20):3781-3801. 8. Brisson AR, Tan S, Linares R, Gounou C, Arraud N. Extracellular vesicles from activated platelets: a semiquantitative cryo-electron microscopy and immuno-gold labeling study. Platelets. 2017;28(3):263271. 9. Fox JE, Boyles JK, Berndt MC, Steffen PK, Anderson LK. Identification of a membrane skeleton in platelets. J Cell Biol. 1988;106(5):1525-1538. 10. Sorimachi H, Mamitsuka H, Ono Y. Understanding the substrate specificity of conventional calpains. Biol Chem. 2012;393(9):853-871. 11. Quach ME, Chen W, Li R. Mechanisms of platelet clearance and translation to improve platelet storage. Blood. 2018;131(14):1512-1521. 12. Tomaiuolo M, Matzko CN, Poventud-Fuentes I, Weisel JW, Brass LF, Stalker TJ. Interrelationships between structure and function during the hemostatic response to injury. Proc Natl Acad Sci U S A. 2019;116 (6):2243-2252.

1701

PERSPECTIVE ARTICLE Ferrata Storti Foundation

Hemophilia A and B: molecular and clinical similarities and differences Giancarlo Castaman1 and Davide Matino2

Center for Bleeding Disorders and Coagulation, Department of Oncology, Careggi University Hospital, Florence, Italy and 2Department of Medicine, McMaster University and the Thrombosis and Atherosclerosis Research Institute, Hamilton, Ontario, Canada

Haematologica 2019 Volume 104(9):1702-1709

Correspondence: GIANCARLO CASTAMAN giancarlo.castaman@unifi.it

Introduction Hemophilia A and B are rare X-linked bleeding disorders caused by mutations in the genes encoding coagulation factor VIII (FVIII) and factor IX (FIX). Hemophilia A (HA) is more common than hemophilia B (HB), with a prevalence of one in 5,000 male live births compared to one in 30,000, respectively.1 The disease severity in hemophilia is classified according to the plasma level of FVIII or FIX activity. The severe form is defined as a factor level <1% of normal, the moderate form as a factor level of 1-5%, and the mild form with a factor level >5 and <40%.2 Patients with severe hemophilia frequently develop hemorrhages into joints, muscles or soft tissues without any apparent cause. They can also suffer from life-threatening bleeding episodes such as intracranial hemorrhages. Persons with mild and moderate factor deficiency rarely experience spontaneous hemorrhages, and excessive bleeding mostly occurs only following trauma or in association with invasive procedures. The residual factor activity generally correlates well with clinical characteristics; however, heterogeneous bleeding phenotypes among individuals with the same factor levels can occur.3 Furthermore, although HA and HB have been usually considered clinically indistinguishable with negligible differences in severity and outcomes, several recent studies are challenging this concept, suggesting that patients with HB could have a less severe bleeding tendency compared to HA patients with the same residual plasma level.4 In this review, we provide an up-to-date summary of evidence highlighting the similarities and differences of these two clotting factor deficiencies.

Comparison of gene defects in hemophilia A and hemophilia B Received: March 5, 2019. Accepted: June 5, 2019. Pre-published: August 8, 2019.

doi:10.3324/haematol.2019.221093 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1702 ÂŠ2019 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.

1702

Both F8 and F9 genes are located on the X chromosome, F8 gene being at the end of the long arm at Xq285and F9 IX gene on the long arm, more towards the centromere, at Xq27.6 F8 gene is extremely large (approx.180 kb) and structurally complex (26 exons), while F9 gene is considerably smaller (approx. 34 kb in length) and structurally simpler, containing only eight exons, the largest of which is only 1,935 bp long. The mutations causing hemophilia A and B have been characterized in several thousands of patients. What is immediately evident from the enormous number of mutations that have been elucidated is that the molecular basis of the hemophilias is extremely diverse. Point mutations, deletions, insertions, and rearrangements/inversions have all been found either in F8 and F9 genes. However, the relative frequency of these mutations differs between HA and HB. In particular, gross genetic abnormalities account for approximately 7% of HB cases in contrast to HA in which gene rearrangements account for almost half of severe cases, with intron 22 inversion being the most common defect. A summary of the differential characteristics of hemophilia A and B is presented in Figure 1. Previous studies have shown that the mutation type in the FVIII and FIX genes correlates with the residual factor activity in plasma and the bleeding tendency in hemophilia patients, with larger gene defects generally associated with a more severe clinical phenotype.3,7 Although one could intuitively argue that HA and HB patients with null mutations could experience a similar bleeding history, such a comparison has never been systematically carried out. The different prevalence of mutations predicting a null allele also explains a higher proportion of HB patients that can be classified as cross-reacting material positive (CRM+). The presence of null mutations prevents the synthesis of any detectable FVIII or FIX antigen. Approximately 5% of HA patients are CRM+ and haematologica | 2019; 104(9)

Hemophilia A and B: sharing differences

have circulating FVIII protein levels at almost 30% of normal. The mutations thought to be responsible for CRM+ HA are generally missense mutations found in the A2-domain of FVIII.8 At variance with HA, almost one-third of patients with HB are classified as CRM+ and can produce variable amounts of FIX protein. The higher prevalence of less severe mutations (missense mutations) in HB could provide a biological basis for a milder bleeding phenotype compared to HA, although clinical evidence is limited. Furthermore, this could also explain the lower prevalence of inhibitors in HB, mostly associated with stop codon or partial/whole gene deletion, probably together with the fact that FIX is smaller than FVIII, with less antigenic epitopes. Interestingly, it is well known that some missense mutations in mild HA are associated with inhibitor occurrence,9 while this has never been reported in patients with mild HB. It is also interesting to note that for some FIX nonsense gene mutations in HB, usually categorized as null mutations, the mechanism of ribosome readthrough could restore translation impaired by mutations and could account for minimal full-length protein biosynthesis. This mechanism could be a modifier of clinical outcomes in this specific patient population.10 Finally, although rare, being implicated in just a small proportion of severe HB cases, it is worth noting the possibility of a particular variant of HB: the hemophilia B Leyden.11 The molecular mechanism is likely to involve disruptions of sites in the proximal promoter of the F9 gene. In this condition, abnormal hemostasis is present after birth but spontaneously ameliorates at puberty, with a progressive recovery of FIX expression and normaliza-

Figure 1. Comparison of characteristics of hemophilia A and B. FIX: factor IX. 54 Nazeef and Sheehan (2016).

haematologica | 2019; 104(9)

tion of FIX level in adulthood.12 This effect is associated with rising post-pubertal growth hormone levels.13 Similar molecular mechanisms that can potentially improve the clinical presentation or outcomes, such as these two mechanisms just discussed for HB, have not yet been identified in HA patients.

Similarities and differences in hemophilia A and B clinical phenotype The numerous bleeding episodes that individuals with severe hemophilia experience can lead to long-term disability. Recurrent joint bleedings can result in severe arthropathy, muscle atrophy, pseudo-tumors, and lead to chronic pain and impaired mobility that often requires surgery and arthroplasty to improve joint function. HA and HB display similar clinical characteristics; however, several studies have reported on possible differences in bleeding frequency and factor consumption,14 clinical scores,15 and the need for orthopedic surgery.16,17 The possible different clinical evolution of HB was initially suggested in 1959 by Quick18 and was based on 24 HB cases he had personally studied. He observed that HB, even in its most severe form, can be less incapacitating and disabling than HA, and that this difference was especially pronounced after adolescence. It should be kept in mind, however, that historically, in some studies, severe HB has been defined with a FIX <2% that could contribute to a less severe bleeding tendency compared to HA, usually defined with a FVIII <1%. However, forty years after Quick, a retrospective study reporting demographic characteristics, hospital admissions, and causes of death of patients with hemophilia was carried out in Scotland by Ludlam et al.19 They retrospectively studied 282 patients

Gouw et al. (2012); 7Belvini et al. (2005); 53Brummel-Ziedins and Mann. (2014);

1703

G. Castaman and D. Matino et al.

with hemophilia during the period between 1980 and 1994 who were treated predominantly with on-demand therapy. The authors found a lower rate of hospital admissions for patients with HB at all levels of severity, suggesting that these individuals have a milder bleeding phenotype compared to HA patients. Results consistent with these were obtained in the US a few years later in a cross-sectional study conducted between May 1998 and May 2002.20 Data collected from 4,343 males with hemophilia aged 2-19 years included age, bleeding frequency, family history, insurance status, orthopedic procedures, prophylaxis use, age at diagnosis and first hemophilia treatment center (HTC) visit, frequency of visits, hemophilia type, inhibitor status, race/ethnicity, body mass index. The authors highlighted the fact that overall, individuals with HB consistently reported fewer bleeding episodes, regardless of age or severity. Interestingly, among individuals with moderate factor deficiency, those affected by HA had a greater degree of range of motion limitation compared to persons with HB. A survey conducted in 2006 aiming to describe prophylaxis use in patients of all ages and severities with HA or HB in Canada also showed some differences between HA and HB treatment.21 Data on 2,663 individuals (2,161 hemophilia A, 502 hemophilia B), were returned by 22 Canadian HTC, totaling 98% of the Canadian hemophilia population. When comparing the use of prophylaxis, the authors reported that 32% of patients with severe HB were receiving prophylaxis compared with 69% of patients with severe HA. However, it is not clear if this difference is the result of a real or perceived difference in the clinical phenotype or just reflect the traditional therapeutic approach to HB patients. However, a subsequent study reported similar results. In a project aimed at constructing a composite score (Hemophilia Severity Score, HSS) to assess the severity of the disease, Schulman et al. evaluated 100 patients affected by HA (n=67) and HB (n=33).15 This was intended as a comprehensive measurement of the clinical severity of the disease and took into account the number of joint bleeds per year, the orthopedic joint score, and the annual consumption of FVIII. Interestingly, the HSS was higher for severe HA [median=0.50; interquartile range (IQR)=0.410.68] than for severe HB (median=0.29; IQR=0.23-0.45) (P= 0.031). This result was not replicated in a subsequent external validation of the score in a smaller, single-center study in Italy. In this case, 65 consecutive hemophilia patients (57 with HA, 8 with HB) were enrolled, and no differences in HSS score were found between HA and HB (median=0.87 for severe HA vs. 0.91 in severe HB patients).22 An additional study that indirectly showed a possible difference in the clinical phenotype of severe HA compared to HB was published a few years later. This singlecenter, case-control study was carried out in Italy to evaluate the role of genotype and endogenous thrombin potential (ETP) as possible predictors of the clinical phenotype of patients affected by severe hemophilia.3 The authors evaluated patients displaying an extremely mild bleeding tendency (n=22) in comparison with those showing a typical bleeding tendency (n=50). In this study, the odds of having a milder form of the disease was five times higher in HB patients compared to persons affected by severe HA.3 1704

More recently, a Canadian single-institute retrospective study evaluating possible differences between bleeding frequency and use of factor concentrate among adult patients with severe and moderate HA and HB was published.14 Sixty-eight HA patients (58 severe, 10 moderate) and 20 patients with HB (15 severe, 5 moderate) were studied between 2001 and 2003. Although no significant difference in terms of factor consumption was observed between the two groups, 10 of 68 (14.7%) HA patients had surgical procedures to correct musculoskeletal complications compared to only 1 of 21 (4.7%) in the HB patient group. The bleeding events were also more frequent in the HA group. A total of 2,800 bleeding events were reported in the severe HA group (average 16/patient/year) while 502 total bleeds were reported among the severe HB patients (average 11/patient/year). The difference in the average number of bleeds per year was even more pronounced when considering patients with moderate factor deficiency: 4.6 for HA (n=10) and 1.06 for HB (n=5) patients. However, a few years later, a study of pediatric HA and HB patients showed apparently contrasting results;23 overall, this study showed a similar severity in the bleeding phenotype during the initial stage of the disease in severe and in moderate hemophilia A and B. The cohort of patients in this analysis was made up of consecutive severe and moderate HA and HB patients from the PedNetHaemophilia Registry study and patients with severe HA from the RODIN study. A total of 582 patients with severe HA and 76 with severe HB were included and there was no difference in age at first exposure to clotting factor (0.81 vs. 0.88 years; P=0.20), age at first bleed (0.82 vs. 0.88 years; P=0.36), or age at first joint bleed (1.18 vs. 1.20 years; P=0.59).23 However, one should bear in mind that this study differed substantially from the others with respect to: a) age (pediatric population vs. adults); b) extensive use of prophylaxis (the authors reported a uniform intention to treat with continuous prophylaxis in 90% of patients born between January 1st 2000 and January 1st 2010); c) type of outcomes evaluated (bleeding characteristics during the early stage of the disease compared to later-in-life bleeding phenotype and musculoskeletal complications). It is interesting to consider that for all parameters in this study there was a non-significant trend towards earlier age at bleeding in HA versus HB patients. A robust support to the different frequencies of bleeding episodes among the two comes from two recent trials recruiting patients with HA and HB, all treated on demand, for phase III studies with recombinant long-acting products.24,25 These studies clearly showed that, at enrollment, the annualized bleeding rates in the year before entering the studies were significantly greater in HA patients. A significant contribution to understanding the possible different evolution of the hemophilic arthropathy in HA and HB was produced by Melchiorre et al. in 2016.26 In this study, including mostly adult patients, the authors showed that the ultrasound score was significantly worse in HA when matched for age and frequency of hemarthrosis. Likewise, the World Federation of Hemophilia clinical score in the HB group was lower [mean and Standard Deviation (SD): 48.6Âą16.2 vs. 22.6Âą16.4; P<0.0001], indicating a less severe arthropathy than in HA patients with a similar total number of hemarthrosis. In addition, the haematologica | 2019; 104(9)

Hemophilia A and B: sharing differences

analysis of circulating osteoprotegerin (which plays a protective role for the subchondral bone) and receptor activator of nuclear factor-kB and RANK ligand (involved in osteoclast activation and bone erosions) showed a more favorable profile in HB patients. Consistent results were obtained with the histological analysis performed on synovial tissue collected from these patients. Taken together, these data confirmed a less severe evolution of the arthropathy in HB patients and widened our understanding of the pathophysiological mechanisms underlying the different rate of joint deterioration and severity of disease. Data published in 2018 by Mancuso et al.,27 reporting a study aimed at the development and validation of criteria to define clinically severe hemophilia (CSH), showed again that FIX deficiency is associated to a milder clinical phenotype when comparing patients with the same residual factor activity. In this study, the authors evaluated the ability of residual circulating FVIII/FIX measured at diagnosis using a one-stage clotting assay to discriminate a severe clinical phenotype (defined a priori as a CSH score >3). Importantly, the results showed a sensitivity of 0.87 [95% Confidence Interval (CI): 0.81-0.91] for FVIII but only 0.68 (95%CI: 0.43-0.87) for FIX, considering a cut-off of 1 IU/dL. In this study, 65.5% (156 of 238) of severe HA patients and 41.2% (13 of 31) of severe HB patients had a CSH score >3. The higher proportion of patients with HA with a severity score >3 suggests also in this cohort of patients the possible milder phenotype in patients with HB. Among patients with severe disease, the odds of having a clinically more severe form of bleeding symptoms in HA was 2.63 (95%CI: 1.23, 5.64). These results have been recently confirmed also in a study on HA and HB patients with mild disease.28

Orthopedic surgery The need for orthopedic surgical treatment can be considered a surrogate of severity of hemophilia disease. Chronic arthropathy is a consequence of recurrent bleeding into joints, hemarthrosis, which is a hallmark of severe hemophilia. The higher the number of bleeds in the joints, the higher the chance that a patient will develop permanent bone and cartilage damage requiring surgical intervention. In a retrospective national collection of data on hemophilia patients who underwent joint arthroplasty, Tagariello et al.17 found an Odds Ratio of 3.38 (95%CI: 1.97-5.77; P<0.001) when considering the risk of undergoing orthopedic surgery in HA compared to HB. This difference was confirmed after adjustment for human immunodeficiency virus, hepatitis C virus, and inhibitor status [Hazard Ratio (HR): 2.65; 95%CI: 1.62-4.33; P<0.001]. It is important to note that neither HA nor HB patients had been on regular primary prophylaxis during their lifetime before arthroplasty. A study on a smaller cohort of patients from the Netherlands could not confirm these results.29 However, this Dutch analysis was based on a substantially lower number of arthroplasty interventions and patients were mostly on factor prophylaxis (77% in HA, 73% in HB). A more recent study from the hemophilia treatment centers in the USA collected data on mild and moderate hemophilia patients who were exclusively treated with on-demand therapy.16 Patients with inhibitors were excluded. A total of 4,771 patients were included in the analysis; 289 (6%) had had orthopedic surgery, such as synovectomy (n=75), joint fusion or joint replacement haematologica | 2019; 104(9)

(n=126), and 123 had a different type of invasive orthopedic procedure. Interestingly, in the regression analysis, the predicted number of joint bleeds for patients with factor activity <30% was greater for patients with HA. Also, the likelihood of undergoing an invasive orthopedic procedure was lower for HB patients (OR: 0.7, 95%CI: 0.5-0.9). These data are consistent with the Italian experience which has also suggested a more frequent progression to orthopedic surgery among patients with HA. Taken together, these results suggest a milder natural history of the disease in the individuals affected by HB. Table 1 summarizes the main clinical findings reported in the studies.

Potential mechanisms affecting the variability in disease severity between hemophilia A and hemophilia B Several possible underlying biological explanations for differences in disease severity can be hypothesized and are presented here. A summary of these mechanisms is reported in Figure 2. Associated prothrombotic abnormalities - the variable severity and frequency of bleeding that patients with hemophilia can experience, even at the same measured factor activity, has long been reported. The presence of an associated hypercoagulable state, such as gain-of-function mutations (FV Leiden or prothrombin G20210A) and other coagulation abnormalities (deficiencies of antithrombin, protein C, protein S), has been hypothesized to modulate the bleeding phenotype. However, the clinical relevance of such factors in modifying the clinical phenotype of severe hemophilia patients is still uncertain. In fact, a low prevalence of such prothrombotic factors in severe HA and HB patients with a milder phenotype30,31 has been reported and conflicting results from different studies have been seen.32-35 A more recent study investigating ETP as predictor of the clinical phenotype in severe hemophilia patients showed no differences in the distribution of FV Leiden or prothrombin G20210A mutations between severe HA and HB patients.3 Extravascular distribution of FIX â&#x20AC;&#x201C; a possible explanation for a milder phenotype in HB patients may lie in differences in the molecular characteristics and different pharmacokinetics of FVIII and FIX proteins. FVIII resides exclusively in the intravascular space, and its residence time is determined exclusively by the rate of plasma clearance. In contrast, FIX also distributes extravascularly. Since the first pharmacokinetic studies of FIX concentrates, it has become evident that the volume of distribution of FIX, unlike FVIII, is around four times greater than the estimated patient plasma volume and is similar to the central compartment plus the volume of the extracellular fluid. Significant extravascular FIX compartmentalization may increase the apparent volume of distribution, and potentially constitutes a mechanism for extended levels of biologically active FIX. In fact, pharmacokinetic studies showed that the PK of FIX is not linear and is most likely best represented by 3-compartment modeling. Assuming a multi-compartmental model implies that the drug in question, here specifically FIX, follows a complex disposition, with receptor binding or compartmentalization in some extra-vascular space, with potential pharmacodynamic implications.36 Even though there has not yet been a direct demonstration of any clinical impact of such a mechanism, pre-clin1705

G. Castaman and D. Matino et al.

ical studies have now provided several lines of evidence of the ability of FIX to distribute extravascularly,37 of binding to the extracellular matrix38 (and in particular to type IV collagen39), an in vivo effect of the binding to collagen IV,40 and of tissue distribution of FIX.41 Although there is some

evidence of tissue distribution in humans,42,43 in hemophilia patients, the extravascular compartment is not readily accessible and further studies are needed in order to obtain robust evidence of the clinical impact of these specific characteristics of FIX.

Table 1. Summary of main clinical findings reported in the studies.

Patients with severe hemophilia Hemophilia A Hemophilia B Ludlam et al.19a N of patients 99 24 Hospital admission rate (bed days/pt/year) 7.3 3.1ᵒ Biss et al.21 N of patients 681 134 N of patients receiving prophylaxis (%) 424/617 (32) 42/131 (32) ᵒ Schulman et al.15 N of patients 37 6 Bleeding score, median (IQR) 0.15 (0.1–0.25) 0.12 (0.6–0.19) ᵀ Joint score, median (IQR) 0.079 (0.033–0.144) 0.057 (0.024–0.122) ᵀ Factor score, median (IQR) 0.15 (0.086–0.267) 0.039 (0.028–0112)* HSS, median (IQR) 0.50 (0.41–0.68) 0.29 (0.23–0.45)* Tagariello et al.17 N of patients 1770 319 Rate of patients undergoing arthroplasty, % (95% CI) 14.3% (12.7%-15.9%) 4.7% (2.4%-7.0%) Odds of undergoing arthroplasty, OR (95% CI) 3.38 (1.97-5.77) 1** Den Uijl et al.29 N of patients 252 30ᵒ Incidence of arthroplasty (%) 78 (31%) 9 (30%)ᵀ Age at 1st treatment (years), medians (5th-95th percentiles) 1.1 (0.2-2.7) 1.3 (0.6-2.9) ᵀ Age at 1st joint bleed (years), medians (5th-95th percentiles) 1.9 (0.5-5.9) 2.4 (0.9-5.5) ᵀ Patients on prophylaxis, N (%) 194 (77%) 22 (73%)ᵀ Annual joint bleeding frequency, medians (5th-95th percentiles) 4.3 (0.3-16.3) 3.8 (0.4-17.8) ᵀ Annual factor use (IU Kg-1) , medians (5th-95th percentiles) 1560 (286-3644) 1260 (302-5826) ᵀ Santagostino et al.3 N of patients 61 11ᵒ Mild bleeding phenotype, N (%) 15 (25) 7 (64) * Median age at first bleed, months (IQR) 12 (12–36) 24 (12–48) ᵀ Median age at first joint bleed, months (IQR) 24 (24–48) 66 (51–102) ** Median number of bleeds per year (IQR) 11 (2–25) 0 (0–12)* Clausen et al.23 N of patients 582 76 Age at diagnosis (years) 0.42 (3 days–0.88) 0.43 (0 days–0.88)ᵀ Age at 1st treatment 0.81 (0.43–1.11) 0.88 (0.60–1.18)ᵀ Age at 1st bleed 0.82 (0.50–1.12) 0.88 (0.60–1.19)ᵀ Patients without bleeding without prophylaxis < 2 years of age, N 105 (17.5) 12 (15.4)ᵀ Median factor (IU Kg-1/year) (IQR) 1280 (369–2170) 333 (38–487)** Median orthopedic joint score (range) 13 (6–20) 5 (2–9)* Median Pettersson score (range) 28 (20–45) 23 (9–31) ᵀ Nagel et al.14 N of patients 58 15 Total bleeds over 36 months, N 2 800 502ᵒ Joint bleeds, N 1 491 332ᵒ Melchiorre et al.26 N of patients 70 35 Hemarthrosis, N (%) <10 11 (15.7) 15 (42.9) ** 10-50 16 (22.8) 3 (8.5) ** >50 43 (61.4) 17 (48.6)* Pettersson score, mean ± SD 6.81±3.99 5.64±4.02ᵀ WFH score, mean ± SD 36.6±21.6 20.2±14.6** US score, mean ± SD 10.91±4.05 4.34±3.39**

Patients with moderate hemophilia Hemophilia A Hemophilia B 69 3.5

33 2.1ᵒ

250 43/244 (18)

188 9/187 (5) ᵒ

21 8 0.05 (0–0.15) 0.05 (0.019–0.11) ᵀ 0.020 (0–0.033) 0.054 (0.016–0.096) ᵀ 0.004 (0.001–0.015) 0.004 (0.002–0.011) ᵀ 0.073 (0.024–0.253) 0.115 (0.039–0.349) ᵀ Na Na Na

Na Na Na

Na Na Na Na Na Na Na

Na Na Na Na Na

97 26ᵒ 0.77 (2 days–1.62) 2.5 days (0 days–1.17)ᵀ 1.42 (0.90–2.92) 1.74 (0.99–4.10)ᵀ 1.47 (0.98–2.82) 1.76 (0.94–3.23)ᵀ 34 (34.7) 15 (57.7)* Na Na Na Na Na Na 10 138 Na Na

5ᵒ 16ᵒ Na Na

Na Na Na

Severe hemophilia was defined as a factor VIII/IX level of <2 U/dL; moderate hemophilia was defined as 2-9 U/dL. +Number (N) of patients at study end in 1994. *P≤0.05, HA versus HB. **P≤0.001, HA versus HB. ᵀNS: not statistically significant, HA versus HB. ᵒComparison of HA versus HB not available. IQR: interquartile range; CI: Confidence Interval; HSS: Hemophilia Severity Score; Pt: patient; SD: Standard Deviation; WFH: World Federation of Hemophilia; US: ultrasound; Na: not available. a

1706

haematologica | 2019; 104(9)

Hemophilia A and B: sharing differences

It is interesting to note that the results of a recent study by Cooley et al.39 also suggest that the total amount of FIX is approximately three times larger than what can be measured in the intravascular compartment. The accessible intravascular compartment from which we collect our plasma samples to measure FIX activity might not, therefore, be reflecting the overall coagulative capacity of the endogenous FIX, but may only reveal information about the fraction that is freely circulating. This adds up to the inter-laboratory differences in FVIII:C and FIX:C measurements and it should be borne in mind that also misclassification might explain, at least in part, the discrepancy in the bleeding tendency between severe HA and HB patients. Altogether, the information available so far on the pharmacokinetic and pharmacodynamic characteristics of FIX suggest some potential mechanisms that could explain a difference in bleeding tendency between HB and HA patients. Inhibitor development - the occurrence of high affinity anti-FVIII or anti-FIX antibodies that neutralize the activity of the infused clotting factor is a major complication of replacement therapy in hemophilia patients. Inhibitor antibodies against FVIII can develop in approximately 2530% of severe HA patients; in contrast, in patients with HB, inhibitors develop in only 3-5% of patients treated with factor concentrates.44 Immune tolerance induction (ITI) via long-term, intravenous administrations of factor concentrates is the only proven strategy to eradicate inhibitors. However, this approach is expensive and impractical for the patients.45 Overall, it is successful in approximately 60-70% of HA patients. The results of a

randomized, controlled study comparing high-dose (200 IU/Kg/day) and low-dose (50 IU/Kg/3 times a week) protocols in a cohort of severe HA patients with high-titer inhibitor showed a similar overall success rate and no statistically significant differences in time to achieve tolerance, but the median time to negative inhibitor titer was 4.6 months (range: 2.8-13.8).46 Although the development of inhibitors to FIX is a much less common event, ITI treatment is not as successful (only approx. 30%), and the development of anti-FIX antibodies may be associated with anaphylactic reactions which may prevent or complicate ITI regimes in a significant proportion of cases.47 HB patients undergoing ITI can also develop nephrotic syndrome.48,49 However, a few anecdotal experiences suggest that ITI could be successfully achieved by adding immunosuppression treatment, especially in patients with anaphylactic reactions.50

Conclusions Different lines of evidence seem to support a difference in bleeding severity between HA and HB. The pathophysiology of the two disorders is, indeed, diverse, with a different distribution of the factors in the body and, in keeping with this, the PK characteristics of infused factors are significantly different. However, because of the rarity of the disorders, no prospective, head-to-head comparative studies have been carried out, and the modern approach in the pediatric hemophilia population (i.e. starting prophylaxis very early) does not allow us to acquire a better understanding of the possible clinical differences during follow up. Improved replacement therapy with extended half-life concentrates with 10- to 14-day intervals between infusions and sustained high FIX troughs51 are greatly

Figure 2. Possible mechanisms involved in the differences in clinical evolution between hemophilia A and hemophilia B patients. FIX: factor IX.

haematologica | 2019; 104(9)

1707

G. Castaman and D. Matino et al.

improving the clinical outcome and, even more so, the quality of life of HB patients when compared to HA patients. The promising perspectives of gene therapy are painting a future scenario in which the decision to offer this option to patients with HB and very mild clinical problems could be challenging considering the costs involved and the yet unknown long-term effects.

References 16. 1. Berntorp E, Shapiro AD. Modern haemophilia care. Lancet. 2012;379(9824):1447-1456. 2. Blanchette VS, Key NS, Ljung LR, MancoJohnson MJ, van den Berg HM, Srivastava A. Definitions in hemophilia: communication from the SSC of the ISTH. J Thromb Haemost. 2014;12(11):1935-1939. 3. Santagostino E, Mancuso ME, Tripodi A, et al. Severe hemophilia with mild bleeding phenotype: molecular characterization and global coagulation profile. J Thromb Haemost. 2010;8(4):737-743. 4. Mannucci PM, Franchini M. Is haemophilia B less severe than haemophilia A? Haemophilia. 2013;19(4):499-502. 5. Gitschier J, Wood WI, Goralka TM, et al. Characterization of the human factor VIII gene. Nature. 1984;312(5992):326-330. 6. Yoshitake S, Schach BG, Foster DC, Davie EW, Kurachi K. Complete nucleotide sequences of the gene for human factor IX (antihemophilic factor B). Biochemistry. 1985;24(14):3736-3750. 7. Belvini D, Salviato R, Radossi P, et al. Molecular genotyping of the Italian cohort of patients with hemophilia B. Haematologica. 2005;90(5):635-642. 8. Amano K, Sarkar R, Pemberton S, KemballCook G, Kazazian HH, Kaufman RJ. The molecular basis for cross-reacting materialpositive hemophilia A due to missense mutations within the A2-domain of factor VIII. Blood. 1998;91(2):538-548. 9. Castaman G, Fijnvandraat K. Molecular and clinical predictors of inhibitor risk and its prevention and treatment in mild hemophilia A. Blood. 2014;124(15):2333-2336. 10. Pinotti M, Caruso P, Canella A, et al. Ribosome readthrough accounts for secreted full-length factor IX in hemophilia B patients with nonsense mutations. Hum Mutat. 2012;33(9):1373-1376. 11. Goodeve AC. Hemophilia B: molecular pathogenesis and mutation analysis. J Thromb Haemost. 2015;13(7):1184-1195. 12. Kurachi S, Huo JS, Ameri A, Zhang K, Yoshizawa AC, Kurachi K. An age-related homeostasis mechanism is essential for spontaneous amelioration of hemophilia B Leyden. Proc Natl Acad Sci U S A. 2009;106(19):7921-7926. 13. Funnell APW, Crossley M. Hemophilia B Leyden and once mysterious cis-regulatory mutations. Trends Genet. 2014;30(1):18-23. 14. Nagel K, Walker I, Decker K, Chan AKC, Pai MK. Comparing bleed frequency and factor concentrate use between haemophilia A and B patients. Haemophilia. 2011;17(6):872-874. 15. Schulman S, Eelde A, Holmström M, Ståhlberg G, Odeberg J, Blombäck M. Validation of a composite score for clinical

1708

17.

18. 19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

Acknowledgments GC has received unrestricted research grants directly to his Institution by CSL Behring, Pfizer and Sobi and has received honoraria as a speaker or for participating in Advisory Boards or educational events from Bayer, Baxalta/Shire, CSL Behring, Kedrion, Novo Nordisk, Pfizer, Roche, Uniqure.

severity of hemophilia. J Thromb Haemost. 2008;6(7):1113-1121. Soucie JM, Monahan PE, Kulkarni R, Konkle BA, Mazepa MA, US Hemophilia Treatment Center Network. The frequency of joint hemorrhages and procedures in nonsevere hemophilia A vs B. Blood Adv. 2018;2(16):2136-2144. Tagariello G, Iorio A, Santagostino E, et al. Comparison of the rates of joint arthroplasty in patients with severe factor VIII and IX deficiency: an index of different clinical severity of the 2 coagulation disorders. Blood. 2009;114(4):779-784. Quick AJ, Hussey CV. Hemophilia B (PTC Deficiency, or Christmas Disease). Arch Intern Med. 1959;103(5):762. Ludlam CA, Lee RJ, Prescott RJ, et al. Haemophilia care in central Scotland 198094. I. Demographic characteristics, hospital admissions and causes of death. Haemophilia. 2000;6(5):494-503. Soucie JM, Cianfrini C, Janco RL, et al. Joint range-of-motion limitations among young males with hemophilia: prevalence and risk factors. Blood. 2004;103(7):2467-2473. Biss TT, Chan AK, Blanchette VS, Iwenofu LN, Mclimont M, Carcao MD. The use of prophylaxis in 2663 children and adults with haemophilia: results of the 2006 Canadian national haemophilia prophylaxis survey. Haemophilia. 2008;14(5):923930. Tagliaferri A, Di Perna C, Franchini M, Rivolta GF, Pattacini C. Hemophilia severity score system: validation from an Italian Regional Hemophilia Reference Center. J Thromb Haemost. 2009;7(4):720-722. Clausen N, Petrini P, Claeyssens-Donadel S, Gouw SC, Liesner R. Similar bleeding phenotype in young children with haemophilia A or B: a cohort study. Haemophilia. 2014;20(6):747-755. Powell JS, Pasi KJ, Ragni MV, et al. B-LONG investigators. Phase 3 study of recombinant factor IX FGc fusion protein in hemophilia B. N Engl J Med. 2013;369(24):2313-2323 Mahlangu J, Powell JS, Ragni MV et al. ALONG Investigators. Phase 3 study of recombinant factor VIII Fc fusion protein in severe hemophilia A. Blood. 2014;123 (3):317-325. Melchiorre D, Linari S, Manetti M, et al. Clinical, instrumental, serological and histological findings suggest that hemophilia B may be less severe than hemophilia A. Haematologica. 2016;101(2):219-225. Mancuso ME, Bidlingmaier C, Mahlangu JN, Carcao M, Tosetto A. The predictive value of factor VIII/factor IX levels to define the severity of hemophilia: communication from the SSC of ISTH. J Thromb Haemost. 2018;16(10):2106-2110. Linari S, Nichele I, Pieri L, Tosetto A, Castaman G. Bleeding tendency and clinical phenotype of mild haemophilia A and

29.

30.

31.

32.

33.

34.

35.

36.

37. 38.

39.

40.

41.

B. Blood Transfusion. 2018;16(Suppl 4):s490. den Uijl IEM, Roosendaal G, Fischer K. Insufficient evidence to suggest less stringent therapy in hemophilia B? Blood. 2009;114(23):4907; author reply 4907-8. Arbini A, Mannucci P, Bauer K. Low Prevalence of the Factor V Leiden Mutation Among “Severe” Hemophiliacs with a “Milder” Bleeding Diathesis. Thromb Haemost. 1995;74(05):1255-1258. Lee DH, Walker IR, Teitel J, et al. Effect of the Factor V Leiden Mutation on the Clinical Expression of Severe Hemophilia A. Thromb Haemost. 2000;83(03):387-391. Shetty S, Vora S, Kulkarni B, et al. Contribution of natural anticoagulant and fibrinolytic factors in modulating the clinical severity of haemophilia patients. Br J Haematol. 2007;138(4):541-544. Van Dijk K, Van Der Bom JG, Fischer K, De Groot PG, Van Den Berg HM. Phenotype of severe hemophilia A and plasma levels of risk factors for thrombosis. J Thromb Haemost. 2007;5(5):1062-1064. Escuriola Ettingshausen C, Halimeh S, Kurnik K, et al. Symptomatic Onset of Severe Hemophilia A in Childhood is Dependent on the Presence of Prothrombotic Risk Factors. Thromb Haemost. 2001;85(02):200-208. Tizzano EF, Soria JM, Coll I, et al. The prothrombin 20210A allele influences clinical manifestations of hemophilia A in patients with intron 22 inversion and without inhibitors. Haematologica. 2002;87(3):279285. Iorio A, Fischer K, Blanchette V, et al. Tailoring treatment of haemophilia B: accounting for the distribution and clearance of standard and extended half-life FIX concentrates. Thromb Haemost. 2017;117(06):1023-1030. Stafford DW. Extravascular FIX and coagulation. Thromb J. 2016;14(S1):35. D M Stern, P P Nawroth, W Kisiel GV and CTE. The binding of factor IXa to cultured bovine aortic endothelial cells. Induction of a specific site in the presence of factors VIII and X. J Biol Chem. 1985;260(11),67176722. Cheung WF, van den Born J, Kühn K, Kjellén L, Hudson BG, Stafford DW. Identification of the endothelial cell binding site for factor IX. Proc Natl Acad Sci U S A. 1996;93(20):11068-11073. Gui T, Reheman A, Ni H, et al. Abnormal hemostasis in a knock-in mouse carrying a variant of factorÂ IX with impaired binding to collagen typeÂ IV. J Thromb Haemost. 2009;7(11):1843-1851. Matino D., Iorio A., Stafford D. CA. Enhanced FIX collagen IV binding shows improved hemostatic effects in a hemophilia B mouse model. Res Pract Thromb Haemost. 2017;1:1-1451.

haematologica | 2019; 104(9)

Hemophilia A and B: sharing differences

42. Cooley B, Funkhouser W, Monroe D, et al. Prophylactic efficacy of BeneFIX vs Alprolix in hemophilia B mice. Blood. 2016;128(2):286-292. 43. Gui T, Lin H-F, Jin D-Y, et al. Circulating and binding characteristics of wild-type factor IX and certain Gla domain mutants in vivo. Blood. 2002;100(1):153-158. 44. Bolton-Maggs PHB, Pasi KJ. Haemophilias A and B. Lancet. 2003;361(9371):1801-1809. 45. Dimichele DM. Inhibitor treatment in haemophilias A and B: inhibitor diagnosis. Haemophilia. 2006;12 Suppl 6:37-42. 46. Hay CRM, DiMichele DM, International Immune Tolerance Study. The principal results of the International Immune Tolerance Study: a randomized dose comparison. Blood. 2012;119(6):1335-1344. 47. Dimichele D. The North American Immune Tolerance registry: contributions

haematologica | 2019; 104(9)

to the thirty-year experience with immune tolerance therapy. Haemophilia. 2009;15 (1):320-328 48. Lee CA, Kessler CM, Varon D, Martinowitz U, Heim M, Warrier I. Management of haemophilia B patients with inhibitors and anaphylaxis. Haemophilia. 1998;4(4):574-576. 49. Tengborn L, Hansson S, Fasth A, LĂźbeck PO, Berg A, Ljung R. Anaphylactoid reactions and nephrotic syndrome--a considerable risk during factor IX treatment in patients with haemophilia B and inhibitors: a report on the outcome in two brothers. Haemophilia. 1998;4(6):854-859. 50. Alexander S, Hopewell S, Hunter S, Chouksey A. Rituximab and desensitization for a patient with severe factor IX deficiency, inhibitors, and history of anaphylaxis. J Pediatr Hematol Oncol. 2008;30

(1):93â&#x20AC;&#x201C;95. 51. Gill JC, Roberts J, Li Y, Castaman G. Sustained high trough factor IX activity levels with continued use of rIX-FP in adult and paediatric patients with haemophilia B. Haemophilia. 2019 Mar 13. [Epub ahead of print] 52. Gouw SC, van den Berg HM, Oldenburg J, et al. F8 gene mutation type and inhibitor development in patients with severe hemophilia A: systematic review and metaanalysis. Blood. 2012;119:2922-2934. 53. Brummel-Ziedins KE, Mann KG. Overview of hemostasis. In Lee CA, Berntorp EE, Hoots WK. Textbook of haemophilia, 3rd edition, pages 1-8, Wiley-Blacwell, London, UK, 2014. 54. Nazeef M, Sheehan JP. New developments in the management of moderate-to-severe hemophilia B. J Blood Med. 2016;7:27-38.

1709

REVIEW ARTICLE Ferrata Storti Foundation

Emerging disease-modifying therapies for sickle cell disease Marcus A. Carden1,2 and Jane Little2

Department of Pediatrics, Division of Pediatric Hematology/Oncology, University of North Carolina Chapel Hill School of Medicine and 2Department of Medicine, Division of Hematology, University of North Carolina Chapel Hill School of Medicine, Chapel Hill, NC, USA

Haematologica 2019 Volume 104(9):1710-1719

ABSTRACT

Correspondence: JANE LITTLE jane_little@med.unc.edu Received: April 10, 2019. Accepted: July 10, 2019. Pre-published: August 14, 2019. doi:10.3324/haematol.2018.207357 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1710

ickle cell disease afflicts millions of people worldwide and approximately 100,000 Americans. Complications are myriad and arise as a result of complex pathological pathways ‘downstream’ to a point mutation in DNA, and include red blood cell membrane damage, inflammation, chronic hemolytic anemia with episodic vaso-occlusion, ischemia and pain, and ultimately risk of cumulative organ damage with reduced lifespan of affected individuals. The National Heart, Lung, and Blood Institute’s 2014 evidence-based guideline for sickle cell disease management states that additional research is needed before investigational curative therapies will be widely available to most patients with sickle cell disease. To date, sickle cell disease has been cured by hematopoietic stem cell transplantation in approximately 1,000 people, most of whom were children, and significantly ameliorated by gene therapy in a handful of subjects who have only limited follow-up thus far. During a timespan in which over 20 agents were approved for the treatment of cystic fibrosis by the Food and Drug Administration, similar approval was granted for only two drugs for sickle cell disease (hydroxyurea and L-glutamine) despite the higher prevalence of sickle cell disease. This trajectory appears to be changing, as the lack of multimodal agent therapy in sickle cell disease has spurred engagement among many in academia and industry who, in the last decade, have developed new drugs poised to prevent complications and alleviate suffering. Identified therapeutic strategies include fetal hemoglobin induction, inhibition of intracellular HbS polymerization, inhibition of oxidant stress and inflammation, and perturbation of the activation of the endothelium and other blood components (e.g. platelets, white blood cells, coagulation proteins) involved in the pathophysiology of sickle cell disease. In this article, we present a crash-course review of disease-modifying approaches (minus hematopoietic stem cell transplant and gene therapy) for patients with sickle cell disease currently, or recently, tested in clinical trials in the era following approval of hydroxyurea.

Introduction ©2019 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.

1710

As the most common monogenic disorder and first defined ‘molecular’ disease,1,2 sickle cell disease (SCD) comprises a complex group of hematologic disorders that share a common genetic link - a missense mutation in the seventh codon of the βglobin gene that leads to adenine being replaced with thymine (GAG→GTG). In turn, at the sixth position of the mature peptide of the β-globin protein the amino acid valine replaces glutamic acid3 which, when inherited in the homozygous state, results in erythroid precursors and mature sickle red blood cells (RBC) that contain abnormal sickle hemoglobin (HbS: a2βS2), rather than normal adult hemoglobin (HbA: a2β2). Compound heterozygous diseases (HbSC: a2βSβC; and HbSβ+ thalassemia: a2βSβ+-Thal) have milder features overall, but can be debilitating and highly morbid as well. Under deoxygenated conditions, HbS polymerizes intracellularly, which makes the sickle RBC fragile, less deformable, and dehydrated, and subsequently more susceptible to endothelial adhesion through activation of adhesion haematologica | 2019; 104(9)

Disease-modifying therapies for SCD

receptors.4-7 Downstream consequences include microvascular occlusion, leukocyte and platelet activation, and a pathologically altered endothelium all existing in a proinflammatory and pro-thrombophilic plasma milieu.8-13 The biomechanical properties of sickle RBC are dependent on intrinsic factors, such as the composition of the hemoglobin [e.g., presence of the anti-sickling fetal hemoglobin (HbF: a2γ2)], membrane integrity, cellular volume and hydration, cytosolic make-up, and extrinsic factors, such as inflammatory cytokines, activated endothelium, and other blood components including platelets, leukocytes, and proteins involved in coagulation.8 Clinical manifestations of the presence of HbS polymerization are wide-ranging and include chronic hemolytic anemia, episodic microcirculatory vaso-occlusion with tissue ischemia and pain, and ultimately chronic end-organ damage that can reduce the lifespan of an individual with SCD.14 Due to its impact on morbidity and mortality, SCD is increasingly being recognized as a global health problem. Researchers in academia and industry have reinvigorated efforts to cure patients with SCD; and where that is not possible because of medical and socioeconomic barriers they aim to prevent, delay, and mitigate its protean complications.15-17 Curing SCD through stem cell transplantation and achieving durable responses through gene therapy have become realities for some patients.18,19 However, as stated by the 2014 evidence-based guidelines from the National Heart, Lung, and Blood Institute (NHLBI), additional research is needed before potentially curative therapies are widely, safely, and inexpensively available to most patients.20 Therefore, in the era following approval of hydroxyurea by the United States Food and Drug Administration (FDA), providers will need to rely on improving patients’ outcomes through utilization of one or more additional emerging novel therapies and advances in care. Although the economic cost benefit of such an approach is difficult to predict, conceptually this may

evolve into a multi-faceted approach to SCD that is similar to that seen with multi-agent chemotherapy for the successful management of cancer.21 In this context, we present emerging non-genetic approaches (i.e. those that do not involve stem cell or gene therapy) currently or recently in clinical trials that offer innovative treatment and palliation in SCD. While we do include agents involved in epigenetic targeting, excellent reviews of other genetic approaches for disease modification or cure (i.e. those receiving stem cell transplants or gene therapy through gene addition, correction, or editing) can be found elsewhere.19,22,23

Methods Relevant literature was identified through various mechanisms, including using search terms ‘sickle cell disease’ and ‘novel treatments’ in MEDLINE, reviewing recent abstracts presented at the American Society of Hematology annual meetings, and examining recent, relevant reviews by others in the field.15-17,21,24-28 Trials actively recruiting pediatric or adult patients with SCD, and which included subjects aged 18 years or older, as of February 15, 2019 were also evaluated through ClinicalTrials.gov. Upon review of each result, we excluded those trials involving gene modification, including stem cell transplant, gene addition, correction, or editing. As we outline in Online Supplementary Tables S1 and S2, we broke down what we thought were novel and important treatments into two groups - those characterized by targeting the abnormal HbS and damaged sickle RBC (i.e. intrinsic to the RBC, including formation of deoxy-HbS and its polymerization in a dehydrated sickle RBC) (Figure 1) and those targeting sequelae downstream from the red cell (i.e. extrinsic to the RBC, including abnormal endothelial and cellular adhesion, vascular tone, other blood components, and inflammation) (Figure 2).

Figure 1. Red cell intrinsic targets. The figure illustrates the therapeutic targets within the red cell precursor [DNA1, and the sickle red blood cell (RBC), HbS5, e.g. voxelator, and polymer formation16], which are most likely to modify sickle cell disease and to affect more than a single downstream sequelae (pain, inflammation, vasculopathy, so forth). RBC image: https://steemit.com/stemng/@gbindinazeez/sickle-cell-anaemia-an-endemic-disease-20181122t215228805z-post]

haematologica | 2019; 104(9)

1711

M.C. Carden and J. Little et al.

Red blood cell intrinsic targets The emerging disease-modifying approaches to SCD that target intrinsic characteristics of RBC are outlined in Online Supplementary Table S1 and discussed below.

Targeting HbS polymerization through the induction of HbF Several recently developed agents aim to reduce deoxyHbS polymerization, the root cause of SCD pathology, through delayed deoxygenation of HbS, reduced intracellular HbS concentration (via cellular hydration), or induction of the anti-sickling HbF.29 Hydroxyurea, a ribonucleotide reductase inhibitor with HbF-inducing properties, is the paradigmatic HbF-inducing agent and was the first drug approved by the FDA for the treatment of adults and children with SCD. Hydroxyurea induces HbF and increases RBC volume, thereby reducing the likelihood of HbS polymerization. Hydroxyurea also decreases neutrophil and platelet counts and increases plasma nitric oxide levels, and is overall associated with decreased morbidity and improved mortality.30,31 The 2014 NHLBI guidelines state that hydroxyurea therapy should be initiated in adults with severe SCD, especially when quality of life is affected, and offered as a prophylactic treatment in young children with sickle cell anemia.20 Novel studies assessing the benefits of hydroxyurea are evaluating individualized pharmacokinetic-based dosing strategies (NCT03789591), the safety and feasibility of adding hydroxyurea to simple transfusions for stroke prevention (NCT03644953), and using patient navigators to reduce barriers to availability and non-adherence (NCT02197845). However, some patients with SCD may not respond adequately to hydroxyurea or refuse treatment because of unwanted side-effects. As such, other agents that modify γ-globin gene silencing and induce HbF are being repurposed or newly investigated. Many drugs being, or previously, investigated work through novel epigenetic mecha-

nisms within erythroid progenitors in the bone marrow. Decitabine with (NCT01685515) or without (NCT01375608) tetrahydrouridine (a cytidine deaminase inhibitor that prevents rapid deactivation of decitabine, thereby allowing the use of an oral formulation of this latter) is a chemotherapy used to treat myelodysplastic syndrome and acute myeloid leukemia. Decitabine and its historic antecedent 5-azacytidine inhibit DNA methyltransferase-1 (DNMT1), thereby reducing overall DNA methylation.32 Perturbed DNA methylation, in animal models and humans, appears to be the major mechanism for derepressed γ-globin expression arising from this class of agents.33 A phase I, first in-human trial of decitabine/tetrahydrouridine found this drug combination to be safe (without cytotoxicity or genotoxicity), well-tolerated, and effective, increasing HbF levels to 4-9%, while doubling F-cell populations.34 Unlike 5-azacytidine, decitabine does not affect RNA metabolism and is likely to have an improved safety profile, although the impact of its irreversible incorporation into DNA has not been fully elucidated and long-term follow-up in large populations is not yet available. Dimethyl butyrate (HQK-1001), an orally bioavailable short-chain fatty acid derivative and inhibitor of histone deacetylases, was active in animal models. However, a phase II double-blind placebo-controlled study (NCT01601340) was terminated early as the drug was associated with an insignificant rise in HbF and more pain episodes when compared to placebo.35 Other histone deacetylase inhibitors that work in part by reversing γ-globin silencing and show promise in phase I trials in SCD include the multiple myeloma drugs panobinostat (NCT01245179) and pomalidomide (NCT01522547).36-38 Again, long-term risk-benefit analyses are not yet available. Lysine-specific demethylase-1 (LSD1) is another enzyme and epigenetic target that modifies histones through demethylation in the process of γ-globin gene

Figure 2. Red cell extrinsic targets. The figures shows the therapeutic targets and pathways extrinsic to the red blood cell (RBC), which are most likely to be useful for managing sickle cell disease, palliating symptoms, and improving organ function. NO: nitric oxide. [RBC image: https://steemit.com/stemng/@gbindinazeez/sickle-cell-anaemia-an-endemic-disease-20181122t215228805z-post]

1712

haematologica | 2019; 104(9)

Disease-modifying therapies for SCD silencing at the γ-promoter DRED complex, which also contains the demethylase DNMT1. LSD1 inhibition in SCD mice increases HbF, reduces reticulocytosis, and decreases organ damage.39 A phase I open-label study (NCT03132324) evaluating the safety, pharmacokinetics, and biological activity of the LSD1 inhibitor INCB059872 in patients with SCD was terminated early due to a ‘business decision’ and the drug is being examined as a treatment for leukemia. Known for its glucose-lowering mechanism in patients with type 2 diabetes, metformin was recently shown to induce HbF in a FOXO3-dependent manner that was additive to hydroxyurea in vitro.40 Metformin is being investigated in a phase I, dose-escalation pilot study in SCD patients with or without the addition of hydroxyurea (NCT02981329).

Targeting HbS polymerization through sickle red blood cell hydration Sickle RBC are naturally imperfect osmometers,41 and their biomechanical properties are dependent on RBCintrinsic properties (e.g., membrane and cytosolic contents) and RBC-extrinsic factors (e.g., environmental osmolality, surface area to volume ratio, oxygen tension).8 As such, SCD pathology results directly from the consequences of red cell dehydration, which increases HbS concentration within the sickle RBC, leading to polymerization under deoxygenated conditions and resultant increased cellular density and stiffness, which reduces sickle RBC deformability and increases adhesion leading to a disruption in microvascular blood flow.5,8,42,43 Theoretically, sickle RBC dehydration is ‘targetable’ and could lead to clinical improvement, but studies of agents directed at cation transport and RBC hydration mechanisms have, to date, been disappointing. While early studies suggested dipyridamole may inhibit sickling-induced cation transport and inhibit sickle RBC dehydration,44 a phase II study (NCT00276146) investigating its use was closed because of poor enrollment. A phase III, randomized, double-blind study of senicapoc (NCT00102791), a small-molecule inhibitor of Ca2+-activated K+ efflux (Gardos channel), reduced sickle RBC dehydration and hemolysis,45 which may be pathophysiologically important. However, the study was terminated early because of lack of improvement in vaso-occlusive event rates in adults with SCD treated with this drug in comparison to those given placebo. Early observations also found magnesium supplementation improved sickle RBC hydration in patients with SCD through inhibition of K-Cl co-transport and reduction of dense sickle RBC.46 However, in the Magnesium for Children in Crisis (MAGiC) study (NCT01197417), which included young adults up to 21 years of age, when compared to the effects of placebo, intravenous magnesium did not reduce duration of stay or opioid use in patients hospitalized for vaso-occlusive events.47

Other anti-sickling agents Allosteric modification of HbS from the low-oxygen affinity tense (T) state to the high-oxygen affinity relaxed (R) state reduces the risk of HbS polymer formation.29 Several novel agents manipulate this biochemical phenomenon and show promise in interrupting the molecular pathogenesis of SCD. Voxelotor (GBT440) is a small molecule that binds to the a-globin chain in HbS, increasing oxygen affinity to favor the R state. Early phase trials haematologica | 2019; 104(9)

demonstrated that GBT440 is well-tolerated in adults with SCD and reduces morphological changes in sickle RBC.48,49 The Hemoglobin Oxygen Affinity Modulation to Inhibit HbS Polymerization (HOPE) study (NCT03036813) is a phase III randomized, double-blind, placebo-controlled, multicenter study evaluating the efficacy and safety of GBT440 in adolescents and adults with SCD. Early results suggest treatment with GBT440 increases hemoglobin levels in patients with SCD who have decreased markers of hemolysis when compared to patients given placebo.50 Importantly, initial evaluations did not suggest physiological oxygen deprivation with GBT440 (i.e. erythropoietin levels do not change); studies of oxygen delivery to the brain and during exercise in patients taking GBT440 are reportedly in process (unpublished, ASH 2018). AES-103 is a naturally occurring small molecule (a 5-hydroxymethyl furfural) that also binds to the a-globin subunit in HbS, increasing oxygen affinity and favoring the R state. AES103 has completed a phase I trial (NCT01597401) and was shown to be well-tolerated in a double-blind, placebocontrolled dose-escalation trial in adults with SCD.51 PEGylated bovine carboxyhemoglobin (PEG-bHb-CO; SANGUINATETM) is a combined oxygen transfer and carbon monoxide-releasing molecule that is given intravenously with the rationale of reducing hypoxemiainduced sickle RBC pathology. It was found to be safe and well-tolerated in a phase I trial among adults with SCD.52,53 A prospective, randomized single-dose, placebo-controlled phase II study (NCT02411708) was recently completed and interim results showed that PEG-bHb-CO administration was associated with reduced pain scores and anti-sickling properties during vaso-occlusive episodes when compared to placebo administration.54 SCD-101 is marketed as a ‘botanical drug’ with anti-sickling activity through an unknown mechanism and was shown in an interim analysis of a recent phase I study (NCT02380079) to be well-tolerated, and may reduce chronic pain and fatigue, improve leg ulcers, and improve sickle RBC morphology in the peripheral blood.55 The kinetics of some of the allosteric-acting agents are much faster (hours to days) than are the predominant effects of HbF inducers, which may take weeks to months to fully alter erythroid precursors. This difference could suggest complementary ways for using these agents.

Targeting intracellular sickle red blood cell oxidative changes (antioxidant therapy) RBC are oxygen carriers, which places them in constant danger from the cumulative impact of reactive oxygen species and free radicals formed by oxygen and hemoglobin metabolism inside the RBC.56 Sickle RBC, due to their unique intracellular milieu with high concentrations of HbS forming and re-forming polymers, are at increased risk of oxidative damage. L-glutamine is an amino acid and a precursor used in the synthesis of glutathione and reduced nicotinamide adenine nucleotide diphosphate, which can protect the sickle RBC from oxidative damage.57 A recent multicenter, randomized, placebo-controlled, double-blind, phase III trial (NCT01179217) found that oral administration of pharmaceutical-grade L-glutamine (EndariTM) reduced the number of vaso-occlusive events and episodes of acute chest syndrome, and hospital admissions compared to those in placebo-treated children and adults with SCD.58 In 2017, EndariTM received orphan drug status and is FDA-approved to reduce acute compli1713

M.C. Carden and J. Little et al.

cations of SCD in patients 5 years or older. However, there remain concerns among many providers regarding the limitations of the study leading to approval for EndariTM, along with its potentially prohibitive cost, limited insurance coverage, and twice-daily powder form, which may reduce adherence.57,59 Another antioxidant agent, N-acetylcysteine, maintains and replenishes glutathione, which is an intracellular antioxidant and scavenger of reactive oxygen species. In a phase II trial, Nacetylcysteine reduced vaso-occlusive episodes and dense sickle RBC formation.60 However, a placebo-controlled, phase III trial (NCT01849016) found N-acetylcysteine had no clinical benefit in reducing pain when given orally, albeit adherence was poor.61 Of note, N-acetylcysteine can also reduce the formation of von Willebrand factor multimers and of von Willebrand factor-dependent platelet aggregation.62 von Willebrand factor reactivity is high during vaso-occlusive episodes, and may sustain microvascular vaso-occlusion.63 A phase I/II trial of Nacetylcysteine administered intravenously during vasoocclusive episodes (NCT01800526) is recruiting adult patients to determine whether this antioxidant can affect von Willebrand factor levels or function and curb pain associated with vaso-occlusion.

Red blood cell extrinsic targets The disease-modifying approaches to SCD that target factors extrinsic to the sickle red blood cells, namely abnormal cellular adhesion and vascular dysfunction, platelet activation and hypercoagulability, and leukocytes, cytokines and other inflammatory mediators are outlined in Online Supplementary Table S2A-C, respectively, and discussed below.

episodes. In a randomized phase II study, crizanlizumab was tested internationally in 198 people with SCD. Compared with placebo, higher dose crizanlizumab resulted in a 45% reduction in annual crises, from 2.98/year to 1.63/year (P=0.01). In addition, the median time to a first crisis was longer in people with SCD who were on high-dose crizanlizumab than in those on placebo (4.07 vs. 1.38 months, respectively; P=0.001). Serious adverse events did not differ between patients treated with the active drug or placebo. However, normal surveillance for infection and platelet function rely on intact function of P-selectin, and this aspect will need to be monitored during more widespread use of this preventive therapy. In a phase II study, 76 people with SCD were treated with intravenous rivapansel or placebo during vaso-occlusive episodes. There were trends toward reductions in mean and median times to resolution of vaso-occlusive episodes in treated patients [41 h and 63 h, respectively (28% and 48% reductions in the mean and median time to resolution, respectively); P=0.19 for both]. These reductions were more substantial than those in the placebotreated group. A secondary endpoint, cumulative intravenous opioid use, was reduced by 83% with GMI-1070 versus placebo (P=0.010). These results suggest that this agent has some efficacy during vaso-occlusive episodes.76 Both crizanlizumab (NCT03814746) and rivipansel (NCT02433158) are now undergoing phase III studies. Intravenous immunoglobulins decrease cellular adhesion in SCD in vitro, likely due to effects on RBC-white blood cell adhesion mediated through the integrin Mac1.77 This observation formed the rationale for a phase I trial of the use of intravenous immunoglobulins in pediatric and adult patients with SCD,78 while a phase II study is currently ongoing only in children (NCT01757418).

Vascular dysfunction

Targeting abnormal cellular adhesion and vascular dysfunction Abnormal cellular adhesion Pioneering work in the 1980s showed that intracellular hemoglobin polymerization in SCD resulted in abnormal RBC adhesion to the endothelium.64 This observation was soon expanded and enhanced by thoughtful investigations, and it is now recognized that many cell types, endothelial and hematopoietic, show abnormal activation and adhesion in SCD. Further, precise experimental identification of adhesive partners, such as integrins,65,66 blood group antigens,67 selectins,68-70 and white cell proteins71 have increased the repertoire of potential therapeutic targets (Online Supplementary Table S2A). BCAM/Lu, expressed on sickle RBC, mediates adhesion to the sub-endothelial protein laminin, and this is augmented by β−adrenergic signaling.72 The effect of β−blockade by propranolol on RBC adhesion and clinical outcomes was tested in people with SCD, with suggestive but inconclusive results (NCT01077921).73 Abnormal cellular adhesion to the endothelium has been shown to be mediated by P- and E-selectins, and early work showed some benefit from an oral agent that blocked P-selectin.74 More recently, the most actively tested agents are crizanlizumab,75 which is an anti-P-selectin monoclonal antibody given prophylactically monthly, and rivipansel, which is an intravenous glycomimetic panselectin antagonist given acutely during vaso-occlusive 1714

People with homozygous SCD experience a lifelong risk of multi-organ vasculopathy, due to the cumulative effects of hemolysis, nitric oxide depletion, inflammation and abnormal cellular adhesion.79 Therapeutic strategies including repletion of nitric oxide via inhalation have not been successful in randomized controlled studies during acute pain episodes or acute chest syndrome,80,81 but when given topically it has shown some quantitative and qualitative success in healing leg ulcers.82,83 A prospective phase II study to test topical nitric oxide as a treatment for leg ulcers is ongoing (NCT02863068). Dietary supplementation to improve nitric oxide availability, including arginine and its precursor citrulline, has also been tried. Citrulline was studied in a small number of people with SCD, and appeared to have some benefit,84 but is not actively under study currently. Arginine has had a more robust clinical history. Thirty-eight children and young adults up to 19 years old with SCD received arginine or placebo intravenously for 5 days during a hospital admission for vaso-occlusive crisis. A significant reduction in opioid use was reported (1.9 ± 2.0 mg/kg vs. 4.1 ± 4.1 mg/kg, respectively; P=0.02) and lower pain scores at discharge (1.9 ± 2.4 vs. 3.9 ± 2.9, respectively; P=0.01).85 A larger phase II study is near completion at Emory in Atlanta (NCT02536170). Targets within the nitric oxide signaling pathway have also been identified and are being tested in clinical trials. Soluble guanylate cyclase catalyzes the production of cyclic GMP, which promotes vascular health.86 One agent, haematologica | 2019; 104(9)

Disease-modifying therapies for SCD

Riociguat®, has been shown to improve function (increased 6-minute walk distance) and decrease vascular resistance in patients with pulmonary hypertension without SCD (NCT00855465).87 The impact of this agent on pain and cardiopulmonary function is currently being tested in 100 people with SCD at multiple sites, in the SterioSCD study (NCT02633397). As a class of drugs, ‘statins’ decrease systemic inflammation and improve vascular health in the general population. Twenty-five people with SCD were treated with low-dose atorvastatin for 1 month: changes were observed in cholesterol and some markers of vascular health, but no conclusive findings were made.88 In a dosefinding study, performed in 26 people with SCD over 3 weeks, simvastatin was safe and increased nitric oxide levels, while decreasing markers of inflammation (C-reactive protein and interleukin-6).89 A 3-month follow-up study in 19 people with SCD showed an excellent safety profile, an improvement in the rate but not the intensity of pain, and salutary effects on some but not all markers of inflammation and vascular health.90 This class of agents is not being actively tested at this time, but its excellent safety profile suggests that it may have a useful role in the management of selected patients with SCD.

Targeting platelet activation and hypercoagulability SCD is a hypercoagulable state with an incidence of thromboembolism comparable to that in individuals with some inherited thrombophilias.12,13,91-93 Overactivation of hemostatic components, including platelets and coagulation factors, contributes to the vasculopathy of SCD through increased endothelial activation, platelet-leukocyte aggregates, and increased inflammation.13,94-96 As such, these factors are seen as potential targets for novel pharmaceutical interventions (Online Supplementary Table S2B). Recent results of studies investigating antiplatelet agents targeting GPIIb/IIIa have been disappointing. A singlecenter, phase II, randomized, double-blind, placebo-controlled study (NCT00834899) found the reversible GPIIb/IIIa inhibitor eptifibatide to be safe but it did not improve time to resolution of vaso-occlusive episodes or hospital discharge.97 A secondary analysis of these data did, however, reveal that eptifibatide reduced ADP-dependent platelet aggregation, platelet-leukocyte aggregates, and inflammatory cytokines.98 Another GPIIb/IIIainhibitor, abciximab, was withdrawn from a patient-oriented study due to insufficient recruitment (NCT01932554). Prasugrel, an irreversible inhibitor of platelet aggregation through the P2Y12 class of ADP receptors has also been studied. A phase II, double-blind, randomized, multicenter trial (NCT01167023) found that prasugrel was safe, reduced markers of platelet activation including P-selectin, and was associated with a trend toward decreased pain when compared to placebo.99 While it was a negative study and patients 18 years or older were not included, the seminal phase III Determining Effects of Platelet Inhibition on VasoOcclusive Events (DOVE) trial was still informative as one of the largest multinational phase III trials in pediatric SCD and found that prasugrel did not reduce the rate of vaso-occlusive episodes in pediatric and adolescent patients up to 17 years of age when compared to placebo.100 In addition, a phase II study (HESTIA2, NCT02482298) assessing the impact of ticagrelor, a reversible P2Y12 antagonist, found that the drug was safe haematologica | 2019; 104(9)

but did not reduce the number of diary-reported pain-free days in adults with SCD.101 While SCD is a procoagulant disorder, with chronic activation of the coagulation system through increased thrombin generation and diminished anticoagulant proteins,12,102 the clinical benefit of routine anticoagulant use in SCD is still largely unknown. For instance, warfarin, a vitamin K antagonist, has been associated with decreased D-dimer levels in adult patients with vaso-occlusive episodes.103 However, a recent phase II study (NCT01036802) evaluating its use in patients with pulmonary hypertension, which in some adults may be due to in-situ thrombosis formation within the pulmonary vasculature, was terminated early due to poor accrual. Similarly, a phase II feasibility study (NCT02098993) investigating unfractionated heparin in acute chest syndrome was also terminated due to poor enrollment. Interestingly, a low-molecular weight heparin, tinzaparin, shortened the durations of vaso-occlusive episodes and hospitalization compared to those in placebo-treated patients in a randomized, double-blind trial.104 A phase III study (NCT02580773) evaluating the effectiveness of tinzaparin on time to resolution of acute chest syndrome is currently recruiting participants. A randomized, doubleblind, phase II study (NCT01419977) evaluating prophylactic dosing of a low molecular weight heparin, dalteparin, during vaso-occlusive episodes has been completed and preliminary results indicate an insignificant impact on markers of coagulation, while reducing pain scores in hospitalized patients given the trial drug compared to those given placebo.105 Novel anti-Xa oral agents, too, are under investigation. A phase II study (NCT02072668) evaluating the impact of rivaroxaban on inflammatory markers during the non-crisis steady state has recently been completed. A phase III, placebo-controlled trial (NCT02179177) investigating the effect of prophylactic dosing of apixaban on daily pain scores is recruiting. Defibrotide, a sodium salt of a single-stranded polydeoxyribonucleotide with anti-thrombotic and antiinflammatory properties, is currently being evaluated primarily for safety (grade III/IV allergic reaction or hemorrhage) in a phase II study (NCT03805581) among adult patients with SCD and acute chest syndrome.

Targeting leukocytes, cytokines, and other inflammatory mediators Patients with SCD are in a constant inflammatory state primarily due to vaso-occlusion-induced hypoxia-reperfusion, endothelial damage, and activated and aging leukocytes, in part regulated by the microbiome.10,106-108 As such, there has been a growing interest among investigators interested in targeting these inflammatory pathways for potential clinical benefit in adult patients with SCD (Online Supplementary Table S2C). Use of inhaled mometasone, a corticosteroid, is being tested in two phase II studies (NCT02061202 – IMPROVE, NCT03758950 – IMPROVE2) to determine its effectiveness in reducing pain and inflammation among patients without asthma. Recent results indicate that the use of inhaled mometasone is feasible and that it can reduce circulating soluble vascular cell adhesion molecule and markers of macrophage activation, while improving daily pain scores.109,110 Invariant natural killer T (iNKT) cell activation is increased in patients with SCD and is regulated by the adenosine A2A receptor.111 Regadenoson, a partially selective adenosine A2A receptor 1715

M.C. Carden and J. Little et al.

Figure 3. Risk-benefit assessments for emerging non-genetic therapies in sickle cell disease. The figure is a visual representation of risk-benefit analyses relative to the proximity of the novel therapy to red blood cell intrinsic targets, such as HbF or polymer formation, which are likely to have the greatest impact on disease modification (green shading). Risk tolerance (red shading) may be greater for treatments that modify disease. Downstream targets may offer significant palliation and may lengthen life, but are primarily management tools and may not warrant as great a risk. We propose a shifting level of risk tolerance (black arrow) based on the projected impact of a therapy.

agonist, is a coronary vasodilator and approved by the FDA for myocardial perfusion imaging. A phase I study (NCT01085201) demonstrated that regadenoson can be safely administered to patients during vaso-occlusive episodes and reduces iNKT cell activation.112 A randomized phase II, placebo-controlled trial (NCT01788631) among patients with vaso-occlusive episodes demonstrated low-dose infusion of regadenoson did not significantly reduce iNKT cell activation, duration of hospital stay, mean total opioid use, or pain scores when compared to placebo.113 NKTT120, a humanized anti-iNKT cell monoclonal antibody, recently completed a phase I study (NCT01783691) and was found to be safe and produced rapid, sustained iNKT cell depletion in adults with SCD.114 Other novel anti-inflammatory agents are also being investigated in trials among patients with SCD. ACZ885 (canakinumab) is a monoclonal antibody that targets interleukin-1Î˛, a cytokine upregulated due to hemolysisinduced free heme and inflammasome activation.115 A phase II trial (NCT02961218) is recruiting adolescent and young adult patients to determine whether canakinumab can reduce daily pain scores. Leukotrienes are biologically active inflammatory mediators produced by leukocytes which are associated with SCD-related morbidity, including asthma.116 Zileuton inhibits 5-lipoxygenase, a leukotriene synthesis enzyme, and was safe and tolerable in a phase I trial (NCT01136941).117 Montelukast, a cysteinyl leukotriene receptor antagonist, is FDA-approved for the treatment of asthma. The aim of a recently com1716

pleted phase II study (NCT01960413) among adolescents and adults with SCD was to determine whether montelukast versus placebo added to hydroxyurea could improve markers of tissue injury associated with vasoocclusion. Results are pending. Omega-3 fatty acids may improve SCD-related pathology through reduction in vaso-occlusion-induced systemic and local inflammation.118 A phase I/II study (NCT02947100) was terminated early due to manufacturing problems. A phase III study (NCT02525107) is recruiting patients to determine whether prophylactic administration of omega-3 fatty acids can reduce the number and severity of vaso-occlusive episodes. Lastly, intriguing research suggests that the microbiome may play a significant role in the pathology of SCD through mechanisms involving translocation of gut flora and inflammation which can affect sickle RBC and leukocytes.108,119 In a single-arm, phase II study (NCT03719729) investigators are recruiting patients with SCD to determine whether gut decontamination with rifaximin, a broad-spectrum antibiotic that inhibits bacterial RNA polymerase, is well-tolerated and can reduce admissions due to vaso-occlusive episodes.

Discussion While stem cell and gene therapies are becoming more commonplace, they are not yet widely available to most haematologica | 2019; 104(9)

Disease-modifying therapies for SCD

patients with SCD. Therefore, optimizing non-curative approaches, i.e. those that do not involve stem cell or gene therapy, but which prevent or abort the complications of SCD, remains an important step in improving health and diminishing symptom burden in most people with SCD, in the USA and worldwide. Collaboration among the government (National Institutes of Health), industry, and academia has led to the development of a range of novel therapies that target many pathways which have been implicated in SCD-related pathophysiology. Through successful enrollment in numerous studies investigating novel therapies, it is also clear that many patients with SCD are eager to explore the potential clinical benefits of these agents. Less clear, unfortunately, are the optimal endpoints that investigators should choose when determining the beneficial role these agents may have in the clinical course and complications of SCD. Risk-benefit analyses by patients, their families, and their healthcare providers are also important. We suggest that treatments â&#x20AC;&#x2DC;closerâ&#x20AC;&#x2122; to the proximate pathophysiological causes of SCD, such as polymer formation, may warrant greater risk exposure than do strategies that solely address downstream consequences (Figure 3), although this assessment will be fur-

References 1. Pauling L, Itano HA, et al. Sickle cell anemia a molecular disease. Science. 1949;110 (2865):543-548. 2. Piel FB, Steinberg MH, Rees DC. Sickle cell disease. N Engl J Med. 2017;377(3):305. 3. Kato GJ, Piel FB, Reid CD, et al. Sickle cell disease. Nat Rev Dis Primers. 2018;4:18010. 4. Eaton WA, Hofrichter J. Hemoglobin S gelation and sickle cell disease. Blood. 1987;70(5):1245-1266. 5. Kaul DK, Fabry ME, Windisch P, Baez S, Nagel RL. Erythrocytes in sickle cell anemia are heterogeneous in their rheological and hemodynamic characteristics. J Clin Invest. 1983;72(1):22-31. 6. Hebbel RP. Beyond hemoglobin polymerization: the red blood cell membrane and sickle disease pathophysiology. Blood. 1991;77(2): 214-237. 7. Kuross SA, Rank BH, Hebbel RP. Excess heme in sickle erythrocyte inside-out membranes: possible role in thiol oxidation. Blood. 1988;71(4):876-882. 8. Barabino GA, Platt MO, Kaul DK. Sickle cell biomechanics. Annu Rev Biomed Eng. 2010;12:345-367. 9. Bunn HF. Pathogenesis and treatment of sickle cell disease. N Engl J Med. 1997;337(11):762-769. 10. Zhang D, Xu C, Manwani D, Frenette PS. Neutrophils, platelets, and inflammatory pathways at the nexus of sickle cell disease pathophysiology. Blood. 2016;127(7):801809. 11. Sparkenbaugh E, Pawlinski R. Interplay between coagulation and vascular inflammation in sickle cell disease. Br J Haematol. 2013;162(1):3-14. 12. Sparkenbaugh E, Pawlinski R. Prothrombotic aspects of sickle cell disease. J Thromb Haemost. 2017;15(7):1307-1316. 13. Noubouossie D, Key NS, Ataga KI. Coagulation abnormalities of sickle cell disease: relationship with clinical outcomes

haematologica | 2019; 104(9)

14.

15. 16.

17. 18.

19. 20.

21.

22.

23. 24.

ther complicated when multi-modality therapies are applied. In addition, pain is such a prominent symptom and burden in SCD that its relief may warrant riskier approaches than would otherwise be considered for a single symptom or organ. It is a time of great optimism, and in the future it may be feasible to take a multimodal approach with a combination of therapies, as seen in cancer treatment regimens, to adequately address both cause and effect in SCD. For example, agents that improve red cell health, such as Lglutamine and hydroxyurea, could be successfully combined with non-overlapping agents, such as anti-selectin therapies, which block inflammatory adhesion during painful crises. As members of the broader sickle cell community (while optimistic and enthusiastic), we must remain vigilant but reasonable, using long-term, prospective data as available to guide assessments about benefits and long-term safety of emerging treatments. We must do better in the coming years to identify complications of all novel therapies rapidly so that we can ensure that (unavoidable) risks and complications are detected, mitigated, and managed in the context of their impact on disease in order to optimize care in SCD.

and the effect of disease modifying therapies. Blood Rev. 2016;30(4):245-256. Platt OS, Brambilla DJ, Rosse WF, et al. Mortality in sickle cell disease. Life expectancy and risk factors for early death. N Engl J Med. 1994;330(23):1639-1644. Telen MJ. Beyond hydroxyurea: new and old drugs in the pipeline for sickle cell disease. Blood. 2016;127(7):810-819. Moerdler S, Manwani D. New insights into the pathophysiology and development of novel therapies for sickle cell disease. Hematology Am Soc Hematol Educ Program. 2018;2018(1):493-506. Kapoor S, Little JA, Pecker LH. Advances in the treatment of sickle cell disease. Mayo Clin Proc. 2018;93(12):1810-1824. Hoban MD, Orkin SH, Bauer DE. Genetic treatment of a molecular disorder: gene therapy approaches to sickle cell disease. Blood. 2016;127(7):839-848. Walters MC. Update of hematopoietic cell transplantation for sickle cell disease. Curr Opin Hematol. 2015;22(3):227-233. Yawn BP, Buchanan GR, Afenyi-Annan AN, et al. Management of sickle cell disease: summary of the 2014 evidence-based report by expert panel members. JAMA. 2014;312(10):1033-1048. Telen MJ, Malik P, Vercellotti GM. Therapeutic strategies for sickle cell disease: towards a multi-agent approach. Nat Rev Drug Discov. 2019;18(2):139-158. Fitzhugh CD, Abraham AA, Tisdale JF, Hsieh MM. Hematopoietic stem cell transplantation for patients with sickle cell disease: progress and future directions. Hematol Oncol Clin North Am. 2014;28(6):1171-1185. Orkin SH, Bauer DE. Emerging genetic therapy for sickle cell disease. Annu Rev Med. 2019;70:257-271. Morrone K, Mitchell WB, Manwani D. Novel sickle cell disease therapies: targeting pathways downstream of sickling. Semin Hematol. 2018;55(2):68-75.

25. Field JJ, Nathan DG. Advances in sickle cell therapies in the hydroxyurea era. Mol Med. 2014;20 Suppl 1:S37-42. 26. Strouse J. Sickle cell disease. Handb Clin Neurol. 2016;138:311-324. 27. Sundd P, Gladwin MT, Novelli EM. Pathophysiology of sickle cell disease. Annu Rev Pathol. 2019;14:263-292. 28. Ataga KI, Desai PC. Advances in new drug therapies for the management of sickle cell disease. Expert Opin Orphan Drugs. 2018;6(5):329-343. 29. Eaton WA, Bunn HF. Treating sickle cell disease by targeting HbS polymerization. Blood. 2017;129(20):2719:2726. 30. Platt OS. Hydroxyurea for the treatment of sickle cell anemia. N Engl J Med. 2008;358(13):1362-1369. 31. McGann PT, Ware RE. Hydroxyurea therapy for sickle cell anemia. Expert Opin Drug Saf. 2015;14(11):1749-1758. 32. Paikari A, Sheehan VA. Fetal haemoglobin induction in sickle cell disease. Br J Haematol. 2018;180(2):189-200. 33. Ginder GD. Epigenetic regulation of fetal globin gene expression in adult erythroid cells. Transl Res. 2015;165(1):115-125. 34. Molokie R, Lavelle D, Gowhari M, et al. Oral tetrahydrouridine and decitabine for non-cytotoxic epigenetic gene regulation in sickle cell disease: a randomized phase 1 study. PLoS Med. 2017;14(9):e1002382. 35. Reid ME, El Beshlawy A, Inati A, et al. A double-blind, placebo-controlled phase II study of the efficacy and safety of 2,2dimethylbutyrate (HQK-1001), an oral fetal globin inducer, in sickle cell disease. Am J Hematol. 2014;89(7):709-713. 36. Bradner JE, Mak R, Tanguturi SK, et al. Chemical genetic strategy identifies histone deacetylase 1 (HDAC1) and HDAC2 as therapeutic targets in sickle cell disease. Proc Natl Acad Sci U S A. 2010;107(28):1261712622. 37. Dulmovits BM, Appiah-Kubi AO, Papoin J, et al. Pomalidomide reverses gamma-globin

1717

M.C. Carden and J. Little et al.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

1718

silencing through the transcriptional reprogramming of adult hematopoietic progenitors. Blood. 2016;127(11):1481-1492. Kutlar A, Swerdlow PS, Meiler SE, et al. Pomalidomide in sickle cell disease: phase I study of a novel anti-switching agent. Blood. 2013;122(21):777. Cui S, Lim KC, Shi L, et al. The LSD1 inhibitor RN-1 induces fetal hemoglobin synthesis and reduces disease pathology in sickle cell mice. Blood. 2015;126(3):386-396. Zhang Y, Paikari A, Sumazin P, et al. Metformin induces FOXO3-dependent fetal hemoglobin production in human primary erythroid cells. Blood. 2018;132(3):321-333. Lew VL, Bookchin RM. Ion transport pathology in the mechanism of sickle cell dehydration. Physiol Rev. 2005;85(1):179200. Nash GB, Johnson CS, Meiselman HJ. Mechanical properties of oxygenated red blood cells in sickle cell (HbSS) disease. Blood. 1984;63(1):73-82. Zhou GP, Anderson KP, Joiner CH, Gallagher PG. Modification of erythrocyte hydration in the treatment of sickle cell disease. Blood Cells Mol Dis. 2001;27(1):65-68. Joiner CH, Jiang M, Claussen WJ, Roszell NJ, Yasin Z, Franco RS. Dipyridamole inhibits sickling-induced cation fluxes in sickle red blood cells. Blood. 2001;97(12):3976-3983. Ataga KI, Reid M, Ballas SK, et al. Improvements in haemolysis and indicators of erythrocyte survival do not correlate with acute vaso-occlusive crises in patients with sickle cell disease: a phase III randomized, placebo-controlled, double-blind study of the Gardos channel blocker senicapoc (ICA17043). Br J Haematol. 2011;153(1):92-104. De Franceschi L, Bachir D, Galacteros F, et al. Oral magnesium supplements reduce erythrocyte dehydration in patients with sickle cell disease. J Clin Invest. 1997;100(7):18471852. Brousseau DC, Scott JP, Badaki-Makun O, et al. A multicenter randomized controlled trial of intravenous magnesium for sickle cell pain crisis in children. Blood. 2015;126(14): 1651-1657. Lehrer-Graiwer J, Howard J, Hemmaway CJ, et al. GBT440, a potent anti-sickling hemoglobin modifier reduces hemolysis, improves anemia and nearly eliminates sickle cells in peripheral blood of patients with sickle cell disease. Blood. 2015;126(23):542. Howard J, Hemmaway CJ, Telfer P, et al. A phase 1/2 ascending dose study and openlabel extension study of voxelotor in patients with sickle cell disease. Blood. 2019;133(17):1865-1875. Vichinsky E, Hoppe C, Howard J, et al. Results from part A of the Hemoglobin Oxygen Affinity Modulation to Inhibit HbS Polymerization (HOPE) trial (GBT440-031), a placebo-controlled randomized study evaluating voxelotor (GBT440) in adults and adolescents with sickle cell disease. Blood. 2018;132(Suppl 1):505. Kato G, Lawrence MP, Mendelsohn L, et al. Phase 1 clinical trial of the candidate antisickling agent Aes-103 in adults with sickle cell anemia. Blood. 2013;122(21):1009. Abuchowski A. PEGylated Bovine carboxyhemoglobin (SANGUINATE): results of clinical safety testing and use in patients. Adv Exp Med Biol. 2016;876:461-467. Misra H, Bainbridge J, Berryman J, et al. A phase Ib open label, randomized, safety study of SANGUINATE in patients with sickle cell anemia. Rev Bras Hematol Hemoter. 2017;39(1):20-27. Buontempo P, Jubin RG, Buontempo C, et al.

55.

56.

57. 58. 59. 60.

61.

62.

63.

64.

65.

66.

67. 68.

69. 70. 71.

72.

73.

Pegylated carboxyhemoglobin bovine (SANGUINATETM) restores RBCs roundness and reduces pain during a sickle cell vaso-occlusive crisis Blood. 2017;130(Suppl 1):969. Swift R, Abdulmalik O, Chen Q, et al. SCD101: a new anti-sickling drug reduces pain and fatigue and improves red blood cell shape in peripheral blood of patients with sickle cell disease. Blood. 2016;128(22):121. Jagadeeswaran R, Rivers A. Evolving treatment paradigms in sickle cell disease. Hematology Am Soc Hematol Educ Program. 2017;2017(1):440-446. Quinn CT. l-Glutamine for sickle cell anemia: more questions than answers. Blood. 2018;132(7):689-693. Niihara Y, Miller ST, Kanter J, et al. A phase 3 trial of l-glutamine in sickle cell disease. N Engl J Med. 2018;379(3):226-235. Minniti CP. l-glutamine and the dawn of combination therapy for sickle cell disease. N Engl J Med. 2018;379(3):292-294. Pace BS, Shartava A, Pack-Mabien A, Mulekar M, Ardia A, Goodman SR. Effects of N-acetylcysteine on dense cell formation in sickle cell disease. Am J Hematol. 2003;73(1):26-32. Sins JWR, Fijnvandraat K, Rijneveld AW, et al. Effect of N-acetylcysteine on pain in daily life in patients with sickle cell disease: a randomised clinical trial. Br J Haematol. 2018;182(3):444-448. Chen J, Reheman A, Gushiken FC, et al. Nacetylcysteine reduces the size and activity of von Willebrand factor in human plasma and mice. J Clin Invest. 2011;121(2):593-603. Sins JWR, Schimmel M, Luken BM, et al. Dynamics of von Willebrand factor reactivity in sickle cell disease during vaso-occlusive crisis and steady state. J Thromb Haemost. 2017;15(7):1392-1402. Hebbel RP, Boogaerts MA, Eaton JW, Steinberg MH. Erythrocyte adherence to endothelium in sickle-cell anemia. A possible determinant of disease severity. N Engl J Med. 1980;302(18):992-995. Takehara K, Soma Y, Igarashi A, Kikuchi K, Moro A, Ishibashi Y. Response of scleroderma fibroblasts to various growth factors. Arch Dermatol Res. 1991;283(7):461-464. Chen G, Chang J, Zhang D, Pinho S, Jang JE, Frenette PS. Targeting Mac-1-mediated leukocyte-RBC interactions uncouples the benefits for acute vaso-occlusion and chronic organ damage. Exp Hematol. 2016;44(10):940-946. Telen MJ. Erythrocyte adhesion receptors: blood group antigens and related molecules. Transfus Med Rev. 2005;19(1):32-44. Telen MJ. Cellular adhesion and the endothelium: E-selectin, L-selectin, and panselectin inhibitors. Hematol Oncol Clin North Am. 2014;28(2):341-354. Okpala I. Investigational selectin-targeted therapy of sickle cell disease. Expert Opin Investig Drugs. 2015;24(2):229-238. Kutlar A, Embury SH. Cellular adhesion and the endothelium: P-selectin. Hematol Oncol Clin North Am. 2014;28(2):323-339. Koehl B, Nivoit P, El Nemer W, et al. The endothelin B receptor plays a crucial role in the adhesion of neutrophils to the endothelium in sickle cell disease. Haematologica. 2017;102(7):1161-1172. Hines PC, Zen Q, Burney SN, et al. Novel epinephrine and cyclic AMP-mediated activation of BCAM/Lu-dependent sickle (SS) RBC adhesion. Blood. 2003;101(8):32813287. De Castro LM, Zennadi R, Jonassaint JC, Batchvarova M, Telen MJ. Effect of propra-

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

nolol as antiadhesive therapy in sickle cell disease. Clin Transl Sci. 2012;5(6):437-444. Kutlar A, Ataga KI, McMahon L, et al. A potent oral P-selectin blocking agent improves microcirculatory blood flow and a marker of endothelial cell injury in patients with sickle cell disease. Am J Hematol. 2012;87(5):536-539. Ataga KI, Kutlar A, Kanter J, et al. Crizanlizumab for the prevention of pain crises in sickle cell disease. N Engl J Med. 2017;376(5):429-439. Telen MJ, Wun T, McCavit TL, et al. Randomized phase 2 study of GMI-1070 in SCD: reduction in time to resolution of vaso-occlusive events and decreased opioid use. Blood. 2015;125(17):2656-2664. Jang JE, Hidalgo A, Frenette PS. Intravenous immunoglobulins modulate neutrophil activation and vascular injury through FcgammaRIII and SHP-1. Circ Res. 2012;110(8):1057-1066. Manwani D, Chen G, Carullo V, et al. Single-dose intravenous gammaglobulin can stabilize neutrophil Mac-1 activation in sickle cell pain crisis. Am J Hematol. 2015;90(5):381-385. Kato GJ, Steinberg MH, Gladwin MT. Intravascular hemolysis and the pathophysiology of sickle cell disease. J Clin Invest. 2017;127(3):750-760. Maitre B, Djibre M, Katsahian S, et al. Inhaled nitric oxide for acute chest syndrome in adult sickle cell patients: a randomized controlled study. Intensive Care Med. 2015;41(12):2121-2129. Gladwin MT, Kato GJ, Weiner D, et al. Nitric oxide for inhalation in the acute treatment of sickle cell pain crisis: a randomized controlled trial. JAMA. 2011;305(9):893-902. Minniti CP, Gorbach AM, Xu D, et al. Topical sodium nitrite for chronic leg ulcers in patients with sickle cell anaemia: a phase 1 dose-finding safety and tolerability trial. Lancet Haematol. 2014;1(3):e95-e103. Connor JL, Jr., Sclafani JA, Kato GJ, Hsieh MM, Minniti CP. Brief topical sodium nitrite and its impact on the quality of life in patients with sickle leg ulcers. Medicine (Baltimore). 2018;97(46):e12614. Waugh WH, Daeschner CW, 3rd, Files BA, McConnell ME, Strandjord SE. Oral citrulline as arginine precursor may be beneficial in sickle cell disease: early phase two results. J Natl Med Assoc. 2001;93(10):363371. Morris CR, Kuypers FA, Lavrisha L, et al. A randomized, placebo-controlled trial of arginine therapy for the treatment of children with sickle cell disease hospitalized with vaso-occlusive pain episodes. Haematologica. 2013;98(9):1375-1382. Stasch JP, Evgenov OV. Soluble guanylate cyclase stimulators in pulmonary hypertension. Handb Exp Pharmacol. 2013;218:279313. Narang BK, Roy S, Sharma R, Singh V, Rawal RK. Riociguat as a treatment regime for pulmonary arterial hypertension: a review. Clin Exp Hypertens. 2015;37(4):323-331. Bereal-Williams C, Machado RF, McGowan V 2nd, et al. Atorvastatin reduces serum cholesterol and triglycerides with limited improvement in vascular function in adults with sickle cell anemia. Haematologica. 2012;97(11):1768-1770. Hoppe C, Kuypers F, Larkin S, Hagar W, Vichinsky E, Styles L. A pilot study of the short-term use of simvastatin in sickle cell disease: effects on markers of vascular dysfunction. Br J Haematol. 2011;153(5):655663.

haematologica | 2019; 104(9)

Disease-modifying therapies for SCD 90. Hoppe C, Jacob E, Styles L, Kuypers F, Larkin S, Vichinsky E. Simvastatin reduces vaso-occlusive pain in sickle cell anaemia: a pilot efficacy trial. Br J Haematol. 2017;177(4):620-629. 91. Naik RP, Streiff MB, Lanzkron S. Sickle cell disease and venous thromboembolism: what the anticoagulation expert needs to know. J Thromb Thrombolysis. 2013;35(3): 352-358. 92. Naik RP, Streiff MB, Haywood C, Jr., Nelson JA, Lanzkron S. Venous thromboembolism in adults with sickle cell disease: a serious and under-recognized complication. Am J Med. 2013;126(5):443-449. 93. Shet AS, Wun T. How I diagnose and treat venous thromboembolism in sickle cell disease. Blood. 2018;132(17):1761-1769. 94. Brittain JE, Knoll CM, Ataga KI, Orringer EP, Parise LV. Fibronectin bridges monocytes and reticulocytes via integrin alpha4beta1. Br J Haematol. 2008;141(6):872-881. 95. Dominical VM, Samsel L, Nichols JS, et al. Prominent role of platelets in the formation of circulating neutrophil-red cell heterocellular aggregates in sickle cell anemia. Haematologica. 2014;99(11):e214-217. 96. Wun T, Cordoba M, Rangaswami A, Cheung AW, Paglieroni T. Activated monocytes and platelet-monocyte aggregates in patients with sickle cell disease. Clin Lab Haematol. 2002;24(2):81-88. 97. Desai PC, Brittain JE, Jones SK, et al. A pilot study of eptifibatide for treatment of acute pain episodes in sickle cell disease. Thromb Res. 2013;132(3):341-345. 98. Brittain JE, Anea C, Desai P, et al. Effect of eptifibatide on inflammation during acute pain episodes in sickle cell disease. Am J Hematol. 2018;93(4):E99-E101. 99. Wun T, Soulieres D, Frelinger AL, et al. A double-blind, randomized, multicenter phase 2 study of prasugrel versus placebo in adult patients with sickle cell disease. J Hematol Oncol. 2013;6:17.

haematologica | 2019; 104(9)

100. Heeney MM, Hoppe CC, Abboud MR, et al. A multinational trial of prasugrel for sickle cell vaso-occlusive Events. N Engl J Med. 2016;374(7):625-635. 101. Kanter J, Abboud MR, Kaya B, et al. Ticagrelor does not impact patient-reported pain in young adults with sickle cell disease: a multicentre, randomised phase IIb study. Br J Haematol. 2019;184(2):269-278. 102. Ataga KI, Key NS. Hypercoagulability in sickle cell disease: new approaches to an old problem. Hematology Am Soc Hematol Educ Program. 2007:91-96. 103. Ahmed S, Siddiqui AK, Iqbal U, et al. Effect of low-dose warfarin on D-dimer levels during sickle cell vaso-occlusive crisis: a brief report. Eur J Haematol. 2004;72(3):213-216. 104. Qari MH, Aljaouni SK, Alardawi MS, et al. Reduction of painful vaso-occlusive crisis of sickle cell anaemia by tinzaparin in a doubleblind randomized trial. Thromb Haemost. 2007;98(2):392-396. 105. Shah N, Willen S, Telen MJ, Ortel TL. Prophylactic dose low molecular weight heparin (dalteparin) for treatment of vasoocclusive pain crisis in patients with sickle cell disease. Blood. 2013;122(21):2241. 106. Hebbel RP. Ischemia-reperfusion injury in sickle cell anemia: relationship to acute chest syndrome, endothelial dysfunction, arterial vasculopathy, and inflammatory pain. Hematol Oncol Clin North Am. 2014;28(2):181-198. 107. Hebbel RP, Osarogiagbon R, Kaul D. The endothelial biology of sickle cell disease: inflammation and a chronic vasculopathy. Microcirculation. 2004;11(2):129-151. 108. Zhang D, Chen G, Manwani D, et al. Neutrophil ageing is regulated by the microbiome. Nature. 2015;525(7570):528-532. 109. Glassberg J, Minnitti C, Cromwell C, et al. Inhaled steroids reduce pain and sVCAM levels in individuals with sickle cell disease: a triple-blind, randomized trial. Am J Hematol. 2017;92(7):622-631.

110. Langer AL, Leader A, Kim-Schulze S, Ginzburg Y, Merad M, Glassberg J. Inhaled steroids associated with decreased macrophage markers in nonasthmatic individuals with sickle cell disease in a randomized trial. Ann Hematol. 2019;98(4): 841-849. 111. Field JJ, Nathan DG, Linden J. Targeting iNKT cells for the treatment of sickle cell disease. Clin Immunol. 2011;140(2):177-183. 112. Field JJ, Lin G, Okam MM, et al. Sickle cell vaso-occlusion causes activation of iNKT cells that is decreased by the adenosine A2A receptor agonist regadenoson. Blood. 2013;121(17):3329-3334. 113. Field JJ, Majerus E, Gordeuk VR, et al. Randomized phase 2 trial of regadenoson for treatment of acute vaso-occlusive crises in sickle cell disease. Blood Adv. 2017;1(20):1645-1649. 114. Field JJ, Majerus E, Ataga KI, et al. NNKTT120, an anti-iNKT cell monoclonal antibody, produces rapid and sustained iNKT cell depletion in adults with sickle cell disease. PloS One. 2017;12(2):e0171067. 115. Owusu-Ansah A, Ihunnah CA, Walker AL, Ofori-Acquah SF. Inflammatory targets of therapy in sickle cell disease. Transl Res. 2016;167(1):281-297. 116. Knight-Perry J, DeBaun MR, Strunk RC, Field JJ. Leukotriene pathway in sickle cell disease: a potential target for directed therapy. Expert Rev Hematol. 2009;2(1):57-68. 117. Mpollo MS, Quarmyne MO, Rayes O, et al. A phase I trial of zileuton in sickle cell disease. Blood. 2013;122(21):993. 118. Kalish BT, Matte A, Andolfo I, et al. Dietary omega-3 fatty acids protect against vasculopathy in a transgenic mouse model of sickle cell disease. Haematologica. 2015;100(7): 870-880. 119. Lim SH, Morris A, Li K, et al. Intestinal microbiome analysis revealed dysbiosis in sickle cell disease. Am J Hematol. 2018;93 (4):E91-E93.

1719

REVIEW ARTICLE Ferrata Storti Foundation

Targeting sickle cell disease root-cause pathophysiology with small molecules Yogen Saunthararajah

Department of Hematology and Oncology, Taussig Cancer Institute, Cleveland Clinic, Cleveland, OH, USA

Haematologica 2019 Volume 104(9):1720-1730

Correspondence: YOGEN SAUNTHARARAJAH saunthy@ccf.org Received: May 10, 2019. Accepted: July 9, 2019. Pre-published: August 8, 2019. doi:10.3324/haematol.2018.207530 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1720 ©2019 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.

1720

ABSTRACT

he complex, frequently devastating, multi-organ pathophysiology of sickle cell disease has a single root cause: polymerization of deoxygenated sickle hemoglobin. A logical approach to disease modification is, therefore, to interdict this root cause. Ideally, such interdiction would utilize small molecules that are practical and accessible for worldwide application. Two types of such small molecule strategies are actively being evaluated in the clinic. The first strategy intends to shift red blood cell precursor hemoglobin manufacturing away from sickle hemoglobin and towards fetal hemoglobin, which inhibits sickle hemoglobin polymerization by a number of mechanisms. The second strategy intends to chemically modify sickle hemoglobin directly in order to inhibit its polymerization. Important lessons have been learnt from the pre-clinical and clinical evaluations to date. Open questions remain, but this review summarizes the valuable experience and knowledge already gained, which can guide ongoing and future efforts for molecular mechanism-based, practical and accessible disease modification of sickle cell disease.

Introduction Sickle cell disease (SCD) demands practical, accessible oral therapies, since it is a problem of global scope. It afflicts millions of people worldwide, and has an especially high prevalence in pediatric populations in low-income, malaria-belt countries.1 Such therapies are technically plausible, since despite the complex and potentially devastating multi-organ pathophysiology of SCD, this condition has a single, well-characterized root cause: polymerization of deoxygenated sickle hemoglobin (HbS). The hemoglobin molecule is an assembly of two a-like protein subunits and two β-like protein subunits (a2β2), each with a heme moiety to transport an oxygen molecule. In SCD, the gene for the β sub-unit (HBB) of adult hemoglobin (HbA) contains an ‘A’ to ‘T’ mutation in the seventh codon. The β sub-units (βS) produced by this mutated gene substitute a hydrophilic glutamate with a hydrophobic valine, predisposing deoxygenated HbS (a2βS2) to polymerization and gelation in red blood cells (RBC). This affects RBC viability, rheology and adhesiveness, promoting hemolysis, endothelial damage, occlusion of small blood vessels, and thromboses of large vessels. The hemolytic anemia is frequently severe, and is only partially and non-sustainably compensated by >10-fold increases in erythropoiesis.2 The net consequence of this anemia and vaso-occlusion is decreased oxygen delivery and hypoxic injury to potentially all tissues of the body, manifest clinically as episodic pain, chronic pain, avascular necrosis of bones, infections, overt and silent strokes, renal/respiratory/cardiac/hepatic failure, and early death. In the USA >$1 billion in annual health care costs is attributed to SCD, and even so, the median life expectancy of affected individuals is shortened by two or more decades on average.3,4 Most children with SCD in low-income countries do not even survive to adulthood.1 By way of emphasis, all this morbidity and mortality begins with a single process, polymerization of deoxygenated HbS in RBC, and it is therefore logical to attempt to interdict this root cause. Two major small molecule drug approaches are in active clinical evaluation: (i) small molecules to shift the hemoglobin manufactured by RBC precursors from HbS to fetal hemoglobin (HbF), and (ii) small molecules to chemically modify HbS to impede its polymerhaematologica | 2019; 104(9)

Targeting sickle cell root-cause pathophysiology

ization. These active efforts are discussed in turn, with an emphasis on lessons learned so far and remaining open questions. Small molecule approaches for which there are no active clinical efforts that we are aware of are not discussed here, e.g., small molecules to decrease HbS concentration by increasing RBC hydration.5,6 Methods to interdict HbS polymerization that are not based on small molecule drugs are also not discussed here, because their application in the areas of the world most affected by SCD will be difficult for reasons of infrastructure and costs, e.g., harvesting of autologous hematopoietic stem cells, their engineering ex vivo, then re-infusion after myeloablative bone marrow conditioning by chemotherapy and/or radiation (gene therapy), or use of hematopoietic stem cells from immune-compatible non-SCD donors for transplant – a valuable approach in the West that has been thoroughly and recently reviewed elsewhere.7

Interdicting HbS polymerization by pharmacological induction of HbF At the fetal stage of life, RBC contain fetal hemoglobin (HbF), an assembly of two a-globin subunits and two γglobin subunits (a2γ2), with the γ-globin subunits being encoded by duplicated γ-globin genes (HBG2 and HBG1). During human development, the switch from HbF to HbA production begins late in fetal gestation (~ 7 months), and the typical adult pattern of <1% HbF and >90% HbA in

RBC is established by ~12 months post-conception.8,9 Several genetic polymorphisms or mutations in humans, some but not all identified, promote persistent, relatively high RBC HbF content beyond infancy. The phenotypes with particularly generous HbF levels (HbF >10%) are referred to as hereditary persistence of fetal hemoglobin (HPFH). SCD patients who co-inherit such genetic variants can, in the best cases, have asymptomatic, normal life-spans.10-12 Notably, HbF has benefits even at lower dynamic ranges than seen in HPFH: HbF levels correlate continuously with fewer vaso-occlusive pain crises, less renal damage, less pulmonary hypertension, fewer strokes and longer survival.4,13-19 In short, nature has demonstrated that HbF is a highly potent modulator of SCD.20 Detailed biochemical studies have demonstrated how: the intracellular concentration of HbS is a major determinant of polymerization kinetics, and HbF substitution for HbS decreases this concentration.20-22 Moreover, HbF does not polymerize with deoxygenated HbS for reasons of molecular structure (the sophisticated biophysics underlying this have recently been reviewed in detail).5 By contrast, HbA can polymerize with deoxygenated HbS.20-22 In short, HbF interdicts the root-cause pathophysiology of SCD. It is logical therefore to attempt to use pharmacology to recapitulate such naturally demonstrated, powerful disease modulation.23

The earliest efforts at HbF induction The earliest efforts built on the observation that HbF is enriched in RBC produced during the recovery phase of

Figure 1. Polymerization of sickle hemoglobin drives the multi-organ cascade of sickle cell disease pathophysiology. This review examines the strategies to interdict the multi-organ cascade of sickle cell disease at its inception using small molecules that shift red blood cell precursor production from sickle hemoglobin (HbS) toward fetal hemoglobin (HbF), and small molecules that chemically modify HbS to decrease its polymerization. We published variations of this figure in Molokie et al.95 and Lavelle et al.23

haematologica | 2019; 104(9)

1721

Y. Saunthararajah et al.

bone marrow from severe insults or stress.24-29 One way of creating such stress is to administer cytotoxic (cell killing) drugs, leading to clinical evaluation in SCD of the oral ribonucleotide reductase inhibitor hydroxyurea.28-30 In the pivotal trial, hydroxyurea (15-35 mg/kg) increased HbF for 2 years in ~50% of the adult SCD patients treated.30,31 As predicted, HbF increases with hydroxyurea correlated strongly with longer RBC half-life,32,33 fewer pain crises,31 and better quality of life34 (the benefits of hydroxyurea therapy in sickle cell mice also depended on HbF induction).35 Trial patients with HbF levels >0.5 g/dL also survived longer15 although a caveat to these analyses was that it was not known whether the higher HbF levels were intrinsic to the patients or a result of the hydroxyurea therapy. There were, however, noteworthy limitations to the induction of HbF by hydroxyurea: (i) average HbF increases at 2 years were modest (3.6%);28-31,36 (ii) HbF increases were particularly unlikely in patients with the lowest baseline HbF levels and thus at highest risk of morbidity and mortality,31,33,37,38 and (iii) HbF increases diminished over time, even in the ~50% of patients with excellent initial HbF induction.31,39 A shared basis for these several limitations was suggested by the correlation between lower HbF increases and fewer reticulocytes (<300,000x109/L) and neutrophils (<7.5x109/L) at baseline: this correlation underscored that HbF induction by cytotoxicity requires sufficient reserves of hematopoietic precursors to mount repeated recoveries from the stress that destroys their counterparts.29,31 Such reserves are circumscribed, subject to attrition via vasoocclusion in the marrow and kidneys, and decline with aging.31,33,37,38 A declining capacity to compensate for hemolytic anemia is a problem even separate from considerations of sustainable HbF induction via cytotoxicity: SCD patients require erythropoiesis at >10-fold the normal rate simply to sustain hemoglobin levels compatible with life, and dwindling compensatory reticulocytosis is a

major cause of early death.2,15,31,40 Therefore, alternative, non-cytotoxic, durable, and more potent methods of inducing HbF are needed.

Directly targeting the enzymes that silence the γ-globin gene

DNA in nuclei is packaged together with RNA and structural proteins – histones - to form chromatin. Chromatin regulates gene transcription by determining accessibility of genes to the massive machinery (~150 proteins) that transcribes genes. Reorganization (‘remodeling’) of chromatin, to facilitate or hinder this machinery, is signaled via post-translational modifications to histones methylation and acetylation of lysine residues, phosphorylation of threonines and serines – and by modifications to DNA, mainly, methylation of cytosines that precede guanines (CpG). These signals determine whether ATPdependent chromatin remodelers shift histones towards or away from gene transcription start sites, repositioning these physical barriers to either welcome or obstruct the gene transcribing basal transcription factor machinery. Thus, induction of HbF, even when it is indirectly via bone marrow stress, implies remodeling γ-globin and β-globin gene loci, to activate one and not the other.41 Specifically, persistent HbF expression requires: (i) decreased operation at HBG2/HBG1 of epigenetic enzymes that create ‘off’ marks and that reposition histones to obstruct transcription start sites, and (ii) increased function of the epigenetic enzymes that create epigenetic ‘on’ marks and that reposition histones away from transcription start sites, with vice versa at HBB. Cytotoxic methods of inducing HbF achieve such chromatin remodeling crudely and indirectly, via bone marrow stress29,41,42 (Figure 2). So why not identify repressing epigenetic enzymes and inhibit them directly43,44 (Figure 2)? Cells contain dozens of epigenetic enzymes mediating gene activation and repression, and not all repressing epigenetic enzymes (corepres-

Figure 2. Induction of fetal hemoglobin (HbF) requires chromatin remodeling, including DNA hypomethylation, of the HbF gene locus. Bone marrow stress, e.g., from cytotoxic drugs such as hydroxyurea, can create chromatin remodeling during the recovery phase of surviving erythroid precursors. An alternative approach is to remodel the hemoglobin F (HbF) gene locus (HBG) directly, e.g., by directly inhibiting/repressing epigenetic enzymes. Enzymes shown are those known to be recruited by BCL11A, TR2 or TR4 (EHMT2 and PRMT5 are not reported participants in the BCL11A/TR2/TR4 hub). The relative efficiencies of these approaches are illustrated by the greater HbF increases produced in the same non-human primates or patients by decitabine ~0.2 mg/kg twice weekly versus hydroxyurea ~20 mg/kg daily. That is, the molar amount of decitabine administered per week is <1/1000th the amount of hydroxyurea administered per week. We published variations of this figure in Molokie et al.95 and Lavelle et al.23

1722

haematologica | 2019; 104(9)

Targeting sickle cell root-cause pathophysiology

sor protein complexes) are logical molecular targets for therapy. Sequence-specific DNA-binding factors are particular in their epigenetic co-regulator usage, e.g., even distinguishing between closely similar BAF and PBAF coactivator complexes.45-48 Logically, the epigenetic enzymes to target for HbF induction are those that have been directly implicated in silencing of HBG2/HBG1. Multi-protein corepressor complexes directed to the HBG loci by the DNA-binding factors DRED and BCL11A have been characterized in great detail.49-53 Druggable epigenetic silencing enzymes contained in these recruited corepressor complexes include DNA methyltransferase 1 (DNMT1), various histone deacetylases (HDAC), lysine demethylase 1 (LSD1, KDM1A), and chromodomain helicase DNA binding protein 4 (CHD4) and other members of the ISWI family of ATP-dependent chromatin remodelers52-55 (Table 1). Other types of biochemical studies have implicated euchromatic histone lysine methyltransferase 2 (EHMT2, G9a),56 and protein arginine methyltransferase 5 (PRMT5) in the silencing of HBG2/HBG1 57,58 (Table 1). Yet another approach to identifying candidate targets has been chemical screens for HbF inducers. This approach has identified histone methyltransferases EHMT1 and EHMT2 as candidates for inhibition59,60 (Table 1). Notably not identified by studies thus far, given that there are clinically available inhibitors for these targets, are epigenetic enzymes in polycomb repressor complex 2 (e.g., EZH2).61 Since the natural genetic experiment of HPFH provides a fundamental rationale for pursuing pharmacological induction of HbF, by extension, can the genetic variants underlying HPFH help to identify or prioritize molecular targets for manipulation? HPFH-linked point mutations cluster in two regions 115 and 200 base-pairs upstream of the HBG2 start site, suggesting these are sites at which key repressors of HBG2/HBG1 bind.62 BCL11A and ZBTB7A have been shown to bind at these locations, and HPFH mutations have been shown to abrogate such binding.63 Moreover, some

HPFH mutations occur at BCL11A rather than Î˛-globin gene loci.51,64 In short, the natural genetic experiment of HPFH also seems to support drugging of the corepressors recruited by BCL11A (and ZBTB7A and DRED).49-53 The candidate targets are discussed below in turn.

Histone deacetylases (HDAC) HDAC were among the first candidate targets identified for HbF induction.65 Moreover, a number of HDAC inhibitors have already been approved by the United States Food and Drug Administration (FDA) to treat peripheral T-cell malignancies (Table 1). Unfortunately, despite exciting pre-clinical results, clinical application of marketed HDAC inhibitors for HbF induction is limited by the pleiotropic roles of HDAC outside of chromatin. That is, clinical side-effects, arising from HDAC participation in multiple cellular and physiological functions, limit the achievement of an epigenetic pharmacodynamic effect in the target compartment, and thus of HbF induction in vivo.66-71 There are efforts to develop HDAC inhibitors that are more selective to specific HDAC than the broad HDAC inhibiting activity of the currently marketed drugs (Table 1), and perhaps these more selective agents will have a more suitable safety profile for HbF induction. The caution remains that even an on-target, specific drug action can generate toxicities if the molecular target of that action has pleiotropic physiological roles.

DNA methyltransferase 1 (DNMT1) DNMT1 is well known to maintain methylation marks on DNA through cell division. In addition, DNMT1 is a corepressor that is recruited by sequence-specific DNAbinding factors, e.g., DRED (TR2/TR4) and BCL11A, which direct epigenetic silencing of HBG.43,53,72-82 The deoxycytidine analog decitabine and its pro-drug 5-azacytidine, FDA-approved to treat the myeloid malignancy myelodysplastic syndrome, can deplete DNMT1: a nitrogen substituted for a carbon in the decitabine pyrimidine

Table 1. Scientifically validated molecular targets for HbF induction and candidate drugs

Target

Recruited by BCL11A

Drugs

Stage

HDAC*

Yes

- Depsipeptide (HDAC1,2,4,6) - Belinostat (broad HDAC inhibitor) - Panobinostat (broad) - Vorinostat (broad)

DNMT1

Yes

- Decitabine - 5-azacytidine (decitabine pro-drug)

KDM1A#

Yes

- ORY-1001 (related to RN-1) - GSK2879552 - 4SC-202 - INCB059872 - GSK3326595 - UNC0638 - not officially designated, patent issued

- Marketed for peripheral T-cell lymphomas - Phase I in SCD (panobinostat) (ClinicalTrials.gov identifier: NCT01245179) - Phase II in SCD and Î˛-thalassemia (HQK-1001)(ClinicalTrials.gov identifiers: NCT01642758, NCT01601340) - Marketed for myelodysplastic syndromes - Oral forms, including in combination with inhibitors of degradation, are in phase I/II for liquid/solid malignancies, and SCD (ClinicalTrials.gov identifier: NCT01685515) - Phase I/II in liquid/solid malignancies - Phase I in SCD (INCB059872) (ClinicalTrials.gov identifier: NCT03132324) (terminated, results not publicly available) - Phase I in liquid/solid malignancies - Pre-clinical in vitro - Pre-clinical in vitro

PRMT5 Not reported EHMT2 Not reported ISWI Yes (CHD4, SMARCA5)

*Only histone deacetylase (HDAC) inhibitors approved in the USA are listed, several other HDAC inhibitors are in clinical trials. #Only KDM1A inhibitors registered in clinical trials in the USA are listed, several other compounds are in development. SCD: sickle cell disease.

haematologica | 2019; 104(9)

1723

Y. Saunthararajah et al.

ring covalently binds to DNMT1 and causes its degradation.83 By depleting DNMT1 protein, decitabine disrupts its scaffolding functions for other epigenetic enzymes such as KDM1A.84,85 That is, decitabine does not just inhibit the enzyme function of DNMT1 but produces a broad corepressor disrupting effect. Because the deoxyribose moiety of decitabine is unmodified, it can incorporate into the elongating DNA strand during the S-phase without terminating chain extension or causing cytotoxicity, contrasting with most nucleoside analogs used in the clinic to treat cancer.86,87 High concentrations of decitabine do, however, produce off-target anti-metabolite effects and cytotoxicity, in significant part via its uridine moiety degradation products that can misincorporate into DNA or inhibit thymidylate synthase.88,89 We designed decitabine dose, schedule and route-of-administration regimens to produce non-cytotoxic depletion of DNMT1 in vivo.43,90-93 These regimens increased HbF by >10% in SCD patients who had no HbF response (~0.3%) to hydroxyurea in the pivotal clinical trial.43,81,94 That is, very small, non-cytotoxic doses of ~0.2 mg/kg twice weekly were sufficient to produce large increases in HbF and total hemoglobins, even in patients in whom hydroxyurea ~20 mg/kg/day, >1000-fold the molar amount of decitabine, did not induce HbF (Figure 2).43,44 Marketed decitabine, however, is a parenteral drug with trivial oral bioavailability, undermining potential for worldwide application. We have therefore combined oral decitabine with tetrahydrouridine to inhibit the enzyme that limits its oral bioavailability, cytidine deaminase.95 This combination was well-tolerated and safe in a phase I study in patients with severe SCD. The target decitabine dose of 0.16 mg/kg produced a wide decitabine concentration-time profile (low Cmax, long Tmax) ideal for non-cytotoxic DNMT1 depletion83,96-98 and decreased DNMT1 protein in peripheral blood mononuclear cells by >75% and repetitive element CpG methylation by ~10%. This increased HbF by 4-9%, doubling HbF-enriched RBC (Fcells) up to ~80% of total RBC. Total hemoglobin increased by 1.2-1.9 g/dL (P=0.01) as reticulocytes simultaneously decreased; that is, better quality and efficiency of HbF-enriched erythropoiesis elevated hemoglobin using fewer reticulocytes. Other indications of better RBC quality, biomarkers of hemolysis, thrombophilia and inflammation (lactate dehydrogenase, bilirubin, D-dimer, C-reactive protein) also improved. The side-effects were a concurrent increase in platelets and decrease in neutrophils, expected with non-cytotoxic DNMT1 depletion. In the relatively short treatment duration of 8 weeks, these blood count shifts did not cross thresholds requiring withholding or modification of treatment, that is, neutrophil counts and platelets remained in ranges observed in SCD patients receiving standard-of-care therapies. The major limitation is the need for longer term studies to demonstrate durable safety and efficacy of the oral tetrahydrouridine/decitabine combination.

