Haematologica, Volume 108, Issue 8

J/"#,-#%3,0%&'()*#5*-K%*-%%

4'+5%+*$/1%6,'0-")%

78&"+$%!"+$,0%9:9;<%;;=:>%%% %?*$/@+,0/%9:9;<%;;=A%%%

!"#$%0/B*/C%&0,+/##%

@'(8*##*,-%→%;#$%1/+*#*,-<%;D%1"E#%

!"#$%&'()*+"$*,-%

."&/0#%*##'/1%2'#$%"3$/0%"++/&$"-+/

F,C%&'()*+"$*,-%+,#$%

G5/%&'()*#5/0%*#%"%-,-H&0,3*$% %!,'-1"$*,-%$5"$%I//&#%$5/%% %+,#$%3,0%"'$5,0#%"#%),C%"#%&,##*()/%

6,'0-")%,3%$5/%!/00"$"H%@$,0$*%!,'-1"$*,-%

haematologica

Editor-in-Chief

Jacob M. Rowe (Jerusalem)

Deputy Editors

Carlo Balduini (Pavia), Jerry Radich (Seattle)

Associate Editors

Shai Izraeli (Tel Aviv), Steve Lane (Brisbane), Pier Mannuccio Mannucci (Milan), Pavan Reddy (Houston), David C. Rees (London), Paul G. Richardson (Boston), Francesco Rodeghiero (Vicenza), Gilles Salles (New York), Kerry Savage (Vancouver), Aaron Schimmer (Toronto), Richard F. Schlenk (Heidelberg), Sonali Smith (Chicago)

Statistical Consultant

Catherine Klersy (Pavia)

Editorial Board

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

Managing Director

Antonio Majocchi (Pavia)

Editorial Office

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

Assistant Editors

Britta Dost (English Editor), Rachel Stenner (English Editor), Anne Freckleton (English Editor), Rosangela Invernizzi (Scientific Consultant), Marianna Rossi (Scientific Consultant), Massimo Senna (Information Technology), Luk Cox (Graphic Artist)

Brief information on Haematologica

Haematologica (print edition, pISSN 0390-6078, eISSN 1592-8721) publishes peer-reviewed papers on all areas of experimental and clinical hematology. The journal is owned by a non-profit organization, the Ferrata Storti Foundation, and serves the scientific community following the recommendations of the World Association of Medical Editors (www.wame.org) and the International Committee of Medical Journal Editors (www.icmje.org).

Haematologica publishes Editorials, Original articles, Review articles, Perspective articles, Editorials, Guideline articles, Letters to the Editor, Case reports & Case series and Comments. Manuscripts should be prepared according to our guidelines (www.haematologica.org/information-for-authors), and the Uniform Requirements for Manuscripts Submitted to Biomedical Journals, prepared by the International Committee of Medical Journal Editors (www.icmje.org).

Manuscripts should be submitted online at http://www.haematologica.org/.

Conflict of interests. According to the International Committee of Medical Journal Editors (http://www.icmje.org/#conflicts), “Public trust in the peer review process and the credibility of published articles depend in part on how well conflict of interest is handled during writing, peer review, and editorial decision making”. The ad hoc journal’s policy is reported in detail at www.haematologica.org/content/policies.

Transfer of Copyright and Permission to Reproduce Parts of Published Papers. Authors will grant copyright of their articles to the Ferrata Storti Foundation. No formal permission will be required to reproduce parts (tables or illustrations) of published papers, provided the source is quoted appropriately and reproduction has no commercial intent. Reproductions with commercial intent will require written permission and payment of royalties.

Subscription. Detailed information about subscriptions is available at www.haematologica.org. Haematologica is an open access journal and access to the online journal is free. For subscriptions to the printed issue of the journal, please contact: Haematologica Office, via Giuseppe Belli 4, 27100 Pavia, Italy (phone +39.0382.27129, fax +39.0382.394705, E-mail: info@haematologica.org).

Rates of the printed edition for the year 2022 are as following:

Institutional: Euro 700

Personal: Euro 170

Advertisements. Contact the Advertising Manager, Haematologica Office, via Giuseppe Belli 4, 27100 Pavia, Italy (phone +39.0382.27129, fax +39.0382.394705, e-mail: marketing@haematologica.org).

Disclaimer. Whilst every effort is made by the publishers and the editorial board to see that no inaccurate or misleading data, opinion or statement appears in this journal, they wish to make it clear that the data and opinions appearing in the articles or advertisements herein are the responsibility of the contributor or advisor concerned. Accordingly, the publisher, the editorial board and their respective employees, officers and agents accept no liability whatsoever for the consequences of any inaccurate or misleading data, opinion or statement. Whilst all due care is taken to ensure that drug doses and other quantities are presented accurately, readers are advised that new methods and techniques involving drug usage, and described within this journal, should only be followed in conjunction with the drug manufacturer’s own published literature.

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

Table of Contents

Volume 108, Issue 8: August 2023

About the Cover

Image taken from the Editorial by Martina Seiffert in this issue.

Landmark Paper in Hematology

1973 Chasing quality remission in acute myeloid leukemia: intensity of induction and residual disease

Yishai Ofran

https://doi.org/10.3324/haematol.2023.283541

Editorials

1975 Choosing between family members is always a balancing act

Rory Bennett and John F. Seymour

https://doi.org/10.3324/haematol.2022.282628

1979 TIGIT: an immune checkpoint beyond T cells in chronic lymphocytic leukemia

Martina Seiffert

https://doi.org/10.3324/haematol.2023.282687

1982 The TRIM31 paradox: an unexpected benefit for leukemia stem cells

Jasmin Straube and Steven W. Lane

https://doi.org/10.3324/haematol.2022.282622

1984 Can we cure relapsed/refractory Hodgkin lymphoma without a stem cell transplant?

Matthew G. Mei and Alex F. Herrera

https://doi.org/10.3324/haematol.2023.282731

1986 The lymphoma microenvironment comes of age in the R-CHOP era?

Alan Cooper and Justin Kline

https://doi.org/10.3324/haematol.2022.282656

1988 CAR T-cell therapy for triple-class exposed relapsed/refractory multiple myeloma

Michele Cavo

https://doi.org/10.3324/haematol.2022.282587

1991 Nandrolone decanoate: new therapeutic option for telomeropathies?

Camilla Frieri

https://doi.org/10.3324/haematol.2022.282540

Review Articles

1993 Deep learning applications in visual data for benign and malignant hematologic conditions: a systematic review and visual glossary

Andrew Srisuwananukorn et al.