Lysine demethylase 1 (LSD1, KDM1A) KDM1A, like DNMT1, is recruited by the HBG2/HBG1 repressing DNA-binding factors DRED and BCL11A, and KDM1A inhibition with either of two specific inhibitors induced HbF in vitro, in sickle mice and in non-human primates.99-102 Several KDM1A inhibitors are in clinical trials for cancer indications (Table 1). At least two of the compounds in trials (ORY-1001, GSK2879552) are built around 1724

a tranylcypropamine warhead that inhibits monoamine oxidases that metabolize catecholamine neurotransmitters in the brain. Although cancer clinical trials are ongoing and unpublished (EudraCT number: 2013-002447-29; ClinicalTrials.gov identifiers: NCT02177812, NCT02034123), there is concern regarding side-effects related to the potential for inhibition of monoamine oxidases other than KDM1A. Thus, there are ongoing efforts to develop and evaluate KDM1A inhibitors with other scaffolds (e.g., ClinicalTrials.gov identifier: NCT01344707). One registered phase I clinical trial evaluated a KDM1A inhibitor for HbF induction in SCD (ClinicalTrials.gov identifier: NCT03132324). This trial has been terminated but results are not publicly available at this time.

Protein arginine methyltransferase 5 (PRMT5) PRMT5 methylation of histone H4 arginine 3 has been implicated as a signal that recruits additional chromatinmodifying enzymes and represses HBG.57 There is a PRMT5 inhibitor in clinical trials (GSK3326595) for cancer indications. No trials of this molecule for HbF induction in SCD have been registered so far.

Euchromatic histone lysine methyltransferase 2 (EHMT2, G9a)

EHMT2 has been shown to be recruited to the Î˛-globin locus by the sequence-specific DNA binding factor NFE2, and the EHMT2 inhibitor UNC0638 has been shown to induce HbF in vitro.59,60 As of this time, there are no registered clinical trials evaluating EHMT2 inhibition to induce HbF.

Chromodomain helicase DNA binding protein 4 (CHD4) and SWI/SNF-related matrix-associated actin-dependent regulator of chromatin subfamily A member 5 (SMARCA5) (ISWI family of ATPdependent chromatin remodelers The culmination of chromatin remodeling for gene repression or activation is nucleosome (histone octamer) repositioning around the transcription start site. This is energetically expensive work executed by SWI/SNF or ISWI family proteins containing the HELICc-DExx ATPase domain, with SWI/SNF moving histones away to facilitate basal transcription factor machinery access and ISWI executing the opposite.46,103,104 Since such nucleosome repositioning is the crux of chromatin remodeling, inhibition of this action should in principle offer corresponding potency. CHD4 and SMARCA5 are HELICcDExx-containing corepressors that are recruited by BCL11A and DRED to repress HBG2/HBG1.54 We have identified a first-in-class drug-like compound series that preliminarily appears to inhibit the HELICc-DExx domains of SMARCA5 and CHD4, and we are actively investigating the potential of this series to induce HbF (US20170253589A1).

Epigenetic targeting - Lessons so far and open questions Pre-clinical and clinical experience to date provide various lessons and raise some questions regarding epigenetic targeting to induce HbF, as described below. The consequences of inhibiting an epigenetic enzyme depend on cellular context The baseline expression pattern of transcription factors is a key determinant of a cellâ&#x20AC;&#x2122;s fate or function response to epigenetic enzyme inhibition, because sequence-specific haematologica | 2019; 104(9)

Targeting sickle cell root-cause pathophysiology

DNA-binding factors direct the function of these epigenetic enzymes and are mandatory for gene activation.105 Stated another way, the consequences of inhibiting a particular epigenetic enzyme depend very much on cellular context.105 A corollary of the above is that although inhibiting silencing epigenetic enzymes can produce cell fate or function shifts, these relate to what the cells were to begin with and are not drastic.105 This is of course critical clinically, since a candidate epigenetic therapeutic for SCD will be distributed systemically. What then about the cellular/transcription factor context of erythropoiesis enables inhibition of DNMT1 etc., to activate HBG2/HBG1? Several groups have found that the developmental switch from HBG2/HBG1 to HBB activation is recapitulated, albeit very rapidly, during erythroid lineage maturation (a ‘maturational switch’ during routine erythropoiesis).106-110 The maturational switch entails removal of activating and acquisition of repressive epigenetic marks at HBG2/HBG1.59,102 with physical migration of the shared enhancer, the locus control region, from the HBG2/HBG1 to the HBB locus.54 These dynamics at HBG2/HBG1 and HBB during erythropoiesis creates an opportunity for pharmacological/biochemical intervention to prevent enhancer migration, to stall the massive gene activating machinery at HBG2/HBG1. That is, HbF induction by inhibiting epigenetic ‘off’ enzymes such as DNMT1 is not predicated on returning the enhancer from HBB back to HBG2/HBG1 (turning a gene that is ‘off’ to ‘on’), but on preventing a switch from HBG2/HBG1 to HBB in the first place (preventing a gene that is ‘on’ from being turned ‘off’). Accordingly, HbF induction by an inhibitor of the silencing epigenetic enzyme EHMT2 (UNC0638) depended on the timing of its addition to cultures of synchronously maturing erythroid progenitors,59 with similar observations in our hands with DNMT1depleting drugs (personal communication).

Why are these drugs being evaluated for, or used, to treat cancer? Some of the most recurrently mutated, deleted or amplified genes in cancers encode for chromatin remodelers. Thus, another concern with epigenetic targeting is whether it might mimic some of these genetic changes and favor activation of oncogenic programs. It is reassuring to some extent, however, that the epigenetic targets and drugs discussed above have been or are being developed to treat and/or prevent cancer. We recently reviewed the biological rationale for this,105 and it is briefly summarized here: cancer cells, including self-replicating cancer cells (cancer or leukemia ‘stem’ cells), contain high amounts of the lineage master transcription factors that normally activate terminal lineage-fates, and depend on specific corepressors (‘addictions’) in order to avoid these terminal fates.111 The pathway of action is activation of the terminal lineage-fates intended by cancer cell lineage master transcription factor content. The same chromatin-‘relaxing’ treatments that trigger terminal lineage-fates of cancer/leukemia stem cells preserve self-renewal of uncommitted tissue stem cells, since these cells express stem cell master transcription factors, not high levels of lineage-specifying transcription factors.92,93 This therapeutic index explains why non-cytotoxic doses and schedules of decitabine can suppress malignant clones and simultaneously improve functional blood counts even in elderly patients with myeloid malignancies.91,105,111-115 Stated simply, several corepressor components (repressing epigenetic haematologica | 2019; 104(9)

enzymes), e.g., DNMT1, HDAC, KDM1A, have been biologically validated as molecular targets for the treatment and prevention of cancer.111

Teratogenic risks Another concern is the potential for teratogenicity: this should be assumed for individual agents, unless shown otherwise by formal toxicological studies.

Drug metabolism is central to the clinical profile of activity Drugs, being biologically active, are metabolized, and this too can contribute substantially to their in vivo profile of activity. For example, DNMT1-depleting decitabine is a pyrmidine nucleoside analog pro-drug that depends absolutely for its activity on the pyrimidine metabolism enzyme deoxycytidine kinase: Deoxycytidine kinase executes the initial phosphorylation of decitabine in cells, which rate-limits its conversion into the nucleotide form that actually depletes DNMT1. Serendipitously, deoxycytidine kinase is most highly expressed in the myeloid compartment, especially erythroid precursors. Thus, the clinical profile of decitabine activity is in major part dictated by its metabolically driven tropism for the myeloid compartment.

Baseline HbF levels dictate final HbF levels There is a wide variation in baseline HbF levels in patients with SCD and even in the general population, reflecting the influence of various genetic polymorphisms on the regulation of this locus.9 Even if a molecular targeted therapy produces similar rates of increase in HbF% (the percentage of total hemoglobin that is HbF) in all patients, the final HbF% will be dictated by the level at which HbF% began. Moreover, in clinical trials we have conducted with DNMT1-depleting decitabine, we have noticed a slightly lower slope to the rate of increase in HbF% in SCD patients with lower HbF% at baseline. Fortunately and importantly, however, HbF induced by this epigenetic strategy was well-distributed among RBC, and the rates of increase of HbF-enriched RBC (F-cells) was actually higher in patients with low F-cells at baseline.43,81,94,95 At some time-point after starting therapy, F-cells entering the circulation are matched by a similar number of Fcells leaving the circulation, producing plateaus in HbF% and F-cell%.

Small molecules to chemically modify HbS to impede polymerization The scientific foundation for efforts to chemically modify HbS is the two-state allosteric Monod-WymanChangeux structural model which characterizes the rapidly reversible equilibrium between the quarternary structure of hemoglobin with low oxygen affinity (fully deoxygenated hemoglobin, ‘‘T’ quarternary structure) and the hemoglobin quarternary structure with high affinity for oxygen (oxygenated hemoglobin, ‘R’ quarternary structure).116,117 The Monod-Wyman-Changeux model demonstrated incompatibility of the R conformation with polymerization, creating a foundation to propose molecules to favor the high oxygen affinity R conformation, as a method to delay HbS polymerization.5 The basic concern with such an approach is that SCD is a disease of decreased oxygen delivery to tissues and, 1725

Y. Saunthararajah et al.

thus, if a chemical modification produces a high oxygen affinity hemoglobin molecule, there is a necessary play off between decreased oxygen supply from increased oxygen affinity of hemoglobin versus increased oxygen supply from less HbS polymerization/higher total hemoglobin. This balancing act is discussed in more detail in the section ‘Lessons learned so far’ below. Ultimately, however, rigorous clinical evaluation is key,5,118 and clinical evaluation has started or is underway for a number of candidate drugs that exploit these principles: Small molecules to convert hemoglobin to methemoglobin The earliest clinical effort in this field evaluated the conversion of hemoglobin to methemoglobin following the administration of sodium nitrite or para-amino-propriophenone (PAPP) to five patients.119 Both agents were able to increase methemoglobin. Methemoglobin levels of >20% (but not less) produced by sodium nitrite extended RBC survival as measured by chromium-labeling. The methemoglobinemia itself was apparently well-tolerated, but there was no evidence of any clinical benefit. Instead, there were significant side-effects from the administered drugs.119 Interestingly, higher methemoglobin levels produced by PAPP did not extend RBC survival, possibly because PAPP was directly hemolytic. Small molecules to convert hemoglobin to carboxyhemoglobin Carbon monoxide can be used to convert hemoglobin to carboxyhemoglobin. Infusion of free pegylated carboxyhemoglobin (MP4CO), as a hemoglobin-based carbon monoxide carrier, was evaluated in a phase I study.120 In an abstract description of results in 18 patients, the maximum increase in carboxyhemoglobin was to 2%, which returned to pre-dosing levels within 8 h of completion of the MP4CO infusion. There was no significant increase in total hemoglobin. No further studies have been reported. Small molecules that delay HbS polymerization by unclear mechanisms Niprisan (Nix-0699) and related small molecules (SCD101) are plant-derived molecules that have been found to delay polymerization of deoxygenated HbS, but by unclear mechanisms.121 SCD-101 has been evaluated in a phase IB clinical trial in 26 SCD patients. There were no major adverse events attributed to the drug taken for 28 days, and it appeared to decrease chronic pain and fatigue at higher doses. However, there were no laboratory data providing evidence of decreased hemolysis or increased total hemoglobin, although analysis of peripheral smears suggested improvements in RBC shape.122 Small molecules to increase hemoglobin oxygen affinity Specific small molecule aldehydes have been found to form reversible Schiff base linkages with the N-terminal amino group of hemoglobin a chains to lock in the high oxygen affinity R conformation, and the polyaromatic adldehyde GBT440 (voxeletor) has been developed through to phase III clinical trial evaluation. In phase I/II randomized, double-blind, placebo-controlled evaluation in SCD patients, some of whom were receiving concurrent therapy with hydroxyurea, there were increases in total hemoglobin of ≥1 g/dL in six of 12 patients who received the drug for 90 days or more.123 There were concurrent decreases in markers of hemolysis (lactate dehy1726

drogenase, total bilirubin). There were no significant adverse events attributed to study drug. Oxygen delivery was evaluated by measurement of oxygen consumption during cardiopulmonary exercise testing, erythropoietin levels, resting heart rate and heart rate during peak exercise, and these parameters did not suggest decreased oxygen delivery to tissues.123 A subsequent double-blind, randomized, placebo-controlled phase III clinical trial evaluated two different doses of the study drug (900 and 1500 mg per day) in 274 SCD patients, two-thirds of whom remained on stable doses of hydroxyurea initiated well before study enrollment.124 A hemoglobin response, defined as an increase from baseline of >1 g/dL at week 24, occurred in 51% of the patients on the 1500 mg dose, 33% on the 900 mg dose, and 7% on placebo, in intention-to-treat analyses. There were also improvements in biomarkers of hemolysis. The frequency of vaso-occlusive crises did not differ between the treatment arms. Breakdown of vaso-occlusive crisis frequency according to whether or not the patients were taking hydroxyurea was not reported. Erythropoietin levels (as a surrogate for oxygen delivery) as well as grade 3 and serious adverse events were similar between the treatment arms.124

Chemical modification of HbS – lessons learned so far and open questions Balancing acts The clinical trial results with GBT440 thus illustrate that chemical modification of hemoglobin to increase its oxygen affinity (promote the hemoglobin R conformation) can indeed significantly decrease hemolysis and significantly increase total hemoglobin. The hope and goal is that higher hemoglobin increases oxygen supply by amounts that exceed any decrease in oxygen supply from the higher oxygen affinity of the modified hemoglobin molecule,5,118 as per the equation: Oxygen Supply = Blood Flow (mL blood/100 g tissue/min) x Arterial Oxygen Saturation (%) x Total Hemoglobin (g/dL).125 Thus, increasing total hemoglobin increases oxygen supply, but chemical modification of some of these hemoglobin molecules to increase oxygen affinity decreases effective arterial oxygen saturation and oxygen supply. Some tissues, e.g., the brain, have limited capacity to increase the ‘blood flow’ component in the equation, and hence, are particularly dependent on the ‘arterial oxygen saturation’ x ‘total hemoglobin’ components, as extensively modeled recently.118,125 Underscoring this point, most silent cerebral infarctions in SCD children have been found to be caused by disruption to oxygen supply that is not caused by large vessel vasculopathy, implying anemia and/or blood oxygen saturation are critical drivers of this hypoxic damage.126-128 Even the ‘blood flow’ component of the equation is a balancing act in SCD patients: whole blood viscosity is a key determinant of blood flow; less HbS polymerization, by increasing (improving) RBC deformability, can decrease whole blood viscosity and thus increase blood flow. On the other hand, higher total hemoglobin/hematocrit can increase blood viscosity which can decrease blood flow, even with hematocrits in an anemia range, because of the contribution of baseline low RBC deformability of SCD to viscosity. This blood flow calculus needs haematologica | 2019; 104(9)

Targeting sickle cell root-cause pathophysiology

to be considered with small molecules aiming to chemically modify HbS, and with small molecules aiming to substitute HbS with HbF. Ultimately, the risk/benefit calculus for any therapeutic approach requires careful clinical trial determination.

Combinatorial approaches In oncology, combinations of drugs are almost mandatory, because the target cell population is evolving, and will select to evade the effects of drugs. Although target cells in SCD are not evolving, other biological realities compel consideration of combination therapies. One reality is that most SCD patients will already have tissue/organ damage that can undermine the potential benefits of novel small molecule therapeutics. For example, diminished bone marrow reserve from vaso-occlusive damage and/or replication-mediated exhaustion, which decreases compensatory reticulocytosis, and which contributes to early death,2,15,31,33,37,38,40 could limit the scope of potential benefit that can be produced by HbF inducers or HbS modifiers. Another biological reality, but potentially positive, is demonstrated by the approval by the FDA of the amino acid glutamine as a treatment to reduce the frequency of vaso-occlusive crises in SCD patients.129 Natural substances, which in most humans can be assumed to be satisfactorily maintained by a normal diet, might actually be important pharmaceuticals for SCD patients. By way of bringing such negative and potentially positive biological realities together, it is noteworthy that the natural substance nicotinamide (vitamin B3) markedly expands hematopoietic stem cells in vitro at concentrations that can be readily and safely produced in vivo with oral supplementation.130-132 Moreover, nicotinamide is a direct precursor for the vital energy currency nicotinamide adenine dinucleotide (NAD) which is depleted in SCD RBC, increasing their susceptibility to oxidative damage. In fact, replenishing NAD is one of the rationales for glutamine administration to SCD patients.129 In short, in considering combination therapy, there could be important, highly feasible, but unexplored opportunities around relatively non-toxic natural substances (glutamine, nicotinamide, vitamin D, etc.). Other under-evaluated natural molecules include the kidney hormone erythropoietin, since declining kidney erythropoietin production also contributes to declining compensatory reticulocytosis.2 Combining small molecules to inhibit more than one co-repressing enzyme in the BCL11A hub, each used at

References 1. Piel FB, Hay SI, Gupta S, Weatherall DJ, Williams TN. Global burden of sickle cell anaemia in children under five, 2010-2050: modelling based on demographics, excess mortality, and interventions. PLoS Med. 2013;10(7):e1001484. 2. Saraf S, Farooqui M, Infusino G, et al. Standard clinical practice underestimates the role and significance of erythropoietin deficiency in sickle cell disease. Br J Haematol. 2011;153(3):386-392. 3. Kauf TL, Coates TD, Huazhi L, Mody-Patel N, Hartzema AG. The cost of health care for

haematologica | 2019; 104(9)

5. 6.

doses low enough to avoid side-effects from off-target actions, and with non-overlapping side-effects from ontarget actions, might produce greater HbF induction than achieved with a single target. Such molecules should have non-cytotoxic mechanisms of action, to avoid potential injury to needed bone marrow capacity. Unfortunately, there are few non-cytotoxic small molecule drugs targeting rational epigenetic targets, and even fewer for which optimal single molecule application has been characterized (Table 1). That is, more non-cytotoxic epigenetic drugs, and more information on their profiles of sideeffects from on-target and off-target actions, are needed to guide any consideration of combination therapy. What about combining HbF inducers with HbS modifiers? This has in effect been evaluated in the clinical trials of GBT440, since this drug was added to stable doses of hydroxyurea in >60% of clinical trial participants. Hemoglobin increases of >1 g/dL occurred in ~40% of patients taking GBT440 1500 mg alone versus ~55% of patients taking GBT440 1500 mg + hydroxyurea in the phase III trial, but whether vaso-occlusive crisis frequency and other adverse events varied between these two groups was not described.124 The efficacy calculus and hope is that increases in oxygen delivery from better RBC deformability and higher total hemoglobin will exceed decreases in oxygen delivery caused by greater blood viscosity and chemical modification of HbS.

Conclusions Clinical proof-of-principle that substantial total hemoglobin increases can be produced by non-cytotoxic inhibition of specific epigenetic enzymes, to shift RBC precursor hemoglobin manufacturing from HbS to HbF, and by chemical modification of hemoglobin to promote the high oxygen affinity â&#x20AC;&#x2DC;Râ&#x20AC;&#x2122; quarternary structure of the hemoglobin molecule, has already been generated in SCD patients. Clinical evaluation to determine the long-term safety, the impact on symptoms and multi-organ pathophysiology, and the durability of any benefits, is ongoing. There is hope that one or more of the small molecules being evaluated will pass rigorous scrutiny and culminate in practical, accessible, cost-effective, safe and potent diseasemodifying therapy for SCD patients worldwide. Funding source: National Heart, Lung and Blood Institute UO1 HL117658.

children and adults with sickle cell disease. Am J Hematol. 2009;84(6):323-327. Platt OS, Brambilla DJ, Rosse WF, et al. Mortality in sickle cell disease. Life expectancy and risk factors for early death. N Engl J Med. 1994;330(23):1639-1644. Eaton WA, Bunn HF. Treating sickle cell disease by targeting HbS polymerization. Blood. 2017;129(20):2719-2726. Ataga KI, Smith WR, De Castro LM, et al. Efficacy and safety of the Gardos channel blocker, senicapoc (ICA-17043), in patients with sickle cell anemia. Blood. 2008;111(8):3991-3997. Tisdale J. Improvements in haploidentical

8. 9. 10.

11.

transplantation for sickle cell disease and beta-thalassaemia. Lancet Haematol. 2019;6(4):e168-e169. Sankaran VG, Xu J, Orkin SH. Advances in the understanding of haemoglobin switching. Br J Haematol. 2010;149(2):181-194. Rochette J, Craig JE, Thein SL. Fetal hemoglobin levels in adults. Blood Rev. 1994;8(4):213-224. Conley CL, Weatherall DJ, Richardson SN, Shepard MK, Charache S. Hereditary persistence of fetal hemoglobin: a study of 79 affected persons in 15 Negro families in Baltimore. Blood. 1963;21:261-281. Perrine RP, Pembrey ME, John P, Perrine S,

1727

Y. Saunthararajah et al.

12.

13. 14.

15.

16.

17.

18.

19.

20. 21.

22.

23. 24.

25.

26.

27.

28.

29.

30.

1728

Shoup F. Natural history of sickle cell anemia in Saudi Arabs. A study of 270 subjects. Ann Intern Med. 1978;88(1):1-6. Ngo DA, Aygun B, Akinsheye I, et al. Fetal haemoglobin levels and haematological characteristics of compound heterozygotes for haemoglobin S and deletional hereditary persistence of fetal haemoglobin. Br J Haematol. 2012;156(2):259-264. Platt OS, Thorington BD, Brambilla DJ, et al. Pain in sickle cell disease. Rates and risk factors. N Engl J Med. 1991;325(1):11-16. Powars DR, Weiss JN, Chan LS, Schroeder WA. Is there a threshold level of fetal hemoglobin that ameliorates morbidity in sickle cell anemia? Blood. 1984;63(4):921-926. Steinberg MH, Barton F, Castro O, et al. Effect of hydroxyurea on mortality and morbidity in adult sickle cell anemia: risks and benefits up to 9 years of treatment. JAMA. 2003;289(13):1645-1651. Atweh GF, Schechter AN. Pharmacologic induction of fetal hemoglobin: raising the therapeutic bar in sickle cell disease. Curr Opin Hematol. 2001;8(2):123-130. Powars DR, Elliott-Mills DD, Chan L, et al. Chronic renal failure in sickle cell disease: risk factors, clinical course, and mortality. Ann Intern Med. 1991;115(8):614-620. Lebensburger J, Johnson SM, Askenazi DJ, et al. Protective role of hemoglobin and fetal hemoglobin in early kidney disease for children with sickle cell anemia. Am J Hematol. 2011;86(5):430-432. Aban I, Baddam S, Hilliard LM, et al. Severe anemia early in life as a risk factor for sicklecell kidney disease. Blood. 2017;129(3):385387. Akinsheye I, Alsultan A, Solovieff N, et al. Fetal hemoglobin in sickle cell anemia. Blood. 2011;118(1):19-27. Goldberg MA, Husson MA, Bunn HF. Participation of hemoglobins A and F in polymerization of sickle hemoglobin. J Biol Chem. 1977;252(10):3414-3421. Nagel RL, Bookchin RM, Johnson J, et al. Structural bases of the inhibitory effects of hemoglobin F and hemoglobin A2 on the polymerization of hemoglobin S. Proc Natl Acad Sci U S A. 1979;76(2):670-672. Lavelle D, Engel JD, Saunthararajah Y. Fetal hemoglobin induction by epigenetic drugs. Semin Hematol. 2018;55(2):60-67. DeSimone J, Biel SI, Heller P. Stimulation of fetal hemoglobin synthesis in baboons by hemolysis and hypoxia. Proc Natl Acad Sci U S A. 1978;75(6):2937-2940. DeSimone J, Heller P, Adams JG. Hemopoietic stress and fetal hemoglobin synthesis: comparative studies in vivo and in vitro. Blood. 1979;54(5):1176-1181. DeSimone J, Heller P, Amsel J, Usman M. Magnitude of the fetal hemoglobin response to acute hemolytic anemia in baboons is controlled by genetic factors. J Clin Invest. 1980;65(1):224-226. DeSimone J, Heller P, Biel M, Zwiers D. Genetic relationship between fetal Hb levels in normal and erythropoietically stressed baboons. Br J Haematol. 1981;49(2):175-183. DeSimone J, Biel M, Heller P. Maintenance of fetal hemoglobin (HbF) elevations in the baboon by prolonged erythropoietic stress. Blood. 1982;60(2):519-523. Lavelle D, DeSimone J, Heller P, Zwiers D, Hall L. On the mechanism of Hb F elevations in the baboon by erythropoietic stress and pharmacologic manipulation. Blood. 1986;67(4):1083-1089. Charache S, Terrin ML, Moore RD, et al. Effect of hydroxyurea on the frequency of

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41. 42.

43.

44.

45.

46.

47.

painful crises in sickle cell anemia. Investigators of the Multicenter Study of Hydroxyurea in Sickle Cell Anemia. N Engl J Med. 1995;332(20):1317-1322. Steinberg MH, Lu ZH, Barton FB, et al. Fetal hemoglobin in sickle cell anemia: determinants of response to hydroxyurea. Multicenter Study of Hydroxyurea. Blood. 1997;89(3):1078-1088. Ballas SK, Marcolina MJ, Dover GJ, Barton FB. Erythropoietic activity in patients with sickle cell anaemia before and after treatment with hydroxyurea. Br J Haematol. 1999;105(2):491-496. Charache S, Dover GJ, Moore RD, et al. Hydroxyurea: effects on hemoglobin F production in patients with sickle cell anemia. Blood. 1992;79(10):2555-2565. Ballas SK, Barton FB, Waclawiw MA, et al. Hydroxyurea and sickle cell anemia: effect on quality of life. Health Qual Life Outcomes. 2006;4:59. Lebensburger JD, Pestina TI, Ware RE, Boyd KL, Persons DA. Hydroxyurea therapy requires HbF induction for clinical benefit in a sickle cell mouse model. Haematologica. 2010;95(9):1599-1603. Segal JB, Strouse JJ, Beach MC, et al. Hydroxyurea for the treatment of sickle cell disease. Evid Rep Technol Assess. 2008;(165):1-95. Machado RF, Anthi A, Steinberg MH, et al. N-terminal pro-brain natriuretic peptide levels and risk of death in sickle cell disease. JAMA. 2006;296(3):310-318. Ware RE, Eggleston B, Redding-Lallinger R, et al. Predictors of fetal hemoglobin response in children with sickle cell anemia receiving hydroxyurea therapy. Blood. 2002;99(1):10-14. Green NS, Manwani D, Qureshi M, et al. Decreased fetal hemoglobin over time among youth with sickle cell disease on hydroxyurea is associated with higher urgent hospital use. Pediatr Blood Cancer. 2016;63(12):2146-2153. West MS, Wethers D, Smith J, Steinberg M. Laboratory profile of sickle cell disease: a cross-sectional analysis. The Cooperative Study of Sickle Cell Disease. J Clin Epidemiol. 1992;45(8):893-909. Goren A, Simchen G, Fibach E, et al. Fine tuning of globin gene expression by DNA methylation. PLoS One. 2006;1:e46. Platt OS, Orkin SH, Dover G, et al. Hydroxyurea enhances fetal hemoglobin production in sickle cell anemia. J Clin Invest. 1984;74(2):652-656. Saunthararajah Y, Hillery CA, Lavelle D, et al. Effects of 5-aza-2 '-deoxycytidine on fetal hemoglobin levels, red cell adhesion, and hematopoietic differentiation in patients with sickle cell disease. Blood. 2003;102(12):3865-3870. Saunthararajah Y, Lavelle D, DeSimone J. DNA hypo-methylating agents and sickle cell disease. Br J Haematol. 2004;126(5):629636. Sikorski TW, Joo YJ, Ficarro SB, et al. Proteomic analysis demonstrates activatorand chromatin-specific recruitment to promoters. J Biol Chem. 2012;287(42):3539735408. Lemon B, Inouye C, King DS, Tjian R. Selectivity of chromatin-remodelling cofactors for ligand-activated transcription. Nature. 2001;414(6866):924-928. Kadam S, Emerson BM. Transcriptional specificity of human SWI/SNF BRG1 and BRM chromatin remodeling complexes. Mol Cell. 2003;11(2):377-389.

48. Kadam S, McAlpine GS, Phelan ML, et al. Functional selectivity of recombinant mammalian SWI/SNF subunits. Genes Dev. 2000;14(19):2441-2451. 49. Sankaran VG, Menne TF, Xu J, et al. Human fetal hemoglobin expression is regulated by the developmental stage-specific repressor BCL11A. Science. 2008;322(5909):18391842. 50. Tanabe O, Katsuoka F, Campbell AD, et al. An embryonic/fetal beta-type globin gene repressor contains a nuclear receptor TR2/TR4 heterodimer. EMBO J. 2002;21(13):3434-3442. 51. Menzel S, Garner C, Gut I, et al. A QTL influencing F cell production maps to a gene encoding a zinc-finger protein on chromosome 2p15. Nat Genet. 2007;39(10):11971199. 52. Cui S, Kolodziej KE, Obara N, et al. Nuclear receptors TR2 and TR4 recruit multiple epigenetic transcriptional corepressors that associate specifically with the embryonic Î˛type globin promoters in differentiated adult erythroid cells. Mol Cell Biol. 2011;31(16): 3298-3311. 53. Xu J, Bauer DE, Kerenyi MA, et al. Corepressor-dependent silencing of fetal hemoglobin expression by BCL11A. Proc Natl Acad Sci U S A. 2013;110(16):65186523. 54. Liu X, Zhang Y, Chen Y, et al. In situ Capture of chromatin interactions by biotinylated dCas9. Cell. 2017;170(5):1028-1043.e19. 55. Constantoulakis P, Knitter G, Stamatoyannopoulos G. On the induction of fetal hemoglobin by butyrates: in vivo and in vitro studies with sodium butyrate and comparison of combination treatments with 5-AzaC and AraC. Blood. 1989;74(6):1963-1971. 56. Chaturvedi CP, Hosey AM, Palii C, et al. Dual role for the methyltransferase G9a in the maintenance of beta-globin gene transcription in adult erythroid cells. Proc Natl Acad Sci U S A. 2009;106(43):18303-18308. 57. Rank G, Cerruti L, Simpson RJ, et al. Identification of a PRMT5-dependent repressor complex linked to silencing of human fetal globin gene expression. Blood. 2010;116(9):1585-1592. 58. Xu Z, He Y, Ju J, et al. The role of WDR5 in silencing human fetal globin gene expression. Haematologica. 2012;97(11):16321640. 59. Krivega I, Byrnes C, de Vasconcellos JF, et al. Inhibition of G9a methyltransferase stimulates fetal hemoglobin production by facilitating LCR/gamma-globin looping. Blood. 2015;126(5):665-672. 60. Renneville A, Van Galen P, Canver MC, et al. EHMT1 and EHMT2 inhibition induces fetal hemoglobin expression. Blood. 2015;126(16):1930-1939. 61. Velcheti V, Wong KK, Saunthararajah Y. EZH2 inhibitors: take it EZy, it is all about context. Cancer Discov. 2019;9(4):472-475. 62. Wang X, Thein SL. Switching from fetal to adult hemoglobin. Nat Genet. 2018;50(4): 478-480. 63. Martyn GE, Wienert B, Yang L, et al. Natural regulatory mutations elevate the fetal globin gene via disruption of BCL11A or ZBTB7A binding. Nat Genet. 2018;50(4):498-503. 64. Uda M, Galanello R, Sanna S, et al. Genomewide association study shows BCL11A associated with persistent fetal hemoglobin and amelioration of the phenotype of betathalassemia. Proc Natl Acad Sci U S A. 2008;105(5):1620-1625. 65. Ginder GD, Whitters MJ, Pohlman JK, Burns

haematologica | 2019; 104(9)

Targeting sickle cell root-cause pathophysiology

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

78.

79.

80.

LJ. Transcriptional activation of an embryonic globin gene in adult erythroid cells in vivo. Prog Clin Biol Res. 1985;191:463-474. Scuto A, Kirschbaum M, Kowolik C, et al. The novel histone deacetylase inhibitor, LBH589, induces expression of DNA damage response genes and apoptosis in Phacute lymphoblastic leukemia cells. Blood. 2008;111(10):5093-100. Lee JH, Choy ML, Ngo L, Foster SS, Marks PA. Histone deacetylase inhibitor induces DNA damage, which normal but not transformed cells can repair. Proc Natl Acad Sci U S A. 2010;107(33):14639-14644. Conti C, Leo E, Eichler GS, et al. Inhibition of histone deacetylase in cancer cells slows down replication forks, activates dormant origins, and induces DNA damage. Cancer Res. 2010;70(11):4470-4480. Gaymes TJ, Padua RA, Pla M, et al. Histone deacetylase inhibitors (HDI) cause DNA damage in leukemia cells: a mechanism for leukemia-specific HDI-dependent apoptosis? Mol Cancer Res. 2006;4(8):563-573. Minucci S, Pelicci PG. Histone deacetylase inhibitors and the promise of epigenetic (and more) treatments for cancer. Nat Rev Cancer. 2006;6(1):38-51. Patthamalai P, Fuchareon S, Chaneiam N, et al. A phase 2 trial of HQK-1001 in HbE-beta thalassemia demonstrates HbF induction and reduced anemia. Blood. 2014;123(12): 1956-1957. van der Ploeg LH, Flavell RA. DNA methylation in the human gamma delta beta-globin locus in erythroid and nonerythroid tissues. Cell. 1980;19(4):947-958. Mavilio F, Giampaolo A, Care A, et al. Molecular mechanisms of human hemoglobin switching: selective undermethylation and expression of globin genes in embryonic, fetal, and adult erythroblasts. Proc Natl Acad Sci U S A. 1983;80(22):6907-6911. Tagle DA, Koop BF, Goodman M, et al. Embryonic epsilon and gamma globin genes of a prosimian primate (Galago crassicaudatus). Nucleotide and amino acid sequences, developmental regulation and phylogenetic footprints. J Mol Biol. 1988;203(2):439-455. DeSimone J, Heller P, Hall L, Zwiers D. 5Azacytidine stimulates fetal hemoglobin synthesis in anemic baboons. Proc Natl Acad Sci U S A. 1982;79(14):4428-4431. Ley TJ, DeSimone J, Anagnou NP, et al. 5azacytidine selectively increases gammaglobin synthesis in a patient with beta+ thalassemia. N Engl J Med. 1982;307(24):14691475. Ley TJ, DeSimone J, Noguchi CT, et al. 5Azacytidine increases gamma-globin synthesis and reduces the proportion of dense cells in patients with sickle cell anemia. Blood. 1983;62(2):370-380. Charache S, Dover G, Smith K, et al. Treatment of sickle cell anemia with 5-azacytidine results in increased fetal hemoglobin production and is associated with nonrandom hypomethylation of DNA around the gamma-delta-beta-globin gene complex. Proc Natl Acad Sci U S A. 1983;80(15):48424846. Dover GJ, Charache S, Boyer SH, Vogelsang G, Moyer M. 5-Azacytidine increases HbF production and reduces anemia in sickle cell disease: dose-response analysis of subcutaneous and oral dosage regimens. Blood. 1985;66(3):527-532. Lowrey CH, Nienhuis AW. Brief report: treatment with azacitidine of patients with end-stage beta-thalassemia. N Engl J Med. 1993;329(12):845-848.

haematologica | 2019; 104(9)

81. Koshy M, DeSimone J, Molokie R, et al. Augmentation of fetal hemoglobin (HbF) levels by low-dose short-duration 5 '-aza-2deoxycytidine (decitabine) administration in sickle cell anemia patients who had no HbF elevation following hydroxyurea therapy. Blood. 1998;92(10):30b-b. 82. Cui S, Kolodziej KE, Obara N, et al. Nuclear receptors TR2 and TR4 recruit multiple epigenetic transcriptional corepressors that associate specifically with the embryonic beta-type globin promoters in differentiated adult erythroid cells. Mol Cell Biol. 2011;31(16):3298-3311. 83. Santi DV, Garrett CE, Barr PJ. On the mechanism of inhibition of DNA-cytosine methyltransferases by cytosine analogs. Cell. 1983;33(1):9-10. 84. Clements EG, Mohammad HP, Leadem BR, et al. DNMT1 modulates gene expression without its catalytic activity partially through its interactions with histone-modifying enzymes. Nucleic Acids Res. 2012;40(10):4334-4346. 85. Brenner C, Luciani J, Bizet M, et al. The interplay between the lysine demethylase KDM1A and DNA methyltransferases in cancer cells is cell cycle dependent. Oncotarget. 2016;7(37):58939-58952. 86. Covey JM, D'Incalci M, Tilchen EJ, Zaharko DS, Kohn KW. Differences in DNA damage produced by incorporation of 5-aza-2'deoxycytidine or 5,6-dihydro-5-azacytidine into DNA of mammalian cells. Cancer Res. 1986;46(11):5511-5517. 87. Schermelleh L, Haemmer A, Spada F, et al. Dynamics of Dnmt1 interaction with the replication machinery and its role in postreplicative maintenance of DNA methylation. Nucleic Acids Res. 2007;35(13):43014312. 88. Zauri M, Berridge G, Thezenas ML, et al. CDA directs metabolism of epigenetic nucleosides revealing a therapeutic window in cancer. Nature. 2015;524(7563):114-118. 89. Almqvist H, Axelsson H, Jafari R, et al. CETSA screening identifies known and novel thymidylate synthase inhibitors and slow intracellular activation of 5-fluorouracil. Nat Commun. 2016;7:11040. 90. Olivieri NF, Saunthararajah Y, Thayalasuthan V, et al. A pilot study of subcutaneous decitabine in beta-thalassemia intermedia. Blood. 2011;118(10):2708-2711. 91. Saunthararajah Y, Sekeres M, Advani A, et al. Evaluation of noncytotoxic DNMT1depleting therapy in patients with myelodysplastic syndromes. J Clin Invest. 2015;125(3):1043-1055. 92. Milhem M, Mahmud N, Lavelle D, et al. Modification of hematopoietic stem cell fate by 5aza 2 ' deoxycytidine and trichostatin A. Blood. 2004;103(11):4102-4110. 93. Hu Z, Negrotto S, Gu X, et al. Decitabine maintains hematopoietic precursor selfrenewal by preventing repression of stem cell genes by a differentiation-inducing stimulus. Mol Cancer Ther. 2010;9(6):1536-1543. 94. DeSimone J, Koshy M, Dorn L, et al. Maintenance of elevated fetal hemoglobin levels by decitabine during dose interval treatment of sickle cell anemia. Blood. 2002;99(11):3905-3908. 95. Molokie R, Lavelle D, Gowhari M, et al. Oral tetrahydrouridine and decitabine for non-cytotoxic epigenetic gene regulation in sickle cell disease: a randomized phase 1 study. PLoS Med. 2017;14(9):e1002382. 96. Santi DV, Norment A, Garrett CE. Covalent bond formation between a DNA-cytosine methyltransferase and DNA containing 5-

azacytosine. Proc Natl Acad Sci USA. 1984; 81(22):6993-6997. 97. Patel K, Dickson J, Din S, et al. Targeting of 5-aza-2'-deoxycytidine residues by chromatin-associated DNMT1 induces proteasomal degradation of the free enzyme. Nucleic Acids Res. 2010;38(13):4313-4324. 98. Creusot F, Acs G, Christman JK. Inhibition of DNA methyltransferase and induction of Friend erythroleukemia cell differentiation by 5-azacytidine and 5-aza-2'-deoxycytidine. J Biol Chem. 1982;257(4):2041-2048. 99. Shi L, Cui S, Engel JD, Tanabe O. Lysine-specific demethylase 1 is a therapeutic target for fetal hemoglobin induction. Nat Med. 2013;19(3):291-294. 100. Cui S, Lim KC, Shi L, et al. The LSD1 inhibitor RN-1 induces fetal hemoglobin synthesis and reduces disease pathology in sickle cell mice. Blood. 2015;126(3):386396. 101. Jagadeeswaran R, Vazquez BA, Thiruppathi M, et al. Pharmacological inhibition of LSD1 and mTOR reduces mitochondrial retention and associated ROS levels in the red blood cells of sickle cell disease. Exp Hematol. 2017;50:46-52. 102. Rivers A, Vaitkus K, Ibanez V, et al. The LSD1 inhibitor RN-1 recapitulates the fetal pattern of hemoglobin synthesis in baboons (P. anubis). Haematologica. 2016;101(6):688697. 103. Hartley PD, Madhani HD. Mechanisms that specify promoter nucleosome location and identity. Cell. 2009;137(3):445-458. 104. Parnell TJ, Huff JT, Cairns BR. RSC regulates nucleosome positioning at Pol II genes and density at Pol III genes. EMBO J. 2008;27(1):100-110. 105. Velcheti V, Schrump D, Saunthararajah Y. Ultimate precision: targeting cancer but not normal self-replication. Am Soc Clin Oncol Educ Book. 2018;38:950-963. 106. Kingsley PD, Malik J, Emerson RL, et al. "Maturational" globin switching in primary primitive erythroid cells. Blood. 2006;107 (4):1665-1672. 107. Mabaera R, Richardson CA, Johnson K, et al. Developmental- and differentiation-specific patterns of human gamma- and beta-globin promoter DNA methylation. Blood. 2007;110 (4):1343-1352. 108. McGrath K, Palis J. Ontogeny of erythropoiesis in the mammalian embryo. Curr Top Dev Biol. 2008;82:1-22. 109. Ni H, Yang XD, Stoeckert CJ, Jr. Maturation and developmental stage-related changes in fetal globin gene expression are reproduced in transiently transfected primary adult human erythroblasts. Exp Hematol. 1999;27 (1):46-53. 110. Papayannopoulou TH, Brice M, Stamatoyannopoulos G. Stimulation of fetal hemoglobin synthesis in bone marrow cultures from adult individuals. Proc Natl Acad Sci U S A. 1976;73(6):2033-2037. 111. Velcheti V, Radivoyevitch T, Saunthararajah Y. Higher-level pathway objectives of epigenetic therapy: a solution to the p53 problem in cancer. Am Soc Clin Oncol Educ Book. 2017;37:812-824. 112. Saleh MFM, Saunthararajah Y. Severe pyoderma gangrenosum caused by myelodysplastic syndrome successfully treated with decitabine administered by a noncytotoxic regimen. Clin Case Rep. 2017;5(12):20252027. 113. Liu Y, Tabarroki A, Billings S, et al. Successful use of very low dose subcutaneous decitabine to treat high-risk myelofibrosis with Sweet syndrome that was

1729

Y. Saunthararajah et al. refractory to 5-azacitidine. Leuk Lymphoma. 2014;55(2):447-449. 114. Tabarroki A, Saunthararajah Y, Visconte V, et al. Ruxolitinib in combination with DNA methyltransferase inhibitors: clinical responses in patients with symptomatic myelofibrosis with cytopenias and elevated blast(s) counts. Leuk Lymphoma. 2015;56(2):497-499. 115. Saunthararajah Y, Triozzi P, Rini B, et al. p53-independent, normal stem cell sparing epigenetic differentiation therapy for myeloid and other malignancies. Semin Oncol. 2012;39(1):97-108. 116. Monod J, Wyman J, Changeux JP. On the nature of allosteric transitions: a plausible model. J Mol Biol. 1965;12:88-118. 117. Eaton WA, Henry ER, Hofrichter J, et al. Evolution of allosteric models for hemoglobin. IUBMB Life. 2007;59(8-9):586-599. 118. Hebbel RP, Hedlund BE. Sickle hemoglobin oxygen affinity-shifting strategies have unequal cerebrovascular risks. Am J Hematol. 2018;93(3):321-325. 119. Beutler E. The effect of methemoglobin formation in sickle cell disease. J Clin Invest. 1961;40:1856-1871. 120. Howard J, Thein SL, Galacteros F, et al. Safety and tolerability of MP4CO: a dose escalation study in stable patients with sick-

1730

le cell disease. Blood. 2013;122(21):2205. 121. Iyamu EW, Turner EA, Asakura T. In vitro effects of NIPRISAN (Nix-0699): a naturally occurring, potent antisickling agent. Br J Haematol. 2002;118(1):337-343. 122. Swift R, Abdulmalik O, Chen QK, et al. SCD-101: a new anti-sickling drug reduces pain and fatigue and improves red blood cell shape in peripheral blood of patients with sickle cell disease. Blood. 2016;128 (22):121. 123. Howard J, Hemmaway CJ, Telfer P, et al. A phase 1/2 ascending dose study and openlabel extension study of voxelotor in patients with sickle cell disease. Blood. 2019;133(17):1865-1875. 124. Vichinsky E, Hoppe CC, Ataga KI, et al. A phase 3 randomized trial of voxelotor in sickle cell disease. N Engl J Med. 2019 Jun 14. [Epub ahead of print] 125. Watchmaker JM, Juttukonda MR, Davis LT, et al. Hemodynamic mechanisms underlying elevated oxygen extraction fraction (OEF) in moyamoya and sickle cell anemia patients. J Cereb Blood Flow Metab. 2018;38(9):1618-1630. 126. Choudhury NA, DeBaun MR, Ponisio MR, et al. Intracranial vasculopathy and infarct recurrence in children with sickle cell anaemia, silent cerebral infarcts and normal

transcranial Doppler velocities. Br J Haematol. 2018;183(2):324-326. 127. Ford AL, Ragan DK, Fellah S, et al. Silent infarcts in sickle cell disease occur in the border zone region and are associated with low cerebral blood flow. Blood. 2018;132(16): 1714-1723. 128. Quinn CT. Silent cerebral infarction: supply and demand. Blood. 2018;132(16):16321634. 129. Niihara Y, Miller ST, Kanter J, et al. A Phase 3 trial of L-glutamine in sickle cell disease. N Engl J Med. 2018;379(3):226-235. 130. Peled T, Shoham H, Aschengrau D, et al. Nicotinamide, a SIRT1 inhibitor, inhibits differentiation and facilitates expansion of hematopoietic progenitor cells with enhanced bone marrow homing and engraftment. Exp Hematol. 2012;40(4):342355.e1. 131. Anand S, Thomas S, Hyslop T, et al. Transplantation of ex vivo expanded umbilical cord blood (NiCord) decreases early infection and hospitalization. Biol Blood Marrow Transplant. 2017;23(7):1151-1157. 132. Horwitz ME, Chao NJ, Rizzieri DA, et al. Umbilical cord blood expansion with nicotinamide provides long-term multilineage engraftment. J Clin Invest. 2014;124(7):31213128.

haematologica | 2019; 104(9)

ARTICLE

Hematopoiesis

Bone marrow adipose tissue-derived stem cell factor mediates metabolic regulation of hematopoiesis

Ferrata Storti Foundation

Zengdi Zhang,1* Zan Huang,1,2,3* Brianna Ong,1 Chinmayi Sahu,1 Hu Zeng,4,5 and Hai-Bin Ruan1,6

Department of Integrative Biology and Physiology, University of Minnesota Medical School, Minneapolis, MN, USA; 2Laboratory of Gastrointestinal Microbiology, Jiangsu Key Laboratory of Gastrointestinal Nutrition and Animal Health, College of Animal Science and Technology, Nanjing Agriculture University, Nanjing, Jiangsu, China; 3National Center for International Research on Animal Gut Nutrition, Nanjing Agriculture University, Nanjing, Jiangsu, China; 4Division of Rheumatology, Department of Medicine, Mayo Clinic, Rochester, MN, USA; 5Department of Immunology, Mayo Clinic, Rochester, MN, USA and 6Center for Immunology, University of Minnesota, Minneapolis, MN, USA 1

*These authors contributed equally.

Haematologica 2019 Volume 104(9):1731-1743

ABSTRACT

ematopoiesis is dynamically regulated by metabolic cues in homeostatic and stressed conditions; however, the cellular and molecular mechanisms mediating the metabolic sensing and regulation remain largely obscure. Bone marrow adipose tissue remodels in various metabolic conditions and has been recently proposed as a niche for hematopoietic stem cells after irradiation. Here, we investigated the role of marrow adipose tissue-derived hematopoietic cytokine stem cell factor in unperturbed hematopoiesis by selectively ablating the Kitl gene from adipocytes and bone marrow stroma cells using Adipoq-Cre and Osx1-Cre, respectively. We found that both Adipoq-Kitl knockout (KO) and Osx1-Kitl KO mice diminished hematopoietic stem and progenitor cells and myeloid progenitors in the bone marrow and developed macrocytic anemia at the steady-state. The composition and differentiation of hematopoietic progenitor cells in the bone marrow dynamically responded to metabolic challenges including high fat diet, Î˛3-adrenergic activation, thermoneutrality, and aging. However, such responses, particularly within the myeloid compartment, were largely impaired in Adipoq-Kitl KO mice. Our data demonstrate that marrow adipose tissue provides stem cell factor essentially for hematopoiesis both at the steady state and upon metabolic stresses.

Correspondence: RUAN HAI-BIN hruan@umn.edu Received: September 3, 2018. Accepted: February 18, 2019. Pre-published: February 21, 2019. doi:10.3324/haematol.2018.205856

Introduction The metabolic and hematopoietic systems demonstrate dynamic and complex interplays in health and disease. On the one hand, a plethora of blood cells including granulocytes, monocytes and macrophages, mast cells, and lymphocytes contribute to the physiological and pathological regulation of energy intake and expenditure, glucose and lipid metabolism, bone remodeling, and the aging process.1-4 On the other hand, hematopoietic stem cells (HSC), myeloid and lymphoid progenitors, and their mature progeny not only impose different bioenergetic demands during development,5-8 but also show flexibility and plasticity in their maintenance, differentiation, and function in response to various metabolic disturbances, such as obesity, hyperglycemia, and aging.9-16 A large body of evidence indicates that both intrinsic and extrinsic factors drive the hematopoietic process; however, the cellular and molecular mechanisms underlying the metabolic regulation of hematopoiesis remain for the most part elusive. In adults, the bone marrow (BM) microenvironment provides niches that support the renewal, quiescence, and differentiation of hematopoietic stem and progenitor cells (HSPC).17-19 Recent studies have started to unveil the complexity and heterogeneity of niche cell types, niche factors, and their actions. BM mesenchymal stem cells [BMSC, haematologica | 2019; 104(9)

Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1731 ÂŠ2019 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.

1731

Z. Zhang et al.

Figure 1. Role of adipose stem cell factor (SCF) in brown fat function. (A and B) Stromal vascular faction (SVF) cells from iWAT of Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl mice were differentiated into adipocytes in vitro and analyzed by Oil O Red staining (A) and western blotting (B). Densitometry of UCP1 shown in Online Supplementary Figure S1C. (C and D) Expression of proteins involved in adipogenesis and thermogenesis in BAT (C) and iWAT (D) from 7-week old Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl male mice. Densitometry of UCP1 shown in Online Supplementary Figure S1D and E. (E and F) Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl mice (n=6) were treated with CL 316,243 for seven days. Ucp1 mRNA levels (E) and UCP1 protein levels (F) in BAT and iWAT were determined. Densitometry of UCP1 shown in Online Supplementary Figure S1J and K. Data are presented as meanÂąStandard Deviation.

also known as skeletal stem cells (SSC)] and their adipogenic, osteogenic, and chondrogenic progeny are major contributors of niche factors, such as stem cell factor (SCF) and CXC chemokine ligand 12 (CXCL12).20-22 The sympathetic nervous system (SNS) extensively innervates the bone and BM to control hematopoietic homeostasis and regeneration via direct actions on HSPC and indirect actions on the niche.23 In addition, signals from the vascular endothelial cells and the HSC progeny such as macrophages and megakaryocytes have also been shown to contribute to different aspects of HSPC regulation.17 Nevertheless, whether these niche constituents mediate the sensing of metabolic cues and subsequent remodeling in hematopoiesis has not yet been determined. White adipose tissue (WAT) that stores excess energy and brown adipose tissue (BAT) that dissipates energy as heat are key determinants of metabolic homeostasis. The role of BM adipose tissue (MAT), the third major adipose depot in the body, is just beginning to be revealed. Developmentally, BM adipocytes arise from the same Osterix+ skeletal lineage as osteoblasts and chondrocytes.24-26 Anatomically, constitutive MAT (cMAT) is found in the most distal portion of the tibia and tail vertebrae while regulated MAT (rMAT) is found in the proximal skeletal sites.27-29 Although cMAT is relatively stable, rMAT expands in conditions like obesity, diabetes, caloric restriction, and aging.27-29 Functionally, there are tripartite interactions between MAT, bone, and hematopoiesis, yet their mechanistic characteristics are still not fully understood.30 An early study taking advantage of the genetic and pharmacological inhibition of adipogenesis suggested MAT to be a negative regulator of the hematopoietic microenvironment.31 In contrast, recent work demonstrated that MAT supports HSC regeneration and myeloerythroid maturation following irradiation and reconstitution, partially by secreting SCF.32,33 1732

The close relationship between hematopoiesis and metabolism is also represented by their regulation by common growth factors and cytokines. SCF and its receptor KIT play an essential role in the survival, migration, and differentiation of multiple stem and progenitor cells including HSPC.34 In the hematopoietic system, loss-offunction mutations in SCF/KIT cause macrocytic anemia while gain-of-function mutations lead to systematic mastocytosis, acute myeloid leukemia, and lymphoma.35,36 In the bone marrow niche, SCF is expressed in LEPR+ stroma cells, endothelial cells, and adipocytes, but not in osteoblasts or hematopoietic cells.22,32 Deleting SCF selectively in these positive niche cells leads to defects in HSC maintenance.22,32 In the metabolic system, SCF has been shown to promote the differentiation of brown adipocytes from human pluripotent stem cells and to be essential to mitochondrial function and energy expenditure in mice.37,38 However, the cellular source of SCF in regulating systemic metabolism has not been determined. Here, we investigated the contribution of adipose-derived SCF in regulating energy and glucose metabolism and in mediating the effect of metabolic stresses on unperturbed hematopoiesis.

Methods Mice All mice used in this study were purchased from the Jackson Laboratory, including Kitlfl/fl (stock n. 017861), Adipoq-Cre (stock n. 010803), Osx1-Cre (stock n. 006361), and KitlEGFP (stock n. 017860). All animals were kept on a 14 hour (h):10 h light:dark cycle in the animal facility at the University of Minnesota, Minneapolis, MN, USA. Mice were group-housed, with free access to water and either a standard chow diet or 60% high fat diet (Research Diets, D12492). 1 mg/kg BW of CL-316, 243 (R&D Systems, #1499, haematologica | 2019; 104(9)

MAT mediates metabolic stress in HSPC

Figure 2. Adipocyte-derived stem cell factor (SCF) is a niche factor for hematopoietic stem and progenitor cells. (A) Co-staining of Perilipin and EGFP in the cMAT and rMAT of KitlEGFP mice. Scale bar=50 μm. (B and C) Levels of total Kitl mRNA (B) and SCF protein (C) in the flushed bone marrow (BM) from tibia of Kitlfl/fl (n = 5) and Adipoq-Cre+;Kitlfl/fl (n=6) 13-week old male mice. (D) BM cellularity in the femur of 13-week old Kitlfl/fl (n = 7) and Adipoq-Cre+;Kitlfl/fl (n=8) male mice. (E) Representative flow cytometry plots showing LSK and MP cells among the lineage– CD127– population in 13-week old Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl male mice. Average frequencies are shown as inserts. (F) Quantification of absolute numbers of LSK and MP cells in the femur of 13-week old Kitlfl/fl (n=7) and Adipoq-Cre+;Kitlfl/fl (n=8) male mice and phenotypic LT-HSC in the femur of 8-month old Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl (n=4) male mice. (G) Representative flow cytometry plots showing CMP, MEP, and GMP cells among the MP population in 13-week old Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl male mice. Average frequencies are shown as inserts. (H) Absolute numbers of CMP, MEP, and GMP cells in the femur of 13-week old Kitlfl/fl (n=7) and Adipoq-Cre+;Kitlfl/fl (n=8) male mice. (I) The ratio of marrow MEP to GMP in 13-week old Kitlfl/fl (n=7) and Adipoq-Cre+;Kitlfl/fl (n=7) male mice. (J) The absolute number of CLP in 13-week old Kitlfl/fl (n=7) and Adipoq-Cre+;Kitlfl/fl (n=7) male mice. (K) The ratio of marrow CMP to CLP in 13-week old Kitlfl/fl (n=7) and Adipoq-Cre+;Kitlfl/fl (n=7) male mice. (L) Colony formation assay of 2x104 BM cells from Kitlfl/fl and AdipoqCre+;Kitlfl/fl (n=4) mice. (M) Spleen weight of Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl (n=4) mice. (N) Colony formation assay of 2x105 splenic cells from Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl (n=4) mice. Data are presented as mean±Standard Deviation. *P<0.05; **P<0.01; ***P<0.001 by unpaired student t-test (B-K) or one-way ANOVA (N).

diluted in saline) were intraperitoneally (i.p.) injected when indicated. For thermoneutral housing, mice born at 22°C were transferred to a room maintained continuously at 30°C at the age indicated. All procedures involving animals were conducted within Institutional Animal Care and Use Committee guidelines under approved protocols.

Flow cytometry Bone marrow cells were isolated by flushing the femur in Ca2+ and Mg2+ free PBS with 1% heat-inactivated bovine serum. Cells were dissociated to a single cell suspension by gently passing through a 25-gauge needle and then filtering through a 70-mm nylon mesh. Red blood cells from BM were removed by ammonium-chloride-potassium lysing buffer. For flow analyses, BM cells were stained with a cocktail of biotin-conjugated lineage antibodies CD3e, B220, Ter119, Mac-1 and Gr-1 (Biolegend, 133307), CD4 (Biolegend, 100403), CD5 (Biolegend, 100603), CD8 (Biolegend, 100703), followed by Streptavidin-AF488 (Biolegend, 405235). Cells were then stained with CD127-APC (eBioscience, 17-1271-82), c-Kit-APC-eFluor780 (eBioscience, 47-1171-82), Sca1-Super Bright 436 (eBioscience, 62-5981-82), CD34-PE (Biolegend, 152204), and FcγR-PerCP-eFluor710 (eBioscience, 460161-80), CD150-BV605 (Biolegend, 115927), and CD48-BUV395 haematologica | 2019; 104(9)

(BDBioscience, 740236). Multicolor analysis for progenitor and stem-cell quantification was performed on a 3-laser-LSRII flow cytometer (BD).12,39 HSPC was defined as Lin–Sca-1+c-Kit+ (LSK); phenotypic LT-HSC was defined as CD150+CD48–Lin–Sca–1+cKit+; myeloid progenitor (MP) was defined as Lin–CD127–Sca–1–cKit+; common lymphoid progenitor (CLP) was defined as Lin–CD127+Sca–1+c-Kit+; common myeloid progenitor (CMP) was defined as Lin–CD127-Sca–1–c-Kit+CD34+ FcγR–; megakaryocyteerythrocyte progenitor (MEP) was defined as Lin–CD127–Sca–1–cKit+CD34–FcγR–; granulocyte-monocyte progenitor (GMP) was defined as Lin–CD127–Sca–1–c-Kit+CD34+FcγR+. SYTOX™ Green Dead Cell Stain (Thermo Fisher Scientific, S34860) was used to exclude dead cells. Absolute number was obtained by using counting beads (Thermo Fisher Scientific, C36950) as instructed by the manufacturer.

Metabolic assays Body composition was assessed using an EchoMRI system by which fat and lean mass measured by magnetic resonance scanning were normalized to body weight for fat percentage and lean percentage. Adipose tissue weight was determined by dissecting and weighing indicated adipose depots. For glucose tolerance tests, 16 h-fasted mice were injected i.p. with glucose (1.5 g/kg 1733

Z. Zhang et al. Table 1. Decline in production of mature blood cells in the periphery of Adipoq-Cre+;Kitlfl/fl mice.

Kitlfl/fl (n = 6) RBC (x1012/L) HGB (g/L) Hematocrit L/L MCV (fL) Reticulocyte (x1012/L) Platelet (x109/L) Neutrophil (x109/L) Eosinophil (x109/L) Basophil( x109/L) Monocyte (x109/L) Lymphocyte (x109/L)

13-week-old male Adipoq-Cre+;Kitlfl/fl (n = 9)

10.54 ± 0.34 150.0 ± 3.7 0.493 ± 0.014 46.73 ± 0.50 0.29 ± 0.28 1193.2 ± 115.6 0.90 ± 0.33 0.23 ± 0.10 0.04 ± 0.02 0.06 ± 0.02 6.10 ± 2.08

8.68 ± 0.28 *** 137.6 ± 4.7 *** 0.450 ± 0.015 *** 51.91 ± 1.04 *** 0.34 ± 0.31 ** 987.6 ± 99.9 ** 0.45 ± 0.20 * 0.10 ± 0.06 * 0.01 ± 0.01 * 0.03 ± 0.01 * 3.71 ± 1.01 *

Kitlfl/fl (n = 8)

14-week-old female Adipoq-Cre+;Kitlfl/fl (n = 7)

10.48 ± 0.40 152.0 ± 3.8 0.502 ± 0.026 47.89 ± 2.56 0.30 ± 0.04 1027.8 ± 104.3 0.87 ± 0.31 0.19 ± 0.11 0.04 ± 0.02 0.13 ± 0.05 5.67 ± 1.38

8.83 ± 0.27 *** 143.6 ± 4.5 ** 0.463 ± 0.018 ** 53.41 ± 2.28 *** 0.43 ± 0.11 * 806.4 ± 82.2 *** 0.38 ± 0.17 ** 0.13 ± 0.11 0.02 ± 0.01 0.04 ± 0.01 *** 3.73 ± 1.14 *

n: number; RBC: red blood cell counts; HGB: hemoglobin concentration; MCV: mean corpuscular volume. Data are presented as mean±Standard Deviation. *P<0.05; **P<0.01; ***P<0.001 by unpaired two-tailed t-test.

body weight). Blood glucose from tail-vein blood collected at the designated times was measured using a Bayer Contour Glucometer (9545C).

Cell culture The stromal vascular faction (SVF) cells derived from iWAT were obtained as previously described.38 SVF cells were cultured in DMEM/F12 (Corning, 10-090-CV) containing 10% FBS (GenClone, 25-514), 1% penicillin/streptomycin (Gibco, 10378016), 20 nM insulin (Sigma), and 1 nM triiodothyronine (T3, Sigma, T6397). Two days after becoming confluent (defined as Day 0), SVF cells were induced with DMEM/F12 containing 10% FBS, 1% penicillin/streptomycin, 0.5 mM isobutylmethylxanthine (IBMX, Sigma, I7018), 125 μM indomethacin (Sigma, I7378), 1 μM dexamethasone (Sigma, D4902), 20 nM insulin, and 1 nM T3 for 48 h. Cells were maintained in DMEM/F12 containing 10% FBS, 1% penicillin/streptomycin, 20 nM insulin, and 1 nM T3 until lipid drop appeared. This medium was replenished every two days.

Oil Red O staining Cells were washed with PBS and then fixed with 10% formaldehyde (Sigma) for 1 h. After washing with 60% isopropanol, fixed cells were stained with Oil Red O solution (2 mg/mL in 60% isopropyl alcohol, Sigma, O-0625) for ten minutes (min), and rinsed five times with pure H2O before photographic images were taken.

Genotyping and quantitative real-time polymerase chain reaction Primers used for DNA amplification were: floxed-Kitl-F, CGAGGTAGGGGAAAAGAACC; floxed-Kitl-R, GGATCTTCCCAGAGGTTGGA; excised-Kitl-F, GGAAAAGAACCAAGTGAAGTC; excised-Kitl-R, ACGGGGAAAGACCTCCGGTCC; Adipoq-Cre-F, GGAAAAGAACCAAGTGAAGTC; Adipoq-CreR, ACGGGGAAAGACCTCCGGTCC. DNA for verifying knockout was isolated using the Quick-DNA Plus Kit (Zymo Research, #D4074). Genotyping was performed according to instructions in the manual provided by the Jackson Laboratory. Total RNA from tissues was isolated using TRIzol (Invitrogen, 15596018). RNA was reverse-transcribed using iScript cDNA Synthesis Kits (Bio-Rad, 170-8891). Quantitative real-time poly1734

merase chain reaction (qRT-PCR) was performed using SYBR Green Supermix (Bio-Rad, 1725124) with a C1000 Thermal Cycler (Bio-Rad) following the manufacturers’ instructions. For thermal cycling: 95°C, 3 min; then 40 cycles of 95°C, 10 seconds (s) and 60°C, 30 s. The primers used for qRT-PCR were long Kitl (flanking exon 5 and 6): 5’-GCCAGAAACTAGATCCTTTACTCCTGA-3’ and 5’-ACATAAATGGTTTTGTGACACTGACTCTG-3’; short Kitl (flanking junctions between exon 5/7 and exon 8/9): 5’-CCCGAGAAAGGGAAAGCCG-3’ and 5’-ATTCTCTCTCTTTCTGTTGCAACATACTT-3’; total Kitl (flanking exon 2 and 3): 5’TCTGCGGGAATCCTGTGACT-3’ and 5’-CGGCGACATAGTTGAGGGTTAT-3’; excised Kitl (flanking exon 1 and 2, exon1 was floxed): 5’-CAGCGCTGCCTTTCCTTATGA-3’ and 5’-ATCAGTCACAGGATTCCCGC-3’, and the housekeeping gene 36b4: 5’-AGATGCAGCAGATCCGCAT-3’ and 5’GTTCTTGCCCATCAGCACC-3’. Delta-delta Ct analysis was used to calculate relative gene expression.

Histology and immunostaining Long bones were fixed in 10% neutral buffered formalin followed by three days of decalcification in 14% EDTA, followed by paraffin or OCT embedding. Bones embedded in paraffin were sectioned at 5 μm thickness using a microtome (Olympus Cut 4060) and stained with Hematoxylin and Eosin (H&E). For quantification of the BM adipocytes, the BM areas 4.5 mm from growth plate were selected. Bones embedded in OCT were sectioned at 7 μm thickness using a cryostat (Leica). Sections were blocked with 3% bovine serum albumin, 0.2% TWEEN 20 in Trisbuffered saline, incubated with chicken-anti-GFP (Aves, GFP1020, 1:1,000), rabbit-anti-perilipin (Cell Signaling Technology, #9349, 1:200) overnight, and secondary antibodies (Alexa Fluor 488 anti-Chicken IgG and Alexa Fluor 674 anti-Rabbit IgG, Life Technologies, 1:400) for 1 h. A Nikon system was used for fluorescence detection.

Western blot Tissue proteins were extracted using RIPA buffer with freshly added proteinase inhibitors. Protein concentrations were determined using BCA Protein Assay Kit (Pierce). Equal amounts of protein samples were subjected to western blot. The following antibodies were used: anti-UCP1 (Abcam, ab209483, 1:5000 dilution), anti-PGC-1a (Bioworld, BS72263, 1:500 dilution), anti-PERhaematologica | 2019; 104(9)

MAT mediates metabolic stress in HSPC

Figure 3. Impaired hematopoiesis when stem cell factor (SCF) is ablated in Osx1+ cells. (A) Bone marrow cellularity in the femur of 8-month old control (n=4), Osx1Cre+;Kitlfl/+ (n=5), and Osx1-Cre+;Kitlfl/fl (n=4) male mice. (B) Frequencies of LSK, phenotypic LT-HSC, MP, CMP, GMP, MEP, and CLP populations in the bone marrow (BM) of 8-month old control (n=4), Osx1-Cre+;Kitlfl/+ (n=5), and Osx1-Cre+;Kitlfl/fl (n=4) male mice, determined by flow cytometry. (C and D) Ratios of MEP to GMP (C) and CMP to CLP (D) in 8-month old control and Osx1-Cre+;Kitlfl/fl (n=4) male mice. (E) Complete blood count of 10-week old control (n=8, including 6 Kitlfl/fl and 2 Osx1-Cre+) and Osx1-Cre+;Kitlfl/fl (n=7) male mice. (F) Colony formation assay of 2x104 BM cells from Kitlfl/fl and Osx1-Cre+;Kitlfl/fl (n=4) mice. (G) Spleen weight of Kitlfl/fl and Osx1-Cre+;Kitlfl/fl (n=4) mice. (H) Colony formation assay of 2x105 splenic cells from Kitlfl/fl and Osx1-Cre+;Kitlfl/fl (n=4) mice. Data are presented as mean±Standard Deviation. *P<0.05; **P<0.01; ***P<0.001 by one-way ANOVA followed with Tukey’s multiple comparison (A, B, and H) or unpaired Student t-test (C, E, and G).

ILIPIN (Cell Signaling Technology, 9349T, 1:1000 dilution), antiPPARγ (CusAb, CSB-PA018424LA01HU, 1:500 dilution), antiCOX4 (Proteintech, 11242-1-AP, 1:1000 dilution), and anti-ACTIN (Sigma, A5441, 1:5000 dilution). Densitometry was performed using Image J. Relative band density was calculated by dividing the densitometry of target protein with loading control from the same membrane.

Colony-forming unit assay The colony-forming unit (CFU) assay was performed using MethoCult™ GF M3434 (Stem cell) according to the manufacturer’s instructions. Briefly, BM cells were flushed from one femur, and filtered through a 40 um cell strainer. Spleen was minced and pressed through a 40 um cell strainer to obtain single cell suspension. 2x104 BM cells or 2x105 spleen cells were plated in methylcellulose, and the CFU were counted after 12 days.

Isolation of bone marrow adipocytes and supernatant A 0.6 mL microcentrifuge tube was cut open at the bottom and placed into a 1.5 mL microcentrifuge tube. Long bones were snipped both ends and placed in the prepared 0.6 mL microcentrifuge tube. BM was flowed out by quick centrifuge (from 0 to 10,000 rpm, RT). Red blood cells from BM were lysed by ammonium-chloride-potassium lysing buffer. After 3,000 rpm centrifugation for 5 min at RT, floating adipocytes were collected as BM adipocytes from the top layer.40 The middle layer was collected and remaining cells were spun down by centrifugation (12,000 rpm, 30 s, RT). The supernatant was collected for SCF measurement using an ELISA (Thermo Fisher Scientific, EMKITL). haematologica | 2019; 104(9)

Statistical analysis All data are presented as mean±Standard Deviation (SD) or mean±Standard Error of Mean (SEM) as indicated in the figure legends. The statistical significance between two groups was determined by unpaired two-tailed Student t-test (Microsoft Excel or GraphPad Prism 7). Datasets involving more than two groups were assessed by one-way ANOVA with Tukey’s correction for multiple comparisons using GraphPad Prism 7. Two-way ANOVA followed by Tukey’s correction or Sidak’s correction for multiple comparisons was performed using GraphPad Prism 7 to examine data with two independent variables. *P<0.05; **P<0.01; ***P<0.001. 1735

Z. Zhang et al.

Results Adipocyte-derived stem cell factor is not essential for brown fat function in vivo To determine whether adipocyte-derived SCF cellautonomously regulates BAT function in vivo, we generated fat-specific SCF KO mice by crossing Adiponectin (Adipoq)-Cre mice with Kitlfl/fl mice. The adipose stromal vascular faction (SVF) cells derived from Kitlfl/fl controls and Adipoq-Cre+;Kitlfl/fl KO mice were isolated and differentiated into adipocytes in culture. The Kitl gene was specifically deleted (Online Supplementary Figure S1A) and the Kitl mRNA reduced its level (Online Supplementary Figure S1B) in KO cells when Adipoq-Cre started to be expressed after adipogenic induction (Online Supplementary Figure S1B). Control and KO SVF cells showed similar capacity in adipogenic differentiation, determined by Oil Red O staining (Figure 1A). Consistent with the previous finding that SCF is required for UCP1 expression,37 we found that differentiated brown adipocytes from Adipoq-Cre+;Kitlfl/fl mice had much lower levels of UCP1 protein compared to Kitlfl/fl cells (Figure 1B and Online Supplementary Figure S1C). Expression of PGC-1a and PERILIPIN was unaffected (Figure 1B). We then sought to determine gene expression in thermogenic fat tissues in vivo. The levels of UCP1, PGC-1a, COX4, and PPARγ proteins in BAT and inguinal WAT (iWAT) were essentially the same between Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl mice (Figure 1C and D and Online Supplementary Figure S1D and E). Kitlfl/fl and AdipoqCre+;Kitlfl/fl mice had similar body weight at 7, 14, and 28 weeks of age (Online Supplementary Figure S1F-H). No changes in the mass of BAT, iWAT, and gonadal WAT (gWAT) (Online Supplementary Figure S1F), composition of lean and fat mass (Online Supplementary Figure S1G), or fasting body weight (Online Supplementary Figure S1H) were observed. In addition, systemic glucose metabolism shown by the glucose tolerance test was also comparable between the two genotypes (Online Supplementary Figure S1I). To assess sympathetic nerve-activated adaptive thermogenesis, we treated mice with a β3-adrenoceptor agonist, CL 316,243 for seven days. qRT-PCR and western blotting showed no difference in levels of Ucp1 mRNA (Figure 1E) or UCP1 protein (Figure 1F and Online Supplementary Figure S1J and K) in BAT or iWAT between Kitlfl/fl and AdipoqCre+;Kitlfl/fl mice. Together, these data demonstrate that SCF secreted by adipocytes is essential for UCP1 expression in vitro but is not essential for energy metabolism in vivo. It is possible that SCF from non-adipose cells or tissues may compensate for the loss of SCF in adipocytes, as no change in serum levels of SCF was observed in AdipoqCre+;Kitlfl/fl mice (Online Supplementary Figure S1L).

Adipocyte-derived stem cell factor controls steady-state hematopoiesis Marrow adipose tissue is an endocrine organ important for hematopoiesis and systemic metabolism.27-30 MAT promotes the regeneration of hematopoietic stem cells after irradiation by secreting SCF.32 We first confirmed that most BM adipocytes (approx. 77% in cMAT and approx. 80% in rMAT) expressed SCF by performing perilipin immunofluorescent staining on bone sections of KitlEGFP knockin mice (Figure 2A and Online Supplementary Figure S2A). Compared to BAT and iWAT, the flushed BM expressed 1736

similar levels of the long isoform of Kitl that can be transcribed and cleaved into the soluble form of SCF, but much less of the short Kitl transcript that encodes the membrane-bound SCF (Online Supplementary Figure S2B).34 Note that there are significant numbers of non-adipocytes in the adipose tissues and BM analyzed. In BM, AdipoqCre was recently shown to only label mature adipocytes, but not bone stroma, adipogenic progenitors, hematopoietic cells, bone lining cells, or osteoblast cells.41 In AdipoqCre+;Kitlfl/fl mice, the Kitl gene was specifically knocked down in marrow adipocytes (Online Supplementary Figure S2C). We could observe a significant loss of Kitl transcript in the flushed BM (Figure 2B) and SCF protein in the BM supernatant (Figure 2C), suggesting that MAT is a major source of SCF in the BM. We then quantified hematopoietic stem and progenitor cells (HSPC) in the BM of Adipoq-Cre+;Kitlfl/fl mice by flow cytometry (Online Supplementary Figure S3).12,39 We first characterized the Adipoq-Cre+ line and could not observe any potential Cre-specific defects in BM cellularity or HSPC numbers when compared to wild-type mice (Online Supplementary Figure S4). Thus, Kitlfl/fl mice were used as controls for comparison in the following experiments. Loss of SCF specifically in adipocytes reduced marrow cellularity in male mice (Figure 2D). The frequency and also absolute number of lineage–Sca–1+c-Kit+ (LSK) HSPC, phenotypic long-term (LT)-HSC (e.g. CD150+CD48– LSK cells), and myeloid progenitors (MP) were all down-regulated in Adipoq-Cre+; Kitlfl/fl male mice (Figure 2E and F). Within MP, common myeloid progenitors (CMP), megakaryocyte-erythrocyte progenitors (MEP), and granulocyte-monocyte progenitors (GMP) all showed decreased frequency and number (Figure 2G and H). Interestingly, there was a reduction in the ratio of MEP to GMP (Figure 2I), suggesting that the extent of dependence on SCF varies between different myeloid progenitors. On the other hand, common lymphoid progenitors (CLP) maintained their number (Figure 2J) and the ratio of CMP/CLP was reduced (Figure 2K) in Adipoq-Cre+;Kitlfl/fl mice. Colony formation assay showed that BM HSPC, though reducing their numbers in the SCF-deficient environment, did not show functional decline when assessed in a complete medium in vitro (Figure 2L). It suggests that loss of adipose SCF results in defects in the niche environment but not intrinsically in HSPC. Whether BM adipocyte-derived SCF supports the long-term proliferation and self-renewal of HSPC requires future investigation. Hematopoietic stresses can mobilize HSPC outside the BM to sites like the spleen to expand hematopoiesis.42 We found that spleen in Adipoq-Cre+;Kitlfl/fl mice was slightly heavier (Figure 2M) and splenic cells from AdipoqCre+;Kitlfl/fl mice formed significantly more colony-forming units (CFU) in vitro (Figure 2N), indicating a compensatory induction of splenic hematopoiesis when BM hematopoiesis was defective. Similar phenotypes were observed in female mice. There was a trending decline in BM cellularity in AdipoqCre+;Kitlfl/fl female mice, compared to their control counterparts (Online Supplementary Figure S5A). Both the absolute number and frequency of LSK cells, MP, CMP, MEP, and GMP were declined when SCF was absent (Online Supplementary Figure S5B and C). CLP remained unchanged (Online Supplementary Figure S5B and C). Adipoq-Cre+;Kitlfl/fl female mice also had reduced ratio of MEP to GMP (Online Supplementary Figure S5D). These data demonhaematologica | 2019; 104(9)

MAT mediates metabolic stress in HSPC

Figure 4. High-fat diet (HFD)-stressed hematopoiesis in control and Adipoq-Cre+;Kitlfl/fl male mice. (A and B) Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl male mice at the age of eight weeks were fed with normal chow (NC) or HFD (n=3-4) for another eight weeks. (A) Representative images of femur sections. (B) Quantification of adipocyte numbers in the BM from the growth plate to 4.5 mm away distally. (C-I) 8-week old Kitlfl/fl (n=6 for each diet) and Adipoq-Cre+;Kitlfl/fl (n=8 for NC and n=6 for HFD) male mice were fed with NC or HFD for eight weeks. BM cellularity (C), LSK number (D), GMP number (E), CLP number (F), CMP number (G), MEP number (H), and the MEP/GMP ratio (I) were determined by flow cytometry. (J-N) Complete blood count of NC- and HFD-fed Kitlfl/fl (n=16 and 13, respectively) and Adipoq-Cre+;Kitlfl/fl (n=12 and 6, respectively) male mice showing granulocyte number (J), monocyte number (K), lymphocyte number (L), megakaryocyte-erythrocyte (MkE) to granulocyte-monocyte (GrMo) ratio (M), and the ratio of lymphocyte to all myeloid cells including MkE and GrMo (N). Data are presented as meanÂąStandard Deviation. *P<0.05; **P<0.01; ***P<0.001 by two-way ANOVA followed by multiple comparison using Tukey's correction (B) or Sidak correction (C-N).

strate that adipocyte-derived niche factor SCF is essential for the maintenance of hematopoietic stem and myeloid progenitor cells. We then performed the complete blood count test of the peripheral blood. Levels of red blood count (RBC), hemoglobin concentration (HGB), and hematocrit were all decreased in both male and female Adipoq-Cre+;Kitlfl/fl mice when compared to Kitlfl/fl controls (Table 1). The mean corpuscular volume (MCV) and the count of reticulocytes were both significantly increased in Adipoq-Cre+;Kitlfl/fl mice (Table 1), showing that Adipoq-Cre+;Kitlfl/fl mice developed the typical macrocytic anemia that is observed in animals and patients with loss-of-function mutations in the SCF/KIT pathway.35,36,43 Meanwhile, Adipoq-Cre+;Kitlfl/fl mice also had less platelets compared to control mice (Table 1). Within white blood cells, neutrophils, monocytes, and lymphocytes reduced their number in both sexes, while eosinophils and basophils were significantly haematologica | 2019; 104(9)

declined in male Adipoq-Cre+;Kitlfl/fl mice and showed a trending reduction in female Adipoq-Cre+;Kitlfl/fl mice (Table 1). Taken together, these data from both male and female mice indicate that adipocytes compose a niche that produce SCF to maintain unperturbed hematopoiesis.

Ablation of stem cell factor in Osterix+ cells impairs hematopoiesis Adipoq-Cre-mediated ablation deleted SCF in all adipose tissues. To exclude the potential confounding effect of SCF from peripheral adipose tissues, we then conditionally deleted the Kitl gene in BM stroma cells using the Osterix (Osx1)-Cre, which marks progenitor cells that can be differentiated into MAT.24-26 Loss of SCF in Osx1+ cells resulted in lower BM cellularity (Figure 3A) and drastically reduced populations of LSK, phenotypic LT-HSC, MP, CMP, GMP, MEP, and CLP in Osx1-Cre+;Kitlfl/fl mice (Figure 3B). Such reductions were much larger than those 1737

Z. Zhang et al. observed in Adipoq-Cre+;Kitlfl/fl mice (Figure 2), suggesting that SCF from non-MAT, Osx1-Cre-derived niche cells also contributed to HSPC defects in Osx1-Cre+;Kitlfl/fl mice. Interestingly, heterozygous Osx1-Cre+;Kitlfl/+ mice also had decreased BM cellularity and frequencies of all HSPC (Figure 3A and B), suggesting the haploinsufficiency of SCF in hematopoietic regulation. Similar to findings in Adipoq-Cre+;Kitlfl/fl mice, the MEP/GMP ratio was reduced (Figure 3C) and the CMP/CLP ratio showed a trending decrease (Figure 3D). As a result, Osx1-Cre+;Kitlfl/fl mice suffered macrocytic anemia (Figure 3E). Furthermore, there were fewer platelets, neutrophils, and monocytes in Osx1Cre+;Kitlfl/fl mice (Figure 3E). Counts of eosinophils, basophils, and lymphocytes were similar between control and Osx1-Cre+;Kitlfl/fl mice (Figure 3E). BM cells from Osx1Cre+;Kitlfl/fl KO mice could efficiently form CFU as wildtype controls (Figure 3F). suggesting defects in the niche environment not intrinsically in HSPC as a result of loss of adipose SCF. However, splenomegaly and profoundly increased CFU formation of splenic cells were observed in Osx1-Cre+;Kitlfl/fl mice (Figure 3G and H), demonstrating a shift of hematopoiesis toward the spleen. Taking data from Adipoq-Cre- and Osx1-Cre-mediated knockout mice, we argue that, in the steady-state, SCF from MAT is essential for homeostasis of hematopoietic progenitors.

MAT-derived stem cell factor contributes to stressed hematopoiesis in obesity Obesity is associated with increased MAT mass and altered hematopoietic and immune functions. We then asked whether MAT-derived SCF mediates the effect of high-fat diet (HFD) on hematopoiesis. Male Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl mice gained similar body weight after HFD feeding for eight weeks (Online Supplementary Figure S6A). HFD significantly increased BM adiposity and cellularity, which could be slightly diminished by the loss of adipose SCF (Figure 4A-C). We did not observe any changes in Kitl gene expression or SCF secretion after HFD in the BM (Online Supplementary Figure S6B and C), indicating a potential compensation from non-adipocytes. Future experiments are required to determine SCF expression by different stroma cells during HFD. Flow cytometric assessment of HSPC showed that numbers of LSK, GMP, and CLP progenitors in wild-type male mice were increased after HFD (Figure 4D-F). Such induction was completely blunted in Adipoq-Cre+;Kitlfl/fl males (Figure 4D-F). It is possible that other yet-to-be defined factors mediate the expansion of HSPC in response to HFD, but their complete function requires MAT-secreted SCF. Even though HFD did not affect numbers of CMP or MEP (Figure 4G and H), there was a reduction in the ratio of MEP to GMP (Figure 4I). The MEP/GMP ratio was lower in AdipoqCre+;Kitlfl/fl males and could not be further reduced by HFD (Figure 4I). Consistent with the changes of HSPC in the BM, there was an increase in the number of granulocytes, monocytes, and lymphocytes in the peripheral blood of wild-type males after HFD (Figure 4J-L). The ratio of megakaryocyte (represented by platelet)-erythrocyte (MkE) to granulocyte-monocyte (GrMo) was down-regulated (Figure 4M), while the ratio of lymphocytes to all myeloid cells was up-regulated by HFD (Figure 4N). These data suggest that HFD preferentially promotes the hematopoietic differentiation toward the GrMo and lymphoid lineages, which may contribute to the development of inflammation and insulin resistance in obesity. In the 1738

peripheral blood of Adipoq-Cre+;Kitlfl/fl male mice, however, HFD could not increase the number of granulocytes, monocytes or lymphocytes to the extent observed in Kitlfl/fl males (Figure 4J-L). HFD-induced downregulation of MkE/GrMo ratio and upregulation of lymphoid/myeloid lineage ratio were both ablated in Adipoq-Cre+;Kitlfl/fl male mice (Figure 4M and N). Other peripheral blood parameters including RBC count, HGB, MCV, and platelet count were either not affected by HFD or were similarly regulated between Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl males (Online Supplementary Figure S6D). These results indicate that SCF from MAT is required for the skewed hematopoietic differentiation toward GrMo and lymphoid lineages during HFD. Sexual dimorphism is observed in obesity-associated inflammation and immune dysfunction, which is partially attributable to difference in hematopoiesis.44,45 We then sought to determine the effect of MAT-derived SCF on obesity-stressed hematopoiesis in females. HFD-induced gain in body weight was similar between Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl female mice (Online Supplementary Figure S7A). We found that, compared to males, HFD feeding in wild-type females did not affect BM cellularity (Online Supplementary Figure S7B) but LSK and all myeloid progenitors including CMP, MEP and GMP were all expanded (Online Supplementary Figure S7C-F), without a change in the relative ratio of MEP to GMP (Online Supplementary Figure S7G). The increase in the frequency of LSK, CMP, and MEP by HFD in females was abolished when SCF was deleted in adipocytes (Online Supplementary Figure S7C-E). However, the increase in GMP frequency after HFD in females seemed to be independent of adipocytederived SCF (Online Supplementary Figure S7F). In the peripheral blood, HFD feeding significantly augmented the red blood cell count and hemoglobin concentration in both Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl female mice (Online Supplementary Figure S7H and I). Mean corpuscular volume (MCV), platelet count, and lymphocyte count were not affected by HFD in either Kitlfl/fl or Adipoq-Cre+;Kitlfl/fl female mice (Online Supplementary Figure S7J-L). Despite the increased frequency of GMP in the BM, there was a declining trend in peripheral granulocytes and significant downregulation of monocytes in HFD wild-type females, which was absent in Adipoq-Cre+;Kitlfl/fl female mice (Online Supplementary Figure S7M and N). Reasons causing these discrepancies between the BM and the periphery are unclear, but may involve the production and turnover of mature cells, their release into the circulation, and recruitment to target tissues. Taken all these data from male and female mice, we conclude that MAT niche factor SCF is required for HFD-induced changes in HSPC maintenance and differentiation, despite the sex differences observed in such responses.

Adipocyte-derived stem cell factor partially mediates the β3-adrenergic regulation of hematopoietic stem and progenitor cells The BM is extensively innervated by the SNS.23,46 β-adrenoceptors are expressed in Nestin+ SSC and their activation by the SNS mediates the circadian mobilization of HSC,47,48 while the neuropathy of the BM niche contributes to the pathogenesis of myeloproliferative neoplasms.49,50 In addition, MAT, particularly the rMAT, expresses all three β-adrenoceptors and undergoes remodeling when the sympathetic tone is elevated by cold or haematologica | 2019; 104(9)

MAT mediates metabolic stress in HSPC

Figure 5. Sympathetic nervous system (SNS)-activated hematopoiesis in control and Adipoq-Cre+;Kitlfl/fl mice. (A and B) 12-week old Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl male mice were treated with the saline vehicle or CL 316,243 for one week. (A) Representative images of femur sections. (B) Quantification of adipocyte numbers in the bone marrow (BM) from the growth plate to 4.5 mm away distally. (C) BM cellularity of 12-week old Kitlfl/fl (n=7 for vehicle and n=6 for CL) and Adipoq-Cre+;Kitlfl/fl (n=8 for vehicle and n=6 for CL) male mice were treated with the saline vehicle or CL 316,243 for one week. (D and E) Relative Kitl mRNA levels in the BM (D) or iWAT (E) of wild-type mice treated with vehicle or CL 316,243 (n=6) for one week. (F-M) 12-week old Kitlfl/fl (n=7 for vehicle and n=6 for CL) and Adipoq-Cre+;Kitlfl/fl (n=8 for vehicle and n=6 for CL) male mice were treated with the saline vehicle or CL 316,243 for one week. LSK number (F), MP number (G), CMP number (H), MEP number (I), GMP number (J), CLP number (K), the MEP/GMP ratio (L), and the CLP/CMP ratio (M) were determined by flow cytometry. (N and O) CL 316,243 treatment-induced changes in MEP/GMP ratio (N) and CLP/CMP ratio (O) in Kitlfl/fl (n=6) and Adipoq-Cre+;Kitlfl/fl (n=6) male mice. Data are presented as mean±Standard Deviation. *P<0.05; **P<0.01; ***P<0.001 by two-way ANOVA followed by multiple comparison using Sidak correction (F-M) or unpaired, two-tailed Student t-test (D, N, and O).

β-adrenergic agonists.51,52 We then sought to determine whether MAT-secreted SCF mediates the effect of β3adrenergic signaling on hematopoiesis. Treatment with CL 316,243 to activate the β3-adrenoceptor for seven days did not change body weight, body fat percentage, MAT mass, or BM cellularity in either Kitlfl/fl or Adipoq-Cre+;Kitlfl/fl mice, when compared to saline controls (Figure 5A-C and Online Supplementary Figure S8A and B). Scheller et al. recently showed that MAT, compared to peripheral WAT, relatively resists lipolysis and remodeling in response to CL 316,243.51 We measured Kitl gene expression and found that CL 316,243 strongly induced Kitl mRNA in the BM but not iWAT (Figure 5D and E). Adipoq mRNA in MAT could be also induced by CL 316,243 (Online Supplementary Figure S8C), suggesting that β3-adrenergic activation is able to remodel MAT. CL 316,243 treatment significantly increased the numbers of LSK, MP, MEP, GMP, and CLP in the wild-type bone marrow (Figure 5FK). Interestingly, the induction of LSK and MP, in particular MEP, by CL 316,243 was diminished in AdipoqCre+;Kitlfl/fl mice (Figure 5F, G, and I). The number of CMP was not affected by CL 316,243 in either control or haematologica | 2019; 104(9)

Adipoq-Cre+;Kitlfl/fl mice (Figure 5H). CL 316,243-induced increase in GMP and CLP was not dependent on adipocyte-derived SCF (Figure 5J and K). As a result of the loss of adipose SCF, the reduction in MEP/GMP ratio was further decreased, while the increase in CLP/CMP ratio was further augmented in Adipoq-Cre+;Kitlfl/fl mice upon CL 316,243 treatment (Figure 5L-O). These data indicate that SCF from MAT mediates some aspects of homeostatic responses in the BM upon the β3-adrenergic activation. Nonetheless, we cannot exclude the possibility that the role of CL 316,243 on hematopoiesis is affected by lipolysis in peripheral WAT. Compared to the standard housing temperature (22°C), thermoneutrality (30°C) suppresses the SNS, promotes HSPC apoptosis, and increases the radiosensitivity of mice.53 Consistent to the previous finding,53 we did not observe any difference in BM cellularity (Online Supplementary Figure S9A) or the frequency of LSK, MP, and CLP (Online Supplementary Figure S9B) between mice housed at 22°C and 30°C. However, thermoneutrality increased the ratio of MEP to GMP and the ratio of CLP to CMP (Online Supplementary Figure S9C and D). After 1739

Z. Zhang et al.

Figure 6. The effect of aging on hematopoietic stem and progenitor cells (HSPC) in Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl mice. (A) Body weight of Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl (n=6) male mice at three [n=9 and 10 for control and knock-out (KO), respectively] and ten (n=6 and 5 for control and KO, respectively) months of age. (B-J) Flow cytometric analyses of bone marrow from Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl (n=6) male mice at three (n=9 and 10 for control and KO, respectively) and ten (n=6 and 5 for control and KO, respectively) months of age, showing total cellularity (B), LSK number (C), MP number (D), CMP number (E), MEP number (F), GMP number (G), CLP number (H), MEP/GMP ratio (I), and CLP/CMP ratio (J). Data are presented as mean±Standard Deviation. Two-way ANOVA followed by multiple comparison using Tukey's correction was performed. (Top) P-values for interaction and between groups. *P<0.05; **P<0.01; ***P<0.001 between 3M and 10M mice in indicated genotypes.

being housed at 30°C for one month, Adipoq-Cre+;Kitlfl/fl mice had similar body weight, body composition, and MAT mass to Kitlfl/fl mice (Online Supplementary Figure S9E and F). Loss of adipose SCF eliminated the rise in MEP/GMP ratio (Online Supplementary Figure S9G) but further augmented the increase in CLP/CMP ratio (Online Supplementary Figure S9H). Collectively, these data show that adipocyte-derived SCF mediates part of the environmental effects, particularly those via the β3-adrenergic signaling, on HSPC function.

MAT-provided stem cell factor in the aged hematopoietic stem and progenitor cell compartment Aging-related changes in the hematopoietic system can be attributed to cell-intrinsic and microenvironmental alterations.54 MAT expands as a function of age in both rodents and humans,27 we then sought to determine whether SCF from MAT contributes to altered HSPC function during aging. Young (3 months old) and middle-aged (10 months old) male mice were analyzed; we did not observe any difference in body weight between Kitlfl/fl and Adipoq-Cre+;Kitlfl/fl mice at either age (Figure 6A). Aging slightly decreased BM cellularity in control mice, which was further down-regulated by the loss of adipose SCF (Figure 6B). Strikingly, the expansion of LSK and various myeloid progenitors including CMP, MEP, and GMP observed in middle-aged mice was totally abolished in Adipoq-Cre+;Kitlfl/fl mice (Figure 6C-G). There were more CLP in 10-month old mice than in 3-month old mice, but there was no difference in these changes between genotypes (Figure 6H). Aging increased the ratios of both MEP to GMP and CLP to CMP; however, Kitlfl/fl and AdipoqCre+;Kitlfl/fl mice showed similar extent of increase as the interaction effect was not statistically significant (Figure 6I 1740

and J). These data demonstrate that SCF, as a MAT niche factor, is essential for the phenotypic expansion of HSC and myeloid progenitors during aging.

Discussion Hematopoietic cytokines support the developmental processes of blood cells, at least partially through rewiring the cellular metabolism. Meanwhile, many of these cytokines also act directly on diverse metabolic tissues and cells. For example, erythropoietin improves glycemic control and insulin sensitivity, prevents obesity by acting on the hypothalamus, and attenuates adipose tissue inflammation.55 Granulocyte-macrophage colony-stimulating factor regulates lipid metabolism in the liver.56 Interleukin 4 inhibits adipogenesis, promotes lipolysis, and also disposes glucose by enhancing insulin action.57,58 Interestingly, a specific hemopoietin cocktail composed of SCF, FLT3 ligand, IL-6, and vascular endothelial growth factor together with bone morphogenic protein 7 induces efficient differentiation of human pluripotent stem cells into functional brown adipocytes.37 Our previous work showed that the expression of SCF in BAT is sensitive to food availability and environmental temperature.38 SCF overexpression activates thermogenesis in BAT and reduces weight gain, while Kit mutant mice become obese as a result of reduced energy expenditure.38 Consistently with these previous findings, in this study we found that adipose-derived SCF was required for thermogenic gene expression in cultured adipocytes, demonstrating a cellautonomous effect of SCF. However, no changes in UCP1 expression, body weight, or glucose metabolism were observed in Adipoq-Cre+;Kitlfl/fl mice, suggesting a possible haematologica | 2019; 104(9)

MAT mediates metabolic stress in HSPC

compensation from SCF or other factors secreted by nonadipose cells. Determining additional sources of SCF in regulating systemic metabolism is warranted in future studies. In the BM, SCF is secreted by endothelial cells,22 stromal cells that can be labeled by LEPR,6,59 CXCL12,60 Nestin-GFPlow,60 PDGFRa,62 N-Cadherin,63 and Prx1-Cre,21 and adipocytes.32 Loss of SCF from these niche cells all leads to reduced HSPC numbers.64,65 It is still highly debatable whether osteoblasts express SCF, but HSC frequency and function were not affected by deleting SCF or CXCL12 from Col2.3+ osteoblasts or ablating Ocn+ osteoblasts cells.21,22,66,67 A recent study by Zhou et al. elegantly showed that BM adipocytes proliferate after irradiation or chemotherapy, and deleting SCF using AdipoqCre/ER inhibits hematopoietic regeneration.32 In this study, using Adipoq-Cre- and Osx1-Cre-mediated knockout of SCF, we determined that MAT-secreted SCF is essential for HSC maintenance and hematopoiesis also in the steady-state. Zhou et al. did not observe any deficiency in HSC frequency in non-irradiated AdipoqCre/ER+;KitlGFP/fl mice, probably because they used heterozygous KitlGFP/fl mice as controls, in which the GFP insert disrupted one allele of the Kitl gene. Haploinsufficiency of the SCF/KIT pathway has been well documented,22,36 and we also observed substantial defects in HSPC in heterozygous Osx1-Cre+;Kitlfl/+ mice. The discrepancies between the results of our studies and those of Zhou et al. could also lie in the different Cre lines used. Adipoq-Cre only labels mature adipocytes, but not bone stroma, adipogenic progenitors, hematopoietic cells, bone lining cells, or osteoblast cells.41 However, AdipoqCreER also recombines in a subset of LEPR+ stromal cells in the BM.32 On the other hand, the Adipoq-Cre line has its limitations as the Cre expression is not restricted to adulthood. However, a minimal amount of MAT is present at the age of one week in mice and MAT rapidly expands afterwards.52 We argue that the hematopoietic defects observed in adult Adipoq-Cre+;Kitlfl/fl mice were largely due to SCF excision during adulthood. In the BM of young mice, adipocytes are relatively rare, compared to other cell types that also express SCF. Even though more than half of Kitl mRNA and SCF protein were lost in the BM of Adipoq-Cre+;Kitlfl/fl mice, it is still unclear why such a profound effect could be observed when adipose SCF was absent. Note that we cannot rule out the contributions of peripheral adipocytes and other Osx1+ progeny to the hematopoietic defects observed in Adipoq-Cre+;Kitlfl/fl mice and Osx1-Cre+;Kitlfl/fl mice, respectively. Tools that can selectively target MAT are needed to solve this enigma. There have been a handful of studies that investigated the role of diet and obesity on the composition of the HSPC compartment and blood cell production.9-13,15,16 Despite the inconsistencies seen in these studies, such as the different mouse models, diets, and length of treatment employed, it is generally accepted that diet-induced obesity promotes the immediate expansion of LSK stem cells but push their differentiation skewed toward the myeloid and lymphoid lineages, which will result in long-term defects in hematopoiesis upon stress and infection. Of

haematologica | 2019; 104(9)

note, only male mice were assessed in these studies. Here, we consistently found that HFD activates myelopoiesis and lymphopoiesis in male mice, which may facilitate tissue inflammation in obesity. However, HFD in females did not change the ratio of MEP to GMP in the BM or the MkE/GrMo ratio in the peripheral blood. These discrepancies may help explain the sexual dimorphism in metabolic dysfunction associated with obesity in animals and humans. It has recently been reported that estrogen signaling could control the sexual dimorphism of HSPC development.44,45 Shown in this study, HFD-induced alterations in the frequency of HSPC populations and the composition of blood cells, despite being divergent between males and females, were both dependent on adipose SCF, underscoring the fundamental niche function of MAT. Similar to obesity, aging is also associated with increased HSC numbers but decreased regenerative potential and skewed differentiation toward myeloid cells.54,68,69 Both cell-intrinsic and cell-extrinsic mechanisms are accountable to these aged-related changes. MAT substantially expand in aged humans and rodents, and has been proposed as a suppressor of hematopoiesis in aging.70 However, in this study, by deleting SCF from adipose tissues, we were able to show that adipose-derived SCF is essential for the expansion of LSK stem cells and all myeloid precursor populations in middle-aged mice. It again supports the notion that MAT provides niche factors for the HSPC, particularly the myeloid compartment. Nevertheless, it is unclear whether such requirement of adipose SCF prevents hematopoietic aging or accelerates the exhaustion of HSPC in the bone marrow. Functional characterization of these HSPC in aged Adipoq-Cre+;Kitlfl/fl mice is required in future experiments. In summary, we demonstrate that MAT is a functionally important source of SCF in steady-state hematopoiesis and required for HSPC to cope with metabolic stresses in obesity and aging. Acknowledgments We thank Dr. Alessandro Bartolomucci, Dr. Maria Razzoli, and Dr. Pilar Ariza Guzman at the Integrative Biology and Physiology Core for EchoMRI analyses and animal housing in temperature-controlled rooms. Funding This work was supported by National Key R&D Program of China (2017YFD0500505), Fundamental Research Funds for the Central Universities (KJQN201604), National Natural and Science Foundation of China (31500944), Natural Science Foundation of Jiangsu Province (BK20150687), and China Scholarship Council postdoctoral fellowship (201606855010) to ZH; Natural Science Foundation of Jiangsu Province (BK20170147) to ZZ; National Natural and Science Foundation of China (81770543), American Diabetes Association (18-IBS-167), and NIAID (R01AId139420 and R21AI140109) to H-BR.

1741

Z. Zhang et al.

References 1. Brestoff JR, Artis D. Immune regulation of metabolic homeostasis in health and disease. Cell. 2015;161(1):146-160. 2. de Haan G, Lazare SS. Aging of hematopoietic stem cells. Blood. 2018;131(5):479-487. 3. Kovtonyuk LV, Fritsch K, Feng X, Manz MG, Takizawa H. Inflamm-Aging of Hematopoiesis, Hematopoietic Stem Cells, and the Bone Marrow Microenvironment. Front Immunol. 2016;7:502. 4. Despars G, St-Pierre Y. Bidirectional interactions between bone metabolism and hematopoiesis. Exp Hematol. 2011; 39(8):809-816. 5. Kohli L, Passegue E. Surviving change: the metabolic journey of hematopoietic stem cells. Trends Cell Biol. 2014;24(8):479-487. 6. Karigane D, Takubo K. Metabolic regulation of hematopoietic and leukemic stem/progenitor cells under homeostatic and stress conditions. Int J Hematol. 2017;106(1):18-26. 7. Buck MD, Sowell RT, Kaech SM, Pearce EL. Metabolic Instruction of Immunity. Cell. 2017;169(4):570-586. 8. Ganeshan K, Chawla A. Metabolic regulation of immune responses. Annu Rev Immunol. 2014;32:609-634. 9. Trottier MD, Naaz A, Li Y, Fraker PJ. Enhancement of hematopoiesis and lymphopoiesis in diet-induced obese mice. Proc Natl Acad Sci U S A. 2012; 109(20):7622-7629. 10. Singer K, DelProposto J, Morris DL, et al. Diet-induced obesity promotes myelopoiesis in hematopoietic stem cells. Mol Metab. 2014;3(6):664-675. 11. van den Berg SM, Seijkens TT, Kusters PJ, et al. Diet-induced obesity in mice diminishes hematopoietic stem and progenitor cells in the bone marrow. FASEB J. 2016;30(5):1779-1788. 12. 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. 13. Lee JM, Govindarajah V, Goddard B, et al. Obesity alters the long-term fitness of the hematopoietic stem cell compartment through modulation of Gfi1 expression. J Exp Med. 2018;215(2):627-644. 14. Adler BJ, Kaushansky K, Rubin CT. Obesity-driven disruption of haematopoiesis and the bone marrow niche. Nat Rev Endocrinol. 2014; 10(12):737-748. 15. Claycombe K, King LE, Fraker PJ. A role for leptin in sustaining lymphopoiesis and myelopoiesis. Proc Natl Acad Sci U S A. 2008;105(6):2017-2021. 16. Nagareddy PR, Kraakman M, Masters SL, et al. Adipose tissue macrophages promote myelopoiesis and monocytosis in obesity. Cell Metab. 2014;19(5):821-835. 17. Wei Q, Frenette PS. Niches for Hematopoietic Stem Cells and Their Progeny. Immunity. 2018;48(4):632-648. 18. Calvi LM, Link DC. The hematopoietic stem cell niche in homeostasis and disease. Blood. 2015;126(22):2443-2451. 19. Morrison SJ, Scadden DT. The bone marrow niche for haematopoietic stem cells. Nature. 2014;505(7483):327-334. 20. Asada N, Kunisaki Y, Pierce H, et al. Differential cytokine contributions of perivascular haematopoietic stem cell niches. Nat Cell Biol. 2017;19(3):214-223.

1742

21. Greenbaum A, Hsu YM, Day RB, et al. CXCL12 in early mesenchymal progenitors is required for haematopoietic stem-cell maintenance. Nature. 2013;495(7440):227230. 22. Ding L, Saunders TL, Enikolopov G, Morrison SJ. Endothelial and perivascular cells maintain haematopoietic stem cells. Nature. 2012;481(7382):457-462. 23. Hanoun M, Maryanovich M, Arnal-Estape A, Frenette PS. Neural regulation of hematopoiesis, inflammation, and cancer. Neuron. 2015;86(2):360-373. 24. Mizoguchi T, Pinho S, Ahmed J, et al. Osterix marks distinct waves of primitive and definitive stromal progenitors during bone marrow development. Dev Cell. 2014;29(3):340-349. 25. Liu Y, Strecker S, Wang L, et al. Osterix-cre labeled progenitor cells contribute to the formation and maintenance of the bone marrow stroma. PLoS One. 2013; 8(8):e71318. 26. Chen J, Shi Y, Regan J, Karuppaiah K, Ornitz DM, Long F. Osx-Cre targets multiple cell types besides osteoblast lineage in postnatal mice. PLoS One. 2014; 9(1):e85161. 27. Scheller EL, Cawthorn WP, Burr AA, Horowitz MC, MacDougald OA. Marrow Adipose Tissue: Trimming the Fat. Trends Endocrinol Metab. 2016;27(6):392-403. 28. Suchacki KJ, Cawthorn WP, Rosen CJ. Bone marrow adipose tissue: formation, function and regulation. Curr Opin Pharmacol. 2016;28:50-56. 29. de Paula FJA, Rosen CJ. Structure and Function of Bone Marrow Adipocytes. Compr Physiol. 2017;8(1):315-349. 30. Li Z, Hardij J, Bagchi DP, Scheller EL, MacDougald OA. Development, regulation, metabolism and function of bone marrow adipose tissues. Bone. 2018;110:134140. 31. 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. 32. Zhou BO, Yu H, Yue R, et al. Bone marrow adipocytes promote the regeneration of stem cells and haematopoiesis by secreting SCF. Nat Cell Biol. 2017;19(8):891-903. 33. Boyd AL, Reid JC, Salci KR, et al. Acute myeloid leukaemia disrupts endogenous myelo-erythropoiesis by compromising the adipocyte bone marrow niche. Nat Cell Biol. 2017;19(11):1336-1347. 34. Lennartsson J, Ronnstrand L. Stem cell factor receptor/c-Kit: from basic science to clinical implications. Physiol Rev. 2012;92(4):1619-1649. 35. Huang Z, Ruan HB, Zhang ZD, et al. Mutation in the first Ig-like domain of Kit leads to JAK2 activation and myeloproliferation in mice. Am J Pathol. 2014; 184(1):122-132. 36. Ruan HB, Zhang N, Gao X. Identification of a novel point mutation of mouse protooncogene c-kit through N-ethyl-Nnitrosourea mutagenesis. Genetics. 2005; 169(2):819-831. 37. Nishio M, Yoneshiro T, Nakahara M, et al. Production of functional classical brown adipocytes from human pluripotent stem cells using specific hemopoietin cocktail without gene transfer. Cell Metab. 2012; 16(3):394-406. 38. Huang Z, Ruan HB, Xian L, et al. The stem cell factor/Kit signalling pathway regulates mitochondrial function and energy expen-

diture. Nat Commun. 2014;5:4282. 39. Herault A, Binnewies M, Leong S, et al. Myeloid progenitor cluster formation drives emergency and leukaemic myelopoiesis. Nature. 2017;544(7648):5358. 40. Fan Y, Hanai JI, Le PT, et al. Parathyroid Hormone Directs Bone Marrow Mesenchymal Cell Fate. Cell Metab. 2017;25(3):661-672. 41. Ambrosi TH, Scialdone A, Graja A, et al. Adipocyte Accumulation in the Bone Marrow during Obesity and Aging Impairs Stem Cell-Based Hematopoietic and Bone Regeneration. Cell Stem Cell. 2017;20(6):771-784 e776. 42. Inra CN, Zhou BO, Acar M, et al. A perisinusoidal niche for extramedullary haematopoiesis in the spleen. Nature. 2015;527(7579):466-471. 43. Broudy VC. Stem cell factor and hematopoiesis. Blood. 1997;90(4):13451364. 44. Singer K, Maley N, Mergian T, et al. Differences in Hematopoietic Stem Cells Contribute to Sexually Dimorphic Inflammatory Responses to High Fat Dietinduced Obesity. J Biol Chem. 2015; 290(21):13250-13262. 45. Nakada D, Oguro H, Levi BP, et al. Oestrogen increases haematopoietic stemcell self-renewal in females and during pregnancy. Nature. 2014;505(7484):555558. 46. del Toro R, Mendez-Ferrer S. Autonomic regulation of hematopoiesis and cancer. Haematologica. 2013;98(11):1663-1666. 47. Mendez-Ferrer S, Lucas D, Battista M, Frenette PS. Haematopoietic stem cell release is regulated by circadian oscillations. Nature. 2008;452(7186):442-447. 48. Mendez-Ferrer S, Battista M, Frenette PS. Cooperation of beta(2)- and beta(3)-adrenergic receptors in hematopoietic progenitor cell mobilization. Ann N Y Acad Sci. 2010;1192:139-144. 49. Arranz L, Sanchez-Aguilera A, MartinPerez D, et al. Neuropathy of haematopoietic stem cell niche is essential for myeloproliferative neoplasms. Nature. 2014; 512(7512):78-81. 50. Drexler B, Passweg JR, Tzankov A, et al. The sympathomimetic agonist mirabegron did not lower JAK2-V617F allele burden, but restored nestin-positive cells and reduced reticulin fibrosis in patients with myeloproliferative neoplasms: results of phase 2 study SAKK 33/14. Haematologica. 2019;104(4):710-716. 51. Scheller EL, Khandaker S, Learman BS, et al. Bone marrow adipocytes resist lipolysis and remodeling in response to beta-adrenergic stimulation. Bone. 2019;118:32-41. 52. Scheller EL, Doucette CR, Learman BS, et al. Region-specific variation in the properties of skeletal adipocytes reveals regulated and constitutive marrow adipose tissues. Nat Commun. 2015;6:7808. 53. Povinelli BJ, Kokolus KM, Eng JW, et al. Standard sub-thermoneutral caging temperature influences radiosensitivity of hematopoietic stem and progenitor cells. PLoS One. 2015;10(3):e0120078. 54. Geiger H, de Haan G, Florian MC. The ageing haematopoietic stem cell compartment. Nat Rev Immunol. 2013;13(5):376-389. 55. Wang L, Di L, Noguchi CT. Erythropoietin, a novel versatile player regulating energy metabolism beyond the erythroid system. Int J Biol Sci. 2014;10(8):921-939. 56. Hunt AN, Malur A, Monfort T, et al.

haematologica | 2019; 104(9)

MAT mediates metabolic stress in HSPC

57.

58.

59.

60.

Hepatic Steatosis Accompanies Pulmonary Alveolar Proteinosis. Am J Respir Cell Mol Biol. 2017;57(4):448-458. Tsao CH, Shiau MY, Chuang PH, Chang YH, Hwang J. Interleukin-4 regulates lipid metabolism by inhibiting adipogenesis and promoting lipolysis. J Lipid Res. 2014; 55(3):385-397. Ricardo-Gonzalez RR, Red Eagle A, Odegaard JI, et al. IL-4/STAT6 immune axis regulates peripheral nutrient metabolism and insulin sensitivity. Proc Natl Acad Sci U S A. 2010;107(52):22617-22622. Comazzetto S, Murphy MM, Berto S, Jeffery E, Zhao Z, Morrison SJ. Restricted Hematopoietic Progenitors and Erythropoiesis Require SCF from Leptin Receptor+ Niche Cells in the Bone Marrow. Cell Stem Cell. 2019;24(3):477-486 Omatsu Y, Sugiyama T, Kohara H, et al. The essential functions of adipo-osteogenic progenitors as the hematopoietic stem and

haematologica | 2019; 104(9)

61.

62.

63.

64.

65.

progenitor cell niche. Immunity. 2010; 33(3):387-399. Mendez-Ferrer S, Michurina TV, Ferraro F, et al. Mesenchymal and haematopoietic stem cells form a unique bone marrow niche. Nature. 2010;466(7308):829-834. Zhou BO, Yue R, Murphy MM, Peyer JG, Morrison SJ. Leptin-receptor-expressing mesenchymal stromal cells represent the main source of bone formed by adult bone marrow. Cell Stem Cell. 2014;15(2):154168. Zhao M, Tao F, Venkatraman A, et al. NCadherin-Expressing Bone and Marrow Stromal Progenitor Cells Maintain Reserve Hematopoietic Stem Cells. Cell Rep. 2019;26(3):652-669 e656. Gao X, Xu C, Asada N, Frenette PS. The hematopoietic stem cell niche: from embryo to adult. Development. 2018;145(2):dev139691 Crane GM, Jeffery E, Morrison SJ. Adult

66.

67.

68.

69.

70.

haematopoietic stem cell niches. Nat Rev Immunol. 2017;17(9):573-590. Ding L, Morrison SJ. Haematopoietic stem cells and early lymphoid progenitors occupy distinct bone marrow niches. Nature. 2013;495(7440):231-235. Yu VW, Saez B, Cook C, et al. Specific bone cells produce DLL4 to generate thymusseeding progenitors from bone marrow. J Exp Med. 2015;212(5):759-774. Young K, Borikar S, Bell R, Kuffler L, Philip V, Trowbridge JJ. Progressive alterations in multipotent hematopoietic progenitors underlie lymphoid cell loss in aging. J Exp Med. 2016;213(11):2259-2267. Rossi DJ, Bryder D, Zahn JM, et al. Cell intrinsic alterations underlie hematopoietic stem cell aging. Proc Natl Acad Sci U S A. 2005;102(26):9194-9199. Veldhuis-Vlug AG, Rosen CJ. Clinical implications of bone marrow adiposity. J Intern Med. 2018;283(2):121-139.

1743

ARTICLE Ferrata Storti Foundation

Hematopoiesis

MicroRNA-127-3p controls murine hematopoietic stem cell maintenance by limiting differentiation

Laura Crisafulli,1,2 Sharon Muggeo,1,2 Paolo Uva,3 Yulei Wang,4 Masayuki Iwasaki,5 Silvia Locatelli,6 Achille Anselmo,7 Federico S. Colombo,7 Carmelo Carlo-Stella,6,8 Michael L. Cleary,5 Anna Villa,1,9 Bernhard Gentner9 and Francesca Ficara1,2

Haematologica 2019 Volume 104(9):1744-1755

UOS Milan Unit, Istituto di Ricerca Genetica e Biomedica (IRGB), CNR, Milan, Italy; Humanitas Clinical and Research Center - IRCCS, Rozzano, Italy; 3CRS4, Science and Technology Park Polaris, Pula, Cagliari, Italy; 4Genentech Inc., South San Francisco, CA, USA; 5Department of Pathology, Stanford University School of Medicine, Stanford, CA, USA; 6Department of Oncology and Hematology, Humanitas Clinical and Research Center - IRCCS, Rozzano, Italy; 7Flow Cytometry Core, Humanitas Clinical and Research Center - IRCCS, Rozzano, Italy; 8Humanitas Huniversity, Department of Biomedical Sciences, Pieve Emanuele, Milan, Italy and 9San Raffaele Telethon Institute for Gene Therapy (SR-TIGET), IRCCS San Raffaele Scientific Institute, Milan, Italy 1 2

ABSTRACT

Correspondence: FANCESCA FICARA francesca.ficara@humanitasresearch.it Received: May 24, 2018. Accepted: February 14, 2019. Pre-published: February 21, 2019. doi:10.3324/haematol.2018.198499 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1744 ÂŠ2019 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.

1744

he balance between self-renewal and differentiation is crucial to ensure the homeostasis of the hematopoietic system, and is a hallmark of hematopoietic stem cells. However, the underlying molecular pathways, including the role of micro-RNA, are not completely understood. To assess the contribution of micro-RNA, we performed microRNA profiling of hematopoietic stem cells and their immediate downstream progeny multi-potent progenitors from wild-type control and Pbx1-conditional knockout mice, whose stem cells display a profound selfrenewal defect. Unsupervised hierarchical cluster analysis separated stem cells from multi-potent progenitors, suggesting that micro-RNA might regulate the first transition step in the adult hematopoietic development. Notably, Pbx1-deficient and wild-type cells clustered separately, linking micro-RNAs to self-renewal impairment. Differential expression analysis of micro-RNA in the physiological stem cell-to-multi-potent progenitor transition and in Pbx1-deficient stem cells compared to control stem cells revealed miR-127-3p as the most differentially expressed. Furthermore, miR-127-3p was strongly stem cell-specific, being quickly down-regulated upon differentiation and not re-expressed further downstream in the bone marrow hematopoietic hierarchy. Inhibition of miR-127-3p function in Lineage-negative cells, achieved through a lentiviral-sponge vector, led to severe stem cell depletion, as assessed with serial transplantation assays. miR-127-3p-sponged stem cells displayed accelerated differentiation, which was uncoupled from proliferation, accounting for the observed stem cell reduction. miR-127-3p overexpression in Lineage-negative cells did not alter stem cell pool size, but gave rise to lymphopenia, likely due to lack of miR-127-3p physiological downregulation beyond the stem cell stage. Thus, tight regulation of miR-127-3p is crucial to preserve the selfrenewing stem cell pool and homeostasis of the hematopoietic system.

Introduction Hematopoietic stem cells (HSC) are characterized by their ability to give rise to all blood lineages for the entire lifespan of an individual. In order to preserve this capability throughout life, a reservoir of quiescent stem cells is maintained in the bone marrow (BM) microenvironment.1 This occurs through transcriptional and epigenetic mechanisms that actively repress cell cycling, differentiation, apoptosis, and senescence, or protect from oxidative stress. To generate the progenitors that haematologica | 2019; 104(9)

miR-127 controls HSC maintenance

are constantly required to replenish differentiated blood cells characterized by high turnover or to respond to peripheral injuries such as bleeding, a portion of HSC must awake from their dormant state and re-activate proliferation and differentiation programs.2 Thanks to their ability to correctly balance self-renewal and multi-potent differentiation the HSC pool size is maintained. However, the molecular pathways underlying this regulation are not completely understood, including the role of micro-RNA (miRNA). These are evolutionary conserved small noncoding RNA (ncRNA) that regulate mRNA stability and translation at the post-transcriptional level through nonperfect binding to target sequences.3 To study the potential role for miRNA in HSC selfrenewal, we took advantage of the Pbx1 conditional knockout (Pbx1-cKO) mouse model. Pbx1 is a homeodomain transcription factor that positively regulates HSC quiescence.4 Its absence in post-natal HSC causes an excessive proliferation that ultimately leads to their exhaustion, indicating a profound self-renewal defect, and a premature myeloid differentiation at the expense of lymphoid differentiation.5 Therefore, the study of Pbx1-deficient HSC provides the opportunity to identify miRNA involved in the maintenance of HSC identity. We employed Pbx1-cKO mice (and controls) to perform miRNA profiling of HSC and their immediate downstream progeny multi-potent progenitors (MPP) characterized by the absence of the Flk2 marker on their cell surface (Flk2-MPP, previously known as short-term HSC), representing one of the very first differentiation steps from HSC, with similar multi-potent differentiation capacity but reduced self-renewal. This approach allowed the identification of an HSC-specific miRNA, miR-127-3p. By modulating its activity through lentiviral vectors we demonstrate that miR-127 acts as a novel brake on differentiation that HSC employ to maintain their pool.

Lentiviral constructs Design and production of lentiviral vectors for stable ectopic miRNA-overexpression (OE vector) or functional downregulation (DR vector) were as previously described.9,10 Briefly, both vectors exploit the spleen focus forming virus (SFFV) promoter and couple miR-127-3p up- or downregulation with the expression of a cotranscribed fluorescence reporter protein (mOrange fluorescent protein, OFP, for OE vector and destabilized Green Fluorescence Protein, dGFP, for DR vector respectively). For the OE vector, a 274bp fragment containing murine pre-mir-127 was PCR-amplified and cloned into the XhoI and MluI sites inside the EF1a intron of lentiviral transfer plasmid #1394 (SFFV.EFintron.OFP; described by Lechman et al.11). In the DR or ‘sponge’ vector, four tandem copies of an imperfectly complementary miR-127-3p target sequence were synthesized as described by Gentner et al.10 and cloned into the 3' untranslated region (XbaI-XmaI sites) of the #1031scrT (LV.SFFV.dGFP) lentiviral backbone.11 Third-generation lentiviral vector particles pseudotyped with VSV-G were generated as described.12,13

Ethics The study was approved by the Humanitas Clinical and Research Center - IRCCS Institutional Ethical Committee (prot. n. CE Humanitas, as per Ministerial Decree 127/14 of 8/2/2013).

Statistical analysis Data are represented as mean±Standard Error (SE) when n>3. The significance of differences was determined by two-tailed Mann-Whitney unpaired test unless otherwise stated. P<0.05 was considered statistically significant (ns: not significant; * P<0.05; ** P<0.005; ***P<0.0005; ****P<0.0001). Statistical analyses were performed with GraphPad Prism (GraphPad Software). Additional methods are presented in the Online Supplementary Methods.

Results Methods

Pbx1-cKO mouse as a model to identify candidate miRNA regulating hematopoietic stem cell self-renewal

Mice

Hematopoietic stem cells and Flk2–MPP were prospectively isolated from the BM of five Pbx1-cKO (Mx1Cre+.Pbx1f/f) and four control (Mx1Cre-.Pbx1f/f) individual mice. The expression level of 376 different miRNA was obtained using a multiplexed Taq-Man-based realtime stem-loop PCR method14 and subjected to normalization, filtering and analysis (Figure 1A). A global analysis through unsupervised hierarchical clustering indicated a clear distinction between HSC and Flk2-MPP at the level of miRNA expression (Figure 1B), suggesting that miRNA regulate the first transition step in adult hematopoietic development. Within each group, Pbx1-deficient and control cells clustered separately, linking miRNA to selfrenewal impairment. Differential expression analysis indicated that 71 miRNA are differentially expressed (DE) during the physiological HSC-to-MPP transition (Figure 1C, right, and Online Supplementary Table S1). A similar analysis on Pbx1-deficient HSC revealed 48 miRNA DE compared to control HSC (Figure 1C, left), half of which (n=23) are concordantly DE in the HSC-to-MPP list (Online Supplementary Table S2), in accordance with the hypothesis that miRNA are involved in HSC self-renewal. This result is similar to that obtained by analyzing mRNA, with Pbx1-deficient HSC exhibiting a transcriptional profile typical of their immediate downstream progeny.4

Tie2Cre+.Pbx1-/f and Mx1Cre+.Pbx1-/f mice have been described.4,6 Briefly, Tie2Cre+.Pbx1+/- and Mx1Cre+.Pbx1+/f mice were bred with Pbx1f/f mice to obtain Tie2Cre+.Pbx1-/f or MxCre+.Pbx1f/f Pbx1-cKO and their littermate controls.

miRNA profile and bioinformatic analysis RNA was extracted with MirVana isolation kit (Ambion, ThermoFisher Scientific) from 1.4-2x103 HSC and 2-6x103 Flk2MPPs (Table 1) sorted from the BM of Polyinosinic-polycytidylic acid [poly(I:C)] (InvivoGen)-treated Mx1Cre+Pbx1f/f and Mx1Cre-Pbx1f/f control mice (four experimental groups, 4-5 samples from individual mice/group). miRNA profiling was performed using the Megaplex TaqMan Assay system coupled with PreAmplification step (Rodent Pool A, Applied Biosystems, ThermoFisher Scientific). The expression level of 376 miRNA plus eight controls was obtained. Non-expressed miRNA (Ct level ≥35) were filtered out, whereas miRNA expressed in at least one of the four groups were further analyzed. Several normalization strategies were applied (quantile, median and endogenous normalization strategy) and quantile normalization was primarily chosen due to its smallest co-efficient of variation among replicates (cv=sdev/mean).7 Differential expression (DE) analysis was performed by Statistical Analysis of Microarray8 (SAM) (FDR with q-value <0.1). haematologica | 2019; 104(9)

1745

L. Crisafulli et al. Table 1. Analyzed and/or sorted cell populations.

Stem and progenitor cells

Lineage restricted progenitors

Myeloid precursors and mature cells

Lymphoid precursors and mature cells

Cell population

Acronyms or abbreviation

Phenotype

Hematopoietic Stem and Progenitor Cells Hematopoietic Stem Cells

LKS

Lin–/c-Kithi/Sca1hi

HSC

Multi-Potent Progenitors (at different stages of maturation)

MPP

Common Lymphoid Progenitors Common Myeloid Progenitors Granulocyte/Macrophage Progenitors Megakaryocyte/Erythrocyte Progenitors Megakaryocyte and Erythrocyte Precursors

CLP CMP GMP MEP Pre-MegE

LKS/CD34–/Flk2– for miRNA profiling; LKS/CD48–/CD150+/Flk2– for analysis of transplanted mice; Sca1hi/CD48– CD150+-/EPCR+ for in vitro differentiation experiments LKS/CD34+/Flk2- for miRNA profiling; LKS/CD127-/CD34+/Flk2int or Flk2high for all other analysis Lin–/CD127+/Flk2high/c-Kitint/Sca1int Lin–/c-Kit+/Sca1-/CD34+/CD16/32Lin–/c-Kit+/Sca1-/CD34+//CD16/32high Lin–/c-Kit+/Sca1-/CD34–/CD16/32– Lin–/c-Kit+/Sca1-/CD41–/CD16/32–/CD105–/CD150+

Erythrocyte Precursors (at different stages of maturation) Poly-Morpho-Nucleated Leukocytes Monoblasts Pro-Monocyte Monocyte Pro-B Pre-B Immature B cells Mature B cells Natural Killer T cells

Pre-E MkP From R1 stage to R41 PMN

ProB PreB B imm B mat NK

Lin–/c-Kit+/Sca1–/CD41–/CD16/32–/CD105+/ Lin–/c-Kit+/Sca1–/CD150+/CD41+/ Ter119–/CD44high (R1); Ter119+/CD44high or CD44int or CD44– (R2 to R4) CD11b+/Ly6Ghigh CD11b–/CD31+/Ly6C– CD11b–/CD31+/Ly6C+ CD11b+/CD31–/Ly6C+ NK1.1–/CD3–/CD43+/B220+ NK1.1–/CD3–/CD43–/B220+/IgM– NK1.1–/CD3–/CD43–/B220+/IgM+ NK1.1–/CD3–/CD43–/B220high /IgM+ NK1.1+/CD3– CD3+/ NK1.1–

Modified from Liu et al..46 Monoclonal antibodies are listed in Online Supplementary Table S4.

Importantly, the expression of most of the DE miRNA changed more than 2-fold. Of note, the most up-regulated miRNA in the normal HSC-to-MPP transition is miR-221 (Online Supplementary Figure S1). This represents a positive control of our analysis, since miR-221 has been predicted to be an important regulator of HSC maturation due to its ability to repress cKit.15 We compared the lists of DE miRNA with those of DE transcripts obtained by microarray.4 None of the DE miRNA between mutant HSC or Flk2–MPP and corresponding controls are located within a transcript detected in the microarray, indicating that their different level of expression is not the result of the level of expression of a host gene. Among the 23 over-lapping miRNA (Figure 1C), we applied stringent criteria to select a few candidates for subsequent studies. We considered only miRNA that: 1) were DE also by applying a different normalization method; 2) had a fold difference higher than five; 3) had anti-correlated predicted targets (PT) among previously described DE mRNA in Pbx1-deficient HSC;4 4) were DE also with independent real-time PCR in wild-type (wt) samples. Only three miRNA candidates fulfilled each 1746

condition, all evolutionary conserved and all down-regulated (miR-127-3p, miR-411-5p, miR-34b-3p). Among these, the most DE both in Pbx1-deficient HSC and in the normal HSC-to-MPP transition is miR-127-3p (Online Supplementary Table S3). This miRNA was therefore chosen as a candidate HSC– self-renewal regulator.

miR-127-3p is an HSC-specific miRNA As further proof that miR-127-3p downregulation correlates with loss of self-renewal, we confirmed its DE in phenotypically-defined HSC from a different previously described Pbx1-cKO mouse model (Tie2Cre+Pbx1-/f)4,5 (Figure 2A). We then prospectively isolated HSC and Flk2-MPP from wt mice and compared the expression level of miR127-3p to that of other miRNA previously associated to HSC, such as miR-99b, miR-125, let-7, miR-221 and miR126.11,15-19 We found that miR-127-3p expression in HSC is not particularly high, being similar to that of other miRNA already reported to play a role in HSC biology (Figure 2B). We confirmed the recently reported downregulation of miR-99b in Flk2–MPP20 (Figure 2C). haematologica | 2019; 104(9)

miR-127 controls HSC maintenance

Figure 1. Strategy for the selection of candidate miRNA regulating hematopoietic stem cell (HSC) self-renewal. (A) Experimental workflow. (B) Heat map shows the unsupervised hierarchical clustering of relative miRNA expression in HSC and multi-potent progenitors (MPP) characterized by the absence of the Flk2 marker on their cell surface (Flk2â&#x20AC;&#x201C;MPP). The color-scale represents Z-score transformed signal intensity. (C) Hierarchical cluster dendrogram is shown for relative expression of differential expression (DE) miRNA in mutant HSC compared to wild-type (wt) HSC (left) and in the normal HSC-to-MPP transition (right). Venn diagram shows the overlap of deregulated miRNA in the two analyses. The significance of overlap was computed by hypergeometric test (P-value 5x10-5). BM: bone marrow.

However, miR-127-3p showed the highest downregulation (>100-fold; miR-99b 3-fold) in the first step of hematopoietic differentiation, whereas the majority of the other miRNA are still expressed at the progenitor level and have only a modest reduction compared to HSC. We then used real-time PCR to investigate whether miR-127-3p is expressed in other hematopoietic cell subpopulations other than HSC within the BM in steadystate conditions. This issue was addressed by analyzing miR-127-3p expression from purified populations including different lineage committed progenitors, megakaryocyte and erythrocyte precursors, myeloid precursors and mature cells, and various lymphoid subsets (Table 1 and Figure 2D), since expression in only one subgroup would not be detectable in pooled populations. Impressively, miR-127-3p was highly HSC-specific, being quickly down-regulated in the earliest step of hematopoietic differentiation, where loss of self-renewal occurs. Furthermore, we confirmed its expression in human HSC-enriched CD34+ cells from different sources (Figure 2E). haematologica | 2019; 104(9)

Inhibition of miR-127-3p function severely impairs HSC self-renewal To investigate the role of miR-127-3p in HSC we transplanted lineage-negative (Lin-) cells from CD45.2+ wt mice into lethally irradiated syngeneic CD45.1+ recipients, after permanently inhibiting binding to its targets through a lentiviral sponge vector carrying the reporter protein dGFP, as previously described for other miRNA.11 Since miR-127-3p targets in HSC are not known, we assessed effective downregulation of miR-127-3p activity on K562 cells. Upregulation of XBP-1 and BLIMP-1, two miR-1273p targets,21 indicated that the sponge vector worked as expected (Online Supplementary Figure S2). Engraftment and multipotent reconstitution by Lintransduced cells was then monitored through periodic blood sampling of transplanted mice and at necropsy several weeks after transplant. The BM of some of the reconstituted mice was transplanted into secondary recipients to assess if, in the absence of miR-127-3p activity, HSC maintain their self-renewal ability (Figure 3A). In primary recipients, survival curves were similar for all 1747

L. Crisafulli et al.

Figure 2. Expression of miR-127-3p. (A) Histogram shows fold difference (FD) of miR-127-3p expression in hematopoietic stem cell (HSC) from 3-4-week old Tie2Cre+Pbx1-/f mice relative to HSC from Tie2Cre-Pbx1+/f littermate controls, as measured by quantitative real-time polymerase chain reaction (qRT-PCR). Due to the extreme paucity of stem cells from the Tie2Cre+Pbx1-/f model,4 HSC from nine mice were grouped in two pools to perform qRT-PCR; bars indicate the range. (B) Expression of the indicated miRNA in HSC from 4-12-week old wild-type (wt) mice as measured by qRT-PCR and expressed in arbitrary units (AU). (C) Expression of the indicated miRNA in multi-potent progenitors (MPP) from wt mice relative to their expression level in HSC. (B and C) miR127-3p and let-7e, n=4; miR-99b, n=8; miR-125a and miR-125b, n=1; miR-221, n=6; miR-126a, n=2; n: number of pools (2-5 mice/pool). For miR-126, bars indicate the range. (D) miR-127-3p expression level in steady state bone marrow cell subpopulations from two pools of wt mice (2-3 mice/pool). (E) qRT-PCR analysis of miR-127 expression in human hematopoietic mature CD34â&#x20AC;&#x201C; cells (or in CD34+CD38+ progenitors) relative to the CD34+ mobilized peripheral blood (MPB) or CD34+CD38- cord blood (CB) compartments. N=2 healthy donors for each source, bars indicate the range. All donors signed informed consent. When miRNA expression is indicated as FD, black bars indicate the sample relative to which FD is calculated.

transplanted mice (Online Supplementary Figure S2B). Similarly, the level of engraftment by sponge-transduced HSC (named 127DR for miR-127-3p functional downregulation), measured as the proportion of cells expressing the donor marker CD45.2 over time, was comparable to that of HSC transduced with a control vector carrying only dGFP (named EV for empty vector) (Figure 3B) or with untransduced cells (data not shown), with the exception of one mouse out of eleven. Sponge and control vectors express dGFP under the SFFV promoter, which is highly active in HSC, progenitors and myeloid cells, particularly monocytes. Due to its rapid turnover, dGFP is only detectable when expressed at very high levels. Therefore, even though the SFFV promoter has good and modest activity in B and T lymphocytes, respectively, the dGFP marker is hardly detectable in these lineages.22 For this reason, we followed by FACS analysis dGFP expression in monocytes; this also served as an indirect measure of dGFP expression in HSC since, due to their high turnover, myeloid cells are continuously generated from 1748

HSC. Virtually all monocytes were GFP+, indicating a very efficient transduction of long-term reconstituting cells (Figure 3C). Importantly, GFP-negative or low B and T lymphocytes isolated from the spleen of transplanted mice contained vector sequences (Online Supplementary Figure S2C), indicating that they derived from transduced cells. We therefore safely monitored the presence of transduced cells in vivo through the donor marker CD45.2, since we could not rely on dGFP expression in all cells. The kinetics of peripheral myeloid, B- and T-cell reconstitution from donor cells was similar in mice transplanted with 127DR- or EV-transduced HSC, as well as the proportion and number of the different cell types (Figure 3D), suggesting that suppression of miR-127-3p activity did not affect multi-lineage reconstitution. However, FACS analysis of the BM at necropsy revealed a significant depletion of donor-derived phenotypically defined HSC, which were GFP+ (Figure 3E), without affecting their output of multi-potent progenitors and their lymphoid and myeloid differentiation potential, as confirmed on BM and spleen haematologica | 2019; 104(9)

miR-127 controls HSC maintenance

Figure 3. Inhibition of miR-127-3p function leads to hematopoietic stem cell (HSC) depletion. (A) Experimental workflow. (B) Level of engraftment of transduced cells over time, measured as percentage of cells expressing the donor marker CD45.2 within total peripheral blood (PB) CD45+ cells. Graphs show engraftment in individual mice transplanted with Lin- cells transduced with 127DR (top, n=11) or empty vector (EV) (bottom, n=9). Tx: transplant. (C) dGFP expression in PB donor monocytes measured by FACS analysis 18 weeks after transplant. (Left) Average of eight mice; (right) representative FACS histogram. (D) Kinetics of multi-lineage differentiation of transduced cells during PB sampling (6-12-18 weeks). Histograms show the absolute cell number, determined by hemocytometer analysis, and stacked columns represent the relative abundance of each population (CD11b+ myeloid cells, CD19+ B cells, CD3+ T cells; CD3R: T cells from recipient), determined by FACS analysis (n=6-8). (E) FACS analysis of 127DR and EV stem and progenitor cells in the bone marrow (BM) of transplanted mice. (Left) Representative FACS analysis of HSC and multi-potent progenitors (MPP) gated on LKS (dot plots) and of their dGFP expression (bottom histograms) compared to that of a representative non-transplanted mouse. (Right) Average of HSC and MPP absolute numbers (n=6-7). (F) Central and peripheral multi-lineage differentiation by transplanted 127DRand EV-transduced cells in primary recipients. Histograms show absolute BM cell numbers (left) and percentage of donor cells in the spleen (right). Stacked columns represent relative abundance of each lineage (n=6-8). All graphs summarize results from two independent experiments, and BM cells counts are relative to one femur and one tibia.

(Figure 3F). None of the transplanted animals displayed extramedullary hematopoiesis (data not shown), indicating that the observed HSC reduction was not due to abnormal egress from the BM. To evaluate whether there was a decrease in functional HSC independently of the immunophenotype used to define them, secondary transplants were performed by injecting a high dose of total BM cells from two individual primary recipients in lethally irradiated CD45.1+ secondary recipient mice. The number of transplanted donor BM cells was sufficient for allowing recovery after irradiation; haematologica | 2019; 104(9)

however, most of the mice transplanted with cells derived from 127DR BM succumbed 4-6 weeks after transplant, soon after the challenge of the first blood withdrawal, suggesting a severe HSC defect (Figure 4A). Four weeks after transplant, donor myeloid chimerism was very high in all mice, without significant differences among the groups of mice transplanted with 127DR or EV transduced BM (Figure 4B), indicating that the early mortality of 127DR BM-transplanted mice was likely not due to homing or engraftment impairment. However, sponging miR127-3p severely compromised myeloid cell and platelet produc1749

L. Crisafulli et al. A

Figure 4. Inhibition of miR-127-3p function severely impairs hematopoietic stem cell (HSC) self-renewal. (A) Kaplan-Mayer survival curves of secondary recipients (n=10-11 from two independent experiments; Log-rank test) of 127DR- and empty vector (EV)-transduced cells. The arrow below the horizontal axis indicates the time of first peripheral blood (PB) withdrawal (4 weeks after transplant). (B-D) PB analysis four weeks after transplant (n=10-11 from two independent experiments). (B) FACS analysis for myeloid chimerism, measured as percentage of donor CD45.2+ cells within CD11b+ cells. (C) Percentage of myeloid and lymphoid cells within CD45+ donor gate determined by FACS analysis. (D) Platelet counts determined by hemocytometer analysis. (E) FACS analysis of 127DR stem and progenitor cells in one of the two secondary recipients that survived throughout the experiment. Dot plots are gated on LKS population and a representative sample is shown for the EV group. Histogram overlay (right) shows dGFP expression within the HSC (non-transplanted and EV) or the LKS (127DR) gate.

tion in secondary recipients (Figure 4C and D), in accordance with a stem cell defect. Two out of eleven mice transplanted with 127DR cells survived throughout the experiment, with very different levels of donor chimerism (Online Supplementary Figure S3A). Several weeks after transplant, we analyzed the most primitive hematopoietic compartment within the BM of the two secondary sponge recipients that were still alive. This analysis revealed that HSC were either absent (Figure 4E) or belonged to the recipient (Online Supplementary Figure S3B). LKS multipotent progenitors were donor-derived GFP+, indicating that HSC had originally engrafted at the time of the secondary transplant and were able to generate differentiated progeny throughout time; however, they failed to self-renew. Therefore, miR127-3p activity is crucial for the maintenance of phenotypically and functionally defined HSC.

miR-127-3p prevents premature differentiation To discern the biological mechanism at the basis of HSC loss, we tested if miR-127-3p is involved in regulating HSC quiescence, metabolic properties, survival, or prevention of premature differentiation, all necessary to preserve self-renewal. All these features were initially analyzed on Lin– cells isolated ex vivo from transplanted mice, as opposed to newly transduced cells, in order to avoid potential influences of cell culture. The proportion of cycling and quiescent HSC was similar in 127DR- and EV-transduced BM (Figure 5A), as well as the proportion of apoptotic cells (Figure 5B), indicating that miR127-3p downregulation does not alter HSC cell cycle or apoptosis. 1750

Since oxidative stress has been described to negatively affect HSC function, we measured reactive oxygen species (ROS) production in response to cytokine stimulation, as described.23 ROS production induced upon short in vitro cytokine exposure was similar in HSC within Lin– cells isolated from 127DR or EV recipients (Figure 5C), suggesting that miR-127-3p downregulation does not lead to an increase in oxidative stress. However, after only one day of culture, a reduced expression level (measured by MFI) of the stem-cell associated cKit marker was observed in 127DR cells (Figure 5D), suggesting that miR-127-3p downregulation leads to accelerated differentiation. Lin– cells isolated from 127DR or EV recipients were also subjected to a standard colony assay (Figure 5E). 127DR stem and progenitor cells generated a significantly lower proportion of immature colonies compared to mature ones, in accordance with the hypothesis that they are more prone to differentiate. We confirmed this phenotype using freshly isolated HSC cultured in vitro. HSC were sorted from wt mice using the EPCR marker,24,25 transduced and analyzed at different time points. Single-cell colony assay performed early after transduction revealed that, despite the fact that the total number of colonies was comparable in the 127DR and EV groups, sponged cells generated fewer CFU-GEMM colonies, which originate from the most immature cells (Figure 5F). FACS analysis did not show any differences in the two groups when the conventional CD48, CD150 staining was performed (Online Supplementary Figure S4A); however, at day 6, the proportion of HSC expressing EPCR was reduced (Figure 5G and Online Supplementary Figure S4B). Moreover, a higher proportion of cells displayed early signs of differenhaematologica | 2019; 104(9)

miR-127 controls HSC maintenance

tiation in cytospin preparation (Figure 5H). The proportion of cKit+ within dGFP+ cells was very high in both experimental groups at day 6. However, at day 9, it dropped significantly in cultures derived from sponged HSC, while this decrease was less dramatic in the EV group. Accordingly, cKit expression level was significantly

lower at day 9 in the 127DR group (Figure 5I). Finally, the expression of genes with myeloid expression pattern (Gene Expression Commons26) was higher in sponged cells from day 9 cultures compared to cells transduced with EV (Figure 5J). Overall, these data suggest that a faster differentiation

Figure 5. miR-127-3p prevents premature differentiation. (A) Cell cycle status of 127DR- and empty vector (EV)-transduced hematopoietic stem cell (HSC) was determined by FACS analysis with Ki67/Sytox staining on the bone marrow (BM) of primary recipients (n= 4 and 5, respectively). (B-D) FACS analysis of 127DR- and EVtransduced Lin– cells isolated ex vivo from the BM of primary recipients (n= 5 and 6). (B) Analysis of early apoptosis (AnnexinV+PI–) in 127DR- and EV-transduced HSC identified by FACS within the Lin– cell population. (C) Reactive oxygen species (ROS) production in 127DR- and EV-transduced HSC in response to short (24 hours) cytokine stimulation in culture with StemSpan medium supplemented with SCF, TPO, Flt-3 Ligand and IL-3, measured by FACS after incubation with cellROX reagent. Vertical axis shows the ratio of the mean fluorescence intensity (MFI) at day (d) 1 versus mean fluorescence intensity (MFI) immediately after Lin– isolation (d0). (D) Early differentiation of ex vivo isolated 127DR and EV-transduced Lin– cells measured as reduction of cKit expression after one day of culture as in (C) (MFI ratio, d0/d1). (E) Colony forming cell (CFC) assay performed on ex vivo isolated 127DR- and EV-transduced Lin– cells (n=6). Colonies were scored as immature (CFU-GM and CFU-GEMM) and mature (CFU-M and CFU-G) at d9. Vertical axis shows the ratio between immature and mature colonies. (F) CFC assay performed on wild-type (wt) HSC two days after lentiviral transduction (d3 of culture, n=4 individual mice). (Left) Total number of colonies; (right) number of each different type of colonies. (G) FACS analysis for endothelial protein C receptor (EPCR) expression on GFP+cKit+CD48–CD150+ cells (n=4). (H) Representative cytospin preparations stained with May-Grunwald and Giemsa of 127DR- and EV-transduced HSC at d6 of culture. (I) FACS analysis for cKit expression at d6 and d9 (n=4). (J) Quantitative real-time polymerase chain reaction (qRT-PCR) analysis of the expression of the indicated genes in transduced HSC at d9 of culture relative to d9 non-transduced cells (n=3).

haematologica | 2019; 104(9)

1751

L. Crisafulli et al.

Figure 6. Forced miR-127 expression does not affect hematopoietic stem cell (HSC) maintenance. (A and B) FACS analysis of multi-lineage differentiation in peripheral blood (PB) 18 weeks after transplant (A) and in bone marrow (BM) (B) of 127OE and empty vector (EV)-recipient mice. Histograms show absolute cell numbers and stacked columns show the relative abundance of each lineage determined by FACS analysis. (A) PB absolute cell number derives from hemocytometer analysis (127OE: n=6; EV: n=9; two independent experiments). (B) BM counts are relative to one femur and one tibia (127OE: n=5; EV: n=7; two independent experiments). (C) FACS analysis of 127OE and EV stem and progenitor cells in the BM of primary recipient mice. Cell numbers are relative to one femur and one tibia. (D-G) FACS analysis of 127OE and EV Lin– transduced cells transplanted in competition with non-transduced cells. Analyses are relative to primary (D and E) and secondary (F and G) recipient mice. (D) Percentage of OFP+ cells within donor CD11b+ cells in PB 18 weeks after transplant of transduced Lin–cells injected in competition with non-transduced cells in a 9:1 ratio (P=0.24). (E) Frequency of OFP+ cells within the indicated BM populations of primary recipients. (F) Frequency of OFP+ cells within the indicated BM populations of secondary recipients. (G) Percentage of OFP+ cells within donor CD11b+ and CD19+ BM cells of secondary recipients (****P<0.0001; Wilcoxon paired test).

to the next maturation step is the underlying cause of the observed HSC loss in vivo when miR-127-3p is not active.

Forced miR-127-3p expression does not affect hematopoietic stem cell maintenance We next asked whether increased miR-127-3p levels affect HSC function in vivo. Lentiviral-mediated miR-127 overexpression was performed in wt CD45.2+ Lin– cells, followed by transplantation in CD45.1+ recipient mice (Online Supplementary Figure S5A). miR-127-overexpression (127OE) vector carried the Orange Fluorescent Protein (OFP) as reporter gene and the corresponding empty vector (EV), carrying only the reporter protein, was used as control. We obtained very high transduction efficiency with both vectors (data not shown). Expression analysis through real-time PCR before and after transduction confirmed that the progeny of infected cells expressed 1752

miR-127-3p, reaching levels similar to those of miR-16, which is present in all hematopoietic cells (Online Supplementary Figure S5B). Periodic PB analysis revealed that 127OE Lin- cells were able to engraft and undergo multi-lineage differentiation, similar to Lin– cells transduced with the EV (data not shown). However, several weeks after transplant, peripheral B- and T-cell counts were significantly lower in mice transplanted with miR127 over-expressing cells, compared to mice transplanted with the EV (Figure 6A). These differences were present also in BM (Figure 6B) and spleen (data not shown), and are consistent with the reported suppression of germinal center regulators by human miR-127,21,27 likely reflecting the lack of miR-127 downregulation from the MPP stage onward rather than upregulation in HSC. In mice transplanted with 127OE cells, the level of miR-127-3p expression in the spleen several weeks after transplantation was haematologica | 2019; 104(9)

miR-127 controls HSC maintenance

similar to that of cells recently infected, and inferior to that of miR-16 (Online Supplementary Figure S5C), thus excluding effects due to supra-physiological expression. The observed lymphopenia was not accompanied by myeloproliferation or splenomegaly, as expected if 127OE were to lead to HSC expansion. Indeed, 127OE did not affect maintenance of stem and progenitor cells (Figure 6C). This is similar to what has been described for miR-99, which did not cause HSC defects when over-expressed, despite its function in regulating HSC self-renewal being revealed after downregulation.20 To test whether miR-127OE HSC show an advantage when subjected to challenging conditions, transduced Lincells were transplanted in competition with un-transduced cells at different ratios, and the presence of OFP+ donor cells was monitored over time on myeloid cells in the PB, as described for the sponge experiments, as well as in BM and spleen at necropsy (Online Supplementary Figure S5D). 127OE OFP+ cells did not expand in primary recipients (Figure 6D and E). Donor-derived OFP+ stem cells were functionally competent, since they engrafted, expanded and gave rise to multi-lineage progeny in secondary recipients (Figure 6F and G), independently from the overexpression of miR-127. However, lower lymphoid reconstitution was detected in secondary recipients (Figure 6G). Taken together, these results suggest that miR-127-3p maintains the HSC pool by limiting premature differentiation, and must be down-regulated upon HSC maturation to maintain correct homeostasis of the hematopoietic system.

Discussion In this report we demonstrate that miR-127-3p is an important novel regulator of the balance between selfrenewal and differentiation in HSC. The miR-127 gene is encoded in a large evolutionary conserved miRNA gene cluster located within an imprinted domain (14q32 in human, 12qF in mouse),28 the Dlk1Gtl2 locus. On the maternally imprinted allele, this locus encodes multiple long ncRNA, snoRNA and a mega-cluster of miRNA organized in two groups, among which the miR-127/miR-136 cluster composed of 7 miRNA genes. RNA-seq data showed that the entire locus, which is regulated by a common cis-element, is specifically transcribed in CD49blo long-term HSC compared to all other BM progenitor and mature cells.28 However, according to our profiling, of the miR-127/miR-136 cluster, only miR127 and miR-434 are expressed as mature forms in adult HSC. Similarly, of all the other miRNA included in the Dlk1-Gtl2 locus beside the miR-127/miR-136 cluster, only miR-411 is expressed in HSC. Of note, three out of the four top miRNA in our HSC-to-MPP list are part of this locus (Online Supplementary Table S2), in line with previous data showing that ncRNA within the Dlk1-Gtl2 locus can be considered HSC markers.28 The miR-127 gene is embedded in a CpG island; therefore, it is conceivable that the rapid downregulation in the earliest step of hematopoietic differentiation is due to epigenetic silencing. Indeed, the expression of miR-127 was experimentally induced in cancer cell lines by chromatin-modifying drugs, which did not affect expression of the other members of the miRNA gene cluster.29 The entire locus has been inactivated in the maternallyhaematologica | 2019; 104(9)

derived allele.30,31 Due to embryonic lethality, the consequences on hematopoiesis were studied using fetal liver stem and progenitor cells. The phenotype observed presented features similar to what we observed in mice transplanted with 127DR Linâ&#x20AC;&#x201C; cells, such as reduction of HSC despite normal progenitor and lineage cell counts, and defective long-term reconstitution capacity.28 Altered PI3KmTOR pathway affecting mitochondrial metabolism explained this phenotype,28 although other molecular mechanisms cannot be excluded. Our data on ROS production and apoptosis do not support the hypothesis that miR-127-3p is involved in regulating oxidative stress in HSC, and indeed miR-127-3p was not one of the miRNA for which a target within the PI3K-mTOR pathway was confirmed.28 A mouse model with specific deletion of miR-127 was generated;32 however, the effects on hematopoiesis have not been investigated. MiR-127 is aberrantly expressed in the context of several solid tumors, potentially acting as either an oncogene (oncomiR)33-35 or as a tumor suppressor.36-41 Within hematopoietic malignancies, miR-127 was found up-regulated, together with other miRNA included within the 14q32 domain, in the acute promyelocytic leukemia (APML) due to PML-RARa translocation.42 It would be worth investigating whether miR-127 aberrant expression contributes to the differentiation block in APML myeloid progenitors. Although miR-127-3p overexpression did not lead to myelo- or lymphoproliferation, we cannot exclude that, in the presence of other genetic abnormalities, miR127-3p may act as an oncomiR. A number of miR-127-3p targets have been found in cancer cells,21,33,35-39,41 in other cell lines,43,44 or during development,32 but they are either not expressed in hematopoietic progenitors, or their expression does not change in the HSC-to-MPP transition. Our search for predicted miR127-3p targets within DE mRNA in the HSC-to-MPP transition and in Pbx1-deficient HSC revealed two potential targets, Gp1bb and Nek2 (Online Supplementary Table S3). We confirmed by qRT-PCR their upregulation in the physiological HSC-to-MPP transition (data not shown); however, the increase in Gp1bb expression was extremely mild, whereas Nek2 seed-sequence is located within an alternatively spliced, non-coding portion of the transcript. It is, therefore, unlikely that Nek2 protein expression is directly regulated by miR-127-3p. Since miR-127-3p binding to its target mRNA might result in translation inhibition rather than mRNA degradation, a global proteomic approach would be required to gain insights on miR-127-3p functional targets mediating the observed phenotype, although protein analysis on rare primary HSC is technically challenging, especially in a model in which stem cells are reduced. The physiological downregulation of miR-127-3p observed in the transition from HSC to MPP coincides with the loss of self-renewal ability that is known to occur during this differentiation step. Moreover, miR-127-3p is absent in HSC obtained from Pbx1-cKO mice, which display a profound self-renewal defect.4 This stimulates future studies to investigate if Pbx1, together with its homeobox partners, directly regulates miR-127 expression, or if miR-127 OE can rescue part of the severe Pbx1cKO phenotype. Our data strongly suggest that HSC must maintain proper levels of miR-127-3p to preserve their long-term self-renewal capacity. We also found that miR29a, miR-99 and miR-126a are expressed in HSC but also 1753

L. Crisafulli et al.

present in Flk2â&#x20AC;&#x201C;MPP although at lower levels (Figure 2B and C and Online Supplementary Table S1). However: 1) none of them was DE in HSC from our self-renewal defective mouse model; or 2) the downregulation was less than 3-fold. Among other miRNA whose function in the hematopoietic system has been previously established, miR-29a and miR-126 were shown to play a role in HSC self-renewal and to be differentially expressed in HSC compared to lineage-committed progenitors, but still highly expressed at the MPP stage,11,45 whereas miR-99 was recently found to be DE in the HSC-to-MPP transition.20 Interestingly, these miRNA regulate HSC pool maintenance through different biological mechanisms. miR-126 restrains HSC cell cycle entry, while miR-99 and miR-29a limit HSC maturation, similar to miR-127. However, the accelerated differentiation displayed by miR-127DR cells was not associated to increased apoptosis or altered cell cycle. This might explain why the effect of miR-127DR was only observed with secondary transplants, since HSC depletion was likely due to a gradual exit of HSC from the self-renewing pool. Likewise, it is not surprising that the accelerated differentiation did not result in an increased differentiated progeny in vivo, since: a) differentiation was not associated with increased prolif-

References 1. Orkin SH, Zon LI. Hematopoiesis: an evolving paradigm for stem cell biology. Cell. 2008;132(4):631-644. 2. Wilson A, Laurenti E, Trumpp A. Balancing dormant and self-renewing hematopoietic stem cells. Curr Opin Genet Dev. 2009;19(5):461-468. 3. Ha M, Kim VN. Regulation of microRNA biogenesis. Nat Rev Mol Cell Bio. 2014;15(8):509-524. 4. Ficara F, Murphy MJ, Lin M, Cleary ML. Pbx1 regulates self-renewal of long-term hematopoietic stem cells by maintaining their quiescence. Cell Stem Cell. 2008;2(5):484-496. 5. Ficara F, Crisafulli L, Lin C, et al. Pbx1 restrains myeloid maturation while preserving lymphoid potential in hematopoietic progenitors. J Cell Sci. 2013; 126(Pt14):3181-3191. 6. Koss M, Bolze A, Brendolan A, et al. Congenital asplenia in mice and humans with mutations in a Pbx/Nkx2-5/p15 module. Dev Cell. 2012;22(5):913-926. 7. Deo A, Carlsson J, Lindlof A. How to choose a normalization strategy for miRNA quantitative real-time (qPCR) arrays. J Bionif Comput Biol. 2011;9(6):795812. 8. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A. 2001;98(9):51165121. 9. Amendola M, Passerini L, Pucci F, Gentner B, Bacchetta R, Naldini L. Regulated and multiple miRNA and siRNA delivery into primary cells by a lentiviral platform. Mol Ther. 2009;17(6):1039-1052. 10. Gentner B, Schira G, Giustacchini A, et al.

1754

11.

12.

13.

14.

15.

16.

17.

18.

19.

eration; and b) miR-127 is not normally expressed beyond the HSC stage, therefore an effect of miR-127DR in more mature cell population is not expected. In conclusion, we identified miR-127 as novel player for preserving the HSC pool. Our discovery has potential implications for several translational aspects of experimental hematology, including hematopoietic malignancies, BM transplantation, and regenerative medicine. Acknowledgments The authors would like to thank Dario Strina, Stefano Mantero and Lucia Susani for technical assistance, Massimiliano Mirolo and Ciro Menale for technical suggestions, and Paolo Vezzoni for helpful discussions and for critically reading the manuscript. Funding We acknowledge support from Marie Curie IRG 256452, from Ricerca Finalizzata GR 2010-2307975 and from AIRCFondazione Cariplo (TRIDEO 15882) to FF, from CNR National Program Aging Project to AV, and NIH grant CA116606 to ML. LC was recipient of a fellowship from Fondazione Nicola del Roscio; SM was recipient of a fellowship from Fondazione Damiano per lâ&#x20AC;&#x2122;Ematologia.

Stable knockdown of microRNA in vivo by lentiviral vectors. Nat Methods. 2009;6(1):63-66. Lechman ER, Gentner B, van Galen P, et al. Attenuation of miR-126 activity expands HSC in vivo without exhaustion. Cell Stem Cell. 2012;11(6):799-811. Follenzi A, Ailles LE, Bakovic S, Geuna M, Naldini L. Gene transfer by lentiviral vectors is limited by nuclear translocation and rescued by HIV-1 pol sequences. Nat Genet. 2000;25(2):217-222. Guenechea G, Gan OI, Inamitsu T, et al. Transduction of human CD34+ CD38bone marrow and cord blood-derived SCID-repopulating cells with third-generation lentiviral vectors. Mol Ther. 2000;1(6):566-573. Chen C, Ridzon DA, Broomer AJ, et al. Real-time quantification of microRNAs by stem-loop RT-PCR. Nucleic Acids Res. 2005;33(20):e179. Felli N, Fontana L, Pelosi E, et al. MicroRNAs 221 and 222 inhibit normal erythropoiesis and erythroleukemic cell growth via kit receptor down-modulation. Proc Natl Acad Sci U S A. 2005; 102(50):18081-18086. Gerrits A, Walasek MA, Olthof S, et al. Genetic screen identifies microRNA cluster 99b/let-7e/125a as a regulator of primitive hematopoietic cells. Blood. 2012; 119(2):377-387. Guo S, Lu J, Schlanger R, et al. MicroRNA miR-125a controls hematopoietic stem cell number. Proc Natl Acad Sci U S A. 2010; 107(32):14229-14234. O'Connell RM, Chaudhuri AA, Rao DS, Gibson WS, Balazs AB, Baltimore D. MicroRNAs enriched in hematopoietic stem cells differentially regulate long-term hematopoietic output. Proc Natl Acad Sci U S A. 2010;107(32):14235-14240. Ooi AG, Sahoo D, Adorno M, Wang Y,

20.

21.

22.

23.

24.

25.

26.

27.

28.

Weissman IL, Park CY. MicroRNA-125b expands hematopoietic stem cells and enriches for the lymphoid-balanced and lymphoid-biased subsets. Proc Natl Acad Sci U S A. 2010;107(50):21505-21510. Khalaj M, Woolthuis CM, Hu W, et al. miR99 regulates normal and malignant hematopoietic stem cell self-renewal. J Exp Med. 2017;214(8):2453-2470. Leucci E, Onnis A, Cocco M, et al. B-cell differentiation in EBV-positive Burkitt lymphoma is impaired at posttranscriptional level by miRNA-altered expression. Int J Cancer. 2010;126(6):1316-1326. Zonari E, Pucci F, Saini M, et al. A role for miR-155 in enabling tumor-infiltrating innate immune cells to mount effective antitumor responses in mice. Blood. 2013;122(2):243-252. Ito K, Hirao A, Arai F, et al. Regulation of oxidative stress by ATM is required for self-renewal of haematopoietic stem cells. Nature. 2004;431(7011):997-1002. Kent DG, Copley MR, Benz C, et al. Prospective isolation and molecular characterization of hematopoietic stem cells with durable self-renewal potential. Blood. 2009;113(25):6342-6350. Kent DG, Dykstra BJ, Eaves CJ. Isolation and Assessment of Single Long-Term Reconstituting Hematopoietic Stem Cells from Adult Mouse Bone Marrow. Curr Protoc Stem Cell Biol. 2016;38:2A.4.12A.4.24. Seita J, Sahoo D, Rossi DJ, et al. Gene Expression Commons: an open platform for absolute gene expression profiling. PloS One. 2012;7(7):e40321. Chen J, Wang M, Guo M, Xie Y, Cong YS. miR-127 regulates cell proliferation and senescence by targeting BCL6. PloS One. 2013;8(11):e80266. Qian P, He XC, Paulson A, et al. The Dlk1Gtl2 Locus Preserves LT-HSC Function by

haematologica | 2019; 104(9)

miR-127 controls HSC maintenance

29.

30.

31.

32.

33.

34.

Inhibiting the PI3K-mTOR Pathway to Restrict Mitochondrial Metabolism. Cell Stem Cell. 2016;18(2):214-228. Saito Y, Liang G, Egger G, et al. Specific activation of microRNA-127 with downregulation of the proto-oncogene BCL6 by chromatin-modifying drugs in human cancer cells. Cancer Cell. 2006;9(6):435-443. Lin SP, Youngson N, Takada S, et al. Asymmetric regulation of imprinting on the maternal and paternal chromosomes at the Dlk1-Gtl2 imprinted cluster on mouse chromosome 12. Nat Genet. 2003;35(1):97-102. Zhou Y, Cheunsuchon P, Nakayama Y, et al. Activation of paternally expressed genes and perinatal death caused by deletion of the Gtl2 gene. Development. 2010; 137(16):2643-2652. Ito M, Sferruzzi-Perri AN, Edwards CA, et al. A trans-homologue interaction between reciprocally imprinted miR-127 and Rtl1 regulates placenta development. Developmen. 2015;142(14):2425-2430. Jiang H, Hua D, Zhang J, et al. MicroRNA127-3p promotes glioblastoma cell migration and invasion by targeting the tumorsuppressor gene SEPT7. Oncol Rep. 2014;31(5):2261-2269. Shi L, Wang Y, Lu Z, et al. miR-127 promotes EMT and stem-like traits in lung cancer through a feed-forward regulatory loop.

haematologica | 2019; 104(9)

Oncogene. 2017;36(12):1631-1643. 35. Wang Y, Kong D. Knockdown of lncRNA MEG3 inhibits viability, migration, and invasion and promotes apoptosis by sponging miR-127 in osteosarcoma cell. Cell Biochem. 2018;119(1):669-679. 36. Bi L, Yang Q, Yuan J, et al. MicroRNA-1273p acts as a tumor suppressor in epithelial ovarian cancer by regulating the BAG5 gene. Oncol Rep. 2016;36(5):2563-2570. 37. Gao X, Wang X, Cai K, et al. MicroRNA127 is a tumor suppressor in human esophageal squamous cell carcinoma through the regulation of oncogene FMNL3. Eur J Pharmacol. 2016;791:603610. 38. Herr I, Sahr H, Zhao Z, et al. MiR-127 and miR-376a act as tumor suppressors by in vivo targeting of COA1 and PDIA6 in giant cell tumor of bone. Cancer Lett. 2017;409:49-55. 39. Wang D, Tang L, Wu H, Wang K, Gu D. MiR-127-3p inhibits cell growth and invasiveness by targeting ITGA6 in human osteosarcoma. IUBMB Life. 2018;70(5):411419. 40. Yu Y, Liu L, Ma R, Gong H, Xu P, Wang C. MicroRNA-127 is aberrantly downregulated and acted as a functional tumor suppressor in human pancreatic cancer. Tumor Biol. 2016;37(10):14249-14257.

41. Zhang J, Hou W, Chai M, et al. MicroRNA127-3p inhibits proliferation and invasion by targeting SETD8 in human osteosarcoma cells. Biochem Biophys Res Commun. 2016;469(4):1006-1611. 42. Dixon-McIver A, East P, Mein CA, et al. Distinctive patterns of microRNA expression associated with karyotype in acute myeloid leukaemia. PloS One. 2008; 3(5):e2141. 43. Ma H, Lin Y, Zhao ZA, et al. MicroRNA127 Promotes Mesendoderm Differentiation of Mouse Embryonic Stem Cells by Targeting Left-Right Determination Factor 2. J Biol Chem. 2016;291(23):12126-12135. 44. Zhai L, Wu R, Han W, Zhang Y, Zhu D. miR-127 enhances myogenic cell differentiation by targeting S1PR3. Cell Death Dis. 2017;8(3):e2707. 45. Han YC, Park CY, Bhagat G, et al. microRNA-29a induces aberrant selfrenewal capacity in hematopoietic progenitors, biased myeloid development, and acute myeloid leukemia. J Exp Med. 2010;207(3):475-489. 46. Liu J, Zhang J, Ginzburg Y, et al. Quantitative analysis of murine terminal erythroid differentiation in vivo: novel method to study normal and disordered erythropoiesis. Blood. 2013;121(8):e43-49.

1755

ARTICLE Ferrata Storti Foundation

Iron Metabolism & its Disorders

Dimeric ferrochelatase bridges ABCB7 and ABCB10 homodimers in an architecturally defined molecular complex required for heme biosynthesis

Nunziata Maio, Ki Soon Kim, Gregory Holmes-Hampton, Anamika Singh and Tracey A. Rouault

Haematologica 2019 Volume 104(9):1756-1767

Molecular Medicine Branch, ‘Eunice Kennedy Shriver’ National Institute of Child Health and Human Development, Bethesda, MD, USA

ABSTRACT

Correspondence: TRACEY A. ROUAULT rouault@mail.nih.gov Received: December 12, 2018. Accepted: February 7, 2019. Pre-published: February 14, 2019. doi:10.3324/haematol.2018.214320 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1756 ©2019 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.

1756

oss-of-function mutations in the ATP-binding cassette (ABC) transporter of the inner mitochondrial membrane, ABCB7, cause X-linked sideroblastic anemia with ataxia, a phenotype that remains largely unexplained by the proposed role of ABCB7 in exporting a special sulfur species for use in cytosolic iron-sulfur (Fe-S) cluster biogenesis. Here, we generated inducible ABCB7-knockdown cell lines to examine the timedependent consequences of loss of ABCB7. We found that knockdown of ABCB7 led to significant loss of mitochondrial Fe-S proteins, which preceded the development of milder defects in cytosolic Fe-S enzymes. In erythroid cells, loss of ABCB7 altered cellular iron distribution and caused mitochondrial iron overload due to activation of iron regulatory proteins 1 and 2 in the cytosol and to upregulation of the mitochondrial iron importer, mitoferrin-1. Despite the exceptionally large amount of iron imported into mitochondria, erythroid cells lacking ABCB7 showed a profound hemoglobinization defect and underwent apoptosis triggered by oxidative stress. In ABCB7-depleted cells, defective heme biosynthesis resulted from translational repression of ALAS2 by iron regulatory proteins and from decreased stability of the terminal enzyme ferrochelatase. By combining chemical crosslinking, tandem mass spectrometry and mutational analyses, we characterized a complex formed of ferrochelatase, ABCB7 and ABCB10, and mapped the interfaces of interactions of its components. A dimeric ferrochelatase physically bridged ABCB7 and ABCB10 homodimers by binding near the nucleotide-binding domains of each ABC transporter. Our studies not only underscore the importance of ABCB7 for mitochondrial FeS biogenesis and iron homeostasis, but also provide the biochemical characterization of a multiprotein complex required for heme biosynthesis.

Introduction ATP-binding cassette (ABC) transporters belong to one of the most abundant families of integral membrane proteins found in all kingdoms of life1,2 and play major roles in several biological processes by mediating the active transport of a variety of molecules across cellular membranes. Three members of the ABC family have thus far been localized to the inner mitochondrial membrane, where they are predicted to act as exporters, since their nucleotide binding domains face the matrix.3,4 These members are ABCB7 (the human ortholog of yeast Atm1), ABCB10 and ABCB8. ABCB6 has been reported to reside either in the outer mitochondrial membrane,5,6 and/or in the Golgi,7 lysosomal,8 and plasma membranes.9 ABCB7 maps to the X-chromosome in mice and humans10 and shows a ubiquitous expression pattern. Knockout studies in mice revealed that expression of ABCB7 was essential for early gestation.11 Mutations in ABCB7 cause X-linked sideroblastic anemia with ataxia (XLSA/A; 301310), which is a recessive disorder characterized by the onset of non- or slowly-progressive cerebellar ataxia and haematologica | 2019; 104(9)

A unique ABCB7-FECH-ABCB10 complex

anemia with hypochromia and microcytosis in infancy or early childhood.12-14 Bone marrow examination showed ringed sideroblasts, which give the condition its name. Complementation assays in yeast suggested that each of the human mutations caused a mild partial loss of function.12,13 Conditional gene targeting in mice showed that ABCB7 was essential for hematopoiesis15 and for the development and function of all tissues and organs analyzed.11 Here, we examined the time-dependent consequences of loss of ABCB7 in multiple cell culture models. We found that knockdown (KD) of ABCB7 led to significant loss of mitochondrial iron-sulfur (Fe-S) proteins, which preceded the development of comparatively milder defects in cytosolic Fe-S enzymes. In erythroid cells, loss of Abcb7 caused defective heme biosynthesis and altered cellular iron distribution with mitochondrial iron overload, which triggered oxidative stress and led to apoptosis of erythroid progenitors. By combining chemical crosslinking with tandem mass spectrometry and mutational analyses, we identified a complex formed of ferrochelatase (FECH), ABCB7 and ABCB10 and characterized its overall architecture. Our studies uncovered the importance of ABCB7 for mitochondrial function and iron homeostasis and identified a previously uncharacterized complex that is required for heme biosynthesis.

Methods Cell lines and cell culture conditions HEK293T and HeLa cells were purchased from the American Type Culture Collection (ATCC) and propagated in Dulbecco modified Eagle medium with 4.5 g/L glucose, 10% fetal bovine serum, and 2 mM glutamine at 37°C, 5% CO2 in a humidified incubator. G1E-ER4 cells were maintained in Iscove modified Dulbecco medium with 15% fetal bovine serum, 100 U/mL penicillin-streptomycin, 2 U/mL erythropoietin (Sino Biological Inc.), monothioglycerol (1:10,000), and 50 ng/mL Kit-ligand (R&D Systems). GATA1-mediated differentiation was induced by the addition of 100 nmol of β-estradiol to a cell culture at a density of 2×105 cells/mL. All cell lines were tested for mycoplasma.

Short hairpin and small interfering RNA-mediated knockdown of ABCB7, MFRN2 and IRP2 in HEK293T, HeLa or G1E-ER4 cells The SMARTvector Inducible Lentiviral short hairpin (sh)RNA system (Dharmacon) was used to generate HEK293T and HeLa stable cell clones with tightly controlled expression of three individual shRNA targeting different regions of the ABCB7 transcript and a scrambled shRNA, used as negative control. Knockdown of Abcb7 or Irp2 in G1E-ER4 cells was achieved with the Accell small interfering RNA delivery system (Dharmacon). Further details of the knockdown procedures, together with information on in vitro and in vivo crosslinking and mass spectrometry, in vitro coupled transcription/translation and pull-down assay of 35S-labeled proteins, the dihydropyrimidine assay, gel electrophoresis, complex I, II and III activity assays, iron and heme measurements, histological staining, flow cytometry studies, assays of superoxide dismutase activity, aconitase and catalase, along with polymerase chain reaction studies and other methods are provided in the Online Supplementary Methods.

haematologica | 2019; 104(9)

Results Knockdown of ABCB7 led to dysfunction of mitochondrial proteins which preceded the development of milder defects in cytosolic Fe-S enzymes We analyzed the effects of progressive loss of ABCB7 in HEK293T, HeLa and G1E-ER4 cells. To interrogate the role of ABCB7 during erythroid differentiation and address the effect of ABCB7 mutations in XLSA patients,12-14 we studied G1E-ER4 cells, a well-validated murine cell model that recapitulates the early stages of terminal red blood cell development.16 We used an inducible shRNA-mediated KD approach to silence the expression of ABCB7 in HEK293T (Online Supplementary Figure S1A) and HeLa cells (Online Supplementary Figure S1B), which enabled us to discern primary from secondary effects by observing when defects appeared during the time-course after induction of KD. KD of ABCB7 resulted in a time-dependent loss of mitochondrial Fe-S proteins, including FECH, glutaredoxin 5 (GLRX5) and multiple subunits of the respiratory complexes I and II (NDUFS1 and NDUFS8 in complex I and SDHB in complex II) in all the cell lines tested (Figure 1A,B, Online Supplementary Figure S2C-E, and S2H). Loss of ABCB7 also elicited profound transcriptional remodeling that included the downregulation of the vast majority of subunits and assembly factors of the mitochondrial respiratory chain within 48 h after KD (Figure 1C) in G1E-ER4 cells and profoundly affected cell morphology (Online Supplementary Figure S1C). Loss of the catalytic subunits resulted in a significant decrease in the activities of complexes I and II (Figures 1A, B, and D). Mitochondrial aconitase (ACO2) activity was also markedly reduced by 80%, whereas cytosolic aconitase (ACO1) decreased to a lesser extent (40%) (Online Supplementary Figure S2H), likely due to the fact that mitochondrial dysfunction altered cytosolic iron status and activated the RNA-binding activity of iron-regulatory protein-1 (IRP1) 17 (see later in the paper), without changing IRP1 protein levels (Online Supplementary Figure S2H). Notably, levels of cytosolic FeS proteins, including CIAPIN1, GLRX3, POLD1, DPYD, PPAT, ERCC2, ELP3 and ABCE1, were unchanged 3 days after KD of ABCB7 (Figure 1E and Online Supplementary Figure S2B). Activity of the cytosolic Fe-S enzyme DPYD (Online Supplementary Figure S2F) and radioactive iron incorporation into the cytosolic Fe-S protein NUBP2 were also unchanged (Online Supplementary Figure S2G). Decreased stability of the cytosolic Fe-S proteins DPYD, PPAT and POLD1 was observed 5 days after KD of ABCB7 (Online Supplementary Figure S2C,D); however, the extent of loss of mitochondrial Fe-S proteins at day 5 was much more profound than the decrease in cytosolic Fe-S protein levels (Online Supplementary Figure S2C,D). Consistent with the importance of mitochondria in performing a global regulatory role in numerous cellular processes linked to iron homeostasis, we found that KD of ABCB7 stabilized iron-regulatory protein-2 (IRP2), a master regulator of iron metabolism18 (Figure 1D and Online Supplementary Figure S2C,D). Levels of the erythroid-specific mitochondrial iron transporter mitoferrin-1 (MFRN1) (Figure 1D) and the ubiquitously expressed mitoferrin-2 (MFRN2) in HEK293T and HeLa cells (Figure 1A and Online Supplementary Figure S2D) increased significantly in ABCB7-KD cells. Overall, stabilization of IRP2 in the 1757

N. Maio et al.

cytosol and increased MFRN1 or MFRN2 in mitochondria likely increased flux of iron into mitochondria in ABCB7depleted cells (further investigated later in this study). Reexpression of wildtype ABCB7, but not the pathogenic XLSA E433K mutant13 in the ABCB7-KD cells restored lev-

els of FECH and SDHB, activities of complex II and aconitases and normalized levels of MFRN1 and IRP2 (Figure 1D and Online Supplementary Figure S2H), demonstrating that loss of ABCB7 caused profound early-onset of mitochondrial dysfunction.

Figure 1. Loss of ABCB7 disrupts mitochondrial function and downregulates genes involved in mitochondrial energy metabolism. (A) Protein blots and respiratory complexes I (CI) and II (CII) in-gel activity assays on mitochondrial lysates from HEK293T cells expressing an inducible control short hairpin (sh)RNA (sh_CTRL) or two shRNA targeting different regions of the ABCB7 transcript (sh_ABCB7-1 and sh_ABCB7-2) for 3 days. Levels of ABCB7, mitoferrin 2 (MTFN2), superoxide dismutase 2 (SOD2), aconitase (ACO2), CII subunits SDHA and SDHB, ferrochelatase (FECH), the CI Fe-S subunit NDUFS8, glutaredoxin 5 (GLRX5), complex IV subunit MTCO1, and citrate synthase (CS) were assayed. Levels of VDAC1 and TOM20 were used as loading controls. (B) Protein blots and CI and CII in-gel activity assays on mitochondrial lysates from G1E-ER4 control cells or from cells depleted of Abcb7 for 3 days. Comparisons between cells at the burst-forming unit erythroid stage (undifferentiated, without Î˛-estradiol) or at the orthochromatophilic stage (differentiated for 72 h with Î˛-estradiol) are shown. Levels of Abcb7, CI Fe-S subunit Ndufs1, CII subunits Sdha and Sdhb, Fech, CIII subunit Uqcrc2, Suclg2, mitochondrial unfoldase Clpx and protease Clpp were assessed by western blot. Levels of Vdac1 and Tom20 were used as loading controls. (C) Log2-fold expression of mitochondrial energy metabolism pathway genes which were differentially expressed in G1E-ER4 cells 48 h after knockdown (KD) of Abcb7 (false discovery rate <0.01, n=3 biological replicates). (D) Complementation assays on G1E-ER4 cells silenced for 3 days to KD the expression of Abcb7 and transfected with wildtype FLAG-tagged ABCB7 or with the X-linked sideroblastic anemia pathogenic mutant ABCB7E433K-F. Levels of Abcb7, Irp2, Pold1, Fech and Mfrn1 are shown, along with CII in-gel activity assay. Tom20 and a-Tubulin (Tub) were used as loading controls for the mitochondrial fractions and total lysates, respectively. (E) Protein blots on total lysates from G1E-ER4 cells. Levels of the cytosolic and nuclear Fe-S proteins Ciapin1, Glrx3, Pold1, Dpyd and Ppat are shown, along with levels of the CIA components Ciao1, Fam96b and Mms19. Tub was used as a loading control. (A, B, D and E, n=6 biological replicates). See also Online Supplementary Figures S13 and S14 for densitometries of immunoblots and statistical analyses. NT: not treated (control).

1758

haematologica | 2019; 104(9)

A unique ABCB7-FECH-ABCB10 complex

Loss of Abcb7 altered cellular iron distribution and caused oxidative stress and apoptosis of erythroid progenitors We investigated the consequences of loss of Abcb7 in developing G1-ER4 cells, which require large amounts of

iron inside mitochondria to support heme synthesis. We first analyzed changes in the activation of the iron responsive element (IRE)-binding activities of Irp1 and Irp2, which post-transcriptionally shape the expression of the mammalian iron metabolism proteome by binding to

F I E

Figure 2. Loss of Abcb7 alters cellular iron distribution and leads to oxidative damage and apoptosis of erythroid progenitors. (A) Iron responsive elemebt (IRE)binding activities of Irp1 and Irp2 in control and Abcb7-knockdown (KD) G1E-ER4 cells before differentiation (-β-estradiol) and 10, 24, 48 or 72 h after differentiation in the presence of β-estradiol. (B) In-gel aconitase activity assay on control or Abcb7-KD G1E-ER4 cells before and after 72 h of differentiation, showing the activities of mitochondrial and cytosolic aconitase (Aco2 and Aco1, respectively). Immunoblots to Irp2, Irp1, Aco2, Tfrc, ferritin (Ft), Hba-a2, Hbb-b1 and Mfrn1 on the same set of samples analyzed in the aconitase activity assays are shown. Tubulin (Tub) was used as a loading control. 55Fe autoradiogram on G1E-ER4 cells treated as for the aconitase assays, showing levels of radiolabeled iron incorporated into the iron storage protein Ft or into hemoglobin (Hb). (C) IRE-binding activities of Irp1 and Irp2 in G1-ER4 cells silenced for Abcb7 for 3 days and differentiated in the presence of β-estradiol for 72 h, and immunoblots to Irp2 and to Tub. (D) Protein levels of Irp2, Alas2 and Bach1 in G1E-ER4 cells before and 30 h after differentiation. (E) Iron content in mitochondria was significantly increased in Abcb7-KD (day 3) cells before (left lanes) differentiation. (F) The cytosolic labile iron pool (LIP) was decreased in cells treated as in (E). The mean fluorescence intensity (MFI) is shown in arbitrary units (a.u.). (G) Representative flow cytometry analysis to sort G1E-ER4 control or Abcb7-KD cells that were double-positive for the apoptotic marker (annexin V) and for mitochondrial reactive oxygen species production (Mitosox) (top, right square). (H) Quantification of G1E-ER4 cells double-positive for annexin V and Mitosox from data shown in (G). (I) In-gel activity assay of Sod1 and Sod2 in G1E-ER4 cells treated as in (E) showed significantly higher activity and protein levels in cells depleted of Abcb7 for 3 days. Data in (E) (F) and (H) are expressed as mean ± standard deviation. (A-I, n=5). See also Online Supplementary Figure S14 for densitometries of immunoblots and statistical analyses. NT: not treated (control). ***P<0.001

haematologica | 2019; 104(9)

1759

N. Maio et al.

IRE in the messenger RNA of transcripts involved in iron homeostasis.18 G1E-ER4 cells robustly activated the IREbinding activities of Irp1 and Irp2 within 10 h after induction of differentiation, with maximal activation at 24 h and 48 h (Figure 2A). By 72 h in differentiation medium, when cells had reached the orthochromatophilic erythroid stage and fulfilled their extracellular iron demands, the IRE-binding activities of Irp1 and Irp2 diminished (Figure 2A). In contrast, in Abcb7-KD G1E-ER4 cells, Irp2 activation was elicited well before induction of differentiation and its levels remained undiminished throughout the time course (Figure 2A). Control cells differentiated for 72 h showed increased sequestration of radiolabeled iron into ferritin (55Fe-autoradiogram) (Figure 2B), as already reported.19 In control cells, iron was also successfully incorporated into protoporphyrin IX (PPIX) by Fech, as radiolabeled hemoglobin (55Fe-Hb) increased significantly (Figure 2B). Compared to control (NT), cells lacking Abcb7 exhibited prolonged activation of the cytosolic iron starvation response with stabilization of Irp2, decreased cytosolic aconitase (Aco1) activity and increased IRE-binding activities of IRP (Figure 2A-C). Hyperactivation of IRP caused translational repression of ferritin (Figure 2B) and decreased radiolabeled-iron sequestration into the heteropolymeric storage form of the protein (Figure 2B). Under the transcriptional control of Gata-1,20 differentiating cells upregulated hemoglobin subunits (Hba-a2 and Hbb-b1) and Alas2 (Figure 2B,D, Online Supplementary Figure S3). Mfrn1 increased during differentiation to fulfill the exceptional iron demand for heme synthesis (Figure 2B and Online Supplementary Figure S3). Abcb7-depleted cells failed to increase levels of Alas2, due to translational repression by IRP at the 5â&#x20AC;&#x2122;-IRE present in its transcript21,22 (Figure 2D), even though Alas2 transcript levels were over 50-fold higher than those in undifferentiated cells (Online Supplementary Figure S3). Importantly, levels of Bach1, an erythroid transcriptional repressor that is rapidly degraded under conditions of heme sufficiency,23 increased upon KD of Abcb7 (Figure 2D), and levels of Hbb-b1 and Hba-a2, which are Bach1 targets,24 were repressed (Figure 2B and Online Supplementary Figure S3). Notably, Tfrc and Mfrn1 levels were elevated in Abcb7-KD cells prior to differentiation (Figure 2B and Online Supplementary Figure S4A,B). Total cellular iron content in early erythroid progenitors depleted of Abcb7 was about 7-fold higher than that in controls (Online Supplementary Figure S5A) and iron accumulated in mitochondria (Figure 2E), whereas the pool of available cytosolic labile iron was significantly reduced (Figure 2F), indicating a condition of cytosolic functional iron deficiency. Mitochondrial iron overload increased the production of reactive oxygen species by 30% (Mitosox+ cells) (Figure 2G,H), which damaged plasma membranes, as shown by the increased percentage of annexin V+ cells (Figure 2G,H and Online Supplementary Figure S5B), even though both mitochondrial and cytosolic superoxide dismutase enzymes (Sod2 and Sod1) were activated (Figure 2I), and were transcriptionally upregulated about 2- and 5-fold, respectively (Online Supplementary Figure S6A-C). Unexpectedly, most of the genes involved in the antioxidant response were downregulated in Abcb7-KD cells (Online Supplementary Figure S6A). Manganese accumulated in mitochondria of Abcb7-KD cells (Online Supplementary Figure S5C), raising the possibility that metalation of Sod2 accounted for the increase in man1760

ganese. KD of ABCB7 in HEK293T cells also caused mitochondrial iron accumulation (Online Supplementary Figure S5D), and activation of SOD enzymes (Online Supplementary Figure S5E,F). As MFRN2 was significantly upregulated upon silencing of ABCB7 in the non-erythroid HEK293T cells (Figure 1A), we knocked down its expression and found that cells depleted of MFRN2 maintained intact mitochondrial function (Online Supplementary Figure S5D,G). KD of MFRN2 in ABCB7depleted cells lowered mitochondrial iron accumulation (Online Supplementary Figure S5D), but levels and activities of Fe-S-dependent enzymes did not return to normal (Online Supplementary Figure S5H), suggesting that mitochondrial iron overload was not the primary cause of the compromise of Fe-S proteins in the matrix.

Heme biosynthesis defect in cells lacking ABCB7 Loss of Abcb7 during erythroid differentiation led to an 80% reduction in the levels of 55Fe-Hb, to the IRP-mediated translational repression of Alas2 and to Bach1-mediated blockage of Hba-a2 and Hbb-b1 transcription, pointing to a potential defect in heme biosynthesis. Treatment with succinylacetone, a potent inhibitor of Alad, confirmed that the defect of 55Fe-Hb synthesis caused by KD of Abcb7 during differentiation was comparable to the inhibition of the heme biosynthetic pathway (Figure 3A, 55 Fe-Hb autoradiogram, lanes 5 and 6). Alas2, Fech and Alad protein levels decreased in cells lacking Abcb7 (Figure 3B and Online Supplementary Figure S7A), whereas levels of the heme biosynthetic enzyme Cpox,25 of Sucla2 and Suclg2, which provide succinyl-CoA to the first ratelimiting step of the pathway.26 and of the unfoldase Clpx, required for the pyridoxal-phosphate-dependent activation of Alas2,27 did not change (Online Supplementary Figure S7A). Abcb7-KD cells exhibited impaired hemoglobinization (Figure 3C). Heme content decreased by 75% (Figure 3D,E) and activities of the heme-containing enzymes cytochrome c oxidase (Figure 3F) and peroxisomal catalase (Figure 3G and Online Supplementary Figure S7C-E) were reduced by more than 80%. Levels of respiratory complexes III and IV, which require three and two heme centers, respectively, were significantly reduced in Abcb7-KD cells (Figure 3F). Heme-bound cytochromes c (Cyc) and c1 (Cyc1) decreased (Holo-cyt) (Figure 3F), without changes in the levels of total apo-cytochromes (Online Supplementary Figure S7B). The heme-regulated eIF2a kinase (HRI), which inhibits the general translation initiation factor eIF2a under conditions of heme deficiency,28 may mediate the translational repression of the globin chains, which are also transcriptionally repressed by Bach1 (Online Supplementary Figure S3), in Abcb7-depleted erythroid progenitors to prevent accumulation of globins in excess of heme, and may also repress translation of the heme biosynthetic enzymes Alas2 and Alad.28 A defect in heme biosynthesis was also observed in HEK293T and HeLa cells upon KD of ABCB7 (Online Supplementary Figure S7F-K).

Irp2 activation in Abcb7-depleted cells sustained mitochondrial iron overload mediated by mitoferrin-1 upregulation We investigated the correlation between loss of Abcb7 and activation of IRP, by analyzing the effect of double KD of Abcb7 and Irp2 in developing erythroid cells. KD of Irp2 significantly diminished the stability of Tfrc haematologica | 2019; 104(9)

A unique ABCB7-FECH-ABCB10 complex

(Figure 4A) and decreased Mfrn1 to levels comparable to those of cells treated with desferrioxamine (Figure 4A). A detectable compensatory activation of the IRE-binding activity of Irp1 in Irp2-KD cells (Figure 4B) was insufficient to stabilize Tfrc to levels comparable to those in

controls. Mitochondrial iron levels decreased by more than 50% (Figure 4C), due to the decreased half-life of MFRN1 under the iron-limiting conditions generated by the lack of Irp2 (Figure 4D). Levels of radiolabeled hemoglobin dropped by more than 80% (Figure 4E,F), causing

G D

Figure 3. Heme biosynthesis defect in cells lacking Abcb7. (A) Protein blots to Irp2, Irp1, Tfrc, Hbb-b1, Hba-a2, and Bach1 in control (NT) and Abcb7-KD G1E-ER4 cells before (-β-estradiol) and after (+β-estradiol) 72 h of differentiation. The effect of the Alad inhibitor succinylacetone (SA) during cell differentiation was also tested. 55Fe incorporated into hemoglobin (Hb) in cells silenced for Abcb7 was profoundly decreased. (B) Mitochondrial fractions from G1E-ER4 cells treated as in (A) were probed with antibodies against Abcb7, Alas2, Abcb10, Fech, and Mfrn1. Tom20 was used as a loading control. (C) Cell pellets of control (NT β-estr) or Abcb7knockdown (KD) (si-Abcb7 β-estr) G1E-ER4 cells differentiated for 72 h showed decreased hemoglobinization in cells depleted of Abcb7. (D) Representative oxidized UV-VIS spectra of heme, with the characteristic Soret band at 414 nm and additional peaks at 541 and 576 nm, in control and Abcb7-KD G1E-ER4 cells 72 h after differentiation showed significantly lower heme levels in cells depleted of Abcb7. (E) Heme levels in control and GIE-ER4 cells depleted of Abcb7 for 3 days. Values are means ± standard deviation. (F) Levels of heme-bound cytochromes c (Cyc) and c1 (Cyc1) were significantly decreased in Abcb7-depleted G1E-ER4 cells. Native immunoblots to Atp5a (CV subunit), Uqcrc2 (CIII subunit) and Mtco1 (CIV subunit) showed that levels of heme-containing complexes CIII and CIV were significantly decreased in Abcb7-KD cells. CIV activity, which requires two heme centers, was also decreased. Tom20 was used as a loading control. Peroxisomal fractions showed comparable levels of catalase protein in control and Abcb7-KD G1E-ER4 cells. Catalase (Cat) protein levels increased to the same extent in control and Abcb7-KD cells during differentiation. Pmp70 (also known as Abcd3) is a marker of peroxisomes and was used as a loading control. (G) Catalase activity in control and Abcb7KD cells before and after differentiation showed profound defects in heme-dependent catalase activity in cells depleted of Abcb7. Values are expressed as % of control. Values are means ± standard error of mean. (A-C, E-G, n=6; D, n=3). See also Online Supplementary Figure S15 for densitometries of immunoblots and statistical analyses. ***P<0.001

haematologica | 2019; 104(9)

1761

N. Maio et al.

an evident hemoglobinization defect in cells lacking Irp2 (Figure 4G). Double KD of Abcb7 and Irp2 reduced iron overload in early erythroid progenitors (Figure 4C). However, it also severely impaired iron delivery to mitochondria during differentiation (Figure 4C) and impaired hemoglobinization (Figure 4G), confirming the essential role of Irp2 in erythropoiesis.29

ABCB7 formed a complex with FECH and ABCB10 The stability of Fech was reduced by the KD of Abcb7. We found that endogenous Abcb7 interacted with Fech and Abcb10 in G1E-ER4 cells during differentiation (Figure 5A), in agreement with published studies that reported the interaction of Fech with Abcb730-32 or with Abcb10.31,32 Using two-dimensional/Blue-Native(BN)/sodium dodecylsulfate

Figure 4. Irp2 activation sustains mitochondrial iron overload mediated by mitoferrin-1 upregulation in erythroid cells depleted of Abcb7. (A) Irp2, Tfrc, ferritin H (Fth), Mfrn1, Alas2, Aco2, Hspa9 and Fam96b levels in G1E-ER4 cells before and during differentiation and either upon treatment with the iron chelator desferrioxamine (DFO) or upon knockdown (KD) of Irp2. Tom20 and tubulin (Tub) were used as loading controls for mitochondrial and cytosolic fractions, respectively. (B) Ironresponsive element (IRE)-binding activities of Irp1 and Irp2 in G1E-ER4 cells differentiated for 72 h and transfected with small interfering (si)-RNA to KD the expression of Irp2, Abcb7 or both for 3 days. DFO-treated samples were run alongside to compare the KD effect with iron deficiency on the activation of iron regulatory proteins (IRP). Irp1 blotting did not show a change in protein levels. (C) Mitochondrial iron in control and Abcb7-, Irp2- or Abcb7/Irp2- double-KD G1E-ER4 cells before and after 72 h of differentiation. Loss of Irp2 reduced iron delivery to mitochondria. (D) A pulse-chase experiment was performed to assess the turnover rate of Mfrn1 under the iron-limiting conditions generated by KD of Irp2. G1E-ER4 cells were silenced for 72 h to KD the expression of Irp2. Cells were then pulsed for 30 min with 35 S-Cys/Met, followed by incubation for the indicated time points in differentiation medium. Radiolabeled Mfrn1 was visualized by autoradiography after immunoprecipitation and sodium dodecylsulfate polyacrylamide gel electropheresis (top panel), whereas total protein levels were assessed by immunoblot (lower panels). (E) Levels of radiolabeled iron incorporated into hemoglobin in Abcb7-, Irp2- and Abcb7/Irp2 double-KD cells. (F) Irp2, Abcb7, Mfrn1 and Hba-a2 levels in G1E-ER4 cells differentiated for 72 h in the presence of Î˛-estradiol and silenced for either Abcb7, or Irp2 or simultaneously for Abcb7 and Irp2. A representative 55Fe-autoradiogram shows significantly decreased levels of radioactive iron incorporated into hemoglobin (Hb) in cells depleted of Abcb7, Irp2 or both. (G) Cell pellets of G1E-ER4 cells differentiated for 72 h and transfected with siRNA to KD the expression of Abcb7, Irp2 or both showed defective hemoglobinization in cells lacking Abcb7, Irp2 or both proteins. (A-D, F, G, n=4; E, n=6). See also Online Supplementary Figure S16 for densitometries of immunoblots and statistical analyses. NT: not treated (control). ***P<0.001

1762

haematologica | 2019; 104(9)

A unique ABCB7-FECH-ABCB10 complex

polyacrylamide gel electrophoresis (2D-SDS PAGE) separation, we further demonstrated that Abcb7, Fech and Abcb10 co-migrated in a single multimeric complex of an approximate molecular weight of 480 kDa (Figure 5B). Interestingly, a second distinctive pool of Fech formed a complex of 250 kDa, which also contained the heme-synthesizing enzymes Alas2 and Ppox (Figure 5B), consistent with previously published results that proposed a direct

transfer of PPIX from Ppox to Fech for iron insertion,32,33 and that suggested a potential regulatory role for the Fech-Alas2 interaction.32 We performed chemical crosslinking of the ABCB7/FECH complex in vitro on purified proteins and in vivo on mitochondria isolated from G1E-ER4 cells coexpressing human ABCB7-FLAG and FECH-HA, coupled with tandem mass spectrometry (XL-MS) to investigate the architecture of the ABCB7/FECH complex. The analysis

E D

G H

Figure 5. ABCB7 forms a complex with ferrochelatase and Abcb10 through direct interaction revealed by crosslinking. (A) Immunoprecipitation (IP) of endogenous Abcb7 in G1E-ER4 cells 30 h after induction of differentiation showed formation of a complex between Abcb7, Fech and Abcb10 that did not include Abcb8 or Mfrn1. (B) Native/two-dimensional (2D) sodium dodecylsulfate polyacrylamide gel electropheresis (SDS-PAGE) analysis on mitochondrial lysates from G1E-ER4 cells showed co-migration of Abcb7, Fech and Abcb10 in a single complex of an approximate molecular weight of 480 kDa. A second distinctive pool of Fech co-migrated with the heme-synthesizing enzymes Ppox and Alas2, indicating formation of a complex at 250 kDa. (C) Map showing the crosslinked sites in the ABCB7/FECH/Abcb10 complex. Inter-subunit crosslinks between ABCB7 and FECH are in blue, and inter-subunit crosslinks between FECH and Abcb10 are in magenta. The lysine residues crosslinked by DDS in the protein sequences are represented by dots. (D) Primary sequence of the C-terminal domain of ABCB7 between residues Val 411 and Cys752. Peptide sequences that were substituted by alanines to test their involvement in the interaction with Fech are highlighted in different colors. (E) Three-dimensional structure of ABCB7 modeled on the structure of S. cerevisiae Atm1 (PDB:4MYC 3) using Swiss-Model.49 The last 44 amino acid residues of ABCB7 are missing from the structure because yeast lacks these terminal residues. One of the two protomers of ABCB7 in the dimeric structure is represented in the surface-mode and the green and magenta sequences in the C terminus of ABCB7 indicate the peptide sequences subjected to alanine scanning mutagenesis in Mut1 and Mut5, respectively, to assess their involvement in the interaction with FECH. (F) Coomassie staining of inputs and in vitro crosslinked products on SDS-PAGE. Magenta asterisks denote dimers of ABCB7 wildtype (WT) or the mutant (Mut1) in which amino acid residues between Val450 and Leu463, involved in binding FECH, were replaced by alanines. Blue asterisks indicate the tetrameric ABCB7-FECH complex. Brown circles indicate FECH dimers. (G) SDS-PAGE analysis after in vivo crosslinking on mitochondria isolated from cells co-transfected with ABCB7-F and FECH-HA, followed by anti-FLAG immunoprecipitation. (H) Native PAGE on purified ABCB7-FLAG, FECH-HA and ABCB10-Myc proteins shows that both ABC transporters dimerized when loaded individually (lanes 1 and 3; magenta and black asterisks correspond to dimers of ABCB7 and ABCB10, respectively). Both ABCB7 and ABCB10 were able to interact physically with dimers of FECH (lanes 4 and 6; blue and orange asterisks denote the hetero-tetrameric ABC transporter-FECH complexes). ABCB7, FECH and ABCB10, when combined together in vitro, formed a multiprotein complex with a 2:2:2 stoichiometry consisting of dimers of each of the components (in lane 7, the complex denoted with the # symbol). ABCB7Mut1, in which amino acid residues between Val450 and Leu463 were replaced by alanines, was unable to interact with FECH (lane 5) and formation of the hexameric complex was disrupted (lane 8), whereas Mut5 showed no defect in interacting with FECH (lane 9). Brown circles indicate FECH dimers (lane 2). (A, B, n=5; F, G and H, n=3).

haematologica | 2019; 104(9)

1763

N. Maio et al.

yielded 147 MS/MS fragmentation spectra corresponding to 29 high-confidence lysine-lysine crosslinks (Online Supplementary Table S1, Figure 5C and Online Supplementary Figure S8A), among which 12 were intersubunit crosslinks between ABCB7 and FECH and ten were crosslinks between endogenous Abcb10 and FECH (Online Supplementary Table S1, Figure 5C and Online Supplementary Figure S8A,B). No crosslinks were detected between ABCB7 and Abcb10, suggesting that FECH mediated an indirect interaction between ABCB7 and ABCB10, as we subsequently confirmed by co-immunoprecipitation experiments. The analysis also identified a high confidence FECH-FECH intersubunit crosslink (K286-K286) (Online Supplementary Table S1). Identification of this crosslinked species is consistent with the distances derived from the crystal structure of FECH, which shows the two lysines at the FECH dimer interface with a Ca-Ca distance of 10.163Å, in agreement with the length of the DSS crosslinker (11.4 Å) (Online Supplementary Figure S9A). The crystal structure of FECH revealed that the PPIX substrate was deeply bound within a pocket that was enclosed by three movable regions34 (Online Supplementary Figure S9B), one of which (residues 90-115 of FECH) was involved in binding the nucleotide-binding domain of ABCB7, according to our data (Online Supplementary Table S1, Figure 5C and Online Supplementary Figure S9C). We performed alanine scanning mutagenesis on the C terminus of ABCB7 (residues V450C752) (Figure 5D,E), and tested the mutants in vivo and in vitro for their ability to interact with FECH. SDS-PAGE analysis on crosslinked purified ABCB7 and FECH proteins, showed three major bands in the presence of the crosslinker (BS3) with approximate molecular weights of 80, 120 and 230 kDa (Figure 5F and Online Supplementary Figure S10A), which met the expected molecular weights of a dimer of FECH (84 kDa), a dimer of ABCB7 (138.4 kDa), and a heterotetrameric complex consisting of a dimer of ABCB7 interacting with a dimer of FECH (222.4 kDa). Importantly, the same complexes at 120 and 230 kDa were also detected on SDS-PAGE after in vivo crosslinking on isolated mitochondria, followed by anti-FLAG immunoprecipitation of ABCB7-FLAG (Figure 5G). We also confirmed the previously reported interaction of ABCB10 with FECH31 in vitro with purified proteins (Online Supplementary Figure S10A,B), which demonstrated that binding of FECH to ABCB10 was direct. BN-PAGE analysis on ABCB7, FECH and ABCB10 purified proteins showed that each ABC transporter dimerized when loaded individually (Figure 5H), confirming the results obtained using the crosslinker (Figure 5F and Online Supplementary Figure S10A). Both ABCB7 and ABCB10 were able to interact physically with dimers of FECH and, when combined, ABCB7, FECH and ABCB10 assembled into a complex of approximately 480 kDa (Figure 5H, complex designated with the # symbol), which conformed to the predicted size of a hetero-hexameric complex composed of dimers of ABCB7, FECH, and ABCB10. The ABCB7-FECH interaction was disrupted upon expression of ABCB7Mut1, in which the peptide sequence from Val450 through Leu463 had been replaced by alanines (Figures 5F, H), whereas ABCB7Mut5 (residues L517-K526 replaced by alanines) was able to bind FECH and form a complex with ABCB10 to the same extent as ABCB7-WT (Figure 5H). ABCB7Mut6, in which the peptide sequence between amino acid residues Gly527 and Asp538 were replaced by alanines, also did not form a complex with FECH (Online Supplementary Figure S10A). Immunoprecipitation of recombinantly expressed 1764

FLAG-tagged ABCB7 wildtype or its mutants (Figure 6A,B) in vivo in G1E-ER4 cells, which co-expressed FECH-HA and that had been silenced to KD the expression of endogenous Abcb7, demonstrated that two short sequences in the C terminus of ABCB7, residues V450-L463 and G527-D538, were essential molecular mediators of the interaction with FECH (Figure 6C-E and Online Supplementary Figure S10C,D), thereby confirming our results obtained in vitro with purified proteins. Residues between Gln464 and Val504 of ABCB7 were also involved in stabilizing the binding of FECH (mutants 2, 3 and 4) (Figure 6C,D, Online Supplementary Figure S10D), whereas the ABCB7 sequences between G505 and Q525 and downstream of Val539 were not involved in the interaction with FECH (Figure 6C-E and Online Supplementary Figure S10C,D). Endogenous Abcb10 co-immunoprecipitated with the ABCB7/FECH complex (Figure 6C,D), consistent with formation of the native complex of 480 kDa (Figure 5H), and the amount recovered in the eluates after immunoprecipitation of ABCB7 decreased upon expression of the ABCB7 mutants, which were defective in binding FECH, suggesting that formation of a multimeric complex containing ABCB7 and ABCB10 homodimers was bridged by a dimeric FECH, rather than through direct physical contacts between the two ABC transporters. In vitro pull-down assays with 35S-radiolabeled ABCB7 wildtype or mutants in the presence of FECH confirmed direct binding of FECH to ABCB7 (Figure 6F-H). The half-life of FECH was significantly reduced in cells lacking endogenous Abcb7 and transfected with ABCB7Mut1 (Online Supplementary Figure S11A,B), which was unable to interact with Fech, indicating that formation of a functional ABCB7/FECH complex was required for FECH stability and completion of heme synthesis.

Discussion Loss-of-function mutations in ABCB7 cause XLSA with ataxia, a recessive disorder characterized by the presence in the patients’ bone marrow of nucleated erythroblasts that exhibit granules of iron accumulated in the mitochondria surrounding the nuclei. To gain insights into the primary effects of loss of ABCB7, we generated cell lines with inducible silencing of ABCB7. KD of ABCB7 elicited a dramatic loss of multiple mitochondrial Fe-S enzymes after only 3 days, whereas defects in cytosolic Fe-S proteins were not observed until 5 days after KD. Similarly, studies in Atm1-depleted yeast cells found a severe growth defect caused by mitochondrial dysfunction, which included loss of oxidative respiration and defective heme biosynthesis.35-37 A growing list of human diseases, including sideroblastic anemia, manifest severe mitochondrial iron overload,17 and many of these disorders affect the core components of the iron-sulfur cluster (ISC) machinery, including frataxin,38 glutaredoxin 522,39 and HSPA9.40 The molecular mechanism underlying the mitochondrial iron accumulation has not been unveiled. Studies in cell culture models have revealed that a feature of defects manifesting mitochondrial iron overload included the activation of the cytosolic iron starvation response,17,38,41 which increased the IRE-binding activities of IRP1 and IRP2, and upregulation of the ubiquitously expressed mitochondrial iron importer MFRN2.41,42 Our studies show that activation of IRP in the cytosol of cells depleted of ABCB7 and upregulation of the mitochondrial iron importers, MFRN1 and MFRN2, haematologica | 2019; 104(9)

A unique ABCB7-FECH-ABCB10 complex

occurred 3 days after KD of ABCB7 and accounted for the increased influx of iron in mitochondria. The decreased cytosolic aconitase activity in ABCB7-depleted cells was therefore likely the result of the conversion of IRP1 from cytosolic aconitase into IRE-binding apo-protein, rather than the result of impaired cytosolic ISC biogenesis, as previously proposed.11 Similarly, yeast cells depleted of Atm1 were reported to activate the iron regulon, which encodes the high affinity iron uptake system.43,44 Despite the large amount of iron imported in mitochondria, cells lacking Abcb7 showed a defect in heme biosynthesis, which resulted not only from translational repres-

sion of Alas2 by IRP, but also from the decreased stability of ferrochelatase. Recent studies proposed that in a mouse model of frataxin deficiency, mitochondrial iron overload was mediated by the upregulation of MFRN2, which was driven by defective heme biosynthesis, due to loss of FECH.45 Interestingly, mitochondrial iron accumulation has also been reported in erythroblasts from patients with erythropoietic protoporphyria due to decreased ferrochelatase levels,46 suggesting that loss of FECH and/or its product, heme, in cells with an intact IRE-IRP system may drive mitochondrial iron overload. The identification and characterization of proteinâ&#x20AC;&#x201C;pro-

Figure 6. Mutational analysis of the C-terminal domain of ABCB7 identified the region V450-D538 as being involved in binding ferrochelatase. (A) Primary sequence of the C-terminal domain of ABCB7. (B) Modeled crystal structure of ABCB7. The numbered peptide sequences in (A) and the corresponding colored regions in (B) refer to the amino acid residues that were subjected to alanine scanning mutagenesis to assess their involvement in interacting with FECH. (C-E) Immunoprecipitation (IP) of FLAG-tagged ABCB7 wildtype (B7) or the mutants, as indicated, expressed in G1E-ER4 cells that had been silenced for 3 days to knockdown the expression of endogenous Abcb7 and that co-expressed HA-tagged FECH. Mutants 1, 2, 3, 4 and 6 of ABCB7 [green peptides and domains in (A) and (B), respectively] showed significantly decreased binding to FECH and to Abcb10. (F-H) In vitro pull-down assays of 35S-labeled FLAG-tagged ABCB7 wildtype (B7) or mutants, as indicated, in the presence of 35Sâ&#x20AC;&#x201C;FECH (F) confirmed the results obtained in vivo (C-E) and demonstrated that binding of FECH to ABCB7 was through direct physical interaction. (C-E, n=6. F-H, n=4). See also Online Supplementary Figure S17 for densitometries of immunoblots and statistical analyses.

haematologica | 2019; 104(9)

1765

N. Maio et al.

Figure 7. Dimeric ferrochelatase bridges ABCB7 and ABCB10 homodimers in a molecular complex required for heme biosynthesis. A functional dimeric ferrochelatase (FECH) bridges ABCB7 and ABCB10 by binding near the nucleotide-binding domains (NDB) of each ATB-binding cassette (ABC) transporter. Two sequences in the C terminus of human ABCB7, regions V450-L463 and G527-D538, were essential for binding to FECH. Amino acid residues 90-115 of FECH, which formed the closing gate of the active site of the enzyme containing protoporphyrin IX (PPIX) were specifically involved in the interaction with ABCB7. The region between lysines 133 and 145 of FECH interacted with both ABCB7 and ABCB10. Likely the 133-145 region of one protomer of FECH interacted with ABCB7 and the corresponding region of the other protomer of the dimeric enzyme interacted with ABCB10. Interestingly, FECH interacted directly with regions of ABCB7 and ABCB10 near the NBD which are enriched in histidines, residues that are known heme-ligating amino acids (see also Online Supplementary Figure S12), raising the possibility that one or both ABC transporters may function as a mitochondrial matrix heme exporter.

tein interactions are fundamental to understanding the mechanisms and regulation of most biological processes, and interacting partners provide insights into biological functions that can be exploited for therapeutic purposes. We found that endogenous Abcb7 formed a complex with Fech and Abcb10 during erythroid differentiation. We performed XL-MS to investigate the architecture of the ABCB7/FECH complex, followed by extensive mutational analyses of ABCB7. Our studies identified two sequences in the C-terminal domain of ABCB7, residues V450-V504 and G527-D538, which were major molecular determinants of the interaction with FECH. Interestingly, these regions are adjacent to the Walker A motif of ABCB7, which is essential for nucleotide binding and transport activity.47 The overall architecture of the ABCB7/FECH/ABCB10 complex is shown in our proposed model (Figure 7), in which a functional dimeric FECH bridges ABCB7 and ABCB10 homodimers. Interestingly, residues 90-115 of FECH enclose the enzymatic active site containing PPIX and specifically bind ABCB7 near its nucleotide-binding domain. It is tempting to speculate that hydrolysis of ATP by ABCB7 may drive a conformational rearrangement on the region 90-115 of FECH, which would enable opening of the pocket and release of the newly synthesized proto-heme. A previous model was proposed based on the identification of a Fech/Abcb10/Mfrn1 complex,31 in which the interaction of Fech with Abcb10 and Mfrn1 was required to integrate mitochondrial iron import with its utilization for heme synthesis. In the same study,31 a separate interaction of Fech with Abcb7 was also reported. We did not identify Mfrn1 as part of the Abcb7/Fech/Abcb10 complex (Figure 5A and Online Supplementary Table S1).

1766

However, since Mfrn1 was reported to be the direct interacting partner of Abcb10,31 it is possible that in our coimmunoprecipitation experiments using an anti-Abcb7 antibody we mainly immunocaptured Abcb7 and its closest interacting partner, FECH, which in turn mediated an interaction with Abcb10; thus, we may not have detected other interactions of Abcb10 with Mfrn1 which are more physically distant from the Abcb7 partners. Nonetheless, Mfrn1 upregulation was essential to meet the exceptionally high iron demand for heme biosynthesis of erythroid cells during differentiation (Figure 4A-G), and decreased Mfrn1 half-life in the iron-deficient conditions generated by loss of Irp2 caused mitochondrial iron deficiency and impaired heme biosynthesis (Figure 4A-G). Our studies offer a potential molecular mechanism for reported cases of erythropoietic protoporphyria in patients harboring mutations between residues 68 and 220,48 which we found were involved in binding ABCB7. Overall, our studies highlight the importance of ABCB7 for mitochondrial function and provide the biochemical characterization of a functional complex formed of ABCB7, FECH and ABCB10, which is required for cellular iron homeostasis, mitochondrial function and heme biosynthesis. Our work suggests that more definitive experiments deploying coordinated activity of the entire complex may aid identification of its physiological substrates. Acknowledgments The authors thank Dr. Manik Ghosh for technical help, members of the Rouault laboratory for helpful discussion and the Eunice Kennedy Shriver NICHD Intramural Research Program for support.

haematologica | 2019; 104(9)

A unique ABCB7-FECH-ABCB10 complex

References 17. 1. Gbelska Y, Krijger JJ, Breunig KD. Evolution of gene families: the multidrug resistance transporter genes in five related yeast species. FEMS Yeast Res. 2006;6(3):345-355. 2. Robey RW, Pluchino KM, Hall MD, Fojo AT, Bates SE, Gottesman MM. Revisiting the role of ABC transporters in multidrug-resistant cancer. Nat Rev Cancer. 2018;18(7):452-464. 3. Srinivasan V, Pierik AJ, Lill R. Crystal structures of nucleotide-free and glutathionebound mitochondrial ABC transporter Atm1. Science. 2014;343(6175):1137-1140. 4. Lee JY, Yang JG, Zhitnitsky D, Lewinson O, Rees DC. Structural basis for heavy metal detoxification by an Atm1-type ABC exporter. Science. 2014;343(6175):11331136. 5. Zutz A, Gompf S, Schagger H, Tampe R. Mitochondrial ABC proteins in health and disease. Biochim Biophys Acta. 2009;1787(6):681-690. 6. Krishnamurthy PC, Du G, Fukuda Y, et al. Identification of a mammalian mitochondrial porphyrin transporter. Nature. 2006;443 (7111):586-589. 7. Tsuchida M, Emi Y, Kida Y, Sakaguchi M. Human ABC transporter isoform B6 (ABCB6) localizes primarily in the Golgi apparatus. Biochem Biophys Res Commun. 2008;369(2):369-375. 8. Kiss K, Brozik A, Kucsma N, et al. Shifting the paradigm: the putative mitochondrial protein ABCB6 resides in the lysosomes of cells and in the plasma membrane of erythrocytes. PLoS One. 2012;7(5):e37378. 9. Paterson JK, Shukla S, Black CM, et al. Human ABCB6 localizes to both the outer mitochondrial membrane and the plasma membrane. Biochemistry. 2007;46(33):94439452. 10. Savary S, Allikmets R, Denizot F, et al. Isolation and chromosomal mapping of a novel ATP-binding cassette transporter conserved in mouse and human. Genomics. 1997;41(2):275-278. 11. Pondarre C, Antiochos BB, Campagna DR, et al. The mitochondrial ATP-binding cassette transporter Abcb7 is essential in mice and participates in cytosolic iron-sulfur cluster biogenesis. Hum Mol Genet. 2006;15(6):953964. 12. Allikmets R, Raskind WH, Hutchinson A, Schueck ND, Dean M, Koeller DM. Mutation of a putative mitochondrial iron transporter gene (ABC7) in X-linked sideroblastic anemia and ataxia (XLSA/A). Hum Mol Genet. 1999;8(5):743-749. 13. Bekri S, Kispal G, Lange H, et al. Human ABC7 transporter: gene structure and mutation causing X-linked sideroblastic anemia with ataxia with disruption of cytosolic ironsulfur protein maturation. Blood. 2000;96(9):3256-3264. 14. Maguire A, Hellier K, Hammans S, May A. X-linked cerebellar ataxia and sideroblastic anaemia associated with a missense mutation in the ABC7 gene predicting V411L. Br J Haematol. 2001;115(4):910-917. 15. Pondarre C, Campagna DR, Antiochos B, Sikorski L, Mulhern H, Fleming MD. Abcb7, the gene responsible for X-linked sideroblastic anemia with ataxia, is essential for hematopoiesis. Blood. 2007;109(8):35673569. 16. Weiss MJ, Yu C, Orkin SH. Erythroid-cellspecific properties of transcription factor GATA-1 revealed by phenotypic rescue of a

haematologica | 2019; 104(9)

18.

19.

20.

21.

22.

23.

24.

25. 26.

27.

28. 29.

30.

31.

32.

33.

gene-targeted cell line. Mol Cell Biol. 1997;17(3):1642-1651. Rouault TA. Mitochondrial iron overload: causes and consequences. Curr Opin Genet Dev. 2016;38:31-37. Rouault TA, Maio N. Biogenesis and functions of mammalian iron-sulfur proteins in the regulation of iron homeostasis and pivotal metabolic pathways. J Biol Chem. 2017;292(31):12744-12753. Ryu MS, Zhang D, Protchenko O, ShakouryElizeh M, Philpott CC. PCBP1 and NCOA4 regulate erythroid iron storage and heme biosynthesis. J Clin Invest. 2017;127(5):17861797. Welch JJ, Watts JA, Vakoc CR, et al. Global regulation of erythroid gene expression by transcription factor GATA-1. Blood. 2004;104(10):3136-3147. Melefors O, Goossen B, Johansson HE, Stripecke R, Gray NK, Hentze MW. Translational control of 5-aminolevulinate synthase mRNA by iron-responsive elements in erythroid cells. J Biol Chem. 1993;268(8):5974-5978. Ye H, Jeong SY, Ghosh MC, et al. Glutaredoxin 5 deficiency causes sideroblastic anemia by specifically impairing heme biosynthesis and depleting cytosolic iron in human erythroblasts. J Clin Invest. 2010;120(5):1749-1761. Zenke-Kawasaki Y, Dohi Y, Katoh Y, et al. Heme induces ubiquitination and degradation of the transcription factor Bach1. Mol Cell Biol. 2007;27(19):6962-6971. Tahara T, Sun J, Igarashi K, Taketani S. Heme-dependent up-regulation of the alphaglobin gene expression by transcriptional repressor Bach1 in erythroid cells. Biochem Biophys Res Commun. 2004;324(1):77-85. Dailey HA, Meissner PN. Erythroid heme biosynthesis and its disorders. Cold Spring Harb Perspect Med. 2013;3(4):a011676. Gibson KD, Laver WG, Neuberger A. Initial stages in the biosynthesis of porphyrins. 2. The formation of delta-aminolaevulic acid from glycine and succinyl-coenzyme A by particles from chicken erythrocytes. Biochem J. 1958;70(1):71-81. Whitman JC, Paw BH, Chung J. The role of ClpX in erythropoietic protoporphyria. Hematol Transfus Cell Ther. 2018;40(2):182188. Chen JJ. Translational control by heme-regulated eIF2alpha kinase during erythropoiesis. Curr Opin Hematol. 2014;21(3):172-178. Cooperman SS, Meyron-Holtz EG, Olivierre-Wilson H, Ghosh MC, McConnell JP, Rouault TA. Microcytic anemia, erythropoietic protoporphyria, and neurodegeneration in mice with targeted deletion of ironregulatory protein 2. Blood. 2005;106(3): 1084-1091. Taketani S, Kakimoto K, Ueta H, Masaki R, Furukawa T. Involvement of ABC7 in the biosynthesis of heme in erythroid cells: interaction of ABC7 with ferrochelatase. Blood. 2003;101(8):3274-3280. Chen W, Dailey HA, Paw BH. Ferrochelatase forms an oligomeric complex with mitoferrin-1 and Abcb10 for erythroid heme biosynthesis. Blood. 2010;116(4):628-630. Medlock AE, Shiferaw MT, Marcero JR, et al. Identification of the mitochondrial heme metabolism complex. PLoS One. 2015;10(8): e0135896. Medlock AE, Dailey TA, Ross TA, Dailey HA, Lanzilotta WN. A pi-helix switch selective for porphyrin deprotonation and product release in human ferrochelatase. J Mol

Biol. 2007;373(4):1006-1016. 34. Medlock A, Swartz L, Dailey TA, Dailey HA, Lanzilotta WN. Substrate interactions with human ferrochelatase. Proc Natl Acad Sci U S A. 2007;104(6):1789-1793. 35. Kispal G, Csere P, Guiard B, Lill R. The ABC transporter Atm1p is required for mitochondrial iron homeostasis. FEBS Lett. 1997;418(3):346-350. 36. Leighton J, Schatz G. An ABC transporter in the mitochondrial inner membrane is required for normal growth of yeast. EMBO J. 1995;14(1):188-195. 37. Miao R, Kim H, Koppolu UM, Ellis EA, Scott RA, Lindahl PA. Biophysical characterization of the iron in mitochondria from Atm1pdepleted Saccharomyces cerevisiae. Biochemistry. 2009;48(40):9556-9568. 38. Puccio H, Simon D, Cossee M, et al. Mouse models for Friedreich ataxia exhibit cardiomyopathy, sensory nerve defect and Fe-S enzyme deficiency followed by intramitochondrial iron deposits. Nat Genet. 2001;27(2):181-186. 39. Camaschella C, Campanella A, De Falco L, et al. The human counterpart of zebrafish shiraz shows sideroblastic-like microcytic anemia and iron overload. Blood. 2007;110(4):1353-1358. 40. Schmitz-Abe K, Ciesielski SJ, Schmidt PJ, et al. Congenital sideroblastic anemia due to mutations in the mitochondrial HSP70 homologue HSPA9. Blood. 2015;126(25): 2734-2738. 41. Crooks DR, Jeong SY, Tong WH, et al. Tissue specificity of a human mitochondrial disease: differentiation-enhanced mis-splicing of the Fe-S scaffold gene ISCU renders patient cells more sensitive to oxidative stress in ISCU myopathy. J Biol Chem. 2012;287(48):40119-40130. 42. Martelli A, Puccio H. Dysregulation of cellular iron metabolism in Friedreich ataxia: from primary iron-sulfur cluster deficit to mitochondrial iron accumulation. Front Pharmacol. 2014;5:130. 43. Rutherford JC, Ojeda L, Balk J, Muhlenhoff U, Lill R, Winge DR. Activation of the iron regulon by the yeast Aft1/Aft2 transcription factors depends on mitochondrial but not cytosolic iron-sulfur protein biogenesis. J Biol Chem. 2005;280(11):10135-10540. 44. Hausmann A, Samans B, Lill R, Muhlenhoff U. Cellular and mitochondrial remodeling upon defects in iron-sulfur protein biogenesis. J Biol Chem. 2008;283(13):8318-8330. 45. Martelli A, Schmucker S, Reutenauer L, et al. Iron regulatory protein 1 sustains mitochondrial iron loading and function in frataxin deficiency. Cell Metab. 2015;21(2):311-323. 46. Rademakers LH, Koningsberger JC, Sorber CW, Baart de la Faille H, Van Hattum J, Marx JJ. Accumulation of iron in erythroblasts of patients with erythropoietic protoporphyria. Eur J Clin Invest. 1993;23(2):130-138. 47. Walker JE, Saraste M, Runswick MJ, Gay NJ. Distantly related sequences in the alpha- and beta-subunits of ATP synthase, myosin, kinases and other ATP-requiring enzymes and a common nucleotide binding fold. EMBO J. 1982;1(8):945-951. 48. Rufenacht UB, Gouya L, Schneider-Yin X, et al. Systematic analysis of molecular defects in the ferrochelatase gene from patients with erythropoietic protoporphyria. Am J Hum Genet. 1998;62(6):1341-1352. 49. Waterhouse A, Bertoni M, Bienert S, et al. SWISS-MODEL: homology modelling of protein structures and complexes. Nucleic Acids Res. 2018;46(W1):W296-W303.

1767

ARTICLE Ferrata Storti Foundation

Iron Metabolism & its Disorders

New thiazolidinones reduce iron overload in mouse models of hereditary hemochromatosis and β-thalassemia Jing Liu,1, 2,# Wei Liu,1,2,# Yin Liu,1,3 Yang Miao,1 Yifan Guo,1 Haoyang Song,1 Fudi Wang,4 Hongyu Zhou,5 Tomas Ganz,6 Bing Yan3,5 and Sijin Liu1, 2

State Key Laboratory of Environmental Chemistry and Ecotoxicology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing, China; 2University of Chinese Academy of Sciences, Beijing, China; 3School of Environmental Science and Engineering, Shandong University, Shandong, China; 4Department of Nutrition, Nutrition Discovery Innovation Center, Institute of Nutrition and Food Safety, School of Public Health, School of Medicine, Zhejiang University, Zhejiang, China; 5Key Laboratory for Water Quality and Conservation of the Pearl River Delta, Ministry of Education, Institute of Environmental Research at Greater Bay, Guangzhou University, Guangzhou, China and 6 Department of Medicine and Department of Pathology, David Geffen School of Medicine at University of California, California, Los Angeles, CA, USA 1

Haematologica 2019 Volume 104(9):1768-1781

JL and WL contributed equally to this work

ABSTRACT

G Correspondence: BING YAN drbingyan@yahoo.com SIJIN LIU sjliu@rcees.ac.cn Received: October 21, 2018. Accepted: February 15, 2019. Pre-published: February 21, 2019. doi:10.3324/haematol.2018.209874 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1768 ©2019 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.

1768

enetic iron-overload disorders, mainly hereditary hemochromatosis and untransfused β-thalassemia, affect a large population worldwide. The primary etiology of iron overload in these diseases is insufficient production of hepcidin by the liver, leading to excessive intestinal iron absorption and iron efflux from macrophages. Hepcidin agonists would therefore be expected to ameliorate iron overload in hereditary hemochromatosis and β-thalassemia. In the current study, we screened our synthetic library of 210 thiazolidinone compounds and identified three thiazolidinone compounds, 93, 156 and 165, which stimulated hepatic hepcidin production. In a hemochromatosis mouse model with hemochromatosis deficiency, the three compounds prevented the development of iron overload and elicited iron redistribution from the liver to the spleen. Moreover, these compounds also greatly ameliorated iron overload and mitigated ineffective erythropoiesis in β-thalassemic mice. Compounds 93, 156 and 165 acted by promoting SMAD1/5/8 signaling through differentially repressing ERK1/2 phosphorylation and decreasing transmembrane protease serine 6 activity. Additionally, compounds 93, 156 and 165 targeted erythroid regulators to strengthen hepcidin expression. Therefore, our hepcidin agonists induced hepcidin expression synergistically through a direct action on hepatocytes via SMAD1/5/8 signaling and an indirect action via eythroid cells. By increasing hepcidin production, thiazolidinone compounds may provide a useful alternative for the treatment of iron-overload disorders.

Introduction Hepcidin, produced by hepatocytes, is a 25-amino acid peptide hormone that plays a central role in systemic iron homeostasis. Hepcidin binds to its receptor, ferroportin, to induce ferroportin degradation, thereby decreasing iron efflux from macrophages and hepatocytes as well as intestinal iron absorption. Hepcidin concentration changes within a physiological range to orchestrate iron absorption, recycling and tissue distribution. However, pathological dysregulation of hepcidin causes diverse iron disorders. In particular, hepcidin deficiency results in iron-loading syndromes such as hereditary hemochromatosis (HH),1 β-thalassemia intermedia2 and other iron-loading anemias.3,4 Thus, enhancing hepcidin production with suitable agonists represents a promising strategy to prevent iron accumulation in HH, β-thalassemia and other iron-loading conditions.5,6 Type 1 HH, caused by mutations of the hemochromatosis (HFE) gene, is the most common form of HH in populations of northern European origin.7 β-thalassemia, haematologica | 2019; 104(9)

Thiazolidinones reduce iron overload

common worldwide in regions where malaria was historically endemic, is a genetic erythrocyte disorder characterized by ineffective erythropoiesis, anemia, and progressive iron overload.8 For both HH and β-thalassemia patients, long-term iron overload causes liver cirrhosis, cardiomyopathy, and endocrinopathies.7 Iron excess is currently managed by phlebotomy in HH and chelation in iron-loading anemias,9 but both treatments have serious limitations, including suboptimal compliance and secondary suppression of hepcidin, which results in a further increase of dietary iron uptake.7,10 Iron-chelating drugs can adversely affect ocular, auditory and renal functions11-13 and their administration can be burdensome, e.g., because of the short half-life of desferrioxamine.14,15 Other approaches, such as mini-hepcidin peptides, are still at the experimental stage.16,17 We therefore searched for hepcidin agonists with favorable characteristics for clinical applications. The thiazolidinone scaffold can be engineered to target diverse pathologies, with derivatives that inhibit tumor growth, repress viral replication and diminish inflammatory responses.18,19 A thiazolidinone derivative [(Z)-5-(4methoxybenzylidene) thiazolidine-2,4-dione] ameliorated liver injury and fibrogenesis,20 suggesting that this class of compounds could target hepatocytes. In the current study, we established a library of thiazolidinone derivatives to look for lead compounds that could increase hepcidin concentration. We report here the identification of three novel compounds that ameliorated iron overload in HH and βthalassemia mice by stimulating the hepatic production of hepcidin.

Methods Synthesis and characterization of thiazolidinone compounds Thiazolidinone compounds were synthesized using the combinatorial library synthesis approach, and a previously described overall synthesis route.21 Briefly, primary thioureas (b) were constructed by reacting aniline (a) with ammonium thiocyanate, in the presence of acid (Figure 1A). Thioureas (b) reacted with ethyl 2-chloroacetate to generate thiazolidinones (c) as a precipitate, which was filtered and washed with absolute ethanol to obtain the product. The final step of the reaction was carried out in piperidine and absolute ethanol at 60°C. Finally, approximately 95% of the product (d) was formed as a precipitate (Online Supplementary Figure S1).

Statistical analysis The differences between individual treated groups relative to the untreated control were assessed using independent t-tests. The significance of mean differences for two or more treatment groups relative to the untreated control was determined by one-way analysis of variance. Data are shown as the mean ± standard deviation (SD). Statistical significance was accepted when P<0.05.

Other experimental details are provided in the Online Supplementary Data.

Results Synthesis and characterization of the combinatorial thiazolidinone library To search for hepcidin agonists, we designed a thiazolidinone compound library by incorporating diverse R1 haematologica | 2019; 104(9)

and R2 groups on the thiazolidinone core.18 Following the procedure illustrated in Figure 1A, a combinatorial library of thiazolidinone compounds containing 210 members was synthesized (Online Supplementary Figure S1 and Online Supplementary Table S2), using protocols that we have previously reported.21 All compounds used for the animal experiments were then purified either by recrystallization or column chromatography to reach a purity ≥98% as measured by liquid chromatography with ultraviolet detection at 214 nm (LC/UV214), and their structures were characterized by 1H-nuclear magnetic resonance and high-resolution mass spectrometry (Online Supplementary Table S3).