https://doi.org/10.3324/haematol.2021.280209

2011

Articles

Strategies to optimize chimeric antigen receptor T-cell therapy in hematologic malignancies: Chinese experience

Wei Sun et al.

https://doi.org/10.3324/haematol.2022.282316

2029 Acute Lymphoblastic Leukemia

RNA helicase DHX15 exemplifies a unique dependency in acute leukemia

Hao Guo et al.

https://doi.org/10.3324/haematol.2022.282066

2044 Acute Myeloid Leukemia

Comprehensive molecular and clinical characterization of NUP98 fusions in pediatric acute myeloid leukemia

Eline J. M. Bertrums et al.

https://doi.org/10.3324/haematol.2022.281653

2058 Acute Myeloid Leukemia

Impact of trisomy 19 on outcome according to genetic makeup in patients with acute myeloid leukemia

Sabine Kayser et al.

https://doi.org/10.3324/haematol.2022.282127

2066 Cell Therapy & Immunotherapy

Hemophagocytic lymphohistiocytosis and disseminated intravascular coagulation are underestimated, but fatal adverse events in chimeric antigen receptor T-cell therapy

Zhiqiang Song et al.

https://doi.org/10.3324/haematol.2022.281455

2079 Cell Therapy & Immunotherapy

Automated production of specific T cells for treatment of refractory viral infections after allogeneic stem cell transplantation

Amadeus T. Heinz et al.

https://doi.org/10.3324/haematol.2022.281996

2090 Chronic Lymphocytic Leukemia

High rate of durable responses with undetectable minimal residual disease with front-line venetoclax and rituximab in young, fit patients with chronic lymphocytic leukemia and an adverse biological profile: results of the GIMEMA phase II LLC1518 – VERITAS study

Francesca R. Mauro et al.

https://doi.org/10.3324/haematol.2022.282116

2100 Chronic Lymphocytic Leukemia

The immunomodulatory molecule TIGIT is expressed by chronic lymphocytic leukemia cells and contributes to anergy

Francesca Arruga et al.

https://doi.org/10.3324/haematol.2022.282177

2115

Hematopoiesis

The E3 ligase TRIM31 regulates hematopoietic stem cell homeostasis and MLL-AF9 leukemia

Kai Zhang et al.

https://doi.org/10.3324/haematol.2022.281955

2129 Hematopoiesis

ANKRD26 is a new regulator of type I cytokine receptor signaling in normal and pathological hematopoiesis

Francesca Basso-Valentina et al.

https://doi.org/10.3324/haematol.2022.282049

2145 Hodgkin Lymphoma

Tislelizumab with gemcitabine and oxaliplatin in patients with relapsed or refractory classic Hodgkin lymphoma: a multicenter phase II trial

Kaiyang Ding et al.

https://doi.org/10.3324/haematol.2022.282266

2154 Multiple Myeloma

PARP1 and POLD2 as prognostic biomarkers for multiple myeloma in autologous stem cell transplant

Melissa Thomas et al.

https://doi.org/10.3324/haematol.2022.282399

2166 Non-Hodgkin Lymphoma

Low T-cell proportion in the tumor microenvironment is associated with immune escape and poor survival in diffuse large B-cell lymphoma

Joo Y. Song et al.

https://doi.org/10.3324/haematol.2022.282265

2177 Non-Hodgkin Lymphoma

Integrated genetic and clinical prognostic factors for aggressive adult T-cell leukemia/lymphoma

Takuro Kameda et al.

https://doi.org/10.3324/haematol.2022.281510

2191 Cell Therapy & Immunotherapy

Adjusted comparison of outcomes between patients from CARTITUDE-1 versus multiple myeloma patients with prior exposure to proteasome inhibitors, immunomodulatory drugs and anti-CD38 antibody from the prospective, multinational LocoMMotion study of real-world clinical practice

Maria-Victoria Mateos et al.

https://doi.org/10.3324/haematol.2022.280482

2204 Red Cell Biology & its Disorders

CD169-CD43 interaction is involved in erythroblastic island formation and erythroid differentiation

Jian Bai et al.

https://doi.org/10.3324/haematol.2022.282192

Letters

2217 Treatment patterns and real-world effectiveness of rituximab maintenance in older patients with mantle cell lymphoma: a population-based analysis

Mengyang Di et al.

https://doi.org/10.3324/haematol.2022.282252

2223 Insights into dasatinib use and outcomes in real-world patients with chronic myeloid leukemia

Josephine A. Adattini et al.

https://doi.org/10.3324/haematol.2022.282237

2228 TIM3, a human acute myeloid leukemia stem cell marker, does not enrich for leukemia-initiating stem cells in B-cell acute lymphoblastic leukemia

Clara Bueno et al.

https://doi.org/10.3324/haematol.2022.282394

2233 Clinical and molecular response of acute myeloid leukemia harboring non-canonical FLT3 N676K driver mutations to contemporary FLT3 inhibitors

Gregory W. Roloff et al.

https://doi.org/10.3324/haematol.2022.282148

2239 Clinical and molecular features of familial chronic lymphocytic leukemia: a pilot monocentric study

Giulia Benintende et al.

https://doi.org/10.3324/haematol.2022.282268

Haematologica | 108 - August 2023

2243 Activity of decitabine combined with all-trans retinoic acid in oligoblastic acute myeloid leukemia: results from a randomized 2x2 phase II trial (DECIDER)

Christoph Rummelt et al.

https://doi.org/10.3324/haematol.2022.282258

2247 Efficacy, safety, and cost of mobilization strategies in multiple myeloma: a prospective, observational study

Binod Dhakal et al.

https://doi.org/10.3324/haematol.2022.282269

Case Reports & Case Series

2253 H syndrome mimicking Erdheim Chester disease: new entity and therapeutic perspectives

Hippolyte Lequain et al.

https://doi.org/10.3324/haematol.2022.282040

2259 Venetoclax for treating refractory autoimmune hemolytic anemia in chronic lymphocytic leukemia: report of two cases in Spain

Pablo Galindo-Navarro et al.

https://doi.org/10.3324/haematol.2022.281850

Chasing quality remission in acute myeloid leukemia: intensity of induction and residual disease

Yishai Ofran

Department of Hematology and Stem Cell Transplantation, Shaare Zedek Medical Center and the Faculty of Medicine, Hebrew University of Jerusalem, Jerusalem, Israel

E-mail: yofran@szmc.org.il

https://doi.org/10.3324/haematol.2023.283541

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

TITLE Improving the remission quality in acute myelogenous leukemia (AML) by increasing cytoreduction during induction therapy.