Screening of thiazolidinone derivatives for hepcidin-stimulatory activity We performed high-throughput screening of the thiazolidinone library using a dual luciferase reporter system developed in the laboratory.22 As shown in Online Supplementary Figure S2, no significant cytotoxicity was detected at 10 μM or 50 μM for these thiazolidinone compounds in SMMC-7721 cells, a hepatocyte cell line used for hepcidin screening. Hence, 10 μM was chosen as the test concentration. Following treatment with thiazolidinone derivatives for 24 h, hepcidin-luciferase activity was measured. As shown in Figure 1B, of the 210 compounds tested, 42 compounds were identified to increase luciferase activity by >1.3 fold relative to the untreated control. Of these, 30 compounds increased luciferase activity by more than 1.5 fold: 12 (compounds 2, 15, 49, 53, 68, 93, 96, 139, 163, 165, 189 and 194) by approximately 2 fold, two (compounds 22 and 23) by 2.5 fold, and one (compound 3) by 3.5 fold, relative to untreated cells (Figure 1B). The 30 compounds that were found to increase luciferase activity by more than 1.5 fold were subsequently rescreened by quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) for stimulation of endogenous hepcidin expression in SMMC-7721 cells. As the qRT-PCR results revealed (Figure 1C), ten of the 30 compounds increased hepcidin mRNA expression by more than 1.5 fold, compared to untreated cells, consistent with the luciferase reporter results (Figure 1B). Hepcidin mRNA expression was increased by nearly 6 fold after treatment with compounds 48 and 165 for 24 h, and compound 69 enhanced hepcidin expression by more than 3 fold, compared to untreated cells (P<0.001) (Figure 1C). Compound 93 stimulated hepcidin expression by approximately 2.5 fold (P<0.05) (Figure 1C) and compounds 49, 53, 139, 140, 142 and 156 increased hepcidin expression by approximately 2 fold (P<0.05) (Figure 1C), compared to untreated cells. By contrast, 13 compounds were not found to alter endogenous hepcidin expression, while compounds 2, 5, 11, 23 and 112 elicited inhibition of endogenous hepcidin transcription (P<0.05) (Figure 1C). Accordingly, compounds 48, 49, 53, 69, 93, 139, 140, 142, 156 and 165 were selected for further assessment. To examine the hepcidin-stimulating activity of the ten potential agonists in vivo, we administered them by intraperitoneal injection to wildtype (Wt) Balb/C mice at a dose of 30 mg/kg body weight. As shown in Online Supplementary Figure S3A, compounds 93 and 156 significantly increased hepatic hepcidin mRNA expression by 1.8 fold, respectively, and 1.5 fold at 6 h following administration of the compound, with a concomitant reduction of serum iron levels (P<0.05) (Online Supplementary Figure 1769

J. Liu et al.

S3B) compared to untreated mice. Compounds 93 and 156 also consistently enhanced hepatic hepcidin levels (P<0.05) (Online Supplementary Figure S3A) and diminished serum iron concentrations in mice 24 h after administration (P<0.05) (Online Supplementary Figure S3B). Additionally, compounds 49, 93, 140, 156 and 165 increased hepcidin mRNA expression by more than 1.5 fold 24 h after administration, compared to the expression in untreated mice (P<0.05) (Online Supplementary Figure S3A). Serum hepcidin was significantly increased by approximately 1.5 fold, relative to that in untreated mice, at 6 h following administration of compounds 69, 93, 139, 156 and 165 (P<0.05) (Online Supplementary Figure S3B),

and remained higher than that in control mice at 24 h after treatment with compounds 93, 156 and 165, but not with compounds 69 and 139 (P<0.05) (Online Supplementary Figure S3B). However, only mice treated with compounds 93, 156 and 165 consistently showed concurrent reductions in serum iron (P<0.05) (Online Supplementary Figure S3B), while mice treated with compounds 49 and 140 did not display such a consistent effect (Online Supplementary Figure S3B). Moreover, compounds 93, 156 and 165 were found to increase hepatic hepcidin expression in mice overall in a dose-dependent manner, from 2, to 10 and 30 mg/kg body weight (P<0.05) (Online Supplementary Figure S4). Considering these results together, compounds 93,

1770

Figure 1. Screening of the thiazolidinone library for compounds stimulating hepcidin expression. (A) Synthesis of the thiazolidinone library. A total of 18 anilines and 20 aromatic aldehydes were used as reactants, and 210 thiazolidinone compounds were obtained with a purity greater than 95%, as determined by liquid chromatography/mass spectrometry. Reagents and conditions: (i) R1-NH2, + NH2SCN, H , H2O, 80°C; (ii) ethyl chloroacetate, NaOAC, EtOH, 60°C; (iii) aldehydes, piperdine, EtOH, 60°C. The bold letters (ad) delineate the synthesis procedure, as described in the Methods section: primary thioureas (b) were constructed by reacting aniline (a) with ammonium thiocyanate, in the presence of acid; thioureas (b) reacted with ethyl 2-chloroacetate to generate thiazolidinones (c) as a precipitate, which was filtered and washed with absolute ethanol to obtain the product; the final step of the reaction was carried out in piperidine and absolute ethanol at 60°C, and approximately 95% of the product (d) was formed as precipitate. (B) A heatmap diagram showing the average fold changes of hepcidin-promoter luciferase activity relative to that of untreated cells (n=4). (C) Endogenous hepcidin mRNA expression in SMMC-7721 cells upon administration of compounds at a concentration of 10 μM for 24 h (n=4). *P<0.05, #P<0.001, compared to untreated control (Ctrl).

haematologica | 2019; 104(9)

Thiazolidinones reduce iron overload

156 and 165 were selected for further investigation in the subsequent experiments.

Compounds 93, 156 and 165 altered iron distribution in wildtype mice As delineated above, after screening, three compounds, 93, 156 and 165, were selected for detailed examination. To determine whether these compounds altered iron dis-

tribution in vivo, we measured the time course of hepatic hepcidin and concomitant changes of iron levels in mice challenged by compounds 93, 156 and 165. The experimental design for the experiments with Wt mice is depicted in Figure 2A. For a short-term study, mice received a single intraperitoneal injection of each compound at a dose of 30 mg/kg body weight, and were then sacrificed at 6, 24, 48, 72 and 96 h. Compounds 93, 156 and 165 rap-

Figure 2. Testing thiazolidinone derivatives for their effects on hepatic hepcidin in wildtype mice. (A) Diagram of the experimental design. (B) Hepatic hepcidin mRNA, (C) serum hepcidin, (D) serum iron and (E) splenic iron in 8-week old Balb/C wildtype (Wt) mice treated with compounds 93, 156 and 165 at a dose of 30 mg/kg body weight at various time points (n=4-6). (F) Splenic iron shown by Perls Prussian blue staining (blue areas evidenced by arrows) of mice treated with compounds 93, 156 and 165 at a dose of 30 mg/kg body weight for 24 h and 12 days. Original magnification, Ă&#x2014;200. *P<0.05; #P<0.001, compared to untreated control (Ctrl).

haematologica | 2019; 104(9)

1771

J. Liu et al.

idly increased hepatic hepcidin mRNA at 6 h, and the peak was observed at 24 h for compound 93 and at 48 h for compounds 156 and 165 with >2-fold increases (P<0.05) (Figure 2B). Thereafter, hepcidin declined and returned to the baseline level at 96 h, suggesting that these compounds could be dosed every 3-4 days for long-term administration. Accordingly, serum hepcidin was increased at 6 h after administration of compounds 93, 156 and 165, and the peak was detected at 24 h for compounds 93 and 165 and at 48 h for compound 156 (P<0.05) (Figure 2C). In another experiment, mice were treated with the compounds at a dose of 30 mg/kg body weight every 3 days for a total of 8 or 12 days. In agreement with the previous observations, treatment with compounds 93, 156 and 165 led to persistent stimulation of hepcidin expression over the 12-day period, especially for compounds 93 and 165 with more than 3-fold induction of hepatic hepcidin mRNA 12 days after administration (P<0.05) (Figure 2B). In parallel to the mRNA changes, similar increases of serum hepcidin relative to the levels in untreated mice were demonstrated (P<0.05) (Figure 2C). Hepcidin induction resulted in reduced serum iron and increased splenic iron content (P<0.05) (Figure 2D, E). Iron staining of spleen sections confirmed the increase of splenic iron content, particularly in macrophages (in blue, indicated by arrows), compared to the content in untreated mice (Figure 2F). In addition, these compounds were evaluated in mice at a lower dose, 10 mg/kg body weight. As shown in Online Supplementary Figure S5, similar results were observed for these three compounds in modulating hepcidin expression and body iron redistribution. With respect to time of peak and duration of effect. these compounds were still active at the lower doses, but less potent than at the higher doses. To screen for potential toxicities of these compounds, liver, spleen, kidney, lung, heart and bone marrow specimens were subjected to histological analysis. No toxic changes were observed after 24 h and 12 days in any of these organs from mice challenged by compounds 93, 156 and 165 at a dose of 30 mg/kg body weight (Online Supplementary Figure S6). No impairment of spontaneous activities (e.g., feeding or movement) was observed. Moreover, no significant changes were found in serum aspartate aminotransferase, alanine aminotransferase or lactate dehydrogenase (Online Supplementary Figure S7A-C). Inflammation increases hepcidin expression,23 in large part through interleukin-6, so we explored the possibility that our compounds acted by increasing inflammatory mediators. As shown in Online Supplementary Figure S7D, serum interleukin-6 was not detectable in serum before or after compound administration for 24 h and 12 days. Bacterial lipopolysaccharide, at a dose of 5 mg/kg body weight, was used as a positive control stimulant of interleukin-6. Furthermore, hepatic inflammation was determined by measuring serum amyloid A1 (SAA1), a downstream target of interleukin-6, and a very sensitive marker of inflammation. Consistently, Saa1 mRNA levels were not significantly altered in livers from mice treated with these compounds (Online Supplementary Figure S7E), and neither was another inflammatory marker, serum tumor necrosis factor-a (Online Supplementary Figure S7F). In an analysis of the complete blood count, no increase of white blood cells was observed in peripheral blood 24 h following administration of a single dose or multiple doses for 12 days (Online Supplementary Figure S8). Collectively, our 1772

findings ruled out that compounds 93, 156 and 165 affect hepcidin through inflammatory mechanisms.

Compounds 93, 156 and 165 target SMAD1/5/8 signaling to promote hepcidin expression Iron/BMP6â&#x20AC;&#x201C;SMAD1/5/8 signaling controls hepcidin expression depending on iron status24,25 and we therefore examined whether our compounds modulated this key pathway. As shown in Figure 3A, the level of phosphorylated SMAD1/5/8 (P-SMAD1/5/8) was increased in livers of mice treated with the three compounds, with compound 156 causing the greatest increase in P-SMAD1/5/8. To interpret this finding, the upstream regulator of PSMAD1/5/8, transmembrane protease serine 6 (TMPRSS6), and its downstream target genes, inhibitor of DNA binding 1 (ID1) and SMAD family member 7 (SMAD7), were examined by qRT-PCR. The mRNA levels of Tmprss6 were suppressed by more than 50% in liver specimens from mice treated with the three compounds, relative to the levels in untreated mice (P<0.05) (Figure 3B). In contrast, Id1 and Smad7 were induced by approximately 1.5 to 2.5 fold in livers from treated mice, relative to the levels in the untreated controls (P<0.05) (Figure 3B). Transferrin receptor 2 (TFR2), HFE, hemojuvelin, TMPRSS6 and bone morphogenetic protein (BMP) receptors interact to activate hepcidin expression by upregulating SMAD1/5/8 signaling. The BMP/BMP receptor interaction enhances SAMD1/5/8 phosphorylation and TMPRSS6 downregulates hepcidin expression by cleaving hemojuvelin and other proteins in the complex.26,27 Consistent with the mRNA changes, the protein levels of TMPRSS6 were reduced in treated mice relative to those in untreated controls (Figure 3A). Additionally, phosphorylated ERK1/2 (P-ERK1/2) was recently found to repress hepcidin expression by suppressing SMAD1/5/8 phosphorylation.28 Here, the protein value of P-ERK1/2 was also diminished upon treatment with the compounds (Figure 3A). These results collectively suggest that compounds 93, 156 and 165 increased hepcidin expression by suppressing TMPRSS6 and ERK1/2 and thereby decreasing their inhibitory effects on SMAD1/5/8 phosphorylation. We next investigated these effects using hepatocytic cell lines, murine Hepa 1-6. Consistent with the in vivo results, compounds 93, 156 and 165 greatly induced hepcidin expression in Hepa 1-6 cells (P<0.05) (Figure 3C) and increased the expression levels of its downstream targets, Id1 and Smad7 (P<0.05) (Figure 3D). Mechanistically, the compounds suppressed ERK1/2 phosphorylation and decreased TMPRSS6 concentration (Figure 3E). Of note, these three compounds differentially suppressed PERK1/2 versus TMPRSS6 protein concentrations, suggesting that they may differ in their mechanisms of effect on SMAD1/5/8 signaling (Figure 3E).

Compounds 93, 156 and 165 target erythroid regulators to strengthen hepcidin Given that erythropoietin and erythropoiesis factors, growth differentiation factor 15 (GDF15), twisted gastrulation BMP signaling modulator 1 (TWSG1) and erythroferrone (ERFE), are also involved in regulating hepcidin expression,29-32 we examined these regulators. As shown in Online Supplementary Figure S9, serum erythropoietin level was unchanged in mice after treatment with the compounds for 24 h and 48 h. However, the expression levels of Erfe, Gdf15 and Twsg1 were significantly repressed in haematologica | 2019; 104(9)

Thiazolidinones reduce iron overload

bone marrow cells from Wt Balb/C mice after the administration of the compounds (P<0.05) (Figure 4A). These results suggest that the investigated compounds might also target Erfe, Gdf15 and Twsg1 in bone marrow erythroid cells to promote hepatic hepcidin expression, and also imply that erythropoietin may not be the only regulator of Erfe, Gdf15 and Twsg1. Subsequently, we analyzed the levels of Gdf15, Twsg1 and Erfe mRNA expression in bone marrow cells from Hbbth3/+ mice through qRT-PCR. As shown in Figure 4B, Gdf15, Twsg1 and Erfe mRNA lev-

els were significantly diminished (P<0.05), in line with the results observed in Wt Balb/C mice (Figure 4A). These results demonstrated that the agonists also targeted erythroid regulators to elevate hepcidin expression, even though the precise molecular mechanisms warrant further detailed investigation.

Hepcidin deficient (Hamp1-/-) mice were resistant to iron changes induced by compounds 93, 156 and 165 To verify that hepatic hepcidin is the iron-relevant target

E Figure 3. Compounds 93, 156 and 165 targeted SMAD1/5/8 signaling. (A) PSMAD1/5/8, P-ERK1/2 and TMPRSS6 levels determined by western blot analysis of liver specimens from 8-week old Balb/C mice 24 h after administration of compounds 93, 156 and 165 at a dose of 30 mg/kg body weight. (B) Changes of TMPRSS6 and downstream target genes of P-SMAD1/5/8 signaling: Id1 and Smad7 determined by quantitative reverse transcriptase polymerase chain reaction analysis (qRT-PCR) (n=4-6) in liver specimens of these mice. (C) Changes of hepcidin mRNA in Hepa 1-6 cells at the indicated times after treatment with compounds 93, 156 and 165 at a concentration of 10 μM (n=4-6). (D) Changes of Id1 and Samd7 were determined by qRT-PCR analysis (n=4) of Hepa 1-6 cells 24 h after treatment with compounds 93, 156 and 165 (10 μM). (E) Variations of P-SAMD1/5/8, SMAD1, P-ERK1/2, ERK1/2 and TMPRSS6 levels analyzed by western blot in Hepa 1-6 cells 24 h after treatment with compounds 93, 156 and 165 at a concentration of 10 μM. *P<0.05; #P<0.001, relative to untreated control (Ctrl).

haematologica | 2019; 104(9)

1773

J. Liu et al.

of the investigated compounds, we examined the changes of iron indices in a hepcidin-deficient mouse model, Hamp1-/- mice, as previously described.33 Hamp1-/- mice do not produce functional hepcidin, and develop severe iron overload after weaning.34 We iron-depleted these mice on a low-iron diet (4 ppm) for 4 weeks according to an established protocol,35 so as to increase the sensitivity of Hamp1/mice to any hepcidin-independent serum iron-lowering effects of the compounds (time line in Online Supplementary Figure S10A). As shown in Online Supplementary Figure S10B-D, serum iron, hepatic and splenic iron levels were not significantly altered in Hamp1-/- mice, 6 or 24h following the administration of compounds 93, 156 and 165 at a dose of 30 mg/kg body weight, relative to the levels in untreated Hamp1-/- mice (P>0.05). These data indicate that

compounds 93, 156 and 165 predominantly target liver hepcidin in their modulation of iron homeostasis.

Compounds 93, 156 and 165 prevented iron overload in HFE-deficient (Hfe-/-) mice The potential therapeutic effect of compounds 93, 156 and 165 was tested in iron overloaded Hfe-/- mice on a normal diet. Consistent with previous reports, serum hepcidin concentration in Hfe-/- mice progressively increased from 6 to 8 weeks (P<0.05) (Figure 5A), as the mice became iron-overloaded,36,37 so that the hepcidin level in Hfe-/- mice at the age of 10 weeks was comparable to that of Wt mice (P<0.05) (Figure 5A). Other iron parameters in Hfe-/- mice indicated that by week 10 rising hepatic iron caused hepcidin upregulation and accelerated iron accu-

Figure 4. Compounds 93, 156 and 165 regulated Gdf15, Twsg1 and Erfe in bone marrow cells. (A) Changes of Gdf15, Twsg1 and Erfe mRNA expression in bone marrow cells from wildtype (Wt) Balb/C mice upon administration of the various compounds at a dose of 30 mg/kg for 24 h. (B) The variations of Gdf15, Twsg1 and Erfe mRNA expression in bone marrow cells from Hbbth3/+ mice treated with the various compounds at a dose of 30 mg/kg body weight for 24 h. *P<0.05; # P<0.001, relative to untreated control (Ctrl).

1774

haematologica | 2019; 104(9)

Thiazolidinones reduce iron overload

mulation in spleens compared to these processes in normal mice (P<0.05) (Online Supplementary Figure S11). Hfe-/- mice (5 weeks old) on a normal diet were treated with compounds 93, 156 and 165 at a dose of 10 mg/kg body weight every other day, and were sacrificed 2 weeks later (Figure 5B). Serum hepcidin was increased 2 fold after

administration of compound 93 and 2.3 fold after administration of compounds 156 and 165, relative to the levels in untreated Hfe-/- mice (P<0.05) (Figure 5C). As an expected consequence of increased hepcidin, serum iron and hepatic iron were reduced (P<0.05) (Figure 5D, E), with a concomitant increase of splenic iron (P<0.05) (Online

Figure 5. Treatment with compounds 93, 156 and 165 redistributed iron in Hfe-/- mice. (A) Serum hepcidin in wildtype (Wt) 129S and Hfe-/- mice at different ages. (B) The experimental scheme. Changes of (C) serum hepcidin, (D) serum iron and (E) hepatic iron in Hfe-/- mice after treatment with compounds 93, 156 and 165 at a dose of 10 mg/kg body weight for 2 weeks (n=4-6). (F) Perls Prussian blue staining of liver and spleen (in blue, indicated by arrows) and 3'-diaminobenzidineenhanced Perls Prussion staining of duodenum (in brown) of 5-week-old Hfe-/- mice treated with compounds 93, 156 and 165 at a dose of 10 mg/kg body weight every other day for 2 weeks. Original magnification, Ă&#x2014;200 for spleen sections; Ă&#x2014;400 for liver and duodenum sections. *P<0.05; #P<0.001, compared to the untreated, control (Ctrl) Hfe-/- mice.

haematologica | 2019; 104(9)

1775

J. Liu et al.

Supplementary Figure S12A). These results were supported by liver and spleen iron staining (Figure 5F). Moreover, iron staining demonstrated increased iron accumulation in the duodenum of mice treated with the compounds (Figure 5F), suggesting inhibition of iron transfer from the intestine to plasma due to increased hepcidin driven by

these compounds. As further evidence of iron redistribution, transferrin saturation and serum ferritin were reduced in mice treated with the compounds (P<0.05) (Online Supplementary Figure S12B,C), and serum transferrin was elevated (P<0.05) (Online Supplementary Figure S12D). Serum interleukin-6, aspartate aminotransferase,

Figure 6. Compound administration to iron-depleted Hfe-/- mice. (A) The experimental design of treatment of iron-depleted Hfe-/- mice with compounds 93, 156 and 165. Changes in (B) serum hepcidin, (C) serum iron, (D) splenic iron and (E) hepatic iron of 9-week old Hfe-/- mice with iron depletion for 3 weeks prior to the administration of compounds 93, 156 and 165 at a dose of 30 mg/kg body weight for another 2 weeks (n=4-6). (F) Tissue iron staining of liver and spleen sections with Perls Prussian blue (in blue, indicated by arrows) and duodenal sections with 3'-diaminobenzidine-enhanced Perls stain (in brown). Original magnification, Ă&#x2014;200 for spleen, and Ă&#x2014;400 for liver and duodenum. *P<0.05; #P<0.001, relative to untreated control (Ctrl) mice.

1776

haematologica | 2019; 104(9)

Thiazolidinones reduce iron overload

alanine aminotransferase and lactate dehydrogenase did not exhibit significant changes (P>0.05) (Online Supplementary Figure S13). We then extended the animal studies for a longer time period, 4 weeks. As shown in Online Supplementary Figure S14A, serum hepcidin concentration was significantly increased by 2.5 fold in Hfe-/- mice after administration of compound 93 at a dose of 10 mg/kg body weight for 4 weeks, compared to the level in untreated Hfe-/- mice (P<0.05). In response to the change in serum hepcidin concentration, serum iron and hepatic iron were reduced by ~20% and ~30%, respectively (P<0.05) (Online Supplementary Figure S14B,C). As a consequence, splenic iron content was increased by ~40% in Hfe-/- mice after treatment with compound 93 relative to the iron content in untreated controls (P<0.05) (Online Supplementary Figure S14D). Tissue iron staining further supported the results of the liver and splenic iron measurements (Online Supplementary Figure S14E). Additionally, no significant increases of serum interleukin-6, aspartate aminotransferase, or alanine aminotransferase levels were found in mice after 4 weeks of treatment with compound 93, ruling out the occurrence of inflammation and hepatic injury (Online Supplementary Figure S15). To model the condition of patients undergoing iron removal, we fed 9-week old male Hfe-/- mice with a low iron diet (4 ppm) for 3 weeks. Immediately after this pretreatment, these Hfe-/- mice on a normal diet were treated with compounds 93, 156 and 165 at a dose of 30 mg/kg body weight every 3 days for 2 weeks (experimental scheme in Figure 6A). After treatment with compounds 93, 156 and 165, the serum hepcidin concentration was increased by approximately 2 fold, relative to that of untreated Hfe-/- mice (P<0.05) (Figure 6B). Consequently, serum iron and transferrin saturation were significantly reduced (P<0.05) (Figure 6C and Online Supplementary Figure S16A), and serum transferrin and splenic iron were increased (P<0.05) (Online Supplementary Figure S16B and Figure 6D), associated with decreased hepatic iron content (P<0.05) (Figure 6E). The tissue iron changes were further confirmed by iron staining, and we also noted increased iron concentration in the duodenum, indicative of inhibition of the transfer of absorbed iron into plasma (Figure 6F). Serum interleukin-6, aspartate aminotransferase and alanine aminotransferase were not significantly altered in Hfe-/- mice by the treatments (Online Supplementary Figure S17).

Compounds 93, 156 and 165 ameliorated iron overload and mitigated ineffective erythropoiesis in Hbbth3/+ mice

Even untransfused patients with β-thalassemia suffer from iron overload and associated organ damage due to inappropriately low production of hepcidin. We therefore tested our compounds in a mouse model of thalassemia intermedia, Hbbth3/+ mice,38 given a 2-week course of treatment. Serum hepcidin was significantly increased by ~45% in Hbbth3/+ mice treated with compounds 93, 156 and 165, relative to the levels in untreated mice (P<0.05) (Figure 7A). As a consequence, the serum iron concentration dropped by ~30% (P<0.05) (Figure 7B). Similarly to what occurred in Hfe-/- mice, the serum ferritin concentration dropped ~28% (P<0.05) (Figure 7C), indicating that treatment with these agonists greatly diminished hyperferritinemia in Hfe-/- and Hbbth3/+ mice, and the iron content haematologica | 2019; 104(9)

in the liver and spleen was reduced by >30% and 35%~50%, respectively (P<0.05) (Figure 7D,E), indicating an effective relief of iron overload in Hbbth3/+ mice by these compounds. Spleen and liver iron staining supported these results (Figure 7F). In β-thalassemia, deficiency of β-globin causes an imbalance between a- and β-globin, so that excess a-globin tetramers aggregate in erythroblasts, leading to apoptosis of orthochromatic erythroblasts and reactive increases in earlier erythroblast forms. Thus, β-thalassemia syndromes often manifest with severe anemia with ineffective erythropoiesis. To understand the effect of our compounds on erythroblast maturation, we examined erythropoiesis in Hbbth3/+ mice. As shown in Figure 8A, the hemoglobin level was elevated by ~15% by these compounds (P<0.05). Additionally, red blood cell content was increased ~10% upon treatment with the compounds compared to the erythrocyte content in untreated controls (P<0.05) (Figure 8B). Furthermore, blood smears revealed an increased number of red blood cells with normal morphology and a decrease of damaged or deformed erythrocytes in treated mice (Figure 8C, denoted by arrows). To corroborate these findings, flow cytometry analysis was performed to define erythroid populations using the erythroid markers TER119 and CD44 and cellular size. As shown in Figure 8D, compounds 93, 156 and 165 increased the percentage of the P7 subpopulation (indicative of mature red blood cells) in bone marrow by 1.2 fold, 1.1 fold and 1.2 fold, respectively, with corresponding declines in the proportions of P6 subpopulations, compared to those in untreated controls. Similar findings were observed for erythropoiesis in spleens from treated mice, as these compounds increased the percentage of the P7 subpopulation in the spleen by 1.4 fold, 1.5 fold and 1.3 fold, respectively, with corresponding declines in the proportions of P5 subpopulations, compared to those in untreated controls (Online Supplementary Figure S18). In addition, we explored the consequences of a longer treatment, for 4 weeks. As shown in Online Supplementary Figure S19A, administration of compound 93 resulted in an approximately 2-fold increase of serum hepcidin in Hbbth3/+ mice relative to the level in untreated Hbbth3/+ mice (P<0.05). As a result, serum iron, hepatic iron and splenic iron were significantly reduced by 20-25% (P<0.05) (Online Supplementary Figure S19B-D). Iron staining supported the changes found in liver and spleen iron (Online Supplementary Figure S19E). There was also a significant reduction of spleen size and weight in Hbbth3/+ mice treated with compound 93 (P<0.05) (Online Supplementary Figure S19F,G), suggestive of an improvement in the previously ineffective erythropoiesis. Furthermore, increases in hemoglobin concentration and red blood cell count were found in Hbbth3/+ mice 4 weeks after administration of compound 93, in comparison to those in untreated Hbbth3/+ mice (P<0.05) (Online Supplementary Figure S20A,B). The findings of blood smears and analysis of representative erythroid populations of bone marrow and spleen were also consistent with the increasing concentration of hemoglobin and red blood cell content (Online Supplementary Figure S20C,D). Additionally, we measured the concentration of malondialdehyde, which is indicative of oxidative stress damage, such as injury by reactive oxygen species,39,40 and of erythropoietin. As show in Online Supplementary Figure S21, the malondialdehyde 1777

J. Liu et al.

level in the spleen was reduced by more than 20% in mice upon treatment with compound 93 for 4 weeks (P<0.001). In the meantime, the serum erythropoietin level remained unchanged upon treatment with compound 93 (Online Supplementary Figure S22). Our data

indicated that all three compounds significantly improved ineffective erythropoiesis in the bone marrow and spleen. Considered together, these results demonstrated that our compounds effectively relieved anemia in Hbbth3/+ mice by correcting ineffective erythropoiesis.

Figure 7. Compounds 93, 156 and 165 alleviated iron overload in Hbbth3/+ mice. (A) Serum hepcidin, (B) serum iron, (C) serum ferritin (D) hepatic iron, (E) splenic iron and (F) 3'-diaminobenzidine-enhanced Perls iron staining of liver sections (in brown) and Perls Prussion staining of spleen sections (in blue, indicated by arrows) after administration of compounds 93, 156 and 165 to Hbbth3/+ mice at a dose of 10 mg/kg body weight every other day for 2 weeks (n=4-6). Original magnification, Ă&#x2014;400. *P<0.05, relative to untreated, control (Ctrl) mice.

1778

haematologica | 2019; 104(9)

Thiazolidinones reduce iron overload

Discussion

pathologies are different.41,42 The common mechanism is lower level of hepcidin compared to that in Wt mice, giving rise to enhanced dietary iron uptake and iron egress out of macrophages and consequently iron deposition in various organs, especially the liver.36,43 However in β-thalassemias, unlike in HH, ineffective erythropoiesis and hemolysis accompany the iron overload, leading to enhanced erythrophagocytosis by macrophages in the spleen and consequently excess iron deposition in splenic macrophages.41 Upon treatment to relieve ineffective erythropoiesis, iron utilization for hemoglobin formation and erythropoiesis would be enhanced, resulting in reduced splenic iron overload.5,44 Our data showed significant reduction of splenic iron and spleen size/weight as well as improvement of hemoglobin and red blood cell indices in Hbbth3/+ mice after treatment with the compounds, demon-

We report here that thiazolidinone derivatives 93, 165 and 156 manifested therapeutically important hepcidinstimulatory activity, and the stimulatory effect on hepcidin synthesis lasted for 3 to 4 days after a single dose in mice. No gross toxicity, systemic or hepatic inflammation, or histologically-detectable tissue toxicity was found following acute or longer-term administration. Importantly, the three compounds greatly increased hepcidin concentration and relieved or prevented iron overload in two mouse models, Hfe-/- mice (representative of type 1 HH) and β-thalassemia intermedia mice, and prevented excessive iron deposition in organs of young Hfe-/- mice. Although Hfe-/- and β-thalassemic mice both exhibit a similar iron overload phenotype, the underlying molecular

Figure 8. Compounds 93, 156 and 165 improved ineffective erythropoiesis in Hbbth3/+ mice. (A) Hemoglobin (HGB) content and (B) red blood cell (RBC) count in peripheral blood samples from 4-week old Hbbth3/+ mice following administration of compounds 93, 156 and 165 at a dose of 10 mg/kg body weight every other day for 2 weeks (n=3-4). (C) Blood smears (original magnification, ×1,000) with damaged or deformed erythrocytes indicated by arrows and (D) representative erythropoiesis profiles of bone marrow cells from these mice (n=3-4). *P<0.05, compared to untreated, control (Ctrl) Hbbth3/+ mice.

haematologica | 2019; 104(9)

1779

J. Liu et al. strating that anemia was markedly ameliorated in β-thalassemia intermedia mice by our compounds through improving ineffective erythropoiesis. Collectively, these data support the potential of compounds 93, 165 and 156 as leads for treating complex iron overload disorders. Our compounds have a distinct therapeutic profile from other hepcidin peptide mimics35 and chemical agonists.45 Hepcidin mimics, mini-hepcidin peptides, prevented iron overload in various mouse models, including Hamp1-/mice,16 and also improved anemia and iron overload in βthalassemic mice by improving ineffective erythropoiesis.17 Compared to these or full-length hepcidin, our compounds are much cheaper to synthesize and may be more readily modified for oral administration. Moreover, the development of hepcidin mimics is still in the experimental stage.16,17 Tmprss6 siRNA and Tmprss6 antisense oligonucleotides also diminished iron overload in Hfe-/- mice and improved ineffective erythropoiesis in β-thalassemic mice,46,47 but face challenges of cost of synthesis and lack of experience with the long-term administration of these classes of compounds. Only a few small molecular agonist candidates have been identified. Although genistein was found to stimulate hepcidin expression,48 it is a compound with a broad array of other activities. Our recent study demonstrated that two natural compounds, icariin and epimedin C, purified from Chinese medicinal plants, stimulated hepcidin expression.33 However, their purification and production on a large scale would be formidable. As previously demonstrated, SMAD1/5/8 signaling fundamentally determines baseline hepcidin expression under normal conditions.49,50 TMPRSS6 is crucially implicated in interactions with TFR2, HFE, hemojuvelin and BMP receptors in the suppression of hepcidin expression.51-54 Moreover, ERK1/2 phosphorylation was recently found to repress hepcidin expression by suppressing SMAD1/5/8 phosphorylation.28 In the current study, we discovered that hepcidin was upregulated at the transcriptional level through enhancement of SMAD1/5/8 phosphorylation as a result of the reduction of TMPRSS6, and we also

References 1. Vujic Spasic M, Kiss J, Herrmann T, et al. Physiologic systemic iron metabolism in mice deficient for duodenal Hfe. Blood. 2007;109(10):4511-4517. 2. Nemeth E. Hepcidin and beta-thalassemia major. Blood. 2013;122(1):3-4. 3. Yilmaz Keskin E, Yenicesu I. Iron-refractory iron deficiency anemia. Turk J Haematol. 2015;32(1):1-14. 4. Nemeth E, Ganz T. Anemia of inflammation. Hematol Oncol Clin North Am. 2014;28(4):671-681. 5. Gardenghi S, Ramos P, Marongiu MF, et al. Hepcidin as a therapeutic tool to limit iron overload and improve anemia in beta-thalassemic mice. J Clin Invest. 2010;120 (12):4466-4477. 6. Ganz T, Nemeth E. The HepcidinFerroportin system as a therapeutic target in anemias and iron overload disorders. Hematology Am Soc Hematol Educ Program. 2011;538-542.

1780

observed reduced ERK phosphorylation in response to these hepcidin agonists. Thus, the reduction in TMPRSS6 and the diminished ERK1/2 phosphorylation presumably cooperated to increase hepcidin level, although their relative contributions are not known thus far. However, no connection between P-ERK and TMPRSS6 was found in current literature. Nevertheless, more work should be done to elucidate the possible relationship between P-ERK and TMPRSS6 through SMAD1/5/8 signaling in fine-tuning hepcidin expression. Erfe, Gdf15 and Twsg1 are stress erythropoiesis-responsive genes that suppress hepcidin expression, presumably to mobilize sufficient iron in support of erythropoiesis under various stresses associated with blood loss.32,55 In the present study, the expression of Erfe, Gdf15 and Twsg1 was significantly repressed in bone marrow cells from Wt and Hbbth3/+ mice responding to the administration of the investigated compounds. These results suggest that our compounds also repressed Erfe, Gdf15 and Twsg1 to promote hepatic hepcidin expression. We conclude that our hepcidin agonists induced hepcidin expression synergistically by a direct action on hepatocytes through SMAD1/5/8 signaling and by an indirect effect through erythroid cells. To summarize, we here identified three promising hepcidin agonists, compounds 93, 156 and 165. Although detailed pharmacological studies still need to be performed, the overall performance of these compounds to date warrants their further development for the treatment of HH, β-thalassemia and likely other iron disorders. Acknowledgments This work was supported by grants from the National Natural Science Foundation of China (grant numbers: 21425731, 21637004, 91543204, 91643204 and 21621064), the Strategic Priority Research Program of the Chinese Academy of Sciences (grant number: XDB14000000), National Key Research and Development Program of China (grant number: 2016YFA0203103) and a grant under the national “973” program (grant number: 2014CB932000).

7. Powell LW, Seckington RC, Deugnier Y. Haemochromatosis. Lancet. 2016;388 (10045):706-716. 8. Galanello R, Origa R. Beta-thalassemia. Orphanet J Rare Dis. 2010;5:11. 9. Kanwar P, Kowdley KV. Diagnosis and treatment of hereditary hemochromatosis: an update. Expert Rev Gastroenterol Hepatol. 2013;7(6):517-530. 10. Origa R. Beta-thalassemia. In: Pagon RA, Adam MP, Ardinger HH, Wallace SE, Amemiya A, Bean LJH, et al., eds. GeneReviews(R). Seattle (WA), 1993-2016. 11. Rahi AH, Hungerford JL, Ahmed AI. Ocular toxicity of desferrioxamine: light microscopic histochemical and ultrastructural findings. Br J Ophthalmol. 1986;70(5):373-381. 12. Cases A, Kelly J, Sabater J, et al. Acute visual and auditory neurotoxicity in patients with end-stage renal disease receiving desferrioxamine. Clin Nephrol. 1988;29(4):176-178. 13. Kontoghiorghes GJ. Deferasirox: uncertain future following renal failure fatalities, agranulocytosis and other toxicities. Expert Opin Drug Saf. 2007;6(3):235-239.

14. Lee P, Mohammed N, Marshall L, et al. Intravenous infusion pharmacokinetics of desferrioxamine in thalassaemic patients. Drug Metab Dispos. 1993;21(4):640-644. 15. Cappellini MD, Musallam KM, Taher AT. Overview of iron chelation therapy with desferrioxamine and deferiprone. Hemoglobin. 2009;33 (Suppl 1):S58-69. 16. Ramos E, Ruchala P, Goodnough JB, et al. Minihepcidins prevent iron overload in a hepcidin-deficient mouse model of severe hemochromatosis. Blood. 2012;120(18): 3829-3836. 17. Casu C, Oikonomidou PR, Chen HY, et al. Minihepcidin peptides as disease modifiers in mice affected by beta-thalassemia and polycythemia vera. Blood. 2016;128(2):265276. 18. Zhang Q, Zhou HY, Zhai SM, Yan B. Natural product-inspired synthesis of thiazolidine and thiazolidinone compounds and their anticancer activities. Curr Pharm Des. 2010;16(16):1826-1842. 19. Jain AK, Vaidya A, Ravichandran V, Kashaw SK, Agrawal RK. Recent developments and

haematologica | 2019; 104(9)

Thiazolidinones reduce iron overload

20.

21.

22.

23.

24. 25.

26. 27.

28.

29. 30.

31.

biological activities of thiazolidinone derivatives: a review. Bioorg Med Chem. 2012;20(11):3378-3395. Chen ZZ, Wang ZL, Deng CY, et al. (Z)-5-(4methoxybenzylidene)thiazolidine-2,4-dione protects rats from carbon tetrachlorideinduced liver injury and fibrogenesis. World J Gastroenterol. 2012;18(7):654-661. Zhou H, Wu S, Zhai S, et al. Design, synthesis, cytoselective toxicity, structure-activity relationships, and pharmacophore of thiazolidinone derivatives targeting drug-resistant lung cancer cells. J Med Chem. 2008;51(5): 1242-1251. Hou YL, Zhang SP, Wang L, et al. Estrogen regulates iron homeostasis through governing hepatic hepcidin expression via an estrogen response element. Gene. 2012;511(2): 398-403. 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. Cherayil BJ. Pathophysiology of iron homeostasis during inflammatory states. J Pediatr. 2015;167(4 Suppl):S15-19. Schmidt PJ, Fleming MD. Modulation of hepcidin as therapy for primary and secondary iron overload disorders preclinical models and approaches. Hematol Oncol Clin North Am. 2014;28(2):387-401. Fung E, Nemeth E. Manipulation of the hepcidin pathway for therapeutic purposes. Haematologica. 2013;98(11):1667-1676. Wahedi M, Wortham AM, Kleven MD, et al. Matriptase-2 suppresses hepcidin expression by cleaving multiple components of the hepcidin induction pathway. J Biol Chem. 2017;292(44):18354-18371. Chen H, Choesang T, Li H, et al. Increased hepcidin in transferrin-treated thalassemic mice correlates with increased liver BMP2 expression and decreased hepatocyte ERK activation. Haematologica. 2016;101(3):297308. Ganz T. Erythropoietic regulators of iron metabolism. Free Radic Biol Med. 2018;pii: S0891-5849(18)31188-31192. Kautz L, Jung G, Valore EV, Rivella S, Nemeth E, Ganz T. Identification of erythroferrone as an erythroid regulator of iron metabolism. Nat Genet. 2014;46(7):678-684. Tanno T, Noel P, Miller JL. Growth differentiation factor 15 in erythroid health and dis-

haematologica | 2019; 104(9)

32.

33.

34.

35.

36.

37.

38.

39. 40.

41.

42.

43.

ease. Curr Opin Hematol. 2010;17(3):184190. Tanno T, Porayette P, Sripichai O, et al. Identification of TWSG1 as a second novel erythroid regulator of hepcidin expression in murine and human cells. Blood. 2009;114 (1):181-186. Zhang M, Liu J, Guo WL, Liu X, Liu SJ, Yin HJ. Icariin regulates systemic iron metabolism by increasing hepatic hepcidin expression through Stat3 and Smad1/5/8 signaling. Int J Mol Med. 2016;37(5):1379-1388. Lesbordes-Brion JC, Viatte L, Bennoun M, et al. Targeted disruption of the hepcidin 1 gene results in severe hemochromatosis. Blood. 2006;108(4):1402-1405. Preza GC, Ruchala P, Pinon R, et al. Minihepcidins are rationally designed small peptides that mimic hepcidin activity in mice and may be useful for the treatment of iron overload. J Clin Invest. 2011;121 (12):4880-4888. Ahmad KA, Ahmann JR, Migas MC, et al. Decreased liver hepcidin expression in the Hfe knockout mouse. Blood Cells Mol Dis. 2002;29(3):361-366. Nicolas G, Viatte L, Lou DQ, et al. Constitutive hepcidin expression prevents iron overload in a mouse model of hemochromatosis. Nat Genet. 2003;34 (1):97-101. Liu J, Zhang J, Ginzburg Y, et al. Quantitative analysis of murine terminal erythroid differentiation in vivo: novel method to study normal and disordered erythropoiesis. Blood. 2013;121(8):e43-49. Pryor WA. On the detection of lipid hydroperoxides in biological samples. Free Radic Biol Med. 1989;7(2):177-178. Esterbauer H, Cheeseman KH. Determination of aldehydic lipid peroxidation products: malonaldehyde and 4hydroxynonenal. Methods Enzymol. 1990;186:407-421. Gardenghi S, Marongiu MF, Ramos P, et al. Ineffective erythropoiesis in beta-thalassemia is characterized by increased iron absorption mediated by down-regulation of hepcidin and up-regulation of ferroportin. Blood. 2007;109(11):5027-5035. Brissot P, Pietrangelo A, Adams PC, de Graaff B, McLaren CE, Loreal O. Haemochromatosis. Nat Rev Dis Primers. 2018;4:18016. Adamsky K, Weizer O, Amariglio N, et al.

44.

45. 46.

47.

48.

49.

50.

51.

52.

53. 54.

55.

Decreased hepcidin mRNA expression in thalassemic mice. Br J Haematol. 2004;124 (1):123-124. Li H, Rybicki AC, Suzuka SM, et al. Transferrin therapy ameliorates disease in beta-thalassemic mice. Nat Med. 2010;16 (2):177-182. Camaschella C. Iron and hepcidin: a story of recycling and balance. Hematology Am Soc Hematol Educ Program. 2013;2013:1-8. Guo S, Casu C, Gardenghi S, et al. Reducing TMPRSS6 ameliorates hemochromatosis and beta-thalassemia in mice. J Clin Invest. 2013;123(4):1531-1541. Schmidt PJ, Toudjarska I, Sendamarai AK, et al. An RNAi therapeutic targeting Tmprss6 decreases iron overload in Hfe(-/-) mice and ameliorates anemia and iron overload in murine beta-thalassemia intermedia. Blood. 2013;121(7):1200-1208. Zhen AW, Nguyen NH, Gibert Y, et al. The small molecule, genistein, increases hepcidin expression in human hepatocytes. Hepatology. 2013;58(4):1315-1325. Corradini E, Meynard D, Wu Q, et al. Serum and liver iron differently regulate the bone morphogenetic protein 6 (BMP6)-SMAD signaling pathway in mice. Hepatology. 2011;54(1):273-284. Kautz L, Meynard D, Monnier A, et al. Iron regulates phosphorylation of Smad1/5/8 and gene expression of Bmp6, Smad7, Id1, and Atoh8 in the mouse liver. Blood. 2008;112(4):1503-1509. Zhang AS, Yang F, Wang JH, Tsukamoto H, Enns CA. Hemojuvelin-neogenin interaction is required for bone morphogenic protein-4induced hepcidin expression. J Biol Chem. 2009;284(34): 22580-22589. Gao JW, Chen JX, Kramer M, Tsukamoto H, Zhang AS, Enns CA. Interaction of the hereditary hemochromatosis protein HFE with transferrin receptor 2 is required for transferrin-induced hepcidin expression. Cell Metabolism. 2009;9(3):217-227. Du X, She E, Gelbart T, et al. The serine protease TMPRSS6 is required to sense iron deficiency. Science. 2008;320(5879):1088-1092. Kent P, Wilkinson N, Constante M, et al. Hfe and Hjv exhibit overlapping functions for iron signaling to hepcidin. J Mol Med (Berl). 2015;93(5):489-498. Kim A, Nemeth E. New insights into iron regulation and erythropoiesis. Curr Opin Hematol. 2015;22(3):199-205.

1781

ARTICLE Ferrata Storti Foundation

Haematologica 2019 Volume 104(9):1782-1788

Myeloproliferative Neoplasms

Long-term outcome after allogeneic hematopoietic cell transplantation for myelofibrosis

Marie Robin,1 Liesbeth C. de Wreede,2 Christine Wolschke,3 Johannes Schetelig,4 Diderik-Jan Eikema,5 Maria Teresa Van Lint,6 Nina Simone Knelange,7 Dietrich Beelen,8 Arne Brecht,9 Dietger Niederwieser,10 Antonin Vitek,11 Wolfgang Bethge,12 Renate Arnold,13 Jürgen Finke,14 Liisa Volin,15 Ibrahim Yakoub-Agha,16 Arnon Nagler,17 Xavier Poiré,18 Hermann Einsele,19 Patrice Chevallier,20 Ernst Holler,21 Per Ljungman,22 Stephen Robinson,23 Alekxandar Radujkovic,24 Donal McLornan,25 Yves Chalandon26 and Nicolaus Kröger3

Hôpital Saint-Louis, APHP, Université Paris 7, Paris, France; 2Department of Biomedical Data Sciences, LUMC, Leiden, the Netherlands and DKMS CTU, Dresden, Germany; 3 University Hospital Eppendorf, Hamburg, Germany; 4Medizinische Klinik und Poliklinik I, Universitätsklinikum Dresden, Dresden, Germany; 5EBMT Statistical Unit, Leiden, the Netherlands; 6Ospedale San Martino, Genova, Italy; 7EBMT Data Office, Leiden, the Netherlands; 8University Hospital, Essen, Germany; 9Helios HSK Wiesbaden, Wiesbaden, Germany; 10University Hospital Leipzig, Leipzig, Germany; 11Institute of Hematology and Blood Transfusion, Prague, Czech Republic; 12Universität Tübingen, Tübingen, Germany; 13 Charité Universitätsmedizin Berlin, Berlin, Germany; 14Division of Medicine I, Hematology, Oncology and Stem Cell Transplantation, University of Freiburg, Freiburg, Germany; 15HUCH Comprehensive Cancer Center, Helsinki, Finland; 16CHU de Lille, INSERM U995, Lille, France; 17Chaim Sheba Medical Center, Tel-Hashomer, Israel; 18 Cliniques Universitaires St. Luc, Brussels, Belgium; 19Department of Internal Medicine II, University Hospital Würzburg, Würzburg, Germany; 20CHU Nantes, Nantes, France; 21 University Regensburg, Regensburg, Germany; 22Karolinska University Hospital, Stockholm, Sweden; 23Bristol Oncology Centre, Bristol, UK; 24University of Heidelberg, Heidelberg, Germany; 25Comprehensive Cancer Centre, Department of Haematology, Kings College, London, UK and 26Hôpitaux Universitaires de Genève and Faculty of Medicine, University of Geneva, Geneva, Switzerland 1

Correspondence: MARIE ROBIN marie.robin@aphp.fr Received: August 24, 2018 Accepted: January 31, 2019. Pre-published: February 7, 2019. doi:10.3324/haematol.2018.205211 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1782 ©2019 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.

1782

ABSTRACT

llogeneic hematopoietic stem cell transplant remains the only curative treatment for myelofibrosis. Most post-transplantation events occur during the first two years and hence we aimed to analyze the outcome of 2-year disease-free survivors. A total of 1055 patients with myelofibrosis transplanted between 1995 and 2014 and registered in the registry of the European Society for Blood and Marrow Transplantation were included. Survival was compared to the matched general population to determine excess mortality and the risk factors that are associated. In the 2-year survivors, disease-free survival was 64% (60-68%) and overall survival was 74% (71-78%) at ten years; results were better in younger individuals and in women. Excess mortality was 14% (8-21%) in patients aged <45 years and 33% (13-53%) in patients aged ≥65 years. The main cause of death was relapse of the primary disease. Graft-versus-host disease (GvHD) before two years decreased the risk of relapse. Multivariable analysis of excess mortality showed that age, male sex recipient, secondary myelofibrosis and no GvHD disease prior to the 2-year landmark increased the risk of excess mortality. This is the largest study to date analyzing long-term outcome in patients with myelofibrosis undergoing transplant. Overall it shows a good survival in patients alive and in remission at two years. However, the occurrence of late complications, including late relapses, infectious complications and secondary malignancies, highlights the importance of screening and monitoring of long-term survivors. haematologica | 2019; 104(9)

Long-term outcome after transplant in MF patients

Introduction Myelofibrosis (MF) is a malignant clonal disease that can be classified as either primary or secondary to either essential thrombocythemia (ET) or polycythemia vera (PV). The clinical phenotype of MF is markedly heterogeneous and disease severity can be assessed by a number of different prognostic scoring systems. For example, utilizing the Dynamic International Prognostic Scoring System (DIPSS-PLUS), low, int-1, int-2 and high-risk patients have a median survival of 15 years, 6.5 years, 35 months and 16 months, repectively.1 JAK-2 inhibitors, specifically ruxolitinib, which remains the only licensed therapeutic agent in MF, alleviate many symptoms and even possibly increase survival, but they are not considered curative.2-4 Only allogeneic hematopoietic stem cell transplantation (HSCT) has been proposed as curative; overall, HSCT has been reported to cure 30-65% of these patients.5-16 One registry paper analyzed the timing to transplant in patients aged <65 years and concluded that those with intermediate-2 or high-risk disease are those who clearly benefit from transplantation strategies.17 This analysis included transplant-episodes prior to the ruxolitinib era, and the role of this agent on transplantation strategies remains under debate.18 Early mortality (within 2 years) after transplantation is known to be 10-30%, but so far no study has analyzed the outcome of transplanted MF patients after this early period. In contrast, long-term outcome studies have been published for HSCT recipients who have more common disease types, such as acute leukemia, lymphoma, and chronic myeloid leukemia.19-24 Understanding the long-term outcome for transplanted MF patients will help to improve monitoring and promote increased awareness of the potential risks of relapse or, indeed, mortality, particularly when compared to the general population.

Methods Patient selection Only patients from countries for which the population mortality tables are available in a uniform format through the Human Mortality Database, allowing a sex- and agematched comparison, and contributing more than twenty allogeneic transplantations for MF were included in the study. Patients aged <18 years and those who were transplanted from an unrelated matched cord blood were excluded. Patients were analyzed at the time of their first allogeneic transplant only. A total of 2,459 patients received a first allogeneic HCT between January 1995 and December 2014 for primary or secondary MF. A total of 1,055 of these 2,459 patients were reported alive and free of their disease at two years after HSCT; these patients were considered for the study and called long-term (disease-free) survivors. These patients were transplanted in 178 centers in 15 countries.

Statistical analysis The end points of interest were overall survival (OS), disease-free survival (DFS), relapse/progression and nonrelapse mortality (NRM) within the first ten years after HSCT for patients alive and disease-free at the 2-year LM after HSCT. For all outcomes, patients were considered to be at risk since this LM. Median follow up was determined using the reverse Kaplan-Meier method. OS was defined as the time since LM until death from any cause, with surviving patients censored at the time of last follow up. Patients still at risk at ten years after HSCT were administratively censored. DFS was defined as time to death or relapse/progression (whichever occurred first). OS and DFS were estimated using the Kaplan-Meier product limit estimation method, and differences in subgroups were assessed by the Log-rank test. The cumulative incidences of relapse/progression (CIR) and NRM were analyzed together in a competing risks framework.25 Competing risks analyses were also applied to estimate the incidences of (extensive) chronic graft-versus-host disease (cGvHD) and secondary malignancies, each with the competing event death, at ten years after HSCT. Previous acute GvHD (aGvHD) in the landmark population was quantified as a simple proportion, since all cases of aGvHD occurred prior to the 2-year LM time point. Cox proportional hazards regression was used to assess the impact of potential risk factors on OS, RFS, CIR and NRM. CIR and NRM were analyzed in a competing risks framework in which the cause-specific hazards (CSH) were modeled. Methods from relative survival were used to estimate the proportion of the deaths observed in our cohort which could be attributed to population causes (population mortality) and which to MF-related causes, including HSCT and pre-treatment (excess mortality).26,27 Patients were matched by age, sex and country and year of HSCT to a cohort from the general population, for whom survival information was available in the population tables in the Human Mortality Database (http:// www.mortality.org/). The excess hazard of death was defined as the difference between the observed hazard in the patient cohort (this myelofibrosis cohort) and the hazard of the matched general population cohort. For multivariable analyses, we estimated Cox proportional hazards models for the excess hazard of death. Risk factors considered were age, sex, MF classification (primary vs. secondary), conditioning intensity, total body irradiation (TBI), donor type, stem cell source, and previous GvHD (defined as the development of any type of GvHD between transplantation and the 2year LM). All estimates are reported with 95% confidence intervals. All analyses were performed in SPSS version 23 and R 3.3.0 (https://cran.r-project.org/), ‘survival’, ‘cmprsk’, ‘prodlim’ and ‘relsurv’ packages.

Results Characteristics of patients and transplant

Definitions Relapse was defined as disease recurrence. Causes of death were classified as related to relapse if the patient experienced a relapse at any period during follow up. Excess mortality was defined as the difference between mortality observed in the myelofibrosis landmark (LM) cohort and mortality in a matched cohort of the general population. haematologica | 2019; 104(9)

Characteristics of the entire patient cohort and the long-term survivors are shown in Table 1. Long-term survivors were transplanted at a median age of 53.5 years; 837 (79%) patients had primary MF at the time of transplantation, 645 (63%) patients received a reduced intensity regimen, and 471 (45%) were transplanted using an HLA-matched sibling donor. 1783

M. Robin et al. Table 1. Patients’ and transplant characteristics.

Whole cohort N Total number of patients Disease at time of transplant Primary myelofibrosis Secondary myelofibrosis Transformation into acute leukemia Median age at HSCT, years < 45 years 45-54 years 55-64 years ≥ 65 years Interval primary diagnosis and transplant, median < 12 months ≥12 months Conditioning regimen Reduced intensity Standard Total body irradiation, Yes No Source of stem cells Marrow Blood Donor type HLA matched sibling donor Other

2-year landmark %

2459

1055

1904 421 134 55 355 729 1137 238

78 17 5

79 18 3

30 70

837 188 30 53.5 193 351 426 85 26.7 308 747

743 1716 1502 877 423 2015

63 37 17 83

645 378 191 855

63 37 18 82

332 2127

14 86

150 905

14 86

1022 1379

43 57

471 565

45 55

14 30 46 10

19 33 40 8 29 71

N: number; HSCT: hematopoietic stem cell transplantation. Unreported data found for regimen and type of donor but always < 4%.

Outcome and predictors for outcome

Table 2. Causes of mortality after two years.

In the entire cohort (2459 patients, without LM), OS and DFS at ten years were 41% (95%CI: 39-44) and 32% (95%CI: 30-35). Median follow up in the LM population was 49.7 months (95%CI: 47-52). In the 1,055 long-term survivors, 166 deaths were registered within ten years after HSCT. For all time periods, the most common cause of death was relapse of MF, followed by GvHD and infection, with a higher occurrence of infection-related deaths between 2- and 5-years post-transplant (Table 2). In the LM population, secondary cancers occurred in 34 patients before the landmark and in 87 patients after the landmark. This translated into a cumulative incidence in the LM population without cancer before the LM at ten years of 14% (11-18) ten years after the transplantation. The most frequent cancer was solid tumor (70%, of whom 3 breast cancers), followed by acute leukemia or myelodysplastic syndrome (17%) and lymphoma (9%). Grade 2-4 acute GvHD had occurred in 23% of the LM patients (n=245). Before LM, 56% (576 patients) of the patients in the LM population had chronic GvHD of whom 263 patients had an extensive chronic GvHD. Among patients without chronic GvHD before the 2-year LM, cumulative incidence of chronic extensive and limited GvHD were 13% (8-18) and 9% (5-12%), respectively. Ten-year OS and DFS for 2-year survivors were 74% (7178%) and 64% (60-68%), respectively (Figure 1). In these patients, relapse incidence and non-relapse mortality ten years after transplant were estimated at 21% (17-24%) and 15% (12-18%) (Figure 1). Risk factors for mortality, DFS and relapse are shown in Table 3. Older age (P<0.001), type of myelofibrosis (higher risk for secondary myelofibrosis, P=0.01), male sex (P=0.004) and no

Years from transplant

1784

Relapse/progression Secondary malignancy* GvHD Infection Organ damage/toxicity Unknown Total

2-5 y %

>5-10 y %

33 9 18 17 4 28 109

41 11 22 21 5

30 8 9 2

61 16 19 4

8 57

y: years; N: number; GvHD: graft-versus-host disease. *Including post-transplant lymphoproliferative disease.

GvHD before LM (P=0.02) were associated with a significantly higher risk of mortality. Older age (P=0.033), reduced intensity conditioning (RIC) (P=0.017), male sex (P=0.003), donor other than an HLA-matched related donor (P=0.01) and no GvHD before landmark were associated significantly with lower DFS. Use of a donor other than HLA-matched related donor (P=0.008), RIC (P=0.042) and no GvHD occurrence before the landmark (P<0.001) significantly increased the risk of relapse.

Comparison to general population The excess mortality of the two-year landmark MF cohort was 21% (18-25%) at ten years; its population mortality was 4% (4-4.2%) (Figure 2). Excess mortality was lower in younger patients and in female gender recipients but remained considerably greater than the haematologica | 2019; 104(9)

Long-term outcome after transplant in MF patients

mortality of the matched population (Figure 2). Excess mortality in the younger cohort (<45 years) was 14% (821%) and population mortality was 1% (1-1.1%) at this age. In contrast, excess mortality in the older cohort (â&#x2030;Ľ65 years) was 33% (13-53%) and population mortality was 12% (10-14%).

Risk factors for late excess mortality A Cox model was developed to estimate the risk factors for excess mortality in the 2-year disease-free survivors. Of note, the interpretation of the influence of variables in this LM model applies to patients alive and free

of the disease two years following transplantation. For instance, patients with severe GvHD may not survive the second year post-transplant but the subset of patients who survived with severe GvHD are incorporated in the model. The multivariable model shows that older age, MF secondary to PV or ET, male gender recipient were risk factors for excess mortality (Table 4). In long-term survival, previous GvHD was protective for mortality (Table 4). The model highlights that age and sex, which were at higher risk in the general model, are still risk factors for excess mortality. Figure 3 shows changes in the hazard of excess mortality of reference patients according

Figure 1. Outcome of myelofibrosis patient from landmark time. (Left) Overall survival (OS; solid line) and disease-free survival (DFS: dashed line) from landmark time. (Right) Incidence of relapse (solid line) and non-relapse mortality (NRM) (dashed line). N: number; Tx: transplant.

Table 3. Multivariable (cause-specific) Cox proportional hazards models for outcomes in the period between two and ten years after hematopoietic stem cell transplantation for patients alive and disease-free at two years after hematopoietic stem cell transplantation.

Variables

Overall survival HR (95%CI)

Age (per decade) 1.45 (1.19 - 1.76) Patient sex Male 1 Female 0.58 (0.4 - 0.84) MF classification PMF 1 SMF 1.66 (1.13 - 2.44) Source of stem cells Marrow 1 PB 0.83 (0.51 - 1.34) Conditioning regimen intensity MAC 1 RIC 1.17 (0.79 - 1.73) Conditioning regimen with Chemo only 1 TBI 1.25 (0.81 - 1.93) Type of donor Matched sibling 1 Unrelated 1.08 (0.77 - 1.51) Any previous GvHD 0.67 (0.48 - 0.94)

Disease-free survival HR (95%CI)

Relapse HR (95%CI)

<0.001

1.18 (1.01 - 1.37)

0.033

1.16 (0.96 - 1.42)

0.131

0.004

1 0.65 (0.49 - 0.87)

0.003

1 0.79 (0.55 - 1.14)

0.205

0.01

1 1.35 (0.97 - 1.88)

0.071

1 1.07 (0.67 - 1.7)

0.442

1 0.77 (0.52 - 1.13)

0.178

1 0.67 (0.41 - 1.09)

0.107

0.434

1 1.48 (1.07 - 2.04)

0.017

1 1.54 (1.02 - 2.35)

0.042

0.322

1 1.28 (0.89 - 1.82)

0.18

1 1.28 (0.8 - 2.06)

0.305

0.669 0.02

1 1.43 (1.09 - 1.89) 0.62 (0.47 - 0.81)

0.011 0.001

1 1.65 (1.14 - 2.39) 0.42 (0.3 - 0.6)

0.008 <0.001

0.78

HR: Hazard Ratio; CI: Confidence Interval; MF: myelofibrosis; PMF: primary myelofibrosis; SMF: secondary myelofibrosis; PB: peripheral blood; MAC: myeloablative conditioning; RIC: reduced intensity conditioning; TBI: total body irradiation; GvHD: graft-versus-host disease.Variables that are significantly associated with the risk are in bold.

haematologica | 2019; 104(9)

1785

M. Robin et al.

to Cox model (variables from Table 4) transplanted at the age of 50 years; the hazards are given for men and for women separately. We can see that there is a decline in hazard of excess mortality over time post HSCT, but after three years (5 years post-transplant), there is a plateau.

Discussion This EBMT report of 1,055 patients alive and in remission at two years after HSCT is the largest study of longterm post-transplant outcome in patients with MF. Results indicate that survival ten years after transplantation in these 2-year survivors is 74%, but also that the

mortality rate does not decrease to that expected in the general population. This is the first long-term study in MF using LM analytical methods. It had previously been reported in other diseases that long-term outcome in transplanted patients remains lower than expected in the general population (except in aplastic anemia).20,21 Our results can be considered disappointing as compared to previous publications, especially from the Center for International Blood and Marrow Transplant Research,20 but the median age was two decades higher in our cohort, which could explain the higher long-term mortality. Indeed, we could confirm that in a subgroup of patients aged under 45 years, OS was very good at 86% ten years after transplantation. Two additional recent long-term analyses in patients with chronic malignancies [chronic lymphocytic leukemia (CLL) and myelodysplastic syndrome (MDS)] from the EBMT registry included patients with a median age closer to MF patients estimated, with long-term survival lower than in this MF cohort.27,24 Similar risk factors for mortality were found with a better OS in women and in younger24 patients. The reason for the higher risk in male recipients is not clear but it is usually thought to be due to behaviors which place the patient at higher risk, and also to a higher propensity towards comorbidities such as cardio-vascular disease.28 In contrast, an EBMT study of patients with acute myeloid leukemia did not show age or sex to be predictors for OS.29 Like in other malignant disorders, late relapse was the leading cause of death in MF patients following HSCT.1922,24,29 Incidence of relapse at ten years after transplant in the long-term survivors is 21%, in agreement with that expected in other malignant disorders. This highlights the fact that, even if the relapse risk decreases over time, it

Table 4. A multivariable Cox proportional hazards model for excess mortality in the period between two and ten years after hematopoietic stem cell transplantation for patients alive and disease-free at two years after hematopoietic stem cell transplantation.

95% confidence interval

1 0.62 1.35

0.41 - 0.93 1.08 - 1.69

0.022 0.008

1 1.81

1.18 - 2.78

0.007

0.74 - 1.82

0.527

0.75 - 2.08

0.384

1 1.1

0.75 - 1.63

0.623

1 0.83

0.48 - 1.44

0.515

1 0.65

0.44 - 0.96

0.031

Hazard ratio

Figure 2. Mortality in myelofibrosis compared to the general population. (Top) Plots show mortality of the disease-free survivors (black line) and of the general population (gray line). (Middle) Plots show mortality of the myelofibrosis patients according to sex (black solid line: female; dashed line: male) and mortality in the general population (gray lines). (Bottom) Plots show mortality of disease-free survivors (black lines) and general population (gray lines) according to age categories. Tx: transplantation.

1786

Patient sex Male Female Age (per decades) Disease Primary myelofibrosis Secondary myelofibrosis Conditioning regimen Standard Reduced intensity regimen No TBI TBI in regimen Donor Matched sibling donor Other donor Source of stem cells Marrow Blood GvHD No Any

1 1.16 1 1.25

TBI: total body irradiation; GvHD: graft-versus-host disease.

haematologica | 2019; 104(9)

Long-term outcome after transplant in MF patients

can still occur late after transplant. Many studies have reported that relapse risk is related to the disease risk at the time of transplant; unfortunately, due to the retrospective registry-based nature of this study, we did not have sufficient data to calculate a relevant International Prognostic Scoring System so this could not be analyzed. However, we observed that the relapse risk was higher in patients who received a RIC, which could be expected. We were, however, surprised that in long-term survivors, the intensity of the regimen still had some impact. In acute myeloid leukemia, the EBMT long-term study did not find that regimen intensity still influences late relapse.29 Occurrence of acute or chronic GvHD before the LM was the strongest factor preventing relapse in long-term survivors. While in many other studies GvHD increased the risk of late deaths, we failed to confirm this in our MF cohort.19,20 GvHD before LM (2 years) in longterm survivors was protective for both relapse risk as well as for mortality. Of course, from this analysis, we cannot extrapolate data confirming that GvHD is needed to improve long-term outcome, because patients with GvHD leading to death in the first two years of transplant had been excluded from the study. The weakness for GvHD analysis within this cohort was that we could not delineate the risk of “active GvHD” because we had no data regarding GvHD resolution, although it is probable that patients still alive at two years with chronic GvHD were those with the less severe GvHD. The vast majority of patients had onset of chronic GvHD before the LM, but some patients had also a late onset. Finally, the majority of survivors suffered (or had suffered) from chronic GvHD which may alter their quality of life, and it is noteworthy that, even if they are in remission from their MF, patients could have a chronic GvHD which can be a cause of death particularly before five years. Infectious complications remained a frequent cause of death between two and five years post transplant. It has previously been reported that splenectomy before transplant increased the risk of late severe infection which may in part contribute to these findings within the MF cohort.30 This high risk of lethal infection should be taken into account in long-term monitoring strategies and highlights the importance of appropriate anti-infective prophylaxis.31,32 Second malignancies were also the cause of very late

References 1. Gangat N, Caramazza D, Vaidya R, et al. DIPSS Plus: A Refined Dynamic International Prognostic Scoring System for Primary Myelofibrosis That Incorporates Prognostic Information From Karyotype, Platelet Count, and Transfusion Status. J Clin Oncol. 2011;29(4):392-397. 2. Verstovsek S, Mesa RA, Gotlib J, et al. A double-blind, placebo-controlled trial of ruxolitinib for myelofibrosis. N Engl J Med. 2012;366(9):799-807. 3. Verstovsek S, Mesa RA, Gotlib J, et al. Efficacy, safety and survival with ruxolitinib in patients with myelofibrosis: results of a median 2-year follow-up of COMFORT-I. Haematologica. 2013;98(12):1865-1871.

haematologica | 2019; 104(9)

Figure 3. Hazard rate for excess risk of mortality over time post transplant. Curves show hazard rates for two reference patients, based on the Cox model for the excess hazard. They were both 50 years (y) of age at time of hematopoietic stem cell transplantation (Tx), had primary myelofibrosis, received standard conditioning, did not receive total body irradiation, had a matched sibling donor, marrow was source of stem cells, and had no previous graft-versus-host disease. Solid line: male patient; dashed line: female patient.

deaths, justifying long-term monitoring and cancer prevention in this population. After five years, 16% of deaths were due to second malignancies, and at ten years, cumulative incidence of secondary cancer was 14%. We could not analyze specific risk factors for second malignancies due to the small numbers involved. There are few long-term survivors for non-transplanted higher risk MF so there are no data for long-term secondary cancers within that population and risk factors are unknown. It is hard to determine how the transplantation process increases the risk of cancer, but chemotherapy, radiotherapy, immune deficiency, chronic GvHD, genetic susceptibility as well as age can cumulatively contribute towards an increased susceptibility. In conclusion, patients with MF have good survival when alive and in remission two years after transplantation, especially younger and female recipients. Severe late complications and late relapses should be monitored and prevention highlighted in order to reduce life-threatening complications. Lifelong follow up is required to optimize long-term outcomes.33

4. Verstovsek S, Mesa RA, Gotlib J, et al. Longterm treatment with ruxolitinib for patients with myelofibrosis: 5-year update from the randomized, double-blind, placebo-controlled, phase 3 COMFORT-I trial. J Hematol Oncol. 2017;10(1):55. 5. Guardiola P, Anderson JE, Bandini G, et al. Allogeneic stem cell transplantation for agnogenic myeloid metaplasia: a European Group for Blood and Marrow Transplantation, Société Française de Greffe de Moelle, Gruppo Italiano per il Trapianto del Midollo Osseo, and Fred Hutchinson Cancer Research Center Collaborative Study. Blood. 1999;93(9):2831-2838. 6. Kerbauy DMB, Gooley TA, Sale GE, et al. Hematopoietic cell transplantation as curative therapy for idiopathic myelofibrosis, advanced polycythemia vera, and essential

thrombocythemia. Biol Blood Marrow Transplant. 2007;13(3):355-365. 7. Patriarca F, Bacigalupo A, Sperotto A, et al. Allogeneic hematopoietic stem cell transplantation in myelofibrosis: the 20-year experience of the Gruppo Italiano Trapianto di Midollo Osseo (GITMO). Haematologica. 2008;93(10):1514-1522. 8. Kröger N, Holler E, Kobbe G, et al. Allogeneic stem cell transplantation after reduced-intensity conditioning in patients with myelofibrosis: a prospective, multicenter study of the Chronic Leukemia Working Party of the European Group for Blood and Marrow Transplantation. Blood. 2009;114 (26):5264-5270. 9. Ballen KK, Shrestha S, Sobocinski KA, et al. Outcome of transplantation for myelofibrosis. Biol Blood Marrow Transplant J Am Soc

1787

M. Robin et al.

10.

11.

12.

13.

14.

15.

16.

17.

1788

Blood Marrow Transplant. 2010;16(3):358367. Robin M, Tabrizi R, Mohty M, et al. Allogeneic haematopoietic stem cell transplantation for myelofibrosis: a report of the Société Française de Greffe de Moelle et de Thérapie Cellulaire (SFGM-TC). Br J Haematol. 2011;152(3):331-339. Gupta V, Malone AK, Hari PN, et al. Reduced-intensity hematopoietic cell transplantation for patients with primary myelofibrosis: a cohort analysis from the center for international blood and marrow transplant research. Biol Blood Marrow Transplant. 2014;20(1):89-97. Rondelli D, Goldberg JD, Isola L, et al. MPDRC 101 prospective study of reduced-intensity allogeneic hematopoietic stem cell transplantation in patients with myelofibrosis. Blood. 2014;124(7):1183-1191. Ditschkowski M, Elmaagacli AH, Trenschel R, et al. Dynamic International Prognostic Scoring System scores, pre-transplant therapy and chronic graft-versus-host disease determine outcome after allogeneic hematopoietic stem cell transplantation for myelofibrosis. Haematologica. 2012;97(10): 1574-1581. Deeg HJ, Gooley TA, Flowers MED, et al. Allogeneic hematopoietic stem cell transplantation for myelofibrosis. Blood. 2003; 102(12):3912-3918. Scott BL, Gooley TA, Sorror ML, et al. The Dynamic International Prognostic Scoring System for myelofibrosis predicts outcomes after hematopoietic cell transplantation. Blood. 2012;119(11):2657-2664. Samuelson Bannow BT, Salit RB, Storer BE, et al. Hematopoietic Cell Transplantation for Myelofibrosis: the Dynamic International Prognostic Scoring System Plus Risk Predicts Post-Transplant Outcomes. Biol Blood Marrow Transplant. 2018;24(2): 386-392. Kröger N, Giorgino T, Scott BL, et al. Impact

18. 19.

20.

21.

22.

23.

24.

25.

of allogeneic stem cell transplantation on survival of patients less than 65 years of age with primary myelofibrosis. Blood. 2015;125(21):3347-3350; quiz 3364. Cervantes F, Pereira A. Does ruxolitinib prolong the survival of patients with myelofibrosis? Blood. 2017;129(7):832-837. Socié G, Stone JV, Wingard JR, et al. Longterm survival and late deaths after allogeneic bone marrow transplantation. Late Effects Working Committee of the International Bone Marrow Transplant Registry. N Engl J Med. 1999;341(1):14-21. Wingard JR, Majhail NS, Brazauskas R, et al. Long-term survival and late deaths after allogeneic hematopoietic cell transplantation. J Clin Oncol. 2011;29(16):2230-2239. Goldman JM, Majhail NS, Klein JP, et al. Relapse and late mortality in 5-year survivors of myeloablative allogeneic hematopoietic cell transplantation for chronic myeloid leukemia in first chronic phase. J Clin Oncol. 2010;28(11):1888-1895. Martin PJ, Counts GW, Appelbaum FR, et al. Life expectancy in patients surviving more than 5 years after hematopoietic cell transplantation. J Clin Oncol. 2010;28(6):10111016. Bhatia S, Francisco L, Carter A, et al. Late mortality after allogeneic hematopoietic cell transplantation and functional status of long-term survivors: report from the Bone Marrow Transplant Survivor Study. Blood. 2007;110(10):3784-3792. van Gelder M, de Wreede LC, Bornhäuser M, et al. Long-term survival of patients with CLL after allogeneic transplantation: a report from the European Society for Blood and Marrow Transplantation. Bone Marrow Transplant. 2017;52(3):372-380. Iacobelli S; EBMT Statistical Committee. Suggestions on the use of statistical methodologies in studies of the European Group for Blood and Marrow Transplantation. Bone Marrow Transplant. 2013;48 Suppl 1S1-37.

26. Pohar Perme M, Estève J, Rachet B. Analysing population-based cancer survival - settling the controversies. BMC Cancer. 2016;16(1):933. 27. Schetelig J, de Wreede LC, van Gelder M, et al. Late treatment-related mortality versus competing causes of death after allogeneic transplantation for myelodysplastic syndromes and secondary acute myeloid leukemia. Leukemia. 2019;33(3):686-695. 28. Pophali PA, Klotz JK, Ito S, et al. Male survivors of allogeneic hematopoietic stem cell transplantation have a long term persisting risk of cardiovascular events. Exp Hematol. 2014;42(2):83-89. 29. Shimoni A, Labopin M, Savani B, et al. Long-term survival and late events after allogeneic stem cell transplantation from HLAmatched siblings for acute myeloid leukemia with myeloablative compared to reduced-intensity conditioning: a report on behalf of the acute leukemia working party of European group for blood and marrow transplantation. J Hematol Oncol. 2016;9 (1):118. 30. Robin M, Porcher R, De Castro Araujo R, et al. Risk factors for late infections after allogeneic hematopoietic stem cell transplantation from a matched related donor. Biol Blood Marrow Transplant. 2007;13(11): 1304-1312. 31. Syrjala KL, Martin PJ, Lee SJ. Delivering care to long-term adult survivors of hematopoietic cell transplantation. J Clin Oncol. 2012; 30(30):3746-3751. 32. Majhail NS, Rizzo JD, Lee SJ, et al. Recommended screening and preventive practices for long-term survivors after hematopoietic cell transplantation. Rev Bras Hematol Hemoter. 2012;34(2):109-133. 33. Bhatia S, Armenian SH, Landier W. How I monitor long-term and late effects after blood or marrow transplantation. Blood. 2017;130(11):1302-1314.

haematologica | 2019; 104(9)

ARTICLE

Chronic Myeloid Leukemia

De novo UBE2A mutations are recurrently acquired during chronic myeloid leukemia progression and interfere with myeloid differentiation pathways

Vera Magistroni,1* Mario Mauri,1* Deborah Dâ&#x20AC;&#x2122;Aliberti,1 Caterina Mezzatesta,1 Ilaria Crespiatico,1 Miriam Nava,1 Diletta Fontana,1 Nitesh Sharma,1 Wendy Parker,2 Andreas Schreiber,2 David Yeung,2,3 Alessandra Pirola,4 Sara Readelli,1 Luca Massimino,1 Paul Wang,2 Praveen Khandelwal,1 Stefania Citterio,5 Michela Viltadi,1 Silvia Bombelli,1 Roberta Rigolio,1 Roberto Perego,1 Jacqueline Boultwood,6,7 Alessandro Morotti,8 Giuseppe Saglio,8 Dong-Wook Kim,9 Susan Branford,2,3,10 Carlo Gambacorti-Passerini1,11** and Rocco Piazza1**

Ferrata Storti Foundation

Haematologica 2019 Volume 104(9):1789-1797

Department of Medicine and Surgery, University of Milano Bicocca, Monza, Italy; Center for Cancer Biology, SA Pathology, Adelaide, Australia; 3University of Adelaide, South Australia, Australia; 4GalSeq s.r.l., Monza, Italy; 5Department of Bioscience and Biotechnology, University of Milano Bicocca, Milano, Italy; 6Bloodwise Molecular Haematology Unit, John Radcliffe Hospital, University of Oxford, Oxford, UK; 7NIHR Biomedical Research Centre, Oxford, UK; 8Department of Clinical and Biological Sciences, San Luigi Hospital, University of Turin, Turin, Italy; 9Department of Hematology, Catholic University, Seoul, South Korea; 10University of South Australia, Adelaide, South Australia, Australia and 11Hematology and Clinical Research Unit, San Gerardo Hospital, Monza, Italy 1 2

*VM and MM contributed equally to this work. **RP and CGP contributed equally to this work

ABSTRACT

espite the advent of tyrosine kinase inhibitors, a proportion of chronic myeloid leukemia patients in chronic phase fail to respond to imatinib or to second-generation inhibitors and progress to blast crisis. Until now, improvements in the understanding of the molecular mechanisms responsible for chronic myeloid leukemia transformation from chronic phase to the aggressive blast crisis remain limited. Here we present a large parallel sequencing analysis of 10 blast crisis samples and of the corresponding autologous chronic phase controls that reveals, for the first time, recurrent mutations affecting the ubiquitin-conjugating enzyme E2A gene (UBE2A, formerly RAD6A). Additional analyses on a cohort of 24 blast crisis, 41 chronic phase as well as 40 acute myeloid leukemia and 38 atypical chronic myeloid leukemia patients at onset confirmed that UBE2A mutations are specifically acquired during chronic myeloid leukemia progression, with a frequency of 16.7% in advanced phases. In vitro studies show that the mutations here described cause a decrease in UBE2A activity, leading to an impairment of myeloid differentiation in chronic myeloid leukemia cells.

Correspondence: VERA MAGISTRONI vera.magistroni@unimib.it ROCCO PIAZZA rocco.piazza@unimib.it Received: January 19, 2018. Accepted: February 26, 2019. Pre-published: February 28, 2019. doi:10.3324/haematol.2017.179937 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1789 ÂŠ2019 Ferrata Storti Foundation

Introduction Chronic myeloid leukemia (CML) is a myeloproliferative disorder with an incidence of 1-2 cases per 100,000/year. It is characterized by the presence of the BCR-ABL1 fusion gene, the product of the reciprocal translocation between chromosomes 9 and 22.1 After the translocation, the coding regions of BCR and ABL1 genes are juxtaposed, leading to an enhanced ABL1 tyrosine kinase activity.2 CML is a multi-step disease, evolving from a mild form that is easy to control, called chronic phase (CP), to a very aggressive and incurable acute phase called blast crisis (BC). The majority of CML-CP patients are successfully treated with drugs able to impair BCR-ABL1 kinase activity (tyrosine kinase inhibitors, TKI), thus confirming the central role of the oncogenic fusion protein in CML pathogenehaematologica | 2019; 104(9)

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.

1789

V. Magistroni et al. sis.3,4 However, a fraction of these patients fail to respond to the treatment (primary resistance) or become resistant after an initial response.4-6 The persistence of BCR-ABL1 activity typically drives the progression to the advanced phase of the disease within 3-5 years. One of the open issues in CML concerns the dissection of the molecular mechanisms driving the transformation to BC, commonly considered as a heterogeneous disease at the molecular level.7-9 BC is mainly characterized by the rapid expansion of the differentiation-arrested BCR-ABL1-positive blast cells,10 therefore resembling an acute leukemia. In most cases of BC cases (approx. 70%), blasts maintain myeloid features, while in 20-30% blast lineage is lymphoid. BCRABL1 expression, which increases during CML progression in conjunction with BCR transcription, seems to have a prominent role in this process, hyperactivating proliferative and anti-apoptotic signals and inducing genetic instability.5,11,12 Previous reports showed the existence of a heterogeneous molecular signature among distinct BC patients.5,7,8 However, these data were limited by the scarcity of matched CP/BC samples, due to the infrequent progression to BC after the advent of TKI. Here we analyzed ten paired CP/BC samples through a wholeexome sequencing (WES) approach, identifying somatic variants specific for BC progression since these were not present in the autologous CP controls. Along with several mutations previously identified as BC driver events,5,7,13 we detected, for the first time, recurrent BC-specific mutations occurring on the UBE2A gene. These data suggest that the appearance of UBE2A variants in CML cells could contribute to BC progression through the impairment of myeloid differentiation.

Plasmids, transfections and lentiviral infections BA/F3_BCR-ABL1-positive cells were transfected with pMIGR1_UBE2A vectors (Online Supplementary Appendix) as by Puttini et al.14 and were analyzed for GFP positivity with a FACSAria flow cytometer (BD Bioscience, San Jose, USA) and FACS-sorted when transfection efficiency was lower than 85%. 32Dcl3-BCR/ABL1 cells were electroporated using a Gene Pulser® II Electroporation System (BIORAD) with pMIGR1_UBE2A WT and I33M vectors as described by Puttini et al.14 To obtain stable UBE2A WT or I33M cell lines, GFP positive population was FACS-sorted with a MoFlo Astrios cell sorter equipped with Summit 6.3 software (both from Beckman Coulter, Miami, FL, USA). For UBE2A silencing, K562 cells were infected with lentivirus obtained from MISSION-shRNA pLKO.1-based vectors (TRCN0000320625) (Sigma-Aldrich, Missouri, USA) and packaged using 293FT cell line. As a control, a pLKO.1MISSION nontarget control vector (SHC002) (Sigma-Aldrich) was used. After infection K562 cells were maintained in 2 µg/mL puromycin for selection of silenced (K562_shUBE2A) and control cells (K562_shNC).

Quantitative real-time polymerase chain reaction

Methods

Total RNA was extracted using Trizol (Thermo-FisherScientific) following the manufacturer’s instructions. 1 μg of total RNA was used to synthesize cDNA using reverse transcription reagents (Thermo-Fisher-Scientific) after pre-treatment with DNAseI (Thermo-Fisher-Scientific) to avoid contamination from genomic DNA. Real-time quantitative polymerase chain reaction (RT-qPCR) was performed using TaqMan® Brilliant II QPCR Master Mix (Agilent Technologies, CA, USA) on a StratageneMX3005P (Agilent-Technologies) under standard conditions. The housekeeping gene glucoronidase β gene (GUSB) was used as an internal reference.12 TaqMan® Gene Expression Assays (ThermoFisher-Scientific) were used (Online Supplementary Table S1).

Cell lines

In vitro translation and ubiquitination assay

The BA/F3-BCR-ABL1 and 32Dcl3-BCR/ABL1 cell lines were generated and maintained as described by Puttini et al.14 and Piazza et al.15 K562 and 293FT were purchased from DSMZ (Braunschweig, Germany) and Thermo-Fisher-Scientific (Waltham, MA, USA) respectively, and were maintained according to the manufacturers’ instructions.

In vitro translation of UBE2A proteins was performed with 1Step Human Coupled IVT Kit-DNA (Thermo-Fischer-Scientific) following the manufacturer’s instructions. For ubiquitination assay we incubated 15 μg of UBE2A proteins with 1 μg of GSTUbiquitin (Enzo-Life-Sciences, NY, USA), 0.2 ng of ubiquitin activating enzyme (E1) (Enzo-Life-Sciences), 2 mM ATP, energy regeneration solution (BostonBiochem, MA, USA), 2 mM MgCl2, 2 mM KCl, 16 µg of BA/F3_BCR-ABL1 whole cell lysate in 50 mM TrisHCl (ph7.5). The reactions were incubated for 20 minutes at 37°C. The products were analyzed by western blotting. For the enzymatic activity of WT and mutated UBE2A, 15 μg of UBE2A in vitro synthesized protein were used. The AMP-Glo Assay (Promega catalog v5011) was used in order to quantify the amount of AMP generated by the ubiquitin conjugation machinery, composed of 170 ng/μL ubiquitin protein, 15 ng/μL UBA1 and 50 µM ATP (SignalChem). The production of AMP from ATP is directly proportional to the enzymatic activity of the ubiquitination machinery and therefore it was used to measure the ubiquitination in the presence of WT and mutated UBE2A. The AMP signal was detected using the AMP detection solution (Promega) and a TECAN reading plate (Infinite F200Pro TECAN).

Patients Diagnosis and staging were performed according to the World Health Organization WHO-2008 classification.16 Peripheral blood (PB) or bone marrow (BM) of ten matched CML chronic phase/blast crisis samples, 31 CP-CML, 14 AP/BC-CML, 38 atypical-CML (aCML), and 40 AML were collected at diagnosis and after obtaining written informed consent approved by the institutional ethics committee. The study was conducted in accordance with the Declaration of Helsinki. Samples were prepared as described by Piazza et al.17

Whole exome sequencing Genomic DNA (gDNA) was extracted from purified cells with PureLink Genomic DNA kit (Thermo-Fisher-Scientific). 1 μg of gDNA from each sample was fragmented (500bp) with a Diagenode-Bioruptor sonicator system (Diagenode, Belgium) and processed according to the standard Illumina protocol. The Illumina TruSeq Exome Enrichment kit (Illumina Inc., San Diego, CA, USA) was used to enrich the genomic libraries for the exonic regions and samples were sequenced as described in the Online Supplementary Appendix. 1790

Neutrophilic differentiation For induction of neutrophilic differentiation, 32Dcl3-BCR/ABL1 UBE2A WT and I33M cells were treated as previously described.18 32Dcl3-BCR/ABL1 cells expressing UBE2A WT or I33M were seeded at a density of 2x105 cells per milliliter and cultured in the haematologica | 2019; 104(9)

UBE2A somatic variants in CML progression

presence of imatinib mesylate (1 ÂľM final concentration) in combination with human recombinant GCSF (10 ng/mL) or IL3 (0.5 ng/mL). At days 3 and 6, cells were analyzed using FACS for CD11b surface expression and imaged with confocal microscopy (Online Supplementary Methods).

Results Single nucleotide variants acquired during chronic myeloid leukemia progression Genomic DNA (gDNA) from matched CP/BC samples was obtained for each patient at diagnosis (CP) and after

Table 1. Single nucleotide variants and indels identified by exome sequencing in blast crisis samples and absent in the paired chronic phase control.

Patient

Disease progression

Time to BC from CP Gene at diagnosis name (months) (aa substitution)

Lymphoid

Myeloid

6 7

M M

Lymphoid Lymphoid

52 -

Myeloid

Lymphoid

Myeloid

RTP2(A190V) KCNH3(A314V) FAT4(R1698W) FUT3(R354C) RUNX1(K194N) SMARCA4(A945T) UBE2A(D114V) ABL1(F486S) PTPN11(G503V) AMER3(R709H) LAMA2(P1025S) GRIN3A(R1024*) SMC5(L1102*) MESDC2(E130del) CCDC40(S17L) NRAS(Q61R) DEFB119(R42H) IKZF1(N159S) AK8(R125H) PPT1(V168A) MDH1B(A272T) GPR98(R1745C) CEL(E216Q) LRP4(D449N) CYP2B6(R145W) BCR(F615W) ASXL1(G641_fs) EPB41L3(P963L) FGFR4(V262M) UBE2A(I33M) ABL1(E255V) BARD1(G527_fs) BSN(R3264H) EFCAB4B(V643M) KRT7(R339W) AP5M1(D289N) XPO1(E571K) HMCN1(S1371L) ABCA13(T2019M) ABL1(T315I)

Mutation ratio

OncoScore39

53% 59% 57% 48% 68% 55% 93% 50% 41% 41% 49% 41% 27% 45% 34% 32% 46% 44% 67% 39% 47% 45% 46% 54% 40% 44% 37% 37% 39% 39% 28% 28% 26% 31% 41% 38% 32% 53% 53% 45%

10.22 5.61 57.19 30.21 73 48.69 25.66 85.8 41.25 34.16 10.79 5.95 24.13 28.07 8.89 81.65 0 72.76 18.39 8.58 0 5.54 18.97 12.28 15.77 81.02 77.48 76.11 46.29 25.66 85.8 83.5 3.32 20.44 70.25 0 47.33 16.1 28.17 85.8

*Stop codon. OncoScore is a text-mining tool that scores genes according to their association with cancer based on available biomedical literature; higher scores correspond to a stronger association with cancer. BC: blast crisis; CP: chronic phase.

haematologica | 2019; 104(9)

1791

V. Magistroni et al. A

Figure 1. Activity of UBE2A mutants. (A) Western blot analysis of total cell lysates from BA/F3_BCR-ABL cell lines stably transfected with pMIGR-UBE2A vectors encoding for wild-type (WT) or mutated (D114V or I33M) UBE2A. Empty vector has been used as negative control. (B) Real-time quantitative polymerase chain reaction (RT-qPCR) of total RNA extracted from BA/F3_BCR-ABL_pMIGR/UBE2A cell lines. The values are normalized on the EMPTY cells (***P<0.001). (C) Western blot of total cell lysates from BA/F3_BCR-ABL_pMIGR/UBE2A cell lines. The signal at 14KDa corresponds to histone H2A. The signal at ~23KDa corresponds to monoubiquitinated histone H2A (mUbH2A) according to Wu et al.37 (D) Western blot analysis of the in vitro ubiquitination reaction performed with in vitro translated UBE2A (WT and mutated forms) and GST-ubiquitin on total BA/F3_BCR-ABL lysate. (Right) The densitometric analysis of GST-ub signal (>170KDa) from three independent experiments obtained with ImageJ software.38 The fold change is obtained normalizing the signal on the WT sample (WT vs. D114V *P=0.022; WT vs. I33M **P=0.0069). (E) Histogram showing the enzymatic activity of in vitro expressed UBE2A using the AMP Glow assay (WT vs. D114V *P=0.0056; WT vs. I33M **P=0.0036).

progression to BC. Initially whole exome sequencing (WES) data from ten patients were analyzed. CP samples were used as baseline controls for each patient to identify somatic variants selectively occurring in BC (Table 1), thus allowing the recognition of molecular events occurring exclusively upon CML progression. By using this approach we identified mutations on genes already associated with BC, such as RUNX1, IKZF1, NRAS, ASXL1 and ABL1.7,13 A total of 41 non-synonymous single nucleotide variants (SNV) and small indels were identified, with a mean of 4.1 mutations/patient acquired upon BC progression. Of these events, 31 were transitions, seven transversions and three indels, with the C>T substitution being the most frequent (63.4%) (Online Supplementary Figure S1). In one patient (patient #7) no acquired exonic SNV could be detected during CML progression. Analysis of SNV data showed the presence of two recurrently mutated genes in this cohort: ABL1, with mutations F486S, E255V and T315I occurring on the BCR-ABL1 fusion gene and leading to TKI resistance (30%, 95%CI: 0.574, 0.026), and UBE2A (Xq24), an E2-ubiquitin conju1792

gating enzyme required for post-replicative DNA damage repair15 (20%, 95%CI: 0.447, 0.000), which has never been previously reported as mutated in CML patients. UBE2A mutations occurred on two non-contiguous residues: D114V and I33M (Tables 1 and 2). Patient #3 (male) showed an UBE2A variant frequency of 93%, as expected given that the gene is localized on the X chromosome. Patient #8 (female) carried a heterozygous UBE2A mutation (mutation ratio: 39%). The high mutation ratio observed in both patients suggests that UBE2A is present in the dominant BC clone (Table 1).

UBE2A mutations are recurrent and acquired in late chronic myeloid leukemia The evidence of recurrent, somatic UBE2A mutations has never been reported in BC cases; however, they had been previously found in other clonal disorders both of solid and hematopoietic origin, confirming their potential role in tumor progression.19 To further characterize the pattern and the frequency of UBE2A mutations in a larger cohort of patients, we sequenced 31 additional CML CP haematologica | 2019; 104(9)

UBE2A somatic variants in CML progression

F G

Figure 2. UBE2A silencing in K562 cells. (A) Real-time quantitative polymerase chain reaction (RT-qPCR) analysis of total RNA extracted from K562 cell lines infected with a lentiviral based system for UBE2A silencing (shNC: scrambled negative control; shUBE2A: UBE2A silenced cells). Values are normalized on shNC cells (***P<0.0001). (B) Western blot analysis of total cell lysates from K562_shNC and K562_shUBE2A cells. (C) Heat map of RNA-sequencing data showing colorcoded expression levels of differentially expressed genes in three distinct populations of K562-shUBE2A compared to control (shNC). (D) RT-qPCR analysis in K562 cell lines of a subset of differentially expressed genes identified by RNA-sequencing. (E) Gene set enrichment analysis of the shUBE2A transcriptome. (F) RT-qPCR analysis in the 32Dcl3 cell line of a subset of differentially expressed genes identified by RNA-sequencing. (G and H) CSF3R protein levels in total cell lysate of K562 cells (G) and of BC/CP samples from patient #3, carrying UBE2A mutation in the BC phase (H).

samples at onset, 14 AP/BC, 40 acute myeloid leukemia, and 38 aCML samples. No evidence of UBE2A mutations could be found in the CP, AML or aCML samples, while in two AP/BC samples, somatic UBE2A variants D114Y and M34fs were detected. Globally, acquired UBE2A mutations could be detected in a total of 16.7% (4 of 24) advanced (AP/BC) CML cases (95%CI: 1.78-31.62) (Table 2).

UBE2A mutations affect protein activity Polyphen-2 (http://genetics.bwh.harvard.edu/pph/),20 11 DANN and FATHMM-MKL21 analyses revealed that all the UBE2A variants identified were potentially damaging, as also suggested by the presence of a N-terminal frameshift variant (M34fs) in one of the patients (Table 2). To gain insight into the functional role of UBE2A mutations, we stably transfected the BA/F3_BCR-ABL1 cell line14 with the wild-type (WT) and the mutated UBE2A variants, I33M and D114V. The level of UBE2A expression haematologica | 2019; 104(9)

in stable transfectants was verified both at protein (Figure 1A) and mRNA (Figure 1B) levels. The analysis of the levels of ubiquitin-conjugated H2A, a known UBE2A substrate,22 in total cell lysate revealed a decreased H2A ubiquitination for both UBE2A variants compared to WT (Figure 1C), with the effect of I33M being more prominent. In line with these findings, suggesting a decreased UBE2A activity for both variants, ubiquitination assay performed with in vitro translated WT and mutated UBE2A proteins confirmed a decrease in ubiquitin-conjugating activity for mutants compared to the WT form (Figure 1D). To further support this indication, we developed a new in vitro assay based on the measurement of the AMP concentration as a proxy to assess the overall level of ubiquitination. This test was performed in the presence of GST-ubiquitin and of the E1 ubiquitin activating enzyme UBA1 together with WT or mutated UBE2A; this revealed a significant decrease in 1793

V. Magistroni et al. A

Figure 3. Induction of erythroid differentiation in UBE2A-silenced K562 cell line. K562 cells were treated with 400 μM hydroxyurea. (A) CD235a immunofluorescence staining for UBE2A-silenced K562 (shUBE2A) and control (shNC) cells after hydroxyurea or mock (-) treatments for the indicated times. (scale bar: 25 μm). (B) Average intensity of CD235a signal obtained acquiring ten fields from two independent experiments for each sample (approx. 80 cells each). (C) Fluorescenceactivated cell sorting analysis (FACS) analysis of CD235a levels in K562 cells in presence (red line) or absence (black line) of hydroxyurea. (D) Quantification of CD235a and hemoglobin mRNA relative levels (HBB: Hemoglobin-subunit-β) through real-time quantitative polymerase chain reaction (RT-qPCR) after hydroxyurea treatment.

ATP consumption, and therefore in ubiquitin-conjugating activity, for UBE2A mutants compared to the WT form [see Figure 1E: enzyme specific activity assay, 1.55-fold (P<0.01) and 1.53-fold (P<0.01) decrease in UBE2A D114V and I33M AMP concentration compared to UBE2A WT].

Transcriptome analysis of UBE2A cellular models shows significant perturbation of downstream pathways related to myeloid development To identify the gene networks perturbed by the UBE2A knock-out, stable lentiviral UBE2A silencing models were generated (Figure 2A and B) in the human myeloid K562 cell line (K562-shUBE2A and K562-shNC cells for UBE2A silencing and scrambled control, respectively) (Online Supplementary Figure S2). Whole-transcriptome analysis 1794

(RNA-Seq) highlighted the presence of 168 differentially expressed genes, with 117 of them being down-regulated and 51 up-regulated (Figure 2C). Gene set enrichment analysis (GSEA) showed significant enrichment for ontologies related to myeloid differentiation (Figure 2D) and neural development (Online Supplementary Figure S3). RT-qPCR on K562shNC/shUBE2A cell lines on a set of five differentially expressed genes (ITGB4, RDH10, CLEC11A, CSF3R, RAP1GAP) confirmed RNA-Seq data (Figure 2E). Interestingly, the colony stimulating factor 3 receptor (CSF3R) was potently down-regulated in shUBE2A both at mRNA (12.5-fold downregulation; Figure 2E) and protein (Figure 2F) levels, hence suggesting that its downmodulation may play a role in the differentiation block that is ultimately responsible for the onset of haematologica | 2019; 104(9)

UBE2A somatic variants in CML progression

Figure 4. Induction of neutrophilic differentiation in UBE2A wild-type (WT) or I33M 32Dcl3 cell line. Cells were treated with IL-3 or granulocyte-colony stimulating factor (GCSF). (A) Fluorescence-activated cell sorting analysis (FACS) analysis of CD11b staining after induction of differentiation at days 3 and 6. (B) CD11b immunofluorescence staining for 32Dcl3 control (CTRL), UBE2A WT and I33M at day 6 showing a clear reduction in UBE2A I33M CD11b staining (scale bar: 20 μm).

Table 2. UBE2A single nucleotide variants and indels identified in blast crisis samples and absent in the paired chronic phase control.

Chromosome

Position

Ref

Var

Codon

AA Change

Polyphen2 HDIV

DANN Score

Fathmm MKL

phastCons7 Vertebrate

chrX chrX chrX chrX

119574955 119583137 119583136 119574957

A A G

G T T Ins AT

ATA->ATG GAT->GTT GAT->TAT ATG->ATATG

Ile33Met Asp114Val Asp114Tyr M34fs

D D D n.a.

0.992 0.993 0.996 n.a.

N D D n.a.

0.999 1 1 n.a.

Chr: Ref: reference; Var: variant; n.a.: not available.

the BC. Immunoblot analysis on CP/BC mononuclear cells from patient #3, which acquired the D114V-UBE2A mutation in BC phase, confirmed CSF3R downmodulation (Figure 2G). To confirm the expression signature identified in the UBE2A silencing models, we stably overexpressed UBE2A WT and I33M in the 32Dcl3-BCR/ABL1 murine myeloid cell line (Online Supplementary Figure S4). In line with the expression profile shown in K562 UBE2A silenced cells, also in these cell lines we observed a comparable modulation in the previous analyzed set of five differentially expressed genes (ITGB4 1.52 P<0.01, RDH10 1.30 P<0.05, CLEC11A 2.84 P<0.01, CSF3R 0.25 P<0.05, RAP1GAP 0.27 P<0.01) (Figure 2H; data are reported as fold-change in UBE2A I33M compared to UBE2A WT), therefore supporting the hypotheses that: 1) UBE2A mutations modulate the activity of the target protein in a loss of function manner; and 2) UBE2A mutations probably act as dominant negative variants. Comparison of our signature with known BC data (GEO _GSE47927 HSC data were used for BC vs. CP calculation) indicated the presence of a moderate positive linear correlation (R2 = 0.234) (Online Supplementary Figure S5). Notably, CSF3R expression level was seen to be markedly decreased also in the reference BC database, with a Log2 fold-change of -2.19. Globally, these data indicate that UBE2A mutations are directly responsible for the modulation of CSF3R, ITGB4, RDH10, CLEC11A and RAP1GAP expression. This hypothesis is also corroborated by the 32Dcl3 cell models. haematologica | 2019; 104(9)

UBE2A activity is involved in myeloid differentiation Erythrocytes and megakaryocyte differentiation can be induced in K562 cells by treating with hydroxyurea or phorbol 12-myristate 13-acetate (PMA), respectively.23,24 Treatment of UBE2A-silenced K562 cells with hydroxyurea showed a significant delay in the ability to differentiate into erythrocytes, as assessed by glycophorin A (GYPA-CD235a) expression levels when compared with the scrambled control (Figure 3A-C) (relative CD235a expression compared to shNC fold-change at day 0: 0.54±0.13, data are reported as fold-change in UBE2A I33M compared to UBE2A WT<0.001; day 1: 0.70 ± 0.12, data are reported as fold-change in UBE2A I33M compared to UBE2A WT <0.05; day 3: 0.61 ± 0.15, data are reported as fold-change in UBE2A I33M compared to UBE2A WT <0.05). Fluorescence-activated cell sorting analysis (FACS) showed a 45% decrease in GYPA surface expression in silenced cells compared to controls after 24 hours (h) of treatment (Figure 3B). In line with these findings, induction of hemoglobin-subunit-β (HBB) production was almost completely suppressed in shUBE2A cells [6.6-fold relative decrease of HBB mRNA level at 24 (h) of treatment: data are reported as fold-change in UBE2A I33M compared to UBE2A WT<0.001] further confirming the negative effect of UBE2A silencing on erythroid differentiation (Figure 3C). Similarly, treatment of K562 cells with the megakaryocytic-inducing agent PMA showed significant impairment of megakaryocyte differentiation in shUBE2A cells, as assessed by the expression levels of CD41 (33% downreg1795

V. Magistroni et al.

ulation of surface protein expression; P<0.001) and CD44 (38% decrease at mRNA level; P<0.01) after PMA treatment (Online Supplementary Figure S6). Neutrophilic differentiation was similarly tested in the 32Dcl3 BCR/ABL1 cell lines over-expressing UBE2A WT or I33M. Treatment of the UBE2A I33M cell line with GCSF + IL-3 showed a delay in neutrophilic differentiation, as assessed by CD11b expression levels when compared with both UBE2A WT or control (Figure 4A). Cells treated with IL-3 alone were used as an internal control. FACS showed no difference in CD11b surface expression in UBE2A I33M cell line compared to controls after three days of treatment but showed a 37% decrease at day 6 which was also confirmed by confocal microscopy analysis (Figure 4B).

Discussion In line with previous results,5,7,9 our analysis performed on matched CP/BC CML samples showed considerable somatic heterogeneity in BC phase. In all the samples, we detected a very low number of acquired SNV, corresponding to an average of 4.1 non-synonymous mutations per patient, a frequency far below the average reported for other hematopoietic neoplasms, such as acute myeloid leukemia (AML: 7.8) and chronic lymphocytic leukemia (CLL: 11.9).25 This can in part be explained by the characteristics of our analysis, where somatic variants occurring in BC were filtered against those in CP, therefore filteringout all the driver and passenger variants pre-existing the evolution to BC. All BC samples showed the prevalence of transition events and, in particular, of C:G>T:A substitutions, accounting for 66.7% of all the SNV (Online Supplementary Figure S1). Approximately 85% of the C:G>T:A transitions were part of a CpG dinucleotide. Cytosines in CpG sites are known to be affected by a high mutation rate, caused by a spontaneous deamination of methylated cytosines.26 This mutation pattern is also in accordance with a BCR-ABL1 dependent mutation signature, characterized by inhibition of the mismatch repair system (MMR) and by accumulation of reactive oxygen species (ROS), as previously reported.27 Mutations in RUNX1 and IKZF1, both involved in hematopoietic differentiation, have already been detected in the advanced stages of CML7 and are confirmed here as specific markers for BC progression. Along with this, the XPO1 gene (exportin-1) mutated here in a single patient with the E571K substitution, is also frequently mutated in clonal hematologic disorders, with the E571K mutation widely represented in chronic lymphocytic leukemia.28 SNV analysis showed the presence of a recurrent mutation affecting the UBE2A gene (Xq24) (pt#3 and pt#8). UBE2A is an E2-ubiquitin conjugating enzyme that has never been found mutated in CML. Interestingly, the two patients harboring UBE2A mutations lacked any recognizable copy number alteration (Table 2). WES and targeted resequencing of a broader cohort showed that somatic UBE2A mutations are found in a significant fraction (16.7%) of advanced CML phases, thus confirming the initial exome analysis and suggesting a driver role for UBE2A loss of function during disease progression. The Saccaromyces Cerevisiae UBE2A homolog Rad6 participates in DNA repair, sporulation and cell cycle regula1796

tion;29 in mammals a role for UBE2A in the regulation of transcription and chromatin reorganization through posttranslational histone modifications has recently been hypothesized.30 Germline mutations of the UBE2A gene in humans have been associated with the X-linked Nascimento-type intellectual disability syndrome.31-33 In order to understand the effect of UBE2A mutations in a BCR-ABL1-positive model, we tested the activity of exogenous UBE2A both in the WT or mutated forms (D114V and I33M) in BA/F3 BCR-ABL1-positive cell lines. We observed a reduced amount of mono-ubiquitinated histone H2A, a known UBE2A substrate, after overexpression of mutated UBE2A compared to the WT (Figure 1C), which indicates that the UBE2A mutations analyzed in this study decrease the activity of the enzyme. This result has been further confirmed by in vitro assays for ubiquitination and enzymatic activity on total cell lysates (Figure 1D and E), thus providing evidence of a damaging effect of the two mutations on UBE2A function. Accordingly, one of the four variants identified in our cohort is a N-terminal frameshift mutation, thus supporting this hypothesis. This evidence is further strengthened by the distribution of UBE2A mutations throughout the entire protein, a pattern that is more common for genes undergoing inactivation. Mutations in the UBE2A paralog UBE2B were not detected in this study, which suggests a specific role for UBE2A in chronic myeloid leukemia. Stable silencing of UBE2A in the BCR-ABLpositive K562 cell line or overexpression of the I33M mutated form in a BCR-ABL1-positive 32Dcl3cl3 myeloid cell line showed profound downmodulation of CSF3R, a critical regulator of myeloid lineage differentiation and development.34,35 CSF3R, also known as granulocyte colony-stimulating factor receptor (GCSFR), is a member of the hematopoietin receptor superfamily35 and plays a key role in promoting neutrophilic differentiation but may also support the development of different types of hematopoietic progenitors.34 This suggests a potential role for CSF3R modulation in the suppression of myeloid differentiation in BC. Although the precise mechanism by which UBE2A controls CSF3R expression is still unknown, our data suggest that UBE2A-mediated CSF3R regulation occurs at transcriptional level. Alteration of CSF3R transcription could occur either by a direct activity of UBE2A on CSF3R promoter through epigenetic mechanisms36 or indirectly by UBE2A-mediated ubiquitination of specific transcription factors. Further studies will be needed to clarify this process and to establish the relevance of CSF3R deregulation in the impairment of CML cell differentiation. In line with these findings, we showed that impairment of UBE2A function induces a delay in the differentiation of K562 and 32Dcl3BCR/ABL1 cells after PMA, hydroxyurea or GCSF treatment, suggesting an important role for UBE2A as a modulator of myeloid differentiation. In conclusion, in this work we identified recurrent, somatic UBE2A mutations occurring in a significant proportion of advanced CML cases. We propose that the acquisition of somatic UBE2A mutations affects myeloid developmental pathways, promoting a differentiation blockade. Further studies will be required to thoroughly dissect the molecular mechanisms responsible for these effects and to define possible therapeutic strategies for UBE2A-mutated BC-CML cases. haematologica | 2019; 104(9)

UBE2A somatic variants in CML progression

Acknowledgments The authors would like to thank Manuela Carrera and Giuliana Laurenza for technical assistance. Funding This work was supported by Associazione Italiana Ricerca sul Cancro (IG-14249 to CGP, IG-17727 to RP, IG-22082 to

References 1. Heisterkamp N, Stam K, Groffen J, de Klein A, Grosveld G. Structural organization of the bcr gene and its role in the Ph' translocation. Nature. 1985;315(6022):758-761. 2. Daley GQ, Van Etten RA, Baltimore D. Induction of chronic myelogenous leukemia in mice by the P210bcr/abl gene of the Philadelphia chromosome. Science. 1990;247(4944):824-830. 3. Gambacorti-Passerini C, Antolini L, Mahon FX, et al. Multicenter independent assessment of outcomes in chronic myeloid leukemia patients treated with imatinib. J Natl Cancer Inst. 2011;103(7):553-561. 4. Hochhaus A, Larson RA, Guilhot F, et al. Long-Term Outcomes of Imatinib Treatment for Chronic Myeloid Leukemia. N Engl J Med. 2017;376(10):917-927. 5. Perrotti D, Jamieson C, Goldman J, Skorski T. Chronic myeloid leukemia: mechanisms of blastic transformation. J Clin Invest. 2010;120(7):2254-2264. 6. Gambacorti-Passerini CB, Gunby RH, Piazza R, Galietta A, Rostagno R, Scapozza L. Molecular mechanisms of resistance to imatinib in Philadelphia-chromosome-positive leukaemias. Lancet Oncol. 2003;4(2):75-85. 7. 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. 8. Boultwood J, Perry J, Zaman R, et al. Highdensity single nucleotide polymorphism array analysis and ASXL1 gene mutation screening in chronic myeloid leukemia during disease progression. Leukemia. 2010;24(6):1139-1145. 9. Johansson B, Fioretos T, Mitelman F. Cytogenetic and molecular genetic evolution of chronic myeloid leukemia. Acta Haematol. 2002;107(2):76-94. 10. Kantarjian HM, Keating MJ, Talpaz M, et al. Chronic myelogenous leukemia in blast crisis. Analysis of 242 patients. Am J Med. 1987;83(3):445-454. 11. Quang D, Chen Y, Xie X. DANN: a deep learning approach for annotating the pathogenicity of genetic variants. Bioinformatics. 2015;31(5):761-763. 12. Marega M, Piazza RG, Pirola A, et al. BCR and BCR-ABL regulation during myeloid differentiation in healthy donors and in chronic phase/blast crisis CML patients. Leukemia. 2010;24(8):1445-1449. 13. Mullighan CG, Miller CB, Radtke I, et al. BCR-ABL1 lymphoblastic leukaemia is characterized by the deletion of Ikaros.

haematologica | 2019; 104(9)

RP), by the European Unionâ&#x20AC;&#x2122;s Horizon 2020 Marie SkĹ&#x201A;odowska-Curie Innovative Training Networks (ITN-ETN) with grant agreement No.: 675712CGP and by Giovani Ricercatori #GR-2011-02351167 to AM. CGP is a member of the European Research Initiative for ALK-Related Malignancies (www.erialcl.net). JB acknowledges support from Bloodwise-UK.

Nature. 2008;453(7191):110-114. 14. Puttini M, Coluccia AM, Boschelli F, et al. In vitro and in vivo activity of SKI-606, a novel Src-Abl inhibitor, against imatinibresistant Bcr-Abl+ neoplastic cells. Cancer Res. 2006;66(23):11314-11322. 15. Piazza RG, Magistroni V, Gasser M, et al. Evidence for D276G and L364I Bcr-Abl mutations in Ph+ leukaemic cells obtained from patients resistant to Imatinib. Leukemia. 2005;19(1):132-134. 16. Vardiman JW, Thiele J, Arber DA, et al. The 2008 revision of the World Health Organization (WHO) classification of myeloid neoplasms and acute leukemia: rationale and important changes. Blood. 2009;114(5):937-951. 17. Piazza R, Valletta S, Winkelmann N, et al. Recurrent SETBP1 mutations in atypical chronic myeloid leukemia. Nat Genet. 2013;45(1):18-24. 18. Schuster C, Forster K, Dierks H, et al. The effects of Bcr-Abl on C/EBP transcriptionfactor regulation and neutrophilic differentiation are reversed by the Abl kinase inhibitor imatinib mesylate. Blood. 2003;101(2):655-663. 19. de Miranda NF, Georgiou K, Chen L, et al. Exome sequencing reveals novel mutation targets in diffuse large B-cell lymphomas derived from Chinese patients. Blood. 2014;124(16):2544-2553. 20. Ramensky V, Bork P, Sunyaev S. Human non-synonymous SNPs: server and survey. Nucleic Acids Res. 2002;30(17):3894-3900. 21. Shihab HA, Rogers MF, Gough J, et al. An integrative approach to predicting the functional effects of non-coding and coding sequence variation. Bioinformatics. 2015;31(10):1536-1543. 22. Sung P, Prakash S, Prakash L. The RAD6 protein of Saccharomyces cerevisiae polyubiquitinates histones, and its acidic domain mediates this activity. Genes Dev. 1988;2(11):1476-1485. 23. Kim KW, Kim SH, Lee EY, et al. Extracellular signal-regulated kinase/90KDA ribosomal S6 kinase/nuclear factorkappa B pathway mediates phorbol 12myristate 13-acetate-induced megakaryocytic differentiation of K562 cells. J Biol Chem. 2001;276(16):13186-13191. 24. Park JI, Choi HS, Jeong JS, Han JY, Kim IH. Involvement of p38 kinase in hydroxyureainduced differentiation of K562 cells. Cell Growth Differ. 2001;12(9):481-486. 25. Vogelstein B, Papadopoulos N, Velculescu VE, Zhou S, Diaz LA Jr, Kinzler KW. Cancer genome landscapes. Science. 2013; 339(6127):1546-1558. 26. Duncan BK, Miller JH. Mutagenic deamination of cytosine residues in DNA. Nature.

1980;287(5782):560-561. 27. Stoklosa T, Poplawski T, Koptyra M, et al. BCR/ABL inhibits mismatch repair to protect from apoptosis and induce point mutations. Cancer Res. 2008;68(8):2576-2580. 28. Puente XS, Pinyol M, Quesada V, et al. Whole-genome sequencing identifies recurrent mutations in chronic lymphocytic leukaemia. Nature. 2011;475(7354):101105. 29. Shekhar MP, Lyakhovich A, Visscher DW, Heng H, Kondrat N. Rad6 overexpression induces multinucleation, centrosome amplification, abnormal mitosis, aneuploidy, and transformation. Cancer Res. 2002;62(7):2115-2124. 30. Roest HP, Baarends WM, de Wit J, et al. The ubiquitin-conjugating DNA repair enzyme HR6A is a maternal factor essential for early embryonic development in mice. Mol Cell Biol. 2004;24(12):5485-5495. 31. Haddad DM, Vilain S, Vos M, et al. Mutations in the intellectual disability gene Ube2a cause neuronal dysfunction and impair parkin-dependent mitophagy. Mol Cell. 2013;50(6):831-843. 32. Budny B, Badura-Stronka M, MaternaKiryluk A, et al. Novel missense mutations in the ubiquitination-related gene UBE2A cause a recognizable X-linked mental retardation syndrome. Clin Genet. 2010;77 (6):541-551. 33. Nascimento RM, Otto PA, de Brouwer AP, Vianna-Morgante AM. UBE2A, which encodes a ubiquitin-conjugating enzyme, is mutated in a novel X-linked mental retardation syndrome. Am J Hum Genet. 2006;79(3):549-555. 34. Yang FC, Tsuji K, Oda A, et al. Differential effects of human granulocyte colony-stimulating factor (hG-CSF) and thrombopoietin on megakaryopoiesis and platelet function in hG-CSF receptor-transgenic mice. Blood. 1999;94(3):950-958. 35. Cosman D. The hematopoietin receptor superfamily. Cytokine. 1993;5(2):95-106. 36. Kim J, Guermah M, McGinty RK, et al. RAD6-Mediated transcription-coupled H2B ubiquitylation directly stimulates H3K4 methylation in human cells. Cell. 2009;137(3):459-471. 37. Wu J, Huen MS, Lu LY, et al. Histone ubiquitination associates with BRCA1-dependent DNA damage response. Mol Cell Biol. 2009;29(3):849-860. 38. Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of image analysis. Nat Methods. 2012;9(7):671-675. 39. Piazza R, Ramazzotti D, Spinelli R, et al. OncoScore: a novel, Internet-based tool to assess the oncogenic potential of genes. Sci Rep. 2017;7(7):46290.