AUTHORS Arlin ZA, Cook P, Ahmed T, et al.

JOURNAL Acute Leukemias IV: Prognostic Factors and Treatment Strategies. Springer Berlin Heidelberg; 1994. Pages 231-234.

Induction of remission is the first and ultimate goal in the treatment of patients with acute myeloid leukemia (AML).

Ever since cure became achievable in AML, remission has been recognized as essential to successful treatment. Ho-

wever, for many years, toxicity associated with intensive chemotherapy limited doctors’ willingness to increase chemotherapy doses. In addition, as long as light microscopy was the only tool for marrow evaluation and mini-

Figure 1. History and future direction of treatment/response evaluation in acute myeloid leukemia. The way we evaluate response to induction therapy in acute myeloid leukemia patients has been modified in the last 50 years. Dr. Arlin's vision regarding the need to achieve "quality remission" had developed when molecular techniques allowed measurement of minimal residual disease. The accumulation of novel therapeutic agents in parallel with better understanding of leukemia biology drives improved patient outcome and potential cure. alloSCT: allogeneic stem cell transplant; MRD: minimal residual disease.

mal residual leukemia could not be measured, relapse after the achievement of morphologic remission was unpredictable. Back in the 1990s, the late Dr. Zalman Arlin was seeking ways to improve the quality of remission. Arguing that presence of unrecognized minimal residual disease (blasts that survived the induction regimen) represents the seeds of relapsed disease, he championed for the potential benefit of intensifying induction regimens, believing that this would result in a deeper or, as he called it, "quality remission”. The rationale was that susceptibility of cycling leukemia cells to the toxic effects of chemotherapy is dose-dependent, while normal hematopoietic stem cells are “dormant” and relatively protected from chemotherapy.

A similar protocol had previously been demonstrated to be effective by Dr. Arlin in acute lymphoblastic leukemia patients. To prove this concept in AML, Dr. Arlin launched a clinical trial exploring induction with higher doses of chemotherapy. Induction included five days of cytarabine 3 g/m2 over 3 hours, mitoxantrone 80 mg/m2 on either day 2 or 3 (as opposed to the usual approach of dividing such a dose over 5-6 days), and varying doses of etoposide (VP16), ranging from 50 to 150 mg/m2 on days 1, 3, and 5. Tragically, after a short illness, Dr. Arlin, succumbed to a brain tumor at the age of 47, and only preliminary results of 19 AML patients (10 newly diagnosed and 9 with relapse) were published under his name as a conference paper.1

References

1. Arlin ZA, Cook P, Ahmed T, et al. Improving the remission quality in acute myelogenous leukemia (AML) by increasing cytoreduction during induction therapy. In: Acute Leukemias IV: Prognostic Factors and Treatment Strategies. Springer Berlin Heidelberg; 1994. p. 231-234.

2. Feldman EJ, Seiter K, Traganos F, et al. Phase II evaluation of a high-dose mitoxantrone based induction regimen in untreated adults with acute myeloid leukemia. Leuk Lymphoma. 2000;38(3-4):309-315.

3. Löwenberg B, Ossenkoppele GJ, van Putten W, et al. High-dose daunorubicin in older patients with acute myeloid leukemia. N

This landmark work proved the feasibility and potential benefit of an intensified induction regimen. Final results of this regimen as a phase II study were published a few years later, demonstrating that with such an intensive regimen remission can be achieved in 36/45 (80%) of newly diagnosed young AML patients with acceptable toxicity.2 The demonstration that intensification of induction is feasible and is potentially beneficial was later confirmed by phase III studies of intensified daunorubicin.3,4 This concept also paved the way for combination therapies as induction (such as adding midostaurin or gemtuzumab ozogamicin to standard induction) and sequential therapy for resistant AML.5

Dr. Arlin's hypothesis that not all remissions are alike, and undetectable MRD is what needs to be eradicated, is now well established. The search for “quality remission” is today known as MRD eradication, and major efforts are being made to measure, monitor and eliminate disease remnants (Figure 1). Current practice and guidelines of therapy in AML incorporate sensitive molecular techniques developed after Dr. Arlin had passed away, but his memory lives on as a visionary clinical scientist who saw beyond his time and was active in the search for major breakthroughs to improve AML therapy.

Disclosure

No conflicts of interest to disclose.

Engl J Med. 2009;361(13):1235-1248.

4. Fernandez HF, Sun Z, Yao X, et al. Anthracycline dose intensification in acute myeloid leukemia. N Engl J Med. 2009;361(13):1249-1259.

5. Stelljes M, Middeke JM, Bug G, et al. In patients with relapsed/refractory AML sequential conditioning and immediate allogeneic stem cell transplantation (allo-HCT) results in similar overall and leukemia-free survival compared to intensive remission induction chemotherapy followed by allo-HCT: results from the randomized phase III ASAP Trial. Blood. 2022;140(Suppl 1):9-11.

Choosing between family members is always a balancing act

Rory Bennett1 and John F. Seymour1,2

1Department of Clinical Haematology, Royal Melbourne Hospital and Peter MacCallum Cancer Centre and 2Department of Medicine, University of Melbourne, Melbourne, Victoria, Australia

Correspondence: J.F. Seymour

john.seymour@petermac.org

Received: January 23, 2023.

Accepted: February 10, 2023. Early view: February 23, 2023.

https://doi.org/10.3324/haematol.2022.282628

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

In this edition of Haematologica, Mauro and co-authors present the results of a phase II study examining the efficacy and safety of frontline fixed-duration (12 months) combination therapy with venetoclax and rituximab in 75 fit, young patients with chronic lymphocytic leukemia (CLL).1 This approach was well-tolerated, with high overall/complete responses rates (95%/76%), a moderate rate of undetectable minimal residual disease (MRD) in the peripheral blood (69%), and no events indicative of disease progression observed after a median follow-up of 20.8 months despite 96% of patients having unmutated IGHV status, but only a small proportion of patients (12%) with TP53 disruption (deletion of 17p and/or TP53 mutation).1

A fixed-duration venetoclax combination regimen is a preferred frontline therapy for younger patients to avoid the treatment and toxicity burden associated with continuous monotherapy with a Bruton tyrosine kinase (BTK) inhibitor. Venetoclax, a B-cell lymphoma-2 inhibitor that restores intrinsic apoptosis in CLL cells, in combination with antiCD20 monoclonal antibodies obinutuzumab (VenO) or rituximab (VenR) are now established standards of care for treatment-naïve2 and relapsed/refractory CLL,3 respectively. Recently, studies (GLOW, CAPTIVATE) evaluating fixedduration frontline venetoclax-ibrutinib treatment have displayed deep remissions with encouraging progressionfree survival (PFS) with short follow-up. While the “all oral” delivery is appealing, such an approach is not clearly superior to venetoclax plus a CD20 monoclonal antibody while retaining the cardiac risks of BTK treatment and exposing the disease to both of our highly effective targeted therapies, perhaps compromising sensitivity to subsequent retreatment. Achieving a status of undetectable MRD (conventionally <10-4 leukemic cells by flow cytometry) is strongly associated with improved PFS with fixed-duration combinations. Until the recent availability of the results of the GAIA/CLL13 studies,4,5 there were no randomized comparisons of the efficacy or safety of VenO and VenR in treatment-naïve patients. The mechanistic and preclinical issues to be considered in the choice of the CD20 partner antibody have been published elsewhere,6 and here we consider the relative clinical merits of VenO and VenR.