1797

ARTICLE Ferrata Storti Foundation

Acute Myeloid Leukemia

Sequential therapy for patients with primary refractory acute myeloid leukemia: a historical prospective analysis of the German and Israeli experience Ron Ram,1 Christof Scheid,2 Odelia Amit,1 Jens Markus Chemnitz,2 Yakir Moshe,1 Michael Hallek,2 Dominik Wolf,3 Irit Avivi1 and Udo Holtick2

Haematologica 2019 Volume 104(9):1798-1803

Bone Marrow Transplantation Unit, Tel Aviv Medical Center and Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel; 2Department I of Internal Medicine, University of Cologne, Cologne, Germany and 3UKIM5, Medical University Innsbruck, Innsbruck, Austria 1

ABSTRACT

Correspondence: RON RAM ronram73@gmail.com Received: August 5, 2018. Accepted: January 28, 2019. Pre-published: February 7, 2019. doi:10.3324/haematol.2018.203869 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1798

rimary refractory acute myeloid leukemia (AML) is associated with a dismal prognosis. The FLAMSA-reduced intensity conditioning protocol (total body irradiation or treosulfan-based) has been described as an effective approach in patients with refractory leukemia undergoing allogeneic hematopoietic cell transplantation. A modified protocol (without amsacrine) has also recently been used. We retrospectively analyzed the transplantation characteristics and outcomes of all consecutive patients between the years 2003 and 2017 (n=51) diagnosed with primary refractory AML who underwent transplantation at the University of Cologne and the Tel Aviv Medical Center. Median age was 54 years and median follow up was 37 months. Median time to neutrophil and platelet engraftment was 13 (range, 8-19) and 13 (range, 7-30) days, respectively. None of the patients had primary graft failure. Incidences of grade 2-4 and grade 3-4 acute graft-versus-host disease (GvHD), overall and moderate-severe chronic GvHD were 50% (95%CI: 41-67%), 12% (95%CI: 3-25%), 61% (95%CI: 47-72%), and 42% (95%CI: 34-51%), respectively. Anti-thymocyte globulin administration was associated with lower incidence of acute GvHD (HR: 0.327; P=0.02). Non-relapse mortality at three months and three years were 6% and 16%, respectively. Relapse incidences were 6% and 29%, respectively. Overall survival rates at three months, three and five years were 90%, 61%, and 53%, respectively. Chronic GvHD disease was associated with a decreased mortality rate (HR: 0.397; P=0.045). We conclude that sequential therapy in patients with primary refractory acute myeloid leukemia is safe and provides a remarkable anti-leukemic effect with durable survival and should be considered for every patient with primary refractory disease.

Introduction ÂŠ2019 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.

1798

Primary refractory acute myeloid leukemia (AML) occurs in approximately 30% of AML patients and is a significant challenge in disease management.1,2 Prognosis is poor and the decision as to how to treat these patients depends on performance status, baseline and concomitant comorbidities, disease characteristics, and patient preference. Salvage chemotherapy, including protocols based on novel agents, results in a dismal median overall survival (OS) of 12.5 months.3 While attempts to reduce the rate of induction failure were seen to have only a modest impact,4-8 the only treatment option shown to prolong survival in patients with primary refractory disease is allogeneic hematopoietic cell transplantation (HCT).9,10 In contrast, a second conventional chemotherapy option offers almost no chance of cure for these patients.11,12 However, outcome after allogeneic HCT remains unfavorable if no response is achieved prior to transplant.10 Moreover, a substantial number of patients will never be able to undergo transplantation. haematologica | 2019; 104(9)

Sequential therapy patients with refractory AML

A sequential therapy approach, based on the FLAMSA protocol was designed to overcome these obstacles.13-15 This both shortened the time to transplant (thus reducing ongoing repeated chemotherapy and neutropenia-associated side effects) and provided significant anti-leukemic activity; it results in remarkable complete remission rates and overall survival. Herein, we describe the GermanIsraeli experience using sequential protocols in patients with AML and primary induction failure, focusing on transplantation outcomes and highlighting the importance of this approach for patients with refractory AML and an available donor.

peting risk. Relapse incidence was evaluated with use of cumulative incidence curves, with NRM treated as a competing risk. The Fine and Gray method was used to evaluate the differences between groups. Multivariate analysis was performed using a Cox proportional hazard regression model for OS and competing risk regression by the Fine and Gray method for NRM and relapse. We used the Duval score to stratify patient populations; patients were scored on a scale of 1-4, considering the fact that all patients were in a primary refractory state.16 SPSS version 23, R 3.1.0 statistical software, and Prism version 5.0 were used for statistical analysis.

Results Methods Only patients with primary refractory AML were included in this study. Patients received either FLAMSA/total body irradiation (TBI), FLAMSA/treosulfan, or FITCy conditioning regimen before HCT. Primary refractory AML was defined as unresponsiveness (at least 20% of blasts in marrow) after at least one course of 7+3 (defined on day 28-35 marrow), four courses of hypomethylating agent, TAD/HAM, or sequential HAM-induction (s-HAM). Percentage of blasts was documented just before the start of conditioning. FLAMSA/ treosulfan instead of FLAMSA/TBI was used in patients older than 65 years or in those with significant comorbidities.15 Patients with relapsed AML were not included in this study. Details about the donor search procedures, the comprehensive treatment, and the supportive care are available in the Online Supplementary Appendix.

Between March 2005 and January 2018, we identified 51 patients who had a primary refractory disease and were treated at the University of Cologne (n=30) or at the Tel Aviv Medical Center (n=21). Table 1 shows patientsâ&#x20AC;&#x2122; characteristics. Median follow up of surviving patients was 37 (2-154) months. Median age was 54 (range, 18-73) years. In the majority of patients, prior chemotherapy was either TAD/HAM or sHAM (n=30, 59%) or 7+3 only (n=11, 22%). Median time from induction chemotherapy to HCT was 84 (25-183) days. The preparative regimen was FLAMSA-TBI (n=13, 26%), FLAMSA-Treosulfan (n=17, 33%) or FITCy-TBI (n=21, 41%). Anti-thymocyte globulin (ATG) was given as part of the conditioning in 6 of 20 (30%) grafts derived from siblings and in 24 of 30 (80%) grafts derived from an unrelated donor. In almost all cases, the graft was derived from peripheral blood stem cells.

Evaluation of response

Early transplantation course

Engraftment was defined as the first of three days with a neutrophil count of >0.5x109/L and a non-transfused platelet count of >20x109/L. Disease response and donor chimerism were assessed at day +30 and day +100 in peripheral blood (PB) and bone marrow (BM). Complete remission (CR) was defined as <5% blast cells in BM by cytomorphology and flow cytometry, and neutrophils of >1.5x109/L in PB. Hematologic relapse was defined by the reappearance of blast cells in the PB, or by >5% blast cells in BM. Death from leukemia was defined as death with refractory disease after transplantation or death from any cause after posttransplantation relapse. Non-relapse mortality (NRM) was defined as death from any cause other than refractory disease or relapse.

All 51 patients were evaluated for early transplantation toxicities. Eleven patients (22%) developed documented infections, including invasive aspergillus (n=2, 4%), gramnegative bacteremia (n=6, 12%), gram-positive bacteremia (n=2, 4%) and C. difficile-associated diarrhea (n=1, 2%). In two patients (4%), the etiology of sepsis could not be determined. Two (4%) patients developed acute respiratory distress syndrome (ARDS) and two (4%) patients developed sinusoidal obstruction syndrome (SOS). In one case, ARDS resulted in early mortality at seven days post transplantation. Data regarding mucositis were available in 21 of 51 patients (41%). Only two (10%) patients experienced grade 3-4 mucositis, whereas five (24%) did not develop any mucositis. Median time to neutrophil engraftment was 13 (range, 8-19) days and median time to platelet engraftment was 13 (range, 7-30) days. None of the patients had primary graft failure. In all patients, analyses of whole marrow chimerism showed 96-100% donorderived cells on day 30 post HCT.

Patients

Statistical analysis All patients gave informed consent to the planned treatment schedule as well as for anonymous data collection. The study was approved by the local ethics committee. Continuous variables were described as the mean, median, standard deviation and range of number of observations, as applicable. Categorical data were described with contingency tables including frequency and percentage. Confidence intervals were calculated at two-sided 95% level of confidence. Two-sided P<0.05 was considered statistically significant. Overall survival (OS) was defined as the time from HCT until the date of death from any cause. For subjects who are still alive, survival data were censored at the last known date of follow up. Disease response and disease progression were assessed according to the previously published response criteria.11 The probabilities of OS were estimated using the Kaplan-Meier method, and the log-rank test was used to evaluate the differences between groups. Probabilities of NRM were estimated with the use of cumulative incidence curves, with relapse treated as a comhaematologica | 2019; 104(9)

Graft-versus-host disease Median onset of grade 2-4 acute GvHD was 26 (range, 7-162) days. Among those patients who developed acute GvHD, involvement of skin, gut, and liver occurred in 17 (50%), 14 (41%), and 10 (29%) of the patients, respectively. By day 100, the cumulative incidences of grade 2-4 and grade 3-4 acute GvHD were 50% (95%CI: 41-67%) and 12% (95%CI: 3-25%), respectively (Figure 1A). In univariate analysis, male gender was associated with a higher incidence of acute GvHD, while older age and ATG administration were associated with a lower incidence of 1799

R. Ram et al.

acute GvHD. In multivariate analysis, only use of ATG had a statistically significant impact on the incidence of acute GvHD (HR 0.327, 95%CI: 0.13-0.82; P=0.02). Median time for the development of chronic GvHD was 9.4 (range, 2.5-12.4) months. At two years post HCT, the cumulative incidence of overall and moderate-severe chronic GvHD were 61% (95%CI: 47-72%) and 42% (95%CI: 34-51%), respectively (Figure 1B). In univariate analysis, ATG administration was associated with lower incidence of overall chronic GvHD, while prior acute GvHD was associated with an increased incidence of chronic GvHD (Table 2). Following multivariate analysis, only prior acute GvHD was associated with a statistically significant impact on chronic GvHD (HR: 1.96, 95%CI: 1.12-3.42; P=0.04).

Non-relapse mortality Ten patients died because of non-relapse mortality. In six patients, death was preceded either by acute or chronic GvHD. There were three deaths within 30 days from HCT: one patient developed grade 3 acute GvHD, one patient who underwent HCT due to transformed chronic myelomonocytic leukemia (CMML) into AML with pretransplant hepato-splenomegaly developed severe SOS, and one patient developed sepsis. The cumulative incidences of NRM at 30 days, three months, one year, and three years were 6%, 6% and 13%, and 16%, respectively (Figure 2A). There were three cases of late (>3 years after transplant) mortality: two patients had prior severe chronic GvHD, while the third, a 73-year old patient, died at the age of 78 with no transplant-associated toxicities. There was no difference between the three protocols in terms of NRM. Because of the low number of events, we did not perform regression analyses for NRM.

Table 1. Patients' characteristcs.

All Cohort (n=51) Domain Sex, n male (%) Age, median, range (years) Median days to HCT, range Cytogenetics Normal karyotype Complex karyotype Other Molecular FLT3-ITD JAK2 BCRABL FLT3-TKD MLL RUNX1 Prior Chemotherapy TAD/HAM sHAM 7+3 only 7+3 plus salvage Azacitidine only % blasts in marrow prior to HCT (range) Preparative regimen FLAMSA-TBI-based FLAMSA-Treo-based FITCy-TBI-based ATG Donor Matched-related Mismatched-related Matched-unrelated Mismatched-unrelated Female-to-male CMV status (D/R) +/+ +/-/+ -/Graft characteristics Peripheral blood (%) CD34, median, range (x106/kg) CD3, median, range (x108/kg)

22 (43%) 54, 18-73 84 (25-183)

Response analysis performed on day 30 showed that only one patient (2% of the 48 patients who survived the

Acute GvHD

28 (55%) 13 (26%) 10 (19%) 14 (28%) 2 (4%) 1 (2%) 6 (12%) 1 (2%) 3 (6%) 7 (14%) 23 (45%) 11 (22%) 8 (13%) 3 (6%) 24 (5-93) %

Chronic GvHD

13 (26%) 17 (33%) 21 (41%) 30 (59%) 19 (37%) 1 (2%) 26 (51%) 5 (10%) 12 (24%) 28 (54%) 5 (10%) 8 (16%) 10 (20%) 50 (98%) 6 (2.6-14.6) 1.9 (1-4.9)

N: number; HCT: hematopoietic cell transplantation; TBI: total body irradiation; D/R: donor/recipient; ATG: anti-thymocyte globulin; CMV: cytomegalovirus.

1800

Disease response and relapse incidence

Figure 1. Incidence of graft-versus-host disease (gvHD). (A) Grade 2-4 and grade 3-4 acute GvHD. (B) Overall and moderate-severe chronic GvHD. HCT: hematopoietic cell transplantation.

haematologica | 2019; 104(9)

Sequential therapy patients with refractory AML

first month after HCT) did not respond to treatment, whereas all other patients obtained complete remission. There were 12 cases of relapse. Relapse incidence at three months, one year, and three years post HCT were 6%, 20%, and 29%, respectively (Figure 2B). There were no cases of late relapses (occurring more than 3 years post HCT). Among the relapsing patients, eight (67%) had a baseline complex karyotype and four (33%) had FLT3-ITD with normal karyotype. There was no difference between the three protocols in terms of incidence of relapse. ATG also did not affect relapse incidence. Patients with baseline complex or monosomal karyotype had a higher risk for relapse (HR: 5.3, 95%CI: 1.6-17.9; P=0.008). Due to the low number of events, we did not perform a full regression analysis for relapse. Intervention at the time of relapse was donor lymphocyte infusion (DLI) (n=5) and a second allogeneic HCT (n=1). Only the patient who had a second transplant survived (73 months post HCT and still alive). Median time from relapse to death among all 11 patients was 1.9 months. Median time from relapse to death in patients given DLI (n=5) was 5.5 months.

Overall survival At the time of data analyses, 23 patients were alive. Incidences of OS at three months, and one, three, and five years post HCT were 90% (95%CI: 84-96%), 71% (95%CI: 60-83%), 61% (95%CI: 54-69%), and 53% (95%CI: 41-66%), respectively (Figure 2C). In univariate analyses, older age and a 9/10 mismatched graft were associated with an increase in the mortality rate, while ATG administration and chronic GvHD were associated with a decreased mortality rate (Table 2). In multivariate analysis, older age remained statistically significant and was associated with increased mortality (HR:

1.32, 95%CI: 1.09-1.71; P=0.046), while only chronic GvHD was associated with decreased mortality (HR: 0.397, 95%CI: 0.07-0.93; P=0.045). Survival was not influenced by time from diagnosis to HCT as a continuous variable, even when we segregated patients into HCT â&#x2030;¤60 days from diagnosis, HCT 61-120 days from diagnosis, and HCT â&#x2030;Ľ120 days from diagnosis (median not reached, 49 months, and 35 months, respectively; P=0.59) (Figure 2D). Survival was also similar between the TBI-FLAMSA, TREO-FLAMSA, and FITCy protocols (1-year OS, 69%, 71%, and 75%, respectively; P=0.91). Patients with Duval score of 4 had a mean OS of 47 (95%CI: 12-82) months, compared to patients with a score of 2 and 3 who had a mean OS of 69 (95%CI: 47-91) months and 61 (95%CI: 44-78) months, respectively (P=0.055).

Discussion We summarized the long-term experience with FLAMSA-based protocols given to patients with primary refractory AML. Median age was 54 years with the oldest patient being 72 years old. Early toxicities were acceptable, with only 4% of the cases developing SOS or ARDS and only one fatal event. Incidence of severe mucositis was rare and occurred in only 10% of the patients. NRM at 30 days was 6%, reflecting the feasibility of the protocol even in older patients, and highlighting the fact that, although all patients experienced prolonged and profound neutropenia, severe infections were not a significant issue. All patients engrafted at a median of 13 days. Acute GvHD occurred in 50% of the patients and 12% developed grade 3-4 acute GvHD. These rates are quite accept-

Table 2. Univariate and multivariate analyses for transplantation outcomes.

Datum Acute GvHD Sex (male) Age ATG Matching (9/10 vs. 10/10) Donor sex (female to male vs. other) Chronic GvHD Sex, male Age ATG Donor sex (female to male vs. other) Prior acute GvHD Overall Mortality Sex (male) Age Time to HCT Conditioning type ATG Matching (9/10 vs. 10/10) Prior acute GvHD Prior chronic GvHD

Univariate analysis 95% CI

Multivariate analysis 95% CI

1.41 0.98 0.39 1.16 1.38

1.10-5.25 0.95-1.00 0.18-0.88 0.39-3.36 0.58-3.29

0.027 0.055 0.02 0.79 0.46

1.75 0.98 0.33

0.55-9.04 0.95-0.99 0.13-0.82

0.13 0.04 0.02

1.2 0.99 0.67 1.7 1.8

0.51-2.82 0.97-1.01 0.29-0.95 0.93-4.45 1.23-4.7

0.66 0.49 0.045 0.08 0.025

0.79

0.44-1.11

0.09

1.96

1.12-3.42

0.04

1.8 1.48 1.00 1.22 0.84 2.25 1.4 0.23

0.76-4.26 1.1-1.9 0.98-1.02 0.42-3.5 0.35-1.01 1.05-6.75 0.44-2.46 0.06-0.83

0.18 0.013 0.23 0.71 0.05 0.048 0.93 0.025

1.32

1.09-1.71

0.046

0.71 2.1

0.22-1.17 0.88-5.3

0.07 0.090

0.397

0.07-0.93

0.045

HR: Hazards Ratios; CI: Confidence Interval; GvHD: graft-versus-host disease; ATG: anti-thymocyte globulin; HCT: hematopoietic cell transplantation. Statistically significant factors in bold.

haematologica | 2019; 104(9)

1801

R. Ram et al.

able, considering the fact that 61% of the patients had an unrelated donor (12% were only 9/10-HLA identical). Nonetheless, there were several cases of late acute GvHD. This may suggest a more cautious approach than the standard post 3-month cyclosporine-tapering schedule. Another important finding in our analysis is the fact that the administration of ATG was found to be associated with a lower incidence of acute GvHD. The impact of ATG on acute GvHD, although still controversial, was demonstrated also by others.17,18 Although the three different protocols used different doses of ATG (15-60 mg/kg), it would appear that there was no significant difference between the various doses, as has been shown previously.19 Interestingly, in our study, the use of ATG was not found to correlate with a lower incidence of chronic GvHD and was not associated with a higher relapse rate. Non-relapse mortality at three months was low (6%) and probably reflects the low incidence of early complications (i.e. severe mucositis and SOS) and of grade 3-4 acute GvHD. The fact that patients with primary refractory AML experience a prolonged and profound neutropenia suggest that this group may benefit from fungal and bacterial prophylaxis, as well as a weekly surveillance testing. This protocol had significant anti-leukemia efficacy

with 98% of the patients responding. Considering that, for the same population, response to salvage chemotherapy is only 30-50%,20 (and is approaching 64% in patients receiving a direct allogeneic HCT3), our results suggest that, for selected patients, the sequential therapy approach is superior. When focusing on durable response and OS, in our cohort that received sequential therapy the 3-year overall survival was 61%. These results are superior to previous reports of patients who had primary refractory disease, achieved complete remission after salvage chemotherapy and proceeded to allogeneic HCT (OS of 48%), to those who were transplanted directly with a refractory disease (OS of 36%), and those with refractory disease who never received a transplant (OS of 25%).3 Yet, although limited by the small sample size, among those patients who had post-transplant relapse, high-risk disease features were common. In addition, patients with a higher Duval score had a shorter OS compared to those with a lower score, and thus our study further validates the original Duval paper.16 In our study, the only factor that was associated with prolonged survival was prior chronic GvHD, suggesting that, although associated with significant morbidity, and sometimes also with increased mortality, in this cohort of high-risk patients, graft-versus-leukemia effect is essential.

Figure 2. Transplantation outcome. (A) Relapse rate, (B) non-relapse mortality, (C) overall survival (OS), and (D) OS by time to hematopoietic cell transplantation (HCT).

1802

haematologica | 2019; 104(9)

Sequential therapy patients with refractory AML

Our study has several limitations. First, as opposed to our definition of primary refractory disease, the recently published European LeukemiaNet (ELN) guidelines recommend considering primary refractory disease after only two induction cycles, irrespective of the schedule used in the second cycle.11 Nevertheless, when comparing previously published literature to our results, these recommendations do not reflect what others have considered to be standard clinical practice. Thus, although a portion of our cohort received only one cycle of induction chemotherapy and may represent a “better prognosis” group, results are still better when compared to other salvage protocols. Second, identifying patients for upfront sequential therapy may be associated with a selection bias for fitter patients. However, the median age of patients in our

References 1. Buchner T, Schlenk RF, Schaich M, et al. Acute Myeloid Leukemia (AML): different treatment strategies versus a common standard arm--combined prospective analysis by the German AML Intergroup. J Clin Oncol. 2012;30(29):3604-3610. 2. Juliusson G, Antunovic P, Derolf A, et al. Age and acute myeloid leukemia: real world data on decision to treat and outcomes from the Swedish Acute Leukemia Registry. Blood. 2009;113(18):4179-4187. 3. Wattad M, Weber D, Dohner K, et al. Impact of salvage regimens on response and overall survival in acute myeloid leukemia with induction failure. Leukemia. 2017;31(6): 1306-1313. 4. Lowenberg B, Ossenkoppele GJ, van Putten W, et al. High-dose daunorubicin in older patients with acute myeloid leukemia. N Engl J Med. 2009;361(13):1235-1248. 5. Fernandez HF, Sun Z, Yao X, et al. Anthracycline dose intensification in acute myeloid leukemia. N Engl J Med. 2009;361(13):1249-1259. 6. Lee JH, Joo YD, Kim H, et al. A randomized trial comparing standard versus high-dose daunorubicin induction in patients with acute myeloid leukemia. Blood. 2011;118 (14):3832-3841. 7. Schlenk RF, Frohling S, Hartmann F, et al. Phase III study of all-trans retinoic acid in previously untreated patients 61 years or older with acute myeloid leukemia. Leukemia. 2004;18(11):1798-1803. 8. Hills RK, Castaigne S, Appelbaum FR, et al. Addition of gemtuzumab ozogamicin to induction chemotherapy in adult patients

haematologica | 2019; 104(9)

10.

11.

12.

13.

14.

cohort was 54 years, which represents a ‘typical’ transplanted leukemia patient. Third, creating and operating a system that can quickly identify donors is challenging and requires substantial resources, which are not always available in all centers. In summary, sequential therapy for patients with primary refractory AML is associated with a substantial antileukemic effect, relatively low toxicity, and 61% 3-year OS. Future protocols should focus on better prevention and control of GvHD, predicting and decreasing relapse incidence in high-risk patients using minimal residual disease tests and post-allogeneic maintenance therapy (i.e. tyrosine kinase inhibitors in patients with FLT3-ITD AML), and prospectively validate these results in a larger cohort of patients.

with acute myeloid leukaemia: a metaanalysis of individual patient data from randomised controlled trials. Lancet Oncol. 2014;15(9):986-996. Craddock C, Labopin M, Pillai S, et al. Factors predicting outcome after unrelated donor stem cell transplantation in primary refractory acute myeloid leukaemia. Leukemia. 2011;25(5):808-813. Schlenk RF, Dohner K, Mack S, et al. Prospective evaluation of allogeneic hematopoietic stem-cell transplantation from matched related and matched unrelated donors in younger adults with high-risk acute myeloid leukemia: German-Austrian trial AMLHD98A. J Clin Oncol. 2010;28 (30):4642-4648. 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. Thol F, Schlenk RF, Heuser M, et al. How I treat refractory and early relapsed acute myeloid leukemia. Blood. 2015;126(3):319327. Schmid C, Schleuning M, Ledderose G, et al. Sequential regimen of chemotherapy, reduced-intensity conditioning for allogeneic stem-cell transplantation, and prophylactic donor lymphocyte transfusion in highrisk acute myeloid leukemia and myelodysplastic syndrome. J Clin Oncol. 2005; 23(24):5675-5687. Chemnitz JM, von Lilienfeld-Toal M, Holtick U, et al. Intermediate intensity conditioning regimen containing FLAMSA, treosulfan, cyclophosphamide, and ATG for allogeneic stem cell transplantation in elder-

15.

16.

17.

18.

19.

20.

ly patients with relapsed or high-risk acute myeloid leukemia. Ann Hematol. 2012; 91(1):47-55. Holtick U, Herling M, Pflug N, et al. Similar outcome after allogeneic stem cell transplantation with a modified FLAMSA conditioning protocol substituting 4 Gy TBI with treosulfan in an elderly population with highrisk AML. Ann Hematol. 2017;96(3):479487. Duval M, Klein JP, He W, et al. Hematopoietic stem-cell transplantation for acute leukemia in relapse or primary induction failure. J Clin Oncol 2010;28(23):37303738. Finke J, Bethge WA, Schmoor C, et al. Standard graft-versus-host disease prophylaxis with or without anti-T-cell globulin in haematopoietic cell transplantation from matched unrelated donors: a randomised, open-label, multicentre phase 3 trial. Lancet Oncol. 2009;10(9):855-864. Kroger N, Solano C, Wolschke C, et al. Antilymphocyte Globulin for Prevention of Chronic Graft-versus-Host Disease. N Engl J Med. 2016;374(1):43-53. Salem G, Ruppert AS, Elder P, et al. Lower dose of antithymocyte globulin does not increase graft-versus-host disease in patients undergoing reduced-intensity conditioning allogeneic hematopoietic stem cell transplant. Leuk Lymphoma. 2015;56(4):10581065. Megias-Vericat JE, Martinez-Cuadron D, Sanz MA, et al. Salvage regimens using conventional chemotherapy agents for relapsed/refractory adult AML patients: a systematic literature review. Ann Hematol. 2018;97(7):1115-1153.

1803

ARTICLE Ferrata Storti Foundation

Haematologica 2019 Volume 104(9):1804-1811

Acute Lymphoblastic Leukemia

Glucocorticoids and selumetinib are highly synergistic in RAS pathway-mutated childhood acute lymphoblastic leukemia through upregulation of BIM Elizabeth C. Matheson,1 Huw Thomas,1 Marian Case,1 Helen Blair,1 Rosanna K. Jackson,1 Dino Masic,1 Gareth Veal,1 Chris Halsey,2 David R. Newell,1 Josef Vormoor1,3 and Julie A.E. Irving1 Newcastle Cancer Centre at the Northern Institute for Cancer Research, Newcastle University, Newcastle upon Tyne; 2Wolfson Wohl Cancer Research Centre, Institute of Cancer Sciences, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow and 3Great North Childrenâ&#x20AC;&#x2122;s Hospital, Newcastle upon Tyne Hospitals NHS Foundation Trust, Newcastle upon Tyne, UK 1

ABSTRACT

Correspondence: JULIE IRVING j.a.e.irving@ncl.ac.uk Received: March 16, 2018. Accepted: January 15, 2019. Pre-published: January 17, 2019. doi:10.3324/haematol.2017.185975 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1804 ÂŠ2019 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.

1804

ew drugs are needed for the treatment of relapsed acute lymphoblastic leukemia and preclinical evaluation of the MEK inhibitor, selumetinib, has shown that this drug has excellent activity in those leukemias with RAS pathway mutations. The proapoptotic protein, BIM is pivotal in the induction of cell death by both selumetinib and glucocorticoids, suggesting the potential for synergy. Thus, combination indices for dexamethasone and selumetinib were determined in RAS pathway-mutated acute lymphoblastic leukemia primagraft cells in vitro and were indicative of strong synergism (combination index <0.2; n=5). Associated pharmacodynamic assays were consistent with the hypothesis that the drug combination enhanced BIM upregulation over that achieved by a single drug alone. Dosing of dexamethasone and selumetinib singly and in combination in mice engrafted with primary-derived RAS pathway-mutated leukemia cells resulted in a marked reduction in spleen size which was significantly greater with the drug combination. Assessment of the central nervous system leukemia burden showed a significant reduction in the drug-treated mice, with no detectable leukemia in those treated with the drug combination. These data suggest that a selumetinib-dexamethasone combination may be highly effective in RAS pathway-mutated acute lymphoblastic leukemia. An international phase I/II clinical trial of dexamethasone and selumetinib (Seludex trial) is underway in children with multiply relapsed/refractory disease.

Introduction Progress in the treatment of childhood acute lymphoblastic leukemia has been exceptional and, using contemporary regimens, sustained remission is achievable in almost 90% of children.1,2 However, the outcome of children who relapse is much poorer and remains a frequent cause of death in children with cancer.3-5 Since further intensification with traditional agents is often associated with significant toxicity and limited success, new therapies are clearly needed. One promising avenue that may deliver novel drugs comes from our previous work showing that mutation in genes which activate the Ras/Raf/Mek/Erk pathway, such as NRAS, KRAS, FLT3, and PTPN11, are highly prevalent in relapsed ALL and, importantly, mutated ALL cells are differentially sensitive to the MEK inhibitor, selumetinib (AZD6244, ARRY-142886).6-8 In contrast, RAS pathway wildtype ALL cells were insensitive to MEK inhibition, both in vitro and in vivo.6 In the IBFMREZ2002 clinical trial for relapsed ALL, RAS pathway mutations were associated with high-risk features such as early relapse, central nervous system (CNS) disease and chemo-resistance and a poorer overall survival was seen in patients with KRAS mutations.6 In haematologica | 2019; 104(9)

Glucocorticoids and selumetinib in RAS pathway-mutated ALL

the UKALLR3 trial, a poorer survival was seen in children with NRAS mutations.7 Thus, this genetic subtype of relapsed ALL clearly warrants exploratory therapies. The Ras/Raf/Mek/Erk cascade regulates diverse cellular functions, including cell proliferation, survival, differentiation, angiogenesis and migration, and is deregulated in numerous cancers, including ALL.9-13 Classic activation is initiated by ligand binding to receptor tyrosine kinases at the cell surface and via Ras, then Raf activates MEK1/2 which has restricted substrate specify for extracellular signal–regulated kinase 1 and 2 (Erk). ERK is a potent kinase with over 200 nuclear and cytoplasmic substrates including transcription factors such as the ETS family and proteins involved in the apoptotic machinery, such as the proapoptotic BIM. Phosphorylation of the predominant form of BIM (BIMEL) by ERK1/2, targets it for ubiquitination and proteasomal degradation and may also directly hinder its interactions with Bax14,15 and selumetinib-induced apoptosis is associated with BIM induction.16 Relapsed ALL is generally more drug resistant than newly diagnosed disease and despite the use of more intensive chemotherapeutic regimens at ALL relapse, there are lower rates of complete remission and end-ofinduction negativity for minimal residual disease.2,3 Assessment of in vitro drug sensitivity of primary ALL samples has shown that blasts at relapse are significantly more resistant to many of the drugs used in upfront treatment protocols, with the highest level of drug resistance seen to glucocorticoids.17,18 Glucocorticoids, such as dexamethasone, are pivotal agents in the treatment of all lymphoid malignancies because of their ability to specifically induce apoptosis in developing lymphocytes and induction of pro-apoptotic BIM is key to this effect.19 Thus, BIM is a common effector in both selumetinib- and dexamethasone-induced apoptosis, suggesting the potential for synergy. In addition, glucocorticoid resistance in ALL has been associated with enhanced activation of the pathway and its inhibition has led to glucocorticoid re-sensitization.20-22 These effects may be more pronounced in the context of RAS pathway-mutated ALL. We, therefore, preclinically evaluated the combination of dexamethasone and selumetinib in vitro and in an orthotopic mouse model engrafted with primary-derived ALL cells and showed pronounced drug synergism in RAS pathway-mutated ALL. These data suggest that this drug combination may

be highly effective in the significant subgroup of patients with this form of leukemia and has led to the Seludex trial, an international phase I/II expansion study on the treatment of relapsed/refractory RAS pathway-mutated ALL.

Methods Compounds and formulation Selumetinib was kindly provided by AstraZeneca (Cheshire, UK). For the in vitro studies, it was dissolved in dimethylsulfoxide to a concentration of 100 mM and stored in single-use aliquots at -20°C. Dexamethasone was purchased from Sigma-Aldrich (Dorset, UK), dissolved in ethanol at 20 mM and stored at -20°C. For in vivo studies, selumetinib was prepared as a suspension in 0.5% hydroxypropyl methylcellulose + 0.1% polysorbate 80.

Patients’ samples

Primagrafts were generated in NOD SCID γ null (NSG) mice using ALL cells from bone marrow samples of children presenting or relapsing with ALL and accessed through the Newcastle Haematology Biobank, after appropriate consent (reference numbers 2002/111 and 07/H0906). Clinical details of the patients are given in Table 1. Mutational screening for RAS pathway mutations and assessment of pathway activation by western blotting of p-ERK was performed as previously described.8,23

In vitro drug sensitivity and synergy Freshly harvested primagraft cells were suspended in RPMI1640 with 15% fetal bovine serum and plated out in triplicate at a density of 5x105 cells/100 μL/well into 96-well plates and treated with a range of concentrations of dexamethasone (0.1 nM to 10 μM) or selumetinib (1 nM to 100 μM). After 96 h, cytotoxicity was assessed using the CellTiter 96 Aqueous One kit (Promega, Southampton, UK). The results were averaged and expressed as a percentage of the cytotoxicity of the control vehicle. Survival curves were plotted and half maximal growth inhibitory values (GI50) were calculated using GraphPad Prism software (GraphPad software Inc., San Diego, CA, USA). Drug combination experiments were analyzed for synergistic, additive, or antagonistic effects using the combination index method developed by Chou and Talalay.24 Briefly, primagraft cells were treated with fixed dose ratios based on the GI50 values of each drug (x0.25, x0.5, x1, x2 and x4) and evaluated by median effect analysis using CalcuSyn software (Cambridge, UK). The dose-effect curve for each drug alone

Table 1. Clinical features of patients and characterization of patient-derived xenografts.

Patient ID

Sex

Age at diagnosis (years)

Cytogenetics

End of induction MRD

Ras pathway mutation

Clonality

pERK

L779 L897a L914

M M F

5.5 16.8 7.3

High hyperdiploid B other High hyperdiploid

Intermediate High risk Low risk

Clonal Clonal Clonal

Positive Positive Positive

L829b relapse L707c LX825 L920 L848

F F F F M

3.1 16.5 14.7 4.4 2.5

High hyperdiploid t(17;19) B other B other t(12;21)

High risk High risk High risk Low risk Low risk

NRAS (Q61R) KRAS (G12D) CBL/FLT3 large del/D836 KRAS (G13D) KRAS (insertion) Wildtype Wildtype Wildtype

Clonal Clonal N/A N/A N/A

Positive Positive Negative Negative Negative

Patient suffered an on-treatment central nervous system relapse. bL829 at diagnosis was NRAS G12D. cPatient relapsed with the same KRAS mutation. B-other group: -; L897 is negative by fluorescence in situ hybridization (FISH) for ETV6-RUNX1, BCR-ABL1, MLL and TCF3-PBX1/HLF. LX825 is negative by FISH for ETV6-RUNX1, BCR-ABL1, MLL, CRLF2, IKZF1, PAX5, IGH and PDGFRB. ID: identity; MRD: minimal residual disease; M: male; F: female; N/A: not available. a

haematologica | 2019; 104(9)

1805

E.C. Matheson et al.

was determined using the median-effect principle and was compared to the effect achieved with a combination of the two drugs to derive a combination index (CI) value.

Pharmacokinetic analyses Plasma pharmacokinetics of selumetinib and dexamethasone were determined using non-compartmental analysis in female CD1 mice after oral dosing. Plasma concentrations of both drugs were measured by liquid chromatographic mass spectrometry (API4000 LCMS/MS, Applied Biosystems, CA, USA), attached to a Perkin Elmer chromatography system (Perkin Elmer Ltd, Beckonsfield, UK) and calibrated using standards prepared in blank mouse plasma. In both cases separation was performed using a Gemini 3μ C18 110A column, (50x3 mm), fitted with a 4x2 mm C18 cartridge (Phenomenex, Macclesfield, UK.

of human leukemia cells reached >1% of total cells, mice were randomized into groups to receive control vehicle (0.5% hydroxypropyl methylcellulose + 0.1% polysorbate 80) or drug treatment (6 mice per group) with dexamethasone, selumetinib or both, administered by oral gavage. Selumetinib was dosed at 25 mg/kg bid, while the dexamethasone dose varied in each study. Tumor burden was monitored weekly by flow cytometry. Pharmacodynamic studies were performed in highly engrafted mice which were dosed for 72 h. Spleens were removed following euthanasia and assessed by flow cytometry to confirm an engraftment of >85%. Cells were lysed and analyzed by western blotting for levels of p-ERK, ERK2, BIM, MCL1 and a-tubulin, as described above. Additional details of the study methods are provided in the Online Supplementary Material.

In vivo experiments All experiments were performed under the UK Home Office NCL-PLL60/4552. Drug efficacy studies were performed as previously described.6 Briefly, primagraft cells were injected intrafemorally and mice were monitored for engraftment every 34 weeks by tail vein bleed. Blood red cells were lysed and analyzed by flow cytometry on a BD FACSCanto II, using antihuman CD10, CD34 and CD19 and anti-mouse CD45 antibodies. Human leukemia cells were gated and their number expressed as a percentage of the total number of nucleated cells. Once the level

Results

The combination of selumetinib and dexamethasone show synergy in vitro in RAS pathway-mutated acute lymphoblastic leukemia and is associated with enhanced induction of BIM To investigate possible synergism, the R3F9 cell line and primagraft ALL cells, with and without Ras pathway mutations (n=8), were treated with dexamethasone,

Figure 1. The combination of selumetinib and dexamethasone shows synergy in vitro in RAS pathway-mutated acute lymphoblastic leukemia and is associated with enhanced levels of BIM. (A) Viability curves of Ras pathway-mutant acute lymphoblastic leukemia (ALL) cells (L829R) with individual drugs and the selumetinib/dexamethasone drug combination. (B) Histogram of combination indices for the selumetinib/dexamethasone combination in wildtype and Ras pathway-mutant ALL cells; mutated genes are shown in brackets. (C) Western analyses of ALL cells (L829R) treated with control vehicle (CV) or GI50 values of selumetinib (10 μM) and dexamethasone (10 μM), singly and in combination. (D) A representative median effect curve (data shown are from L897) after simultaneous drug dosing and with each drug added 24 h prior to the partner drug, followed by a further 72 h incubation. CV: control vehicle; Sel: selumetinib; Dex: dexamethasone; CI: combination index.

1806

haematologica | 2019; 104(9)

Glucocorticoids and selumetinib in RAS pathway-mutated ALL

selumetinib and the drug combination at 0.25x, 0.5x, 1x, 2x and 4x their respective GI50 concentrations and viability data were evaluated by median effect analysis. The CI for all RAS pathway-mutated samples were indicative of strong synergy with a mean of 0.1 (range, 0.02-0.15) (Figure 1A,B and Online Supplementary Figure S1). Synergism was not observed in ALL primagraft cells without RAS pathway activation (CI >1.2). Mechanistic assessments were performed with the GI50 concentration of both drugs for 24 h. As expected, ALL cells treated with selumetinib resulted in almost complete inhibition of ERK phosphorylation and downregulation of MCL1 levels. Dexamethasone treatment also downregulated pERK levels. Treatment with dexamethasone or selumetinib was associated with increased levels of BIM which were further enhanced with the drug combination. A representative western blot and a histogram of the combined densitometry values (n=4 patient-derived xenografts) are shown in Figure 1C and Online Supplementary Figure S2A. The apoptotic marker, cleaved PARP, was enhanced with the drug combination in some, but not all, patient-derived xenograft samples at this time point (Online Supplementary Figure S2B). While the loss in cell viability in non-dividing patient-derived xenograft ALL cells must be due to increased cell death, we also showed enhanced apoptosis with the drug combination for the NRAS-mutated R3F9 cell line (Online Supplementary Figure S2C,D), an effect reduced by BIM knockdown (Online Supplementary Figure S2E,F). There were similar levels of induction of the glucocorticoid receptor target gene, GILZ, in cells treated with both dexamethasone and with the drug combination suggesting that enhanced glucocorticoid receptor transcriptional activity is not a component of the synergism (Online Supplementary Figure S3). Synergism between selumetinib and other drugs for example, gemcitabine, is highly schedule dependent and sequential rather than simultaneous dosing appears optimal.25 Thus, we assessed synergism in primagraft ALL cells dosed simultaneously or with only selumetinib or dexamethasone for 24 h followed by both drugs for an additional 72 h, prior to cell viability assessments. We saw similar synergism across all experimental parameters (Figure 1D) and thus we selected simultaneous drug administration in subsequent in vivo studies.

Pharmacokinetic studies to define a clinically relevant oral dose and exclude drug-drug interactions To determine the optimal oral dose of dexamethasone that will achieve clinically relevant serum levels, pharmacokinetic studies were performed in CD1 mice. Mice (n=27) were dosed with 0.5, 1 and 5 mg/kg dexamethasone by oral gavage. Blood samples were taken after 15 min, 30 min, 1 h, 3 h, 6 h and 24 h and serum dexamethasone levels were analyzed (Online Supplementary Figure S4A). A Tmax of 60 min was observed, with Cmax values of 48.9, 94.7 and 766.5 ng/mL following the 0.5, 1 and 5 mg/kg doses, respectively. Given the reported Cmax average of 40-90 ng/mL in recent UK and American ALL trials, 1 mg/kg was deemed the most appropriate dose level.26,27 Dexamethasone can induce cytochrome P450 forms, including CYP3A4, the principal isoform responsible for selumetinib oxidative metabolism; we therefore performed selumetinib pharmacokinetic analyses, following administration of the drug alone (25 mg/kg) and after coadministration with 1 mg/kg dexamethasone (Online haematologica | 2019; 104(9)

Supplementary Figure S4B,C). A Tmax of 60 min was observed, with Cmax values for selumetinib of 4.74 μg/mL compared to 5.49 μg/mL, respectively (P>0.05, Student t test). Other parameters were also similar (Online Supplementary Figure S4C), indicative of no drug-drug interaction (P>0.05 for all).

Selumetinib and dexamethasone show synergy in vivo and clear central nervous system disease The drug combination was evaluated in vivo and its effects compared to those of single drugs and control vehicle in primagrafts derived from diagnostic ALL (NRAS Q61R and KRAS G12D) and relapse (KRAS G13D) samples. Scheduling and dosing, by oral gavage, are shown in Figure 2A-G. Given the significant weight loss (>20%) associated with dexamethasone, dosing could not be prolonged, even when the dose was lowered from 1 mg/kg bid to 0.25 mg/kg sid. There was no additional toxicity observed in mice given the drug combination. Nevertheless, at the end of the dosing period, there was a significant reduction in spleen size with selumetinib or dexamethasone alone but a statistically greater reduction in mice given the drug combination, with spleen weights approaching those of healthy mice (P<0.001) (combined data are shown in Figure 2G). In addition, brains were assessed for the depth of leukemia infiltration in the leptomeninges. For mice engrafted with L897 and L779 primagraft cells, there was a significant reduction in leukemic infiltration in drug-treated mice, with a mean and standard deviation of 66.3 μ ± 100.6 for animals given the control vehicle, compared to 3.1 μ ± 12.5 for those treated with dexamethasone and 5.37 μ ± 21.475 for those treated with selumetinib (Online Supplementary Figure S5A). Mice treated with the drug combination showed no leukemic infiltration (P<0.05 for all by the Student t test). For L779, there was demonstrable CNS disease once peripheral ALL exceeded 1% i.e. pre-dosing (Online Supplementary Figure S5B). Clearance of CNS disease in mice engrafted with L829R cells was unevaluable because of minimal CNS leukemia in both control and drug-treated mice. The results of the pharmacodynamic assessment of engrafted spleens after short-term dosing were consistent with observations in vitro; inhibition of ERK phosphorylation and lower MCL1 levels associated with selumetinib dosing, similar induction of GILZ with dexamethasone dosing, and modest enhancement of BIM levels with the drug combination (Figure 3A-D). Annexin V binding in circulating ALL cells, as detected by multi-parameter flow cytometry, increased at both 24 h and 48 h in all drug-treated mice and was highest for those given the drug combination (Figure 3E).

Discussion Selumetinib is a potent, selective, allosteric inhibitor of MEK1/2 with demonstrated anti-tumor activity and a favorable toxicity profile. It has progressed to phase III clinical trials for several types of adult solid cancers.28-30 In the pediatric setting, selumetinib has recently undergone phase I clinic testing as a monotherapy in children with BRAF-driven recurrent/refractory pediatric low-grade glioma which defined a maximum tolerated dose of 25 mg/m2/dose bid.31 Sustained responses (1 complete, 7 partial) were observed in some children and selumetinib was well tolerated, with the most common toxicity being rash. 1807

E.C. Matheson et al.

G Figure 2. Selumetinib and dexamethasone show synergy in vivo in RAS pathwaymutated acute lymphoblastic leukemia. (A-F) In vivo drug efficacy studies of single drugs and their combination in RAS pathway mutated-acute lymphoblastic leukemia (ALL) showing dose scheduling and peripheral blood monitoring before and during dosing and spleen weights at the end of dosing for mice with L779-NRAS (A and B, respectively), L897-KRAS (C and D) and L829 relapse- KRAS (E and F) ALL. For L779, mice were dosed with selumetinib at 25 mg/kg and dexamethasone at 1 mg/kg twice daily and then once daily after a recovery period. For L897, the dosage of selumetinib was 25 mg/kg and that of dexamethasone 0.5 mg/kg (bid), with the dexamethasone being increased to 1 mg/kg (sid) following a recovery period. For L829R, selumetinib was dosed at 25 mg/kg (bid) and dexamethasone at 0.25 mg/kg (sid). (G) The mean and standard deviation are shown for combined spleen weight data for all three efficacy experiments (one-way analysis of variance with the Tukey multiple comparison test, ***P<0.001, ****P<0.0001; n=17 mice treated with control vehicle, n=17 treated with selumetinib, n=15 treated with dexamethasone and n=14 treated with the combination. CV: control vehicle; Sel: selumetinib; Dex: dexamethasone.

1808

haematologica | 2019; 104(9)

Glucocorticoids and selumetinib in RAS pathway-mutated ALL

Figure 3. Pharmacodynamic analyses in acute lymphoblastic leukemia cells after drug dosing in vivo support in vitro data. (A,B) Western blot analyses of spleen cells from mice engrafted with RAS pathway-mutant acute lymphoblastic leukemia cells 72 h after dosing: L779; NRAS; 25 mg/kg selumetinib and 1 mg/kg dexamethasone bid (A) and L897; KRAS; 25 mg/kg selumetinib and 0.5 mg/kg dexamethasone bid (B). (C) Histograms of densitometry from western blot analyses, showing mean Âą standard error of mean (SEM) (3-4 mice per treatment) [one-way analysis of variance (ANOVA) with the Tukey multiple comparison test, *P<0.05, **P<0.01]. (D) Relative expression of GILZ mRNA (mean and SEM) in treated mice compared to those given the control vehicle, as quantified by real-time polymerase chain reaction analysis in all three patient-derived xenograft experiments, again after 72 h dosing (ANOVA as before **P<0.01; ns, not significant. (E) Histograms of annexin V-positive ALL cells (mean Âą SEM) determined by flow cytometric analyses of peripheral blood at 24 h and 48 h after dosing (2 mice per group). CV: control vehicle; Sel: selumetinib; Dex: dexamethasone.

haematologica | 2019; 104(9)

1809

E.C. Matheson et al.

In addition, a phase I trial of selumetinib in children with neurofibromatosis type 1 and inoperable plexiform neurofibromas showed partial responses in 17 of 24 children (71%) and no excess toxicity was reported.32 While we have previously shown preclinical activity of selumetinib in ALL, this drug, like other MEK inhibitors, is likely to show maximal therapeutic benefit in combination. Indeed, in phase III clinical trials for advanced nonsmall-cell lung cancer and uveal melanoma, selumetinib has been evaluated in combination with docetaxel and dacarbazine, respectively.33 In this current study, we show significant synergy of selumetinib with the synthetic glucocorticoid dexamethasone in vitro and in an orthotopic mouse model engrafted with RAS pathway-activated primary-derived ALL cells. Importantly, we demonstrate this across a range of cytogenetic subgroups, including high hyperdiploidy, B-other, t(17;19) and t(1;19) ALL. Pharmacokinetic data showed clinically relevant drug levels and optimal scheduling and in vivo pharmacodynamic analyses confirmed an impact on drug targets and apoptosis. Mechanistically, the synergism was associated with enhanced induction of the pro-apoptotic protein, BIM, and decreases in the anti-apoptotic BH3-only protein, MCL1. BIM is a BH3-only protein that binds to anti-apoptotic BCL2 family members, including MCL1 and BCL2, to liberate and directly activate bax and bak, which then elicit caspase-dependent apoptosis. BIM is an effector protein in both glucocorticoid and MEK inhibitor responses and reducing pERK activity enhances BIM levels as well as decreasing MCL1 protein levels by increasing its turnover.34 Therefore, we propose that the drug combination enables BIM to inhibit anti-apoptotic BH3-only proteins more completely and directly activate BAX and BAK. Others have reported a direct effect of MEK inhibition on glucocorticoid receptor transcriptional activity which may also contribute to the synergism, but this did not appear significant in our experiments at the time point chosen.22 Epigenetic regulation of the BIM locus due to acetylation has been described in a subgroup of glucocorticoid-resistant ALL and is associated with BIM under-expression. Such individuals may be expected to have a suboptimal response to the selumetinib/dexamethasone drug combination.35 However, the incidence of acetylated BIM in the relapsed setting and in the context of RAS pathway mutations has not been described to date. Our synergism data are supported by a study by Jones et al., who used an integrated approach to understand glucocorticoid resistance and relapse and identified MAPK pathways as a contributory factor.22 In this study, knockdown of MEK2 or MEK inhibition enhanced responses not only to glucocorticoids but also to other chemotherapeutics and was not dependent on the presence of RAS pathway mutations, a finding suggestive of activation of the pathway through alternative routes. We have previously shown an excellent correlation between pERK activation and the presence of RAS pathway mutations, although we too noted some rare exceptions which in our study were in part explained by the presence of chromosomal translocations, including the Philadelphia chromosome and 11q23.6 Drug synergies have also been shown for MEK inhibitors with both traditional chemotherapeutics such as gemcitabine and targeted agents including PI3K/AKT inhibitors36,37 and the BCL-XL inhibitor, navitoclax (ABT263).38 Inhibiting the other effector pathways of RAS is clearly a rational strategy; however, while we have 1810

observed synergism of MEK and AKT inhibition in RAS pathway-mutated ALL in vitro, the synergism was considerably weaker than that observed with dexamethasone (unpublished observations). In solid cancers, increased levels of BIM protein are also observed with MEK inhibition, but the protein is inactive due to sequestration by high levels of BCL-XL. In the presence of navitoclax, BIM is released, triggering an apoptotic response.38 We have previously reported a reduction of CNS leukemia in selumetinib-treated mice and now confirm this in additional primagraft samples and show complete absence of leukemic infiltration in the leptomeninges of mice treated with the selumetinib/dexamethasone drug combination.6 The identification of CNS disease in mice with similar levels of ALL engraftment prior to drug dosing suggests that the drug combination completely eradicated the leukemia in situ. This is a highly significant finding given the association of RAS pathway mutations and CNS disease at relapse that we previously reported in the IBFMREZ2002 clinical trial and the fact that in contemporary regimens, the proportion of CNS relapses is increasing.39 A key question, relevant to MEK inhibitor therapy, is whether Ras pathway mutations are initiating events in ALL or secondary, cooperating genetic events and there is evidence for both.13 However, for targeted therapies to be successful, the target is ideally present on all tumor cells and we and others have reported that mutations can be subclonal, particularly at diagnosis, and can be gained or lost at relapse.6,40-42 Importantly, we have also shown that mutations at relapse are in the major ALL clone, are often selected from a minor subclone at diagnosis and that apparent ‘loss’ of a Ras pathway mutation can be ‘replacement’ of one for another.6,8,43 This suggests a dependence on the pathway that can be exploited by MEK inhibition and, as we show here, is enhanced with co-exposure to dexamethasone. Based on these promising data, an international phase I/II clinical trial of oral dexamethasone and selumetinib (Seludex) is underway in RAS pathway-mutated, multiply relapsed/refractory ALL. A parallel, national study in adult disease at first relapse is also ongoing, since the prevalence of RAS pathway mutations and association with poor prognosis has also been noted.44 One relevant observation from selumetinib and other Mek inhibitor trials is that the most common toxicity is inflammatory rash. In severe cases, the recommended treatment is oral glucocorticoids and no adverse effects of drug co-administration have been reported.45 Thus, if efficacy is seen in the proposed clinical trials, selumetinib and other Mek inhibitors may be a much needed novel therapy for a substantial number of children with high-risk, relapsed disease. There may also be a role for the drug combination in the upfront treatment of RAS-driven, high-risk ALL, to avert relapse. Acknowledgments The authors gratefully acknowledge Cancer Research UK (project grant to JAEI, HN and JV, number 18780), Bloodwise (previously known as the Leukaemia and Lymphoma Research Fund, project grant to JAEI, number 11007), the North of England Children’s Cancer Research Fund and the Newcastle Haematology Biobank for ALL samples. We are grateful to AstraZeneca for their kind donation of selumetinib. CH is funded by the Chief Scientist Office (ETM/374). We thank Clare Orange and Lynn Stevenson, University of Glasgow and Think Pink, Scotland for help with histology and slide scanning. haematologica | 2019; 104(9)

Glucocorticoids and selumetinib in RAS pathway-mutated ALL

References 1. Vora A, Goulden N, Mitchell C, et al. Augmented post-remission therapy for a minimal residual disease-defined high-risk subgroup of children and young people with clinical standard-risk and intermediate-risk acute lymphoblastic leukaemia (UKALL 2003): a randomised controlled trial. Lancet Oncol. 2014;15(8):809-818. 2. Vora A, Goulden N, Wade R, et al. Treatment reduction for children and young adults with low-risk acute lymphoblastic leukaemia defined by minimal residual disease (UKALL 2003): a randomised controlled trial. Lancet Oncol. 2013;14(3):199-209. 3. Parker C, Waters R, Leighton C, et al. Effect of mitoxantrone on outcome of children with first relapse of acute lymphoblastic leukaemia (ALL R3): an open-label randomised trial. Lancet. 2010;376(9757):2009-2017. 4. Hof J, Krentz S, van Schewick C, et al. Mutations and deletions of the TP53 gene predict nonresponse to treatment and poor outcome in first relapse of childhood acute lymphoblastic leukemia. J Clin Oncol. 2011;29(23):3185-3193. 5. Malempati S, Gaynon PS, Sather H, La MK, Stork LC, Children's Oncology Group. Outcome after relapse among children with standard-risk acute lymphoblastic leukemia: Children's Oncology Group study CCG1952. J Clin Oncol. 2007;25(36):5800-5807. 6. Irving J, Matheson E, Minto L, et al. Ras pathway mutations are prevalent in relapsed childhood acute lymphoblastic leukemia and confer sensitivity to MEK inhibition. Blood. 2014;124(23):3420-3430. 7. Moorman AV, Irving J, Enshaei A, et al. Composite index for risk prediction in relapsed childhood acute lymphoblastic Leukaemia. Haematologica. 2015;100(s1): 195-196. 8. Case M, Matheson E, Minto L, et al. Mutation of genes affecting the RAS pathway is common in childhood acute lymphoblastic leukemia. Cancer Res. 2008;68 (16):6803-6809. 9. Chung E, Kondo M. Role of Ras/Raf/MEK/ERK signaling in physiological hematopoiesis and leukemia development. Immunol Res. 2011;49(1-3):248-268. 10. Pylayeva-Gupta Y, Grabocka E, Bar-Sagi D. RAS oncogenes: weaving a tumorigenic web. Nat Rev Cancer. 2011;11(11):761-774. 11. Ahearn IM, Haigis K, Bar-Sagi D, Philips MR. Regulating the regulator: post-translational modification of RAS. Nat Rev Mol Cell Biol. 2012;13(1):39-51. 12. Ward AF, Braun BS, Shannon KM. Targeting oncogenic Ras signaling in hematologic malignancies. Blood. 2012;120(17):33973406. 13. Knight T, Irving JA. Ras/Raf/MEK/ERK Pathway activation in childhood acute lymphoblastic leukemia and its therapeutic targeting. Front Oncol. 2014; 4:160. 14. Balmanno K, Cook SJ. Tumour cell survival signalling by the ERK1/2 pathway. Cell Death Differ. 2009;16(3):368-377. 15. Ley R, Ewings KE, Hadfield K, Cook SJ. Regulatory phosphorylation of Bim: sorting out the ERK from the JNK. Cell Death Differ. 2005;12(8):1008-1014. 16. Meng J, Fang B, Liao Y, Chresta CM, Smith PD, Roth JA. Apoptosis induction by MEK inhibition in human lung cancer cells is medi-

haematologica | 2019; 104(9)

ated by Bim. PLoS One. 2010;5(9):e13026. 17. Hongo T, Fujii Y. In vitro chemosensitivity of lymphoblasts at relapse in childhood leukemia using the MTT assay. Int J Hematol. 1991;54(3):219-230. 18. Klumper E, Pieters R, Veerman AJ, et al. In vitro cellular drug resistance in children with relapsed/refractory acute lymphoblastic leukemia. Blood. 1995;86(10):3861-3868. 19. Lu J, Quearry B, Harada H. p38-MAP kinase activation followed by BIM induction is essential for glucocorticoid-induced apoptosis in lymphoblastic leukemia cells. FEBS Lett. 2006;580(14):3539-3544. 20. Rambal AA, Panaguiton ZL, Kramer L, Grant S, Harada H. MEK inhibitors potentiate dexamethasone lethality in acute lymphoblastic leukemia cells through the proapoptotic molecule BIM. Leukemia. 2009;23(10):1744-1754. 21. Polak A, Kiliszek P, Sewastianik T, et al. MEK inhibition sensitizes precursor B-cell acute lymphoblastic leukemia (B-ALL) cells to dexamethasone through modulation of mTOR activity and stimulation of autophagy. PLoS One. 2016;11(5):e0155893. 22. 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. 23. Nicholson L, Knight T, Matheson E, et al. Casitas B lymphoma mutations in childhood acute lymphoblastic leukemia. Genes Chromosomes Cancer. 2012;51(3):250256. 24. Chou TC, Talalay P. Quantitative analysis of dose-effect relationships: the combined effects of multiple drugs or enzyme inhibitors. Adv Enzyme Regul. 1984;22:2755. 25. Xu J, Knox JJ, Ibrahimov E, et al. Sequence dependence of MEK inhibitor AZD6244 combined with gemcitabine for the treatment of biliary cancer. Clin Cancer Res. 2013;19(1):118-127. 26. Yang L, Panetta JC, Cai X, et al. Asparaginase may influence dexamethasone pharmacokinetics in acute lymphoblastic leukemia. J Clin Oncol. 2008;26(12):1932-1939. 27. Jackson RK, Irving JAE, Veal GJ. Pharmacokinetics of standard versus short high-dose dexamethasone therapy in childhood acute lymphoblastic leukemia: results from the UKALL 2011 trial. Cancer Res. 2016;76(14 Suppl):CT115. 28. Bennouna J, Lang I, Valladares-Ayerbes M, et al. A phase II, open-label, randomised study to assess the efficacy and safety of the MEK1/2 inhibitor AZD6244 (ARRY-142886) versus capecitabine monotherapy in patients with colorectal cancer who have failed one or two prior chemotherapeutic regimens. Invest New Drugs. 2011;29(5):1021-1028. 29. Davies BR, Logie A, McKay JS, et al. AZD6244 (ARRY-142886), a potent inhibitor of mitogen-activated protein kinase/extracellular signal-regulated kinase kinase 1/2 kinases: mechanism of action in vivo, pharmacokinetic/pharmacodynamic relationship, and potential for combination in preclinical models. Mol Cancer Ther. 2007;6 (8):2209-2219. 30. Janne PA, Shaw AT, Pereira JR, et al. Selumetinib plus docetaxel for KRASmutant advanced non-small-cell lung cancer: a randomised, multicentre, placebo-con-

31.

32.

33. 34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

trolled, phase 2 study. Lancet Oncol. 2013;14(1):38-47. Banerjee A, Jakacki R, Onar-Thomas A, et al. A phase 1 study of AZD6244 in children with recurrent or refractory low-grade gliomas: a Pediatric Brain Tumor Consortium report. J Clin Oncol. 2014;32(5s):10065. Dombi E, Baldwin A, Marcus LJ, et al. Activity of selumetinib in neurofibromatosis type 1-related plexiform neurofibromas. N Engl J Med. 2016;375(26):2550-2560. Website-https://clinicaltrials.gov/. Domina AM, Vrana JA, Gregory MA, Hann SR, Craig RW. MCL1 is phosphorylated in the PEST region and stabilized upon ERK activation in viable cells, and at additional sites with cytotoxic okadaic acid or taxol. Oncogene. 2004;23(31):5301-5315. Bachmann PS, Piazza RG, Janes ME, et al. Epigenetic silencing of BIM in glucocorticoid poor-responsive pediatric acute lymphoblastic leukemia, and its reversal by histone deacetylase inhibition. Blood. 2010;116(16): 3013-3022. Engelman JA, Chen L, Tan X, et al. Effective use of PI3K and MEK inhibitors to treat mutant Kras G12D and PIK3CA H1047R murine lung cancers. Nat Med. 2008;14 (12):1351-1356. Haagensen EJ, Kyle S, Beale GS, Maxwell RJ, Newell DR. The synergistic interaction of MEK and PI3K inhibitors is modulated by mTOR inhibition. Br J Cancer. 2012;106(8): 1386-1394. Corcoran RB, Cheng KA, Hata AN, et al. Synthetic lethal interaction of combined BCL-XL and MEK inhibition promotes tumor regressions in KRAS mutant cancer models. Cancer Cell. 2013;23(1):121-128. Krishnan S, Wade R, Moorman AV, et al. Temporal changes in the incidence and pattern of central nervous system relapses in children with acute lymphoblastic leukaemia treated on four consecutive Medical Research Council trials, 1985-2001. Leukemia. 2010;24(2):450-459. Molteni CG, Te Kronnie G, Bicciato S, et al. PTPN11 mutations in childhood acute lymphoblastic leukemia occur as a secondary event associated with high hyperdiploidy. Leukemia. 2010;24(1):232-235. Mullighan CG, Zhang J, Kasper LH, et al. CREBBP mutations in relapsed acute lymphoblastic leukaemia. Nature. 2011;471 (7337):235-239. Ma X, Edmonson M, Yergeau D, et al. Rise and fall of subclones from diagnosis to relapse in pediatric B-acute lymphoblastic leukaemia. Nat Commun. 2015;6:6604. Tartaglia M, Martinelli S, Cazzaniga G, et al. Genetic evidence for lineage-related and differentiation stage-related contribution of somatic PTPN11 mutations to leukemogenesis in childhood acute leukemia. Blood. 2004;104(2):307-313. Trinquand A, Tanguy-Schmidt A, Ben Abdelali R, et al. Toward a NOTCH1/FBXW7/RAS/PTEN-based oncogenetic risk classification of adult T-cell acute lymphoblastic leukemia: a Group for Research in Adult Acute Lymphoblastic Leukemia study. J Clin Oncol. 2013;31 (34):4333-4342. Welsh SJ, Corrie PG. Management of BRAF and MEK inhibitor toxicities in patients with metastatic melanoma. Ther Adv Med Oncol. 2015;7(2):122-136.

1811

ARTICLE Ferrata Storti Foundation

Haematologica 2019 Volume 104(9):1812-1821

Acute Lymphoblastic Leukemia

Asparagine levels in the cerebrospinal fluid of children with acute lymphoblastic leukemia treated with pegylated-asparaginase in the induction phase of the AIEOP-BFM ALL 2009 study

Carmelo Rizzari,1^ Claudia Lanvers-Kaminsky,2^ Maria Grazia Valsecchi,3 Andrea Ballerini,4 Cristina Matteo,4 Joachim Gerss,5 Gudrun Wuerthwein,2 Daniela Silvestri,3 Antonella Colombini,1 Valentino Conter,1 Andrea Biondi,1 Martin Schrappe,6 Anja Moericke,6 Martin Zimmermann,7 Arend von Stackelberg,8 Christin Linderkamp,9 Michael C. Frühwald,10 Sabine Legien,11 Andishe Attarbaschi,12 Bettina Reismüller,12 David Kasper,13 Petr Smisek,14 Jan Stary,14 Luciana Vinti,15 Elena Barisone,16 Rosanna Parasole,17 Concetta Micalizzi,18 Massimo Zucchetti4* and Joachim Boos2*

Pediatric Hematology-Oncology Unit, Department of Pediatrics, University of MilanoBicocca, MBBM Foundation, Monza, Italy; 2Department of Pediatric Hematology and Oncology, University Childrens’ Hospital of Münster, Münster, Germany; 3Medical Statistics Unit, Department of Clinical Medicine and Prevention, University of Milano-Bicocca, Milan, Italy; 4Department of Oncology, Laboratory of Cancer Pharmacology, Istituto di Ricerche Farmacologiche Mario Negri IRCCS, Milan, Italy; 5Institute of Biostatistics and Clinical Research, University of Münster, Münster, Germany; 6Department of Pediatrics, University Medical Center Schleswig-Holstein, Campus Kiel, Kiel, Germany; 7Department of Pediatric Hematology and Oncology, Hannover Medical School, Hannover, Germany; 8Pediatric Hematology and Oncology, Charité, Berlin, Germany; 9Pediatric Hematology and Oncology, Hannover Medical School, Hannover, Germany; 10Children’s Hospital, Augsburg, Germany; 11 Pediatrics 5 (Oncology, Hematology, Immunology); Stuttgart Cancer Center; Klinikum Stuttgart – Olgahospital, Stuttgart, Germany; 12Department of Pediatric Hematology and Oncology, St. Anna Children's Hospital, Vienna, Austria; 13Department of Pediatrics and Adolescent Medicine, Medical University of Vienna, Vienna, Austria; 14Czech Paediatric Haematology/Oncology, Charles University and University Hospital Motol, Prague, Czech Republic; 15Department of Pediatric Hemato-Oncology, Ospedale Bambino Gesù, Rome, Italy; 16Department of Pediatric Hemato-Oncology, Regina Margherita Children’s Hospital, Turin, Italy; 17Department of Pediatric Hematology-Oncology, Ospedale Pausillipon, Naples, Italy and 18Department of Pediatric Hematology-Oncology, IRCCS I.G. Gaslini, Genova, Italy. 1

Correspondence: CARMELO RIZZARI c.rizzari@asst-monza.it Received: September 14, 2018. Accepted: January 31, 2019. Pre-published: January 31, 2019. doi:10.3324/haematol.2018.206433 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1812 ©2019 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.

1812

^CR and CL-K share first authorship. *MZ and JB share last authorship.

ABSTRACT

sparagine levels in cerebrospinal fluid and serum asparaginase activity were monitored in children with acute lymphoblastic leukemia treated with pegylated-asparaginase. The drug was given intravenously at a dose of 2,500 IU/m2 on days 12 and 26. Serum and cerebrospinal fluid samples obtained on days 33 and 45 were analyzed centrally. Since physiological levels of asparagine in the cerebrospinal fluid of children and adolescents are 4-10 μmol/L, in this study asparagine depletion was considered complete when the concentration of asparagine was ≤0.2 μmol/L, i.e. below the lower limit of quantification of the assay used. Over 24 months 736 patients (AIEOP n=245, BFM n=491) and 903 cerebrospinal fluid samples (n=686 on day 33 and n=217 on day 45) were available for analysis. Data were analyzed separately for the AIEOP and BFM cohorts and yielded superimposable results. Independently of serum asparaginase activity levels, cerebrospinal fluid asparagine levels were significantly reduced during the investigated study phase but only 28% of analyzed samples showed complete asparagine depletion while relevant levels, ≥1 µmol/L, were still detectable in around 23% of them. Complete cerebrospinal fluid asparagine depletion was found in around 5-6% and 3337% of samples at serum asparaginase activity levels <100 and ≥1,500 IU/L, haematologica | 2019; 104(9)

Effects of PEG-asparaginase on CSF asparagine levels

respectively. In this study cerebrospinal fluid asparagine levels were reduced during pegylated-asparaginase treatment, but complete depletion was only observed in a minority of patients. No clear threshold of serum pegylated-asparaginase activity level resulting in complete cerebrospinal fluid asparagine depletion was identified. The consistency of the results found in the two independent data sets strengthen the observations of this study. Details of the treatment are available in the European Clinical Trials Database at https://www.clinicaltrialsregister.eu/ctr-search/trial/2007-004270-43/IT.

Introduction Asparaginase is one of the major anticancer drugs used in the treatment of acute lymphoblastic leukemia (ALL). The enzyme reduces the levels of asparagine in serum by hydrolyzing it to aspartic acid and ammonia. Currently there are three commercially available asparaginase products.1 The oldest one is the purified native enzyme extracted from Escherichia coli, subsequently also available in a polyethylene glycol conjugated form (PEG-asparaginase) commonly used as the first-line preparation in the treatment of children with ALL throughout Europe and USA. A third asparaginase product derived from Erwinia chrysanthemi (ERW-asparaginase) exists and, due to its structural differences with respect to the E. coli asparaginase products, has been primarily used as a second-line treatment in children with hypersensitivity reactions to the E. coli products.2 Since leukemic cells need exogenous asparagine for their survival much more than the normal host cells do, the depletion of asparagine in serum serves as a surrogate for the anti-leukemic action of asparaginase, no matter which type of product is used. Due to this mechanism of action and to the pharmacodynamic ability of asparaginase products to reduce asparagine pools also in the cerebrospinal fluid (CSF), it has been questioned whether profound and prolonged asparagine depletion, as that determined in the serum, could be of relevance in preventing central nervous system (CNS) relapses.3 Of course, it is exceedingly difficult to ascertain the role of a single drug in the prevention of ALL relapses, especially in an extramedullary compartment such as the CNS where relapses are quite rare. However, in a previous study, patients with higher CSF asparagine levels (>1 µmol/L) during asparaginase treatment were more likely to have isolated CNS relapse.4 Available studies reporting data on CSF asparagine depletion during asparaginase treatment have been mostly performed in limited cohorts of patients and using different asparaginase products, schedules and assays. In the past it has been consistently reported that profound and prolonged CSF asparagine depletion in children treated with standard induction chemotherapy treatment schedules5–9 is achieved when native forms of asparaginase are used. To this end the conceptual question on how asparaginase products may determine asparagine depletion in the CSF remains unanswered. One possible explanation for the asparagine depletion observed in the CSF could lie in a continuous balance between the serum and CSF asparagine pools.10,11 Another possible explanation is that, at peak levels, very small amounts of asparaginase products could cross the blood-brain barrier;12 however, activity levels have never been directly measured in the CSF during the administration of native forms of asparaginase. Given that PEG-asparaginase has a far greater molecular weight than that of the native forms of asparaginase, it is conceivable that it is even more difficult for the pegylated form to cross the blood-brain barrier. Different haematologica | 2019; 104(9)

results have been reported in patients treated with PEGasparaginase wherein detectable CSF asparagine levels have been almost invariably reported thus suggesting that pharmacodynamic differences exist between the different asparaginase products.4,13–15 We very recently demonstrated, even with the limitations of the experimental preclinical model adopted, that in the CSF of rats asparaginase activity levels could be measured, consistently even if transiently, only for non-pegylated formulations.16 In the international AIEOP-BFM ALL 2009 trial protocol (https://www.clinicaltrialsregister.eu/ctr-search/trial/2007004270-43/IT), conducted by members of the Associazione Italiana di Ematologia e Oncologia Pediatrica (AIEOP) and Berlin-Frankfurt-Münster (BFM) group, children with newly diagnosed ALL have been treated with multiple antileukemic agents, including PEG-asparaginase as the first-line preparation. Since PEG-asparaginase has been used in the AIEOP-BFM ALL 2009 study protocol for the first time as a front-line asparaginase agent instead of the previously used native E. coli asparaginase product and since two different randomized studies, consisting of PEG-asparaginase-intensified schedules, are the most relevant treatment questions of the AIEOP-BFM ALL 2009 study, a tight therapeutic drug monitoring study of PEGasparaginase treatment has been implemented to better understand the pharmacological phenomena underlying asparaginase treatment in this therapeutic context. The main findings of the above-mentioned therapeutic drug monitoring specifically related to CSF asparagine levels and asparaginase serum activity, analyzed in parallel after the administration of PEG-asparaginase in the induction phase of the AIEOP-BFM ALL 2009 study, are the focus of the present report.

Methods Patients’ eligibility and treatment schedule

Children ≥1 year and <18 years old diagnosed with ALL and eligible for the AIEOP-BFM ALL 2009 protocol were investigated in this study. PEG-asparaginase (Oncaspar®, Shire) was given during the induction phase (namely protocol IA) to children diagnosed and treated in the participating centers. The drug was given intravenously as a 2 h infusion at the dosage of 2,500 IU/m2 with a maximum dose of 3,750 IU/m2 on days 12 and 26. Details of the treatment are available in the European Clinical Trials Database at https://www.clinicaltrialsregister.eu/ctr-search/trial/2007-004270-43/IT. CSF asparagine levels were evaluated when lumbar punctures relevant for this part of the PEG-asparaginase study were scheduled in protocols IA and the subsequent consolidation phase protocol IB, i.e., on protocol days +33 and +45, respectively, which correspond to days 7 and 19 after the second PEG-asparaginase dose of protocol IA. Serum asparaginase activity levels were measured at the same two time points. Mainly because of the long half-life of PEG-asparaginase and the slow decay of the related activity levels, and in order to have a larger set of samples to be analyzed, data 1813

C. Rizzari et al.

analyses were focused on these two time points only but with the following adjustments: (i) CSF samples were considered analyzable when collected at a distance of 7±3 days and of 19±3 days after the second PEG-asparaginase dose (day +26) of protocol IA; (ii) the serum samples had to be collected the same day (±1) as the CSF samples. In the text, tables and figures of this report the CSF asparagine values obtained at day +33 ±3 and day +45 ±3 are simply referred to as day +33 and day +45.

Sample collection Serum and CSF sample collection started on June 1st, 2010; CSF collection ended on December 31st, 2012. Samples were collected from patients treated according to the AIEOP-BFM ALL 2009 protocol in Italy (AIEOP), Germany (BFM-G), Austria (BFM-A), and the Czech Republic (CPH). CSF samples were immediately frozen at -80°C, shipped on dried ice and stocked at -80°C until amino acid analysis. Blood samples from a peripheral vein or central venous catheter were collected according to the treatment schedule. Serum was separated in 2 mL tubes and immediately frozen at -20°C until shipment. CSF asparagine levels and asparaginase serum activity were determined in the Laboratory of Cancer Pharmacology at the Department of Oncology of the “Mario Negri” Pharmacology Research Institute IRCCS (Milan, Italy) for the AIEOP samples, in the Clinical Pharmacology Laboratory of the Department of Pediatric Hematology and Oncology at the University Hospital of Muenster (Germany) for the BFM-G and CPH samples and in the Department of Pediatrics and Adolescent Medicine of the Medical University of Vienna (Austria) for the BFM-A CSF samples.

Determination of serum pegylated-asparaginase activity PEG-asparaginase activity was evaluated with the commercially available enzymatic medac asparaginase activity test (MAAT) (medac GmbH, Hamburg, Germany) or with the L-aspartic βhydroxamate (AHA) test.17

The medac asparaginase activity test Briefly, the MAAT is an IVD-CE-certified test which is commercially available. It is a homogeneous microplate assay that analyses the catalytically active asparaginase in serum by detecting the amount of hydrolyzed substrate analogue of asparagine, quantified by photometric reading at 700 nm. The assay uses calibrators containing a native enzyme preparation from E. coli (ASP medac) and has a lower limit of quantification (LLOQ) of 30 U/L. All the values below the LLOQ were considered 0 U/L for the statistical analysis. The MAAT was used for the determination of asparaginase activity in the serum samples of AIEOP patients.

The L-aspartic β-hydroxamate test

AHA is the substrate for the quantification of native E. coli, pegylated E. coli, and Erwinia chrysanthemii asparaginase in human serum. Asparaginase hydrolyzes AHA to L-aspartic acid and hydroxylamine, which is determined at 710 nm after condensation with 8-hydroxyquinoline and oxidation to indooxine. The LLOQ is 5 U/L.17 All the values below the LLOQ were considered 0 U/L for the statistical analysis. The AHA test was used for the determination of asparaginase activity in serum of CPH and BFMG samples. Since the AHA test calibrates against known amounts of PEG-asparaginase in contrast to the MAAT, which uses native E. coli asparaginase as the calibrator, it considers the different substrate turnover rates of PEG-asparaginase compared to native E. coli asparaginase under the assay conditions. Thus, the PEGasparaginase activity determined by the MAAT is a mean of 1.42 higher than that determined by the AHA test, as recently demonstrated.18 1814

Determination of cerebrospinal fluid asparagine levels CSF asparagine levels were measured using a high performance liquid chromatographic technique after derivatization with ophthaldialdehyde as described by Turnell and Cooper19 and already used in previous pharmacological studies performed by the AIEOP-BFM group.15,20 The LLOQ was 0.2 μmol/L and all the analyzed data with results below this limit were considered 0 μmol/L for the statistical analysis. Since bloody CSF punctures might have altered the quantification of asparagine, either though the release of asparagine present in the erythrocytes or through possible contamination by asparaginase, CSF samples contaminated with blood were excluded from the analysis.

Informed consent All patients and their parents or legal guardians signed appropriate informed consent for the biological study procedure encompassed in the AIEOP-BFM ALL 2009 study for the asparaginase therapeutic drug monitoring. Assent was given by patients according to ethical standards and national guidelines. Protocol studies were approved by each national and local review board, in accordance with the Declaration of Helsinki and national laws.

Statistical analysis Descriptive analyses include the distribution of patients’ characteristics and dot plots on CSF asparagine concentration and PEGasparaginase serum activity. Box plots and scatter plots were used to describe continuous values, with the Wilcoxon test to compare medians. Data are presented separately on the original scale according to the type of enzymatic test used, which was MAAT for AIEOP samples and AHA for all other samples, collectively identified as BFM samples.

Results Between June 2010 and December 2012 1,764 patients were unselectively enrolled in Italian, German, Czech, and Austrian centers adopting the AIEOP-BFM ALL 2009 study protocol. Overall, 736 patients were included in the present study, 245 of whom belonged to the AIEOP cohort and 491 to the BFM cohort. Their main biological and clinical characteristics are presented in Table 1. The distribution of these characteristics is superimposable to that of the entire cohort of 1,764 patients enrolled in the study in the same period (data not shown). The total number of CSF samples collected in the two groups was 903. Overall, 903 CSF samples were collected on days 33 and/or 45, of which 314 in the AIEOP cohort and 589 in the BFM cohort.

Asparagine levels in cerebrospinal fluid Of the 903 CSF samples analyzed for asparagine levels, 686 (AIEOP n=230 and BFM n=456) were collected on protocol day +33 and 217 (AIEOP n=84 and BFM n=133) on protocol day +45. The distribution of different CSF asparagine levels detected at the CSF punctures (on days +33 and +45) is presented in Table 2 and Figure 1 (A and B for the AIEOP and BFM cohorts, respectively). Given that the physiological concentration of asparagine in the CSF of children and adolescents ranges between 4 and 10 μmol/L, the CSF asparagine levels found in this study were overall quite consistently reduced at both time points, as depicted in Figure 1. Independently of the levels of asparaginase activity, CSF asparagine levels were significantly reduced during the investigated study phase but haematologica | 2019; 104(9)

Effects of PEG-asparaginase on CSF asparagine levels

only 28% of analyzed samples showed complete asparagine depletion (i.e. below the LLOQ) while relevant levels (≥1 μmol/L) were still detectable in around 23% of them. In particular, asparagine levels ≥1 μmol/L were detected at days +33 and +45 in 16.9% and 34.6% (AIEOP, Table 2A) and in 18.9% and 41.4% (BFM, Table 2B) of analyzed CSF samples, respectively. Median levels were significantly higher at day +45 than at day +33 (AIEOP, P<0.001; BFM, P<0.001).

Asparaginase activity in serum and correlation with cerebrospinal fluid asparagine levels Overall there were 753 serum samples (AIEOP n=271 and BFM n=482) corresponding to the available CSF samples of which 574 (AIEOP n=198 and BFM n=376) were collected on day +33 and 179 (AIEOP n=73 and BFM n=106) on day +45. The mean PEG-asparaginase activity levels measured in these serum samples were 1,839 (±685)

Table 1. Main biological and clinical characteristics at the onset of acute lymphoblastic leukemia in 736 patients enrolled in the AIEOPBFM ALL 2009 protocol with at least one cerebrospinal fluid sample. Data are reported separately for the cohorts of patients belonging to the Associazione Italiana di Ematologia e Oncologia Pediatrica (AIEOP) and Berlin-Frankfurt-Münster (BFM) groups.

AIEOP Number of patients Gender Male Female Age 1-5 years 6-9 years 10-14 years 15-17 years WBC (x 106/L) <20 20-100 ≥100 Final risk group T non HR BCP SR BCP MR HR CNS status CNS 1/2 CNS 3 Not known Immunophenotype BCP T Not known Genetics TEL-AML positive MLL-AF-4 positive CSF sample available At day +33 only At day +45 only At both days

BFM

245

100

491

100

149 96

60.8 39.2

291 200

59.3 40.7

145 40 43 17

59.2 16.3 17.6 6.9

272 96 78 45

55.4 19.6 15.9 9.2

164 60 21

66.9 24.5 8.6

315 112 64

64.2 22.8 13.0

18 67 100 60

7.3 27.4 40.8 24.5

45 164 172 110

9.2 33.4 35.0 22.4

234 3 8

95.5 1.2 3.3

426 16 49

86.8 3.3 10.0

214 31

87.3 12.7

414 76 1

84.5 15.5

IU/L and 314 (±266) IU/L at days +33 and +45 in AIEOP samples (MAAT) and 1,226 (±470) IU/L and 222 (±141) IU/L in BFM samples (AHA test), respectively. PEGasparaginase activity <100 IU/L was found in 1.9% of serum samples (2.5% in AIEOP and 1.6% in BFM) taken at day +33 and in 19.6% of serum samples (17.8% in AIEOP and 20.8% in BFM) taken at day +45 (Table 5A,B) As shown in Table 3A (for the AIEOP cohort) and 3B (for the BFM cohort), at serum asparaginase activity levels <100 IU/L, 100% and 89.3% of the respective patients had CSF asparagine levels >0.2 μmol/L, while this rate decreased to approximately 70% at asparaginase activity ≥100 IU/L. Figures 2 and 3 (subdivided in A and B for the AIEOP and BFM cohorts, respectively) show the CSF asparagine levels at days +33 and +45 in relationship to the asparaginase activity detected at the same time points. At serum asparaginase activity levels lower than 100 IU/L the median CSF asparagine concentration was higher than that in the cohorts with serum activity above 100 IU/L and below 500 IU/L (P<0.003 for AIEOP and P<0.002 for BFM) and that with an overall activity level above 100 IU/L (P<0.001). When asparaginase activity and CSF asparagine levels were determined using progressively higher activity level intervals, even at high asparaginase serum levels of ≥1,500 IU/L, CSF asparagine levels below the LLOQ were found in roughly one-third of the samples (Table 3A and B), indicating that the asparagine level in the CSF was above the LLOQ even at higher levels of asparaginase activity in serum.

Table 2. Distribution of asparagine levels in cerebrospinal fluid, at each sampling point (days +33 and +45) in the (A) Associazione Italiana di Ematologia e Oncologia Pediatrica (AIEOP) and (B) Berlin-FrankfurtMünster (BFM) cohorts. A

Day +33

Day +45

N N. of samples with CSF asparagine levels (μM) <LLOQ >0.2 ≤0.5 >0.5 ≤1 >1 ≤4 >4 Mean* (SD) Median* (IQR)

230 77 74 40 33 6

33.5 32.2 17.4 14.3 2.6

0.61 (1.2) 0.30 (0-0.76)

84 20 23.8 18 21.4 17 20.2 24 28.6 56.0 1.12 (1.6) 0.58 (0.22-1.43)

Day +33

41 1

16.7 0.4

96 2

19.6 0.4

161 15 69

65.7 6.1 28.2

358 35 98

72.9 7.1 20.0

AIEOP: Associazione Italiana di Ematologia e Oncologia Pediatrica; BFM: BerlinFrankfurt-Münster. WBC: white blood cells; HR: high risk; BCP: B-cell precursor; SR: standard risk; MR: medium risk; CNS: central nervous system; CSF: cerebrospinal fluid.

haematologica | 2019; 104(9)

N. of samples with CSF asparagine levels (μM) <LLOQ >0.2 ≤0.5 >0.5 ≤1 >1 ≤4 >4 Mean* (SD) Median* (IQR)

Day +45

456 133 138 99 76 10

29.2 30.3 21.7 16.7 2.2

133 26 30 22 52 3

0.83(2.4) 0.4 (0-0.85)

19.6 22.6 16.5 39.1 2.3 0.9 (0.9) 0.75 (0.3-1.3)

(*) samples with values below the lower limit of quantification are assigned a value of 0. CSF: cerebrospinal fluid; LLOQ: lower limit of quantification; SD: standard deviation; IQR: interquartile range.

1815

C. Rizzari et al.

Of note, the data on the relationship between serum asparaginase activity and CSF asparagine levels obtained in the two cohorts of patients were consistent within subgroups defined by clinical and biological characteristics, such as sex, age, white blood cell count at diagnosis and immunophenotype (Table 4). Table 5 shows asparagine levels in CSF according to

whether the asparaginase activity in the serum was less than 100 IU/L or above this level at days +33 and +45 in (A) AIEOP and (B) BFM cohorts. The asparagine levels in CSF samples were significantly higher when serum asparaginase activity was below 100 IU/L than when it was higher than 100 IU/L in both cohorts (AIEOP and BFM).

Figure 1. Asparagine levels in the cerebrospinal fluid during the induction phase. Box plots of asparagine levels (μmol/L) of cerebrospinal (CSF) punctures scheduled on days +33 and +45 in the (A) Associazione Italiana di Ematologia e Oncologia Pediatrica (AIEOP) and (B) Berlin-Frankfurt-Münster (BFM) cohorts.

Table 3. Asparagine levels in cerebrospinal fluid punctures scheduled on days +33 and +45 sorted according to respective serum asparaginase activity levels in the (A) Associazione Italiana di Ematologia e Oncologia Pediatrica (AIEOP) and (B) Berlin-Frankfurt-Münster (BFM) cohorts. A

0 <100 N

N. of samples with CSF and serum data 18 6.6 CSF ASN Concentration <LLOQ 0 0.0 >0.2 ≤0.5 2 11.1 >0.5 ≤1 4 22.2 >1 ≤4 6 33.3 >4 6 33.3 Mean* (SD) 3.36 (3.1) Median* (IQR) 2.08 (0.57-6.35)

PEG-asparaginase activity in serum (IU/L) ≥100 <500 ≥500 <1000 ≥1000 <1500 N % N % N % 50

18.5

10 20.0 10 20.0 10 20.0 17 34.0 3 6.0 1.12 (1.4) 0.63 (0.28-1.44)

7.0

8 42.1 7 36.8 3 15.8 1 5.3 0 0.0 0.35 (0.5) 0.25 (0-0.4)

14.7

≥1500 %

144

12 30.0 10 22.5 9 22.5 9 22.5 0 0.0 0.59 (0.6) 0.39 (0-0.89)

53.1

48 33.3 52 36.1 26 18.1 18 12.5 0 0.0 0.41 (0.5) 0.29 (0-0.65)

Total N

271 78 28.8 81 29.9 52 19.2 51 18.8 9 3.3 0.75 (1.3) 0.36 (0-0.87)

0 <100 N

N. of samples with CSF and serum data 28 5.8 CSF asparagina concentration <LLOQ 3 10.7 >0.2 ≤0.5 3 10.7 >0.5 ≤1 8 28.6 >1 ≤4 12 42.9 >4 2 7.1 Mean* (SD) 1.50 (1.52) Median* (IQR) 1.02 (0.56-2.12)

PEG-asparaginase activity in serum (IU/L) ≥100 <500 ≥500 <1000 ≥1000 <1500 N % N % N % 90

18.7

21 23.3 27 30.0 15 16.7 27 30.0 0 0.0 0.67 (0.59) 0.45 (0.22-1.10)

116

24.1

31 26.7 38 32.8 28 24.1 18 15.5 1 0.9 0.70 (2.03) 0.40 (0-0.83)

142

29.4

42 29.6 46 32.4 33 23.2 21 14.8 0 0.0 0.49 (0.48) 0.40 (0-0.80)

≥1500 %

106

22.0

39 36.8 35 33.0 18 17.0 13 12.3 1 0.9 0.43 (0.59) 0.30 (0.01-0.65)

Total N

482 136 28.2 149 30.9 102 21.2 91 18.9 4 0.8 0.62 (1.17) 0.40 (0.01-0.88)

(*) samples with values below the lower limit of quantification are assigned a value of 0. PEG: pegylated; CSF: cerebrospinal fluid; LLOQ: lower limit of quantification; SD: standard deviation; IQR: interquartile range.

1816

haematologica | 2019; 104(9)

Effects of PEG-asparaginase on CSF asparagine levels

Discussion Survival rates obtained with chemotherapy schedules applied over the last three decades in childhood ALL have increased progressively and currently approach 90%.3 Asparaginase has been shown to play a key role in obtaining such excellent results and for this reason the drug has been invariably included, since its introduction in clinical practice, in the polychemotherapy schedules designed for childhood ALL. The enzyme exerts its antileukemic activity by depleting the systemic pools of asparagine, an amino acid essential for the rapid proliferation of malignant lymphoblasts.21–23 It is well known that the activity of any asparaginase product may be pharmacologically monitored by measuring serum asparaginase activity levels, which reflect the enzyme’s ability to deplete circulating asparagine pools and are considered a surrogate marker of its clinical efficacy.1 There is currently a wide agreement that activity lev-

els of at least 0.1 IU/mL (i.e. 100 IU/L) should be achieved and maintained during the whole planned asparaginase treatment to ensure maximal asparagine depletion in serum (<0.2 μmol/L) and maximal therapeutic efficacy.9,21 Whether a similarly profound and prolonged asparagine depletion in the CSF is needed to prevent CNS relapses is not known; however, there are reports associating the two phenomena.4 The physiological concentration of asparagine in human CSF varies depending on the age of the patient. In children and adolescents its concentration ranges between 4 and 10 μmol/L.24 Although these values are lower than those found in serum (about 40-80 μmol/L), it has been reported that these levels ensure the growth of leukemic blasts.25 For this reason, in principle, it is reasonable to study asparagine depletion in CSF to evaluate its clinical relevance better and prospectively. The main purpose of the study reported here was to evaluate the level of asparagine depletion in the CSF and

Figure 2. Effect of pegylated-asparaginase activity on asparagine level in the cerebrospinal fluid. Distribution of asparagine levels (μmol/L) detected in cerebrospinal (CSF) samples collected during puncture on days +33 (blue dots) and +45 (red dots) versus pegylated asparaginase (PEG-ASP) activity levels (IU/L) detected in serum at the same sampling day (±1) in the (A) Associazione Italiana di Ematologia e Oncologia Pediatrica (AIEOP) and (B) Berlin-Frankfurt-Münster (BFM) cohorts.

Figure 3. Cerebrospinal fluid asparagine concentration at different levels of pegylated-asparaginase activity. Box plots of asparagine cerebrospinal fluid (CSF) concentrations (μmol/L) versus categorized pegylated-asparaginase (PEG-ASP) activity levels (IU/L) in serum collected at the same sampling points (±1 day) in the (A) Associazione Italiana di Ematologia e Oncologia Pediatrica (AIEOP) and (B) Berlin-Frankfurt-Münster (BFM) cohorts.

haematologica | 2019; 104(9)

1817

C. Rizzari et al.

therefore the pharmacodynamic effect exerted by PEGasparaginase on the CNS of children with ALL. The interest in this aspect derives from various studies showing that different degrees of asparagine depletion in the CSF may depend not only on the asparaginase product used4–9 but, apparently, also on the levels of asparaginase activity achieved in serum.26–28 In this study the administration of PEG-asparaginase in the induction phase of the ALL treatment adopted was associated with a significant, even if widely variable, reduction of CSF asparagine below the physiological levels. However, complete asparagine depletion was observed overall in only about 28% of the analyzed CSF samples. Considerable CSF asparagine levels, greater than 1 μmol/L, and thus higher than the LLOQ of the assay used, which is considered the threshold of complete asparagine depletion, were detected overall in 23% of the analyzed CSF samples, with this latter figure becoming 17-19% and 35-41% for samples taken 7 and 19 days,

respectively, after PEG-asparaginase administration (Table 2A,B). The findings of the large therapeutic drug monitoring program regarding asparaginase activity levels in serum conducted in the frame of the AIEOP-BFM ALL 2009 study and here reported regard exclusively the subset of patients studied to evaluate asparagine depletion in the CSF. These data show that after a dose of 2,500 IU/m2, PEG-asparaginase activity levels much higher than 100 IU/L are achieved in induction in the vast majority of patients and are maintained both 7 and 14 days following a standard administration schedule including PEGasparaginase, thus fully confirming the prolonged half-life of this product (Table 5A,B). In the serum samples collected along with CSF samples, PEG-asparaginase activity less than 100 IU/L was only detected in around 5-6% of samples, when the samples were taken ≤19 days after administration of PEG-asparaginase (Tables 3A,B and 5A,B). We can, therefore, assume

Table 4. Asparagine levels in cerebrospinal fluid sorted by asparaginase activity levels in serum, at scheduled cerebrospinal fluid punctures (days +33 and +45) in the (A) Associazione Italiana di Ematologia e Oncologia Pediatrica (AIEOP) and (B) Berlin-Frankfurt-Münster (BFM) cohorts. A

Asparagine levels in CSF (µmol/L) Serum PEG-ASP activity Serum PEG-ASP activity 0 <100 IU/L ≥100 IU/L Mean Median Mean Median (SD) (IQR) (SD) (IQR) Gender Male Female Age at diagnosis 1-9 years 10-17 years WBC count at diagnosis <20 x 106/L ≥20 x 106/L Immunophenotype B lineage T ALL

Total Mean (SD)

Median (IQR)

4.0 (3.5) 2.6 (2.6)

3.0 (0.8-7.7) 1.7 (0.5-4.6)

0.7 (0.8) 0.4 (0.8)

0.4 (0.2-0.9) 0.2 (0-0.6)

0.9 (1.4) 0.6 (1.2)

0.4 (0.2-1.0) 0.3 (0-0.7)

3.3 (3.0) 3.4 (4.4)

2.1 (0.6-6.4) 1.4 (0.4-8.5)

0.6 (0.8) 0.5 (0.8)

0.4 (0-0.8) 0.3 (0-0.7)

0.8 (1.3) 0.7 (1.3)

0.4 (0-1.0) 0.3 (0-0.8)

3.5 (3.0) 3.1 (3.5)

2.8 (0.5-6.4) 1.4 (0.6-7.7)

0.5 (0.8) 0.6 (0.9)

0.3 (0-0.7) 0.4 (0-1.0)

0.7 (1.2) 0.8 (1.5)

0.4 (0-0.8) 0.4 (0-1.0)

2.7 (2.5) 6.5 (4.5)

2.0 (0.5-4.8) 8.5 (1.3-9.6)

0.6 (0.8) 0.6 (0.7)

0.3 (0-0.8) 0.3 (0-0.9)

0.7 (1.1) 1.2 (2.2)

0.4 (0-0.8) 0.4 (0-1.4)

Asparagine levels in CSF (µmol/L) Serum PEG-ASP activity Serum PEG-ASP activity 0 <100 IU/L ≥100 Mean Median Mean Median (SD) (IQR) (SD) (IQR) Gender Male Female Age at diagnosis 1-9 years 10-17 years WBC count at diagnosis <20 x 106/L ≥20 x 106/L Immunophenotype B lineage T ALL

Total Mean (SD)

Median (IQR)

1.7 (1.8) 1.1 (0.5)

1.3 (0.5-2.2) 0.9 (0.7-1.2)

0.5 (0.6) 0.6 (1.6)

0.4 (0.0-0.8) 0.4 (0.0-0.8)

0.6 (0.8) 0.6 (1.5)

0.4 (0.0-0.9) 0.4 (0.0-0.9)

1.7 (1.7) 0.8 (0.6)

1.3 (0.6-2.2) 0.9 (0.4-1.1)

0.6 (1.3) 0.5 (0.5)

0.4 (0.0-0.9) 0.4 (0.0-0.7)

0.7 (1.3) 0.5 (0.5)

0.4 (0.0-0.9) 0.4 (0.0-0.8)

1.8 (1.8) 1.0 (0.8)

1.3 (0.7-2.1) 1.0 (0.4-1.4)

0.6 (1.4) 0.5 (0.5)

0.4 (0.0-0.8) 0.4 (0.0-0.8)

0.7 (1.4) 0.6 (0.6)

0.4 (0.0-0.9) 0.4 (0.0-0.9)

1.4 (1.3) 1.9 (2.2)

1.0 (0.5-2.1) 1.3 (0.9-1.6)

0.6 (1.2) 0.5 (0.5)

0.4 (0.0-0.8) 0.4 (0.0-0.8)

0.6 (1.2) 0.6 (0.8)

0.4 (0.0-0.9) 0.4 (0.0-0.9)

CSF: cerebrospinal fluid; PEG-ASP: pegylated asparaginase; SD: standard deviation; IQR: interquartile range; WBC: white blood cell; ALL: acute lymphoblastic leukemia.

1818

haematologica | 2019; 104(9)

Effects of PEG-asparaginase on CSF asparagine levels

that serum asparagine levels of patients enrolled in this pharmacological study were continuously depleted for at least 14 days after each PEG-asparaginase dose in the majority of the patients. In our study when CSF asparagine levels were evaluated in relationship to serum asparaginase activity levels, a considerable number of samples was found not to have levels below the LLOQ. In patients with serum asparaginase activity levels below 100 IU/L (which is considered insufficient to consistently obtain complete asparagine depletion in serum) only 6.5% of the corresponding CSF samples had levels below the LLOQ. When the level of asparaginase activity was 100 IU/L or higher, 70% of the samples had asparagine levels higher than 0.2 μmol/L (i.e. higher than the LLOQ). Furthermore, at serum asparaginase activity levels of 100 IU/L or higher – including samples with activity levels greater than 1500 IU/L - only about one third of the corresponding CSF samples had asparagine levels below the LLOQ. Some studies have been conducted in the past on aspects related to CSF asparagine depletion along with administration of different asparaginase products. Dibenedetto et al. evaluated CSF asparagine levels 3 days after the administration of the fourth dose of ERWTable 5. Asparagine levels in cerebrospinal fluid by levels of asparaginase activity in serum, at the cerebrospinal fluid sampling points (days +33 and +45) in the (A) Associazione Italiana di Ematologia e Oncologia Pediatrica (AIEOP) and (B) Berlin-Frankfurt-Münster (BFM) cohorts. A

PEG-asparaginase activity in serum (IU/L) 0 <100 ≥100 Total N % N % N % N. of samples with 18 6.6 CSF and serum data CSF asparagine levels at day +33 N. 5 Mean* (SD) (μmol/L) 5.9 (3.3) Median* (μmol/L) 6.59 IQR (μmol/L) 4.85-7.65 CSF asparagine levels at day +45 N. 13 Mean* (SD) (µmol/L) 2.4 (2.5) Median* (μmol/L) 1.44 IQR (μmol/L) 0.53-2.76

253

93.4

271

193 0.4 (0.5) 0.31 0-0.71

198 0.6 (1.1) 0.32 0-0.76

60 1.0 (1.3) 0.60 0.22-1.30

73 1.2 (1.7) 0.61 0.3-1.4

PEG-asparaginase activity in serum (IU/L) 0 <100 ≥10 Total N % N % N % N. of samples with 28 5.8 CSF and serum data CSF asparagine levels at day +33 N. 6 Mean* (SD) (μmol/L) 1.50 (2.39) Median* (μmol/L) 0.68 IQR (μmol/L) 0.37-0.91 CSF asparagine levels at day +45 N. 22 Mean* (SD) (μmol/L) 1.50 (1.27) Median* (μmol/L) 1.27 IQR (μmol/L) 0.67-2.13

454

94.2

482

370 0.54 (1.22) 0.38 0.01-0.79

376 0.56 (1.25) 0.38 001-0.79

84 0.69 (0.60) 0.49 0.23-1.12

106 0.86 (0.84) 0.68 0.32-1.23

PEG: pegylated; CSF: cerebrospinal fluid; SD: standard deviation; IQR: interquartile range.

haematologica | 2019; 104(9)

asparaginase (given at a dosage of 10,000 IU/m2 intramuscularly every 72 h) and found them to be below the LLOQ (≤0.2 μmol/L) in 75% of treated children. Despite the small number of cases analyzed in that experience and based on the fact that asparaginase is believed not to be able to cross the barrier between blood and CSF, it was concluded that this phenomenon was the reflection of the asparagine depletion observed in serum.7 On the other hand, Ahlke et al. showed that 2,500 IU/m2 un-pegylated E. coli asparaginase led to complete depletion of CSF asparagine 2 or 3 days after application. Median trough plasma activity levels in this dose-group were 106 IU/L (26-349 IU/L).29 Among the findings of our study, we observed that even at asparaginase activity levels greater than 1500 IU/L CSF asparagine levels below the LLOQ were found in roughly 30 to 40% of the samples (Table 3). In the following years, two additional studies investigated this phenomenon in patients treated with asparaginase. The first, conducted in 1996 by Gentili et al.,5 evaluated 44 patients with newly diagnosed ALL treated in the induction phase of a BFM-oriented protocol with 10,000 IU/m2 of ERW-asparaginase every 3 days. The analysis of CSF and serum asparagine levels, measured on average 3 days after each dose, revealed CSF asparagine levels similar to those reported in the previously reported study by Dibenedetto et al.7 The second study, performed by Rizzari et al.,6 compared the ability of ERW-asparaginase and native E. coli asparaginase to deplete asparagine in the CSF. In all the 62 patients treated in the induction phase with either intravenous or intramuscular ERW-asparaginase or native E. coli asparaginase (10,000 IU/m2 every 3 days), asparagine levels in both the serum and CSF remained below the LLOQ (≤0.2 μmol/L) even if asparaginase activity levels were higher in the group treated with E. coli asparaginase than in that treated with ERW-asparaginase. Similar results were found in a study by Woo et al.8 A different trend has been found in studies performed so far in patients treated with PEG-asparaginase. Vieira Pinheiro et al.30 studied patients treated with PEG-asparaginase within the German Cooperative Acute Lymphoblastic Leukemia (COALL) study and Rizzari et al.15 patients treated with the same product within the AIEOP ALL 2000 study. Overall, both studies showed that CSF asparagine levels in patients treated with PEGasparaginase were undetectable (i.e., below the detection limit) only in a fraction of patients, no matter if serum asparaginase activity levels were much higher than 100 IU/L. Additional studies reported by the Nordic Society of Pediatric Hematology and Oncology and even more recently by the Dutch Childhood Leukemia Study Group (DCLSG) confirmed these observations.14,27,31,32 Based on the most updated scientific evidence it is not possible to provide a clear and incontrovertible explanation on how asparaginase products may achieve the observed asparagine depletion in the CSF. It has been hypothesized that the asparagine depletion observed in the CSF could result from a continuous balance between the serum and CSF asparagine pools.10,11 Another possible explanation can be inferred from the specific physicochemical properties of the native asparaginase products compared to the PEG-asparaginase product. It is conceivable that native asparaginase formulations, given their lower molecular weight and steric size, might have some capacity to penetrate, even in very low amounts, into the CSF thus providing local asparaginase activity. 1819

C. Rizzari et al.

Nevertheless, it has been postulated that this activity never exceeds 0.2% of that present in serum.11 Conversely, this may not be possible for PEG-asparaginase, mainly because of its tertiary structure.16,32,33 However, so far there is no clear proof that any asparaginase product determines any degree of CSF asparagine depletion in humans by directly penetrating the CSF. To contribute to this issue a preclinical study was recently conducted to evaluate whether the three commercially available asparaginase formulations have different abilities to enter the CSF and reduce local asparagine levels. Even with the limitations of the model used in that preclinical experience, the enzymatic activity measured in CSF demonstrated that asparaginase products, in particular both the native forms derived from Erwinia chrysanthemi and E. coli, may transiently penetrate the CNS when administered at high doses, whereas the PEG-asparaginase product does not, most probably because of the differences in molecular weight.16,34,35 To conclude, the findings of the therapeutic drug monitoring performed in our study and reported here indicate that: (i) the administration of PEG-asparaginase was able to cause a broad reduction of physiological CSF asparagine levels (normally 4-10 μmol/L) but complete asparagine depletion was observed overall in only about 28% of the analyzed CSF samples; (ii) CSF asparagine levels greater than 1 μmol/L (thus higher than the LLOQ of the assay adopted) were detectable in 23% of the analyzed samples; (iii) at serum asparaginase activity levels less than 100 IU/L only 6.5% of the CSF samples had asparagine levels below

References 1. Labrou NE, Papageorgiou AC, Avramis VI. Structure-function relationships and clinical applications of L-asparaginases. Curr Med Chem. 2010;17(20):2183–2195. 2. Pieters R, Hunger SP, Boos J, et al. Lasparaginase treatment in acute lymphoblastic leukemia: a focus on Erwinia asparaginase. Cancer. 2011;117(2):238–249. 3. Pui C-H, Campana D, Pei D, et al. Treating childhood acute lymphoblastic leukemia without cranial irradiation. N Engl J Med. 2009;360(26):2730–2741. 4. Avramis VI, Sencer S, Periclou AP, et al. A randomized comparison of native Escherichia coli asparaginase and polyethylene glycol conjugated asparaginase for treatment of children with newly diagnosed standard-risk acute lymphoblastic leukemia: a Children’s Cancer Group study. Blood. 2002;99(6):1986–1994. 5. Gentili D, Conter V, Rizzari C, et al. LAsparagine depletion in plasma and cerebro-spinal fluid of children with acute lymphoblastic leukemia during subsequent exposures to Erwinia L-asparaginase. Ann Oncol. 1996;7(7):725–730. 6. Rizzari C, Zucchetti M, Conter V, et al. Lasparagine depletion and L-asparaginase activity in children with acute lymphoblastic leukemia receiving i.m. or i.v. Erwinia c. or E. coli L-asparaginase as first exposure. Ann Oncol. 2000;11(2):189–193. 7. Dibenedetto SP, Di Cataldo A, Ragusa R, Meli C, Lo Nigro L. Levels of L-asparagine in CSF after intramuscular administration

1820

10.

11.

12. 13.

14.

the LLOQ; and (iv) at serum asparaginase activity levels of 100 IU/L and higher, up to 1,500 IU/L and beyond, CSF asparagine levels were lower than the LLOQ in only about 33-37% of the samples. Thus, a further increase of the PEG-asparaginase dose would not help to obtain complete CSF asparagine depletion. The consistent results found in the two independent data sets presented here strengthen the observations inferred from this study. Acknowledgments The authors thank the patients and families who participated in this trial, the physicians and nurses of all hospitals for their contribution in performing this study, and the members of the AIEOP-BFM ALL 2009 Asparaginase Working Party for productive discussions during the development and progress of the study. They also thank the partners in the reference laboratories and all the technicians for their expert work in cytology, genetics, and MRD diagnostics; and the data managers for their careful study conduction. Funding This study was supported by Comitato M. L. Verga and Fondazione Tettamanti (Monza) and by Stiftung Deutsche Krebshilfe (Bonn). The TDM program performed in the international AIEOPBFM ALL 2009 study has been supported by an unrestricted grant from the Shire company (and previously from medac GmBH, Sigma-Tau, Baxalta, which marketed the drug during the period of the present study).

of asparaginase from Erwinia in children with acute lymphoblastic leukemia. J Clin Oncol. 1995;13(2):339–344. Woo MH, Hak LJ, Storm MC, et al. Cerebrospinal fluid asparagine concentrations after Escherichia coli asparaginase in children with acute lymphoblastic leukemia. J Clin Oncol. 1999;17(5):1568–1568. Pinheiro JPV, Boos J. The best way to use asparaginase in childhood acute lymphatic leukaemia--still to be defined? Br J Haematol. 2004;125(2):117–127. Schwartz MK, Lash ED, Oettgen HF, Tomato FA. L-asparaginase activity in plasma and other biological fluids. Cancer. 1970;25(2):244–252. Riccardi R, Holcenberg JS, Glaubiger DL, Wood JH, Poplack DG. L-asparaginase pharmacokinetics and asparagine levels in cerebrospinal fluid of rhesus monkeys and humans. Cancer Res. 1981;41(11 Pt 1):4554–4558. Müller HJ, Boos J. Use of L-asparaginase in childhood ALL. Crit Rev Oncol Hematol. 1998;28(2):97–113. Avramis VI, Panosyan EH. Pharmacokinetic /pharmacodynamic relationships of asparaginase formulations The past, the present and recommendations for the future. Clin Pharmacokinet. 2005;44(4): 367–393. Appel IM, Pinheiro JP V, den Boer ML, et al. Lack of asparagine depletion in the cerebrospinal fluid after one intravenous dose of PEG-asparaginase: a window study at initial diagnosis of childhood ALL. Leukemia. 2003;17(11):2254–2256.

15. Rizzari C, Citterio M, Zucchetti M, et al. A pharmacological study on pegylated asparaginase used in front-line treatment of children with acute lymphoblastic leukemia. Haematologica. 2006;91(1):24– 31. 16. Ballerini A, Moro F, Nerini IF, et al. Pharmacodynamic effects in the cerebrospinal fluid of rats after intravenous administration of different asparaginase formulations. Cancer Chemother Pharmacol. 2017;79(6):1267–1271. 17. Lanvers C, Vieira Pinheiro JP, Hempel G, Wuerthwein G, Boos J. Analytical validation of a microplate reader-based method for the therapeutic drug monitoring of Lasparaginase in human serum. Anal Biochem. 2002;309(1):117–126. 18. Lanvers-Kaminsky C, Rüffer A, Würthwein G, et al. Therapeutic drug monitoring of asparaginase activity-method comparison of MAAT and AHA test used in the international AIEOP-BFM ALL 2009 trial. Ther Drug Monit. 2018;40(1):93–102. 19. Turnell DC, Cooper JD. Rapid assay for amino acids in serum or urine by pre-column derivatization and reversed-phase liquid chromatography. Clin Chem. 1982;28 (3):527–531. 20. Boos J, Werber G, Ahlke E, et al. Monitoring of asparaginase activity and asparagine levels in children on different asparaginase preparations. Eur J Cancer. 1996;32A(9):1544–1550. 21. Moricke A, Zimmermann M, Valsecchi MG, et al. Dexamethasone vs prednisone in induction treatment of pediatric ALL:

haematologica | 2019; 104(9)

Effects of PEG-asparaginase on CSF asparagine levels

22.

23. 24. 25.

26.

27.

results of the randomized trial AIEOP-BFM ALL 2000. Blood. 2016;127(17):2101–2112. Asselin B, Rizzari C. Asparaginase pharmacokinetics and implications of therapeutic drug monitoring. Leuk Lymphoma. 2015;56(8):2273–2280. van den Berg H. Asparaginase revisited. Leuk Lymphoma. 2011;52(2):168–178. Zeidan A, Wang ES, Wetzler M. Pegasparaginase: where do we stand? Expert Opin Biol Ther. 2009;9(1):111–119. Gerrits GP, Trijbels FJ, Monnens LA, et al. Reference values for amino acids in cerebrospinal fluid of children determined using ion-exchange chromatography with fluorimetric detection. Clin Chim Acta. 1989;182(3):271–280. Asselin BL, Lorenson MY, Whitin JC, et al. Measurement of serum L-asparagine in the presence of L-asparaginase requires the presence of an L-asparaginase inhibitor. Cancer Res. 1991;51(24):6568–6573. Henriksen LT, Nersting J, Raja RA, et al. Cerebrospinal fluid asparagine depletion during pegylated asparaginase therapy in

haematologica | 2019; 104(9)

28.

29.

30.

31.

children with acute lymphoblastic leukaemia. Br J Haematol. 2014;166(2):213– 220. Hawkins DS. Asparaginase pharmacokinetics after intensive polyethylene glycolconjugated L-asparaginase therapy for children with relapsed acute lymphoblastic leukemia. Clin Cancer Res. 2004;10(16): 5335–5341. Ahlke E, Nowak-Göttl U, SchulzeWesthoff P, et al. Dose reduction of asparaginase under pharmacokinetic and pharmacodynamic control during induction therapy in children with acute lymphoblastic leukaemia. Br J Haematol. 1997;96(4):675–681. Vieira Pinheiro JP, Wenner K, Escherich G, et al. Serum asparaginase activities and asparagine concentrations in the cerebrospinal fluid after a single infusion of 2,500 IU/m(2) PEG asparaginase in children with ALL treated according to protocol COALL-06-97. Pediatr Blood Cancer. 2006;46(1):18–25. Tong WH, Pieters R, de Groot-Kruseman

32.

33.

34.

35.

HA, et al. The toxicity of very prolonged courses of PEGasparaginase or Erwinia asparaginase in relation to asparaginase activity, with a special focus on dyslipidemia. Haematologica. 2014;99(11):1716– 1721. van der Sluis IM, de Groot-Kruseman H, Te Loo M, et al. Efficacy and safety of recombinant E. coli asparaginase in children with previously untreated acute lymphoblastic leukemia: a randomized multicenter study of the Dutch Childhood Oncology Group. Pediatr Blood Cancer. 2018;65(8):e27083. Pasut G, Veronese FM. State of the art in PEGylation: the great versatility achieved after forty years of research. J Control Release. 2012;161(2):461–472. Serlin Y, Shelef I, Knyazer B, Friedman A. Anatomy and physiology of the blood– brain barrier. Semin Cell Dev Biol. 2015;38:2–6. Pardridge WM. Drug transport across the blood-brain barrier. J Cereb Blood Flow Metab. 2012;32(11):1959–1972.

1821

ARTICLE Ferrata Storti Foundation

Haematologica 2019 Volume 104(9):1822-1829

Non-Hodgkin Lymphoma

Burkitt-like lymphoma with 11q aberration: a germinal center-derived lymphoma genetically unrelated to Burkitt lymphoma Blanca Gonzalez-Farre,1,2,3* Joan Enric Ramis-Zaldivar,2,3* Julia SalmeronVillalobos,2 Olga Balagué,1,2,3 Verónica Celis,4 Jaime Verdu-Amoros,5 Ferran Nadeu,2,3 Constantino Sábado,6 Antonio Ferrández,7 Marta Garrido,8 Federico García-Bragado,9 María Dolores de la Maya,10 José Manuel Vagace,10 Carlos Manuel Panizo,11 Itziar Astigarraga,12 Mara Andrés,13 Elaine S. Jaffe,14 Elias Campo1,2,3* and Itziar Salaverria2,3*

Hematopathology Unit, Hospital Clínic de Barcelona, University of Barcelona, Barcelona, Spain; 2Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Barcelona, Spain; 3Centro de Investigación Biomédica en Red de Cáncer (CIBERONC), Madrid, Spain; 4 Pediatric Oncology Department, Hospital Sant Joan de Déu, Esplugues de Llobregat, Spain; 5Pediatric Oncology Department, Hospital Clínico Universitario de Valencia, Valencia, Spain; 6Pediatric Oncology Department, Hospital Universitari Vall d'Hebron, Barcelona, Spain; 7Pathology Department, Hospital Clínico de Valencia, Valencia, Spain; 8 Pathology Department, Hospital Universitari Vall d'Hebron, Barcelona, Spain; 9Pathology Department, Complejo Hospitalario de Navarra, Pamplona, Spain; 10Pediatric Hematology Department, Hospital Materno Infantil de Badajoz, Badajoz, Spain; 11Department of Hematology, Clínica Universidad de Navarra and Instituto de Investigación Sanitaria de Navarra (IdiSNA), Pamplona, Spain; 12Pediatrics Department, Hospital Universitario Cruces, IIS Biocruces Bizkaia, UPV/EHU, Barakaldo, Spain; 13Pediatric Oncology Department, Hospital La Fe, Valencia, Spain and 14Hematopathology Section, Laboratory of Pathology, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA 1

*BGF, JERZ, EC and IS contributed equally to this work.

ABSTRACT

Correspondence: ITZIAR SALAVERRIA isalaver@clinic.cat Received: October 5, 2018. Accepted: February 7, 2019. Pre-published: February 7, 2019. doi:10.3324/haematol.2018.207928 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1822 ©2019 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.