Compared to chemoimmunotherapy, VenO has demonstrated superior efficacy in treatment-naïve patients regardless of fitness. In frail patients in the CLL14 trial, rates of undetectable MRD in the peripheral blood at 15 months following VenO were 75.5% versus 35.2% with obinutuzumab-chlorambucil (P<0.001).2 At the 5-year follow-up, the PFS of the VenO-treated patients had improved over that of the obinutuzumab-chlorambucil treated ones: 62.6% versus 27.0%.7 In fit patients in GAIA/CLL13, the rates of undetectable MRD in the peripheral blood were superior with VenO than with chemoimmunotherapy (age-stratified fludarabine-cyclophosphamide-rituximab or bendamustine-rituximab) (86.5% vs. 52.0%; P<0.0001) as was the PFS (hazard ratio [HR]=0.42, 97.5% confidence interval [CI]: 0.26-0.68; P<0.0001) after a median follow-up of 38.8 months.4,5

VenR has demonstrated superior efficacy over bendamustine-rituximab in relapsed/refractory CLL; however, no benefit over chemoimmunotherapy was observed in treatment-naïve patients. In the MURANO trial, rates of peripheral blood undetectable MRD after 9 months were 62.4% (VenR) versus 13.3% (bendamustine-rituximab), and after 5 years of follow-up a PFS benefit was observed favoring VenR (median 53.6 months vs . 17.0 months; P<0.0001).8 However, in GAIA/CLL13 rates of peripheral blood undetectable MRD at 15 months (57.0% vs 52.0%; P=0.317) and PFS at a median of 38.8 months follow-up (HR=0.79, 97.5% CI: 0.53-1.18; P=0.183) were not significantly different between fit, treatment-naïve patients without TP53 disruption who were given VenR or chemoimmunotherapy.4,5 Although in different treatment settings, the apparent disparity may have been affected by extended 24-month venetoclax therapy in MURANO, or the use of an age-stratified chemoimmunotherapy comparator in GAIA/CLL13. Although not directly compared, observed 3-year PFS rates were lower with VenR (80.8%) than with VenO (87.7%) in the latter study.5

Compared with VenR, VenO may exhibit greater efficacy in unmutated IGHV disease. Patients with unmutated IGHV, TP53 disruption, and/or genomic complexity treated with VenR in the MURANO trial had an inferior PFS to those

CLL14

Phase III: treatmentnaïve, unfit patients with CLL requiring therapy (N=432)

TP53 disruption: 13.8%

Unmutated IGHV: 59.8%

VenO (N=216) vs. ChlO (N=216)

VenO: 78.8% (N=167) vs ChlO: 76.6% (N=164)

Fatal AE

VenO: 2.4%, ChlO: 1.9%

VERITAS

Phase II: treatmentnaïve, fit patients with CLL requiring therapy, with unmutated IGHV and/or TP53 disruption

Unmutated IGHV: 96% TP53 disruption: 9 (12%)

or BR, N=79)

12-month VenR 45.3% (N=34)

Fatal AE - N=3: clinical TLS (N=1), COVID19 (N=2)

VenR: 68.0% (N=51)

VenR: 34.7% (N=26)

VenR: 12% (N=9), including 6.7% (N=5) due to COVID-19^

All - VenO: 3 pts. vs. ChlO: 5 pts.&

12.2% (N=29)

Grade ≥3VenO: 8.8% (N=20), VenR: 10.1% (N=24)

Grade ≥3 -

VenR: 1.3% (N=1)

&All biochemical tumor lysis syndrome events in the venetoclax–obinutuzumab group occurred during treatment with obinutuzumab, before exposure to venetoclax. No clinical tumor lysis syndrome events. ^Near beginning of the SARS-CoV2 pandemic prior to vaccination availability. IRR: infusion-related reactions; TLS tumor lysis syndrome; CLL: chronic lymphocytic leukemia; VenO: venetoclax-obinutuzumab; ChlO: obinutuzumab-chlorambucil; pts: patients; AE: adverse events; R/R: relapsed/refractory; VenR: venetoclax-rituximab; BR: bendamustine-rituximab; CIT: chemoimmunotherapy; FCR: fludarabine-cyclophosphamide-rituximab; COVID-19: coronavirus disease 2019.

without the markers present in a 5-year follow-up, although benefit of VenR over bendamustine-rituximab was retained in each subgroup.8 Lower rates of undetectable MRD were also observed in the subgroup with genomic complexity than in the subgroup without complex genomics.9 The TP53-disrupted subgroup had the lowest 5year PFS (70.2%). In CLL14, VenO patients with del(17p) (and/or TP53 mutation) had an inferior median PFS (median 28 months follow-up) to those without TP53 disruption, and with longer follow-up (median 52.4 months) PFS was also longer for patients with mutated IGHV than those with unmutated IGHV following VenO (HR=0.47, 95% CI: 0.25-0.87; P=0.02).10 However, only del(17p) was associated with inferior PFS on multivariate analysis.10 Although

not directly compared in GAIA/CLL13, 3-year PFS rates for both unmutated IGHV and mutated IGHV patients appear higher with VenO than with VenR (unmutated IGHV 82.9% vs. 76.4%; mutated IGHV 93.6% vs. 87.0%, respectively). Collectively these data support the superior efficacy of VenO compared with VenR for patients with treatmentnaïve CLL, including those with unmutated IGHV status. The relative efficacies in TP53-disrupted, treatment-naïve CLL remain unclear. While 24 months of treatment with VenR is still effective in relapsed/refractory TP53-disrupted CLL, data supporting 12 months of VenR in TP53disrupted treatment-naïve patients are limited.