1822

urkitt-like lymphoma with 11q aberration is characterized by pathological features and gene expression profile resembling those of Burkitt lymphoma but lacks the MYC rearrangement and carries an 11q-arm aberration with proximal gains and telomeric losses. Whether this lymphoma is a distinct category or a particular variant of other recognized entities is controversial. To improve the understanding of Burkitt-like lymphoma with 11q aberration we performed an analysis of copy number alterations and targeted sequencing of a large panel of B-cell lymphomarelated genes in 11 cases. Most patients had localized nodal disease and a favorable outcome after therapy. Histologically, they were high grade B-cell lymphoma, not otherwise specified (8 cases), diffuse large B-cell lymphoma (2 cases) and only one was considered as atypical Burkitt lymphoma. All cases had a germinal center B-cell signature and phenotype with frequent LMO2 expression. The patients with Burkitt-like lymphoma with 11q aberration had frequent gains of 12q12-q21.1 and losses of 6q12.1-q21, and lacked common Burkitt lymphoma or diffuse large B-cell lymphoma alterations. Potential driver mutations were found in 27 genes, particularly involving BTG2, DDX3X, ETS1, EP300, and GNA13. However, ID3, TCF3, or CCND3 mutations were absent in all cases. These results suggest that Burkitt-like lymphoma with 11q aberration is a germinal center-derived lymphoma closer to high-grade B-cell lymphoma or diffuse large B-cell lymphoma than to Burkitt lymphoma.

Introduction Our knowledge of lymphomas in children and young adults has increased dramatically in the last years with the identification of several subtypes that predominantly occur in this age subgroup.1-4 One of these recently recognized categories is Burkittlike lymphoma with 11q aberration (BLL-11q) which has morphological, phenotyphaematologica | 2019; 104(9)

Burkitt-like lymphoma with 11q aberration

ic, and gene expression profiles resembling those of Burkitt lymphoma (BL), but lacks MYC rearrangements according to standard methods of detection, such as fluorescence in situ hybridization (FISH). Alternatively, these tumors carry an 11q-arm aberration characterized by proximal gains and telomeric losses.4 In comparison with BL, BLL-11q seems to have more complex karyotypes, a certain degree of cytological pleomorphism, sporadically a follicular pattern and a high incidence of nodal presentation.4,5 Very similar cases have also been reported in the post-transplant setting,6 although its incidence in other immunocompromised conditions, such as human immunodeficiency infection, is still unclear.7,8 BLL-11q has been incorporated in the revised World Health Organization (WHO) classification as a provisional category1 because its precise taxonomy as a particular variant of BL, diffuse large B-cell lymphoma (DLBCL) or a distinct form of high-grade B-cell lymphoma (HGBCL) is still controversial.1,4-6,9-11 The clarification of the biological nature of this uncommon subtype of lymphoma is clinically relevant because of increasing interest in defining the most appropriate management strategies for specific subtypes of lymphomas in pediatric and young adult patients.12 Recent DNA copy number alteration and nextgeneration sequencing studies have provided a comprehensive catalog of genomic aberrations in BL and DLBCL which clearly distinguish these entities.13-17 In this study we performed an integrated analysis of genomic and mutational alterations with a complete annotation of clinical and pathological features of BLL-11q with the goal of obtaining insights to refine the understanding of the pathogenesis of these tumors and improve their diagnosis.

per probe, while the pattern corresponding to the 11q gain/loss or gain/amplification/loss aberration would be two blue, three up to five green signals and one red signal. The probe was tested in an independent series of eight non-Hodgkin B-cell lymphomas and four MYC-negative HGBCL with lack of the 11q alteration by array analysis and all showed the normal pattern described above.

Copy number analysis DNA was hybridized on Oncoscan FFPE or SNP array platform (ThermoFisher Scientific, Waltham, MA, USA) and analyzed as described previously (Online Supplementary Methods).2 Published copy number data on MYC-positive BL20 and DLBCL13 were reanalyzed for comparison.

Sequencing approaches The mutational status of 96 B-cell lymphoma-related genes (Online Supplementary Table S2) was examined by target next-generation sequencing in ten BLL-11q cases and four MYC-negative 11q-negative cases using a NGS SureSelect XT Target Enrichment System Capture strategy (Agilent Technologies, Santa Clara, CA, USA) before sequencing in a MiSeq instrument (Illumina, San Diego, CA, USA) (Online Supplementary Methods). Additionally, analysis of hotspots of mutation in ID3, TCF3 and CCND3 genes, ETS1 exon 1 (transcript NM_005238) and verification of variants in specific cases was performed by Sanger sequencing using primers described in Online Supplementary Table S3.

Gene expression analysis Cell of origin was determined by Lymph2Cx assay (Nanostring, Seattle, WA, USA) as previously published.21 Gene expression levels of MYC and ETS1 were investigated by real-time quantitative polymerase chain reaction (Online Supplementary Methods) using Taqman assays described in Online Supplementary Table S4.

Methods Statistical methods

Sample selection and DNA/RNA extraction To identify BLL-11q cases we initially reevaluated the presence of MYC translocations in 95 cases diagnosed as BL, atypical BL or HGBCL, not otherwise specified (NOS), in our Hematopathology Unit between 2000-2016. Three consultation cases from centers belonging to the Sociedad Española de Hematología y Oncología Pediátricas (SEHOP) were also analyzed. Cases were reviewed by three pathologists (BG-F, EC, ESJ). DNA and RNA were extracted using standard protocols (Qiagen, Hilden, Germany). This study was approved by the Institutional Review Board of the Hospital Clinic of Barcelona. Informed consent was obtained from all patients in accordance with the Declaration of Helsinki.

Immunohistochemistry and fluorescence in situ hybridization Immunohistochemical analysis using a panel of antibodies detecting common B- and T-cell markers as well as LMO2 and MYC was performed and interpreted as previously reported (Online Supplementary Table S1).18,19 MYC breaks and MYC/IGH fusions were analyzed by FISH using the XL MYC BA Probe (Metasystems, Altlussheim, Germany) and LSI IGH/MYC/CEP 8 TriColor Dual Fusion Probe Kit (Vysis-Abbott, Abbott Park, IL, USA) respectively. The 11q alteration was studied with a custom FISH probe using BAC clones (Invitrogen Inc.) for proximal gains (RP11-414G21-spectrum green) and terminal losses (RP11-629A20-spectrum red) combined with CEP11-spectrum aqua (Vysis-Abbott Inc.). The FISH constellation in a normal case is characterized by two signals haematologica | 2019; 104(9)

The χ2 method was used for categorical variables and Student ttests for continuous variables. Non-parametric tests were applied when necessary. The P-values for multiple comparisons were adjusted using the Benjamini–Hochberg correction. Survival curves were estimated with the Kaplan-Meier method. Statistical analyses were carried out with SPSS v22 and R software v3.1.3.

Results Identification of cases with Burkitt-like lymphoma with 11q aberration To identify BLL-11q cases we reevaluated the presence of MYC translocations in 95 cases diagnosed as having BL, atypical BL or HGBCL, NOS. We confirmed the presence of MYC rearrangements in 78 cases (82.1%), of which 67 (70.5%) were classified as BL. Since the 11q aberration has been found mainly in children and young adults (<40 years old),4 we analyzed the 60 patients under 40 years and the 35 older patients separately (Online Supplementary Figure S1). In the younger cohort (n=60), the 46 (76.7%) cases with MYC translocations were classified as having BL. To find BLL-11q cases, we initially used the Oncoscan platform in the remaining 14 MYC-negative patients and detected the presence of the 11q gain/loss alteration in eight of them. Additionally, we found a copy number pattern consistent with the presence of 11q alteration in three recent consultation cases from SEHOP (Online Supplementary Figures S1 and S2). Then, among those BLL-11q cases we were able 1823

B. Gonzalez-Farre et al.

to verify the presence of the 11q aberration by FISH in all ten evaluable cases (Online Supplementary Figure S3 and Supplementary Table S5). The morphological, clinical, and genetic features and consensus diagnosis of the 11 BLL11q identified in our files are summarized in Table 1. The six cases negative for the MYC rearrangement and 11q aberrations by Oncoscan were re-classified as DLBCL (3 cases) or HGBCL, NOS (3 cases). The DLBCL had predominant centroblastic morphology, germinal center phenotype, very high proliferative index and a focal “starry sky” pattern (see Online Supplementary Results). The absence of 11q alterations was also verified using the 11q FISH probe in four of these MYC/11q-negative cases with available material (Online Supplementary Figure S1A). In the 35 older patients (≥40 years old), a MYC translocation was found in 32 cases; one was classified as DLBCL, 21 as BL, and ten were HGBCL with double- or triple-hit aberrations (BCL2 and/or BCL6 translocations). Only three cases were negative for MYC translocations and were classified as HGBCL, NOS (Online Supplementary Figure S1B and Online Supplementary Results). We screened these cases with the 11q FISH probe and all three were negative for the 11q aberration.

tumors were reclassified morphologically as HGBCL, NOS, two as DLBCL and only one case was considered atypical BL. None of the cases was considered as typical BL (Figure 1). Six cases showed a “starry sky” pattern and two had a nodular growth pattern with the presence of a disrupted follicular dendritic cell meshwork (Figure 1C). Ki67 was very high in all the samples, as in BL. All cases had a germinal center (GC) phenotype and GC B-cell (GCB) signature by Nanostring Lymph2Cx assay. MUM1/IRF4 was negative in all 11 cases. One case expressed BCL2 (Figure 1D). LMO2, a germinal center marker that is usually seen in GCB-DLBCL but not in BL18 was expressed in five cases (Figure 1A, B). Interestingly, using a 40% cutoff,19 five cases were positive for MYC expression. However, only one case showed diffuse and intense positivity while the other four cases had either only positivity in around 50% of the cells or the intensity was not that expected in typical BL. Additionally, MYC RNA levels were significantly lower in BLL-11q than in MYC-positive BL (relative expression 0.07 vs. 0.36, P=0.019) (Online Supplementary Figure S4A). Epstein-Barr virus hybridization was negative in the nine cases tested. Clinically, BLL-11q frequently had a nodal localized presentation (8/11) in the head and neck region. Two cases had an extranodal presentation, one in the context of an acute appendicitis and the other debuted as an omental mass. Eight patients (73%) had stage I-II, and one patient presented in an advanced stage (IV-E) with

Clinical and morphological results of cases of Burkitt-like lymphoma with 11q aberration The 11 patients with BLL-11q had a mean age of 15 years (range, 8-37 years); eight were male (Table 1). Eight

Table 1. Pathological and clinical features of 11 cases of Burkitt-like lymphoma with 11q aberration.

Case

Age/ gender

Biopsy site

Original diagnosis

Final diagnosis

Immunophenotype CD10& IRF4/ BCL2 LMO2 BCL6 MUM1

27, M

#2**

37,M

8,F

17,F

14,F

#6 #7

14,M 8, M

#14

8,M

#15

12,M

#16

14, M

#17

16, M

Laterocervical LN Axillary LN Tonsil Submaxilar LN Laterocervical LN

Appendix Laterocervical LN Laterocervical LN Laterocervical mass Laterocervical LN Omentum

Atypical BL Atypical BL HGBCL

Stage* MYC

COO Chemo- Rituximab Outcome/ Nanostring therapy follow-up (Lymph2Cx)

HGBCL, NOS

GCB

Yes

CR, 72m

DLBCL

IV-E

GCB

Yes

CR, 112m

DLBCL & HGBCL blastoid HGBCL HGBCL, NOS

GCB

CR, 54m

GCB

Yes

CR, 22m

HGBCL

GCB

CR, 29m

HGBCL BL

HGBCL with features intermediate between BL and DLBCL DLBCL Atypical BL

+ +

+ -

II I

GCB GCB

P P

No No

CR, 25m CR, 113m

HGBCL blastoid

Weak +

GCB

CR, 15m

DLBCL

HGBCL, NOS

GCB

CR, 35m

DLBCL

HGBCL, NOS

III

GCB

Yes

CR, 12m

HGBCL

HGBCL, NOS

III

GCB

Yes

CR, 4m

*Stage was established according to the St. Jude/International Pediatric NHL Staging System (IPNHLSS) or Ann Arbor staging system for pediatric and adult patients, respectively. COO: cell of origin; M: male; F: female; LN: lymph node; BL: Burkitt lymphoma; HGBCL: high-grade B-cell lymphoma; NOS: not otherwise specified; DLBCL: diffuse large B-cell lymphoma; Epstein-Barr virus in situ hybridization (EBER) was negative in all nine tested cases. E: extranodal; GCB: germinal center B-cell; A: adult schema protocol (R-CHOP or burkimab); P: pediatric schema protocol; CR: complete response; m: months. All patients received central nervous system prophylaxis. **Human immunodeficiency virus (HIV)-positive patient

1824

haematologica | 2019; 104(9)

Burkitt-like lymphoma with 11q aberration

Figure 1. Morphological features of cases of Burkitt-like lymphoma with 11q aberration. (A1-A3) Case #2 shows the typical morphology of diffuse large B-cell lymphomas with large and irregular cells resembling centroblasts. This case was positive for (A2) LMO2 and negative for (A3) MYC. (B1-B3) Case #4 corresponds to a tumor with high-grade B-cell lymphoma (HGBCL) morphology. It is composed mostly of medium-sized cells with mild heterogeneity. Note the “starry sky” pattern. This case was positive for (B2) MYC and (B3) LMO2 expression. (C1-C2; case #7) Lymph node with nodular architecture and a “starry sky” pattern with large follicles and a disrupted follicular cell meshwork highlighted with (C2) CD21. (D1-D2; case #5) This is a case with HGBCL features with expression of (D2) BCL2 in the neoplastic cells.

haematologica | 2019; 104(9)

1825

B. Gonzalez-Farre et al.

widespread disease in the context of chronic human immunodeficiency virus infection. All cases were treated with chemotherapy, including rituximab in five. All patients were alive with no disease after a median followup of 29 months (Table 1).

Copy number analysis The copy number analysis of all the 11 BLL-11q cases showed a total of 78 alterations (mean 7.1; range, 2-15) (Online Supplementary Tables S5 and S6). Seven cases had the typical 11q gain/loss pattern (Figure 2A-B, Online

Figure 2. Genetic features of cases of Burkitt-like lymphoma with 11q aberration. (A) Global copy number profile of the 11 cases of Burkitt-like lymphoma (BLL) with 11q aberration. The horizontal axis indicates chromosomes from 1 to Y and p to q. The vertical axis indicates the frequency of the genomic aberration among the cases analyzed. Gains are depicted in blue, losses are depicted in red. (B) Individual copy number profile of case #16 showing a prototypical, gain, loss and amplification in the 11q region. Each probe is aligned from chromosome 1 to Y and p to q arm. (C) Mutational overview of ten cases of BLL with 11q aberration. The heat map shows the casespecific pattern of driver mutations found by next-generation sequencing. Each column represents a case and each row represents a gene. The right bar graph illustrates the mutation frequency of each gene. (D) A diagram of the relative positions of driver mutations is shown for BTG2, ETS1 and GNA13 genes. Domains BTG2: BTG family domain. Domains ETS1: PNT: pointed domain; TAD: transactivation domain; H-1/2: inhibitory a-helices 1/2; DBD: DNA binding domain; H4-5: a-helix 4/5. Domains GNA13: G-alpha: G protein a subunit. Circles indicate missense mutations, triangles indicate truncating mutations and rhombi indicate splicing mutations.

1826

haematologica | 2019; 104(9)

Burkitt-like lymphoma with 11q aberration

Supplementary Figure S2), two cases had only an 11q terminal deletion, one case showed a complex 11q alteration with two gains and two losses, and finally one case had an 11q23.3-q25 copy number neutral loss of heterozygosity in addition to gain (Online Supplementary Figure S2). Two minimal regions of gain were identified (chr11:103326831111737912/11q22.3-q23.1 and chr11:114767237116764582/11q23.3,hg19) whereas the minimal region of loss was in chr11:128214400-132020453/11q24.3-q25 (including the ETS1 and FLI1 genes). No cases with homozygous deletions of these two targets were observed in our series. The breakpoint region between gain and loss was not conserved and spanned from chr11:118352769 to chr11:121062860. Amplifications in the 11q arm were observed in four cases, with a minimal region chr11:118347020-120155799/11q23.3, including the USP2 gene (Online Supplementary Figure S5). The most recurrent copy number aberrations other than 11q were 12q12-q21.1 gains and 6q12.1-q21 losses (Figure 2A). BLL-11q cases displayed similar levels of complexity as MYC-positive BL (7.1 vs. 6 alterations/case),20 but significantly lower than those of GCB-DLBCL (7.1 vs. 19, alterations/case P<0.0001).13 The BLL-11q genomic profile differed from that of BL and DLBCL (Online Supplementary Figure S6). BLL-11q had frequent gains of 5q21.3-q32 and losses of 6q12.1-q21 and lacked the 1q gains seen in MYCpositive BL. BLL-11q also lacked alterations typically seen in GCB-DLBCL such as gains of 2p16.1 and 7p and losses of 1p36.32. The six tumors negative for both MYC and 11q-aberrations in patients younger than 40 years had similar levels of genomic complexity as those observed in BLL-11q (11.8 vs. 7.1 alterations/case) (Online Supplementary Figure S7A). The unique significant aberration that distinguished the two groups was the presence/absence of the 11q aberration. The review of the literature regarding other lymphoid neoplasms confirmed that the 11q alteration observed in BLL-11q is mainly absent in other lymphoma entities with the exception of transformed follicular lymphoma (16%) (Online Supplementary Results).22

Next-generation sequencing and gene expression analysis Target next-generation sequencing showed a total of 49 potential driver mutations affecting 27 different genes (mean=4.9 mutations per case) (Figure 2C, D, Online Supplementary Figures S8 and S9; Online Supplementary Table S7). Interestingly, all cases lacked the typical BL mutations in ID3, TCF3, or CCND3 genes, and their mutational profile was more similar to that of other GCderived lymphomas with recurrent mutations affecting BTG2 (4 cases), DDX3X, ETS1, EP300, and GNA13 (3 cases each) (Online Supplementary Table S8). Five cases had mutations in epigenetic modifier genes such as EP300, CREBBP, KMT2C, EZH2, ARID1A, KMT2D, HIST1H1D and HIST1H2BC. Two cases had concomitant TMEM30A deleterious mutations associated with a 6q14.1 deletion as seen in DLBCL but not in BL (Figure 2C).14-16 BTG2 mutations, found in four cases, comprised three missense mutations and one deletion in a splicing site. BTG2 is a tumor suppressor gene with an important role in G1/S transition through inhibition of CCND1 in a pRbdependent mechanism.23 These BTG2-inactivating mutations could release CCND1 inhibition and accelerate G1/S transition. GNA13 mutations were found in three cases haematologica | 2019; 104(9)

including four missense and two nonsense mutations, and one missense mutation in a splicing site. Two MYC missense mutations occurred in the central domain of the protein, but did not affect threonine phosphorylation sites (Online Supplementary Table S7).24 ETS1 mutations have been previously described in BLL-11q and activated B-cell DLBCL13,17 but not in conventional BL (Online Supplementary Table S8).14,15 We detected three coding mutations located on the winged helix-turn-helix DNAbinding domain but the previously described exon 1 mutations (NM_005238) were absent in this series. ETS1 RNA expression was lower in BLL-11q than in MYC-positive BL (relative expression 6.6 vs. 19.3, P<0.001) and was also lower in ETS1-mutated than wild-type BLL-11q (relative expression 1.9 vs. 8.6, P=0.03) (Online Supplementary Figure S4B). The mutational profile of four MYC-negative/11q alteration-negative cases with material available was analyzed using the same approach. No mutations in BTG2, EP300 or ETS1 genes were observed. Moreover, three out of four did not harbor any BL-related mutation on ID3, TCF3 and CCND3 whereas the fourth case had a mutational profile commonly seen in BL with MYC, DDX3X, SMARC4, CCND3 and TP53 mutations (Online Supplementary Figure S7B).

Discussion BLL-11q was initially recognized as a particular subset of HGBCL that had an expression profile and some pathological characteristics similar to those of BL but lacked MYC-translocations and alternatively shared a common pattern of gains at 11q23 associated with losses at 11q24qter. The particular features of these cases raise some uncertainty on their precise categorization as a variant of BL or a tumor related to other HGBCL.1,4-6,9-11 On the other hand, the limited number of cases reported and the different methodologies used for their recognition do not provide a clear view of their incidence and clinico-pathological characteristics. In this study we searched our files for cases that could be reclassified as BLL-11q among 95 tumors previously classified as BL, atypical BL, or HGBCL, NOS and found eight (8%) cases with the chromosomal aberration. These cases, together with three additional cases received on consultation, were investigated for copy number alterations and mutational profiles and compared to the genomic aberrations recently identified in BL, DLBCL, and HGBCL.13-17 The complexity of BLL-11q was similar to that of MYC-positive BL,20 but significantly lower than that of GCB-DLBCL.13 The BLL-11q genomic profile differed from that of BL and DLBCL (Online Supplementary Figure S6). BLL-11q had frequent gains of 5q21.3-q32 and losses of 6q12.1-q21 and lacked the 1q gains seen in MYC-positive BL. BLL-11q also lacked alterations typically seen in GCB-DLBCL such as gains of 2p16.1 and 7p and losses of 1p36.32. Additionally, we identified a mutational profile in BLL-11q which was different from that of MYC-positive BL since all cases lacked the typical BL mutations in ID3, TCF3, or CCND3 genes and had mutations in BTG2, DDX3X, and ETS1 not seen in BL. In addition, BLL-11q had mutations in epigenetic modifier genes such as EP300, CREBBP, KMT2C, EZH2, ARID1A, KMT2D, HIST1H1D and HIST1H2BC, which are common in DLBCL, particu1827

B. Gonzalez-Farre et al.

larly of the GCB subtype. Other genes frequently mutated in GCB-DLBCL but not in BL were GNA13 and TMEM30A associated with 6q14.1.14-16 We also compared our results with those of two recent studies on HGBCL (including double- and triple-hit lymphomas).25,26 These cases also have recurrent mutations in histone modifier genes such as KMT2D, CREBBP and EZH2 (Online Supplementary Table S8). Intriguingly, HGBCL, NOS, mainly with MYC-translocations, shared mutations in genes frequently mutated in both BL and GC-DLBCL.25,26 All these observations suggest that BLL11q is a neoplasm closer to other GC-derived lymphomas rather than BL in which the 11q aberration together with other mutations may play a relevant role in their pathogenesis. While this manuscript was under revision, Wagener et al. published a mutational study of 15 BLL-11q cases. Similar to our findings, no mutations in ID3/TCF3 were identified and those cases carried frequent mutations in GC-DLBCL-associated genes such as GNA13, FOXO1 and EZH2. Intriguingly, that study did not find mutations in BTG2, KMT2D, KMT2C or CREBBP observed in our study.27 Collectively, these findings indicate that the genomic and mutational profiles of BLL-11q are different from those of BL and more similar to those of other GCderived lymphomas. In addition to the genetic differences, our BLL-11q cases differed clinically, morphologically and phenotypically from conventional BL and instead showed features more consistent with HGBCL or DLBCL. As in previous studies, all our patients were younger than 40 years, although occasional cases in older patients have been reported.4,5,27 Contrary to BL, most of our cases of BLL-11q presented with localized lymphadenopathy.4,5,27 These cases have a favorable outcome after therapy, although the optimal clinical management remains to be determined. Morphologically, our cases had a prominent “starry sky” pattern and high proliferation rate (>90%) but did not have the typical cytological features of BL since they were better classified as HGBCL with blastoid or intermediate features between HGBCL (8 cases) and DLBCL (2 cases) and only one had features of atypical BL. As previously reported,4 two of our cases displayed a follicular growth pattern, with an underlying meshwork of follicular dendritic cells, raising the differential diagnosis with other pediatric lymphomas such as large B-cell lymphoma with IRF4 rearrangement.3 However, BLL-11q did not express IRF4/MUM1 and often exhibited a “starry sky” pattern with frequent mitotic figures, features that are not usual in large B-cell lymphoma with IRF4 rearrangement. We also identified different immunohistochemical stains that could help in the differential diagnosis from other lymphomas entities. LMO2, a GC marker that is typically downregulated in BL and other lymphomas with MYC translocation,18 was detected in 46% of our BLL-11q cases. In addition, and contrary to BL, MYC expression with a diffuse and intense pattern was only detected in one of our cases while the other four positive cases either exhibited partial positivity or the intensity was weak contrary to the pattern seen in BL. Negativity for MYC rearrangement is a crucial element for the recognition of these cases. The recommended technique for detecting MYC translocations in clinical practice is FISH analysis using break-apart probes, with the limitation that a subset of 4% of MYC-positive cases are not detected with this method but picked up using 1828

MYC/IGH probes.28 The genetic feature that distinguishes BLL-11q is an alteration of the 11q arm that prototypically is characterized by an 11q23.2-q23.3 gain/amplification and 11q24.1-qter loss. Additionally, isolated cases have been recognized with single 11q24.1-qter terminal loss or 11q23 gain with 11q24-qter copy number neutral loss of heterozygosity.4,11 In our study we identified the presence of these 11q alterations using copy number arrays. We also confirmed the presence of 11q alterations by FISH analysis with a custom probe in all tested cases, suggesting that this approach may be useful in clinical practice to identify such cases (Online Supplementary Table S8). The specificity of this FISH approach was also confirmed by the fact that no false positive cases were observed in the 12 lymphoma cases in which the array showed a normal 11q pattern. Nevertheless, more studies on the clinical value of this probe are needed and, for the time being, confirmation of the finding by copy number array would be desirable. The specific 11q alteration observed in BLL11q should be distinguished from other 11q aberrations such as 11q gains of the 11q24 region that include ETS1 and FLI1, detected in DLBCL,29 or the 11q25 losses missing ETS1 and FLI1 described in some post-transplant lymphoproliferative disorders.30,31 On the other hand, although the 11q23 gain/11q24-qter loss of BLL-11q is mainly absent in other lymphoma entities, its detection should not be considered as a unique feature to diagnose BLL-11q cases since some transformed follicular lymphomas may carry a similar 11q aberration pattern.22 In summary, BLL-11q is a GC-derived lymphoma with genomic and mutational profiles closer to those of HGBCL or GCB-DLBCL rather than BL in which the 11q aberration, together with other mutations, may play a relevant role in pathogenesis. These observations support a reconsideration of the “Burkitt-like” term for these tumors. Although, the most appropriate name is not easy to propose and requires broader discussion and consensus, we think that the term “aggressive B-cell lymphoma with 11q aberration” captures their pathological features. To identify these cases we suggest performing copy number arrays or FISH with the 11q probe in cases with BL, DLBCL, and HGBCL morphology, a GC phenotype and very high proliferative index (>90%), without MYC rearrangements, in young patients. The recognition of these tumors is clinically relevant because they have a favorable outcome after therapy, although further studies are needed to determine the optimal clinical management. Funding This work was supported by Asociación Española Contra el Cáncer (AECC CICPFI6025SALA), Fondo de Investigaciones Sanitarias Instituto de Salud Carlos III (Miguel Servet program CP13/00159 and PI15/00580, to IS), Spanish Ministerio de Economía y Competitividad, Grant SAF2015-64885-R (EC), Generalitat de Catalunya Suport Grups de Recerca (2017SGR-1107 I.S. and 2017-SGR-1142 to EC), and the European Regional Development Fund “Una manera de fer Europa”. JERZ was supported by a fellowship from the Generalitat de Catalunya AGAUR FI-DGR 2017 (2017 FI_B01004). EC is an Academia Researcher of the "Institució Catalana de Recerca i Estudis Avançats" (ICREA) of the Generalitat de Catalunya. This work was developed at the Centro Esther Koplowitz, Barcelona, Spain. The group is supported by Acció Instrumental d’Incorporació de Científics i Tecnòlegs PERIS 2016 (SLT002/16/00336) from the Generalitat de Catalunya. haematologica | 2019; 104(9)

Burkitt-like lymphoma with 11q aberration

The copy-number data reported in this article have been deposited at the GEO database under accession number GSE116527. Sequencing data have been deposited at the European Nucleotide Archive (ENA, accession number ERP110085). Acknowledgments We thank the centers of the Sociedad Española de

References 1. Swerdlow SH, Campo E, Harris NL, Jaffe ES, Pileri SA, Stein H, Thiele J (Eds) WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues. (Revised 4th edition) IARC: Lyon 2017. 2. Schmidt J, Gong S, Marafioti T, et al. Genome-wide analysis of pediatric-type follicular lymphoma reveals low genetic complexity and recurrent alterations of TNFRSF14 gene. Blood. 2016;128(8):11011111. 3. Salaverria I, Philipp C, Oschlies I, et al. Translocations activating IRF4 identify a subtype of germinal center-derived B-cell lymphoma affecting predominantly children and young adults. Blood. 2011;118(1):139-147. 4. Salaverria I, Martin-Guerrero I, Wagener R, et al. A recurrent 11q aberration pattern characterizes a subset of MYC-negative high-grade B-cell lymphomas resembling Burkitt lymphoma. Blood. 2014;123(8): 1187-1198. 5. Rymkiewicz G, Grygalewicz B, Chechlinska M, et al. A comprehensive flow-cytometry-based immunophenotypic characterization of Burkitt-like lymphoma with 11q aberration. Mod Pathol. 2018;31(5):732-743. 6. Ferreiro JF, Morscio J, Dierickx D, et al. Post-transplant molecularly defined Burkitt lymphomas are frequently MYC-negative and characterized by the 11q-gain/loss pattern. Haematologica. 2015;100(7):e275e279. 7. Capello D, Scandurra M, Poretti G, et al. Genome wide DNA-profiling of HIV-related B-cell lymphomas. Br J Haematol. 2010;148(2):245-255. 8. Deffenbacher KE, Iqbal J, Liu Z, Fu K, Chan WC. Recurrent chromosomal alterations in molecularly classified AIDS-related lymphomas: an integrated analysis of DNA copy number and gene expression. J Acquir Immune Defic Syndr. 2010;54(1):18-26. 9. Poirel HA, Cairo MS, Heerema NA, et al. Specific cytogenetic abnormalities are associated with a significantly inferior outcome in children and adolescents with mature Bcell non-Hodgkin's lymphoma: results of the FAB/LMB 96 international study. Leukemia. 2009;23(2):323-331. 10. Havelange V, Ameye G, Theate I, et al. The peculiar 11q-gain/loss aberration reported in a subset of MYC-negative high-grade B-

haematologica | 2019; 104(9)

11.

12.

13.

14.

15.

16. 17.

18.

19.

20.

21.

Hematología y Oncología Pediátricas that submitted cases for consultation and Noelia Garcia, Silvia Martín, and Helena Suarez for their excellent technical assistance. We are indebted to the IDIBAPS Genomics Core Facility and to HCB-IDIBAPS Biobank-Tumor Bank and Biobanc de l’HospitaI Infantil Sant Joan de Déu, both integrated in the National Network Biobanks of ISCIII for the sample and data procurement. We thank Prof. Reiner Siebert from the University of Ulm for sharing the 11q

cell lymphomas can also occur in a MYCrearranged lymphoma. Cancer Genet. 2016;209(3):117-118. Grygalewicz B, Woroniecka R, Rymkiewicz G, et al. The 11q-gain/loss aberration occurs recurrently in MYC-negative Burkitt-like lymphoma with 11q aberration, as well as MYC-positive Burkitt lymphoma and MYC-positive high-grade B-cell lymphoma, NOS. Am J Clin Pathol. 2017;149(1):17-28. Dunleavy K, Gross TG. Management of aggressive B-cell NHLs in the AYA population: an adult vs pediatric perspective. Blood. 2018;132(4):369-375. Karube K, Enjuanes A, Dlouhy I, et al. Integrating genomic alterations in diffuse large B-cell lymphoma identifies new relevant pathways and potential therapeutic targets. Leukemia. 2018;32(3):675-684. Richter J, Schlesner M, Hoffmann S, et al. Recurrent mutation of the ID3 gene in Burkitt lymphoma identified by integrated genome, exome and transcriptome sequencing. Nat Genet. 2012;44(12):13161320. Schmitz R, Young RM, Ceribelli M, et al. Burkitt lymphoma pathogenesis and therapeutic targets from structural and functional genomics. Nature. 2012;490(7418):116120. Love C, Sun Z, Jima D, et al. The genetic landscape of mutations in Burkitt lymphoma. Nat Genet. 2012;44(12):1321-1325. Morin RD, Mungall K, Pleasance E, et al. Mutational and structural analysis of diffuse large B-cell lymphoma using wholegenome sequencing. Blood. 2013;122(7): 1256-1265. Colomo L, Vazquez I, Papaleo N, et al. LMO2-negative expression predicts the presence of MYC translocations in aggressive B-cell lymphomas. Am J Surg Pathol. 2017;41(7):877-886. Johnson NA, Slack GW, Savage KJ, et al. Concurrent expression of MYC and BCL2 in diffuse large B-cell lymphoma treated with rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone. J Clin Oncol. 2012;30(28):3452-3459. Scholtysik R, Kreuz M, Klapper W, et al. Detection of genomic aberrations in molecularly defined Burkitt's lymphoma by array-based, high resolution, single nucleotide polymorphism analysis. Haematologica. 2010;95(12):2047-2055. Scott DW, Wright GW, Williams PM, et al. Determining cell-of-origin subtypes of dif-

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

fuse large B-cell lymphoma using gene expression in formalin-fixed paraffinembedded tissue. Blood. 2014;123(8):12141217. Bouska A, McKeithan TW, Deffenbacher KE, et al. Genome-wide copy-number analyses reveal genomic abnormalities involved in transformation of follicular lymphoma. Blood. 2014;123(11):16811690. Guardavaccaro D, Corrente G, Covone F, et al. Arrest of G(1)-S progression by the p53inducible gene PC3 is Rb dependent and relies on the inhibition of cyclin D1 transcription. Mol Cell Biol. 2000;20(5):17971815. Bahram F, von der LN, Cetinkaya C, Larsson LG. c-Myc hot spot mutations in lymphomas result in inefficient ubiquitination and decreased proteasome-mediated turnover. Blood. 2000;95(6):2104-2110. Evrard SM, Pericart S, Grand D, et al. Targeted next generation sequencing reveals high mutation frequency of CREBBP, BCL2 and KMT2D in high-grade B-cell lymphoma with MYC and BCL2 and /or BCL6 rearrangements. Haematologica. 2019;104(4):e154-e157. Momose S, Weissbach S, Pischimarov J, et al. The diagnostic gray zone between Burkitt lymphoma and diffuse large B-cell lymphoma is also a gray zone of the mutational spectrum. Leukemia. 2015;29(8): 1789-1791. Wagener R, Seufert J, Raimondi F, et al. The mutational landscape of Burkitt-like lymphoma with 11q aberration is distinct from that of Burkitt lymphoma. Blood. 2019;133 (9):962-966. King RL, McPhail ED, Meyer RG, et al. False-negative rates for MYC FISH probes in B-cell neoplasms. Haematologica. 2019; 104(6):e248-e251. Bonetti P, Testoni M, Scandurra M, et al. Deregulation of ETS1 and FLI1 contributes to the pathogenesis of diffuse large B-cell lymphoma. Blood. 2013;122(13):2233-2241. Rinaldi A, Kwee I, Poretti G, et al. Comparative genome-wide profiling of post-transplant lymphoproliferative disorders and diffuse large B-cell lymphomas. Br J Haematol. 2006;134(1):27-36. Rinaldi A, Capello D, Scandurra M, et al. Single nucleotide polymorphism-arrays provide new insights in the pathogenesis of post-transplant diffuse large B-cell lymphoma. Br J Haematol. 2010;149(4):569577.

1829

ARTICLE Ferrata Storti Foundation

Chronic Lymphocytic Leukemia

Energy metabolism is co-determined by genetic variants in chronic lymphocytic leukemia and influences drug sensitivity

Junyan Lu,1* Martin Böttcher,2* Tatjana Walther,3 Dimitrios Mougiakakos,2 Thorsten Zenz3,4 and Wolfgang Huber1

European Molecular Biology Laboratory (EMBL), Heidelberg, Germany; 2Department of Internal Medicine 5, Hematology and Oncology, University of Erlangen-Nuremberg, Erlangen, Germany; 3Molecular Therapy in Hematology and Oncology, National Center for Tumor Diseases and German Cancer Research Centre, Heidelberg, Germany and 4 Department of Medical Oncology and Hematology, University Hospital Zürich and University of Zürich, Zürich, Switzerland 1

Haematologica 2019 Volume 104(9):1830-1840

*JL and MB contributed equally to this work.

ABSTRACT

Correspondence: WOLFGANG HUBER wolfgang.huber@embl.de THORSTEN ZENZ thorsten.zenz@usz.ch DIMITRIOS MOUGIAKAKOS dimitrios.mougiakakos@uk-erlangen.de Received: July 30, 2018. Accepted: February 14, 2019. Pre-published: February 21, 2019. doi:10.3324/haematol.2018.203067 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1830

hronic lymphocytic leukemia cells have an altered energy metabolism compared to normal B cells. While there is a growing understanding of the molecular heterogeneity of the disease, the extent of metabolic heterogeneity and its relation to molecular heterogeneity has not been systematically studied. Here, we assessed 11 bioenergetic features, primarily reflecting cell oxidative phosphorylation and glycolytic activity, in leukemic cells from 140 chronic lymphocytic leukemia patients using metabolic flux analysis. We examined these bioenergetic features for relationships with molecular profiles (including genetic aberrations, transcriptome and methylome profiles) of the tumors, their ex vivo responses to a panel of 63 compounds, and with clinical data. We observed that leukemic cells with mutated immunoglobulin variable heavy-chain show significantly lower glycolytic activity than cells with unmutated immunoglobulin variable heavy-chain. Accordingly, several key glycolytic genes (PFKP, PGAM1 and PGK1) were found to be down-regulated in samples harboring mutated immunoglobulin variable heavy-chain. In addition, 8q24 copy number gains, 8p12 deletions, 13q14 deletions and ATM mutations were identified as determinants of cellular respiration. The metabolic state of leukemic cells was associated with drug sensitivity; in particular, higher glycolytic activity was linked to increased resistance towards several drugs including rotenone, navitoclax, and orlistat. In addition, we found glycolytic capacity and glycolytic reserve to be predictors of overall survival (P<0.05) independently of established genetic predictors. Taken together, our study shows that heterogeneity in the energy metabolism of chronic lymphocytic leukemia cells is influenced by genetic variants and this could be therapeutically exploited in the selection of therapeutic strategies.

©2019 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.

1830

Introduction Resistance to apoptosis rather than aberrant proliferation is regarded as the reason for chronic lymphocytic leukemia (CLL) cell accumulation. However, active proliferation also contributes to CLL pathogenesis, as sizable clonal birth rates were observed in this disease.1,2 This suggests a substantial bioenergetic demand for proliferating subsets of CLL cells in order to support cell growth and division. Deregulated energy metabolism is considered to be one of the hallmarks of cancer.3 While molecular mechanisms promoting survival and proliferation of CLL cells have been extensively studied, fewer studies have addressed energy metabolism in CLL. Garcia-Manteiga et al. suggested oxidative phosphorylation as the primary haematologica | 2019; 104(9)

Characterizing energy metabolism of CLL

Table 1. Results of multivariate Cox regression model for overall survival (n=119, events =18) by including either glycolytic reserve or glycolytic capacity as a predictor.

Multivariate Cox model including glycolytic reserve Factor Glycolytic reserve U-CLL Treatment Trisomy12 Age TP53 mutations 11q22.3 deletions 17p13 deletions

Multivariate Cox model including glycolytic capacity Factor Gycolytic capacity U-CLL Treatment Trisomy12 TP53 mutation 11q22.3 deletions Age 17p13 deletions

Hazard Ratio

Lower 95% CI

Upper 95% CI

0.033 0.095 0.206 0.265 0.413 0.504 0.629 0.790

1.10 3.00 2.50 2.40 1.20 1.60 0.71 0.80

1.00 0.83 0.61 0.52 0.79 0.42 0.17 0.16

1.20 11.00 9.90 11.00 1.80 5.90 2.90 4.00

Hazard Ratio

Lower 95% CI

Upper 95% CI

0.046 0.101 0.178 0.312 0.469 0.494 0.546 0.644

1.10 2.90 2.60 2.20 1.70 0.61 1.10 0.68

1.00 0.81 0.65 0.48 0.42 0.15 0.76 0.13

1.10 10.00 10.00 9.70 6.50 2.50 1.70 3.60

CI: Confidence Interval; U-CLL: chronic lymphocytic leukemia cells with unmutated IGHV genes.

source of energy.4 This hypothesis is supported by subsequent findings that aerobic mitochondrial respiration results in high levels of oxidative stress of circulating CLL cells5 and that targeting the respiratory machinery can be therapeutically exploited to achieve selective toxicity.6 However, Maclntyre et al. reported increased concentrations of pyruvate and glutamate in serum samples from CLL patients as compared to healthy donors, which suggests active glycolysis.7 It has been well established that genetic heterogeneity contributes to the variable clinical outcomes of CLL. Based on the somatic mutation status in the variable regions of the immunoglobulin (Ig) heavy chain (IGHV) genes, CLL can be divided into two subgroups with distinct prognosis: CLL cells with unmutated IGHV genes (U-CLL) display higher B-cell receptor (BCR) signaling activity and are more aggressive than CLL cells with mutated IGHV genes (M-CLL). Serum samples from U-CLL patients were found to contain higher levels of lactate, fumarate, and uridine than those from M-CLL patients,7 suggesting U-CLL cells might have higher rates of aerobic glycolysis. This finding is in line with the observation that normal B cells undergo a metabolic switch from oxidative phosphorylation towards glycolysis upon BCR stimulation.4 However, considering the number of clinically relevant genetic alterations documented in CLL,8,9 the relationship between genetic heterogeneity and energy metabolism remains largely unexplored. Our previous work showed that many of the recurrent mutations influence drug sensitivities of CLL.10 As metabolic reprogramming has been shown to affect drug responsiveness of various cancers,2,11,12 metabolism may serve as a promising target for overcoming drug resistance in CLL. To gain a better understanding of the metabolic landscape of CLL tumor cells in relation to their genetic profile, haematologica | 2019; 104(9)

and to determine the role of metabolism in the response to drug treatments, we assessed the bioenergetic features of primary CLL samples (n=140 patients) through extracellular flux assays investigating two major metabolic processes: 1) aerobic glycolysis; and 2) oxidative phosphorylation. We performed an integrative analysis of these data with previously recorded ex vivo responses of the same samples to a panel of 63 drugs, somatic genome mutations, tumor transcriptomes, DNA methylomes, and clinical data.10 We found multiple associations between the mutational status and bioenergetic features, and found glycolysis activity of CLL cells contributed to resistance towards compounds targeting mitochondria-related biological processes that include rotenone, orlistat, venetoclax, and navitoclax. In addition, glycolytic capacity and glycolytic reserve features were shown to provide additional information to known genomic markers, such as IGHV and TP53, for predicting overall survival (OS).

Methods Extracellular flux assays Extracellular flux analyses (illustrated in Online Supplementary Figure S1) were performed on 152 CLL samples and nine B-cell samples from healthy donors on a Seahorse XFe96 system as previously described.13 The resulting data files (*.asyr) were converted to comma-separated value (CSV) files using the Wave Desktop software package (Agilent/Seahorse Bioscience) and imported into R for quality assessment and further analysis. The data for 140 of the 152 CLL samples passed quality control and were used for subsequent analyses. A detailed description of the workflow and criteria for quality control are provided in the Online Supplementary Methods. 1831

J. Lu et al. A

Figure 1. Difference in energy metabolism between chronic lymphocytic leukemia (CLL) cells and normal B cells. (A) Scatterplot of the top two principal components of the 11 tested bioenergetic features. Each dot represents a CLL patient sample (blue) or a healthy-donor derived B cell (red). (B) Beeswarm plots showing differences of six of the bioenergetic features between B-cell samples (n=9) and CLL samples (n=140).

Integrative data analysis Analyses were performed using R 3.4 and included univariate association tests, multivariate regression with and without lasso penalization, Cox regression, generalized linear models, principal component analysis, and gene set enrichment analysis. For association tests between bioenergetic features and genetic variants (i.e. copy number variants and gene mutations), only those with five or more variant cases were included. Summary statistics of patientsâ&#x20AC;&#x2122; demographic and clinical features are provided in Online Supplementary Table S1. All P-values from association tests were adjusted for multiple testing by applying the Benjamini-Hochberg procedure to control false discovery rate (FDR). Further details are provided in the Online Supplementary Methods.

Data availability Our data and analysis are provided as a reader-reproducible pipeline supported by the R package seahorseCLL (https://github.com/lujunyan1118/seahorseCLL). An online platform based on R Shiny (http://mozi.embl.de/public/seahorseCLL) is also provided for reference and to visualize our dataset.

Study approval The study was approved by the Ethics Committee Heidelberg (University of Heidelberg, Germany; S-206/2011; S-356/2013). Patients who donated tumor material provided written informed consent prior to study.

Results Chronic lymphocytic leukemia cells and B cells show distinct energy metabolic phenotypes We first compared the energy metabolic profiles of the 140 CLL samples and nine B-cell samples from healthy donors. In a principal component analysis (PCA) (Figure 1A), the CLL samples were clearly separated from the Bcell samples, which indicates that CLL cells have a distinct metabolic phenotype. Nine of the 11 bioenergetic features 1832

showed altered levels between CLL cells and B cells (ANOVA test, Benjamini and Hochberg multiple testing method for FDR = 5%) (Online Supplementary Table S2). In accordance with a previous report,6 mitochondrial respiration-related features, including basal respiration, maximal respiration, and ATP production were increased in CLL cells (Figure 1B). With regard to aerobic glycolysis, no significant differences were seen in basal glycolysis activity between CLL and B cells. However, CLL cells showed elevated glycolytic capacity and glycolytic reserve (Figure 1B). As these two features measure the maximum capability of cells for glycolysis and the flexibility of cells to respond to energetic demands, this observation suggests an increased adaptability of CLL cells to use glycolysis as an energy source when needed, although they do not primarily rely on it.

Molecular determinants of energy metabolism in chronic lymphocytic leukemia Figure 1 shows a variability among the bioenergetic profiles of the CLL samples. We hypothesized that this variability may be related to the molecular heterogeneity of CLL.8,9 Therefore, we tested the tumor-to-tumor variations of the bioenergetic features for possible correlations with 20 molecular features, including recurrent somatic mutations and copy number variations, IGHV status and methylation clusters (Figure 2A and Online Supplementary Figure S2). The most prominent association identified was IGHV status: IGHV mutated CLL (M-CLL) samples had lower glycolytic activity and glycolytic capacity than IGHV unmutated CLL (U-CLL) samples (Figure 2B). Patients with M-CLL and U-CLL have been observed to have distinct serum metabolite profiles; U-CLL patients have higher lactate level in serum, which can be considered a sign of elevated glycolysis.7 To our knowledge, our large sample size study provides the first direct proof that U-CLL do indeed have a higher glycolytic activity than M-CLL. haematologica | 2019; 104(9)

Characterizing energy metabolism of CLL

Figure 2. Associations between genetic variants and bioenergetic features. (A) The distribution of P-values of the associations between each genetic variant and each energy metabolic feature (ANOVA test). Gray: associations that did not pass a threshold corresponding to a 5% false discovery rate (FDR) (Benjamini and Hochberg method); red: associations with higher bioenergetic values in mutated cases; blue: associations with lower bioenergetic values in mutated cases (or highprogrammed subtype). (B and C) Examples of associations, visualized in beeswarm plots. (B) Glycolysis and IGHV status. (C) Glycolysis and DNA methylation cluster.

IGHV status is strongly associated with three subtypes of CLL defined by their global levels of CpG methylation.14 Accordingly, we found that the high-programmed CLL (HP-CLL) subtype, which has higher global methylation level, had a lower glycolysis activity than the low-programmed CLL (LP-CLL) subtype (Figure 2C). To further dissect the role of IGHV status in metabolic reprogramming, we analyzed transcriptome data that we had measured for 120 of these patient samples (of which 111 had annotation for IGHV status). We performed gene set enrichment analysis on the genes that were differentially expressed between M-CLL and U-CLL samples using the Hallmark gene sets from Molecular Signature Database (MsigDB).15 We found that genes down-regulated in M-CLL were enriched in the glycolysis pathway (Figure 3A). Thirty-four glycolysis-related genes were down-regulated in M-CLL (Figure 3B), including several that encode key enzymes PFKP (Phosphofructokinase, platelet), PGAM1 (Phosphoglycerate Mutase 1), and PGK1 (Phosphoglycerate kinase 1) (Figure 3C).16-18 This analysis suggests that IGHV status directly influences the expression of genes related to glycolysis resulting in the observed difference in glycolytic parameters between M-CLL and U-CLL. As IGHV status reflects the B-cell receptor (BCR) signaling activity,19 we referred to two published datasets for the transcriptomic signatures of BCR stimulation in CLL, either by anti-IgM antibody20 (GEO ID: GSE49695) or unmethylated bacterial DNA (CpG) (GEO ID: GSE30105). In both conditions, genes that were up-regulated after BCR stimulation were significantly enriched in the glycolysis pathway (Online Supplementary Figure S3). haematologica | 2019; 104(9)

Together these results indicate a causal link from BCR signaling to glycolysis activity in CLL, in line with previous evidence.21,22 We also identified several other novel associations between bioenergetic features and genetic variants (Online Supplementary Figure S4). Gain of 8q24, deletion of 8p12, ATM mutation, EGR2 mutation and MED12 mutation were found to be associated with higher values of respiration-related features such as ATP production and maximal respiration, while tumors with chromothripsis showed lower oxygen consumption rate (OCR) values.

Glycolytic activity contributes to drug resistance in chronic lymphocytic leukemia Sensitivity to drugs is an informative cellular phenotype that reflects pathway dependencies of tumor cells.10 Therefore, we asked how the 11 intrinsic bioenergetic features were related to the vulnerabilities of CLL cells towards a panel of 63 drugs applied ex vivo. This panel comprised clinically used drugs as well as small molecule probes of pathways important in leukemia. Using the Pearson correlation test, we identified 118 significant (FDR=10%) associations between drug sensitivities and bioenergetic features (Figure 4A and Online Supplementary Figure S5). Thirty-two drugs had at least one significant association with a bioenergetic feature. A significant association between a bioenergetic feature and an ex vivo drug response indicates that the sensitivity or resistance of CLL samples to the drug is affected by the intrinsic activity of the bioenergetic feature. At an aggregate level, glycolysis-related features of the 1833

J. Lu et al.

CLL cells were positively correlated with the viabilities of those cells after drug treatment, while respiration-related features were negatively correlated. This suggests that higher glycolysis activity of CLL cells reduces sensitivity to drugs, while higher respiration activity contributes to increased sensitivity ex vivo. There were more specific patterns for drugs with different target profiles. CLL samples with higher respiration activity were more sensitive to kinase inhibitors, including the inhibitors of Brutonâ&#x20AC;&#x2122;s tyrosine kinase (BTK), ibrutinib, and of spleen tyrosine kinase, tamatinib, both of which target the BCR pathway. In addition, two checkpoint kinase 1 (Chk1) inhibitors, AZD7762 and PF-477736, and the heat shock protein 90 (Hsp90) inhibitor AT13387 showed similar association patterns, which is in line with the report that they also target the BCR signaling cascade.10 Viabilities after treatment of drugs targeting mitochondria-related biological processes (rotenone, venetoclax and navitoclax) were positively correlated with the glycolysisrelated features (Figure 4A and Online Supplementary Figure S6) for most of the drug concentrations (Online Supplementary Figure S5); the multivariate test results show that this finding is not merely due to confounding by IGHV status (Online Supplementary Figure S7). Rotenone is a mitochondrial complex I inhibitor, which disrupts the electron transport chain and thus blocks cellular respiration. Therefore, the correlation between rotenone response and glycolysis activity can be explained by the fact that higher glycolysis activity or potential (with increased metabolic flexibility) can compensate for cytotoxic effects of respiration inhibition by providing an alternative way of producing ATP. Venetoclax and navitoclax

are BH3-mimetics that target the BCL2 protein and lead to mitochondrial damage and the inhibition of oxidative respiration.23 Thus, lower reliance on oxidative respiration is a plausible explanation for the resistance to BH3-mimetics of CLL cells with high glycolysis activity. We also observed associations between glycolysis-related features and the responses to orlistat, an anti-obesity drug, which has also been identified as a pro-apoptotic agent in CLL by inhibiting lipoprotein lipase (LPL),24 and KX2-391, an inhibitor of the proto-oncogene tyrosine-protein kinase Src (Online Supplementary Figure S6). We previously showed that although drug response phenotypes of CLL cells were largely influenced by genetic variants, there was still substantial variance in the drug response phenotypes that were not explained by genetics. Thus, we asked whether the energy metabolism profile could add additional predictive information. For each drug, we built two multivariate linear regression models to predict its response profile: one included only the 20 genetic features shown in Online Supplementary Figure S2 as predictors, the other included these genetic features plus 11 bioenergetic features. As a measure of predictive strength, we compared the variance explained (R2 value adjusted by numbers of predictors) between the two models. For most drugs, including bioenergetic features in the model did not increase explanatory power (Figure 4B, dots on diagonal line); moreover, responses to individual kinase inhibitors were well explained by the genetic features (blue dots in Figure 4B and Online Supplementary Figure S8). However, for five drugs, including venetoclax and rotenone, the variance explained increased by 10% or more upon inclusion of the bioenergetic features (red dots

Figure 3. Genes from the glycolysis pathway are down-regulated in immunoglobulin variable heavy-chain (IGHV) gene mutated chronic lymphocytic leukemia (MCLL) samples. (A) Hallmark gene sets that are significantly (10%; Benjamini and Hochberg method for false discovery rate) enriched among genes differentially expressed between M-CLL and unmutated CLL (U-CLL). (B) Heatmap showing z-score of the expression values of glycolysis pathway genes that are differentially expressed between M-CLL and U-CLL samples. (C) Beeswarm plots for the expression values of three key genes in the glycolysis pathway: PFKP (Phosphofructokinase, platelet), PGAM1 (Phosphoglycerate Mutase 1), and PGK1 (Phosphoglycerate kinase 1).

1834

haematologica | 2019; 104(9)

Characterizing energy metabolism of CLL

in Figure 4B). In addition, except for cephaeline, bioenergetic features were more significant than genetic features in the multivariate models (Figure 4C).

Association between clinical course and energy metabolism of chronic lymphocytic leukemia The use of primary patient cells enabled us to investigate the associations between bioenergetic features with patient history or outcome in CLL. In our study cohort, 43 patients had received treatment before sample collection,

in all cases with chemotherapeutic agents (Online Supplementary Table S1), and none of them was undergoing treatment when samples were collected. Therefore, we first asked whether these completed treatments prior to sample collection affected the energy metabolism of primary tumor samples, as studies have shown chemotherapy or targeted therapy could drive clonal evolution leading to drug resistance or oxidative stress.25-27 We found two bioenergetic features, namely glycolytic capacity and glycolytic reserve, associated with pretreatment

Figure 4. Correlation test results between drug response phenotypes and bioenergetic features. (A) y-axis: negative logarithm of the Pearson correlation test P-values. Only drugs with at least one significant association with bioenergetic features are shown (Benjamini and Hochberg method for false discovery rate (FDR)] = 10%). Viabilities across different drug concentrations were aggregated using Tukeyâ&#x20AC;&#x2122;s median polish method. Correlations with glycolysis-related features are in warm colors and correlations with respiration-related features are in cold colors. The dotted line indicates the P-value threshold given by the Benjamini and Hochberg method for FDR (10%). (B) Comparison of explained variance of drug responses between the multivariate model, including only genetic features, and the model including genetic and bioenergetic features. (C) Red: predictors with significant (<0.05) P-values in multivariate models for the drugs; red bar: a positive association with drug responses (higher drug sensitivity is associated with presence of the mutation or higher value of the bioenergetic feature); blue bar: negative association.

haematologica | 2019; 104(9)

1835

J. Lu et al.

status at a significance threshold of P<0.05 (Online Supplementary Table S3 and Online Supplementary Figure S9). However, pretreatment status was also highly correlated with IGHV (P=0.0006, Ď&#x2021;2 test). This reflects the fact that U-CLL patients more frequently receive treatment due to faster progression. Furthermore, glycolytic capacity and reserve are correlated with IGHV status based on our analysis (see above). Thus, to dissect confounding from more direct association, we included IGHV status as a blocking factor in a multivariate model. In this more indepth analysis, no significant association between pretreatment status and bioenergetic features was detected (P<0.05). In a second analysis to assess potential roles of pretreatment status on the biology of the tumor samples, we revisited our association tests between the bioenergetic features and: i) the genetic variants; and ii) the drug responses. Including pretreatment status as a blocking factor had negligible impact on directions, strengths and

P-values of these associations (Online Supplementary Figure S10). Together, these results indicate that the treatments experienced by 43 of our patients led to no detectable differences between the metabolic phenotypes of their circulating CLL cell samples and those of the other 97 patients. Therefore, we proceeded with the subsequent analysis using the combined dataset of 140 samples. Returning to clinical outcomes, we considered two end points: time to treatment (TTT) and OS. Univariate Cox regression models indicated that glycolytic reserve, maximal respiration, and spare respiratory capacity were associated with TTT, and glycolytic capacity and glycolytic reserve were associated with OS (P<0.05) (Online Supplementary Figure S11). Samples with higher values of these features were associated with worse clinical outcomes, i.e. shorter time to treatment and OS. In multivariate Cox models including age, trisomy 12, deletion of 11q22.3, deletion of 17p13, TP53 mutation and IGHV sta-

C D

Figure 5. Associations between bioenergetic features and clinical course. (A and B) Kaplan-Meier plots for overall survival (OS) stratified by immunoglobulin variable heavy-chain (IGHV) gene status and glycolytic capacity (A) or glycolytic reserve (B). The cutoff to define high and low bioenergetic groups was estimated by maximally selected rank test. The cutoff value and number of samples in each group are shown inside the parentheses in the figure panels. (C and D) Scatter plots for associations of CD38 expression with glycolytic capacity and glycolytic reserve.

1836

haematologica | 2019; 104(9)

Characterizing energy metabolism of CLL

tus as co-variates, bioenergetic features were not picked up as predictive for TTT (Online Supplementary Tables S4S6). However, glycolytic capacity and glycolytic reserve were the most significant predictors for OS also in the multivariate Cox models (Table 1), indicating that these two glycolysis-related features provide additional OSrelated information to established variables such as IGHV status, one of the most reliable prognostic markers in CLL. M-CLL patients with low glycolytic capacity or reserve showed best prognosis, U-CLL patients with high glycolytic capacity or reserve showed worst prognosis, while the other two groups lie in between (Figure 5A and B). We also investigated associations of each bioenergetic feature to clinically relevant phenotypes including CD38

expression, CD49d (IGTA4) expression, and lymphocyte doubling time (LDT), which are considered as indicators for CLL progression.28-31 Again, we considered IGHV status as a potential confounder (Online Supplementary Tables S7 and S8). There were significant correlations between CD38 gene expression with glycolytic capacity and glycolytic reserve (5% FDR) (Figure 5C and D). As well as the known fact that CD38 expression is highly associated with IGHV status,32 we found that it was positively correlated to glycolytic capacity or glycolytic reserve in both M-CLL and U-CLL disease subgroups (Online Supplementary Figure S12). This result suggests an IGHV status-independent link between CD38 activity and adaptability of CLL cells to glycolysis as an energy source.

Figure 6. Multivariate regression models for energy metabolism features. (A) Explanatory power (cross-validated R2) of datasets of different data types for the prediction of the energy metabolic features. Error bars represent standard deviations of R2 over 100 repeated cross-validations. Numbers in parentheses after dataset names indicate the number of variables in the dataset. (B) Visualization of fitted adaptive L1 (lasso) regularization multivariate models using drug responses, gene mutations, immunoglobulin variable heavy-chain (IGHV) gene status, pretreatment status, and the top 20 principal components of the gene expression (RNASeq) data. Numbers in parentheses indicate the number of samples used for the regression. (Bottom) Scatterplot of Z-scores of the energy metabolic features (i.e. centered by mean and scaled by standard deviation). (Middle) Predictor values. The continuous variables [drug responses and gene expression principal components (PC)] are shown centered and scaled using the red-white-blue color representation, the binary variables (genetic variants, IGHV status) in black and white (black: mutation present). (Left) Horizontal bars show average model coefficients over 100 repeated cross-validations. Only the features that were selected at least 80 times out of 100 repeats are shown.

haematologica | 2019; 104(9)

1837

J. Lu et al.

The complex network of chronic lymphocytic leukemia energy metabolic predictors While the analyses presented so far provide insights on pairwise associations between bioenergetic features and other tumor properties, we next aimed to create a systems-level map of the network of gene mutations, DNA methylation, gene expression, ex vivo drug responses, and bioenergetic features. We used multivariate linear regression with lasso regularization to predict each bioenergetic feature by other available biological features and measured prediction performance using cross-validated R2 (Figure 6). We first assessed to what extent each omics data type alone, or the combination of all the datasets, explained each bioenergetic features. The gene expression data and the drug response data performed best in predicting bioenergetic features (Figure 6A). Combining all datasets slightly increased the predictive power for each metabolic feature, indicating that each set contains non-redundant information. Notably, the glycolysis-related features were better explained by the multi-omics data than the respiration-related features (Figure 6A and Online Supplementary Figure S13). We visualized predictor profiles for individual bioenergetic features, focusing on the ex vivo drug responses, gene expressions, and genetic variants (Figure 6B and Online Supplementary Figure S13). In accordance with the above univariate analysis, the multivariate model identified IGHV status and response to mitochondria-targeting drugs like venetoclax and rotenone as important predictors for glycolysis-related features. In addition, SF3B1 mutation was identified as one of the top predictors for glycolytic capacity and reserve, as its presence is associated with higher values. SF3B1 is an mRNA splicing factor that is frequently mutated in CLL and associated with more aggressive disease and worse survival, but its oncogenic mechanism is still elusive.33 Another genomic aberration, deletion of 13q14, was selected as one of the top predictors for basal respiration and ATP production. Several principal components (PC) from the gene expression datasets were also identified by the multivariate modeling. PC8 was the top predictor with positive coefficient for all respiration-related features. As the genes with high positive loadings on PC8 are enriched in E2F targets, this suggests that higher expression of E2F targets associates with higher respiratory activity in CLL cells. On the other hand, PC10 was the top predictor, with negative coefficient, for maximal respiration and spare respiratory capacity (Online Supplementary Figure S14). Based on enrichment analysis, genes with high negative loadings on PC10 are enriched in the mTOR pathway and therefore this also suggests higher mTOR pathway activity associates with high respiration capability. These findings are in line with previous reports that E2F transcription factors and mTOR pathway are key players in regulating mitochondrial activity.34,35 PC 2, 4, 6 and 11 were identified as predictors for several glycolysis-related features (Figure 6B and Online Supplementary Figure S13). Gene set enrichment analysis highlighted TNFa-NFκB signaling as the most enriched pathway for genes with high loadings on PC2, 4 and 6 (Online Supplementary Figure S14). This finding is consistent with previous reports that NFκB signaling pathway controls energy homeostasis in inflammatory and cancer cells.36 As we also found NFκB activation signatures in the 1838

two published transcriptomic profiling datasets of BCR stimulation (Online Supplementary Figure S3), which is in line with previous reports that BCR stimulation activate NFκB, we suggest that NFκB activation may play a role in increased glycolysis after BCR activation.37,38

Discussion In this study, we identified molecular features that underlie the heterogeneity of energy metabolism in CLL and linked bioenergetic features with ex vivo drug responses and clinical course. We found that, although CLL cells and B cells have a similar basal glycolytic activity, CLL cells had a significantly higher glycolytic capacity and glycolytic reserve, which are both indicators for the cell’s potential to switch to glycolysis as an energy source when necessary. Interestingly, we also found glycolytic capacity and reserve, but not basal glycolysis, to be novel predictors for OS in our cohort; CLL patients with higher glycolytic capacity and reserve showed worse prognosis. In addition, higher glycolytic capacity and reserve were also found to be correlated with high expression of the CD38 gene, a cell surface marker of B-cell activation and a negative prognostic marker in CLL. These observations can be viewed in the context of a recent report of the increased reliance of CLL cells on aerobic glycolysis to produce energy after a glycolytic switch induced by their contact with stromal cells.39 Although we assayed circulating CLL cells for our study, the glycolytic capacity and reserve in the flux assay may actually measure the ability of CLL cells to adapt to glycolysis in a stimulated state, similar to the stimulation by stromal cells. Our findings thus imply that circulating CLL cells may have previously undergone such metabolic reprogramming and carry the metabolic repertoire that allows them to quickly switch to glycolysis when a suitable stimulation occurs, e.g. upon stromal contact. Our findings also suggest that the magnitude and efficiency of this switch can further impact the prognosis of CLL patients. We showed that U-CLL has significantly higher glycolytic rates, which validates the previous hypothesis that U-CLL may have higher reliance on aerobic glycolysis due to higher BCR signaling pathway activity.4,7 In addition, we illustrated that the glycolysis pathway is more active in U-CLL than M-CLL, accompanied by an upregulation of key enzymes regulating cellular glycolysis. This indicates that M-CLL and U-CLL have intrinsically different energy metabolisms and that the BCR signaling pathway may have a direct impact on the metabolic reprogramming. We had previously attempted to monitor circulating CLL cells in vivo by using fluorodeoxyglucose positron emission tomography (FDG-PET), which pinpoints anatomical locations with high rate of glycolysis.40 This attempt failed due to insufficient sensitivity, and our results suggest that considering the difference between the M-CLL and U-CLL subtypes could increase the sensitivity of this diagnostic approach. We found that the CLL patient samples with gain of 8q24 showed increased respiratory activity. The likely reason for this is the oncogenic activity of the extra copy of the MYC proto-oncogene. Previous studies have shown that MYC substantially contributes to mitochondrial biogenesis, and the overexpression of MYC leads to increased respiratory capability in several cell line models, which is haematologica | 2019; 104(9)

Characterizing energy metabolism of CLL

in line with our observation.41 In our study, we also highlighted the possibility of exploiting heterogeneity of energy metabolism to improve individualized patient care. We show that higher glycolytic flexibility can contribute to the resistance of CLL samples to treatment with drugs that affect mitochondria, such as rotenone, venetoclax, and navitoclax. We postulate that the cytotoxic effects of these drugs may partially result from restricting the energy supply by blocking cellular respiration and thus, cells with higher glycolytic potential can counteract their effect due to higher metabolic flexibility. The current study has certain limitations. Firstly, while most of the proliferative activity of CLL cells appears in lymph node and bone marrow, in this study we only used circulating CLL cells due to the easier availability of patient material, which was instrumental in providing an adequate study size. In addition, although we observed many biologically meaningful associations, these are generally weak, as indicated by the relatively small effect sizes or correlation coefficients. While it is possible that biological variables not measured by us contribute to the heterogeneity in energy metabolism, a likely explanation could be biological noise (since we are using patient samples instead of cell lines) and technical noise of the Seahorse extracellular flux measurements, and the other

References 1. Messmer BT, Messmer D, Allen SL, et al. In vivo measurements document the dynamic cellular kinetics of chronic lymphocytic leukemia B cells. J Clin Invest. 2005; 115(3):755-764. 2. van Gent R, Kater AP, Otto SA, et al. In vivo Dynamics of Stable Chronic Lymphocytic Leukemia Inversely Correlate with Somatic Hypermutation Levels and Suggest No Major Leukemic Turnover in Bone Marrow. Cancer Res. 2008;68(24):10137-10144. 3. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011; 144(5):646-674. 4. Garcia-Manteiga JM, Mari S, Godejohann M, et al. Metabolomics of B to plasma cell differentiation. J Proteome Res. 2011; 10(9):4165-4176. 5. Moran EC, Kamiguti AS, Cawley JC, Pettitt AR. Cytoprotective antioxidant activity of serum albumin and autocrine catalase in chronic lymphocytic leukaemia. Br J Haematol. 2002;116(2):316-328. 6. Jitschin R, Hofmann AD, Bruns H, et al. Mitochondrial metabolism contributes to oxidative stress and reveals therapeutic targets in chronic lymphocytic leukemia. Blood. 2014;123(17):2663v2672. 7. MacIntyre DA, Jiménez B, Lewintre EJ, et al. Serum metabolome analysis by 1HNMR reveals differences between chronic lymphocytic leukaemia molecular subgroups. Leukemia. 2010;24(4):788-797. 8. Zenz T, Mertens D, Küppers R, Döhner H, Stilgenbauer S. From pathogenesis to treatment of chronic lymphocytic leukaemia. Nat Rev Cancer. 2010;10(1):37-50. 9. Fabbri G, Dalla-Favera R. The molecular pathogenesis of chronic lymphocytic leukaemia. Nat Rev Cancer. 2016;1

haematologica | 2019; 104(9)

assays used. Indeed, our study is, to our knowledge, the first that uses such a dynamic assay to systematically interrogate energy metabolism on such a large scale. Taken together, our in-depth characterization of energy metabolism and integrative analyses provide valuable insights on mechanisms underlying the metabolic regulation of CLL cells, and reveal the possibilities of guiding clinical diagnosis and individualized patient care based on metabolic profiles. Our large-scale energy metabolism dataset complements the current traditional omics datasets, such as RNA sequencing, DNA sequencing, and methylation profiling, and contribute to a better understanding of CLL biology. Acknowledgments The authors thank the reviewers for helpful suggestions and comments, which improved the quality of this work. Funding The work was supported by the European Union (Horizon 2020 project SOUND) and GCH-CLL project co-founded by the European Commission/DG Research and Innovation. DM was supported by the Else Kröner-Fresenius-Stiftung. DM and MB were supported by the Erich und Gertrud RoggenbuckStiftung. TZ was supported by the Monique Dornonville de la Cour Stiftung and the Cancer Research Center (CRC) Zurich.

6(3):145-162. 10. Dietrich S, Oleś M, Lu J, et al. Drug-perturbation-based stratification of blood cancer. J Clin Invest. 2018;128(1):427-445. 11. Wu S-H, Bi J-F, Cloughesy T, Cavenee WK, Mischel PS. Emerging function of mTORC2 as a core regulator in glioblastoma: metabolic reprogramming and drug resistance. Cancer Biol Med. 2014; 11(4):255-263. 12. Long Y, Tsai W-B, Wangpaichitr M, et al. Arginine Deiminase Resistance in Melanoma Cells Is Associated with Metabolic Reprogramming, Glucose Dependence, and Glutamine Addiction. Mol Cancer Ther. 2013;12(11):2581-2590. 13. Böttcher M, Renner K, Berger R, et al. D-2hydroxyglutarate interferes with HIF-1a stability skewing T-cell metabolism towards oxidative phosphorylation and impairing Th17 polarization. Oncoimmunology. 2018;e1445454. 14. Oakes CC, Seifert M, Assenov Y, et al. DNA methylation dynamics during B cell maturation underlie a continuum of disease phenotypes in chronic lymphocytic leukemia. Nat Genet. 2016;48(3):253-264. 15. Liberzon A, Birger C, Thorvaldsdóttir H, Ghandi M, Mesirov JP, Tamayo P. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst. 2015;1(6):417-425. 16. Semenza GL, Roth PH, Fang HM, Wang GL. Transcriptional regulation of genes encoding glycolytic enzymes by hypoxiainducible factor 1. J Biol Chem. 1994; 269(38):23757-23763. 17. Hitosugi T, Zhou L, Elf S, et al. Phosphoglycerate Mutase 1 Coordinates Glycolysis and Biosynthesis to Promote Tumor Growth. Cancer Cell. 2012; 22(5):585-600. 18. Dunaway GA, Kasten TP, Sebo T, Trapp R.

19.

20.

21.

22.

23.

24.

25.

26.

Analysis of the phosphofructokinase subunits and isoenzymes in human tissues. Biochem J. 1988;251(3):677-683. Stevenson FK, Krysov S, Davies AJ, Steele AJ, Packham G. B-cell receptor signaling in chronic lymphocytic leukemia. Blood. 2011;118(16):4313-4320. Tavolaro S, Colombo T, Chiaretti S, et al. Increased chronic lymphocytic leukemia proliferation upon IgM stimulation is sustained by the upregulation of miR-132 and miR-212. Genes Chromosom Cancer. 2015; 54(4):222-234. Doughty CA, Bleiman BF, Wagner DJ, et al. Antigen receptor-mediated changes in glucose metabolism in B lymphocytes: Role of phosphatidylinositol 3-kinase signaling in the glycolytic control of growth. Blood. 2006;107(11):4458-4465. Qorraj M, Bruns H, Böttcher M, et al. The PD-1/PD-L1 axis contributes to immune metabolic dysfunctions of monocytes in chronic lymphocytic leukemia. Leukemia. 2017;31(2):470-478. Lagadinou ED, Sach A, Callahan K, et al. BCL-2 inhibition targets oxidative phosphorylation and selectively eradicates quiescent human leukemia stem cells. Cell Stem Cell. 2013;12(3):329-341. Pallasch CP, Schwamb J, Königs S, et al. Targeting lipid metabolism by the lipoprotein lipase inhibitor orlistat results in apoptosis of B-cell chronic lymphocytic leukemia cells. Leukemia. 2008;22(3):585-592. Ahn IE, Underbayev C, Albitar A, et al. Clonal evolution leading to ibrutinib resistance in chronic lymphocytic leukemia. Blood. 2017;129(11):1469-1479. Udensi UK, Tchounwou PB. Dual effect of oxidative stress on leukemia cancer induction and treatment. J Exp Clin Cancer Res. 2014;33:106.

1839

J. Lu et al.

27. Malcikova J, Stano-Kozubik K, Tichy B, et al. Detailed analysis of therapy-driven clonal evolution of TP53 mutations in chronic lymphocytic leukemia. Leukemia. 2015; 29(4):877-885. 28. Bulian P, Shanafelt TD, Fegan C, et al. CD49d is the strongest flow cytometrybased predictor of overall survival in chronic lymphocytic leukemia. J Clin Oncol. 2014;32(9):897-904. 29. Ibrahim S, Keating M, Do KA, et al. CD38 expression as an important prognostic factor in B-cell chronic lymphocytic leukemia. Blood. 2001;98(1):181-186. 30. Molica S, Alberti A. Prognostic value of the lymphocyte doubling time in chronic lymphocytic leukemia. Cancer. 1987; 60(11):2712-2716. 31. Viñolas N, Reverter JC, Urbano-Ispizua A, Montserrat E, Rozman C. Lymphocyte doubling time in chronic lymphocytic leukemia: an update of its prognostic significance. Blood Cells. 1987;12(2):457-470.

1840

32. Cruse JM, Lewis RE, Webb RN, Sanders CM, Suggs JL. Zap-70 and CD38 as predictors of IgVH mutation in CLL. Exp Mol Pathol. 2007;83(3):459-461. 33. Cazzola M, Rossi M, Malcovati L, Associazione Italiana per la Ricerca sul Cancro Gruppo Italiano Malattie Mieloproliferative. Biologic and clinical significance of somatic mutations of SF3B1 in myeloid and lymphoid neoplasms. Blood. 2013;121(2):260-269. 34. Benevolenskaya EV, Frolov MV. Emerging links between E2F control and mitochondrial function. Cancer Res. 2015;75(4):619623. 35. Morita M, Gravel S-P, Chénard V, et al. mTORC1 controls mitochondrial activity and biogenesis through 4E-BP-dependent translational regulation. Cell Metab. 2013; 18(5):698-711. 36. Tornatore L, Thotakura AK, Bennett J, Moretti M, Franzoso G. The nuclear factor kappa B signaling pathway: integrating

37.

38.

39.

40.

41.

metabolism with inflammation. Trends Cell Biol. 2012;22(11):557-566. Saba NS, Liu D, Herman SEM, et al. Pathogenic role of B-cell receptor signaling and canonical NF-κB activation in mantle cell lymphoma. Blood. 2016;128(1):82–92. Weil R, Israël A. T-cell-receptor- and B-cellreceptor-mediated activation of NF-κB in lymphocytes. Curr Opin Immunol. 2004; 16(3):374-381. Jitschin R, Braun M, Qorraj M, et al. Stromal cell-mediated glycolytic switch in CLL cells involves Notch-c-Myc signaling. Blood. 2015;125(22):3432-3436. Karam M, Novak L, Cyriac J, Ali A, Nazeer T, Nugent F. Role of fluorine-18 fluorodeoxyglucose positron emission tomography scan in the evaluation and follow-up of patients with low-grade lymphomas. Cancer. 2006;107(1):175–183. Morrish F, Hockenbery D. MYC and mitochondrial biogenesis. Cold Spring Harb Perspect Med. 2014;4(5):a014225–a014225.

haematologica | 2019; 104(9)

ARTICLE

Cell Therapy & Immunotherapy

CD20 and CD37 antibodies synergize to activate complement by Fc-mediated clustering

Simone C. Oostindie,1,2 Hilma J. van der Horst,3 Margaret A. Lindorfer,4 Erika M. Cook,4 Jillian C. Tupitza,4 Clive S. Zent,5 Richard Burack,5 Karl R. VanDerMeid,5 Kristin Strumane,1 Martine E. D. Chamuleau,3 Tuna Mutis,3 Rob N. de Jong,1 Janine Schuurman,1 Esther C. W. Breij,1 Frank J. Beurskens,1 Paul W. H. I. Parren2,6 and Ronald P. Taylor4

1 Genmab, Utrecht, the Netherlands; 2Department of Immunohematology and Blood Transfusion, Leiden University Medical Center, Leiden, the Netherlands; 3Department of Hematology, Amsterdam University Medical Center, Amsterdam, the Netherlands; 4 Department of Biochemistry and Molecular Genetics, University of Virginia School of Medicine, Charlottesville, Virginia, USA; 5Wilmot Cancer Institute, University of Rochester Medical Center, Rochester, NY, USA and 6Lava Therapeutics, Utrecht, the Netherlands

Ferrata Storti Foundation

Haematologica 2019 Volume 104(9):1841-1852

ABSTRACT

D20 monoclonal antibody therapies have significantly improved the outlook for patients with B-cell malignancies. However, many patients acquire resistance, demonstrating the need for new and improved drugs. We previously demonstrated that the natural process of antibody hexamer formation on targeted cells allows for optimal induction of complement-dependent cytotoxicity. Complement-dependent cytotoxicity can be potentiated by introducing a single point mutation such as E430G in the IgG Fc domain that enhances intermolecular Fc-Fc interactions between cell-bound IgG molecules, thereby facilitating IgG hexamer formation. Antibodies specific for CD37, a target that is abundantly expressed on healthy and malignant B cells, are generally poor inducers of complement-dependent cytotoxicity. Here we demonstrate that introduction of the hexamerization-enhancing mutation E430G in CD37-specific antibodies facilitates highly potent complement-dependent cytotoxicity in chronic lymphocytic leukemia cells ex vivo. Strikingly, we observed that combinations of hexamerization-enhanced CD20 and CD37 antibodies cooperated in C1q binding and induced superior and synergistic complement-dependent cytotoxicity in patient-derived cancer cells compared to the single agents. Furthermore, CD20 and CD37 antibodies colocalized on the cell membrane, an effect that was potentiated by the hexamerization-enhancing mutation. Moreover, upon cell surface binding, CD20 and CD37 antibodies were shown to form mixed hexameric antibody complexes consisting of both antibodies each bound to their own cognate target, so-called hetero-hexamers. These findings provide novel insights into the mechanisms of synergy in antibody-mediated complement-dependent cytotoxicity and provide a rationale to explore Fc-engineering and antibody hetero-hexamerization as a tool to enhance the cooperativity and therapeutic efficacy of antibody combinations.

Introduction Monoclonal antibodies (mAbS) have become the backbone of treatment regimens for several cancer indications. The chimeric immunoglobulin (Ig)G1 CD20 mAb rituximab was the first mAb approved for clinical use in cancer therapy. CD20 is expressed on more than 90% of mature B cells and rituximab is widely used to treat B-cell malignancies.1-3 However, many patients do not experience complete remission or acquire resistance to rituximab treatment, thereby demonstrating the need for improved mAb therapeutics or alternative tumor-targeting strategies.4-6 haematologica | 2019; 104(9)

Correspondence: SIMONE C. OOSTINDIE sio@genmab.com RONALD P. TAYLOR rpt@virginia.edu Received: October 2, 2018. Accepted: February 19, 2019. Pre-published: February 21, 2019. doi:10.3324/haematol.2018.207266 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1841 ÂŠ2019 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.

1841

S.C. Oostindie et al.

mAbs employ various mechanisms to eliminate cancer cells, such as induction of programmed cell death or Fcmediated effector functions, including antibody-dependent cell-mediated cytotoxicity (ADCC), antibody-dependent cellular phagocytosis (ADCP), and complementdependent cytotoxicity (CDC), which can be increased by Fc engineering.7 ADCC and ADCP, for example, can be enhanced by improving FcγR binding through Fc glycoengineering or amino acid modifications.8-11 Likewise, C1q binding and CDC can be increased by amino acid substitutions in Fc domains.12,13 CDC is initiated when membrane-bound antibodies bind the hexavalent C1q molecule, which together with C1r and C1s forms the C1 complex, the first component of the classical complement pathway. C1 activation triggers an enzymatic cascade that leads to covalent attachment of opsonins to target cells, and the generation of potent chemoattractants, anaphylatoxins and membrane attack complexes (MAC).14 IgG antibodies bound to cell surface antigens assemble into ordered hexamers, providing high avidity docking sites to which C1 binds and is activated.15 IgG hexamer formation and complement activation can be enhanced by single point mutations in IgG Fc domains, such as E430G, which increase interactions between Fc domains of cell-bound IgG.16 The hexamerization-enhanced (Hx) CD20-targeting mAb 7D8 displayed strongly enhanced CDC of B cells from patients with chronic lymphocytic leukemia (CLL), which often demonstrate complement resistance due to low CD20 and high membrane complement regulatory protein (mCRP) expression.16-18 In polyclonal antibody responses, antibodies against distinct epitopes or antigens are thought to cooperate resulting in increased effector functions against target cells. This increase can be mimicked in mAb combinations or cocktails. For example, mAbs targeting epidermal growth factor receptor (EGFR) do not induce CDC in vitro, but combinations of mAbs against multiple EGFR epitopes induced potent CDC.19,20 CD37, which is abundantly expressed on B cells, represents a promising therapeutic target for the treatment of Bcell malignancies.21,22 Currently known CD37 mAbs in clinical development, however, are generally poor inducers of CDC.23-27 Here we show that introducing Hx mutations into CD37 mAbs strongly potentiated CDC of CLL cells, and that combinations of CD20 and CD37 targeting mAbs could further enhance CDC of tumor cell lines and primary patient cells. We investigated the mechanism behind the synergistic CDC activity of CD20 and CD37 mAbs, and found that the mAb combinations activate complement cooperatively. The two mAbs formed mixed hexameric antibody complexes consisting of both antibodies each bound to their cognate targets, which we termed heterohexamers. The concept of hetero-hexamer formation and the use of Fc-Fc interaction enhancing mutations could serve as a tool to enhance cooperativity, and thereby the tumor killing capacity, of mAb combinations.

Methods

and stored using protocols approved by the institutional review boards in accordance with the Declaration of Helsinki (Online Supplementary Methods).

Antibodies and reagents mAb IgG1-CD20-7D8, IgG1-CD20-11B8, IgG1-CD37 clone 37.3 and IgG1-gp120 were recombinantly produced at Genmab.18,28-30 The HIV-1 gp120 mAb b12 was used to determine assay background signal. Mutations to enhance or inhibit Fc-Fc interactions were introduced in expression vectors encoding the antibody heavy chain by gene synthesis (GeneArt). Rituximab (MabThera®), ofatumumab (Arzerra®), and obinutuzumab (Gazyvaro®) were obtained from the institutional pharmacy (UMC Utrecht). See Online Supplementary Methods for details on reagents used.

CDC assays CDC assays with CLL patient cells were performed with human complement as described.31 CDC assays with B-lymphoma cell lines and patient-derived B-lymphoma cells were per-formed using 100,000 target cells incubated [45 minutes (min) at 37˚C] with a mAb concentration series and pooled normal human serum (NHS, 20% final concentration) as a complement source. Killing was calculated as the percentage of propidium idodide (PI) or 7-AAD positive cells determined by flow cytometry. See Online Supplementary Methods for details on cell markers used to define cell populations.

Expression analysis Expression levels of cellular markers were determined using an indirect immunofluorescence assay (QIFIKIT®, BioCytex) according to the manufacturer‘s instructions (Online Supplementary Methods).

C1q binding and CDC efficacy

Daudi cells (3x106 cells/mL) were incubated with 10 μg/mL mAb and a concentration series of purified human C1q for 45 min at 37˚C. After washing, cells were incubated with FITC-labeled rabbit anti-human C1q antibody for 30 min at 4˚C and analyzed on a FACS Canto II flow cytometer (BD Biosciences, CA, USA). The efficiency of C1q binding and subsequent CDC was assessed as described above using fixed mAb concentrations, a concentration series of purified C1q and 20% C1q depleted serum.

Confocal microscopy Raji cells were opsonized with A488 labeled Hx-CD20-7D8 and A594 labeled Hx-CD37 mAbs (2.5 µg/mL final concentrations), and incubated for 15 min at room temperature. After washing, cells were placed on a poly-D lysine-coated slide and images were captured with a Zeiss AxioObserver LSM 700 microscope using plan-Apochromat 63X/1.40 Oil DIC M27 objective lenses and acquired/processed using Zen software.

Förster resonance energy transfer analysis Proximity-induced Förster resonance energy transfer (FRET) analysis was determined by measuring energy transfer between cells incubated with A555-conjugated donor and A647-conjugated acceptor mAbs using flow cytometry (Online Supplementary Methods). The dynamic range of FRET analysis by flow cytometry was determined using control mAbs (Online Supplementary Figure S1).

Cells Daudi, Raji and WIL2-S B-lymphoma cell lines were obtained from the American Type Culture Collection (ATCC n. CCL-213, CCL-86 and CRL-8885, respectively). All primary patient cells used in this study were obtained after written and informed consent 1842

Data processing and statistical analyses All values are expressed as the mean±Standard Deviation of at least two independent experiments. Graphs were generated and analyzed using GraphPad Prism 7.0 (CA, USA). Differences haematologica | 2019; 104(9)

CD20 and CD37 antibodies show synergy in CDC

between two groups were analyzed using paired Student t-test with two-tailed 95% Confidence Intervals and between more groups by paired or unpaired one-way ANOVA followed by a Tukey’s post-hoc multiple comparisons test. See Online Supplementary Methods for details on synergy and colocalization analysis.

Results Hexamerization-enhancing mutations in CD20 and CD37 mAbs substantially enhance complement-dependent cytotoxicity of chronic lymphocytic leukemia B cells We previously reported increased CDC with engineered mAbs containing Hx mutations in the Fc domain.15,16 We therefore investigated whether introducing the Hx mutation E430G into the CD37 chimeric IgG1 mAb 37.3 could potentiate CDC in B cells isolated from chronic lymphocytic leukemia (CLL) patients and compared this to the CD20 mAb IgG1-CD20-7D8 with and without a Hx mutation. Wild-type (WT) IgG1-CD20-7D8 promoted considerable CDC of CLL B cells and CDC was increased by the E430G mutation (Figure 1A). While WT IgG1-CD37 efficiently binds to CLL B cells, it was ineffective at inducing CDC (Figure 1B and Online Supplementary Figure S2), in contrast to Hx-CD37 (Figure 1B). For both Hx-CD20-7D8 and Hx-CD37, high levels of cell killing largely required active complement, since CDC was almost absent in heat-inactivated NHS, NHS supplemented with EDTA or medium alone (Figure 1A and B). Background killing of cells from patient A mediated by Hx-CD37 in the absence of complement, was slightly higher than expected. However, in C1q-depleted serum, background killing was 16%, compared to 6% for cells reacted with Hx-CD20-7D8. Background killing in C1qdepleted serum for six other CLL patient samples averaged 13% and 14% for cells reacted with Hx-CD20-7D8 and Hx-CD37, respectively. Reaction in NHS increased CDC to averages of 91% and 95%, respectively. Introduction of the Hx mutation E430G into CD20 and CD37 mAbs did not affect pharmacokinetic profiles and binding to FcRn (data not shown).16 At the highest concentration (16 μg/mL) Hx-CD37 induced ≥95% CDC of tumor B cells for 9 of 12 patients (Figure 1C). At concentrations of 0.25 and 2 μg/mL, Hx-CD37 generally demonstrated higher potency than Hx-CD20-7D8 (Figure 1D and E), which may be explained by higher expression levels of CD37 (approximately 2-fold) in the majority of CLL samples (Online Supplementary Figure S3A and B).

IgG1-CD20-7D8 and WT IgG1-CD37 did not demonstrate enhanced CDC. However, while neither WT IgG1CD20-11B8 nor WT IgG1-CD37 induced CDC as single agents, the combination promoted strong lysis of approximately 60% (Figure 2B). Minimal cell lysis was observed in experiments with heat-inactivated serum, indicating that that cell killing was largely dependent on complement (Online Supplementary Figure S4). We also examined whether combinations of CD20 and CD37 mAbs with Hx mutations also showed cooperativity in CDC by testing mAb combinations using a full dose-response matrix (8x8 serial dilution grid) based on the EC50 values of the different mAbs. Surprisingly, both Hx-CD20-7D8 and Hx-CD20-11B8 in combination with Hx-CD37 showed enhanced CDC of Daudi cells compared to the single agents (Figure 2C and Online Supplementary Figure S5A). We next assessed whether the observed combination effect was synergistic using the Loewe additivity-based combination index (CI) score calculated by CompuSyn, whereby effects were categorized as synergistic (CI<1), additive (CI=1), or antagonistic (CI>1).33 The Loewe additivity-based model assumes synergy when the effect of a drug combination is higher than the effect of a drug combined with itself, and takes into account both the potency and the shape of the dose-effect curve of each drug in the dose-response matrix. Synergy was observed for both Hx-CD20-7D8 and Hx-CD20-11B8 when combined with Hx-CD37, with average CI values of 0.37 and 0.31 (effective dose - ED95), respectively (Figure 2D, Online Supplementary Table S1 and Online Supplementary Figure S5B). At the lower tested mAb concentrations, synergy was more profound (lower CI values) for combinations of Hx-CD37 with type II CD20 mAbderived Hx-CD20-11B8 than with type I CD20 mAbderived Hx-CD20-7D8. In addition to Daudi cells, we used two other B-cell lines expressing various levels of CD20 and CD37 to further examine the cooperativity in CDC between combinations of Hx-CD37 with Hx-CD20 mAbs or with clinically validated CD20 mAbs. Across all B-cell lines tested, enhanced CDC activity was observed for combinations of Hx-CD37 with Hx-CD20 mAbs, as well as for combinations of HxCD37 with ofatumumab, rituximab and obinutuzumab (Figure 2E). Even in WIL2-S cells expressing low levels of CD37, a combination of Hx-CD37 with obinutuzumab induced 72% lysis, whereas the single agents induced only 5% and 10% lysis respectively. Despite high single agent activity of Hx-CD37 and Hx-CD20 mAbs at 10 μg/mL (per mAb) in Daudi cells, the cooperativity between HxCD37 and Hx-CD20 mAbs became apparent at the lower mAb concentrations.

CD20 and CD37 mAbs synergistically induce complement-dependent cytotoxicity of malignant B cells

Enhanced binding and use of C1q by combinations of hexamerization-enhanced CD20 and CD37 mAbs

We investigated the CDC activity of combinations of WT CD20 and WT CD37 mAbs using two different CD20 mAbs. The ability to activate complement represents a key distinction between type I CD20 mAbs, which mediate strong CDC, and type II CD20 mAbs, which only mediate weak CDC.32 The effect of combining WT type I CD20 mAb 7D8 or WT type II CD20 mAb 11B8 with WT CD37 mAbs on CDC was assessed using Daudi cells. As expected, WT IgG1-CD20-7D8 showed potent CDC activity (96.6% cell lysis), whereas WT IgG1-CD37 did not induce CDC (Figure 2A). The combination of WT

We hypothesized that the observed synergy in CDC between Hx-CD20 and Hx-CD37 mAbs resulted from more efficient use of complement proteins, starting with binding of C1q. Therefore, we determined whether combinations of Hx-CD20 and Hx-CD37 differed in their C1q binding capacity. We incubated Daudi cells with fixed mAb concentrations and titrated C1q, and measured C1q binding and the concentration of C1q required to induce CDC, referred to here as CDC efficacy. Hx-CD20-7D8 already induced efficient C1q binding as a single agent, while Hx-CD37 showed limited C1q binding (Figure 3A).

haematologica | 2019; 104(9)

1843

S.C. Oostindie et al.

Combinations of Hx-CD20-7D8 and Hx-CD37 did not significantly increase C1q binding. However, the combination showed higher CDC efficacy as demonstrated by the lower EC50 value in the C1q dose-response curves compared to the single mAbs (0.03 μg/mL for the combination vs. 0.12 μg/mL for Hx-CD20-7D8 and 0.34 μg/mL for Hx-CD20-11B8) (Figure 3B). In contrast to the results with the type I CD20 mAb-derived variant, combining type II CD20 mAb-derived Hx-CD20-11B8 with HxCD37 resulted in increased C1q binding compared to the single mAbs, as well as increased CDC efficacy (Figure 3C and D). Collectively, these data suggest that combinations of both type I and type II CD20 mAb-derived Hx-CD20 mAb with Hx-CD37 mAbs activate complement more effectively.

Confocal microscopy was used to determine whether the CD20- and CD37-specific antibodies associate on the cell surface upon target binding. Cell-bound Hx variants of CD20 mAb 7D8 and CD37 mAb 37.3 were detected using A488 and A594 fluorescent labeling, respectively, and antibody colocalization was quantified by calculating spatial overlap (Manders’ co-efficients) between the two fluorescent labels. The merged A488/A594 image showed that membrane-bound Hx-CD20 and Hx-CD37 mAbs indeed colocalized on the surface of Raji cells (Figure 4A), which was confirmed by quantitative analysis, giving Manders' coefficients of M1=0.805 (fraction of image 1 overlapping image 2) and M2=0.751 (fraction of image 2 overlapping image 1).

1844

CD20 and CD37 mAbs colocalize on B cells

Figure 1. Hexamerization-enhancing mutations in CD20 and CD37 mAbs substantially enhance complement-dependent cytotoxicity (CDC) of chronic lymphocytic leukemia (CLL) B cells. (A and B) CDC of B cells obtained from patient A with CLL. Cells were opsonized with different concentrations of CD20 mAb 7D8 as wild type (IgG1-CD20-7D8) or with a hexamerization-enhancing mutation (Hx-CD20-7D8) (A); or CD37 mAb 37.3 as wild type (IgG1-CD37) or with a hexamerization-enhancing mutation (Hx-CD37) (B) in the presence of 50% pooled normal human serum (NHS), heat-inactivated (HI) NHS, NHS + EDTA or medium. Representative examples of three replicate experiments are shown. (C-E) CDC of B cells obtained from 12 different CLL patients (patient B-M). CLL B cells were opsonized with 16 μg/mL (C), 2 μg/mL (D) or 0.25 μg/mL (E) HxCD20-7D8 or Hx-CD37. The dashed line represents 95% cell lysis. CDC induction is expressed as the percentage lysis determined by the fraction of TO-PRO3 positive cells and data shown are mean and Standard Deviation of duplicate measurements.

haematologica | 2019; 104(9)

CD20 and CD37 antibodies show synergy in CDC

Figure 2. CD20 and CD37 mAbs synergistically induce complement-dependent cytotoxicity (CDC) of malignant B cells. (A and B) CDC on Daudi cells opsonized with 30 μg/mL wild-type (WT) type I CD20 mAb 7D8 (IgG1-CD20-7D8) (A) or type II CD20 mAb 11B8 (IgG1-CD20-11B8) (B), CD37 mAb 37.3 (IgG1-CD37), or a combination thereof (15 + 15 μg/mL) in the presence of 20% NHS. CDC induction is expressed as the percentage lysis determined by the fraction of propidium iodide (PI)-positive cells. Data shown are mean and Standard Deviation of triplicate measurements. (C) 8 x 8 CDC dose response matrix plot for the combination of hexamerizationenhanced CD37 mAb Hx-CD37 (0-0.8 μg/mL) with hexamerization-enhanced CD20 mAb Hx-CD20-11B8 (0-8 μg/mL), tested on Daudi cells and categorized as a color gradient from green (0% lysis) to yellow (50% lysis) to red (100% lysis). HIV gp120-specific mAb b12 (IgG1-gp120) was used as a negative control human mAb. (D) Loewe additivity-based combination index (CI) values calculated by CompuSyn for the CDC dose response matrix as described in (C) and categorized as synergistic (<1, red), additive (1, white) and antagonistic (>1, blue). Representative examples of two replicate experiments are shown. (E) CDC and CD37 expression analysis on Daudi, Raji and WIL2-S cells. For the CDC assay, cells were opsonized with Hx-CD37 (10 μg/mL), different CD20 mAb variants (10 μg/mL) or combinations thereof (10 + 10 μg/mL). Data show the mean of nine replicates collected from three independent experiments. Expression levels were determined using QIFIKIT analysis. The number of antibody molecules per cell was calculated from the antibody-binding capacity (mean fluorescence intensity, MFI) normalized to a calibration curve, according to the manufacturer’s guidelines. Expression data show the mean of four replicates collected from two independent experiments. ****P<0.0001.

haematologica | 2019; 104(9)

1845

S.C. Oostindie et al.

Colocalization of cell-bound CD20 and CD37 mAbs was further examined by directly assessing molecular proximity using fluorescence resonance energy transfer (FRET) analysis. We examined FRET on Daudi cells between WT and Hx variants of CD20 and CD37 mAbs alone and in combination. Consistent with its CDC activity (Figure 2A), WT IgG1-CD20-7D8 induced high FRET, which suggests antibody hexamer formation (Figure 4B). WT IgG1-CD2011B8 did not demonstrate proximity-induced FRET (Figure 4C), and WT IgG1-CD37 induced approximately 15% FRET (Figure 4B and C). Introducing a Hx mutation resulted in increased FRET levels for each of the single agents, indicating that enhancing Fc-Fc interactions increases mAb colocalization at the cell surface (P<0.0001) (Figure 4D and E). Introduction of the Hx mutation did not affect target binding (data not shown), thereby excluding the possibility that increased FRET would be due to more mAb being available on the cell surface. Combinations of WT IgG1CD20-7D8 and WT IgG1-CD37 induced approximately 30% FRET, which was increased compared to the WT IgG1-CD37 single mAb (P<0.0001, Figure 4B). Combinations of WT IgG1-CD20-11B8 and WT IgG1CD37 substantially increased FRET compared to each single mAb (P<0.0001) (Figure 4C), consistent with the enhanced CDC induction (Figure 2B). Combinations of Hx-CD20-7D8 or Hx-CD20-11B8 with Hx-CD37 further enhanced FRET compared to the FRET levels induced by the WT mAb combinations (P<0.0001) (Figure 4D and E). These results confirm that CD20 and CD37 IgG1 mAbs bind in close proximity on the cell membrane, which can be enhanced by introducing the E430G mutation.

Hexamerization-enhanced CD20 and CD37 mAbs cooperate in complement-dependent cytotoxicity through Fc-mediated clustering in hetero-hexamers Both enhancing Fc-Fc interactions in the CD20 or CD37 mAbs and combining the two B-cell target mAbs resulted in enhanced mAb colocalization. Together with the dependency of CDC on the formation of hexameric IgG complexes on the cell surface,15 this suggests that the CD20 and CD37 mAbs might not only form hexamers composed of mAbs bound to identical surface targets, but may cooperate by also forming mixed hexameric complexes of mAbs bound to either target, referred to here as heterohexamers. The contribution of Fc-Fc interactions between Hx-CD20-11B8 and Hx-CD37 to the CDC activity of the mAb combination was examined by the introduction of the complementary Fc-Fc interface mutations K439E and S440K. K439E and S440K suppress Fc-Fc interactions between antibody molecules containing the same mutation, whereas Fc-Fc interactions are restored in K439K and S440K antibody mixtures.15 The capacity of Hx-CD2011B8 and Hx-CD37 variants with K439E and S440K mutations to induce CDC was tested using Daudi and WIL2-S cells. The CDC activity of Hx-CD20-11B8 was completely inhibited by introducing either the K439E or S440K Fc-Fc inhibiting mutation using Daudi and WIL2-S cells (Figure 5A and B). CDC activity was restored when Fc-Fc inhibition was neutralized by mixing the two CD20 mAbs. Similar results were observed for Hx-CD37 on Daudi cells, while on WIL2-S cells, Hx-CD37 did not induce CDC, most likely due to low CD37 expression (Figure 5C and D). Combining Hx-CD20-11B8 and Hx-CD37 mAbs harbor-

Figure 3. Enhanced binding and use of C1q by combinations of hexamerization-enhanced CD20 and CD37 mAbs. The capacity to bind C1q (A, C) and the efficiency to bind C1q and promote Complement-dependent cytotoxicity (CDC) (B and D) was assessed using Daudi cells opsonized with 10 Îźg/mL of hexamerization-enhanced variants of type I CD20 mAb-derived Hx-CD20-7D8 (A-B) or type II CD20 mAb-derived Hx-CD20-11B8 (C and D), CD37 mAb 37.3-derived Hx-CD37, or a combination thereof (5 + 5 Îźg/mL). Binding was detected using a FITC-labeled rabbit anti-human C1q secondary antibody and is expressed as mean fluorescence intensity. CDC induction was assessed in C1q-depleted serum by calculating the percentage of propidium idodide (PI)-positive cells as determined by flow cytometry. Representative examples of three replicate experiments are shown.

1846

haematologica | 2019; 104(9)

CD20 and CD37 antibodies show synergy in CDC

ing the same Fc-Fc inhibiting mutation (K439E or S440K) strongly reduced CDC activity on Daudi and WIL2-S cells (Figure 5E and F). However, CDC of both cell lines was restored by mixing Hx-CD20-11B8 and Hx-CD37, each

carrying one of the complementary mutations K439E or S440K. These data suggest that Hx-CD20-11B8 and HxCD37 can indeed form hetero-hexameric complexes, thereby cooperating to activate complement.

Figure 4. CD20 and CD37 mAbs colocalize on B cells. (A) Confocal fluorescence microscopy analysis to detect colocalization of cell-bound CD20 and CD37 mAbs. Raji cells were opsonized with hexamerization-enhanced A488-conjugated CD20 mAb 7D8-derived Hx-CD20-7D8 (image 1, green) and hexamerization-enhanced A594-conjugated CD37 mAb 37.3-derived Hx-CD37 (image 2, red), and incubated for 15 minutes (min) at room temperature. Images were captured in PBS imaging medium at ambient temperature using a Zeiss Axi-oObserver LSM 700 microscope with Plan-Apochromat 63X/1.40 Oil DIC M27 objective lenses and acquired/processed using Zen software. Two excitation lasers were used at 488 and 555 nm. In the merged image, overlap of red and green produces orange or yellow. A representative example of two replicate experiments is shown. (B and C) FRET analysis to detect the molecular proximity of (B) WT type I CD20 mAb 7D8 (IgG1-CD20-7D8) or (C) WT type II CD20 mAb 11B8 (IgG1-CD20-11B8), WT CD37 mAb 37.3 (IgG1-CD37) or a combination thereof on the cell membrane of Daudi cells. (D and E) FRET analysis to detect the molecular proximity of hexamerization enhanced variants of (D) type I CD20 mAb 7D8-derived Hx-CD20-7D8 or (E) type II CD20 mAb 11B8-derived Hx-CD20-11B8, CD37 mAb 37.3-derived Hx-CD37 or a combination thereof on the cell membrane of Daudi cells. Daudi cells were opsonized with 10 μg/mL A555-conjugated- and 10 μg/mL A647-conjugated antibody variants for 15 min at 37˚C. FRET was calculated from mean fluorescence intensity values as determined by flow cytometry. Data shown are mean and Standard Deviation of six replicates collected from three independent experiments. ****P<0.0001.

haematologica | 2019; 104(9)

1847

S.C. Oostindie et al. A

Figure 5. Hexamerization-enhanced CD20 and CD37 mAb cooperate in complement-dependent cytotoxicity (CDC) through Fc-mediated clustering in hetero-hexamers. The effect of introducing Fc-Fc inhibiting mutations S440K and K439E on the CDC induction of hexamerization-enhanced type II CD20 mAb 11B8-derived HxCD20-11B8 on Daudi cells (A) and WIL2-S cells (B), hexamerization-enhanced CD37 mAb 37.3-derived Hx-CD37 on Daudi (C) and WIL2-S cells (D) and the mAb combinations thereof on Daudi (E) and WIL2-S cells (F). Cells were opsonized with concentration series of Hx-CD20-11B8 and Hx-CD37 variants in the presence of 20% NHS. CDC induction is expressed as the percentage lysis determined by the fraction of propidium iodide (PI)-positive cells. Representative examples of two (WIL-2S) and three replicates (Daudi) are shown. (G) The effect of introducing Fc-Fc inhibiting mutations S440K and K439E on the molecular proximity of Hx-CD20-11B8 and Hx-CD37 variants on the cell membrane of Daudi cells. Daudi cells were incubated with 10 μg/mL A555-conjugated Hx-CD20-11B8 variants and 10 μg/mL A647conjugated Hx-CD37 variants for 15 minutes at 37˚C. FRET was calculated from the mean fluorescence intensity values as determined by flow cytometry. Data shown are mean and Standard Deviation of six replicates collected from three independent experiments. ****P<0.0001.

1848

haematologica | 2019; 104(9)

CD20 and CD37 antibodies show synergy in CDC

Next, the effect of the Fc-Fc interaction-inhibiting mutations on colocalization of Hx-CD20-11B8 and Hx-CD37 mAbs on the cell membrane of Daudi cells was evaluated using FRET analysis. mAb combinations with Hx-CD2011B8 and Hx-CD37 variants, both harboring the same FcFc inhibiting mutation (K439E or S440K) showed reduced FRET on Daudi cells (Figure 5G and Online Supplementary Figure S6). FRET levels were restored when Fc-Fc inhibition was neutralized by combining Hx-CD20-11B8 and Hx-CD37 mAbs, each having one of the complementary mutations K439E or S440K. Thus, donor- and acceptorlabeled Hx-CD20-11B8 and Hx-CD37 mAb variants come together in close proximity on the cell membrane of Daudi cells, which appears to be, at least in part, mediated by the Fc domain.

Combinations of hexamerization-enhanced CD20 and CD37 mAbs induce superior ex vivo complement-dependent cytotoxicity of tumor cells obtained from patients with B-cell malignancies We next examined the capacity of combinations of HxCD20 and Hx-CD37 mAbs to induce CDC ex vivo in tumor cells obtained from patients with B-cell malignancies. First, the CDC activity of Hx-CD20-7D8, Hx-CD37 or combinations was evaluated using tumor cells obtained from 15 patients diagnosed with CLL. Both Hx-CD207D8 and Hx-CD37 induced substantial CDC of CLL tumor cells from all 15 tested donors (Figure 6A), in accordance with that seen in Figure 1. Hx-CD37 was more effective in CDC than Hx-CD20-7D8, which may be explained by higher expression CD37 on CLL cells

Figure 6. Combinations of hexamerization-enhanced CD20 and CD37 monoclonal antibodies (mAbs) induce superior ex vivo complement-dependent cytotoxicity (CDC) of tumor cells obtained from patients with B-cell malignancies. (A) B cells obtained from 15 patients diagnosed with chronic lymphocytic leukemia (CLL) were opsonized with fixed concentrations of hexamerization-enhanced type I CD20 mAb 7D8-derived Hx-CD20-7D8 or hexamerization-enhanced CD37 mAb 37.3-derived Hx-CD37 (open symbols: 2.5 μg/mL, closed symbols: 2 μg/mL; each presented as 100%), or 1:1 mixtures thereof (open symbols: 0.625 μg/mL of each mAb, closed symbols: 0.5 μg/mL of each mAb; each represented as 50%) in the presence of 50% NHS. CDC induction is presented as the percentage lysis determined by the fraction of TO-PRO-3 positive cells. (B) B cells of a representative CLL patient sample (patient G) were opsonized with different total mAb concentrations of Hx-CD207D8 or Hx-CD37 (single agents indicated as 100%) and combinations thereof at different antibody ratios (indicated as 75%:25%, 50%:50% and 25%:75%) in the presence of 50% NHS. CDC induction is presented as the percentage lysis determined by the fraction of TO-PRO-3 positive cells. Data shown are mean and Standard Deviation of duplicate measurements. (C) B cells obtained from ten patients diagnosed with different B-cell malignancies [B-cell non-Hodgkin lymphoma (B-NHL) not otherwise specified (NOS), follicular lymphoma (FL), marginal zone lymphoma (MZL) and mantle cell lymphoma (MCL)] were opsonized with 10 μg/mL of hexamerization-enhanced type II CD20 mAb 11B8-derived Hx-CD20-11B8 or Hx-CD37, and the combination thereof (5 + 5 μg/mL) in the presence of 20% NHS. CDC induction is presented as the percentage lysis determined by the fraction of 7-AAD positive B-lymphoma cells. (D) CDC assay with B-cell patient samples representative for BNHL (NOS), FL, MZL and MCL as described in (C). Data shown are mean and Standard Deviation of duplicate measurements. *P<0.05; **P<0.01; ****P<0.0001.

haematologica | 2019; 104(9)

1849

S.C. Oostindie et al.

(P<0.05) (Online Supplementary Figure S3A and B). Importantly, significantly increased CDC levels were observed in 9 of 15 tested CLL donors upon treatment with the combination of Hx-CD20-7D8 and Hx-CD37. Even at modest total concentrations of Hx-CD20-7D8 and Hx-CD37 (â&#x2030;¤1.25 Îźg/mL for each mAb), >90% CDC of B cells was induced in 12 of the 15 tested CLL donors (Figure 6A). Enhanced CDC by the mAb combination was observed over a range of mAb concentrations and at different mAb ratios, and was more apparent at lower mAb concentrations, as illustrated for one representative donor (Figure 6B). Next, we evaluated the cytotoxic capacity of Hx-CD20-11B8, Hx-CD37, and the combination thereof using tumor cells of ten patients diagnosed with different B-cell lymphomas, including B-cell non-Hodgkin lymphoma (B-NHL) not otherwise specified (NOS), follicular lymphoma (FL), marginal zone lymphomas (MZL) and mantle cell lymphoma (MCL). While for the single agents a large variation in CDC efficacy was observed between the donors, the combination of Hx-CD20-11B8 and Hx-CD37 consistently showed enhanced CDC activity

compared to the individual mAbs (Figure 6C). Representative figures from each tested lymphoma subtype show that combinations of Hx-CD20-11B8 and Hx-CD37 at the tested 1:1 ratio may enhance CDC, even when CDC induced by the individual mAbs was low or absent (Figure 6D). Furthermore, analysis of CD20 and CD37 target expression levels on primary B cells from 24 CLL patients and ten patients with different NHL subtypes illustrated a large diversity in target expression levels and ratios (Online Supplementary Figure S3A-C). These results suggest that combinations of Hx-CD20 and HxCD37 mAbs may generally increase the therapeutic potential of CDC-inducing mAbs in B-cell malignancies across different target expression levels and ratios.

Discussion Improving therapeutic efficacy against (heterogeneous) tumors has been the focus of intense preclinical and clinical development. We previously showed that the therapeutic

Figure 7. Model for Fc-mediated clustering of CD20 and CD37 monoclonal antibodies (MAbs) in hetero-hexamers upon binding to the cell surface. (A) mAbs naturally cluster into hexameric complexes upon antibody binding to a cognate antigen on a target cell, thereby providing a docking site for C1q binding and complementdependent cytotoxicity (CDC) induction. (B) Upon binding of mAbs targeting two different coexpressed antigens on the plasma membrane that (are able to) colocalize, hetero-hexameric antibody complexes are formed consisting of both mAbs, providing a docking site for C1q binding and CDC induction. Introducing hexamerizationenhancing mutations can increase Fc-mediated clustering of mAbs, both into homo- and hetero-hexameric antibody complexes on the cell surface, thereby increasing the number of C1q docking sites and further potentiating CDC.

1850

haematologica | 2019; 104(9)

CD20 and CD37 antibodies show synergy in CDC

potential of IgG1 mAbs can be enhanced by introducing mutations that improve hexamerization by Fc-mediated clustering and thereby increase CDC activity.15,16 In the present study, we introduced a Hx mutation, E430G, into CD20 and CD37 mAbs and observed an impressive increase in CDC activity in primary CLL samples. Moreover, combinations of CD20 and CD37 mAbs showed enhanced and synergistic CDC activity, including combinations of Hx-CD37 mAbs with the approved mAbs rituximab, ofatumumab and obinutuzumab. With several CD20 and CD37 mAbs approved or in clinical development, it is attractive to study the mechanism behind the cooperativity between mAbs targeting these two antigens.34,35 It was recently reported that expression levels of CD20 and CD37 mRNA and protein are correlated on lymphoma B cells.36 Here, using confocal microscopy and FRET analysis we show that CD20 and CD37 mAbs colocalize on surfaces of B cells and that enhancing Fc-Fc interactions increases mAb colocalization. The observed synergistic CDC activity of CD20 and CD37 mAbs was supported by increased C1q binding and increased CDC efficacy, as illustrated by enhanced CDC at relatively low C1q concentrations. Synergy in complement activation was most evident for CD37 mAbs in combination with type II CD20 mAbs, than with type I CD20 mAbs which are already effective at clustering as WT mAbs. De Winde et al.37 recently suggested that the organization of the B-cell plasma membrane is shaped by dynamic protein-protein interactions and that this organization might be altered by targeted mAb therapies. It has previously been described that membrane proteins can cluster into lipid rafts or tetraspanin-enriched microdomains (TEM), enabling efficient signal transduction.38,39 We hypothesized that the synergistic interactions in CDC between CD20 and CD37 mAbs could be driven by clustering of both target-bound mAbs into oligomeric complexes. By introducing Fc-Fc inhibiting mutations, we were able to demonstrate that CD20 and CD37 mAbs do not only permit the formation of homo-hexamers consisting of mAbs bound to either single target separately, but also allow the formation of hetero-hexamers composed of alternating CD20 and CD37 mAbs, each bound to their own cognate target, explaining the synergistic effects. Other hetero-hexamer variants may occur, although the presence of such alternative variants remains to be demonstrated (see Figure 7). We therefore propose a model for Fc-mediated clustering of synergistic mAb combinations on malignant B cells (Figure 7). Upon binding of mAbs targeting two different coexpressed antigens on the plasma membrane that are able to colocalize, hetero-hexameric complexes are formed, provid-

References 1. Deans JP, Li H, Polyak MJ. CD20-mediated apoptosis: signalling through lipid rafts. Immunology. 2002;107(2):176-182. 2. Gopal AK, Press OW. Clinical applications of anti-CD20 antibodies. Transl Res. 1999; 134(5):445-450. 3. Leget GA, Czuczman MS. Use of rituximAb, the new FDA-approved antibody. Curr Opin Oncol. 1998;10(6):548-551.

haematologica | 2019; 104(9)

ing a docking site for C1q binding and CDC induction. Introducing hexamerization-enhancing mutations can increase Fc-mediated clustering of mAbs into both homoand hetero-hexameric complexes, thereby increasing the number of hexamers and further potentiating CDC. Increasing the therapeutic potency of mAb combinations, driven by hetero-hexamerization, could be of clinical relevance, as illustrated by a combination of Hx-CD20 and Hx-CD37 mAbs that showed strong CDC of tumor B cells obtained from patients with different B-cell malignancies. Hicks et al.40 recently reported that the antitumor activity of IMGN529, a CD37-targeting antibody-drug conjugate in clinical development, was potentiated in combination with rituximab in vivo in B-NHL xenograft models, which was associated with increased CD37 internalization rates. Other mechanisms of synergy between CD20 and CD37 have also been reported, such as upregulation of CD20 expression in Daudi cells after treatment with the radiolabeled anti-CD37 mAb 177Lu‐lilotomab.41 The concept of antibody hetero-hexamer formation may hold relevance for a broader range of targets and effector mechanisms. An emerging therapeutic approach is the development of designer polyclonals, consisting of multiple mAbs in one product of which several are in clinical development, such as MM-151, targeting three epitopes on EGFR and Sym013, a mixture of six mAbs targeting all three HER family members (EGFR, HER2 and HER3).42,43 One could speculate that enhancing Fc-mediated antibody clustering involving different coexpressing targets on hematologic or solid tumors may induce synergistic efficacy, providing a rationale for application in designer polyclonals. Whether effector mechanisms other than CDC, such as ADCC or ADCP are also enhanced by combinations of CD20 and CD37 mAbs remains to be elucidated. One may speculate that the engagement of two mAbs binding coexpressed targets allows for higher total antibody binding on the cell surface, allowing more efficient engagement of FcγRs on effector cells. In the present work, we have demonstrated that synergy in CDC induced by combinations of CD20 and CD37 mAbs is likely driven by Fc-mediated clustering into hetero-hexameric antibody complexes on the cell surface. Enhancing hetero-hexamerization between mAb combinations using Fc engineering represents a powerful tool to increase the therapeutic efficacy of mAb combinations directed against hematologic and other tumor targets. Acknowledgments The authors would like to thank Joost Bakker (SCICOMVISUALS) for designing Figure 7.

4. Beurskens FJ, Lindorfer MA, Farooqui M, et al. Exhaustion of cytotoxic effector systems may limit mAb-based immunotherapy in cancer patients. J Immunol. 2012; 188(7):3532-3541. 5. Taylor RP, Lindorfer MA. Analyses of CD20 Monoclonal Antibody–Mediated Tumor Cell Killing Mechanisms: Rational Design of Dosing Strategies. Mol Pharmacol. 2014;86(5):485-491. 6. Weiner GJ. Building better monoclonal antibody-based therapeutics. Nat Rev Cancer. 2015;15(6):361-370.

7. Wang X, Mathieu M, Brezski RJ. IgG Fc engineering to modulate antibody effector functions. Protein Cell. 2018;9(1):63-73. 8. Lazar GA, Dang W, Karki S, et al. Engineered antibody Fc variants with enhanced effector function. Proc Natl Acad Sci U S A. 2006;103(11):4005-4010. 9. Richards JO, Karki S, Lazar GA, Chen H, Dang W, Desjarlais JR. Optimization of antibody binding to Fc RIIa enhances macrophage phagocytosis of tumor cells. Mol Cancer Ther. 2008;7(8):2517-2527. 10. Shields RL, Namenuk AK, Hong K, et al.

1851

S.C. Oostindie et al.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

1852

High Resolution Mapping of the Binding Site on Human IgG1 for Fcgamma RI, Fcgamma RII, Fcgamma RIII, and FcRn and Design of IgG1 Variants with Improved Binding to the Fcgamma R. J Biol Chem. 2000;276(9):6591-6604. Stavenhagen JB, Gorlatov S, Tuaillon N, et al. Fc Optimization of Therapeutic Antibodies Enhances Their Ability to Kill Tumor Cells In vitro and Controls Tumor Expansion In vivo via Low-Affinity Activating Fc Receptors. Cancer Res. 2007;67(18):8882-8890. Idusogie EE, Wong PY, Presta LG, et al. Engineered antibodies with increased activity to recruit complement. J Immunol. 2001;166(4):2571-2575. Natsume A, In M, Takamura H, et al. Engineered Antibodies of IgG1/IgG3 Mixed Isotype with Enhanced Cytotoxic Activities. Cancer Res. 2008;68(10):3863-3872. Melis JPM, Strumane K, Ruuls SR, Beurskens FJ, Schuurman J, Parren PWHI. Complement in therapy and disease: Regulating the complement system with antibody-based therapeutics. Mol Immunol. 2015;67(2, Part A):117-130. Diebolder CA, Beurskens FJ, de Jong RN, et al. Complement Is Activated by IgG Hexamers Assembled at the Cell Surface. Science. 2014;343(6176):1260-1263. de Jong RN, Beurskens FJ, Verploegen S, et al. A Novel Platform for the Potentiation of Therapeutic Antibodies Based on AntigenDependent Formation of IgG Hexamers at the Cell Surface. PLoS Biol. 2016; 14(1):e1002344. Lindorfer MA, Cook EM, Tupitza JC, et al. Real-time analysis of the detailed sequence of cellular events in mAb-mediated complement-dependent cytotoxicity of B-cell lines and of chronic lymphocytic leukemia B-cells. Mol Immunol. 2016;70:13-23. Teeling JL, French RR, Cragg MS, et al. Characterization of new human CD20 monoclonal antibodies with potent cytolytic activity against non-Hodgkin lymphomas. Blood. 2004;104(6):1793-1800. Dechant M, Weisner W, Berger S, et al. Complement-Dependent Tumor Cell Lysis Triggered by Combinations of Epidermal Growth Factor Receptor Antibodies. Cancer Res. 2008;68(13):4998-5003. Klausz K, Berger S, Lammerts van Bueren JJ, et al. Complement mediated tumor specific cell lysis by antibody combinations targeting epidermal growth factor receptor (EGFR) and its variant III (EGFRvIII).

Cancer Sci. 2011;102(10):1761-1768. 21. Link MP, Bindl J, Meeker TC, et al. A unique antigen on mature B cells defined by a monoclonal antibody. J Immunol. 1986;137(9):3013-3018. 22. Schwartz-Albiez R, DĂśrken B, Hofmann W, Moldenhauer G. The B cell-associated CD37 antigen (gp40-52). Structure and subcellular expression of an extensively glycosylated glycoprotein. J Immunol. 1988; 140(3):905-914. 23. Deckert J, Park PU, Chicklas S, et al. A novel anti-CD37 antibody-drug conjugate with multiple anti-tumor mechanisms for the treatment of B-cell malignancies. Blood. 2013;122(20):3500-3510. 24. Heider K-H, Kiefer K, Zenz T, et al. A novel Fc-engineered monoclonal antibody to CD37 with enhanced ADCC and high proapoptotic activity for treatment of B-cell malignancies. Blood. 2011;118(15):41594168. 25. Pereira DS, Guevara CI, Jin L, et al. AGS67E, an Anti-CD37 Monomethyl Auristatin E Antibodyâ&#x20AC;&#x201C;Drug Conjugate as a Potential Therapeutic for B/T-Cell Malignancies and AML: A New Role for CD37 in AML. Mol Cancer Ther. 2015; 14(7):1650-1660. 26. Repetto-Llamazares AHV, Larsen RH, Patzke S, et al. Targeted Cancer Therapy with a Novel Anti-CD37 Beta-Particle Emitting Radioimmunoconjugate for Treatment of Non-Hodgkin Lymphoma. PLoS One. 2015;10(6):e0128816. 27. Zhao XB, Biswas S, Mone A, et al. Novel Anti-CD37 Small Modular Immunopharmaceutical (SMIP) Induces BCell-Specific, Caspase-Independent Apoptosis in Human CLL Cells. Blood. 2004;104(11):2515. 28. Burton DR, Pyati J, Koduri R, et al. Efficient neutralization of primary isolates of HIV-1 by a recombinant human monoclonal antibody. Science. 1994;266(5187):1024-1027. 29. Deckert. J. CD37-binding molecules and immunoconjugates thereof. WO 2011/112978A1.(2011). 30. Vink T, Oudshoorn-Dickmann M, Roza M, Reitsma J-J, de Jong RN. A simple, robust and highly efficient transient expression system for producing antibodies. Methods. 2014;65(1):5-10. 31. Cook EM, Lindorfer MA, van der Horst H, et al. Antibodies That Efficiently Form Hexamers upon Antigen Binding Can Induce Complement-Dependent Cytotoxicity under Complement-Limiting

32.

33.

34.

35.

36.

37.

38. 39. 40.

41.

42.

43.

Conditions. J Immunol. 2016;197(5):17621775. Cragg MS, Morgan SM, Chan HTC, et al. Complement-mediated lysis by anti-CD20 mAb correlates with segregation into lipid rafts. Blood. 2003;101(3):1045-1052. Chou T-C. Theoretical Basis, Experimental Design, and Computerized Simulation of Synergism and Antagonism in Drug Combination Studies. Pharmacol Rev. 2006;58(3):621-681. Beckwith KA, Byrd JC, Muthusamy N. Tetraspanins as therapeutic targets in hematological malignancy: a concise review. Front Physiol. 2015;6:91. Teo EC-y, Chew Y, Phipps C. A review of monoclonal antibody therapies in lymphoma. Crit Rev Oncol Hematol. 2016;97:72-84. Xu-Monette ZY, Li L, Byrd JC, et al. Assessment of CD37 B-cell antigen and cell of origin significantly improves risk prediction in diffuse large B-cell lymphoma. Blood. 2016;128(26):3083-3100. de Winde CM, Elfrink S, van Spriel AB. Novel Insights into Membrane Targeting of B Cell Lymphoma. Trends Cancer. 2017;3(6):442-453. Hemler ME. Tetraspanin functions and associated microdomains. Nat Rev Mol Cell Biol. 2005;6(10):801-811. Simons K, Sampaio JL. Membrane Organization and Lipid Rafts. Cold Spring Harb Perspect Biol. 2011;3(10):a004697. Hicks SW, Lai KC, Gavrilescu LC, et al. The Antitumor Activity of IMGN529, a CD37Targeting Antibody-Drug Conjugate, Is Potentiated by Rituximab in Non-Hodgkin Lymphoma Models. Neoplasia (New York, NY). 2017;19(9):661-671. Repetto-Llamazares AHV, Malenge MM, O'Shea A, et al. Combination of (177) Lulilotomab with rituximab significantly improves the therapeutic outcome in preclinical models of non-Hodgkin's lymphoma. Eur J Haematol. 2018;101(4):522-531. Dienstmann R, Patnaik A, GarciaCarbonero R, et al. Safety and Activity of the First-in-Class Sym004 Anti-EGFR Antibody Mixture in Patients with Refractory Colorectal Cancer. Cancer Discov. 2015;5(6):598-609. Jacobsen HJ, Poulsen TT, Dahlman A, et al. Pan-HER, an Antibody Mixture Simultaneously Targeting EGFR, HER2, and HER3, Effectively Overcomes Tumor Heterogeneity and Plasticity. Clin Cancer Res. 2015;21(18):4110-4122.

haematologica | 2019; 104(9)

ARTICLE

Platelet Biology & its Disorders

CD45 expression discriminates waves of embryonic megakaryocytes in the mouse

Ferrata Storti Foundation

Isabel Cortegano,1 Natalia Serrano,2 Carolina Ruiz,1 Mercedes Rodríguez,1 Carmen Prado,1 Mario Alía,1 Andrés Hidalgo,3 Eva Cano,4 Belén de Andrés1 and María-Luisa Gaspar1

Department of Immunology, Centro Nacional de Microbiología, Instituto de Salud Carlos III (ISCIII), Majadahonda; 2Centro de Biología Molecular Severo Ochoa, Consejo Superior de Investigaciones Científicas (CBMSO-CSIC), Madrid; 3Area of Cell and Developmental Biology, Centro Nacional de Investigaciones Cardiovasculares, Madrid and 4 Neuroinflamation Unit, Chronic Diseases Research Program, Instituto de Salud Carlos III (ISCIII), Majadahonda, Spain 1

Haematologica 2019 Volume 104(9):1853-1865

ABSTRACT

mbryonic megakaryopoiesis starts in the yolk sac on gestational day 7.5 as part of the primitive wave of hematopoiesis, and it continues in the fetal liver when this organ is colonized by hematopoietic progenitors between day 9.5 and 10.5, as the definitive hematopoiesis wave. We characterized the precise phenotype of embryo megakaryocytes in the liver at gestational day 11.5, identifying them as CD41++CD45CD9++CD61+MPL+CD42c+ tetraploid cells that express megakaryocyte-specific transcripts and display differential traits when compared to those present in the yolk sac at the same age. In contrast to megakaryocytes from adult bone marrow, embryo megakaryocytes are CD45- until day 13.5 of gestation, as are both the megakaryocyte progenitors and megakaryocyte/erythroid-committed progenitors. At gestational day 11.5, liver and yolk sac also contain CD41+CD45+ and CD41+CD45- cells. These populations, and that of CD41++CD45-CD42c+ cells, isolated from liver, differentiate in culture into CD41++CD45-CD42c+ proplatelet-bearing megakaryocytes. Also present at this time are CD41-CD45++CD11b+ cells, which produce low numbers of CD41++CD45-CD42c+ megakaryocytes in vitro, as do fetal liver cells expressing the macrophage-specific Csf receptor-1 (Csf1r/CD115) from MaFIA transgenic mice, which give rise poorly to CD41++CD45-CD42c+ embryo megakaryocytes both in vivo and in vitro. In contrast, around 30% of adult megakaryocytes (CD41++CD45++CD9++CD42c+) from C57BL/6 and MaFIA mice express CD115. We propose that differential pathways operating in the mouse embryo liver at gestational day 11.5 beget CD41++CD45-CD42c+ embryo megakaryocytes that can be produced from CD41+CD45- or from CD41+CD45+ cells, at difference from those from bone marrow.

Correspondence: ISABEL CORTEGANO icortegano@isciii.es MARÍA-LUISA GASPAR mlgaspar@isciii.es Received: March 8, 2018. Accepted: December 14, 2018. Pre-published: December 20, 2018. doi:10.3324/haematol.2018.192559

Introduction Megakaryocytes are the hematopoietic cells responsible for the production of platelets. In adults, these cells are generated in the bone marrow (BM) from hematopoietic stem cells (HSC) via a common megakaryocyte and erythroid progenitor (MEP) that expresses the receptor for SCF (c-Kit) and is negative for lineage-specific antigens (Lin-), for the stem cell antigen-1 (Sca1), and for the fms-related tyrosine kinase 3 receptor, Flt3/CD135 (Flt3-LS-K cells).1 Nevertheless differentiation into all hematopoietic lineages, including the megakaryocyte/erythroid, from Flt3+ progenitors was also obtained.2 The hierarchical model of hematopoiesis defines progressively restricted lineage-committed progenitors. From HSC a pool of multipotent progenitors (MPP) produces common lymphoid progenitors (CLP) and common myeloid progenitors (CMP), these latter giving rise to MEP and granulocyte/macrophage progenitors (GMP).3,4 However, several recent reports suggest that megakaryocyte/erythroid-commitment may happen directly from HSC or from MPP, supporting a model of multiple lineage commitments occurring in parallel within the HSC/MPP cell pool.5-8 Among CMP, CD41+ cells that express a megakaryocyte-specific signahaematologica | 2019; 104(9)

Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1853 ©2019 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.

1853

I. Cortegano et al.

ture, platelet factor 4 (PF4), CD9, von Willebrand factor (VWF), are separated from erythroid progenitors, and closer to other myeloid progenitors expressing Flt3 and the macrophage colony-stimulating factor-1 receptor (Csf1r/CD115).7 Clonal unilineage megakaryocyte progenitors (MKP) were defined as burst-forming unit megakaryocytes (BFU-MK) and as colony-forming unit megakaryocytes (MK-CFU), and were Lin-c-Kit+Sca1FcγRII/IIIloCD127-Thy1.1-CD9++CD41+ cells expressing the thrombopoietin receptor (myeloproliferative leukemia virus, MPL).1,9 Other studies revealed distinct lineage potentials among erythromyeloid progenitors,10 defining megakaryocyte/erythroid-committed progenitors (PreMegE) as Lin-Sca1-c-Kit+FcγR-CD105-CD150+CD41- and MKP, exclusively associated with megakaryocyte generation, as Lin-Sca1-c-Kit++CD150+CD41+. Embryo hematopoiesis proceeds in two phases, primitive and definitive, which are conserved among different species, including mice and humans.11,12 In the mouse, a primitive wave of erythromyeloid cells forms in the yolk sac (YS) at E7.5.13,14 At E8.5 erythromyeloid progenitors are generated in the YS and the intraembryonic paraaortic splanchnopleura/aorta-gonads-mesonephros region (PSp/AGM), the latter also containing progenitors with lymphoid activity.15-17 Definitive HSC that are the source of all adult hematopoietic cell lineages are present in the PSp/AGM at E10.5.18 The emergence of these definitive HSC in the embryo is dependent on the expression of the transcription factor RUNX1,19 which is required for progression of CD41+ embryonic precursors into HSC.20 The fetal liver (FL) represents the major hematopoietic organ during gestation, receiving extrinsic HSC and MPP from the YS, PSp/AGM and the placenta at E10.5. MEP involved in primitive and definitive megakaryopoiesis appear in the YS at E7.25 and at E9.5, respectively.21,22 RUNX1-independent diploid platelet-forming cells have been identified in the YS at E8.5/10.5.23 Moreover, CD42c+ megakaryocytes can be identified in the YS, in circulation and in the FL from E9.5 onwards, and large reticulated immature platelets circulate at E10.5.21,24 Embryo-derived megakaryocytes differ from those from the adult BM, as illustrated by the in vitro effects of thrombopoietin,25 cell-intrinsic differences in vivo after transplantation26 and the smaller size of those from YS.22 In the FL from E10.5-E11.5 mice, megakaryocytes progressively increase in size and ploidy.27 However, despite several reports on BMderived megakaryopoiesis published recently, the intermediate cells that appear during this process early in life, and the changes in surface phenotype, have yet to be fully defined. We found previously that at E10.5/E11.5, FL megakaryocytes are c-KitDCD49f++CD41++CD9++CD42c+VWF+ and they rapidly produce, independently of thrombopoietin stimulation, proplatelet-bearing megakaryocytes (P-MK) in vitro.28 Strikingly, these FL megakaryocytes were CD41++CD45-, as were the diploid platelet-forming cells found in the YS.23 Here we show that, unlike those from BM, FL megakaryocytes remain CD45- until E13.5, as do the PreMegE and MKP. However, both CD41+CD45+ and CD41+CD45- cells are present in the FL, these populations bearing MK-CFU, megakaryocyte gene expression, and containing Lin-Sca1-c-Kit++CD150+CD41+ MKP. These cells develop into CD41++CD45-CD42c++ P-MK in vitro. The E11.5 FL also contains CD41-CD45++CD11b+ cells that produce CD41+CD45+ cells in vitro, although they do not develop 1854

into P-MK. Accordingly, CD45++EGFP+ cells from E11.5 FL ex vivo preparations from MaFIA transgenic mice, which trace cells expressing Csf1r/CD115,29 give origin poorly to CD41++ cells both in vivo and in vitro. Interestingly, a high proportion of adult BM CD41++CD45++CD9++CD42c++ megakaryocytes from C57BL/6 mice express CD115 and are EGFP+ in MaFIA mice. Our results identify different pathways of megakaryopoiesis in the mouse embryo FL and in adult BM, driven by distinct MKP expressing or not CD45.

Methods Mice and embryo microsurgery and cell suspensions BALB/c, C57BL/6 and C57BL/6-Tg(Csf1r-EGFPNGFR/FKBP1A/TNFRSF6) 2Bck/J MaFIA29 mice were maintained at the animal facilities of the Instituto de Salud Carlos III. All animal studies were approved by the Animal Health Ethics Authority from the Autonomous Government of Madrid (PROEX 080/15). Embryo microsurgery and cell suspensions were obtained as described previously28 and in the Online Supplementary Methods.

Flow cytometry and cell purification Cells were stained as reported elsewhere,28 with the fluorochrome-labeled antibodies described in the Online Supplementary Methods and Online Supplementary Tables S1 and S2.

Quantitative real-time polymerase chain reaction analysis RNA was extracted, oligo(dT)-primed cDNA samples were prepared and quantitative real-time polymerase chain reaction (RTqPCR) amplifications were performed with the primers and protocols described,30,31 as indicated in the Online Supplementary Methods and Online Supplementary Table S3.

Colony-forming cell assays and cell cultures Clonal semisolid cultures and cultures of purified cell populations were performed as indicated in the Online Supplementary Methods.

Immunofluorescence Immunostaining was performed on cryosections from YS and embryos as indicated in the Online Supplementary Methods. The preparations were analyzed by confocal microscopy (Leica DMRD) and the images were processed with ImageJ software.

Statistical analysis GraphPad Prism 4.0 software was used to calculate the means and standard error of the mean (SEM). Comparisons were performed with unpaired and paired Student t tests, with the χ2 test or the Kruskal-Wallis test, to obtain the P values. Data are expressed as mean ± SEM. A P-value less than 0.05 was defined as statistically significant; statistical significance is shown as *P<0.05, **P<0.01 and ***P<0.001.

Results Megakaryocyte lineage cells are present in hematopoietic organs and blood vessels during post-gastrulation embryo development Co-expression of the CD41/aIIa integrin (GPIIb) and CD42c/GPIb-β chains was used to trace megakaryocytes and platelets by flow cytometry. CD41++CD42c+ haematologica | 2019; 104(9)

CD45-negative megakaryopoiesis in the mouse embryo

megakaryocytes were detected from E9.5 in the YS, and in the circulating peripheral blood mononuclear cells and the P-Sp/AGM from this moment on (Figure 1A). Similarly, CD41++CD42c+ cells were found from E10.5 in the FL, although the mean fluorescence intensity (MFI) of CD41 in CD41++CD42c+ megakaryocytes was weaker than that at other locations, as particularly evident in E11.5 samples (Figure 1B). CD41++CD42c+ megakaryocytes were more abundant in the YS than in the P-Sp/AGM and FL at E10.5, and their numbers increased along gestation (Figure 1C). Cells in the window corresponding to platelets (low forward scatter/side scatter on a log scale) were also detected in the YS and FL from E8.5 and E10.5, respectively, and they were in the circulation from E9.5 (Figure 1D) as described previously.23 Consistent with previous reports,21 the earliest platelets were large and although their size

diminished in the YS and FL as gestation progressed, those circulating remain large until E13.5 (Figure 1E).

Megakaryocytes in the E9.5-E11.5 mouse embryo accumulate in the fetal liver To trace the anatomic distribution of CD41++ megakaryocytes at E11.5 in the YS and embryo, immunofluorescence analyses were performed. YS preparations contained clusters of small CD41++ cells that may correspond to aggregated platelets, as well as individual megakaryocytes (Figure 2A). In the embryo proper, CD41++ megakaryocytes were enriched in the FL. From E10.5 to E15.5, FL CD41++ megakaryocytes increased in volume and enhanced the complexity of the so-called membrane demarcation system (DMS, involved in proplatelet formation), indicating maturation to megakaryo-

Figure 1. Megakaryocyte lineage cells are present from E9.5 in the mouse embryo. Cell suspensions from hematopoietic locations in the embryo were prepared and stained with anti-CD41-PE and anti-CD42c-FITC for cytometry. (A) Representative dot-plots of the staining of cell suspensions from the yolk sac (YS), paraaortic splanchnopleura/aorta-gonads-mesonephros region (P-Sp/AGM), peripheral blood mononuclear cells (PBMC) and fetal liver (FL) at the gestational ages indicated (from E8.5 to E11.5). The quadrants define positive cells (determined by using fluorescence-minus-one control isotypes) and the numbers inside the plots are the frequencies of CD41++CD42c+ cells [mean Âą standard error of mean (SEM) n=5, n=6, n=7, and n=16 for E8.5, E9.5, E10.5 and E11.5, respectively]. (B) The bar graphs show the mean fluorescence intensity (MFI) of the CD41 fluorescence among CD41++CD42c+ cells from the locations indicated at E11.5. (C) Absolute numbers of CD41++CD42c+ cells in the YS, P-Sp/AGM and FL from E8.5-E13.5, as derived from the frequencies displayed in the histograms in panel (A) and the total number of cells recovered per organ in each preparation. (D) Representative dot-plots of the CD41 and CD42c staining in cells gated in the low side scatter (SSC)/forward scatter (FSC) window shown in the left dot-plot with the scale FSC as logarithmic, corresponding to platelets. (E) The graph represents the size (determined by the mean FSC channel) of the cells identified in the low SSC/FSC platelet window, in the YS, PBMC and FL from E8.5 to E13.5. The data in the graphs in (B, C and E) are the means Âą SEM [numbers as in panel (A), and n=8 and 17 for E12.5 and E13.5, respectively]. Fluorescence scales are logarithmic. The group comparisons were performed with a two-tailed Student t-test. **P<0.01 and ***P<0.001.

haematologica | 2019; 104(9)

1855

I. Cortegano et al.

cyte stages with increased ploidy32 (Figure 2B-E). In fact, most CD41++CD42c+ megakaryocytes from E11.5 FL were tetraploid (4N or more) (Figure 2F). By contrast, YS CD41++ megakaryocytes were smaller and mostly diploid, as described for the diploid platelet-forming cells.23 At E10.5, FL megakaryocytes displayed a less complex membrane demarcation system than the concurrent YS diploid platelet-forming cells, although the latter did not reach greater membrane demarcation system complexity at E11.5, as did those from FL (Figure 2D). As we found previously,28 E11.5 CD41++VWF+ megakaryocytes were consistently CD45-, and only weak CD45 signals were detected in CD41lo cells (Figure 3A). Flow cytometry analyses of YS and FL cell suspensions from E10.5-E15.5 embryos showed that CD41++CD42c+ megakaryocytes were mainly CD45- in the YS and FL until E13.5 (population #1 in Figure 3B-C), whereas CD45+ cells were detected among the CD41+CD42c- cells (population #2) in these same preparations. Also, most embryo megakaryocytes from E11.5 placenta were CD45- (Figure 3C). From E13.5 onwards, CD41++CD42c+ megakaryo-

cytes displayed low levels of CD45, which increased at E15.5, and megakaryocytes from adult BM were CD45+ (Figure 3C). Accordingly, CD45 was not detected by RTqPCR in CD41++CD42c+ samples from the YS and FL at E11.5, while they expressed Runx1 (Figure 3D). The fact that CD41++CD42c+ megakaryocytes were mostly CD45- in the YS and FL until E13.5 suggested that the initial FL megakaryocytes may be derived by the homing of CD45- megakaryocytes from the YS. The expression of selected markers by the CD41++CD45- cells present in the YS and FL between E10.5 to E11.5 (c-Kit and CD42c) (Table 1) indicated a maturation of these cells in the YS (the brightness of c-Kit dropped and the expression of CD42c increased) that was not so evident in the FL during the same period, and consequently at E11.5 CD41++CD45- cells in the FL showed weaker CD41 fluorescence (Figure 1B), and stronger c-Kit expression (Table 1) than those in the YS at E11.5. Therefore, our data show that at E11.5 YS and FL megakaryocytes differ in their cell volume and ploidy. Those from YS are diploid platelet-forming cells while

Figure 2. Topographical and morphological characteristics of embryo megakaryocytes. Immunofluorescence analyses on E11.5 yolk sac (YS) and embryo tissue slices (10 μm) stained with anti-CD41-FITC (green), and counterstained with DAPI. Representative photomicrographs are shown. The scale bar indicates the magnification of the photomicrographs. (A) The upper photomicrographs show two views of YS samples. The white boxes define the areas magnified in the bottom photomicrographs. (B) The photomicrographs show fetal liver (FL) samples at the gestational ages indicated. (C) Higher magnification of representative megakaryocytes displaying the increased complexity of the membrane demarcation system (DMS) used to define the Pre-DMS (i), Inter-DMS (ii and iii) and Late-DMS (iv) stages. (D) The bar graph displays the relative number (percentage) of YS and FL megakaryocytes presenting the DMS stages identified in the CD41++ megakaryocytes at the gestational ages indicated. (E) The graph displays the cell volume of CD41++ cells. The volume (πD3x1.33) was calculated after measuring the cell diameter “D”. One hundred cells were counted in 20 different photomicrographs for each time point. Data in panels (D) and (E) are the mean ± standard error of mean (SEM) for YS preparations at E9.5 (n=6), E10.5 (n=9) and E11.5 (n=9), and for FL samples at E10.5 (n=3), E11.5 (n=16), E13.5 (n=9) and E15.5 (n=8). (F) Left, representative histograms showing the DAPI staining in nuclei of electronically gated CD41++CD42c+ embryo-derived megakaryocytes (EMK) present in cell suspensions from YS and FL at E11.5; right, quantification of cells with 2N, 4N or >4N ploidy among CD41++CD42c+ EMK. The mean ± SEM are shown (n=3). Comparisons were performed in contingency tables with χ2 and Fisher exact tests. **P<0.01 and ***P<0.001.

1856

haematologica | 2019; 104(9)

CD45-negative megakaryopoiesis in the mouse embryo

those in FL are bigger and mostly tetraploid cells, and express higher levels of Runx1, even though they bear less CD41 and CD42c intensity than those in the YS. From E11.5 to E15.5, FL megakaryocytes increase in size, as well as membrane complexity and become CD45+.

Weak CD45 expression in megakaryocyte progenitors in the E11.5 embryo In the adult BM, all nucleated stages of megakaryocyte differentiation are CD45+.33 Since most megakaryocytes were CD45- at E11.5, we wondered whether megakaryo-

cyte lineage-committed progenitors were also CD45- at these embryonic stages. We therefore determined the number of cells expressing CD45, and the levels of CD45, among LSK, PreMegE, MKP, CMP, GMP and CLP cells (their phenotypes are defined in Online Supplementary Table S1) in E11.5 FL cell suspensions, comparing these with cells from the liver and BM of newborn mice (postnatal day 3, PD3) and from those in the adult BM. The level of CD45 expression was weaker in E11.5 samples than in those from adult mice. Remarkably, both the frequency of cells expressing CD45 and the levels of CD45

C E

Figure 3. Megakaryocytes and megakaryocyte-lineage committed progenitors are CD45- in the yolk sac and embryo at E10.5-E13.5. (A) Left photomicrograph: the fetal liver (FL) in an embryo preparation stained with anti-CD41 (green) and anti-CD45 (red). The boundaries of the vessel (V) are indicated by the dotted line. Right photomicrographs: higher magnification of cells indicated by the white boxes showing overlaid signals and separated in channels. Green CD41++ cells negative for the red CD45 stain are shown. (B) Yolk sac (YS) and FL cell suspensions from E10.5, E11.5, E13.5 and E15.5 embryos were stained with anti-CD41-PE, anti-CD42cFITC and anti-CD45-PE-Cy7. The upper-left dot-plot displays a representative CD41/CD42c staining showing the CD41++CD42c+ megakaryocytes and CD41+CD42ccell populations (labeled as 1 and 2, respectively) analyzed for expression of CD45 in the histograms. The vertical lines in the histograms indicate the fluorescenceminus-one (FMO) isotype control limit. Numbers inside the histograms are the percentages of positive cells. (C) Bar graphs showing the quantification (relative number) of CD45+ cells among the CD41+CD42c- cells and CD41++CD42c+ megakaryocytes. The mean ± standard error of mean (SEM) for E10.5 (n=9), E11.5 (n=9), E13.5 (n=9), E15.5 (n=8), placenta (n=4) and adult bone marrow (BM) (n=4) is shown. (D) CD45 and Runx1 expression analyzed by real-time quantitative polymerase chain reaction on samples of purified CD41+CD42c- and CD41++CD42c+ cells from the E11.5 YS and FL. The results were calculated relative to the expression of the HPRT housekeeping gene using the 2-DCt method. The data are the mean ± SEM (n=4). Results for total FL at E11.5 are shown as C+. (E) After tracing and electronically excluding Lin+ cells with biotin-labeled antibodies against Ter119, B220, CD19, CD11b and anti-CD90.2 revealed with the fluorochrome-labeled streptavidin indicated below, progenitor populations in E11.5 FL and adult BM cell suspensions were identified by multicolor flow cytometry by using combinations of antibodies, as follows: (i) anti-Sca1-PE-Cy7, anti-c-Kit-APC, anti-Flt3-PE, and streptavidin-FITC to identify LSK (Lin-c-Kit++Sca1+) cells and common lymphoid progenitors (CLP: Lin-c-Kit+Sca1+); and (ii) anti-c-Kit-APC, anti-CD34-BV421, anti-FcγRII/III-FITC, anti-CD150-PerCp-Cy5.5, and anti-CD41-PE, with anti-Sca1-PE-Cy7 and streptavidin-PE-Cy7, to identify granulocyte/macrophage progenitors (GMP: Lin-c-Kit++Sca1-CD34+FcγRII/III++), common myeloid progenitors (CMP: Lin-c-Kit++Sca1CD34++FcγRII/III-), megakaryocyte/erythroid-committed progenitors (PreMegE: Lin-Sca1-c-Kit+CD150++CD41-) and megakaryocyte progenitors (MKP: Lin-Sca1-cKit+CD150++CD41+). CD45 expression was monitored with anti-CD45-APC-Cy7. The histograms show the expression of CD45 by progenitor cells in the E11.5 FL and adult BM (filled gray histograms). The FMO isotype signal is shown overlaid (dotted line). The data shown are from one representative experiment. Fluorescence scales are logarithmic. (F) The quantification (frequency) of CD45+ cells and their mean fluorescence intensity (MFI) in the CD45 channel are shown in the bar graphs. The horizontal dotted line represents the isotype background limit. The data in the graphs are the means ± SEM (n=5), comparing the groups with the twotailed Student t-test. *P<0.05, **P<0.01 and ***P<0.001.

haematologica | 2019; 104(9)

1857

I. Cortegano et al. Table 1. Expression of surface receptors in CD45-CD41++ megakaryocytes from the yolk sac and fetal liver.

E10.5 YS

E11.5 YS

E10.5 FL

E11.5 FL

P value E11.5 YS/FL

P value YS E10.5/E11.5

P value FL E10.5/E11.5

246±35 (3) 402±11 (8)

188±10 (12) 474±10 (10)

281±10 (3) 326±9 (7)

273± 14 (12) 383±12 (8)

P<0.001

P<0.05

P<0.001

P<0.01

60±11 (4) 90±1 (8)

63±2 (6) 95±1 (8)

75±7 (5) 82±2 (8)

77±4 (6) 90±1 (8)

P<0.01

MFI c-Kit-APC CD42c-FITC Frequency c-Kit-APC CD42c-FITC

P<0.01 P<0.01

The mean fluorescence intensity (MFI) of electronically gated CD45-CD41++ cells was determined.Values are the mean ± standard error of mean and (number). The statistical significances were calculated using unpaired and paired t tests (the latter when values were obtained from the same pools of embryos). E: embryonic day; YS: yolk sac; FL: fetal liver.

were much lower in LSK cells, PreMegE and MKP at E11.5 than in adult BM (Figure 3E-F). In neonates, CD45 expression increased in progenitors from liver, although remaining lower in PreMeg and MKP than in those from neonatal and adult BM (Online Supplementary Figure S1). Hence, our data show that at E11.5, not only megakaryocytes, but also LSK, PreMegE and MKP display less CD45 than at PD3 and in adult mice.

CD45+ and CD45- megakaryocyte lineages are present at E11.5 Adult BM megakaryocytes are Lin-CD41++CD45+ acetylcholinesterase (AChE)+, while immature megakaryocytes are Lin-CD45+CD41+AChE-.34 To identify whether CD41+CD45+ immature megakaryocytes equivalent to those from BM were present in the E11.5 embryo, we analyzed CD41/CD45 expression in cell preparations from the YS and FL. At E11.5 there are CD45+ cell populations that are negative or positive for CD41 (R1/CD41-CD45++ and R2/CD41+CD45+, respectively). Among CD45- cells there are cells expressing low or high levels of CD41 (R3/CD41+CD45- and R4/CD41++CD45-, respectively), or negative for it (DN cells). R2/CD41+CD45+ cells were highly prominent in the YS at E9.5 and E10.5, and R1/CD41CD45++ and R3/CD41+CD45- cells, the first apparent from E10.5, and increasing as development proceeded (Figure 4A-B). Signals for AChE were obtained only for the purified R2/CD41+CD45+ and R4/CD41++CD45- cell subsets (Figure 4C). The R2/CD41+CD45+ cells in FL can be further subdivided based on higher or lower CD45 level (Online Supplementary Figure S2C; R2a and R2b, respectively), with few CD45+ cells displaying high levels of CD41 (Online Supplementary Figure S2C,D; R2c). Expression of the megakaryocyte-related cell surface markers CD42c, MPL, CD9 and CD61 was found in R4/CD41++CD45- and R3/CD41+CD45- cells in YS and FL at E10.5/E11.5, and also in the R2c/CD41++CD45+ cell subset in FL (Figure 4D and Online Supplementary Figures S2 and S3). Since the number of R2c/CD41++CD45+ cells was low (Online Supplementary Figure S2D), there were fewer CD41++CD45+CD42c+ megakaryocytes than CD41++CD45-CD42c+ megakaryocytes at E10.5-E11.5, in agreement with the results displayed in Figure 3B,C. Megakaryocyte-lineage-specific transcripts NF-E2, PF4, VWF and Fli1 were expressed by R4/CD41++CD45- cells, which displayed myeloid-specific 1858

transcripts (PU1 and myeloperoxidase) very weakly (Figure 4E, and data not shown). From now on we will refer to the CD41++CD45-CD42c+ megakaryocytes present in the R4 region in FL samples as embryo-derived megakaryocytes (EMK), and to the CD41++CD45+CD42c+ cells as adult-type megakaryocytes (AMK). When analyzed for the presence of earlier hematopoietic progenitors by flow cytometry (Figure 4F), R4/CD41++CD45- cells comprise only few MKP besides the EMK. Accordingly, when the differentiation potential of purified R4/CD41++CD45- cells from E11.5 FL cell suspensions was analyzed on clonal MegaCult and MethoCult assays (Figure 4G), they only produced megakaryocyte lineage colonies (MK-CFU), and myeloid lineage colonies (M-CFU) in which P-MK were detected as individual cells, like CD45+CD41++CD42c+ megakaryocytes from adult BM (Online Supplementary Figure S4C). The R2/CD41+CD45+ and R3/CD41+CD45- cell subsets also expressed VWF, yet they had a mixture of other progenitors, containing Lin-c-Kit++ subpopulations with the phenotype of GMP, MKP, low numbers of CMP, and in the case of R3/CD41+CD45- cells, also PreMegE, as did CD41+CD45+ cells from adult BM (Online Supplementary Figure S4A,B). Consequently, R2/CD41+CD45+ cells from FL and BM produced both MK-CFU and M-CFU, and the R3/CD41+CD45- cell population from FL produced E/MCFU (Figure 4G and Online Supplementary Figure S4C). On the other hand, R1/CD41-CD45++ cells were mainly Lin+CD11b+, but also contained CLP, CMP and GMP, and produced M-CFU and E/M-CFU. Likewise they accumulated PU1 and myeloperoxidase myeloid-specific transcripts, as did the R2/CD41+CD45+ cells (Figure 4E-G). By contrast, purified DN/CD41-CD45- cells mostly produced E-CFU progenitors and no MK-CFU (Figure 4G). In summary, the expression of AChE and other surface and molecular markers, as well as clonal megakaryocyte and hematopoietic lineage differentiation assays indicated that R4/CD41++CD45-CD42c+ cells are EMK with low proliferative activity and prone to develop proplatelets in vitro, as expected for mature megakaryocytes. The R2/CD41+CD45+ and R3/CD41+CD45- cells contain oligoclonal progenitors including MKP, which could represent, respectively, CD45+ adult-like intermediate stages (iAMK) and CD45- embryo intermediate stages (iEMK) in the differentiation of the megakaryocyte lineage. haematologica | 2019; 104(9)

CD45-negative megakaryopoiesis in the mouse embryo

G E

Figure 4. CD45+ and CD45- megakaryocyte subsets are present at E11.5 in the yolk sac and fetal liver. Flow cytometry studies were performed on preparations from E11.5 fetal liver (FL) cell suspensions using Ter119-PerCP.Cy5.5, CD41-PE, CD45-PE-Cy7 and CD42c-FITC antibodies to analyze CD41/CD45/CD42c expression after electronically excluding Ter119+ cells. Fluorescence scales are logarithmic. (A) Representative contour plots of the yolk sac (YS) and FL cell preparations at the indicated gestational ages. The boxes inside the dot plots identify four cell subsets expressing CD45 and/or CD41, which are labeled (R1-R4 and DN) as indicated in the YS E11.5 dot-plot. (B) The bar graphs represent the frequency of each population (R1-R4) as the means ± standard error of mean (SEM) (E9.5, n=6; E10.5, n=8; E11.5, n=9). (C) Acetylcholinesterase (AChE) expression (brown dots) was determined in the purified cell populations indicated from the E11.5 FL. Representative photomicrographs from one of three experiments of hematoxylin-eosin counterstained cells are shown. C+ cells are purified CD9++CD41++CD42c+ cells from E17.5 FL cell suspensions. Bar, 10 μm. (D) Representative histograms showing the expression of CD42c (gray histograms) in the populations defined in (A), in which each point was analyzed at least three times. The fluorescence-minus-one (FMO) isotype signal is shown overlaid (dotted line). The numbers inside the plots represent the frequency of CD42c+ cells in this experiment, representative of three performed with similar results. (E) Expression of NF-E2, VWF, PF4, PU1 and MPO transcripts in cDNA samples from the E11.5 FL CD45/CD41 cell populations purified by flow cytometry as indicated in panel (A). The values for each transcript were calculated relative to the HPRT gene using the 2-DCt method. The bars represent the means ± SEM. R1, n=5; R2, n=4; R3, n=9; R4, n=6. (F) Relative numbers of progenitor cells present in the indicated CD45/CD41 cell subsets. The data are means ± SEM. (n=3). Progenitor cell populations were identified as in Figure 3E-F and Online Supplementary Table S1. (G) Clonal differentiation assays. Purified cells from the E11.5 FL populations indicated were seeded in semisolid MegaCult medium (upper graph) and in semisolid MethoCult medium (bottom graph). The colonies grown in MethoCult (erythroid and myeloid colony-forming units: E-CFU and M-CFU) and MegaCult (megakaryocyte colony-forming units: MK-CFU) were counted at 3, 7 and 10 days, respectively. The data are the means ± SEM (n=4). Comparisons among groups were performed with the two-tailed Student t-test. *P<0.05, **P<0.01 and ***P<0.001.

In vitro megakaryocyte differentiation stages from CD45+ and CD45- megakaryocyte lineages in the fetal liver at E11.5 In order to reproduce the steps of megakaryocyte differentiation in vitro, we used short-term liquid cultures (STLC) to trace the differentiation of cells from purified R1-R4 E11.5 FL cell suspensions defined in Figure 4A, and from purified adult BM CD41+CD45+CD42c- and CD41++CD45+CD42c+ cells. These STLC have the advantage of rapidly producing sufficient cells for phenotypic and genetic analyses while allowing morphological changes to be observed. After 24 h in STLC, EMK (R4) from E11.5 FL produced adherent cells and other elongated and mobile cells which, after 48 h, emitted proplatelets and were CD41++CD45-CD42c++ P-MK (Figures 5A-D). haematologica | 2019; 104(9)

STLC of R2/CD41+CD45+ cells also contained CD41++CD42c+ megakaryocytes that were either CD45+ or CD45-, and very few CD41+CD45+ cells (3.6% ± 1.5%, n=4). Similar results were obtained in STLC with R3/CD41+CD45- progenitors, although the CD41++CD42c+ cells that developed there were mostly CD45- (Figure 5B). DNA content analysis of the CD41++CD42c++ megakaryocytes growing in the STLC from FL R2-R4 cell subsets showed that they accumulated in the 8N stage although the megakaryocytes generated in R4/CD41++CD45- cell STLC reached higher ploidy (Figure 5C). Consistent with the myelo/monocyte traits expressed by R1/CD41-CD45++ cells, these cells generated CD41CD45++CD11b++ cells in STLC, and also CD41++CD45+/CD42c+CD11b- cells (Figure 5B and Online Supplementary 1859

I. Cortegano et al.

Figure S5A). Accordingly, after STLC there was a bias towards myeloperoxidase expression in cultures from R1/CD41-CD45++ cells and towards PF4 in those from R2/CD41+CD45+ cells, R3/CD41+CD45- cells and EMK (Figure 5E). The R4/CD41++CD45CD41++CD45+CD42c+ megakaryocytes generated in R1/CD41-CD45++ STLC were not elongated nor did they present a P-MK morphology (Figure 5D). Therefore, during the differentiation to CD41++CD45- megakaryocytes from CD41-CD45++ and CD41+CD45+ cells, a reduction of CD45 levels occurred in conjunction with an increase of CD42c (Figure 5F). Although we found a reduction in the CD45 transcript levels from CD41+CD45+ cells in STLC

(Online Supplementary Figure S5B), to rule out that the decrease in expression of membrane-bound CD45 was due to the use of anti-CD45 for the isolation of the cells, we performed STLC with isolated CD31++CD42c- cells that contained most CD45+/++ cells.28 After 48 h, the CD31++CD42c- cells gave rise to CD42c+ cells that had reduced their CD45 levels (Online Supplementary Figure S5C). By contrast, STLC from CD41+CD45+CD42c- and BM-purified cells (Online CD41++CD45+CD42c+ Supplementary Figure S4D-E) allowed the growth of large cells and P-MK displaying large proplatelets after 96 h, reaching ploidy stages up to 64N. The megakaryocytes (CD41++CD42c+) in these cultures were consistently

Figure 5. The CD41+CD45- and CD41+CD45+ cell populations from E11.5 fetal liver produce megakaryocytes in vitro. Cells from the CD41/CD45 populations indicated in Figure 4A were isolated by flow cytometry from E11.5 fetal liver (FL) cell suspensions, and cultured in short-term liquid cultures (STLC) in the presence of 50 ng/mL recombinant murine thrombopoietin, either in 96-well plates or 8-well culture slides. After 24 and 48 h the cultures were photographed under light microscopy and the cells were counted in function of their morphology. After 48 h the cells were recovered and stained in suspension with anti-CD41-PE, anti-CD45PE-Cy7 and anti-CD42c-FITC antibodies for cytometry analyses, or stained in the slides with anti-CD41-biotin/Tyr.Cy3 and anti-CD42c-FITC antibodies. (A) Representative contour plots of cells after the sorting procedure for each population are shown in the upper plots. Below-left, a representative photomicrograph (left panel) of R2/CD41+CD45+ cells growing in STLC showing an adherent cell (asterisk, ADH); elongated, mobile cell (arrowhead, EMC); and proplatelet-bearing megakaryocyte (arrow, P-MK). Bar, 10 μm. The bar chart in the right panel shows the frequency of the cells with these morphologies in cultures from the indicated purified cells. The data are means ± standard error of mean (SEM) (200 cells counted in 8-9 photographs for each culture, from 4 different experiments counted by 2 independent investigators). (B) Representative contour plots of the indicated cells after 48 h in culture. The numbers in the plots represent the frequency of the cells in the boxes. Data are means ± SEM, n=4. Fluorescence scales in the plots in panels (A) and (B) are logarithmic. (C) Left, representative histograms showing the DAPI staining in nuclei of electronically gated CD41++CD42c+ from R2 and R4 cultured cells; right, quantification of cells with 2N-16N in CD41++CD42c+ megakaryocytes in 48 h STLC from R2 (white), R3 (hatched) and R4 (black) cells. The means ± SEM are shown (n=5). Group comparisons were performed with the two-tailed Student t-test. (D) Photomicrographs showing CD41 (red) and CD42c (green) staining in cells grown on culture slides (upper panels): bars represent 25 μm (R1, R2 and R4) and 10 μm (R3). White squares indicate the areas amplified in the bottom panels. (E) Expression of PF4 and MPO transcripts in cDNA samples from E11.5 FL cell populations after STLC. The values for each transcript were calculated relative to the HPRT gene using the 2-DCt method as in Figure 4E. The bars represent the means ± SEM. R1, n=5; R2, n=4; R3, n=9; R4, n=6. (F) The horizontal bar chart displays the mean fluorescence intensity (MFI) values obtained by cytometry for the CD45 and CD42c fluorescent labeling of cells after 48 h in cultures from the different subpopulations isolated from the E11.5 FL. The data are means ± SEM, n=4. Group comparisons were performed with the two-tailed Student t-test. *P<0.05, **P<0.01 and ***P<0.001.

1860

haematologica | 2019; 104(9)

CD45-negative megakaryopoiesis in the mouse embryo

CD45+, indicating that adult megakaryocytes maintain CD45 expression along their differentiation. In summary, the iAMK and iEMK present among R2/CD41+CD45+ and R3/CD41+CD45- cells produced CD41++CD45-CD42c+ EMK in culture, and in the case of R2/CD41+CD45+ cells also CD41++CD45+CD42c+ AMK, with a characteristic P-MK morphology, while CD41++CD42c+ megakaryocytes growing in BM STLC remained CD45+. Hence, we conclude that in the FL at

E11.5, CD45 levels are modulated in the differentiation towards CD41++CD42c+ megakaryocytes.

CD45++CD11b+CD115+ cells can differentiate into intermediate megakaryocyte stages As expected from their megakaryocyte potential in CFU assays (Figure 4G), CD41-CD45++ cells also differentiated into CD41++CD45+CD42c+ cells, although at the times of analysis they did not undergo great morphological

C A

Figure 6. FL CD11b+CD115+ cells have a low megakaryocyte-lineage potential at E11.5. (A) E11.5 fetal liver (FL) cell suspensions were stained with anti-Ter119PerCP-Cy5.5, anti-CD11b-APC, anti-CD45-PE-Cy7, anti-CD41-PE and anti-CD42c-FITC antibodies. A representative contour plot (upper left) of Ter119- cells (electronically excluded) displaying CD11b expression on CD45++ cells is shown. The box inside (R5) shows the CD45+CD11b+ cell population that is analyzed for the expression of CD41 in the bottom left contour plot, identifying cells expressing or not CD41 (R6 and R7 boxes inside the plot). The purified populations of R5/CD45+CD11b+, R6/CD45+CD11b+CD41- and R7/CD45+CD11b+CD41+ cells were cultured in short-term liquid cultures (STLC) for 48 h, recovered and stained as in Figure 5B. Representative contour plots (upper right) of the cells recovered for the STLC performed with R5/CD45+CD11b+ purified cells are shown. The boxes in the plots indicate the cell subsets analyzed in the bottom graph, which displays the frequency of the cells grown in STLC from R5, R6 and R7 cells. The data are the means ± standard error of mean (SEM) (n=3). (B) Cell suspensions were prepared from bone marrow (BM) from 2-month old C57BL/6 and MaFIA mice, and were stained with anti-CD45-PE-Cy7, anti-CD41-PE, anti-CD9-APC and CD115-BV605. The C57BL/6 preparations also included an anti-CD42c-FITC. Fluorescence-minus-one (FMO) isotype controls for the CD115 and the CD42c antibodies were included. Representative contour plots display the CD41/CD42c/CD115 (upper plot and histogram) or the EGFP/CD41 (bottom plot) signals corresponding to CD9+++ cells (left plot) from C57BL/6 and MaFIA mice, respectively. The histogram shows the CD115 signal (filled in gray) displayed by CD9++CD41++CD42c+ BM megakaryocytes from C57BL/6 mice. The corresponding FMO isotype control is overlaid (dotted line). Boxes inside the bottom plot indicate the EGFP- and EGFP+ CD9+++CD41++ cells that were purified by cell sorting and analyzed by real-time quantitative polymerase chain reaction (RT-qPCR) for the expression of NE-F2 and VWF transcripts as in Figure 4E. The results of the RT-qPCR are displayed in the bottom histograms as means ± SEM (n=3). The upper right graph displays the frequency of the cells in EGFP- and EGFP+ regions (data are the means ± SEM, n=4). (C) The expression of CD115 in E11.5 FL cell preparations from MaFIA mice embryos stained with anti-CD45-PE-Cy7, anti-CD41-PE and anti-CD115-APC antibodies is shown. The quantification of the CD115+ cells among the R1-R4 cell populations (identified as in Figure 4A, left dot-plot) is shown in the graph as means ± SEM (n=3). A representative contour plot (right) of CD115 expression on EGFP+ cells on electronically gated R1/CD41-CD45++ cells is shown. The number inside the plot represents the frequency of CD115+ cells in the box. (D) Representative contour plots of the FL cell preparations (n =3) from MaFIA mice embryos at E11.5. Electronically selected EGFP+ and EGFP- cells (indicated by boxes inside the plots) were analyzed for expression of CD45 and CD41 by staining with anti-CD45-PE-Cy7 and anti-CD41-PE antibodies. The boxes inside the CD45/CD41 plots indicate the CD45++CD41-, CD45+CD41+ and CD45-CD41++ cells, and the number inside is the percentage of CD45CD41++ cells. (E) The contour plots show the analyses of the purified EGFP- and EGFP+ cells after sorting (upper contour plots) and after 48 h in STLC (middle contour plots) stained as in panel (D). The bottom graphs represent the frequency of the cell populations identified in the boxes depicted in the CD45/CD41 plots for the EGFP- and EGFP+ cells that were growing in the STLC. Data are the means ± SEM (n=3). All fluorescence scales are logarithmic. Results were compared with the two-tailed Student t-test. *P<0.05, **P<0.01.

haematologica | 2019; 104(9)

1861

I. Cortegano et al.

Figure 7. Proposed pathways for megakaryopoiesis in the mouse E11.5 embryo liver and in adult bone marrow. Schematic differentiation pathways proposed from data obtained in the experiments performed with cell populations identified ex vivo in the E11.5 fetal liver (FL) and adult bone marrow (BM), and with the short-term liquid culture (STLC) differentiation analysis of purified subpopulations after staining with CD45, CD41 and CD42c antibodies. (A) Purified CD45- embryonic cKit++CD41+CD45-CD42c- immature embryo-type CD45- megakaryocytes (iEMK) differentiate into c-Kit+CD41++CD45-CD42c+ embryo-type megakaryocytes (EMK) (STLC data in Figure 5B,D), and at 24 h become elongated mobile cells (EMC) and differentiate into proplatelet-bearing megakaryocytes (P-MK) after 48 h (bottom left). (B) Purified adult-type c-Kit++CD41+CD45+CD42c- embryonic immature adult-type CD45+ megakaryocytes (iAMK) [highlight for the authors] (that are CD11b-CD115-, see Figure 6C) differentiate in STLC (data in Figure 5B and 5D) towards CD45- embryonic-type megakaryocytes (EMK) (middle left plot) and to CD45+ adult-type megakaryocytes (AMK) (middle-right plot). (C) Cultures from CD41+/-CD45++CD42c-CD11b+CD115+ embryonic progenitors (data in Figure 6E) give rise to CD11b+ myelo/monocytic cells and to a few CD115+ iAMK (right plot). The bottom-right data correspond to the results from ex vivo staining of cell suspensions from adult BM (Figure 6B). The c-Kit data were obtained from staining with anti-c-Kit, anti-CD45 and anti-CD41 as shown in Online Supplementary Figures S5 and S6. Black solid connecting lines indicate results from STLC; red dotted lines are differentiation in adult BM.

changes nor did they develop proplatelets, and most CD45++ cells were myelo/monocyte-committed CD11b+ cells ex vivo (Figure 6A and Online Supplementary Figures S3 and S5D). We reasoned that CD45++ cells could be a heterogeneous population, containing cells able to differentiate into megakaryocytes. Indeed, low levels of CD41 were expressed in CD45++CD11b+ cells (Figure 6A). We postulated that these CD41loCD45++CD11b+ cells may be able to produce CD41++CD45+CD42c+ megakaryocyte-lineage cells. In fact, more CD41++ cells were obtained in STLC from purified CD41loCD45++CD11b+ cells than in those from CD41-CD45++CD11b+ cells (Figure 6A), and there was a bias towards PF4 expression in cells from CD41loCD45++CD11b+ cultures (Online Supplementary Figure S5D). To confirm these results we used samples from MaFIA transgenic mice, which allow tracking of cells expressing the macrophage-specific promoter for Csf1r/CD115.29 We analyzed BM preparations from adult mice, in which megakaryocytes were identified as Ter119CD45+CD9++CD41++CD42c+ (Figure 6B). BM megakaryocytes from C57BL/6 mice expressed CD115 (26% Âą 5.4 1862

%; n=4). Accordingly, around 30% Âą 3.5 % (n=4) of BM megakaryocytes from MaFIA mice were CD45+EGFP++ cells expressing NF-E2 and VWF transcripts, although at higher and lower levels, respectively, than those from EGFP- megakaryocytes. When E11.5 FL preparations were analyzed, one fifth of CD41-CD45++ cells actually expressed CD115 brightly at E11.5, and the CD45++CD115+ were EGFP+ (Figure 6C). However, among these EGFP+ cells only around 1% were CD41++CD45- EMK ex vivo (Figure 6D), and when plated in STLC these EGFP+ cells mainly produced CD41CD45++EGFP+ cells and low numbers of CD41+CD45+ EGFP+ cells (Figure 6E). Surprisingly, after STLC some EGFP- cells became CD41+CD45+/-EGFP+ and CD41++CD45-EGFP+ cells. This observation indicates that the FL EGFP- population contains cells with the potential to become CD115+ in vitro and to generate CD41++ megakaryocytes, which may be reflecting what happens in vivo in the adult BM. Overall, our findings support the notion of a low potential of embryo CD45++ cells to produce CD41+ megakaryocytes in vivo, at difference from the adult BM situation. haematologica | 2019; 104(9)

CD45-negative megakaryopoiesis in the mouse embryo

Discussion The morphological, functional and molecular changes that take place in the differentiation of megakaryocytes have been assessed here using bulk in vitro cultures of megakaryocyte-committed progenitors from the E11.5 FL. One striking finding was that embryonic CD41++CD42c+CD61++CD9++ megakaryocytes are negative, until E13.5, for the leukocyte common CD45 antigen, a large transmembrane glycoprotein expressed on the surface of all hematopoietic cells and their precursors, except mature erythrocytes and platelets.33,35,36 CD45 accounts for up to 10% of lymphocyte cell surface proteins and is involved in the dephosphorylation of the regulatory tyrosine of Src family kinases, negatively modulating cell signaling.33,35,36 The CD45 protein sets the threshold for signal transduction, and CD45 deficiency produces developmental defects and extended phosphorylation of the JAK/STAT cascade.37 The absence of CD45 or diminished levels of this protein have been associated with a hyperadhesive phenotype and impairment of progenitor mobilization from the BM.38,39 It could be that the low expression of CD45 may favor the observed accumulation of megakaryocytes in FL at E11.5, together with interactions through integrin receptors that are expressed highly by megakaryocytes. We used CD41 expression to trace megakaryocytes, since CD41 is expressed strongly by cells of the megakaryocyte lineage, including platelets, in the adult mouse.40 CD41 was defined as a marker for the early stages of primitive and definitive hematopoiesis in the mouse embryo,13 and as a marker of HSC in mice and zebrafish,41,42 tracing the divergence of definitive hematopoiesis from endothelial cells in mouse c-Kit+ progenitors.40,43 CD41++CD45- megakaryocytes are found in the YS and embryo (P-Sp/AGM, FL) from E9.5 and in the circulating blood, as also reported by others.23 Interestingly, E11.5, PreMegE and MKP also display less CD45 than those from newborn and adult BM, whereas CD45 levels appear to be similar in other lineage progenitors, revealing a linkage of the CD45-/dim trait to embryo erythroid/megakaryocyte-lineage cells. Since CD41++ megakaryocytes remain CD45- until E13.5 in the FL, it is tempting to speculate that CD45- EMK may correspond to the primitive wave of megakaryopoiesis generating CD41+CD42c+ Runx1- diploid platelet-forming cells described in the YS at E10.5.23 The progression of primitive HSC to definitive HSC is dependent on RUNX1.20 At E11.5 RUNX1-deficient mice have primitive erythrocytes but lack hematopoietic cells in FL and identifiable platelets in blood.44 They also lack definitive HSC and CD45+ cells, and have very few CD41++CD45- cells.20 It would thus be conceivable that CD45- Runx1+ megakaryocytes present in the FL at E11.5 belong to the definitive wave of megakaryopoiesis. However, RUNX1 is essential for megakaryocyte maturation in the adult BM.45 Therefore, the fact that EMK in the E11.5 FL are Runx1+, and that many of them are tetraploid cells with larger size than those in the contemporaneous YS, may indicate that the local environment in the FL provides conditions allowing maturation of primitive wave CD45- megakaryocytes. It has been described that megakaryocytes require MPL in order to reach >8N maturation stages after E14.5.24 At E11.5, after 2 days in culture, the cell subpopulations isolated from FL produced mostly megakaryocytes with 8N haematologica | 2019; 104(9)

ploidy, which may represent the in vitro differentiation of MPL-independent megakaryocytes. Also, at E11.5 FL R4/CD41++CD45- megakaryocytes express the transcription factors NF-E2 and Fli1, in agreement with the findings on a megakaryocyte transcription factor core for YS diploid platelet-forming cells at E10.5 and for FL megakaryocytes at E13.5.24 In the FL, CD41 and CD45 expression define several cell subsets at E11.5. CD41++CD45- cells are already megakaryocyte-committed CD42c+MPL+CD9+CD61++AChE+ cells that develop rapidly in culture to P-MK, whereas CD41CD45++ cells are mostly CD11b+ myelo/monocyte-committed cells. On the other hand, CD41+CD45+ and CD41+CD45- cells have a more immature phenotype than the aforementioned populations. The phenotypic data and the gene expression profile ex vivo, as well as in vitro studies of these purified populations, prompt us to propose two major pathways of megakaryocyte differentiation operating in the E11.5 FL (Figure 7): (i) from CD41+CD45- iEMK, CD41 is upregulated and CD42c is expressed, producing EMK (CD41++CD45-CD42c+) that develop proplatelets with no evidence of CD45 expression (P-MK); (ii) from CD41+CD45+ iAMK (that are CD115-) (Figure 6C), cells enter a CD41+CD45+CD42c+ stage from which CD41++CD45-CD42c+ EMK arise. Therefore, CD45 diminishes when the levels of CD41 of these increase and they acquire CD42c to become EMK. The first pathway is common before E13.5 but becomes rare after E15.5, and it is currently unknown whether it is even retained at low levels in the BM, while the reverse applies to the CD45derived pathway, although in this case CD45 is retained in BM CD41++CD42c+ megakaryocytes. Moreover, our data reveal the involvement of Csf1r-expressing cells in adult BM CD45+ megakaryopoiesis, which may represent a third pathway of megakaryopoiesis (Figure 7C), minor or absent in the embryo, and opens the issue of the generation of CD41++ megakaryocytes from CD45++CD115expressing cells in the adult. Csf1r/CD115 is considered a mature monocytic differentiation receptor,46 but besides the high expression of Csf1r in monocytes, macrophages, osteoclasts and myeloid dendritic cells, it is also expressed at low levels on HSC, CMP and CLP, as well as among several non-hematopoietic embryonic cells.47 More work is needed to clarify the differential contribution of these CD45++CD11b+CD115+ cells to adult and embryo megakaryopoiesis and its relevance. Embryo-fetal-derived megakaryocytes engraft poorly into adult mice and produce low number of platelets.48,49 It is presently unknown whether the different subsets of megakaryocyte progenitors identified in the embryo may give rise to functional or immature platelets in vivo, but they may represent new tools to uncover the mechanisms underlying the maturation of the membrane demarcation system assembly machinery that yields platelets, similarly as the recently described mechanisms by which BM megakaryocytes sense extracellular matrix rigidity to release platelets.50 In summary, we present a number of findings proving that embryo megakaryocytes are hematopoietic CD45- nucleated cells that are produced from CD45- and CD45+ progenitor cells, findings that may be extended to human cord blood samples in order to probe the existence of a human CD45- megakaryocyte counterpart. These issues have relevant implications for understanding aberrant megakaryopoiesis processes and megakaryocyte-derived tumors, and also represent a tool 1863

I. Cortegano et al.

that may provide clues to improve megakaryocyte reconstitution by using cord blood-derived progenitors for transplantation and for designing better conditions to increase platelet production to treat thrombocytopenic pathologies.

Molecular Severo Ochoa (CBMSO) and IC received a fellowship from the MICINN. The CBMSO receives institutional funding from Fundación Ramón Areces. The CNIC is supported by the MEIC and the Pro CNIC Foundation, and is a Severo Ochoa Center of Excellence (MEIC award SEV-2015-0505).

Funding This work was supported by grants from the Ministerio de Ciencia e Innovación (MICINN SAF2009-12596) and from the Ministerio de Economía y Competitividad (MINECO SAF2012-33916 and SAF2015-70880-R MINECO/FEDER). NS was the recipient of a fellowship from the Centro de Biología

Acknowledgments The authors would like to thank Miriam Pérez-Crespo and Eduardo Martorell for help with animal care, Fernando Gonzalez for support with the confocal microscopy and Mark Shefton, medical writer from BioMedRed Company, for editing the manuscript.

References 1. Luc S, Anderson K, Kharazi S, et al. Downregulation of Mpl marks the transition to lymphoid-primed multipotent progenitors with gradual loss of granulocyte-monocyte potential. Blood. 2008;111(7):3424-3434. 2. Boyer SW, Schroeder AV, Smith-Berdan S, Forsberg EC. All hematopoietic cells develop from hematopoietic stem cells through Flk2/Flt3-positive progenitor cells. Cell stem cell. 2011;9(1):64-73. 3. Akashi K, Traver D, Miyamoto T, Weissman IL. A clonogenic common myeloid progenitor that gives rise to all myeloid lineages. Nature. 2000;404(6774):193-197. 4. Kondo M, Weissman IL, Akashi K. Identification of clonogenic common lymphoid progenitors in mouse bone marrow. Cell. 1997;91(5):661-672. 5. Sanjuan-Pla A, Macaulay IC, Jensen CT, et al. Platelet-biased stem cells reside at the apex of the haematopoietic stem-cell hierarchy. Nature. 2013;502(7470):232-236. 6. Nishikii H, Kanazawa Y, Umemoto T, et al. Unipotent megakaryopoietic pathway bridging hematopoietic stem cells and mature megakaryocytes. Stem Cells. 2015;33(7):2196-2207. 7. Paul F, Arkin Y, Giladi A, et al. Transcriptional heterogeneity and lineage commitment in myeloid progenitors. Cell. 2015;163(7):1663-1677. 8. Perie L, Duffy KR, Kok L, de Boer RJ, Schumacher TN. The branching point in erythro-myeloid differentiation. Cell. 2015; 163(7):1655-1662. 9. Nakorn TN, Miyamoto T, Weissman IL. Characterization of mouse clonogenic megakaryocyte progenitors. Proc Natl Acad Sci U S A. 2003;100(1):205-210. 10. Pronk CJ, Rossi DJ, Mansson R, et al. Elucidation of the phenotypic, functional, and molecular topography of a myeloerythroid progenitor cell hierarchy. Cell Stem Cell. 2007;1(4):428-442. 11. Martinez-Agosto JA, Mikkola HK, Hartenstein V, Banerjee U. The hematopoietic stem cell and its niche: a comparative view. Genes Dev. 2007;21(23):3044-3060. 12. Mikkola HK, Orkin SH. The journey of developing hematopoietic stem cells. Development. 2006;133(19):3733-3744. 13. Ferkowicz MJ, Starr M, Xie X, et al. CD41 expression defines the onset of primitive and definitive hematopoiesis in the murine embryo. Development. 2003;130(18):43934403. 14. Palis J, Robertson S, Kennedy M, Wall C,

1864

15.

16.

17.

18. 19.

20.

21.

22.

23.

24.

25.

26.

Keller G. Development of erythroid and myeloid progenitors in the yolk sac and embryo proper of the mouse. Development. 1999;126(22):5073-5084. Godin IE, Garcia-Porrero JA, Coutinho A, Dieterlen-Lievre F, Marcos MA. Para-aortic splanchnopleura from early mouse embryos contains B1a cell progenitors. Nature. 1993;364(6432):67-70. Marcos MA, Godin I, Cumano A, et al. Developmental events from hemopoietic stem cells to B-cell populations and Ig repertoires. Immunol Rev. 1994;137:155-171. Medvinsky AL, Samoylina NL, Muller AM, Dzierzak EA. An early pre-liver intraembryonic source of CFU-S in the developing mouse. Nature. 1993;364(6432):64-67. Medvinsky A, Dzierzak E. Definitive hematopoiesis is autonomously initiated by the AGM region. Cell. 1996;86(6):897-906. North T, Gu TL, Stacy T, et al. Cbfa2 is required for the formation of intra-aortic hematopoietic clusters. Development. 1999;126(11):2563-2575. Liakhovitskaia A, Rybtsov S, Smith T, et al. Runx1 is required for progression of CD41+ embryonic precursors into HSCs but not prior to this. Development. 2014;141(17): 3319-3323. Tober J, Koniski A, McGrath KE, et al. The megakaryocyte lineage originates from hemangioblast precursors and is an integral component both of primitive and of definitive hematopoiesis. Blood. 2007;109(4): 1433-1441. Xu MJ, Matsuoka S, Yang FC, et al. Evidence for the presence of murine primitive megakaryocytopoiesis in the early yolk sac. Blood. 2001;97(7):2016-2022. Potts KS, Sargeant TJ, Markham JF, et al. A lineage of diploid platelet-forming cells precedes polyploid megakaryocyte formation in the mouse embryo. Blood. 2014;124 (17):2725-2729. Potts KS, Sargeant TJ, Dawson CA, et al. Mouse prenatal platelet-forming lineages share a core transcriptional program but divergent dependence on MPL. Blood. 2015;126(6):807-816. Pastos KM, Slayton WB, Rimsza LM, Young L, Sola-Visner MC. Differential effects of recombinant thrombopoietin and bone marrow stromal-conditioned media on neonatal versus adult megakaryocytes. Blood. 2006;108(10):3360-3362. Slayton WB, Wainman DA, Li XM, et al. Developmental differences in megakaryocyte maturation are determined by the microenvironment. Stem Cells. 2005;23(9): 1400-1408.

27. Matsumura G, Sasaki K. The ultrastructure of megakaryopoietic cells of the yolk sac and liver in mouse embryo. Anat Rec. 1988;222(2):164-169. 28. Serrano N, Cortegano I, Ruiz C, et al. Megakaryocytes promote hepatoepithelial liver cell development in E11.5 mouse embryos by cell-to-cell contact and by vascular endothelial growth factor A signaling. Hepatology. 2012;56(5):1934-1945. 29. Burnett SH, Kershen EJ, Zhang J, et al. Conditional macrophage ablation in transgenic mice expressing a Fas-based suicide gene. J Leukoc Biol. 2004;75(4):612-623. 30. Marcos MA, Morales-Alcelay S, Godin IE, Dieterlen-Lievre F, Copin SG, Gaspar ML. Antigenic phenotype and gene expression pattern of lymphohemopoietic progenitors during early mouse ontogeny. J Immunol. 1997;158(6):2627-2637. 31. Gozalbo-Lopez B, Andrade P, Terrados G, et al. A role for DNA polymerase mu in the emerging DJH rearrangements of the postgastrulation mouse embryo. Mol Cell Biol. 2009;29(5):1266-1275. 32. Eckly A, Heijnen H, Pertuy F, et al. Biogenesis of the demarcation membrane system (DMS) in megakaryocytes. Blood. 2014;123(6):921-930. 33. Charbonneau H, Tonks NK. 1002 protein phosphatases? Annu Rev Cell Biol. 1992; 8:463-493. 34. Matsumura-Takeda K, Sogo S, Isakari Y, et al. CD41+/CD45+ cells without acetylcholinesterase activity are immature and a major megakaryocytic population in murine bone marrow. Stem Cells. 2007;25(4):862870. 35. Thomas ML. The leukocyte common antigen family. Annu Rev Immunol. 1989;7:339-369. 36. Hermiston ML, Xu Z, Weiss A. CD45: a critical regulator of signaling thresholds in immune cells. Annu Rev Immunol. 2003;21:107-137. 37. Irie-Sasaki J, Sasaki T, Matsumoto W, et al. CD45 is a JAK phosphatase and negatively regulates cytokine receptor signalling. Nature. 2001;409(6818):349-354. 38. Shivtiel S, Kollet O, Lapid K, et al. CD45 regulates retention, motility, and numbers of hematopoietic progenitors, and affects osteoclast remodeling of metaphyseal trabecules. J Exp Med. 2008;205(10):2381-2395. 39. Shivtiel S, Lapid K, Kalchenko V, et al. CD45 regulates homing and engraftment of immature normal and leukemic human cells in transplanted immunodeficient mice. Exp Hematol. 2011;39(12):1161-1170.e1. 40. Zhang J, Varas F, Stadtfeld M, Heck S, Faust N, Graf T. CD41-YFP mice allow in vivo

haematologica | 2019; 104(9)

CD45-negative megakaryopoiesis in the mouse embryo

labeling of megakaryocytic cells and reveal a subset of platelets hyperreactive to thrombin stimulation. Exp Hematol. 2007;35(3): 490-499. 41. Bertrand JY, Kim AD, Teng S, Traver D. CD41+ cmyb+ precursors colonize the zebrafish pronephros by a novel migration route to initiate adult hematopoiesis. Development. 2008;135(10):1853-1862. 42. Mitjavila-Garcia MT, Cailleret M, Godin I, et al. Expression of CD41 on hematopoietic progenitors derived from embryonic hematopoietic cells. Development. 2002;129 (8):2003-2013. 43. Boisset JC, van Cappellen W, Andrieu-Soler C, Galjart N, Dzierzak E, Robin C. In vivo imaging of haematopoietic cells emerging

haematologica | 2019; 104(9)

from the mouse aortic endothelium. Nature. 2010;464(7285):116-120. 44. Okuda T, van Deursen J, Hiebert SW, Grosveld G, Downing JR. AML1, the target of multiple chromosomal translocations in human leukemia, is essential for normal fetal liver hematopoiesis. Cell. 1996;84 (2):321-330. 45. Ichikawa M, Asai T, Saito T, et al. AML-1 is required for megakaryocytic maturation and lymphocytic differentiation, but not for maintenance of hematopoietic stem cells in adult hematopoiesis. Nat Med. 2004;10(3): 299-304. 46. Lieu YK, Reddy EP. Impaired adult myeloid progenitor CMP and GMP cell function in conditional c-myb-knockout mice. Cell

Cycle. 2012;11(18):3504-3512. 47. Stanley ER, Chitu V. CSF-1 receptor signaling in myeloid cells. Cold Spring Harb Perspect Biol. 2014;6(6):a021857. 48. Fuentes R, Wang Y, Hirsch J, et al. Infusion of mature megakaryocytes into mice yields functional platelets. J Clin Invest. 2010;120(11):3917-3922. 49. Sim X, Poncz M, Gadue P, French DL. Understanding platelet generation from megakaryocytes: implications for in vitroderived platelets. Blood. 2016;127(10):12271233. 50. Abbonante V, Di Buduo CA, Gruppi C, et al. A new path to platelet production through matrix sensing. Haematologica. 2017;102(7): 1150-1160.

1865

ARTICLE Ferrata Storti Foundation

Platelet Biology & its Disorders

Fatal dysfunction and disintegration of thrombin-stimulated platelets

Oleg V. Kim,1,2 Tatiana A. Nevzorova,3 Elmira R. Mordakhanova,3 Anastasia A. Ponomareva,3,4 Izabella A. Andrianova,3 Giang Le Minh,3 Amina G. Daminova,3,4 Alina D. Peshkova,3 Mark S. Alber,2 Olga Vagin,5,6 Rustem I. Litvinov1,3 and John W. Weisel1

University of Pennsylvania Perelman School of Medicine, Department of Cell and Developmental Biology, Philadelphia, PA, USA; 2University of California Riverside, Department of Mathematics, Riverside, CA, USA; 3Kazan Federal University, Institute of Fundamental Medicine and Biology, Kazan, Russian Federation; 4Kazan Institute of Biochemistry and Biophysics, FRC Kazan Scientific Center of RAS, Kazan, Russian Federation; 5Geffen School of Medicine at UCLA, Department of Physiology, Los Angeles, CA, USA and 6VA Greater Los Angeles Healthcare System, Los Angeles, CA, USA 1

Haematologica 2019 Volume 104(9):1866-1878

ABSTRACT

Correspondence: JOHN WEISEL weisel@pennmedicine.upenn.edu Received: July 18, 2018. Accepted: February 14, 2019. Pre-published: February 21, 2019. doi:10.3324/haematol.2018.202309 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1866

latelets play a key role in the formation of hemostatic clots and obstructive thrombi as well as in other biological processes. In response to physiological stimulants, including thrombin, platelets change shape, express adhesive molecules, aggregate, and secrete bioactive substances, but their subsequent fate is largely unknown. Here we examined late-stage structural, metabolic, and functional consequences of thrombin-induced platelet activation. Using a combination of confocal microscopy, scanning and transmission electron microscopy, flow cytometry, biochemical and biomechanical measurements, we showed that thrombin-induced activation is followed by time-dependent platelet dysfunction and disintegration. After ~30 minutes of incubation with thrombin, unlike with collagen or ADP, human platelets disintegrated into cellular fragments containing organelles, such as mitochondria, glycogen granules, and vacuoles. This platelet fragmentation was preceded by Ca2+ influx, integrin aIIbβ3 activation and phosphatidylserine exposure (activation phase), followed by mitochondrial depolarization, generation of reactive oxygen species, metabolic ATP depletion and impairment of platelet contractility along with dramatic cytoskeletal rearrangements, concomitant with platelet disintegration (death phase). Coincidentally with the platelet fragmentation, thrombin caused calpain activation but not activation of caspases 3 and 7. Our findings indicate that the late functional and structural damage of thrombin-activated platelets comprise a calpain-dependent platelet death pathway that shares some similarities with the programmed death of nucleated cells, but is unique to platelets, therefore representing a special form of cellular destruction. Fragmentation of activated platelets suggests that there is an underappreciated pathway of enhanced elimination of platelets from the circulation in (pro)thrombotic conditions once these cells have performed their functions.

1866

Introduction Platelets are blood cells that play a pivotal role in preventing bleeding (hemostasis) and obstructing blood vessels (thrombosis), in addition to other biological functions.1,2 Activated platelets mechanically contract blood clots and thrombi,3 which is an important pathogenic mechanism in thrombosis.4–6 Under (patho)physiological conditions, platelets activated by stimulants change their morphology, express adhesive molecules, undergo aggregation, and secrete bioactive substances. Disruption of these functions can have pathological consequences, including heart attack and stroke. However, the mechanisms underlying the subsequent fate of activated platelets, including platelet clearance, remain poorly defined. The structural and funchaematologica | 2019; 104(9)

Dysfunction and disintegration of platelets

tional consequences of platelet activation and the survival of these cells are related to fundamental aspects of cell biology, including death pathways of anucleate cells. One of the main platelet activators is thrombin, which causes exposure of activated adhesive proteins (integrins aIIbβ3, aνβ3, a2β1, P-selectin, ephrins, etc.), secretion, contraction, and changes in energy metabolism.7–11 Stimulation leads to the platelets changing from a discoid shape to a star-like cell with multiple protrusions, with this morphological change being driven by reorganization of actin, tubulin, spectrin, and filamin.12–16 Another major consequence of platelet activation is the release of microvesicles segregated into plasma membrane-derived ectosomes and exosomes, originating from intracellular structures, both of which have important biological functions.17,18 Apoptosis plays an essential role in the survival of platelets in vivo.19 The role of apoptosis in platelet lifespan was discovered as a result of the use of BH3-mimetic anticancer drugs, which induce apoptosis in cancer cells, but also cause thrombocytopenia.20 The role of apoptosis in platelet lifespan has been delineated mainly through various mouse models.19 In contrast, the fate of platelets after activation is still mired in controversy21 and, therefore, deserves further study. In this study we tested the hypothesis that thrombininduced platelet activation later results in metabolic exhaustion, dysfunction, and breakup of platelets. Using light and electron microscopy combined with flow cytometry and rheometry, we revealed dynamic alterations of platelet morphology and cytoskeletal rearrangements that accompany biochemical and biomechanical changes in platelets treated with thrombin. Thrombin causes delayed agonistspecific dose-dependent platelet dysfunction and fragmentation associated with reorganization of actin. Thrombininduced exposure of phosphatidylserine and active integrin aIIbβ3, as well as Ca2+ influx characteristic of initial platelet activation, are followed by mitochondrial depolarization, formation of reactive oxygen species and metabolic ATP depletion concomitant with platelet disintegration and activation of calpain, but not effector caspases, suggesting a calpain-dependent pathway of platelet death.

lowing shape changes characteristic of platelet activation, many platelets and small platelet aggregates fell apart into subcellular particles in a time-dependent manner. One hour after addition of thrombin, a substantial fraction (~25%) of platelets underwent disintegration, revealed as various stages of fragmentation (Figure 1A,B). After 3 h of thrombin-induced activation, multiple platelet fragments (>90% of platelets) were dispersed within the fibrin network (Figure 1C). Notably, the platelet fragments displayed a bimodal size distribution with two distinct peaks at about 200 nm and 900 nm (Figure 1D), which are much smaller than intact platelets (2-4 μm). The platelet fragmentation observed by microscopy started ~30 min following thrombin treatment and progressed linearly (Figure 1E). The degree of platelet fragmentation was dose-dependent in the range of 0.1-10 U/mL thrombin (Online Supplementary Figure S1A) and was partially inhibited by rivaroxaban, an inhibitor of factor Xa (Online Supplementary Figure S1B), suggesting the combined destructive effect on platelets of added and endogenously generated thrombin. Remarkably, the onset of platelet fragmentation was the same at 1 U/mL and 10 U/mL concentrations of thrombin, suggesting that platelet disintegration is a delayed process that occurs only after activation is completed and platelets enter a new (dys)functional stage. Flow cytometry of thrombin-treated platelets revealed that the number of CD41-positive fragments smaller than 1 μm increased 10- to 20-fold after 60 min (Figure 1F). Concomitantly, the number of platelets (gated as CD41-positive particles >1 µm) decreased 2.5- to 3-fold compared to the number of untreated platelets (Figure 1G). Time-lapse confocal microscopy of platelets in PRP clots revealed that after initial activation, individual platelets and their aggregates attached to fibrin and extended filopodia to pull on fibrin fibers, causing clot contraction that lasted about 30 min. Subsequently, platelets began decomposing into subcellular fragments (Figure 2). There were two types of platelet-derived particles, which differed in size and cellular origin. One type separated from the ends of the filopodia, while the other type arose from the cell bodies undergoing fragmentation. The tips of filopodia formed smaller platelet fragments (Figure 2A), while much larger subcellular fragments parted from the platelet bodies (Figure 2A,B).

Methods Blood was collected and processed in accordance with a protocol approved by the University of Pennsylvania Institutional Review Board and in compliance with the Helsinki Declaration of ethical principles for medical research involving human subjects. Formation of platelet-rich plasma (PRP) clots, distinct fluorescent labeling of components of the PRP clots for confocal microscopy, dynamic rheometry of contracting clots, platelet isolation, staining of actin and other intracellular structures, scanning and transmission electron microscopy, flow cytometry, measurement of ATP and mitochondrial transmembrane potential, caspase and calpain activity, western blot, image analysis and statistical analyses are described in the Online Supplementary Methods.

Results Thrombin-induced fragmentation of platelets revealed with confocal microscopy and flow cytometry Confocal microscopy of platelets in fresh, hydrated plasma clots formed with thrombin and Ca2+ revealed that, folhaematologica | 2019; 104(9)

Ultrastructural visualization of thrombin-induced platelet fragmentation To see structural details of thrombin-induced platelet fragmentation, isolated platelets treated with thrombin for 15 and 60 min and control untreated platelets were imaged using transmission and scanning electron microcopy (Figure 3). Resting platelets displayed a typical discoid shape with a smooth plasma membrane and cytoplasm containing mitochondria, glycogen granules, lysosomes, an open canalicular system, and dense and α-granules (Figure 3A,C). In contrast, many platelets treated with thrombin for 15 min broke apart into more or less separated membranous particles, some of which contained inclusions, such as mitochondria, glycogen granules and vacuoles (Figure 3B, see fragments 1-4). Scanning electron microscopy of thrombintreated platelets revealed multiple blebs on the surface, likely reflecting fragmentation of the cell body (Figure 3E), quite differently from the untreated control (Figure 3G). At the longer incubation time with thrombin (60 min), platelets underwent further disintegration, producing small fragments (Figure 3D,F) with size distributions peaking at 430 1867

O.V. Kim et al.

nm according to both the transmission and scanning electron micrographs (Figure 3H,I). Importantly, platelets treated with soluble collagen or ADP maintained their integrity during 2 h of incubation, although they did show morphological signs of activation (Online Supplementary Figure S2).

Cytoskeletal rearrangements associated with thrombin-induced platelet fragmentation To see whether fragmentation of thrombin-activated platelets involves rearrangement of actin cytoskeleton, we

used fluorescent confocal microscopy to follow the dynamics of rhodamine-labeled F-actin in resting platelets and thrombin-treated platelets for 1 h. In most resting platelets, F-actin was detected throughout the area of a cell section and was distributed evenly rather than forming patches (Figure 4A and Online Supplementary Figure S3A). In some resting platelets, F-actin was preferentially distributed along the cell periphery. By contrast, in thrombin-activated platelets F-actin was typically detected as intense fluorescent clusters shifted toward the cell center

Figure 1. Thrombin-induced platelet fragmentation. (A-C) Representative confocal microscopy images of thrombin-activated platelets undergoing shape changes and fragmentation. The images were acquired 1 h (A, B) and 3 h (C) after clotting of citrated human platelet-rich plasma (PRP) with 1 U/mL thrombin and 31 mM CaCl2 at 37°C. (A) Fibrin-attached activated platelets with filopodia (1, 2, 5) and platelets at different stages of fragmentation: initial (4), intermediate (3), and final (6). Platelets are labeled with calcein (green) and fibrin is labeled with Alexa Fluor-647 (blue). (B) The same platelets as in (A) are shown without the fibrin channel. (see Online Supplementary Movie S1 for the thrombin-induced fragmentation dynamics at a large scale in three dimensions). (C) Confocal microscopy-based three-dimensional reconstruction of widespread fragmentation of thrombin-treated platelets within a PRP clot prepared as in (A) after 3 h of incubation. Scale bars in (A-C) = 3 μm. (See Online Supplementary Movie S2 for the large-scale three-dimensional reconstructed image). (D) The subcellular particle size distribution in a PRP clot (n=180, bars) fitted with a multi-peak Gaussian function (solid line). (E) Increasing fraction of fragmented platelets in a PRP clot under conditions described in (A-C) determined from time-lapse confocal microscopy (mean ± standard error of mean, n=4). (F) Flow cytometry-based enumeration of platelet-derived CD41-positive platelet fragments normalized by the number of gated platelets after 60 min of incubation of isolated platelets with 1 U/mL thrombin in Tyrode buffer with 2 mM CaCl2 at 37°C (mean ± standard deviation, n=10). (G) Relative numbers of the platelet-gated signals (corresponding to the intact platelets remaining in a sample) after 60 min of incubation under the same conditions as in (E) (mean ± standard deviation, n=12). *P<0.05, **P<0.01.

1868

haematologica | 2019; 104(9)

Dysfunction and disintegration of platelets

(Figure 4B,D and Online Supplementary Figure S3B). F-actin was detected in platelet fragments but the fluorescence intensity was non-uniform (Figure 4E and Online Supplementary Figure S3C). Inhibition of actin polymerization by latrunculin A or cytochalasin D before activation with thrombin prevented platelet fragmentation, as revealed by both confocal microscopy of platelets in plasma clots (Figure 4G,H) and transmission electron microscopy of isolated platelets (Online Supplementary Figure S4A-E). Remarkably, latrunculin A in combination with thrombin caused significant enlargement of the open canalicular system, associated with formation of large vacuoles as well as platelet degranulation, but without disintegration of platelets (Figure 4H and Online Supplementary Figure S4C,D). Pretreatment of platelets with paclitaxel, which inhibits depolymerization of microtubules, also prevented throm-

bin-induced platelet disintegration (not shown), indicating involvement of the plasma membrane cytoskeleton built of microtubules.

Condensation of F-actin in thrombin-stimulated platelets and its extinction during fragmentation In order to quantify changes in the intracellular concentration or compactness of actin, the intensity of the F-actin stain was compared in untreated and thrombin-treated cells, using confocal microscopy at identical microscope settings. Thrombin caused an ~300-fold increase in the overall intracellular F-actin intensity (Figure 4C), which was accompanied by a reduction in the size of individual platelets. The X-Y cross-sectional area of resting platelets was between 3 and 12 μm2 with a median of 7.5 μm2, while in the vast majority of thrombin-treated platelets, the cross-sectional area was less than 5 μm2 with a median

Figure 2. Time-lapse confocal microscopy of thrombin-induced platelet fragmentation in a plasma clot. (A) A single thrombin-activated platelet and (B) an aggregate of three thrombin-activated platelets fall apart and form cellular fragments over time. The upper rows in (A) and (B) show platelets without fibrin. Subcellular fragments originating from the filopodia (white arrowheads) and cell bodies (white arrows) are indicated. The dashed rectangular area in (B) at 24 min and 60 min timepoints show disintegration of a larger part of the platelet aggregate (24 min) into smaller fragments (60 min). Platelets are green and fibrin is red. Scale bars in (A) and (B) = 3 μm.

haematologica | 2019; 104(9)

1869

O.V. Kim et al.

Figure 3. Ultrastructural characterization of thrombin-induced platelet fragments. (A-D) Representative transmission electron micrographs of isolated control untreated platelets (A, C) and platelets incubated with 1 U/mL of thrombin for 15 min (B) and 60 min (D) at 37°C in Tyrode buffer with 2 mM CaCl2. (1-4) Particles formed during thrombin-induced fragmentation of a platelet. The designated structures are a-granules (a), δ-granules (δ), glycogen granules (g), lysosomal vacuoles (l), mitochondria (m), open canalicular system (ocs), vacuole (v). Scale bars in (A-D) = 0.5 µm. (E-G) Representative scanning electron micrographs of isolated platelets incubated at 37°C with 1 U/mL of thrombin for 15 min (E) and 60 min (F) and a control untreated platelet (G) in Tyrode buffer with 2 mM CaCl2. Scale bars in (E-G) = 1 μm. (H, I) Size distributions of thrombin-induced platelet fragments from micrographs obtained with transmission (H, n=400 fragments) and scanning (I, n=210 fragments) electron microscopy. The experimentally determined particle size distributions (bars) in (H) and (I) are fitted with a Gaussian function (solid lines).

1870

haematologica | 2019; 104(9)

Dysfunction and disintegration of platelets

Figure 4. Rearrangement of actin during thrombin-induced platelet fragmentation. (A, B) Representative confocal microscopy z-stack projections of fixed isolated platelets stained for F-actin with Alexa Fluor 488-phalloidin before (A) and 60 min after (B) treatment with thrombin showing the intracellular redistribution of actin and a substantial increase in the intensity of the F-actin stain in the thrombin-treated platelets. Scale bars = 2 μm. (C) The F-actin-related fluorescence intensity in resting versus thrombin-treated platelets was quantified under identical staining conditions and microscope settings (mean ± standard deviation, n=3). (D, E) F-actinbased three-dimensional reconstruction of thrombin-activated platelets in platelet-rich plasma (PRP) clots before (D) and after (E) fragmentation captured at 5 min and 60 min after addition of thrombin, respectively. Scale bars = 2 μm. (F) Histogram of F-actin-containing cellular particles of various size in untreated (n=107) and thrombin-treated (n=107) platelets, showing a shift towards smaller platelet fragments after 1 h of incubation with thrombin. (G, H) Representative confocal microscopy z-stack projections of calcein-labeled platelets (green) treated with thrombin in the absence (G) and presence (H) of latrunculin A, which prevented platelet fragmentation. Scale bars = 10 µm. (I) Mean fluorescence intensity of individual whole platelets or cellular fragments stained for F-actin (total 107 particles analyzed) as a function of their size (mean ± standard deviation). ***P<0.001 and ****P<0.0001 compared to the untreated platelets, two-tailed Mann-Whitney U test.

Figure 5. Time-lapse measurement of intracellular calcium in thrombin-treated platelets. (A, B) Representative confocal micrographs of platelets pre-incubated with a Ca2+-sensitive fluorophore Fluo-4AM at 15 min (A) and 120 min (B) after addition of thrombin to platelet-rich plasma in the presence of extracellular Ca2+. Scale bars = 4 μm. (C) Confocal microscopy-based dynamics of intracellular Ca2+ levels in thrombin-activated platelets (mean ± standard deviation, n=3) in the presence (black circles) and absence (open circles) of extracellular Ca2+ superimposed over the time course of platelet fragmentation in the presence (blue dashed line) and absence (blue open squares) of extracellular Ca2+. The arrow indicates the onset of platelet disintegration in the presence of Ca2+. There were statistically significant differences (P<0.05) for 30 min versus 60 min and for 60 min versus 120 min.

haematologica | 2019; 104(9)

1871

O.V. Kim et al. of 3.5 μm2 (Figure 4F). Remarkably, the intensity of F-actin staining decreased in fragments (Figure 4I), so that in many particles that had a cross-sectional area <4 μm2 Factin staining was not detected at all, suggesting a breakdown of the actin network during platelet fragmentation. The high intensity of F-actin staining in thrombin-activated platelets and its extinction in fragments is clearly seen in three-dimensional reconstructions (Figure 4D,E).

Thrombin-induced platelet fragmentation is preceded by Ca2+ influx and is Ca2+-dependent Because platelet functionality and structural rearrangements depend strongly on Ca2+, we correlated thrombininduced platelet fragmentation with intracellular Ca2+ concentration using confocal microscopy with a Ca2+-dependent fluorophore (Figure 5A,B). The intracellular Ca2+ spiked during the first minutes of thrombin treatment in the presence of extracellular calcium. After spiking, the average platelet calcium content remained constant up to about 30 min, and then decreased greatly, with a strong, inverse correlation with the time course of platelet fragmentation (r=-0.93, P<0.01). In contrast, in the absence of extracellular Ca2+, the intracellular calcium (released from intracellular depots) increased monotonically 4-fold after 120 min of incubation with thrombin, with no evidence of platelet fragmentation within 2 h (Figure 5C).

Mitochondrial depolarization, ATP depletion, and generation of reactive oxygen species in platelets undergoing thrombin-induced fragmentation Thrombin induced a time-dependent reduction of the mitochondrial transmembrane potential (DΨm) in platelets. As revealed by time-lapse confocal microscopy, the overall fluorescence intensity of the DΨm-sensitive MitoTracker dye in freshly formed thrombin-initiated PRP-clots dropped 2- and 4-fold after 60 min and 90 min, respectively (Figure 6A-D). A similar gradual decrease was observed with another DΨm-sensitive dye, tetramethylrhodamine (data not shown). Remarkably, in activated platelets some mitochondria were translocated towards the platelet periphery, localized within filopodia and either remained inside filopodia-derived vesicles or got released as free mitochondria into the extracellular space (Figure 6E,F). The reducing mitochondrial membrane potential strongly and inversely (r=-0.93, P<0.01) correlated with an increase of the fraction of disintegrated platelets, which reached 55% of the total number of visualized platelets by 90 min (Figure 6G). Remarkably, the initial drop of DΨm was almost concurrent with the beginning of platelet fragmentation at about 30 min after thrombin-induced clot formation and platelet activation. This time point also corresponded to the termination of contraction of a PRP clot, measured as a decrease by 90% of platelet-generated contractile stress (Figure 6H), suggesting that platelet disintegration is responsible for the cessation of clot contraction. To establish whether thrombin treatment disturbs energy metabolism, we performed time-lapse quantification of the ATP content in thrombin-treated platelets using confocal microscopy in the presence of an ATP-sensitive fluorophore. The intracellular ATP decreased progressively during the 2 h following platelet activation (Figure 6I). Importantly, such steady kinetics is characteristic of the metabolic depletion of ATP rather than ATP secretion, which occurs as a burst within the first seconds or minutes following platelet activation with thrombin.22 1872

Accordingly, when ATP was measured both in platelet lysates and in activated platelet supernatants during a 3 h incubation of PRP with thrombin, the ATP in the supernatant first increased (over 15 min), reflecting the fast ATP secretion from activated platelets, and then remained constant, while the intracellular ATP decreased progressively over time (Figure 6J). These results confirm that the continuing decrease in thrombin-treated platelet ATP is not due to secretion, but results from gradual metabolic exhaustion. The decline of the intracellular ATP level and reduction of DΨm in response to thrombin were concomitant with reactive oxygen species (ROS) formation, which occurred at about 30 min after activation (Figure 6K-M). Importantly, the increasing mitochondrial ROS production showed a strong temporal correlation (r=-0.98, P<0.01) with the platelet disintegration dynamics (Figure 6M).

Lack of caspase activation in platelets undergoing thrombin-induced fragmentation Previous studies suggested involvement of executor caspases in thrombin-induced platelet activation and further dysfunction.23–25 To test whether thrombin-treated platelets underwent a caspase-dependent death pathway, we used time-lapse confocal microscopy of PRP clots and flow cytometry of thrombin-activated isolated platelets pre-incubated with a fluorogenic substrate of caspases 3 and 7. In both experimental settings, there were no signs of caspase activation in response to thrombin stimulation, at least within 1.5 h (Figure 7A-F). Cleavage of the fluorogenic caspase substrate, as revealed by confocal microscopy, was detected in less than 3% of platelets after 90 min of treatment with thrombin (Figure 7A,C), while treatment for 90 min with calcium ionophore A23187, used as a positive control, led to the activation of caspases 3 and 7 in about one-third of platelets (Figure 7B,C). Flow cytometry also did not reveal any increase of caspase activity in response to thrombin treatment up to 3 h, with less than 0.5% of platelets exhibiting substrate-related fluorescence, while stimulation of platelets with A23187 led to the activation of caspases in 32% of platelets (Figure 7D,E). It is worth noting that an increase of thrombin concentration from 1 U/mL to 5 U/mL did not cause caspase activation (data not shown). In addition, no procaspase 3 cleavage was revealed in western blots of platelet lysates obtained from thrombin-treated platelets (Figure 7G). The lack of caspase activation suggests that thrombin-induced platelet dysfunction and fragmentation represent a caspase-independent pathway of platelet death.

Calpain activity increases in platelets undergoing thrombin-induced fragmentation As an alternative to caspases, calpains may also contribute to platelet fragmentation by cleaving cytoskeletal proteins such as actin and fodrin.26,27 We evaluated activity of calpains in thrombin-treated platelets incubated with a fluorogenic calpain substrate using both flow cytometry of isolated cells and time-lapse confocal microscopy of platelets in PRP clots. As revealed by flow cytometry, treatment of platelets with thrombin resulted in 1.5- and 2-fold increases of a fluorescent signal of the calpain cleavage product after 60 min and 180 min incubation, respectively, as compared to that in untreated cells (Figure 8A). At the same time, the signal from the calpain cleavage product was 7-fold weaker than that of the strongly haematologica | 2019; 104(9)

Dysfunction and disintegration of platelets

Figure 6. Concurrent thrombin-induced platelet fragmentation, cessation of contractility, mitochondrial depolarization, metabolic ATP depletion, and reactive oxygen species production. (A-D) Representative confocal microscopy images of platelet mitochondria stained with a red MitoTracker (A, B) and the same platelets marked with green calcein (C, D) before (A, C) and 90 min after (B, D) adding thrombin to platelet-rich plasma (PRP). (E, F) Confocal microscopy images showing both activated green platelets and red mitochondria (E) or mitochondria only (F) located at the platelet periphery and in filopodia (1, 2, 5) as well as in separated platelet fragments (3, 4, 6). White arrows indicate typical examples of a mitochondrion inside a platelet fragment and a free extracellular mitochondrion. Scale bars in (A-D) = 5 μm, those in (E) and (F) = 3 μm. (G) Time-lapse confocal microscopy of the DΨm-sensitive MitoTracker fluorescence intensity (red symbols) (mean ± standard error of mean, n=4), reflecting a gradual decrease of DΨm inversely correlating with the increasing average fraction of fragmented platelets (blue symbols). There were statistically significant differences for the initial time point versus 30 min and for 30 min versus 90 min (P<0.05, two-tailed Mann-Whitney U test). (H) Superimposition of the platelet-generated contractile stress over platelet fragmentation dynamics in contracting blood clots. The shaded areas in (G) and (H) show the time-frame of cessation of clot contraction concurrent with mitochondrial depolarization and fragmentation. (I) Gradual metabolic reduction of the intracellular ATP content in individual thrombin-stimulated platelets (mean ± standard error of mean, 50-100 platelets analyzed from 3 PRP clots). There were statistically significant differences for the initial time point versus 30 min and for 30 min versus 120 min (P<0.05, two-tailed Mann-Whitney U test). (J) Parallel dynamic measurements of the ATP content in lysates of thrombin-activated platelets and in the activated platelet supernatant following clot formation normalized by the thrombin-free baseline (mean ± standard deviation, n=3). There were statistically significant differences for the initial time point versus 30 min for supernatant, and for the initial time point versus 30 min and 30 min versus 120 min for platelet lysates (P<0.05, twotailed Mann-Whitney U test). Whole blood (G, H) or PRP (I, J) was activated with 1 U/mL thrombin and CaCl2 at 37°C. (K, L) Representative confocal micrographs of platelets pre-incubated with MitoSOXTM Red (mitochondrial super oxide indicator) at 3 min (K) and 60 min (L) after addition of thrombin to PRP in the presence of Ca2+. Scale bars = 4 μm. (See Online Supplementary Movies S3 and S4 for the full 120 min long time sequences at low and high magnifications) (M) Confocal microscopy-based dynamics of reactive oxygen species (ROS) production superimposed on the fragmentation dynamics, both determined in PRP clots (mean ± standard deviation, n=3).

haematologica | 2019; 104(9)

1873

O.V. Kim et al.

increased fluorescence observed after treatment of platelets with A23187, used as a positive control. In accordance with the flow cytometry data, the time-lapse fluorescent confocal microscopy of PRP clots pre-incubated for 90 min with a calpain substrate also revealed activation of platelet calpains in response to thrombin. Specifically, fluorescence of the calpain cleavage product detected in platelets 90 min after thrombin treatment was 6.5-fold higher than that detected in untreated platelets (Figure 8B).

Remarkably, the rate of enzymatic reaction calculated as the first derivative of the dynamic fluorescence signal reached a maximum (the highest calpain activity) after 35 min of thrombin-induced platelet activation, coinciding with a decrease in the mitochondrial membrane potential (Figure 8C) and the beginning of platelet fragmentation (Figure 8D). Accordingly, pre-treatment of platelets with a calpain inhibitor ALLN prior to platelet activation resulted in a 3-

Figure 7. The lack of caspase 3/7 activity in thrombin-stimulated platelets. (A, B) Representative confocal micrographs of activated platelets pre-incubated with a fluorogenic substrate of caspases 3 and 7. Platelets were visualized after activation with thrombin (A) or the calcium ionophore A23187 (B, positive control) in platelet-rich plasma (PRP) clots in the presence of Ca2+. Scale bars = 4 μm. (C) Quantitative estimation of the dynamic caspase-3/7 activity in platelets based on confocal microscopy (mean ± standard deviation, n=3). P<0.01 for the difference between thrombin-treated platelets and the positive control (two-tailed MannWhitney U test). (D-F) Representative flow cytometry dot plots of isolated platelets analyzed for caspase-3/7 activity with a fluorogenic substrate after incubation with thrombin (D), the calcium ionophore A23187 (E) and without treatment (F). The inset boxed numbers (mean ± standard deviation) indicate the average caspase-3/7positive fractions of platelets stimulated with thrombin (D) and A23187 (E) and untreated platelets (F) (n=3, P<0.0001). (G) Western blot analysis of caspase-3 cleavage in platelet lysates obtained from control untreated platelets (Ctrl), platelets treated with thrombin (Thr) and platelets treated with the calcium ionophore A23187 used as a positive control. No caspase-3/7 activation was revealed in platelets treated with 1 U/mL thrombin at 37ºC either in PRP (A-C) or Tyrode buffer (D-G) in the presence of Ca2+.

1874

haematologica | 2019; 104(9)

Dysfunction and disintegration of platelets

fold decrease in the rate of accumulation of the calpain cleavage product (Figure 8B). The residual calpain activity in thrombin-treated platelets exposed to ALLN was still higher than that in untreated platelets because ALLN did not suppress calpain completely, while in the resting platelets calpain was not activated at all. Importantly, pretreatment of platelets with ALLN prior to platelet activation caused a significant initial increase in fluorescence of the DΨm-sensitive MitoTracker by about 35% in freshly formed PRP clots (Figure 8C). Furthermore, the timedependent drop of DΨm was attenuated by ALLN; after 90 min of thrombin-induced platelet activation in the presence of ALLN, DΨm was reduced by only 20%, while without the calpain inhibitor the reduction of DΨm was 65% (P<0.05). To establish whether active calpain is involved in platelet fragmentation, platelets were observed over time in PRP clots in the absence or presence of ALLN (Figure 8D). Calpain inhibition caused an ~30 min delay in the commencement of thrombin-induced platelet fragmentation, which was followed by an increase in the number of

fragmented cells at about the same rate as in the absence of ALLN (27%/min vs. 28%/min with and without ALLN, respectively), suggesting that calpain is involved in platelet fragmentation during the first 30 min, followed by seemingly calpain-independent platelet disintegration. After 90 min of incubation, the fraction of thrombin-activated fragmented platelets pre-treated with ALLN was 55% smaller than the fraction of disintegrated platelets in the absence of ALLN. Thus, inhibition of calpain partially protects thrombin-treated platelets from a progressive decrease of DΨm and delays fragmentation, suggesting involvement of calpain in thrombin-induced platelet dysfunction and disintegration.

Discussion While the early stages of platelet activation have been studied extensively,2,7,10 alterations of platelets occurring at the later stages, after their functions have been fulfilled, are much less well understood. Meanwhile, such late-

Figure 8. Calpain activation in thrombin-treated platelets and the effects of calpain inhibition on the mitochondrial membrane potential and platelet fragmentation. (A) Flow cytometry-based measurement of calpain activity assessed by the fluorogenic substrate. Isolated platelets were analyzed after 15, 60, and 180 min incubation with thrombin and the calcium ionophore A23187 (positive control) in the presence of 2 mM Ca2+ (mean ± standard deviation, n=6). The data were normalized by the values for untreated platelets (negative control) at the corresponding time points (mean ± standard error of mean, n=4). *P<0.05, **P<0.01, ****P<0.0001 (two-tailed Mann-Whitney U test). (B) Confocal microscopy-based time-lapse fluorescence intensity of the calpain cleavage product, reflecting calpain activity or the rate of enzymatic reaction (dashed curves, comprising the first derivative of the experimental plots). Platelets in re-calcified platelet-rich plasma (PRP) were activated with thrombin in the presence or absence of ALLN, a calpain inhibitor, and compared with untreated platelets in PRP (mean ± standard error of mean, n=4). (C) Time-lapse confocal microscopy of the MitoTracker fluorescence intensity, reflecting the platelet mitochondrial transmembrane potential in the presence or absence of ALLN during 90 min incubation with thrombin and Ca2+ at 37°C (mean ± standard error of mean, n=4). (D) Confocal microscopy-based platelet fragmentation dynamics in the PRP clots tracked over time in the absence or presence of ALLN (mean ± standard error of mean, n=4). P<0.05 (B-D) for the difference between thrombin-treated platelets and thrombin-treated platelets in the presence of ALLN (two-tailed Mann-Whitney U test).

haematologica | 2019; 104(9)

1875

O.V. Kim et al.

stage alterations can be critical for platelet fate and may be important for remodeling and resolution of clots and thrombi. In this work, we tracked delayed structural, biochemical, and biomechanical changes in human platelets exposed to thrombin, a potent platelet stimulant in (pro)thrombotic states. Based on the literature and our own observations,18,28 we hypothesized that after a period of thrombin-induced augmented functionality, platelets would become dysfunctional and lose their structural integrity. Our present findings support this hypothesis and shed light on the mechanisms underlying thrombin-induced platelet death. Structurally, a substantial fraction of thrombin-stimulated platelets undergo disintegration into subcellular organellecontaining fragments (Figures 1-3). This effect is specific for thrombin because no fragmentation was induced by collagen or ADP (Online Supplementary Figures S2). Thrombin-induced platelet fragmentation is associated with changes in intracellular Ca2+ concentration and reorganization of the platelet cytoskeleton (Figures 4,5). It is also concurrent with cessation of platelet contractility, metabolic ATP depletion, formation of ROS, and mitochondrial depolarization (Figure 6). Notably, thrombin does not induce detectable activation of the effector caspases 3 and 7 in platelets, suggesting a caspase-independent platelet death pathway (Figure 7). Thrombin-treated platelets break up into extracellular particles of various sizes, origin and composition. It is tempting to identify the platelet body fragments as exosomes, but they are distinct in a number of features. First, particles formed during break up of thrombin-stimulated platelets are relatively large and heterogeneous (0.1-1 μm) (Figures 1-3), whereas exosomes are smaller and more uniform (0.03-0.10 μm).29 Second, the larger platelet fragments are produced by dysfunctional and energetically exhausted platelets, whereas exosomes are generated by metabolically active and structurally intact platelets from their endosomal multi-vesicular bodies.29 Therefore, the particles formed during thrombin-induced platelet disintegration comprise a special type of platelet-derived particles looking like “apoptotic cellular bodies” rather than membrane-derived microvesicles or secreted exosomes and are likely to be cleared by phagocytes from the blood flow.30 Furthermore, structurally, the changes observed in platelets do not resemble those that occur in necrosis since there is no swelling and no rupture of the cell membrane or that of cellular organelles and platelet fragments all retain intact membrane (Figure 3). Fragmentation of thrombin-treated platelets is accompanied by a dramatic increase in F-actin-related fluorescence intensity and redistribution of F-actin towards the center of platelets with formation of highly fluorescent patches (Figure 4), suggesting polymerization and/or clustering of actin. Remarkably, the smallest platelet fragments do not show F-actin staining, suggesting that actin filaments are depolymerized or destroyed at the late stages of platelet disintegration. The essential role of F-actin dynamics in thrombin-induced platelet fragmentation is confirmed by the complete prevention of platelet disintegration after blocking actin polymerization with cytochalasin D or latrunculin A (Figure 4G,H and Online Supplementary Figure S4). The membrane cytoskeleton formed of microtubules is also involved in platelet disintegration, because inhibition of tubulin dynamics with paclitaxel prevented the break-up of platelets caused by thrombin (not shown). 1876

Platelet disintegration induced by thrombin correlates strongly with energy exhaustion and dysfunction that includes mitochondrial depolarization, a drop in the ATP content, and the loss of contractility, all seen at about 30 min after addition of thrombin (Figure 6). The observed mitochondrial depolarization may be attributed to an increase in cytosolic and mitochondrial Ca2+ upon thrombin treatment (Figure 5), which can trigger cyclophilin Ddependent mechanisms of the mitochondrial potential collapse.31 This decrease in DΨm was shown to be associated with generation of ROS (Figure 6), which can damage intracellular structures and could cause platelet death.32,33 Remarkably, some mitochondria in activated platelets localize in filopodia and may be released in the extracellular milieu. Although the mechanism of mitochondria translocation toward the tips of platelet filopodia is not clear, ROS-dependent actin polymerization and mitochondria-bound myosin were shown to participate in intracellular mitochondria transport.34,35 This mechanism may also be important to translocate mitochondria to sites of high ATP utilization, such as contracting filopodia. The metabolic ATP depletion in platelets may be due to mitochondrial depolarization as well as to impaired glycolysis, both important sources of ATP in activated platelets.36–38 The insufficiency of ATP is aggravated by its consumption due to energy-demanding platelet functions, such as contractility. Irrespective of the mechanism of the decrease of ATP content, it is a signature of energy exhaustion and impaired platelet functionality. Not surprisingly, the significant (~54%) decrease of ATP content in platelets after 30 min of thrombin treatment coincided with the termination of platelet-driven clot contraction that depends on the activity of non-muscle myosin II, which is a major part of the ATP-dependent-cell contractile machinery.39 The inhibitory effect of blebbistatin on platelet fragmentation (not shown) implies that the contractile myosin IIaactin complex contributes to platelet disintegration, perhaps facilitating mechanical disconnection or dislodging of fragments. It is not clear whether the revealed biochemical and biomechanical alterations associated with platelet thrombin-induced disintegration provide the conditions for the structural decomposition of platelets or comprise its consequences. However, it is evident that the magnitude and kinetics of the observed late-stage functional and metabolic alterations following thrombin treatment of platelets are distinct from the non-specific changes related to platelet aging or “storage lesions” (see the Online Supplementary Results section). Neither time-lapse confocal microscopy, nor flow cytometry nor western blotting detected procaspase 3/7 activation in platelets in response to thrombin stimulation (Figure 7), indicating that the thrombin-induced platelet death pathway is caspase-independent and is likely nonapoptotic, which is in agreement with previous results40 that demonstrated no caspase-3 activity in platelets treated with thrombin. Alternatively, our data suggest the involvement of Ca2+-dependent protease calpain in platelet disintegration, perhaps due to calpain-catalyzed cleavage of cytoskeletal proteins.41 Calpain may compensate for the lack of active caspases by cleaving caspase substrates such as gelsolin, protein kinase C-δ and fodrin. It can also cleave other cytosolic proteins and many important regulators of apoptosis, including the anti-apoptotic XIAP and Bcl-xL, thereby recapitulating apoptotic events in activated platelets.26,42–44 For most cells, changes in the nucleus are haematologica | 2019; 104(9)

Dysfunction and disintegration of platelets

an essential part of apoptosis, but for the anucleate platelet the cell’s development and lifespan are determined by the intrinsic pathway of apoptosis.19 On the other hand, whether or not apoptosis is involved in the fate of platelets after activation has been controversial, but other death pathways cannot be excluded.45 Irrespective of the mechanisms underlying platelet death as an outcome of thrombin-induced activation, our data suggest a pathway for enhanced elimination of activated platelets from the circulation in (pro)thrombotic conditions associated with thrombinemia. In severe thrombotic conditions, such as disseminated intravascular coagulation46 or trauma-induced coagulopathy,47 platelets may vanish due to removal of platelet fragments from the blood, perhaps by monocytes, dendritic cells and macrophages.48 Thrombin-induced platelet disintegration may therefore be a pathogenic mechanism that modulates platelet counts, functionality, and fate in disease states associated with hypercoagulability and high thrombin activity in blood. In conclusion, following thrombin-induced platelet activation, a substantial fraction of platelets later undergo structural disintegration into subcellular particles. This fragmentation of platelets is accompanied by dramatic rearrangements of platelet cytoskeletal components, including redistribution of actin and microtubule dynamics. Thrombin-induced platelet fragmentation is concur-

References 1. Rosen ED, Raymond S, Zollman A, et al. Laser-induced noninvasive vascular injury models in mice generate platelet-and coagulation-dependent thrombi. Am J Pathol. 2001;158(5):1613–1622. 2. Brass LF, Ma P, Tomaiuolo M, Diamond SL, Stalker TJ. A systems approach to the platelet signaling network and the hemostatic response to injury. In: Platelets in Thrombotic and Non-Thrombotic Disorders. Springer. 2017;p.367–378. 3. Kim OV, Litvinov RI, Alber MS, Weisel JW. Quantitative structural mechanobiology of platelet-driven blood clot contraction. Nat Commun. 2017;8(1):1274. 4. Le Minh G, Peshkova AD, Andrianova IA, et al. Impaired contraction of blood clots as a novel prothrombotic mechanism in systemic lupus erythematosus. Clin Sci. 2018;132(2):243–254. 5. Peshkova AD, Minh GL, Tutwiler V, Andrianova IA, Weisel JW, Litvinov RI. Activated monocytes enhance platelet-driven contraction of blood clots via tissuefactor expression. Sci Rep. 2017;7(1):5149. 6. Tutwiler V, Wang H, Litvinov RI, Weisel JW, Shenoy VB. Interplay of platelet contractility and elasticity of fibrin/erythrocytes in blood clot retraction. Biophys J. 2017;112(4):714– 723. 7. Kamath S, Blann A, Lip G. Platelet activation: assessment and quantification. Eur Heart J. 2001;22(17):1561–1571. 8. Prevost N, Kato H, Bodin L, Shattil SJ. Platelet integrin adhesive functions and signaling. Methods Enzymol. 2007;426:103–115. 9. Shattil SJ. Integrins and Src: dynamic duo of adhesion signaling. Trends Cell Biol. 2005;15(8):399–403.

haematologica | 2019; 104(9)

rent with severe impairment of platelet functionality, including mitochondrial depolarization, metabolic ATP depletion, generation of ROS, and loss of platelet contractility. The lack of caspase activity and increased calpain activity in energy-exhausted thrombin-treated platelets undergoing fragmentation suggests a calpain-dependent platelet death pathway. Fragmentation of activated platelets suggests that platelet death is an underappreciated mechanism for enhanced elimination of platelets from the circulation in (pro)thrombotic conditions or under other conditions once these cells have performed their functions. Analogous to eryptosis, suicidal death of erythrocytes, the platelet death pathway described here could be named “thromboptosis” or “plateleptosis”. Acknowledgments We would like to thank Dr. Xiaolu Yang for fruitful discussions. The work was supported by National Institutes of Health grants UO1HL116330 and R01 HL135254, National Science Foundation grant DMR1505662, the Program for Competitive Growth at Kazan Federal University, grant 18-415-160004 from the Russian Foundation for Basic Research, American Heart Association grants 17SDG33680177 and 16PRE30260002, and grant EPSRC EP/C513037/1 to P. R. Williams (Swansea University, Wales, UK) for the TA Instruments ARG2 rheometer. Transmission electron microscopy was carried out on the equipment of CSF-SAC FRC KSC RAS (Kazan, Russia).