Both VenO and VenR have demonstrated favorable side effect profiles without significant late adverse effects being

Table 1. Comparison of key safety data of venetoclax plus obinutuzumab or venetoclax plus rituximab in pivotal studies.

observed. The incidences of all grade ≥3 adverse events, including neutropenia, infections, infusion-related reactions, and tumor lysis syndrome, in the studies discussed are summarized in Table 1. Presented data from GAIA/CLL13 document a higher overall incidence of grade ≥3 adverse events with VenO than with VenR (84.6% vs 71.3%, respectively), including higher rates of grade ≥ 3 neutropenia, infections, and infusion-related reactions. No clear difference in the small numbers of fatal adverse events was observed between the treatment arms. These observations are similar to those in a large phase III study of obinutuzumab chemotherapy versus rituximab chemotherapy (both with maintenance) in follicular lymphoma.11 The incidence of severe pulmonary infection with both anti-CD20 therapies is of particular relevance in the context of the pandemic of severe acute respiratory syndrome coronavirus-2; however, detailed infection data from GAIA/CLL13 have not been reported. The risks of antiCD20 B-cell-depleting therapy are highlighted by the two COVID-19-related deaths in the VERITAS study.1 The more prolonged B-cell depletion observed after obinutuzumab may translate into a more sustained increased risk of severe infection.

Significant tumor lysis syndrome is now an uncommon occurrence in clinical trials and there is not a clear difference in incidence between patients treated with VenO or VenR. The three laboratory tumor lysis syndrome events (1.5%) observed in CLL14 occurred following obinutuzumab prior to exposure to venetoclax. Six grade ≥3 tumor lysis syndrome events (3.1%) were observed in MURANO following VenR, one of which was fatal during venetoclax ramp-up. In GAIA/CLL13, venetoclax ramp-up commenced on day 22 of cycle 1 for both VenO and VenR with similar grade ≥3 tumor lysis syndrome incidences observed by Cairo-Bishop criteria (VenO 8.8% vs. VenR 10.1%). The majority of tumor lysis syndrome events with VenO occurred prior to venetoclax ramp-up, while for VenR most occurred after initi-

References

1. Mauro FR, Starza ID, Messina M, et al. High rate of durable responses with undetectable minimal residual disease with frontline venetoclax and rituximab in young, fit patients with chronic lymphocytic leukemia and an adverse biological profile: results of the GIMEMA phase II LLC1518 - VERITAS study. Haematologica. 2023;108(8):2091-2100.

2. Fischer K, Al-Sawaf O, Bahlo J, et al. Venetoclax and obinutuzumab in patients with CLL and coexisting conditions. N Engl J Med. 2019;380(23):2225-2236.

3. Seymour JF, Kipps TJ, Eichhorst B, et al. Venetoclax–rituximab in relapsed or refractory chronic lymphocytic leukemia. N Engl J Med. 2018;378(12):1107-1120.

4. Eichhorst B, Niemann C, Kater AP, et al. A randomized phase III study of venetoclax-based time-limited combination treatments (RVe, GVe, GIVe) vs standard chemoimmunotherapy (CIT: FCR/BR) in frontline chronic lymphocytic leukemia (CLL) of fit patients: first co-primary endpoint analysis of the International

ation of venetoclax ramp-up. With this dosing schedule, one patient (1.3%) experienced tumor lysis syndrome using the more stringent Howard criteria in VERITAS.

Collectively these data suggest a trend to greater adverse events experienced with VenO compared with VenR without clearly increased treatment-related deaths or tumor lysis syndrome. Because of its more favorable safety profile, VenR may be better suited as first therapy in frail, unfit patients, and those with significant infection risk factors including risk factors for COVID-19.

Obinutuzumab is more efficacious than rituximab in combination with venetoclax as frontline therapy for CLL, independently of patients’ fitness, at the cost of increased adverse events. Although both combinations (VenO and VenR) are active against TP53-disrupted and unmutated IGHV CLL, a frontline continuous BTK inhibitor may be preferred. While the potential benefit of frontline VenR over chemoimmunotherapy in non-TP53-disrupted CLL is unclear, VenR is an effective treatment and may be preferred over VenO when safety is the highest priority. It remains uncertain whether VenR with 24 months of venetoclax would improve PFS outcomes in the frontline setting.

Disclosures

RB has no conflicts of interest to disclose. JFS is a member of the board of directors or advisory committees for Abbvie, F. Hoffman-La Roche Ltd., BMS, Gilead, Janssen, and Genor Biopharma; has acted as a consultant for TG Therapeutics, Celgene and F. Hoffman-La Roche Ltd.; has received honoraria from Abbvie, BMS, Gilead, F. Hoffman-La Roche Ltd., and Janssen; and has received research funding from and participated in speaker’s bureau for AbbVie, F. Hoffman-La Roche Ltd., and Celgene.

Contributions

RB and JFS co-wrote the manuscript.

Intergroup GAIA (CLL13) trial. Blood. 2021;138(Suppl 1):71.

5. Eichhorst B, Niemann C, Kater A. Time-limited venetoclaxobinutuzumab +/- ibrutinib is superior chemoimmunotherapy in frontline chronic lymphocytic leukaemia (CLL): PFS co-primary endpoint of the randomised phase 3 GAIA/CLL13 trial. Presented at: 2022 EHA Congress Vienna, Austria. EHA Library. Eichhorst B. 06/12/2022; 366209; LB2365.

6. Butler LA, Tam CS, Seymour JF. Dancing partners at the ball: rational selection of next generation anti-CD20 antibodies for combination therapy of chronic lymphocytic leukemia in the novel agents era. Blood Rev. 2017;31(5):318-327.

7. Al-Sawaf O, Zhang C, Robrecht S, et al. S148: Venetoclaxobinutuzumab for previously untreated chronic lymphocytic leukemia: 5-year results of the randomized CLL14 study. Hemasphere. 2022;6(S3):49-50.

8. Seymour JF, Kipps TJ, Eichhorst BF, et al. Enduring undetectable MRD and updated outcomes in relapsed/refractory CLL after

fixed-duration venetoclax-rituximab. Blood. 2022;140(8):839-850.

9. Kater AP, Wu JQ, Kipps T, et al. Venetoclax plus rituximab in relapsed chronic lymphocytic leukemia: 4-year results and evaluation of impact of genomic complexity and gene mutations from the MURANO phase III study. J Clin Oncol. 2020;38(34):4042-4054.