10. Siess W. Molecular mechanisms of platelet activation. Physiol Rev. 1989;69(1):58–178. 11. Pang A, Cui Y, Chen Y, et al. Shear-induced integrin signaling in platelet phosphatidylserine exposure, microvesicle release and coagulation. Blood. 2018;132 (5):533–543. 12. Aslan JE. Platelet shape change. In: Platelets in Thrombotic and Non-Thrombotic Disorders. Springer. 2017;p 321–336. 13. Shin E-K, Park H, Noh J-Y, Lim K-M, Chung J-H. Platelet shape changes and cytoskeleton dynamics as novel therapeutic targets for anti-thrombotic drugs. Biomol Ther. 2017;25(3):223–230. 14. Seifert J, Rheinlaender J, Lang F, Gawaz M, Schäffer TE. Thrombin-induced cytoskeleton dynamics in spread human platelets observed with fast scanning ion conductance microscopy. Sci Rep. 2017;7(1):4810. 15. Sandmann R, Köster S. Topographic cues reveal two distinct spreading mechanisms in blood platelets. Sci Rep. 2016;6:22357. 16. Diagouraga B, Grichine A, Fertin A, Wang J, Khochbin S, Sadoul K. Motor-driven marginal band coiling promotes cell shape change during platelet activation. J Cell Biol. 2014;204(2):177–185. 17. Edelstein LC. The role of platelet microvesicles in intercellular communication. Platelets. 2017;28(3):222–227. 18. Ponomareva A, Nevzorova T, Mordakhanova E, et al. Intracellular origin and ultrastructure of platelet-derived microparticles. J Thromb Haemost. 2017;15(8):1655–1667. 19. Mason KD, Carpinelli MR, Fletcher JI, et al. Programmed anuclear cell death delimits platelet life span. Cell. 2007;128(6):1173– 1186. 20. Oltersdorf T, Elmore SW, Shoemaker AR, et al. An inhibitor of Bcl-2 family proteins

21.

22.

23.

24. 25. 26.

27.

28.

29.

30.

induces regression of solid tumours. Nature. 2005;435(7042):677–681. McArthur K, Chappaz S, Kile BT. Apoptosis in megakaryocytes and platelets: life and death of a lineage. Blood. 2018;131(12):605– 610. Beigi R, Kobatake E, Aizawa M, Dubyak GR. Detection of local ATP release from activated platelets using cell surface-attached firefly luciferase. Am J Physiol. 1999;276(1):C267–C278. Leytin V, Allen DJ, Mykhaylov S, Lyubimov E, Freedman J. Thrombin-triggered platelet apoptosis. J Thromb Haemost. 2006;4(12):2656–2663. Leytin V. Apoptosis in the anucleate platelet. Blood Rev. 2012;26(2):51–63. White MJ, Kile BT. Apoptotic processes in megakaryocytes and platelets. In: Seminars in Hematology. Elsevier. 2010;p 227–234. Vanags DM, Pörn-Ares MI, Coppola S, Burgess DH, Orrenius S. Protease involvement in fodrin cleavage and phosphatidylserine exposure in apoptosis. J Biol Chem. 1996;271(49):31075–31085. Wood D, Thomas A A, Devi L, et al. Bax cleavage is mediated by calpain during druginduced apoptosis. Oncogene. 1998;17(9): 1069–1078. Agbani EO, Williams CM, Hers I, Poole AW. Membrane ballooning in aggregated platelets is synchronised and mediates a surge in microvesiculation. Sci Rep. 2017;7(1):2770. Heijnen HF, Schiel AE, Fijnheer R, Geuze HJ, Sixma JJ. Activated platelets release two types of membrane vesicles: microvesicles by surface shedding and exosomes derived from exocytosis of multivesicular bodies and alfa-granules. Blood. 1999;94(11):3791– 3799. Brown SB, Clarke MCH, Magowan L,

1877

O.V. Kim et al.

31.

32.

33.

34.

35.

1878

Sanderson H, Savill J. Constitutive death of platelets leading to scavenger receptor-mediated phagocytosis. A caspase-independent cell clearance program. J Biol Chem. 2000;275(8):5987–5996. Jobe SM, Wilson KM, Leo L, et al. Critical role for the mitochondrial permeability transition pore and cyclophilin D in platelet activation and thrombosis. Blood. 2008;111(3): 1257–1265. Choo H-J, Saafir TB, Mkumba L, Wagner MB, Jobe SM. Mitochondrial calcium and reactive oxygen species regulate agonist-initiated platelet phosphatidylserine exposure. Arterioscler Thromb Vasc Biol. 2012;32(12):2946–2955. Yamagishi S, Edelstein D, Du X, Brownlee M. Hyperglycemia potentiates collageninduced platelet activation through mitochondrial superoxide overproduction. Diabetes. 2001;50(6):1491–1494. Shneyer BI, Ušaj M, Wiesel-Motiuk N, Regev R, Henn A. ROS induced distribution of mitochondria to filopodia by Myo19 depends on a class specific tryptophan in the motor domain. Sci Rep. 2017;7(1):11577. Boudreau LH, Duchez A-C, Cloutier N, et al. Platelets release mitochondria serving as

36.

37.

38.

39.

40.

41.

substrate for bactericidal group IIA-secreted phospholipase A2 to promote inflammation. Blood. 2014;124(14):2173–2183. Aibibula M, Naseem KM, Sturmey RG. Glucose metabolism and metabolic flexibility in blood platelets. J Thromb Haemost. 2018;16(11):2300–2314. Ravi S, Chacko B, Sawada H, et al. Metabolic plasticity in resting and thrombin activated platelets. PLoS One. 2015;10(4): e0123597. Nayak MK, Dhanesha N, Doddapattar P, et al. Dichloroacetate, an inhibitor of pyruvate dehydrogenase kinases, inhibits platelet aggregation and arterial thrombosis. Blood Adv. 2018;2(15):2029–2038. Adelstein RS, Eisenberg E. Regulation and kinetics of the actin-myosin-ATP interaction. Annu Rev Biochem. 1980;49(1):921– 956. Wolf BB, Goldstein JC, Stennicke HR, et al. Calpain functions in a caspase-independent manner to promote apoptosis-like events during platelet activation. Blood. 1999;94(5): 1683–1692. Goll DE, Thompson VF, Li H, Wei W, Cong J. The calpain system. Physiol Rev. 2003;83(3):731–801.

42. Martin SJ, O’Brien GA, Nishioka WK, et al. Proteolysis of fodrin (non-erythroid spectrin) during apoptosis. J Biol Chem. 1995;270(12):6425–6428. 43. Kothakota S, Azuma T, Reinhard C, et al. Caspase-3-generated fragment of gelsolin: effector of morphological change in apoptosis. Science. 1997;278(5336):294–298. 44. Harwood SM, Yaqoob MM, Allen DA. Caspase and calpain function in cell death: bridging the gap between apoptosis and necrosis. Ann Clin Biochem. 2005;42(6): 415–431. 45. Nogusa S, Thapa RJ, Dillon CP, et al. RIPK3 activates parallel pathways of MLKL-driven necroptosis and FADD-mediated apoptosis to protect against influenza A virus. Cell Host Microbe. 2016;20(1):13–24. 46. Gando S, Levi M, Toh C-H. Disseminated intravascular coagulation. Nat Rev Dis Primer. 2016;2:16037. 47. Chang R, Cardenas JC, Wade CE, Holcomb JB. Advances in the understanding of trauma-induced coagulopathy. Blood. 2016;128 (8):1043–1049. 48. Elliott MR, Ravichandran KS. Clearance of apoptotic cells: implications in health and disease. J Cell Biol. 2010;189(7):1059–1070.

haematologica | 2019; 104(9)

ARTICLE

Platelet Biology & its Dsorders

Ferrata Storti Foundation

Dina Vara,1 Eugenia Cifuentes-Pagano,2 Patrick J. Pagano2 and Giordano Pula1 Institute of Biomedical and Clinical Science, University of Exeter Medical School, Exeter, UK and 2Department of Pharmacology and Chemical Biology and Vascular Medicine Institute, School of Medicine, University of Pittsburgh, Pittsburgh, PA, USA

Haematologica 2019 Volume 104(9):1879-1891

ABSTRACT

he regulation of platelets by oxidants is critical for vascular health and may explain thrombotic complications in diseases such as diabetes and dementia, but remains poorly understood. Here, we describe a novel technique combining electron paramagnetic resonance spectroscopy and turbidimetry, which has been utilized to monitor simultaneously platelet activation and oxygen radical generation. This technique has been used to investigate the redox-dependence of human and mouse platelets. Using selective peptide inhibitors of NADPH oxidases (NOXs) on human platelets and genetically modified mouse platelets (NOX1-/- or NOX2-/-), we discovered that: 1) intracellular but not extracellular superoxide anion generated by NOX is critical for platelet activation by collagen; 2) superoxide dismutation to hydrogen peroxide is required for thrombin-dependent activation; 3) NOX1 is the main source of oxygen radicals in response to collagen, while NOX2 is critical for activation by thrombin; 4) two platelet modulators, namely oxidized low density lipoproteins (oxLDL) and amyloid peptide β (Aβ), require activation of both NOX1 and NOX2 to pre-activate platelets. This study provides new insights into the redox dependence of platelet activation. It suggests the possibility of selectively inhibiting platelet agonists by targeting either NOX1 (for collagen) or NOX2 (for thrombin). Selective inhibition of either NOX1 or NOX2 impairs the potentiatory effect of tested platelet modulators (oxLDL and Aβ), but does not completely abolish platelet hemostatic function. This information offers new opportunities for the development of disease-specific antiplatelet drugs with limited bleeding side effects by selectively targeting one NOX isoenzyme.

Introduction Platelets are anucleated circulating cells responsible for initiating hemostasis via thrombus formation and blood clotting. The regulation of platelets is of primary importance for cardiovascular medicine and for the discovery of new drugs to treat cardiovascular diseases.1 In addition to canonical signaling pathways depending on protein kinase activity,2 platelets are regulated in a redox-dependent manner. Several lines of evidence suggest that platelets are modulated by extracellular reactive oxygen species (ROS)3 and that platelet activation is essentially dependent on the generation of endogenous ROS.4-6 Therefore, the study of platelet regulation and hemostasis is shedding light on the interface between ROS biochemistry and cellular physiology. Superoxide anion (O2•−) from exogenous sources or endogenously produced by platelets is shown to significantly increase platelet aggregation and thrombus formation.7 O2•− has a pre-eminent role in biology and pathophysiology, as it serves as a progenitor for formation of hydrogen peroxide (H2O2), peroxynitrite (ONOO•−) and hydroxyl radical (HO•), and thereby plays a key role in the post-translational oxidative modification of proteins.8 The work of several research groups has focused on haematologica | 2019; 104(9)

Correspondence: GIORDANO PULA g.pula@exeter.ac.uk Received: October 8, 2018. Accepted: January 18, 2019. Pre-published: January 24, 2019. doi:10.3324/haematol.2018.208819 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1879 ©2019 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.

1879

D. Vara et al.

NADPH oxidases (NOX) as important sources of ROS in platelets responsible for the regulation of platelet responsiveness.9-15 Despite the increased interest in this aspect of platelet biology and hemostasis regulation, progress within this field is hampered by the lack of reliable and quantitative techniques for the analysis of platelet oxidative status.16,17 This has made it challenging to completely appreciate the importance of endogenous and exogenous oxidants on the regulation of platelet signaling pathways and on the balance between hemostasis and thrombosis in health and disease. Indirectly, this has impeded the development of pharmacological treatments for thrombotic conditions based on the control of ROS generation. We addressed this biomedical need by combining the measurement of platelet activation (using turbidimetry18) and the simultaneous measurement of intracellular or extracellular oxygen radicals [using electron paramagnetic resonance (EPR) or EPR spectroscopy19,20] into one multiplex technique that allows the accurate study of the oxidative status and function of human platelets. This technique is likely to find application in clinical practice, where the simultaneous analysis of platelet responsiveness and oxidative stress can help develop more advanced diagnostics for patients at risk of thrombotic diseases. It could also find application in drug discovery, as NOX modulation is becoming an important therapeutic strategy in several diseases.21,22 In the cardiovascular field, in order to avoid side effects and bleeding complications of antithrombotic drugs, modern drug discovery aims to develop targeted approaches that interfere with the contribution of platelets to pathological alterations of the vascular system while preserving their vascular protective functions.23,24 Within this context, it is imperative to deepen our understanding of the regulation of platelets in both health and disease, as redox-dependent regulation of platelets remains poorly understood.25 Our novel approach can help to clarify redox-dependent mechanisms regulating platelets and hemostasis, validate new drug discovery targets, and identify novel antiplatelet drug candidates. In this study, we utilized the EPR/turbidimetry technique to clarify the dynamics of the generation and activation of oxygen radicals in human platelets in response to physiological and pathological stimuli. The use of NOX1- and NOX2-selective peptide inhibitors allowed the identification of key differences in the involvement of these two enzymes in the response to platelet agonists and modulators. The application of this technique will further our understanding of redox-dependent platelet regulation and may have important consequences for antiplatelet drug discovery, where the quest for truly pathway-specific inhibitors targeting pathological platelet activation without interfering with their physiological hemostatic function remains an unmet objective.

Methods Platelet preparation Human blood was drawn from healthy volunteers by median cubital vein venepuncture following Royal Devon and Exeter NHS Foundation Trust Code of Ethics and Research Conduct and under National Research Ethics Service South West – Central Bristol approval (Rec. n. 14/SW/1089). Sodium citrate was used as anticoagulant (0.5% w/v). Platelet rich plasma (PRP) was separated from 1880

whole blood by centrifugation [250xg, 17 minutes (min)], and platelets were separated from PRP by a second centrifugation step (500xg, 10 min), in the presence of prostaglandin E1 (PGE1, 40 ng/mL) and indomethacin (10 μM). For mouse platelets, blood was taken with intracardiac puncture from 12-week old females under the Home Office license PPL30/3348 and anticoagulated with 0.5% w/v citrate. PRP was separated from whole blood by centrifugation (160xg, 15 min), and platelets were separated from PRP by a second centrifugation step (600xg, 10 min), in the presence of prostaglandin E1 (PGE1, 40 ng/mL) and indomethacin (10 μM).

Electron paramagnetic resonance/turbidimetry assay 2x108 platelets/mL were prepared as described above. Before adding stimuli, 200 μM CMH or PPH was added to platelets in the presence of 25 μM deferroxamine and 5 μM diethyldithiocarbamate (DETC). Platelet suspensions were loaded onto a Chronolog 700-2 aggregometer with continuous stirring and the turbidimetry readings were immediately started. 50 μL of platelet-free supernatant were transferred into the Hirschmann precision micropipettes and read using an e-scan (Noxygen, Germany).

Thrombus formation under physiological flow assay The Bioflux200 system (Fluxion, South San Francisco, CA, USA) was used to analyze thrombus formation in human and mouse whole blood under flow. Heparin-anticoagulated whole blood was anticoagulated with 5 μ/mL heparin plus 40 μM D-Phenylalanyl-prolyl-arginyl Chloromethyl Ketone or PPACK and incubated with scrambled or the NOX inhibitory peptides, NoxA1ds and Nox2ds-tat before the addition of 1 μM 3,3'-dihexyloxacarbocyanine iodide (DiOC6) for 10 min before the blood was added to the wells. Thrombus formation was visualized by fluorescence microscopy at a shear rate of 1000 sec−1. A more detailed description of the methods used is available in the Online Supplementary Appendix.

Results Superoxide anion-dependence of platelet activation by collagen but not thrombin Generation of ROS in a living cell can be examined by EPR spectroscopy using 1-hydroxy-3-methoxycarbonyl2,2,5,5-tetramethylpyrrolidine (CMH), an oxygen radicalspecific spin probe.16 This spin probe crosses the plasma membrane and directly reacts with intracellular oxygen radicals to generate a nitroxide adduct that can be detected by EPR spectroscopy (Online Supplementary Figure S1A and B). We combined classical aggregometry (also known as turbidimetry) with CMH-dependent and EPR spectroscopy by analyzing CMH oxidation in the platelet supernatant while the aggregation reaction is taking place (Online Supplementary Figure S1C). EPR is considered a gold standard for ROS detection and offers the important advantage of providing a quantification of the generation rate of oxygen radicals. It is in fact possible to build a calibration curve using known concentrations of the nitroxide adduct (CM●) (Online Supplementary Figure S2A and B). Once platelet suspension density and incubation time is known, it is possible to interpolate experimental data of resonance intensity to determine CMH oxidation rates (moles per platelet per min; see formula in Online Supplementary Figure S2C). This assay allowed us to correlate platelet aggregation induced by collagen and thrombin with the rate of oxygen radical generation (measured as rate of CMH oxidation). Collagen haematologica | 2019; 104(9)

Redox-dependent regulation of platelets

was tested at concentrations ranging from 1 to 30 µg/mL and both aggregation kinetics (Figure 1A) and oxygen radical generation rates at 10 min from stimulus delivery (Figure 1C) were concentration dependent. Similarly to collagen, a synthetic collagen-related peptide (CRP) that selectively engages GPVI receptor (but not other collagen-binding receptors on platelets), also led to concentration-dependent O2•− formation that was necessary for platelet aggregation

(Online Supplementary Figure S3). Thrombin also showed concentration dependence of both EPR (Figure 1B) and aggregation (Figure 1D) responses between 0.03 and 1 unit/mL. Online Supplementary Table S1 shows the EC50 values of these three agonists for EPR detection of superoxide anion generation and aggregation response. In order to confirm the chemical nature of the oxygen radicals generated by platelets upon activation and detected

Figure 1. Oxygen radical generation by platelets activated by physiological stimuli collagen and thrombin. Washed human platelets were prepared as described. Collagen was tested at concentrations ranging from 1 to 30 μg/mL (A and C), while thrombin was tested from 0.03 to 1 unit/mL (B and D). Electron paramagnetic resonance (EPR) was used to measure oxygen radicals formation as 1hydroxy-3-methoxycarbonyl-2,2,5,5-tetramethylpyrrolidine (CMH) oxidation rates [32.7 and 37.3 attomoles per platelet per minute (min), respectively] (A and C) and aggregation was simultaneously assessed by turbidimetry (B and D). For either technique, representative examples are shown in top panels, while quantification is shown in the bottom panels. Aggregation curves up to 5 min are shown, while EPR resonance readings were taken after 10 min of stimulation. Examples of EPR traces and aggregation curves are representative of 4 independent experiments. Statistical analysis was performed by one-way ANOVA with Bonferroni post-hoc test. *P<0.05 compared to resting platelets. N=4 for (A-D).

haematologica | 2019; 104(9)

1881

D. Vara et al.

by EPR, we utilized cell-permeant pegylated superoxide dismutase (PEG-SOD), which scavenges O2•− by dismuting it into H202. The EPR signal detected in human platelets stimulated by collagen and thrombin was abolished by PEG-SOD, suggesting that O2•− is the oxygen radical species generated under these conditions (Online Supplementary Figure S4A and C, respectively). The generation of O2•− radicals appeared necessary for platelet activation by collagen, as the scavenging of this oxygen radical by PEG-SOD significantly inhibited aggregation in response to this agonist (Online Supplementary Figure S4B). On the other hand, thrombin-dependent aggregation was not affected by PEG-SOD, suggesting that O2•− is not an essential component of the signaling cascades induced by this agonist. Similarly to collagen, both EPR and aggregation signals in response to CRP were abolished by PEGSOD (Online Supplementary Figure S3C and D). The data obtained with PEG-SOD were confirmed using another O2•− scavenger, 4-hydroxy-2,2,6,6-tetramethylpiperidin-1oxyl (or TEMPOL). TEMPOL abolishes the EPR signal in response to either collagen or thrombin (Online Supplementary Figure S5A and C, respectively), while only the aggregation response to collagen but not thrombin (Online Supplementary Figure S5C and D, respectively) is inhibited. Interestingly, the non-selective antioxidant and cell permeable amino acid N-acetyl cysteine (NAC) inhibits aggregation induced by both collagen and thrombin, suggesting that the thrombin responses depend on oxidative reactions, but that O2•−, per se, is not the oxidant species involved in this response (Online Supplementary Figure S5C and D). In order to confirm that the CMH-based EPR did not detect H202 (another major ROS) in our hands, we performed experiments in the presence of catalase (CAT, noncell permeable) and pegylated catalase (PEG-CAT, cell permeable), which convert H202 to water (Online Supplementary Figure S6). No effect of either enzymes on the collagen- or thrombin-induced EPR signal was observed, proving the specificity of the CMH-based EPR measurements for oxygen radicals and O2•− in particular.

Low levels of superoxide anion are released by platelets although they are not required for platelet activation We demonstrated that the CMH-based EPR spectroscopy predominantly measures intracellular but not extracellular platelet O2•− in response to either collagen or thrombin (Online Supplementary Figure S7), as the non-pegylated and thus non-cell permeable version of the enzyme superoxide dismutase (SOD) did not affect the responses. Interestingly, we also utilized a non-cell permeable EPR probe called 1Hydroxy-4-phosphono-oxy-2,2,6,6-tetramethylpiperidine (or PPH) to detect extracellular O2•− released by platelets (Online Supplementary Figure S8A-D). Noticeably, the scavenging of extracellular O2•− with SOD was effective in abolishing the PPH-based signal (Online Supplementary Figure S8E and F), but did not affect collagen- or thrombin-induced aggregation (Online Supplementary Figure S8G and H), suggesting that extracellular O2- does not participate in the process of platelet activation. In addition, we confirmed the formation of H202 in response to collagen or thrombin using the H2O2-specific probe Amplex Red (Online Supplementary Figure S9A and B) and that catalase effectively quenched the signal. More importantly, the degradation of H202 by catalase or pegylated catalase inhibited platelet aggregation stimulated by thrombin, but not collagen (Online Supplementary Figure S9C-F). This suggests that H202 is a crit1882

ical component of the signaling triggered by thrombin but not collagen.

Differential role of NADPH oxidases 1 and 2 in collagen- and thrombin-dependent activation of human platelets We utilized the combinatorial EPR/aggregometry that we developed to assess the role of NOX1 and NOX2 in platelet activation by collagen and thrombin. In experiments illustrated in Figure 2, we used the NOX1-specific inhibitory peptide NoxA1ds to assess the role of this enzyme in platelet activation.26 NOX1 inhibition almost completely inhibited oxygen radical formation in response to collagen (Figure 2A), but not thrombin (Figure 2B). In parallel, collagen- (Figure 2C) but not thrombin-dependent (Figure 2D) aggregation was inhibited by NoxA1ds. Importantly a scrambled version of the peptide was used as control. Taken together, these data suggest that NOX1 is activated and participates in the signaling of platelet activation in response to collagen but not thrombin. We also tested the role of NOX2 using the specific inhibitory peptide Nox2ds-tat.27,28 Both collagen- and thrombindependent oxygen radical formation are significantly impaired by NOX2 inhibition (Figure 3A and B, respectively). Interestingly, although the inhibition of the thrombin response by Nox2ds-tat reduces oxygen radical formation to basal levels, the inhibition of collagen-induced oxygen radical levels is only partial (i.e. in the presence of Nox2ds-tat the oxygen radical levels induced by collagen are significantly higher than resting controls). This is reflected in the aggregation data, which show complete inhibition by Nox2ds-tat of the aggregation induced by thrombin but only marginal inhibition of the response to collagen (Figure 3D and C, respectively). This is consistent with NOX1 playing a larger role than NOX2 in collageninduced platelet aggregation. The data obtained with collagen were essentially confirmed using the synthetic ligand for GPVI CRP (Online Supplementary Figure S10), which suggests that GPVI is the key receptor linking collagen-dependent platelet activation to oxygen radical generation and determines the redox patterns triggered by collagen in platelets. The active engagement of NOX1 and NOX2 in collagen and thrombin signaling was confirmed by co-immunoprecipitation of these two core subunits with respective essential cytosolic components of each complex. NOX1 is co-immunoprecipitated with its canonical activating subunit NOXA129 in the presence of collagen, which is consistent with NOX1 being post-translationally activated in response to collagen (Figure 4A). NOX2 appears to be co-immunoprecipitated with its canonical organizing subunit p47phox in the presence of thrombin and weakly in the presence of collagen (Figure 4B). This aligns with the conclusions reached using our EPR and aggregation experiments, i.e. that collagen activates a primarily NOX1-dependent response, while thrombin activates primarily NOX2. Accordingly, whole blood thrombus formation experiments on collagen showed that NOX1 inhibition with NoxA1ds abolishes thrombus formation (Figure 4C), while NOX2 inhibition by Nox2ds-tat induces only a marginal inhibition (Figure 4D). Experiments in platelets from wild-type, NOX1-/- or NOX2-/- mice confirmed the centrality of NOX1 for collagen signaling with marginal involvement of NOX2 in the aggregation response (Figure 5A and B), while NOX2 is critical for thrombin signaling (Figure 5C). haematologica | 2019; 104(9)

Redox-dependent regulation of platelets

Role of NOX1 and NOX2 in mediating the effect on platelets of weak agonists/positive modulators oxidized LDL (oxLDL) and amyloid beta (Aβ1-42) Using these tools (NoxA1ds26 and Nox2ds-tat27,28), we determined that NOX1 or NOX2 play an equivalent role in the generation of superoxide anion in response to oxLDL (Figure 6A) and that both enzymes are required for the stimulation of the modest aggregation induced by this modulator (Figure 6B). As suggested by the inhibition of aggregation by PEG-SOD, O2•− is the oxidant molecule required for the activation of platelets by oxLDL. It is important to note how oxLDL elicits a very modest aggregation (both as size and kinetic of the response), which conforms to the designation of oxLDL as a modulator rather

haematologica | 2019; 104(9)

than agonist (i.e. exerting its effect by enhancing the responsiveness of platelets to low agonist levels). In fact, low concentrations of collagen (Figure 6C) or thrombin (Figure 6D) characterized by the ability to induce no/low aggregation, resulted able to induce a robust platelet aggregation in the presence of ox-LDL. The synergistic effect of oxLDL in combination with thrombin or collagen was inhibited by NoxA1ds (Figure 6C) or Nox2ds-tat (Figure 6D), respectively. We could not test NoxA1ds on ox-LDL + collagen or Nox2ds-tat on oxLDL + thrombin, because, as shown above, NoxA1ds inhibits collagen directly and Nox2ds-tat inhibits thrombin directly. Similarly, Aβ1-42 induces O2•− formation via activation of NOX1 and NOX2 (Figure 7A), which leads to a modest platelet aggregation

Figure 2. NOX1 is specifically involved in platelet activation by collagen but not thrombin. 1-hydroxy-3-methoxycarbonyl2,2,5,5-tetramethylpyrrolidine (CMH) was utilized for the detection of oxygen radicals generated by platelets. 10 μg/mL collagen (A) or 0.1 unit/mL thrombin (B) were tested. 10 μM of NoxA1ds abolished the electron paramagnetic resonance (EPR) response measured in the presence of collagen. The scrambled peptide at the same concentration (scNoxA1ds) was used as a negative control. Interestingly, the inhibition of NOX1 by NoxA1ds also inhibited collagen-dependent (C), but not thrombin-dependent (D) platelet aggregation. Aggregation curves for up to 5 minutes (min) are shown, while EPR resonance readings were taken after 10 min of stimulation. Examples of EPR traces and aggregation curves are representative of 3 or more independent experiments. Statistical analysis was performed by one-way ANOVA with Bonferroni post-hoc test for EPR. *P<0.05 compared to resting platelets or t-test for aggregation. *P<0.05 compared to scrambled control. N=3 for (A-D).

1883

D. Vara et al.

(Figure 7B). On the other hand, similarly to oxLDL, Aβ1-42 displays the ability to synergistically increase the aggregation response to low concentrations of collagen (Figure 7C) or thrombin (Figure 7D). As proved with inhibitory peptides NoxA1ds26 and Nox2ds-tat,27,28 the synergistic effect on collagen- or thrombin-induced aggregation is NOX1- or NOX2-dependent, respectively. We could not test NoxA1ds on Aβ1-42 + collagen or Nox2ds-tat on Aβ1-42 + thrombin, because, as shown above, NoxA1ds inhibits collagen directly and Nox2ds-tat inhibits thrombin directly. Transgenic mice NOX1-/- and NOX2-/- were utilized to assess the NOXdependence of the responses to oxLDL and Aβ1-42. As oxLDL coating of surfaces is not commonly used and there is no accepted protocol for this procedure, we tested the

1884

effect of oxLDL added to mouse blood on thrombus formation on low levels (i.e. 0.05 mg/mL) of collagen coating (Figure 8A). These data showed that oxLDL potentiates the thrombus formation on collagen in a NOX2-dependent manner (as it was not evident in NOX2-/- blood). NOX1 ablation inhibits collagen responses directly, therefore we could not investigate the role of NOX1 in the potentiation of responses to this agonist by oxLDL. The ablation of either NOX1 or NOX2 silencing significantly also impairs thrombus formation at physiological arterial shear on absorbed Aβ1-42 (1,000 sec-1) (Figure 8C). At low shear (200 sec-1), mouse platelets display low levels of adhesion to absorbed Aβ1-42 without formation of thrombi, which was inhibited by NOX1 genetic silencing but unaffected by

Figure 3. NOX2 is activated by both collagen and thrombin, but essential only for platelet aggregation induced by thrombin. 1hydroxy -3-methoxycarbonyl2,2,5,5-tetramethylpyrrolidine (CMH) was utilized for the detection of oxygen radicals generated by platelets. 10 μg/mL collagen (A) or 0.1 unit/mL thrombin (B) were tested. 10 μM of Nox2ds-tat inhibited the electron paramagnetic resonance (EPR) response measured in the presence of either collagen or thrombin, although the collagen-dependent response remained significantly higher than resting levels of oxygen radical formation. The scrambled peptide at the same concentration (scNox2ds-tat) was used as a negative control. Interestingly, the inhibition of NOX2 by Nox2ds-tat also inhibited thrombin-dependent (D), but not collagen-dependent (C) platelet aggregation. Examples of EPR traces and aggregation curves are representative of 3 or more independent experiments. Statistical analysis was performed by one-way ANOVA with Bonferroni post-hoc test for EPR. *P<0.05 compared to resting platelets or t-test for aggregation. *P<0.05 compared to scrambled control. N=4 for (B and D), n=5 for (A), and n=7 for (C).

haematologica | 2019; 104(9)

Redox-dependent regulation of platelets

NOX2 ablation (Online Supplementary Figure S11). As these modulators induce little or no aggregation on their own, we analyzed the modulatory effect of oxLDL and Aβ1-42 in aggregation experiments using NOX1-/- and NOX2-/- mouse platelets. We showed that oxLDL (Figure 8B) and Aβ1-42

(Figure 8D) potentiate the aggregation induced by collagen in wild-type mice but not in NOX2-/- mice, and the aggregation stimulated by thrombin in wild-type mice but not in NOX1-/- mice. The potentiation of collagen response could not be tested in NOX1-/- mice, which do not respond to this

Figure 4. NOX1 is the main source of superoxide anions in human platelet response to collagen, while NOX2 is the main source of thrombin-dependent reactive oxygen species (ROS). The activation of NOX1 (A) and NOX2 (B) was assessed by co-immunoprecipitation with their canonical activating and organizing subunits NOXA1 and p47phox, respectively. 400 μL of platelet suspension (4x108 platelets/mL) were stimulated with 10 μg/mL collagen or 0.1 unit/mL thrombin or vehicle solution (Tyrode’s buffer) for 10 minutes (min) before gentle cell lysis (NP40 buffer). Specific NOX1 or NOX2 antibodies and Protein A/G were used to immunoprecipitate the NOX complexes (which by extension should include their regulatory subunits after activation). The immunoprecipitates were tested by immunoblotting using NOX1, NOXA1, NOX2 or p47phox antibodies (as indicated). The data are representative of 4 independent experiments. The functional role of NOX1 and NOX2 in collagen-dependent platelet activation was assessed in a whole blood flow assay (C and D). Platelets were stained with DiOC6 as described and the Bioflux platform (Fluxion, San Francisco, CA, USA) was utilized to assess the thrombus formation induced by collagen under physiological flow (1000 sec-1). The experiments were performed in the presence of NoxA1ds or its negative scrambled control (C) or Nox2ds-tat or its negative scrambled control (D). Images were taken at 10 min of flow and are representative of 4 independent experiments. They were quantified by assessing the surface area coverage by platelets (C and D, bottom). Statistical significance was tested by t-test. *P<0.05 compared to scrambled control. N=4 for (C and D).

haematologica | 2019; 104(9)

1885

D. Vara et al.

stimulus, while the potentiation of thrombin responses could not be tested in NOX2-/-, for the same reason.

Discussion The study of platelet function is critical to understanding vascular homeostasis and disease. A thorough understanding of the redox-dependent regulation of platelet function has been hampered by the lack of a reliable technique to measure intracellular ROS.16,17 We have resolved this problem by optimizing an EPR-based technique for the detection of oxygen radicals and combining it with a classical turbidimetric assay for the simultaneous measurement of platelet aggregation. Although EPR has previously been used for the analysis of ROS formation in live cells;16,17 this technique has never been applied to the study of platelet redox signaling. We utilized the assay developed here to

clarify several unknown aspects of the regulation of platelet activity by endogenous oxidants. The molecular mechanisms underlying the redox dependence of platelet activation in response to collagen, thrombin, and oxidized LDL are summarized in Online Supplementary Figure S12. Although, the dependence of platelet activation on the generation of endogenous ROS has been described,1-6 here for the first time we elucidated the chemical nature of the ROS involved in the signaling of different platelet agonists and modulators. O2•− are generated in response to all tested agonists and modulators, but although these ROS are directly involved in the signaling of collagen, oxLDL and Aβ1-42, their dismutation to hydrogen peroxide is necessary for the signaling of thrombin. The literature on this aspect of platelet biology is quite inconclusive because of the use of different techniques and conditions. Although previous studies reported the generation of hydrogen peroxide in response to thrombin (and leading to

1886

Figure 5. NOX1- and NOX2-dependence of collagen and thrombin aggregation and superoxide generation tested in transgenic mice. Platelets were isolated for wild-type (C57BL6/J), NOX1-/- or NOX2-/- mice exsanguinated via intracardiac puncture and resuspended at 2x108 platelets/mL density. Platelets were stimulated with 3 μg/mL collagen (A) or 0.1 unit/mL thrombin (C). Aggregation (left) and superoxide anion formation (right) were measured as described for 5 and 10 minutes (min), respectively. (B) The functional role of NOX1 and NOX2 in collagen-dependent platelet activation was also assessed in a whole blood flow assay. Platelets were stained with DiOC6 and the Bioflux platform (Fluxion, San Francisco, CA, USA) was utilized to assess the thrombus formation induced by collagen under physiological flow (1,000 sec-1). Images were taken at 10 min of flow and are representative of 4 independent experiments. They were quantified by assessing the surface area coverage by platelets with Image J. Data are representative of 4 independent experiments. Statistical analysis was performed by one-way ANOVA with Bonferroni post-hoc test. *P<0.05. N=4 for (A-C).

haematologica | 2019; 104(9)

Redox-dependent regulation of platelets

apoptosis),30 this study for the first time highlights a significant difference in the role of hydrogen peroxide in the responses to collagen and thrombin. Our observations on the link between thrombin-dependent activation and hydrogen peroxide contradict previous reports pointing to a role for hydrogen peroxide in collagen but not thrombin signaling.31 The experimental differences in our and previ-

ous studies are extensive and the poor specificity of the tools used for older studies (e.g. 2',7'–dichlorofluorescein diacetate or DCFDA) are potentially responsible for these discrepancies. Another important addition to our understanding of redox regulation of platelets is the clarification of whether oxidants act intracellularly or extracellularly. Our data clear-

Figure 6. NOX1 and NOX2 are required for the induction of superoxide anion formation and platelets activation by oxidized low density lipoprotein (oxLDL). Oxygen radical generation (A) and platelet aggregation (B) in response to oxLDL were measured as described. 10 μM of scrambled NoxA1ds (scNoxA1ds), NoxA1ds, scrambled Nox2ds-tat (scNox2ds-tat), Nox2ds-tat or 100 unit/mL of PEG-SOD were pre-incubated 10 minutes (min) before platelet stimulation with 50 ng/mL oxLDL. Examples of electron paramagnetic resonance (EPR) traces and aggregation curves are representative of 4 independent experiments. Statistical analysis was performed by one-way ANOVA with Bonferroni post-hoc test. *P<0.05. N=3 for (B) and n=4 for (A). The effect of oxLDL as platelet modulator was investigated by traditional aggregometry (C and D). Pre-incubation with 50 ng/mL oxLDL for 10 min was followed by stimulation with 3 μg/mL collagen (C) or 0.03 unit/mL thrombin (D). In order to test the dependence of the modulatory effect of oxLDL on NOX1 and NOX2, 10 μM scrambled NoxA1ds (scNoxA1ds), NoxA1ds, scrambled Nox2ds-tat (scNox2ds-tat), Nox2ds-tat were pre-incubated (10 min before addition of oxLDL). Aggregation curves are representative of 4 independent experiments.

haematologica | 2019; 104(9)

1887

D. Vara et al. ly suggest that platelet O2•− acts intracellularly in these responses, as only cell permeable scavenger can affect their activity (i.e. PEG-SOD), while hydrogen peroxide is likely to be formed by dismutation of intracellular O2•− and released extracellularly, where it potentiates platelet responses induced by thrombin (as proved by the effect of non-cell permeable catalase shown in Online Supplementary

Figure S6). These findings add significantly to our understanding of platelet redox regulation and are in agreement with previous suggestions of an intracellular function for O2•− .32,33 It is also in agreement with the observation of a positive regulatory role for extracellular hydrogen peroxide in thrombin-induced responses34 and a role for extracellular oxidants in the regulation of platelet surface receptor func-

Figure 7. NOX1 and NOX2 are required for the induction of superoxide anion formation and platelet activation by amyloid peptide β 1-42. Oxygen radical generation (A) and platelet aggregation (B) in response to amyloid peptide β 1-42 (Aβ1-42) were measured as described. 10 μM of scrambled NoxA1ds (scNoxA1ds), NoxA1ds, scrambled Nox2ds-tat (scNox2ds-tat), Nox2ds-tat or 100 unit/mL of PEG-SOD were pre-incubated 10 minutes (min) before platelet stimulation with 20 μM Aβ1-42. Examples of electron paramagnetic resonance (EPR) traces and aggregation curves are representative of 4 independent experiments. Statistical analysis was performed by one-way ANOVA with Bonferroni post-hoc test. *P<0.05. N=3 for (A and B). The effect of Aβ1-42 as platelet modulator was investigated by traditional aggregometry (C and D). Pre-incubation with 20 μM Aβ1-42 for 10 min was followed by stimulation with 3 μg/mL collagen (C) or 0.03 unit/mL thrombin (D). In order to test the dependence of the modulatory effect of Aβ1-42 on NOX1 and NOX2, 10 μM scrambled NoxA1ds (scNoxA1ds), NoxA1ds, scrambled Nox2ds-tat (scNox2dstat), Nox2ds-tat were pre-incubated (10 min before addition of Aβ1-42). Aggregation curves are representative of 4 independent experiments.

1888

haematologica | 2019; 104(9)

Redox-dependent regulation of platelets

tion by protein disulphide isomerases (PDI).35 Regarding the redox-dependence of the collagen signaling in platelets, although further studies are required, protein tyrosine phosphatases (PTP) are the most likely link between redox and conventional signaling in platelets.36 The oxidative inactivation of protein phosphatases has in fact been suggested to play a key role in the activation and regulation of the pathophysiological roles of platelets.37 The Src Homology Phosphatase 2 (SHP2) has been shown to play an important role as a negative regulator of platelet activation,38 and

recent studies demonstrated that ROS generated during platelet activation oxidize and inhibit SHP2. This, in turns, facilitates the activation of protein kinase-mediated signaling pathways and drives the processes associated with cell activation (e.g. adhesion receptor activation, shape change, etc.).37,39 In this study, we also highlight the differential involvement of NOX1 and NOX2 in physiological agonist signaling (i.e. collagen and thrombin). NOX1 is essential for the signaling of collagen with NOX2 inhibition only partially

haematologica | 2019; 104(9)

Figure 8. Experiments on transgenic mice suggest that NOX1 and NOX2 are required for the effect of amyloid peptide β 142 and oxidized low density lipoprotein (oxLDL). Platelets from wild-type, NOX1-/- or NOX2-/mice were stained with DiOC6 and the Bioflux platform (Fluxion, San Francisco, CA, USA) was utilized to assess the thrombus formation induced by collagen under physiological flow. (A) Ibidi Vena8+ flow chambers were coated with 0.05 mg/mL collagen and whole blood was treated with 50 ng/mL native LDL (nLDL) or oxLDL. (C) Ibidi Vena8+ flow chambers were coated with 20 μM Aβ1-42 or scrambled control peptide (ScAβ1-42). The shear rate utilized was 1,000 sec-1, which leads to thrombus formation. Images were taken at 10 minutes (min) of flow and are representative of 4 independent experiments. Images were quantified by assessing the surface area coverage by platelets with Image J. Data are representative of 4 independent experiments. Statistical analysis was performed by oneway ANOVA with Bonferroni post-hoc test. *P<0.05. N=4 for (A and C). Aggregation experiments were performed by pretreating platelets with 50 ng/mL native LDL (nLDL) or oxLDL (B) or 20 μM Aβ1-42 or scrambled control peptide (ScAβ1-42) (D) for 10 min. Low level aggregation was then stimulated with either 10 μg/mL collagen or 0.03 unit/mL thrombin, as indicated. Aggregation data are representative of 3 independent experiments.

1889

D. Vara et al. reducing collagen-dependent aggregation and O2•− formation, in both human (Figure 3A and B) and mouse (Figure 5A) platelets. The secondary role of NOX2 in collagen responses is further demonstrated by thrombus formation experiments under physiological flow, which is marginally but significantly reduced by NOX2 inhibition/silencing in human but not mouse blood (compare Figure 4D and Figure 5B). This discrepancy between thrombus formation in human and mouse blood could be suggestive of a higher involvement of NOX2 in collagen responses of human platelets. In any case, the dominant role of NOX1 in collagen response is clear in our experiments both in human and mouse platelets. Similarly, NOX2 is essential for the signaling of thrombin without NOX1 involvement in human platelets (Figure 2C and D) and limited yet significant involvement in mouse platelets (Figure 5C). This is in agreement with previous indications from our and other groups,14,40 but in essential disagreement with a recent study by Delaney et al. in mouse platelets reaching opposite conclusions (i.e. NOX1 involvement in thrombin signaling and NOX2 involvement in collagen signaling).13 Some differences can be expected between signaling pathways and redox dependence of human and mouse platelets, which may explain this discrepancy. Overall, combining human and mouse platelet data, it is safe to state that thrombindependent aggregation depends very heavily on NOX2 activity (although a role of NOX1 in thrombin was detected in mouse platelets, which we did not observe in human platelets), while collagen-induced aggregation was predominantly NOX1-dependent with marginal involvement of NOX2 (as shown in our human platelet data). So, although mouse platelets display some engagement of NOX1 in thrombin responses, species-specific differences cannot fully explain the discrepancy between our report and Delaney et al.’s work. The fact that Delaney et al. used male animals for NOX1 studies and females for NOX2 studies may explain some of the differences with our study (performed entirely on female animals). Other potentially important differences are in the platelet isolation procedure, the concentration of agonists (very low concentration of thrombin used), and the use of collagen-related peptide instead of collagen. In addition to physiological stimuli, in this study we analyzed the effect of oxLDL and Aβ1-42, platelet modulators associated with the thrombotic complications of atherosclerosis41 and cerebrovascular amyloid angiopathy (CAA),42 respectively. Both oxLDL and Aβ1-42 have been shown to activate platelets and act as positive modulators.15,43 We confirmed the ability of these molecules to induce partial platelet aggregation. On the other hand, they were able to significantly increase the responses to low levels of physiological agonists. This mode of action is consistent with the definition of positive modulators or primers,44 which are characterized by the ability to trigger an unwanted hemostatic response and clot formation leading to thrombosis. The involvement of platelet positive modulators in thrombotic complications associated with diseases is particularly important for vascular health and relative pharmacotherapy.45 Interestingly, the aggregation induced by these primers was also redox-dependent and inhibited by the O2•− scavenger PEG-SOD. This is in agreement with previous literature on oxLDL15,46 and Aβ1-42,47 and supports the hypothesis that platelet primers may act in a redox-dependent manner.48 1890

Also intriguing were our conclusions regarding the involvement of NOX1 and NOX2 in the signaling of oxLDL and Aβ1-42. We provide compelling evidence in human platelets with NOX-selective inhibitors Nox2dstat and NoxA1ds or in genetically modified mouse platelets (NOX1-/- or NOX2-/-) that both NOX1 and NOX2 are activated by oxLDL and Aβ1-42 and that they are both required for the functional effects of these pathology-associated modulators on platelets. These conclusions were confirmed by experiments with transgenic mouse platelets, both in thrombus adhesion and aggregation experiments. Only adhesion to Aβ1-42 under low shear seemed exclusively NOX1-dependent. This is a low level adhesion response incapable of properly triggering thrombus formation (Online Supplementary Figure S11). These data may, therefore, suggest a differential involvement of NOX1 and NOX2 in different molecular events triggered by Aβ1-42. This hypothesis merits further study for satisfactory elucidation. The involvement of NOX2 in the signaling of oxLDL has been suggested previously,15,46 while the involvement of NOX1 is novel. The possibility of abolishing the effect of oxLDL and Aβ1-42 on platelets by inhibiting only one of the two NOX may suggest that both enzymes are required for reaching a threshold in the superoxide anion levels leading to platelet stimulation. The similarities between the activity and redox-dependence of oxLDL and Aβ1-42 may suggest that they act on similar receptors. As suggested by literature for both molecules, the receptors for these platelet modulators are likely to be CD36.15,49 Importantly, the fact that both platelet NOX are required for the modulatory activity of oxLDL or Aβ1-42 offers the opportunity of targeted intervention without the complete inhibition of the response to physiological agonists. In other words, the inhibition of only one NOX isozyme could reduce the pro-thrombotic tendencies associated with vascular inflammation without completely impairing the hemostatic response (i.e. NOX1 inhibition will not impair thrombin response, while NOX2 inhibition will not impair collagen responses). This is the ultimate goal of modern antiplatelet drug development and may help to resolve the persisting problem of bleeding risks associated with all existing antiplatelet treatment.50 In summary, herein we describe the development and application of a novel approach to simultaneously monitor the generation of oxygen radicals and platelet aggregation. This technique has the potential to become a standard technique to assess platelet responsiveness and thrombotic risks associated to vascular conditions in a clinical setting. In addition, this study allowed the identification of patterns of platelet regulation and the clarification of the mode of action of physiological agonists and pathological modulators of platelets. The results of this study also highlight the potential of NADPH oxidase targeting for the development of novel antiplatelet drugs with better pharmacodynamic profiles (i.e. limited bleeding side effects). Acknowledgments The authors would like to thank the British Heart Foundation for sponsoring this research (PG/15/40/31522), the Clinical Research Facility (CRF) of the University of Exeter and in particular Dr Bridget Knight for blood collections and Dr Bruno Fink from Noxygen Science Transfer & Diagnostics GmbH for the technical support. haematologica | 2019; 104(9)

Redox-dependent regulation of platelets

References 1. Chaudhari K, Hamad B, Syed BA. Antithrombotic drugs market. Nat Rev Drug Discov. 2014;13(8):571-572. 2. Stegner D, Nieswandt B. Platelet receptor signaling in thrombus formation. J Mol Med (Berl). 2011;89(2):109-121. 3. Iuliano L, Colavita AR, Leo R, Pratico D, Violi F. Oxygen free radicals and platelet activation. Free Radic Biol Med. 1997;22(6):999-1006. 4. Watt J, Ewart M-A, Greig FH, Oldroyd KG, Wadsworth RM, Kennedy S. The effect of reactive oxygen species on whole blood aggregation and the endothelial cellplatelet interaction in patients with coronary heart disease. Thromb Res. 2012;130(2):210-215. 5. Freedman JE. Oxidative stress and platelets. Arterioscler Thromb Vasc Biol. 2008;28(3):s11-16. 6. Ferroni P, Vazzana N, Riondino S, Cuccurullo C, Guadagni F, Davi G. Platelet function in health and disease: from molecular mechanisms, redox considerations to novel therapeutic opportunities. Antioxid Redox Signal. 2012;17(10):1447-1485. 7. Krotz F, Sohn HY, Pohl U. Reactive oxygen species: players in the platelet game. Arterioscler Thromb Vasc Biol. 2004;24(11):1988-1996. 8. Chondrogianni N, Petropoulos I, Grimm S, et al. Protein damage, repair and proteolysis. Mol Aspects Med. 2014;35:1-71. 9. Violi F, Pignatelli P. Platelet NOX, a novel target for anti-thrombotic treatment. Thromb Haemost. 2014;111(5):817-823. 10. Bartimoccia S, Carnevale R, Sanguigni V, et al. NOX 5 is expressed in platelets from patients with chronic granulomatous disease. Thromb Haemost. 2016;116(1):198200. 11. Krotz F, Sohn HY, Gloe T, et al. NAD(P)H oxidase-dependent platelet superoxide anion release increases platelet recruitment. Blood. 2002;100(3):917-924. 12. Begonja AJ, Gambaryan S, Geiger J, et al. Platelet NAD(P)H-oxidase-generated ROS production regulates alphaIIbbeta3-integrin activation independent of the NO/cGMP pathway. Blood. 2005; 106(8):2757-2760. 13. Delaney MK, Kim K, Estevez B, et al. Differential Roles of the NADPH-Oxidase 1 and 2 in Platelet Activation and Thrombosis. Arterioscler Thromb Vasc Biol. 2016;36(5):846-854. 14. Vara D, Campanella M, Pula G. The novel NOX inhibitor 2-acetylphenothiazine impairs collagen-dependent thrombus formation in a GPVI-dependent manner. Br J Pharmacol. 2013;168(1):212-224. 15. Magwenzi S, Woodward C, Wraith KS, et al. Oxidized LDL activates blood platelets through CD36/NOX2-mediated inhibition of the cGMP/protein kinase G signaling cascade. Blood. 2015;125(17):2693-2703. 16. Dikalov SI, Harrison DG. Methods for detection of mitochondrial and cellular reactive oxygen species. Antioxid Redox Signal. 2014;20(2):372-382. 17. Griendling KK, Touyz RM, Zweier JL, et al. Measurement of Reactive Oxygen Species, Reactive Nitrogen Species, and RedoxDependent Signaling in the Cardiovascular System: A Scientific Statement From the American Heart Association. Circ Res. 2016;119(5):e39-75. 18. Koltai K, Kesmarky G, Feher G, Tibold A,

haematologica | 2019; 104(9)

19.

20.

21.

22.

23. 24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

Toth K. Platelet Aggregometry Testing: Molecular Mechanisms, Techniques and Clinical Implications. Int J Mol Sci. 2017;18(8):1803. Davies MJ. Detection and characterisation of radicals using electron paramagnetic resonance (EPR) spin trapping and related methods. Methods. 2016;109:21-30. Hawkins CL, Davies MJ. Detection and characterisation of radicals in biological materials using EPR methodology. Biochim Biophys Acta. 2014;1840(2):708-721. Bedard K, Whitehouse S, Jaquet V. Challenges, Progresses, and Promises for Developing Future NADPH Oxidase Therapeutics. Antioxid Redox Signal. 2015;23(5):355-357. Diebold BA, Smith SM, Li Y, Lambeth JD. NOX2 As a Target for Drug Development: Indications, Possible Complications, and Progress. Antioxid Redox Signal. 2015;23(5):375-405. Pudusseri A, Shameem R, Spyropoulos AC. A new paradigm shift in antithrombotic therapy. Front Pharmacol. 2013;4:133. Wong PC, Seiffert D, Bird JE, et al. Blockade of protease-activated receptor-4 (PAR4) provides robust antithrombotic activity with low bleeding. Sci Transl Med. 2017;9(371). Pietraforte D, Vona R, Marchesi A, et al. Redox control of platelet functions in physiology and pathophysiology. Antioxid Redox Signal. 2014;21(1):177-193. Ranayhossaini DJ, Rodriguez AI, Sahoo S, et al. Selective recapitulation of conserved and nonconserved regions of putative NOXA1 protein activation domain confers isoform-specific inhibition of Nox1 oxidase and attenuation of endothelial cell migration. J Biol Chem. 2013;288(51):3643736450. Rey FE, Cifuentes ME, Kiarash A, Quinn MT, Pagano PJ. Novel competitive inhibitor of NAD(P)H oxidase assembly attenuates vascular O(2)(-) and systolic blood pressure in mice. Circ Res. 2001;89(5):408-414. Csanyi G, Cifuentes-Pagano E, Al Ghouleh I, et al. Nox2 B-loop peptide, Nox2ds, specifically inhibits the NADPH oxidase Nox2. Free Radic Biol Med. 2011;51(6):1116-1125. Valente AJ, Jamali AE, Epperson TK, Gamez MJ, Pearson DW, Clark RA. NOX1 NADPH oxidase regulation by the NOXA1 SH3 domain. Free Radic Biol Med. 2007;43(3):384-396. Lopez JJ, Salido GM, Gomez-Arteta E, Rosado JA, Pariente JA. Thrombin induces apoptotic events through the generation of reactive oxygen species in human platelets. J Thromb Haemost. 2007;5(6):1283-1291. Pignatelli P, Pulcinelli FM, Lenti L, Gazzaniga PP, Violi F. Hydrogen peroxide is involved in collagen-induced platelet activation. Blood. 1998;91(2):484-490. Qiao J, Arthur JF, Gardiner EE, Andrews RK, Zeng L, Xu K. Regulation of platelet activation and thrombus formation by reactive oxygen species. Redox Biol. 2018;14:126-130. Chlopicki S, Olszanecki R, Janiszewski M, Laurindo FRM, Panz T, Miedzobrodzki J. Functional Role of NADPH Oxidase in Activation of Platelets. Antioxid Redox Signal. 2004;6(4):691-698. Redondo PC, Jardin I, Hernandez-Cruz JM, Pariente JA, Salido GM, Rosado JA. Hydrogen peroxide and peroxynitrite enhance Ca2+ mobilization and aggregation in platelets from type 2 diabetic

35. 36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

patients. Biochem Biophys Res Commun. 2005;333(3):794-802. Furie B, Flaumenhaft R. Thiol isomerases in thrombus formation. Circ Res. 2014; 114(7):1162-1173. Jang JY, Min JH, Wang SB, et al. Resveratrol inhibits collagen-induced platelet stimulation through suppressing NADPH oxidase and oxidative inactivation of SH2 domaincontaining protein tyrosine phosphatase-2. Free Radic Biol Med. 2015;89:842-851. Jang JY, Min JH, Chae YH, et al. Reactive oxygen species play a critical role in collagen-induced platelet activation via SHP-2 oxidation. Antioxid Redox Signal. 2014;20(16):2528-2540. Mazharian A, Mori J, Wang YJ, et al. Megakaryocyte-specific deletion of the protein-tyrosine phosphatases Shp1 and Shp2 causes abnormal megakaryocyte development, platelet production, and function. Blood. 2013;121(20):4205-4220. Wang SB, Jang JY, Chae YH, et al. Kaempferol suppresses collagen-induced platelet activation by inhibiting NADPH oxidase and protecting SHP-2 from oxidative inactivation. Free Radic Biol Med. 2015;83:41-53. Walsh TG, Berndt MC, Carrim N, Cowman J, Kenny D, Metharom P. The role of Nox1 and Nox2 in GPVI-dependent platelet activation and thrombus formation. Redox Biol. 2014;2:178-186. Obermayer G, Afonyushkin T, Binder CJ. Oxidized low-density lipoprotein in inflammation-driven thrombosis. J Thromb Haemost. 2018;16(3):418-428. Jaunmuktane Z, Mead S, Ellis M, et al. Evidence for human transmission of amyloid-beta pathology and cerebral amyloid angiopathy. Nature. 2015;525(7568):247250. Donner L, Falker K, Gremer L, et al. Platelets contribute to amyloid-beta aggregation in cerebral vessels through integrin alphaIIbbeta3-induced outside-in signaling and clusterin release. Sci Signal. 2016; 9(429):ra52. Blair TA, Moore SF, Hers I. Circulating primers enhance platelet function and induce resistance to antiplatelet therapy. J Thromb Haemost. 2015;13(8):1479-1493. Gresele P, Falcinelli E, Momi S. Potentiation and priming of platelet activation: a potential target for antiplatelet therapy. Trends Pharmacol Sci. 2008;29(7):352-60. Carnevale R, Bartimoccia S, Nocella C, et al. LDL oxidation by platelets propagates platelet activation via an oxidative stressmediated mechanism. Atherosclerosis. 2014;237(1):108-116. Abubaker AA, Vara D, Eggleston I, Canobbio I, Pula G. A novel flow cytometry assay using dihydroethidium as redoxsensitive probe reveals NADPH oxidasedependent generation of superoxide anion in human platelets exposed to amyloid peptide beta. Platelets. 2017;5:1-9. Dayal S, Wilson KM, Motto DG, Miller FJ Jr., Chauhan AK, Lentz SR. Hydrogen peroxide promotes aging-related platelet hyperactivation and thrombosis. Circulation. 2013;127(12):1308-1316. Herczenik E, Bouma B, Korporaal SJ, et al. Activation of human platelets by misfolded proteins. Arterioscler Thromb Vasc Biol. 2007;27(7):1657-1665. McFadyen JD, Schaff M, Peter K. Current and future antiplatelet therapies: emphasis on preserving haemostasis. Nat Rev Cardiol. 2018;15(3):181-191.

1891

ARTICLE Ferrata Storti Foundation

Haematologica 2019 Volume 104(9):1892-1905

Hemostasis

Tspan18 is a novel regulator of the Ca2+ channel Orai1 and von Willebrand factor release in endothelial cells

Peter J. Noy,1 Rebecca L. Gavin,1 Dario Colombo,2 Elizabeth J. Haining,2 Jasmeet S. Reyat,1 Holly Payne,2 Ina Thielmann,3 Adam B. Lokman,2 Georgiana Neag,2 Jing Yang,1 Tammy Lloyd,1 Neale Harrison,1 Victoria L. Heath,2 Chris Gardiner,4 Katharine M. Whitworth,5 Joseph Robinson,5 Chek Z. Koo,1 Alessandro Di Maio,1 Paul Harrison,6,7 Steven P. Lee,5 Francesco Michelangeli,8 Neena Kalia,2,9 G. Ed Rainger,2 Bernhard Nieswandt,3 Alexander Brill,2,9,10 Steve P. Watson2,9 and Michael G. Tomlinson1,9

School of Biosciences, College of Life and Environmental Sciences, University of Birmingham, Birmingham, UK; 2Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Birmingham, UK; 3University Hospital Würzburg and Rudolf Virchow Center for Experimental Biomedicine, Würzburg, Germany; 4Department of Haematology, University College London, London, UK; 5 Institute of Immunology and Immunotherapy, Cancer Immunology and Immunotherapy Centre, University of Birmingham, Birmingham, UK; 6Scar Free Foundation for Burns Research, Queen Elizabeth Hospital Birmingham, University Hospitals Birmingham National Health Service (NHS) Foundation Trust, Birmingham, UK; 7Institute of Inflammation and Ageing, University of Birmingham, Birmingham, UK; 8Department of Biological Sciences, University of Chester, Chester, UK; 9Centre of Membrane Proteins and Receptors (COMPARE), Universities of Birmingham and Nottingham, BirminghamNottingham, UK and 10Department of Pathophysiology, Sechenov First Moscow State Medical University, Moscow, Russia 1

Correspondence:

ABSTRACT

MICHAEL G. TOMLINSON m.g.tomlinson@bham.ac.uk

a2+ entry via Orai1 store-operated Ca2+ channels in the plasma membrane is critical to cell function, and Orai1 loss causes severe immunodeficiency and developmental defects. The tetraspanins are a superfamily of transmembrane proteins that interact with specific ‘partner proteins’ and regulate their trafficking and clustering. The aim of this study was to functionally characterize tetraspanin Tspan18. We show that Tspan18 is expressed by endothelial cells at several-fold higher levels than most other cell types analyzed. Tspan18-knockdown primary human umbilical vein endothelial cells have 55-70% decreased Ca2+ mobilization upon stimulation with the inflammatory mediators thrombin or histamine, similar to Orai1-knockdown. Tspan18 interacts with Orai1, and Orai1 cell surface localization is reduced by 70% in Tspan18-knockdown endothelial cells. Tspan18 overexpression in lymphocyte model cell lines induces 20-fold activation of Ca2+ -responsive nuclear factor of activated T cell (NFAT) signaling, in an Orai1-dependent manner. Tspan18-knockout mice are viable. They lose on average 6-fold more blood in a tail-bleed assay. This is due to Tspan18 deficiency in non-hematopoietic cells, as assessed using chimeric mice. Tspan18knockout mice have 60% reduced thrombus size in a deep vein thrombosis model, and 50% reduced platelet deposition in the microcirculation following myocardial ischemia-reperfusion injury. Histamine- or thrombin-induced von Willebrand factor release from endothelial cells is reduced by 90% following Tspan18-knockdown, and histamine-induced increase of plasma von Willebrand factor is reduced by 45% in Tspan18knockout mice. These findings identify Tspan18 as a novel regulator of endothelial cell Orai1/Ca2+ signaling and von Willebrand factor release in response to inflammatory stimuli.

Received: March 26, 2018. Accepted: December 19, 2018. Pre-published: December 20, 2018. doi:10.3324/haematol.2018.194241 Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/104/9/1892 ©2019 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.

1892

haematologica | 2019; 104(9)

Tspan18 regulates Orai1 in endothelial cells

Introduction The tetraspanins are a superfamily of proteins containing four transmembrane regions that interact with and regulate the trafficking, lateral mobility and clustering of specific ‘partner proteins’. These include signaling receptors, adhesion molecules and metalloproteinases.1-3 Recently, the first crystal structure of a tetraspanin, CD81, demonstrated a cone-shaped structure with a cholesterol-binding cavity within the transmembranes.4 Molecular dynamics simulations suggest that cholesterol removal causes a dramatic conformational change, whereby the main extracellular region swings upwards.4 This raises the possibility that tetraspanins function as ‘molecular switches’ to regulate partner protein function via conformational change, and suggests that tetraspanins are viable future drug targets. Tetraspanin Tspan18 was previously studied in chick embryos, in which it stabilizes expression of the homophilic adhesion molecule cadherin 6B to maintain adherens junctions between premigratory epithelial cranial neural crest cells.5,6 Transcriptional Tspan18 downregulation is required for loss of cadherin 6B expression, breakdown of epithelial junctions, and neural crest cell migration. However, Tspan18 knockdown has no major effect on chick embryonic development.5,6 The function of Tspan18 in humans or mice has still not been studied. Store-operated Ca2+ entry (SOCE) through the plasma membrane Ca2+ channel Orai1 is essential for the healthy function of most cell types.7 Loss of SOCE results in severe immunodeficiency that requires a bone marrow transplant for survival. Further symptoms include ectodermal dysplasia and impaired development of skeletal muscle.7 The process of SOCE is biphasic. The first step is initiated following the generation of the second messenger inositol trisphosphate (IP3) from upstream tyrosine kinase or G protein-coupled receptor signaling. IP3 induces the transient release of Ca2+ from endoplasmic reticulum (ER) stores via IP3 receptor channels.8 Depletion of Ca2+ is detected by the ER-resident dimeric Ca2+-sensor protein STIM1, which then undergoes a conformational change and interacts with Orai1 hexamers in the plasma membrane.9,10 STIM1 binding induces Orai1 channel opening and clustering via a mechanism that is not fully understood, allowing Ca2+ entry across the plasma membrane.9,10 The resulting increase in intracellular Ca2+ concentration is relatively large and sustained, sufficient to activate a variety of signaling proteins, including the widely-expressed nuclear factor of activated T-cell (NFAT) transcription factors.8 Endothelial cells line all blood and lymphatic vessels and play a central role in hemostasis and in thromboinflammation, in which inflammatory cells contribute to thrombosis.11,12 In the thrombo-inflammatory disease deep vein thrombosis, blood flow stagnation induced by prolonged immobility, for example, is the trigger for endothelial cells to exocytose Weibel-Palade storage bodies via a mechanism involving Ca2+ signaling.13,14 This releases the multimeric glycoprotein von Willebrand factor (vWF) and the adhesion molecule P-selectin, which recruit platelets and leukocytes, respectively. vWF-bound platelets provide a pro-coagulant surface for activation of clotting factors and thrombin generation, neutrophils release neutrophil extracellular traps, and mast cells release endothelial-activating substances.15-17 This series haematologica | 2019; 104(9)

of thrombo-inflammatory events leads to formation of a blood clot which occludes the vein, and can cause death by pulmonary thromboembolism. The aim of this study was to determine the function of tetraspanin Tspan18 in humans and mice. We found that Tspan18 is highly expressed by endothelial cells, interacts with Orai1, and is required for its cell surface expression and SOCE function. As a consequence, Tspan18deficient endothelial cells have impaired Ca2+ mobilization and release of vWF upon activation induced by inflammatory mediators, and Tspan18-knockout mice are protected from deep vein thrombosis and myocardial ischemia-reperfusion injury, and have defective hemostasis.

Methods Ethics statement Procedures in Birmingham were approved by the UK Home Office according to the Animals (Scientific Procedures) Act 1986, and those in Würzburg by the district government of Lower Frankonia (Bezirksregierung Unterfranken).

Mice Tspan18-/- mice were generated by Genentech/Lexicon Pharmaceuticals on a mixed genetic background of 129/SvEvBrd and C57BL/6J.18 They were purchased from the Mutant Mouse Regional Resource Center and bred as heterozygotes to generate litter-matched Tspan18-/- and Tspan18+/+ pairs. Radiation fetal liver chimeric mice were generated as described.19

Antibodies Anti-epitope tag antibodies were mouse anti-Myc 9B11 and rabbit anti-Myc 71D10 (Cell Signaling Technology), mouse anti-FLAG M2 and rabbit anti-FLAG (Sigma). Other antibodies were mouse anti-human calnexin AF18 (Abcam), rat antimouse CD16/32 (BioLegend), CD41 (eBioscience) and panendothelial cell antigen MECA-32 (BD Pharmingen), mouse antiERK1/2 and rabbit anti-phospho-ERK1/2 (Cell Signaling Technology) and rabbit anti-human vWF (GE Healthcare). Biotinylated isolectin GS-IB4 glycoprotein was from ThermoFisher Scientific.

Expression constructs The NFAT/AP1-luciferase transcriptional reporter construct has been described previously.20,21 pEF6/Myc-His (mock) and pEF6/Myc-His/lacZ were from Invitrogen. N-terminal FLAGtagged tetraspanin constructs were generated in pEF6/Myc-His as described.22,23 pcDNA3.1 Myc-His-tagged human Orai1 and MO70-FLAG-tagged human Orai1 E106Q were from Addgene24 and the dominant-active calcineurin was as described.25

Cell culture and transfections Detailed information on cell cultures is provided in the Online Supplementary Appendix. Wild-type and IP3 receptor-deficient DT40 chicken B-cell lines,26 and Jurkat human T-cell line, were transfected by electroporation.21 Human embryonic kidney (HEK)-293T (HEK-293 cells expressing the large T-antigen of simian virus 40) and the human HeLa epithelial cell line were transfected using polyethylenimine (Sigma)27 and Lipofectamine 2000 (Invitrogen),28 respectively. Human umbilical vein endothelial cells (HUVEC)29 were transfected with 10 nM 1893

P.J. Noy et al.

Silencer Select siRNA duplexes Lipofectamine RNAiMAX (Invitrogen).

(Invitrogen)

using

Quantitative real-time polymerase chain reaction Quantitative real-time polymerase chain reaction (qPCR) was performed using TaqMan assays for Tspan18, Orai1, Orai2, Orai3, 18S and GAPDH.30 Details are available in the Online Supplementary Appendix.

Lentiviral transduction Human umbilical vein endothelial cells were lentivirally transduced with Orai1-Myc as described.31 Details are available in the Online Supplementary Appendix.

Nuclear factor of activated T-cell/AP-1-luciferase transcriptional reporter assay

The NFAT/AP-1-luciferase assay, and β-galactosidase assay to normalize for transfection efficiency, were as described.21

Co-immunoprecipitation A detailed description of co-immunoprecipitation from transfected HEK-293T cell lysates22 is provided in the Online Supplementary Appendix.

Immunofluorescence microscopy Detailed information is provided in the Online Supplementary Appendix. In brief, cells were prepared as described29 and the Manders’ coefficients (M1 and M2) were used as the co-localization measure.32 Ear vasculature was imaged and quantified as described.33,34

Immunohistochemistry Immunohistochemistry was as described;35 details are available in the Online Supplementary Appendix.

Intracellular Ca2+ Human umbilical vein endothelial cells were loaded with Fluo-4 NW dye according to the manufacturer’s instructions (Molecular Probes). Fluorescence was measured every 3 seconds for 5 minutes using a FlexStation fluorescence reader (Molecular Devices), and thrombin (1 U/mL), histamine (20 μM) or ionomycin (10 μM) were injected after acquiring a baseline for 30 seconds.

ELISA and coagulation time assays Detailed information is provided in the Online Supplementary Appendix.

Platelet aggregation and adhesion to human umbilical vein endothelial cells Platelet assays were as described;36,37 detailed information is provided in the Online Supplementary Appendix.

In vivo assays Mouse models were as described;13,37-39 detailed information is in the Online Supplementary Appendix.

Results Tspan18 is expressed by endothelial cells A lack of effective antibodies to many tetraspanins is a current problem in the tetraspanin field. This may be due to their relatively small size, high degree of sequence conservation during evolution, and compact 4-transmem1894

brane structure.4 For example, no Tspan18 antibodies have been published, and commercially-available antibodies are made to Tspan18 peptides and do not detect full-length Tspan18 when rigorously tested (MG Tomlinson, 2019, unpublished manuscript). Therefore, to characterize the Tspan18 expression profile, mouse tissues were tested by qPCR. Tspan18 mRNA was most highly expressed in lung and at lower levels in other tissues (Figure 1A). Analyses of published transcriptomic data40 showed that Tspan18 was most highly expressed by endothelial cells compared to other mouse lung cell types (Figure 1B). Similar analyses of transcriptomic data from mouse brain41 also showed relatively strong endothelial expression of Tspan18 (Figure 1C). Consistent with this, qPCR showed that Tspan18 was expressed by primary HUVEC and the human microvascular endothelial HMEC-1 cell line (Figure 1D). Tspan18 expression was low or absent on most other cell types tested, although peripheral blood leukocytes expressed comparable levels to HUVEC (Figure 1D). In transcriptomic data from the Human Protein Atlas (www.proteinatlas.org), Tspan18 was expressed by most human tissues at a level between 10 and 70 tags per million, but in cell lines was only expressed at 10 or greater tags per million by HUVEC and 8 of the other 64 cell types analyzed.42

Tspan18 is required for Ca2+ signaling in primary human endothelial cells To investigate Tspan18 function, its expression in HUVEC, which is a widely-used primary human endothelial cell model, was knocked-down using two different siRNA duplexes. Subsequent analyses revealed a 60% reduction in peak Ca2+ elevation in response to the inflammatory mediator thrombin (Figure 2A). A similar defect was observed in response to histamine (Figure 2B). Positive control ionomycin treatment gave a sustained intracellular Ca2+ response in all samples (Figure 2C) and effective knockdown was confirmed by qPCR (Figure 2D). Functionality of thrombin and histamine receptors was confirmed by anti-phospho-ERK1/2 mitogen-activated protein kinase (MAPK) western blotting, as this was not affected by Tspan18 knockdown (Figure 2E).

Tspan18 promotes Ca2+-responsive nuclear factor of activated T-cell signaling in lymphocyte cell lines To investigate the mechanism by which Tspan18 regulates Ca2+ signaling, a more tractable cell line system was established, namely DT40 cells that are derived from chicken B cells. In this cell line, a transfected NFAT/adapter protein 1 (AP-1) transcriptional luciferase reporter can be used as a readout for Ca2+ signaling downstream of transfected membrane proteins.21,43 Transfection of a FLAG epitope-tagged Tspan18 expression construct was sufficient to induce robust NFAT/AP-1 activation (Figure 3A). As controls, five other FLAGtagged tetraspanins (CD9, CD63, CD151, Tspan32 and Tspan9) were chosen because they represent a diverse range of tetraspanins based on sequence identities.22 These did not induce NFAT/AP-1 activation, despite their substantially higher expression than Tspan18 as assessed by anti-FLAG western blotting (Figure 3A). Despite the fact that the NFAT/AP-1 promoter can be activated by Ca2+ signaling, it is maximally activated by combined Ca2+ signaling and MAPK; Ca2+ activates NFAT and MAPK activates AP-1. To determine whether haematologica | 2019; 104(9)

Tspan18 regulates Orai1 in endothelial cells

Figure 1. Tspan18 is highly expressed by endothelial cells. (A) Quantitative real-time polymerase chain reaction (qPCR) was carried out for Tspan18 using cDNA derived from a panel of mouse tissues. Data were normalized for the HPRT housekeeping gene and adjusted such that the lung value was 100 in each experiment. Error bars represent Standard Error of Mean from three independent tissue samples. (B) RNA-Seq data from major cell types in mouse lung, generated by Du et al.,40 was used to show Tspan18 mRNA expression levels as fragments per kilobase of transcript sequence per million mapped fragments (FPKM). (C) RNA-Seq data from major cell types in mouse brain, generated by Zhang et al.,41 was used to show Tspan18 mRNA expression levels as described in (B). (D) qPCR was carried out for Tspan18 on cDNA derived from a panel of primary human cells [dermal fibroblasts, aortic smooth muscle, hepatocytes, peripheral blood leukocytes from buffy coat and human umbilical vein endothelial cells (HUVEC)], and human cell lines (HEK-293T human embryonic kidney cells, MDA-MB-231 epithelial cells, DAMI megakaryocytic cells, HEL and K562 erythroleukemia cells, U937 monocytic cells, Jurkat and HPB-ALL T cells, DG75 and Raji B cells and HMEC-1 microvascular endothelial cells). Data were normalized for actin and adjusted such that the HUVEC value was 100 in each experiment. Error bars represent Standard Error of Mean from two independent cell samples.

Tspan18 activates Ca2+ signaling, MAPK or both, Tspan18-transfected DT40 cells were stimulated with the Ca2+ ionophore ionomycin or phorbol ester PMA to activate the MAPK pathway. PMA synergized with Tspan18 expression in activating NFAT/AP-1, but ionomycin did not (Figure 3B). As a positive control, combined PMA and ionomycin induced relatively strong NFAT/AP-1 activation in the presence or absence of Tspan18 (Figure 3B). The capacity of Tspan18 overexpression to induce NFAT/AP-1 activation was not restricted to DT40 B cells, since similar data were obtained in the human Jurkat Tcell line (Figure 3C). Taken together, these data suggest that Tspan18 promotes Ca2+ signaling and NFAT activation via a mechanism that is common to endothelial cells, B cells and T cells.

Tspan18-induced nuclear factor of activated T-cell activation requires functional Orai1 store-operated Ca2+ entry channels To understand the mechanism by which Tspan18 promotes Ca2+ signaling, a series of NFAT/AP-1 reporter experiments were conducted in gene-knockout DT40 cells and using inhibitors and a dominant-interfering construct. Firstly, Tspan18-induced NFAT/AP-1 activation was found to be independent of the three IP3 receptors haematologica | 2019; 104(9)

(Figure 3D). IP3 receptors release Ca2+ from ER stores in response to tyrosine kinase and G protein-coupled receptor activation, suggesting that Tspan18 does not operate on these pathways or IP3 receptors themselves. However, Tspan18 did not activate NFAT/AP-1 following chelation of extracellular Ca2+ (Figure 3E), or following treatment with the immunosuppressive drug cyclosporin A (Figure 3F), which prevents NFAT translocation to the nucleus by inhibiting its activatory phosphatase calcineurin. These data suggest that Tspan18 might induce Ca2+ entry via the SOCE channel Orai1, a major entry route for extracellular Ca2+ in non-excitable cells.8 Consistent with this possibility, a dominant interfering form of Orai1 (E106Q), which multimerizes with endogenous Orai1 to yield a non-functional channel,44-46 inhibited Tspan18-induced NFAT/AP-1 activation (Figure 3G). As a positive control to confirm that downstream NFAT signaling was still intact in the presence of dominant interfering Orai1, its inhibitory effect was overcome by the expression of an active form of calcineurin (Figure 3G). Therefore, Tspan18 may activate Ca2+ entry through the Orai1 SOCE pathway.

Tspan18 interacts with Orai1 To investigate whether Tspan18 interacts with Orai1, transfected epitope-tagged proteins were used because of 1895

P.J. Noy et al.

the lack of effective antibodies to Tspan18. To test for an interaction using co-immunoprecipitation, transfected HEK-293T cells were lyzed in 1% digitonin, a stringent detergent that has been used previously to identify tetraspanin-partner protein interactions.22,47 FLAG-tagged Tspan18 co-immunoprecipitated with Myc-tagged Orai1,

but five other control tetraspanins did not (Figure 4A). Moreover, Tspan18 and Orai1 co-localized when expressed in HeLa cells, at a level of approximately 90% pixel co-localization when assessed using the Manders’ coefficient (Figure 4B). These data suggest that Tspan18 interacts with Orai1.

Figure 2. Tspan18-knockdown endothelial cells have impaired Ca2+ mobilization. (A-D) Human umbilical vein endothelial cells (HUVEC) were transfected with a negative control siRNA (CON) or with one of two independent siRNA targeting Tspan18 (T18 KD). After 48 hours, HUVEC were loaded with the Ca2+-sensitive dye Fluo-4 NW and Ca2+ measurements taken using a FlexStation fluorescence reader during addition (arrow) of (A) 1 U/mL thrombin, (B) 20 μM histamine, or (C) 10 μM ionomycin. Representative Ca2+ traces are shown for Tspan18-knockdown HUVEC (left), with quantitation of maximum intracellular Ca2+ concentrations (right). Data were analyzed by one-way ANOVA with Dunnett’s multiple comparisons test. Error bars represent the Standard Error of Mean (SEM) from three independent experiments. *P<0.05; ***P<0.001. (D) siRNA-transfected HUVEC from (A) to (C) were harvested, mRNA extracted, cDNA produced and Tspan18 mRNA levels were assessed by quantitative real-time polymerase chain reaction (qPCR). Data were normalized to 18S and actin as internal controls and adjusted such that the non-siRNA-transfected mock value was 1 in each experiment. Data were then normalized by logarithmic transformation, and analyzed by one-way ANOVA and Tukey’s multiple comparison test. Error bars represent the Standard Error of the Mean from three independent experiments. ***P<0.001. (E) HUVEC were subjected to Tspan18 siRNA knockdown as described for (A-D), stimulated with 1 U/mL thrombin or 20 μM histamine for 5 minutes, then whole cell lysates were analyzed by western blotting with phospho-ERK1/2 and total ERK1/2 antibodies. (Left) Representative blots. (Right) Quantitation of three independent experiments. Error bars represent SEM. Knockdown efficiency was similar to that shown in (D) (data not shown). secs: seconds; RFU: relative fluorescence unit.

1896

haematologica | 2019; 104(9)

Tspan18 regulates Orai1 in endothelial cells

Orai1-knockdown endothelial cells have impaired Ca2+ mobilization and Orai1 surface expression requires Tspan18 To determine whether knockdown of Orai1 could phenocopy Tspan18 knockdown, intracellular Ca2+ mobilization was measured following siRNA-mediated knockdown of Orai1. This resulted in impaired Ca2+ mobilization in response to thrombin (Figure 5A) or histamine (Figure 5B). As a control, knockdown of the other Orai family members, Orai2 and Orai3, did not affect Ca2+ mobilization (Figure 5A and B), in agreement with previous studies on Orai proteins in HUVEC.48,49 Positive control ionomycin treatment gave a sustained intracellular Ca2+ response in all samples (Figure 5C) and effective

knockdown was confirmed by qPCR (Figure 5D-F). A common mechanism of tetraspanin function is to regulate their partner proteins by facilitating their exit from the endoplasmic reticulum (ER) and trafficking to the cell surface.1,50,51 To determine whether Orai1 localization could be regulated by Tspan18 in this manner, HUVEC were lentivirally transduced with Myc-Orai1 and transfected with control or Tspan18 siRNA duplexes, and Orai1 subcellular localization assessed by confocal microscopy. Orai1 was localized primarily to the cell periphery in control cells, but this was reduced following Tspan18 knockdown, with Orai1 partially co-localized with the ER marker calnexin (Figure 5G). Quantitative analyses showed that approximately 40% of Orai1 was

Figure 3. Tspan18 overexpression in cell lines activates Ca2+-responsive NFAT signaling in an Orai1-dependant manner. (A) The DT40 B cell line was transfected with an NFAT/AP-1-luciferase reporter construct, a β-galactosidase expression construct driven by the elongation factor (EF)-1a promoter to control for transfection efficiency, and FLAG-tagged mouse tetraspanin constructs or empty vector control. At 24 hours (h) post transfection, cells were lyzed and assayed for luciferase and β-galactosidase. Luciferase data were normalized for β-galactosidase values (left). Whole cell lysates from these cells were separated by SDS-PAGE and blotted with an anti-FLAG antibody. (Right) Representative blot. (B and C) The DT40 B cell line (B) and the human Jurkat T cell line (C) were transfected with an NFAT/AP-1luciferase reporter construct and β-galactosidase expression construct with (+) or without (-) FLAG-tagged mouse Tspan18. At 24 h post transfection cells were stimulated for 6 h with 50 ng/mL PMA or 1 μM ionomycin (Iono) (left), or both together (right). Luciferase assays were then performed as described in (A). (D) DT40 cells with (+) or without (-) expression of FLAG-tagged mouse Tspan18 were tested for NFAT/AP-1 luciferase activity as described in (A), but using cells with gene knockouts of the three IP3 receptors (IP3R-) in comparison to wild-type (WT) cells (left). Whole cell lysates were western blotted with an anti-FLAG antibody (right). (E and F) DT40 cells with (+) or without (-) expression of FLAG-tagged mouse Tspan18 were tested for NFAT/AP-1 luciferase activity as described in (A), except that cells were treated with 4 mM EGTA as a Ca2+ chelator (E) or with 2 μM cyclosporin A as a calcineurin inhibitor (F). (G) DT40 cells were transfected with FLAG-tagged human Tspan18 in the presence or absence of a dominant negative human Orai1 E106Q mutant construct, or a consitutively active human calcineurin construct. The experiment was conducted as described for (A). All luciferase data were corrected for β-galactosidase values, normalized by logarithmic transformation, and analyzed by one-way ANOVA and Tukey’s multiple comparison test. *P<0.05; **P<0.01; ***P<0.001. Error bars represent the Standard Error of the Mean from at least three independent experiments. ns: not significant.

haematologica | 2019; 104(9)

1897

P.J. Noy et al.

ER-localized in Tspan18 knockdown cells compared to 10-15% in control cells. This partial co-localization could be due to some Orai1 localization in the Golgi and/or trans-Golgi network, as shown by staining close to the nucleus, rather than the more extended perinuclear reticular staining of the ER (Figure 5G). These data support a role for Tspan18 in regulating Orai1 exit from the ER, and/or Golgi, and trafficking to the cell surface.

Tspan18 deficient mice have impaired hemostasis due to a defect in non-hematopoietic cells To investigate Tspan18 function in vivo, Tspan18knockout mice were acquired from Genentech/Lexicon Pharmaceuticals. These mice had been generated as part of a library of 472 knockouts,18 but were functionally uncharacterized. Breeding of heterozygotes gave an equal proportion of male and female mice with Mendelian genotype ratios, and the mice bred successfully as homozygote knockouts (data not shown). Furthermore, Tspan18-knockout mice had normal body weights and whole blood cell counts (data not shown).

Tspan18-knockout mice were first evaluated for a hemostasis phenotype using a tail bleed assay. Most Tspan18-knockout mice bled more than wild-type littermates, demonstrating a significant disruption to hemostasis (Figure 6A). Some Tspan18-knockout mice did not bleed excessively (Figure 6A), indicating that the bleeding phenotype was variable. This variability could be due to genetic modifier loci, as demonstrated in mice deficient for the platelet collagen/fibrin receptor GPVI.52 To determine whether impaired hemostasis was due to loss of Tspan18 from hematopoietic or non-hematopoietic cells, tail bleeding assays were performed on irradiated fetal liver chimeric mice. These demonstrated that the bleeding phenotype was due to Tspan18 loss from nonhematopoietic cells (Figure 6A), and suggests that a platelet defect is not responsible. Consistent with this, Tspan18-knockout platelets aggregated normally in response to collagen (Figure 6B) or thrombin (Figure 6C). Furthermore, prothrombin time and activated partial thromboplastin time tests showed that coagulation was similar for wild-type and Tspan18-knockout plasma

1898

Figure 4. Tspan18 interacts with Orai1. (A) HEK-293T cells were transfected with a Myctagged human Orai1 expression construct and one of a panel of FLAG-tagged human tetraspanin constructs. Cells were lyzed in 1% digitonin and immunoprecipitated with an anti-FLAG antibody. Samples were separated by SDS-PAGE and both immunoprecipitated (IP) and whole cell lysate (WCL) samples were blotted with anti-FLAG and anti-Myc antibodies. A representative blot for each is shown (left) with quantitation of Myc-tagged Orai1 immunoprecipitated with the tetraspanins (right). Data were nomalized by logarithmic transformation before analysis by one-way ANOVA and Dunnettâ&#x20AC;&#x2122;s post test. Error bars represent Standard Error of Mean (SEM) from three independent experiments. *P<0.05. (B) HeLa cells were transfected with Myc-tagged human Orai1, FLAG-tagged human Tspan18, or both constructs. Cells were fixed and stained with an anti-Myc antibody (green), an anti-FLAG antibody (red), and imaged by confocal microscopy (left). The Manders' coefficients (M1 and M2) were calculated from the confocal stacks to quantify the degree of overlap (right). Error bars represent the SEM from three independent experiments.

haematologica | 2019; 104(9)

Tspan18 regulates Orai1 in endothelial cells

(Figure 6D and E). These data suggest that the impaired hemostasis phenotype is due to a defect in a nonhematopoietic vascular cell type such as the endothelial cell. This did not appear to be due to any observable structural defects in the vasculature, because immunohistochemistry analyses of blood vessels in organs such as kidney and pancreas were similar for wild-type and Tspan18-knockout mice (Figure 6F), as were immunofluorescence analyses of blood vessels in the ear (Figure 6G).

Tspan18 and Orai1 are required for endothelial release of von Willebrand factor in response to inflammatory mediators Endothelial cell stimulation by inflammatory agonists induces vWF release from Weibel-Palade bodies via a process that involves Ca2+ signaling.14 To investigate whether Tspan18 could be required for vWF release, HUVEC were subjected to Tspan18 knockdown and stimulated in culture medium with thrombin or hista-

Figure 5. Orai1-knockdown endothelial cells have impaired Ca2+ mobilization and Orai1 surface expression requires Tspan18. (A-F) Human umbilical vein endothelial cells (HUVEC) were transfected with a negative control siRNA (CON) or with one of two independent siRNA targeting Orai1, Orai2 or Orai3 (Orai1-3 KD). After 48 hours (h), Ca2+ measurements were taken as described in Figure 2A-C, following addition of 1 U/mL thrombin (A), 20 ÎźM histamine (B), or 10 ÎźM ionomycin (C), and quantitation of maximum intracellular Ca2+ concentrations is shown. Error bars represent Standard Error of the Mean (SEM) from three independent experiments. **P<0.01; ***P<0.001. (D-F) siRNA-transfected HUVEC from (A-C) were subjected to quantitative real-time polymerase chain reaction (qPCR) for Orai1 (D), Orai2 (E) or Orai3 (F), as described for Figure 2D. Error bars represent SEM from three independent experiments. ***P<0.001. (G) HUVEC lentivirally-transduced to express Myc-tagged Orai1 were treated with control or Tspan18 siRNA. Cells were stained with anti-Myc (white) or anti-calnexin endoplasmic reticulum marker (red) antibodies, and representative confocal microscopy images are shown (top). In the line graphs below the images (bottom), the percentage expression of Orai1 (black) and calnexin (red) across the yellow line in the top panel was determined using ImageJ. The percentage of Orai1 that localized to a calnexin endoplasmic reticulum mask was then quantified (right). Data were generated from 15 cells per condition from three independent experiments (five cells per condition per experiment). Error bars represent SEM. ***P<0.001. RFU: relative fluorescence unit.

haematologica | 2019; 104(9)

1899

P.J. Noy et al.

mine. Soluble vWF release, as detected by ELISA, was reduced by approximately 90% compared to control cells (Figure 7A). This was corroborated by immunofluorescent staining of vWF that showed minimal release of Weibel-Palade bodies after Tspan18 knockdown (Figure 7B). Similar to Tspan18, Orai1 knockdown reduced Weibel-Palade body release after thrombin stimulation,

but knockdown of Orai2 or Orai3 had no effect (Figure 7C). Furthermore, Tspan18 knockdown reduced platelet adhesion to a thrombin-activated HUVEC monolayer by 85-90% (Figure 7D). These data support a role for Tspan18 and Orai1 in vWF release and platelet capture following endothelial cell activation by inflammatory mediators.

Figure 6. Tspan18-knockout mice have a hemostasis defect due to the absence of Tspan18 expression by non-hematopoietic cells. (A) Tail bleeding assays were performed by amputating 3 mm tail tips of anesthetized mice and the weight of blood lost was measured. The mice were Tspan18+/+, Tspan18-/-, or lethally irradiated Tspan18-/- or Tspan18+/+ mice reconstituted with fetal liver cells from Tspan18+/+ or Tspan18-/- embryos. Each symbol represents one animal. All data were analyzed by Fisher’s exact test. *P<0.05; ***P<0.001. Note that bleeding was stopped by cauterizing the tails of some mice, because of regulations limiting the amount of blood loss on our Home Office License. (B and C) Washed platelets from Tspan18+/+ or Tspan18-/- mice were activated with 3 μg/mL collagen (B) or 0.3 U/mL thrombin (C), and aggregation was measured by light transmission with stirring. Quantitated percentage aggregation each minute is shown. Error bars represent the Standard Error of Mean (SEM) from at least three pairs of litter-matched mice. (D and E) Plasma samples from Tspan18+/+ and Tspan18-/- mice were subjected to a prothrombin time test with human placental thromboplastin (D) and an activated partial thromboplastin time test with purified soy phosphatides with ellagic acid (E). Error bars represent SEM from four pairs of litter-matched mice. (F) Immunohistochemistry was used to show a grossly normal vasculature in Tspan18+/+ and Tspan18-/- mice formalin-fixed paraffin-embedded 5 μm sections from kidney and pancreas, using the MECA32 anti-mouse panendothelial cell antibody. Images are representative of three pairs of litter-matched mice. (G) Confocal microscopy was used to show a grossly normal vasculature in Tspan18+/+ and Tspan18-/- mice ears, by staining anterior ear tissue with biotinylated isolectin GS-IB4 glycoprotein followed by Alexa647-conjugated streptavidin. Images are representative of three pairs of litter-matched mice. ImageJ quantitation of 3 fields of view (500 x 500 pixels) per mouse showed a mean total vessel length of 11131±1271 pixels for Tspan18+/+ and 11684±283 pixels for Tspan18-/- (n=3; error represents SEM). mins: minutes.

1900

haematologica | 2019; 104(9)

Tspan18 regulates Orai1 in endothelial cells

Figure 7. Tspan18-knockdown endothelial cells have impaired histamine- and thrombin-induced release of von Willebrand Factor (vWF). (A) Human umbilical vein endothelial cells (HUVEC) were transfected with a negative control siRNA (CON) or with one of two independent siRNA targeting Tspan18 (T18 KD). After 48 hours, HUVEC were stimulated with 1 U/mL thrombin or 20 ÎźM histamine for 5 minutes (min). Cultured media was removed and assayed for vWF by ELISA. Pre-stimulation levels of vWF were subtracted from these data. Error bars represent the Standard Error of Mean (SEM) from three experiments. **P<0.01. (B) HUVEC transfected as described in (A) were stimulated with 1 U/mL thrombin for 5 min, and the cells were fixed and stained with an anti-vWF antibody followed by Alexa488-conjugated secondary antibody. Representative confocal microscopy images are shown (left). Z-stack images were de-noised, background-subtracted and analyzed for the number of vWF cellular bodies, using ImageJ. Counts were made on 5-10 cells per experiment for four independent experiments (right panel). Error bars represent Standard Error of Mean (SEM). ***P<0.001; ns: not significant. (C) The experiment and quantitation was conducted as for (B), except that HUVEC were mock-transfected with no siRNA (Mock), transfected with negative control siRNA (CON), or siRNA to Orai1, Orai2 or Orai3 (KD). (D) HUVEC were siRNA-transfected and stimulated with 1 U/mL thrombin for 5 min as described in (A). Human washed platelets were fluorescently labeled and incubated with the HUVEC monolayers. Non-adherent platelets were removed by washing and images were collected. Representative fluorescent images of adhered platelets (top) and phase contrast images of the HUVEC monolayers and adhered platelets (bottom) are shown, with quantitation of platelet adhesion from three independent experiments (right). Error bars represent SEM. *P<0.05.

haematologica | 2019; 104(9)

1901

P.J. Noy et al. A

1902

Figure 8. Tspan18-knockout mice have impaired histamineinduced von Willebrand Factor (vWF) release and impaired thrombo-inflammatory responses. (A) Tspan18+/+ and Tspan18-/- mice were intraperitoneally-injected with histamine and plasma vWF levels were measured 30 minutes (min) later by ELISA. Error bars represent Standard Error of Mean (SEM) from eight Tspan18+/+ and seven Tspan18-/mice. *P<0.05. (B) Tspan18+/+and Tspan18-/- mice were anesthetized, and the abdominal aorta exposed and mechanically injured through a single firm compression with forceps. Blood flow was subsequently monitored with a Doppler flowmeter to calculate the time until complete occlusion of the vessel. Each symbol represents one animal. (C) Tspan18+/+ and Tspan18-/- mice were anesthetized and the mesentery was exteriorized through an abdominal incision. Platelets were fluorescently labeled with Dylight 488-conjugated anti-GPIX derivative. Small mesenteric arterioles were exposed to FeCl3-induced chemical injury via topical application. Time to appearance of the first thrombi was recorded (left), and the time until complete occlusion of the vessel was measured using fluorescence intravital microscopy (right). Each symbol represents one animal. (D) Tspan18+/+ and Tspan18-/- mice were anesthetized and surgery performed to stenose the inferior vena cava. After 48 hours (h), thrombus length (left) and weight (right) were measured. Each symbol represents one animal and horizontal bars represent the median. *P<0.05. All data in (A-D) were analyzed by one-way ANOVA with Dunnettâ&#x20AC;&#x2122;s multiple comparisons test. (E) Myocardial ischemia-reperfusion injury was induced in the left ventricle of the beating heart of anesthetized mice by occluding the left anterior descending artery for 45 min with a suture. Reperfusion was instigated for 2 h by removal of the ligature, after which the organ was harvested. Frozen sections were analyzed for the presence of platelets in the microcirculation by immunofluorescence microscopy. Three litter-matched pairs of Tspan18+/+ and Tspan18-/- mice were used, with three sections per mouse and five images analyzed per section. Each symbol represents one image. For each image, the integrated density value was calculated as a representation of the total number of platelets (left), and the average aggregate size was also calculated (right). Error bars represent the SEM and data were analyzed by MannWhitney test. ****P<0.0001.

haematologica | 2019; 104(9)

Tspan18 regulates Orai1 in endothelial cells

Tspan18-knockout mice have impaired histamine-induced release of endothelial von Willebrand factor and impaired thrombo-inflammatory responses To determine whether Tspan18 has a role in vWF release in vivo, Tspan18-knockout mice were intra-peritoneally injected with histamine, and plasma vWF levels were analyzed by ELISA. Induced plasma vWF release was reduced by approximately 50% in the absence of Tspan18 (Figure 8A). Basal plasma vWF was normal in Tspan18-knockout mice (Figure 8A), indicating a requirement for Tspan18 in regulated, but not basal, vWF release. To investigate the role of Tspan18 in thrombosis, two arterial thrombosis models and two thrombo-inflammatory models were used. In a platelet-driven aorta injury arterial thrombosis model,38 no difference in time to complete occlusion of the vessel between Tspan18-knockout and wild-type littermate control mice was observed (Figure 8B). Similarly, there was no thrombosis defect in mesenteric arterioles following application of FeCl3 (Figure 8C), which is also a platelet-driven model,38 but shows reduced platelet deposition and thrombus formation in the complete absence of vWF.53,54 In a deep vein thrombosis thrombo-inflammatory model that is dependent on endothelial vWF,13 thrombus length and weight were reduced by approximately 60% in the Tspan18-knockout mice, compared to wild-type littermate controls (Figure 8D). Moreover, 4 of 9 Tspan18-knockout mice failed to develop a thrombus compared to 100% thrombus formation in wild-type mice (Figure 8D). Macroscopically, thrombi from Tspan18-knockout mice had similar red and white parts to those from wild-type mice (data not shown). Finally, in a vWF-dependent myocardial ischemia-reperfusion thrombo-inflammatory model,55,56 platelet deposition and aggregate size in the microcirculation were reduced by approximately 50% (Figure 8E). The reduction in severity in the two thrombo-inflammatory models is consistent with the requirement of Tspan18 for endothelial vWF release in response to inflammatory mediators.

Discussion We have discovered that Tspan18 is expressed by endothelial cells and interacts with the SOCE channel Orai1. Tspan18-knockdown endothelial cells had reduced Orai1 expression at the cell surface and impaired Ca2+ signaling. This is consistent with the established role of tetraspanins in interacting with specific partner proteins in the ER, and promoting their trafficking to the cell surface,1,50,51 albeit via mechanisms that are yet to be defined. Tspan18 is not particularly related to any of the other 32 mammalian tetraspanins,22 suggesting that it may be unique amongst tetraspanins in regulating Orai1. Indeed, none of the five tetraspanins that were selected as controls interacted with Orai1, or induced Ca2+-responsive NFAT activation, when over-expressed. At the cell surface, tetraspanins can regulate the lateral diffusion and clustering of their partner proteins.2,3 A question that arises from the present study is whether Tspan18 regulates Orai1 clustering at the endothelial cell surface. Interestingly, a unimolecular coupling model of Orai1 activation was recently proposed, whereby one molecule of a STIM1 dimer is sufficient to induce opening of the Orai1 hexamer channel.10 This would enable the haematologica | 2019; 104(9)

other STIM1 molecule in the dimer to cross-link with a second Orai1 hexamer and form a lattice of clustered Orai1 channels. The degree of cluster formation could dictate the kinetics of channel activation and could concentrate Ca2+ influx to particular regions of the plasma membrane.10 This could affect the extent to which downstream effectors are activated. Tetraspanins have been reported to exist as nanodomains of approximately ten tetraspanins of a single type,57 therefore Tspan18 may cluster Orai1 into pre-formed nanodomains, so modulating Orai1 lattice formation by STIM1. This may provide a means by which endothelial cells fine-tune SOCE and downstream functional responses. It remains to be determined whether Tspan18 also regulates Orai2 and Orai3, but we found no role for these Orai family members in inflammatory mediator-induced HUVEC Ca2+ mobilization, consistent with other studies.48,49 The inflammatory mediators thrombin and histamine activate G protein-coupled receptors to induce downstream Ca2+ mobilization and the release of vWF from Weibel-Palade bodies.14 Consistent with impaired Ca2+ signaling, Tspan18-knockout endothelial cells had impaired inflammatory mediator-induced vWF release in vitro and in vivo. In contrast, basal release of vWF appeared to be normal, because basal plasma vWF levels were unaffected in Tspan18-knockout mice. We hypothesize that impaired vWF release, in response to inflammatory mediators, explains the in vivo phenotypes observed in Tspan18-knockout mice. The protection from deep vein thrombosis is consistent with the central role of vWF in this disease.13 Furthermore, the reduced platelet deposition in the microcirculation during myocardial ischemia-reperfusion injury is consistent with the role of vWF in this process.55,56 The hemostasis defect still needs to be explained, because although a tail bleeding phenotype has been demonstrated in endothelial-specific vWFknockout mice, these animals also had low plasma vWF,58,59 unlike Tspan18 knockouts. The tail bleeding assay measures blood loss following excision of the tip of the tail, which contains the two lateral veins, the dorsal vein, and the ventral artery. We speculate that endothelial cells in the veins and artery, adjacent to the site of excision, are activated and release vWF via Ca2+-dependent signaling. The vWF may trap platelets, facilitating their aggregation and preventing excessive blood loss from the site of tail injury. Therefore, our data suggest that acute release of vWF adjacent to a site of injury might be important for hemostasis, at least for some types of injury. Finally, the lack of a phenotype in the two arterial thrombosis models is consistent with the importance of platelets in these models,38 and we found no defect in aggregation in vitro for Tspan18-knockout platelets. In summary, we have identified Tspan18 as a novel regulator of endothelial Orai1 and SOCE. Our in vivo data show that Tspan18 regulates inflammation-induced vWF release but not basal release, and promotes hemostasis and thrombo-inflammatory processes but not arterial thrombosis. Acknowledgments We are grateful to Carl Blobel, Chris Bunce, Dean Kavanagh, Neil Morgan and Steve Publicover for their helpful comments on this project. We thank the Birmingham Biomedical Sciences Unit for maintaining mice, and the Birmingham Advanced Light Microscope Facility for imaging expertise. 1903

P.J. Noy et al.

Funding This work was funded by a British Heart Foundation Project Grant (PG/13/92/30587) which supported PJN, Biotechnology and Biological Sciences Research Council PhD Studentships which supported RLG and EJH, British Heart Foundation PhD Studentships which supported DC, JSR and CZK (FS/05/048, FS/12/79/29871 and FS/18/9/33388), a British Heart

References 1. Matthews AL, Szyroka J, Collier R, Noy PJ, Tomlinson MG. Scissor sisters: regulation of ADAM10 by the TspanC8 tetraspanins. Biochem Soc Trans. 2017;45(3):719-730. 2. Termini CM, Gillette JM. Tetraspanins Function as Regulators of Cellular Signaling. Front Cell Dev Biol. 2017;5:34. 3. van Deventer SJ, Dunlock VE, van Spriel AB. Molecular interactions shaping the tetraspanin web. Biochem Soc Trans. 2017;45(3):741-750. 4. Zimmerman B, Kelly B, McMillan BJ, et al. Crystal Structure of a Full-Length Human Tetraspanin Reveals a Cholesterol-Binding Pocket. Cell. 2016;167(4):1041-1051. 5. Fairchild CL, Conway JP, Schiffmacher AT, Taneyhill LA, Gammill LS. FoxD3 regulates cranial neural crest EMT via downregulation of tetraspanin18 independent of its functions during neural crest formation. Mech Dev. 2014;132:1-12. 6. Fairchild CL, Gammill LS. Tetraspanin18 is a FoxD3-responsive antagonist of cranial neural crest epithelial-to-mesenchymal transition that maintains cadherin-6B protein. J Cell Sci. 2013;126(Pt 6):1464-1476. 7. Putney JW, Steinckwich-Besancon N, Numaga-Tomita T, et al. The functions of store-operated calcium channels. Biochim Biophys Acta. 2017;1864(6):900-906. 8. Trebak M, Putney JW Jr. ORAI Calcium Channels. Physiology (Bethesda). 2017; 32(4):332-342. 9. Yeung PS, Yamashita M, Prakriya M. Pore opening mechanism of CRAC channels. Cell Calcium. 2017;63:14-19. 10. Zhou Y, Cai X, Nwokonko RM, Loktionova NA, Wang Y, Gill DL. The STIM-Orai coupling interface and gating of the Orai1 channel. Cell Calcium. 2017;63:8-13. 11. Gragnano F, Sperlongano S, Golia E, et al. The Role of von Willebrand Factor in Vascular Inflammation: From Pathogenesis to Targeted Therapy. Mediators Inflamm. 2017;2017:5620314. 12. Kawecki C, Lenting PJ, Denis CV. von Willebrand factor and inflammation. J Thromb Haemost. 2017;15(7):1285-1294. 13. Brill A, Fuchs TA, Chauhan AK, et al. von Willebrand factor-mediated platelet adhesion is critical for deep vein thrombosis in mouse models. Blood. 2011;117(4):1400-1407. 14. McCormack JJ, Lopes da Silva M, Ferraro F, Patella F, Cutler DF. Weibel-Palade bodies at a glance. J Cell Sci. 2017;130(21):3611-3617. 15. Brill A, Fuchs TA, Savchenko AS, et al. Neutrophil extracellular traps promote deep vein thrombosis in mice. J Thromb Haemost. 2012;10(1):136-144. 16. Ponomaryov T, Payne H, Fabritz L, Wagner DD, Brill A. Mast Cells Granular Contents Are Crucial for Deep Vein Thrombosis in

1904

Foundation Senior Fellowship (FS/08/062/25797) to MGT which also supported JY, a Biotechnology and Biological Sciences Research Council Project Grant (BB/P00783X/1) which supported NH and a Medical Research Council New Investigator Award (RRAK10717) which supported MGT. SPW is a British Heart Foundation Chair (CH03/003).

Mice. Circ Res. 2017;121(8):941-950. 17. von Bruhl ML, Stark K, Steinhart A, et al. Monocytes, neutrophils, and platelets cooperate to initiate and propagate venous thrombosis in mice in vivo. J Exp Med. 2012;209(4):819-835. 18. Tang T, Li L, Tang J, et al. A mouse knockout library for secreted and transmembrane proteins. Nat Biotechnol. 2010;28(7):749-755. 19. Hughes CE, Navarro-Nunez L, Finney BA, Mourao-Sa D, Pollitt AY, Watson SP. CLEC2 is not required for platelet aggregation at arteriolar shear. J Thromb Haemost. 2010;8(10):2328-2332. 20. Shapiro VS, Truitt KE, Imboden JB, Weiss A. CD28 mediates transcriptional upregulation of the interleukin-2 (IL-2) promoter through a composite element containing the CD28RE and NF-IL-2B AP-1 sites. Mol Cell Biol. 1997;17(7):4051-4058. 21. Tomlinson MG, Calaminus SD, Berlanga O, et al. Collagen promotes sustained glycoprotein VI signaling in platelets and cell lines. J Thromb Haemost. 2007;5(11):2274-2283. 22. Haining EJ, Yang J, Bailey RL, et al. The TspanC8 subgroup of tetraspanins interacts with A disintegrin and metalloprotease 10 (ADAM10) and regulates its maturation and cell surface expression. J Biol Chem. 2012;287(47):39753-39765. 23. Protty MB, Watkins NA, Colombo D, et al. Identification of Tspan9 as a novel platelet tetraspanin and the collagen receptor GPVI as a component of tetraspanin microdomains. Biochem J. 2009;417(1):391-400. 24. Gwack Y, Srikanth S, Feske S, et al. Biochemical and functional characterization of Orai proteins. J Biol Chem. 2007;282(22):16232-16243. 25. O'Keefe SJ, Tamura J, Kincaid RL, Tocci MJ, O'Neill EA. FK-506- and CsA-sensitive activation of the interleukin-2 promoter by calcineurin. Nature. 1992;357(6380):692-694. 26. Sugawara H, Kurosaki M, Takata M, Kurosaki T. Genetic evidence for involvement of type 1, type 2 and type 3 inositol 1,4,5-trisphosphate receptors in signal transduction through the B-cell antigen receptor. Embo J. 1997;16(11):3078-3088. 27. Ehrhardt C, Schmolke M, Matzke A, et al. Polyethylenimine, a cost-effective transfection reagent. Signal Transduction. 2006;6:179â&#x20AC;&#x201C;184. 28. Noy PJ, Yang J, Reyat JS, et al. TspanC8 Tetraspanins and A Disintegrin and Metalloprotease 10 (ADAM10) Interact via Their Extracellular Regions: EVIDENCE FOR DISTINCT BINDING MECHANISMS FOR DIFFERENT TspanC8 PROTEINS. J Biol Chem. 2016;291(7):3145-3157. 29. Wilson E, Leszczynska K, Poulter NS, et al. RhoJ interacts with the GIT-PIX complex and regulates focal adhesion disassembly. J Cell Sci. 2014;127(Pt 14):3039-3051. 30. Pfaffl MW. A new mathematical model for

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42. 43.

44.

relative quantification in real-time RT-PCR. Nucleic Acids Res. 2001;29(9):e45. Reyat JS, Tomlinson MG, Noy PJ. Utilizing Lentiviral Gene Transfer in Primary Endothelial Cells to Assess LymphocyteEndothelial Interactions. Methods Mol Biol. 2017;1591:155-168. Manders EMM, Verbeek FJ, Aten JA. Measurement of Colocalization of Objects in Dual-Color Confocal Images. J MicroscOxford. 1993;169:375-382. Gardenier JC, Hespe GE, Kataru RP, et al. Diphtheria toxin-mediated ablation of lymphatic endothelial cells results in progressive lymphedema. JCI Insight. 2016;1(15): e84095. Simms V, Bicknell R, Heath VL. Development of an ImageJ-based method for analysing the developing zebrafish vasculature. Vascular Cell. 2017;9(1):2. Parsonage G, Machado LR, Hui JW, et al. CXCR6 and CCR5 localize T lymphocyte subsets in nasopharyngeal carcinoma. Am J Pathol. 2012;180(3):1215-1222. Pollitt AY, Poulter NS, Gitz E, et al. Syk and Src family kinases regulate C-type lectin receptor 2 (CLEC-2)-mediated clustering of podoplanin and platelet adhesion to lymphatic endothelial cells. J Biol Chem. 2014;289(52):35695-35710. Senis YA, Tomlinson MG, Ellison S, et al. The tyrosine phosphatase CD148 is an essential positive regulator of platelet activation and thrombosis. Blood. 2009;113:49424954. Braun A, Varga-Szabo D, Kleinschnitz C, et al. Orai1 (CRACM1) is the platelet SOC channel and essential for pathological thrombus formation. Blood. 2009; 113(9):2056-2063. Xu Z, Alloush J, Beck E, Weisleder N. A murine model of myocardial ischemia-reperfusion injury through ligation of the left anterior descending artery. J Vis Exp. 2014(86). Du Y, Kitzmiller JA, Sridharan A, et al. Lung Gene Expression Analysis (LGEA): an integrative web portal for comprehensive gene expression data analysis in lung development. Thorax. 2017;72(5):481-484. Zhang Y, Chen K, Sloan SA, et al. An RNAsequencing transcriptome and splicing database of glia, neurons, and vascular cells of the cerebral cortex. The Journal of neuroscience : the official journal of the Society for Neuroscience. 2014;34(36):11929-11947. Uhlen M, Fagerberg L, Hallstrom BM, et al. Proteomics. Tissue-based map of the human proteome. Science. 2015;347(6220):1260419. Fuller GL, Williams JA, Tomlinson MG, et al. The C-type lectin receptors CLEC-2 and Dectin-1, but not DC-SIGN, signal via a novel YXXL-dependent signaling cascade. J Biol Chem. 2007;282(17):12397-12409. Prakriya M, Feske S, Gwack Y, Srikanth S,

haematologica | 2019; 104(9)

Tspan18 regulates Orai1 in endothelial cells

45.

46.

47.

48.

49.

Rao A, Hogan PG. Orai1 is an essential pore subunit of the CRAC channel. Nature. 2006;443(7108):230-233. Vig M, Beck A, Billingsley JM, et al. CRACM1 multimers form the ion-selective pore of the CRAC channel. Curr Biol. 2006;16(20):2073-2079. Yeromin AV, Zhang SL, Jiang W, Yu Y, Safrina O, Cahalan MD. Molecular identification of the CRAC channel by altered ion selectivity in a mutant of Orai. Nature. 2006;443(7108):226-229. Dornier E, Coumailleau F, Ottavi JF, et al. TspanC8 tetraspanins regulate ADAM10/Kuzbanian trafficking and promote Notch activation in flies and mammals. J Cell Biol. 2012;199(3):481-496. Li J, Cubbon RM, Wilson LA, et al. Orai1 and CRAC channel dependence of VEGFactivated Ca2+ entry and endothelial tube formation. Circ Res. 2011;108(10):11901198. Zhou MH, Zheng H, Si H, et al. Stromal interaction molecule 1 (STIM1) and Orai1 mediate histamine-evoked calcium entry and nuclear factor of activated T-cells

haematologica | 2019; 104(9)

50.

51.

52.

53.

54.

(NFAT) signaling in human umbilical vein endothelial cells. J Biol Chem. 2014; 289(42):29446-29456. Saint-Pol J, Eschenbrenner E, Dornier E, Boucheix C, Charrin S, Rubinstein E. Regulation of the trafficking and the function of the metalloprotease ADAM10 by tetraspanins. Biochem Soc Trans. 2017; 45(4):937-944. Vences-Catalan F, Duault C, Kuo CC, Rajapaksa R, Levy R, Levy S. CD81 as a tumor target. Biochem Soc Trans. 2017; 45(2):531-535. Cheli Y, Jensen D, Marchese P, et al. The Modifier of hemostasis (Mh) locus on chromosome 4 controls in vivo hemostasis of Gp6-/- mice. Blood. 2008;111(3):1266-1273. Denis C, Methia N, Frenette PS, et al. A mouse model of severe von Willebrand disease: defects in hemostasis and thrombosis. Proc Natl Acad Sci U S A. 1998;95(16):95249529. Ni H, Denis CV, Subbarao S, et al. Persistence of platelet thrombus formation in arterioles of mice lacking both von Willebrand factor and fibrinogen. J Clin

Invest. 2000;106(3):385-392. 55. De Meyer SF, Savchenko AS, Haas MS, et al. Protective anti-inflammatory effect of ADAMTS13 on myocardial ischemia/reperfusion injury in mice. Blood. 2012; 120(26):5217-5223. 56. Gandhi C, Motto DG, Jensen M, Lentz SR, Chauhan AK. ADAMTS13 deficiency exacerbates VWF-dependent acute myocardial ischemia/reperfusion injury in mice. Blood. 2012;120(26):5224-5230. 57. Zuidscherwoude M, Gottfert F, Dunlock VM, Figdor CG, van den Bogaart G, van Spriel AB. The tetraspanin web revisited by super-resolution microscopy. Sci Rep. 2015; 5:12201. 58. Kanaji S, Fahs SA, Shi Q, Haberichter SL, Montgomery RR. Contribution of platelet vs. endothelial VWF to platelet adhesion and hemostasis. J Thromb Haemost. 2012; 10(8):1646-1652. 59. Verhenne S, Denorme F, Libbrecht S, et al. Platelet-derived VWF is not essential for normal thrombosis and hemostasis but fosters ischemic stroke injury in mice. Blood. 2015; 126(14):1715-1722.

1905