10. Al-Sawaf O, Zhang C, Lu T, et al. Minimal residual disease

dynamics after venetoclax-obinutuzumab treatment: extended off-treatment follow-up from the randomized CLL14 study. J Clin Oncol. 2021;39(36):4049-4060.

11. Marcus R, Davies A, Ando K, et al. Obinutuzumab for the firstline treatment of follicular lymphoma. N Engl J Med. 2017;377(14):1331-1344.

TIGIT: an immune checkpoint beyond T cells in chronic lymphocytic leukemia

Martina Seiffert

Molecular Genetics, German Cancer Research Center (DKFZ), Heidelberg, Germany

Malignant B cells from patients with chronic lymphocytic leukemia (CLL) are known for their expression of genes that are typically found in T cells, the most prominent one being ZAP70. In this issue of Haematologica, Arruga and colleagues add to the list of these genes another interesting molecule which is TIGIT: T-cell immunoreceptor with Ig and ITIM domains.1 TIGIT is an inhibitory checkpoint receptor that is up-regulated by T cells and natural killer cells upon antigen recognition and limits their response. Like PD-1, it is associated with dysfunctional or exhausted T cells in cancer and is, therefore, currently being tested as a novel immunotherapy target in several clinical trials where dual blockade of PD-(L)1 and TIGIT shows promising early results in cancer patients.2

While TIGIT functions as an inhibitory receptor, it is often co-expressed with the activating co-stimulatory receptor CD226 that works in parallel to CD28.3 CD226 contributes to the formation of the immune synapse through its interaction with its ligand CD155, which is widely expressed on tumor cells and antigen-presenting cells in the tumor microenvironment. Binding of CD226 to CD155 leads to Tcell activation, which is blocked by the presence of TIGIT via its interaction with CD155 (Figure 1A).4 Recent studies showed that both PD-1 and TIGIT disrupt activation of the co-stimulatory receptor CD226 through distinct mechanisms.5 This provides a mechanistic rationale for the dual blockade of PD-(L)1 and TIGIT to reinvigorate anti-tumor activity of CD8+ T cells in cancer immunotherapy.

Besides its abundant expression in activated and exhausted T cells, TIGIT expression was observed in memory B cells where it is essential for effective immune regulation.6 TIGIT-positive memory B cells also express additional inhibitory molecules, including IL-10, PD-L1, and CD39/CD73, molecules that are expressed also by CLL cells. In line with this, Arruga et al. show that TIGIT expression in CLL cells correlates with IL-10 expression. As CLL cells resemble activated, antigen-experienced B cells and share a transcriptional profile with memory B cells,7 the expression of TIGIT might be part of this signature

Correspondence: M.

Seiffert m.seiffert@dkfz.de

Received: February 21, 2023.

Accepted: February 27, 2023.

Early view: March 9, 2023.

https://doi.org/10.3324/haematol.2023.282687

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

and, therefore, inherited from the cell-of-origin of CLL. Furthermore, TIGIT expression in memory B cells is induced via B-cell receptor (BCR) or CD40 engagement, and the latter results in the strongest TIGIT upregulation.8 CD40 stimulation of B cells occurs in secondary lymphoid tissues via CD40-ligand expressed by follicular helper T cells (TFH). Interestingly, TIGIT on B cells limits the proliferation of TFH, likely acting as a negative feedback mechanism to prevent overshooting immune responses (Figure 1B). In agreement with this, mice lacking TIGIT expression in B cells develop severe experimental autoimmune encephalomyelitis, suggesting an important role for TIGITexpressing B cells in immune tolerance.9

Survival and proliferation of CLL cells is driven by stimuli from the microenvironment, with BCR and CD40 stimulation being the most prominent signals.10 CD4+ T cells, including TFH in lymph nodes expressing CD40-ligand, are important mediators of these stimuli. Considering that TIGIT expression in memory B cells is triggered by such signals, it is likely that the expression of TIGIT in CLL cells is also induced in the lymph node niche via the interaction with TFH and other immune cells. In line with this, Arruga and colleagues observed that blockade of BCR signaling by the Bruton's tyrosine kinase inhibitor ibrutinib leads to a downregulation of TIGIT expression in CLL cells. They further showed that TIGIT expression in CLL cells was inversely correlated with BCR expression and, therefore, associated with non-responsiveness or anergy of CLL cells upon BCR stimulation. Accordingly, CLL cells from patients treated with ibrutinib showed an up-regulated surface IgM expression associated with a loss of TIGIT expression and recovery from anergy. Interestingly, high TIGIT expression in lymph node biopsies of CLL patients was associated with a lower proliferation rate of CLL cells, whereas CD226 expression was linked to greater proliferation, which is in line with the concept that TIGIT induces anergy of CLL cells. In support of this, Arruga et al. show that a high TIGIT to CD226 ratio was predictive for good prognosis in CLL,1 whereas TIGIT expression was low or absent and CD226

Figure 1. TIGIT has a modulatory role in T cells, memory B cells, and chronic lymphocytic leukemia cells. (A) The interaction of TIGIT expressed on T cells with CD155 expressed on antigen-presenting cells (APC) or tumor cells limits T-cell activity and thereby contributes to T-cell exhaustion. (B) TIGIT is also expressed on a subset of memory B cells where it limits proliferation of CD155-expressing follicular helper T cells (TFH) and tissue inflammation. (C) In patients with chronic lymphocytic leukemia (CLL), expression of TIGIT is observed on malignant B cells and CD155 mainly on monocytes. Here, TIGIT expression is associated with B-cell anergy and better outcome of patients as it limits B-cell receptor (BCR) signaling and proliferation of CLL cells.

expression was high in Richter’s syndrome, an aggressive transformation of CLL that is driven by highly proliferative B cells.11 Altogether, this suggests that TIGIT in CLL cells is linked to anergy and a limited proliferation rate (Figure 1C). However, a causal relationship and the underlying mechanisms of this link remain unresolved.

As in T cells, TIGIT might limit B-cell activation and proliferation as part of an immune shutdown mechanism that is important to limit immune-mediated tissue damage. Exploring the role of TIGIT, and probably also other immune checkpoint molecules in B cells, will be important to better understand their potential role in B-cell transformation. The observations of Arruga et al. raise the question as to if and how the expression of TIGIT and other immune checkpoint

molecules in malignant B cells influences treatment response to immune checkpoint inhibitors. While blockade of TIGIT might increase T-cell activity, and, therefore, immune control of CLL, it might further release anergy of CLL cells, and increase BCR and other signaling activity. This might lead to an increased proliferation rate and transformation to a more aggressive disease. It is hard to predict what the net outcome of such counteracting activities of TIGIT blockade will be, and future studies are necessary to improve our understanding of the role of immune checkpoints including TIGIT in B-cell malignancies.

Disclosures

No conflicts of interest to disclose.

1. Arruga F, Rubin M, Papazoglou D, et al. The immunomodulatory molecule TIGIT is expressed by chronic lymphocytic leukemia cells and contributes to anergy. Haematologica. 2023;108(8):2101-2115.

2. Cho BC, Abreu DR, Hussein M, et al. Tiragolumab plus atezolizumab versus placebo plus atezolizumab as a first-line treatment for PD-L1-selected non-small-cell lung cancer (CITYSCAPE): primary and follow-up analyses of a randomised, double-blind, phase 2 study. Lancet Oncol. 2022;23(6):781-792.

3. Joller N, Hafler JP, Brynedal B, et al. Cutting edge: TIGIT has T cell-intrinsic inhibitory functions. J Immunol. 2011;186(3):1338-1342.

4. Yu X, Harden K, Gonzalez LC, et al. The surface protein TIGIT suppresses T cell activation by promoting the generation of mature immunoregulatory dendritic cells. Nat Immunol. 2009;10(1):48-57.

5. Banta KL, Xu X, Chitre AS, et al. Mechanistic convergence of the TIGIT and PD-1 inhibitory pathways necessitates co-blockade to optimize anti-tumor CD8(+) T cell responses. Immunity. 2022;55(3):512-526.e9.

6. Hasan MM, Nair SS, O'Leary JG, et al. Implication of TIGIT(+) human memory B cells in immune regulation. Nat Commun. 2021;12(1):1534.

7. Klein U, Tu Y, Stolovitzky GA, et al. Gene expression profiling of B cell chronic lymphocytic leukemia reveals a homogeneous phenotype related to memory B cells. J Exp Med. 2001;194(11):1625-1638.

8. Asashima H, Axisa P-P, Pham THG, et al. Impaired TIGIT expression on B cells drives circulating follicular helper T cell expansion in multiple sclerosis. J Clin Invest. 2022;132(20):e156254.

9. Xiao S, Bod L, Pochet N, et al. Checkpoint receptor TIGIT expressed on Tim-1+ B cells regulates tissue inflammation. Cell Rep. 2020;32(2):107892.

10. Ten Hacken E, Burger JA. Microenvironment interactions and Bcell receptor signaling in chronic lymphocytic leukemia: implications for disease pathogenesis and treatment. Biochim Biophys Acta. 2016;1863(3):401-413.

11. Condoluci A, Rossi D. Richter Syndrome. Curr Oncol Rep. 2021;23(3):26.

The TRIM31 paradox: an unexpected benefit for leukemia stem cells

Jasmin Straube1,2 and Steven W. Lane1,2

1QIMR Berghofer Medical Research Institute and 2The University of Queensland, Brisbane, Queensland, Australia

Correspondence: S.W. Lane

steven.lane@qimrberghofer.edu.au

Received: February 27, 2023.

Accepted: March 7, 2023. Early view: March 16, 2023.

https://doi.org/10.3324/haematol.2022.282622

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Acute myeloid leukemia (AML) is an aggressive blood cancer that arises due to mutations in hematopoietic stem cells (HSC). AML is genetically heterogeneous, and this genetic information is now used in the clinic to determine prognosis and for targeted therapeutic intervention where appropriate. Conversely, post-translational factors that regulate AML growth and survival, such as the role of E3 ubiquitin ligases, are less well understood.

E3 ubiquitin ligases are mediators of protein ubiquitination and consequent degradation, acting via recruitment of an E2 ubiquitin conjugating enzyme to a recognized substrate. The polyubiquinated substrate is then marked among others for proteasome degradation and its role is described in cell cycle regulation and numerous other biological processes.

Among the hundreds of E3 ligases, few have been studied

Figure 1. Loss of Trim31 has distinct effects on normal and malignant hematopoietic stem cells. Trim31-loss leads to accumulation of CDK8 in the cells leading to upregulation of Pbx1 and Ccnd1. In normal hematopoietic stem cells (HSC) this results in increased cell cycling with loss of proliferation and self-renewal capacity. On the contrary, in MLL-AF9-driven acute myeloid leukemia (AML) cells, Trim31-loss leads to increased proliferation, expansion of leukemic stem cells, and reduced survival through more aggressive disease.

in the hematopoietic system but these have an important role in HSC self-renewal capacity and hematopoietic stem cell homeostasis. For example, loss of E3 ubiquitinase Fbw7 leads to differential expression of mediators of cell cycle entry (Myc, Notch1, cyclin E), loss of stem cell quiescence and HSC self-renewal capacity, and depletion of HSC.1 Loss of E3 ubiquitin ligase Itch, although also driving a hyperproliferative HSC phenotype, was also superior in maintaining HSC repopulation capacity.2 Levels of the E3 ubiquitin ligase TRIM31 impact prognosis and progression in solid cancers.

In the report published in Haematologica, Zhang et al.3 sought to determine whether the loss of TRIM31 impacted normal hematopoiesis and the development of leukemia. Using a series of phenotypical analysis and transplantation assays, they demonstrated that TRIM31 was required for long-term (LT)-HSC maintenance and competitive repopulation capacity. This functional change was mediated by increased cell cycling and failure to retain a quiescent LT-HSC population that is required to maintain HSC recovery.4 Next, the authors sought to use publically available patient gene expression and clinical data to validate these findings. This revealed that patients with low TRIM31 expression had inferior overall survival and, interestingly, this was mostly found in patients with MLL-translocated AML. This unexpected finding prompted the authors to assess the effect of TRIM31 loss in a model of MLL-translocated AML. Consistent with this human data, but in contrast to the findings in normal hematopoiesis, loss of TRIM31 actually accelerated MLL-AF9-induced AML.

The authors examined the underlying mechanism by which TRIM31 loss drives HSC functional decline. First, to

References

1. Thompson BJ, Jankovic V, Gao J, et al. Control of hematopoietic stem cell quiescence by the E3 ubiquitin ligase Fbw7. J Exp Med. 2008;205(6):1395-1408.

2. Rathinam C, Matesic LE, Flavell RA. The E3 ligase Itch is a negative regulator of the homeostasis and function of hematopoietic stem cells. Nat Immunol. 2011;12(5):399-407.

3. Zhang K, Liu D, Li Y, et al. The E3 ligase TRIM31 regulates hematopoietic stem cell homeostasis and MLL-AF9 leukemia. Haematologica. 2023;108(8):2116-2129.

identify substrates of TRIM31, they performed mass spectrometry in wild-type (WT) and TRIM31-/- cells. CDK8 was a direct interaction partner of TRIM31 and was substantially more abundant upon loss of TRIM31. Consistently, CDK8 ubiquitination was decreased in TRIM31-/- cells. To confirm the role of CDK8 activation in LT-HSC functional decline, Zhang et al. used a CDK8 inhibitor and genetic approaches to modulate CDK8. In both cases they were able to restore LT-HSC function, which was evident through increased donor-derived chimerism in blood and the HSC compartment. It remains to be determined what the key factors are that regulate TRIM31 expression in AML.

Altogether, this work shows that loss of TRIM31 has distinct and opposite effects in normal and malignant hematopoiesis (Figure 1). As these pathways can now be targeted therapeutically, this work also identifies a need for further study examining how various E3 ubiquitin ligases control cell cycle and HSC/AML function.

With the advent of next generation sequencing, there have been major breakthroughs in understanding recurrent genetic mutations and prognostic schemas,5 and targeted therapies (e.g., FLT3 inhibitors). This work highlights the importance of understanding protein stability and degradation through ubiquitin ligases in AML, particularly how this might lead to cell cycle dysregulation and, potentially, chemotherapy resistance in AML.6

Disclosures

No conflicts of interest to disclose.

Contributions

Both authors contributed equally.

4. Austin RJ, Straube J, Bruedigam C, et al. Distinct effects of ruxolitinib and interferon-alpha on murine JAK2V617F myeloproliferative neoplasm hematopoietic stem cell populations. Leukemia. 2020;34(4):1075-1089.

5. Straube J, Ling VY, Hill GR, Lane SW. The impact of age, NPM1(mut), and FLT3(ITD) allelic ratio in patients with acute myeloid leukemia. Blood. 2018;131(10):1148-1153.

6. Ling VY, Straube J, Godfrey W, et al. Targeting cell cycle and apoptosis to overcome chemotherapy resistance in acute myeloid leukemia. Leukemia. 2023;37(1):143-153.

Can we cure relapsed/refractory Hodgkin lymphoma without a stem cell transplant?

Matthew G. Mei and Alex F. Herrera

Division of Lymphoma, Department of Hematology and Hematopoietic Cell Transplantation, City of Hope National Medical Center, Duarte, CA, USA

Correspondence: A.F. Herrera

aherrera@coh.org

Received: February 1, 2023.

Accepted: March 28, 2023.

Early view: April 6, 2023.

https://doi.org/10.3324/haematol.2023.282731

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

A remarkable development in the treatment of classic Hodgkin lymphoma (cHL) has been the striking efficacy observed when PD-1 blockade is combined with cytotoxic chemotherapy. In this issue of Haematologica, Ding and colleagues publish results of a single-arm trial testing the safety and effi cacy of tislelizumab, gemcitabine, and oxaliplatin (T-GemOx) in relapsed and/or refractory (R/R) cHL.1 All 30 evaluable patients responded, and only one patient did not achieve complete remission (CR). To date, this is the third published data set of combined PD-1 blockade and combination chemotherapy in R/R cHL featuring a best CR rate exceeding 90%. However, in addition to its role in the evolving story of PD-1 blockade and chemotherapy in cHL, this study also provides a potential glimpse into the future, as no patient was consolidated with autologous stem cell transplant (autoSCT), yet the 12-month progression-free survival (PFS) was 96%.

Published trials of PD-1 with chemotherapy in the R/R setting include pembrolizumab with gemcitabine, vinorelbine, and liposomal doxorubicin (pembro-GVD), as well as nivolumab with ifosfamide with carboplatin, ifosfamide, and etoposide (NICE). 2,3 The high CR rate with TGemOx is consistent with the data from pembro-GVD and NICE, as all three studies featured best CR rates exceeding 90%. Nonetheless, pembro-GVD and NICE were tested strictly in the second-line setting with a goal of bridging patients to curative autoSCT. These are the first prospective data evaluating PD-1 blockade with chemotherapy beyond second-line treatment and also as a destination treatment without consolidative autoSCT. T-GemOx was administered to a patient population that was signi ficantly more heavily pre-treated than those in the pembro-GVD and NICE trials, nearly half of whom (43.3%) had received at least two prior lines of therapy, and 16.7% of whom had undergone prior autoSCT. Even still, in this high-risk group of R/R cHL patients, the best CR rate and 12-month PFS both exceeded 95% with a median followup of 15.8 months, remarkable results that are comparable to those obtained with pembro-GVD and NICE.

The results of this trial raise at least two important questions. One is whether or not there is, in fact, an optimal choice of anti-PD-1 antibody and speci fi c cytotoxic chemotherapies, and the other is whether or not anti-PD1 combined with chemotherapy could potentially represent curative therapy in the R/R setting without subsequent consolidative autoSCT. Pembro-GVD, NICE, and T-GemOx all use different anti-PD-1 agents and very different chemotherapy backbones, yet return very similar results. As patients enrolled on this trial were more heavily pre-treated than in NICE and pembro-GVD, both of which restricted enrollment to patients after one prior line of therapy, is it possible that there is something special about T-GemOx in comparison to the other two regimens? Tiselizumab is specifically engineered to decrease FcγR binding on macrophages via a mutated Fc region with a single-agent CR rate of 62.9%. This is numerically higher than the CR rates observed with either nivolumab or pembrolizumab.4-6 Gemcitabine and oxaliplatin have immunogenic properties including, but not limited to, depletion of myeloid-derived suppressor cells and promotion of cytotoxic lymphocyte-driven responses,7 mechanisms which may augment synergy with PD-1 blockade. However, a trial comparing these regimens head-to-head is impractical, and off-protocol use of T-GemOx is restricted by the fact that tislelizumab is currently not approved for use outside China. GemOx is also not a widely used standard salvage regimen for R/R cHL, although the various regimens used in studies of anti-PD1-based salvage therapy refl ect the lack of a clear standard chemotherapy salvage backbone in cHL. One important question that will be answered definitively with a randomized trial is whether the addition of PD-1 blockade to salvage chemotherapy improves effi cacy compared to chemotherapy alone as initial salvage therapy for R/R cHL (Co-operative Group Trial EA4211; https://clinicaltrials.gov/ct2/show/NCT05711628).

Perhaps the more provocative and compelling question is whether or not a subset of patients with R/R cHL can be

cured without consolidative autoSCT. As modern salvage regimens built around PD-1 blockade deliver unprecedented rates of complete metabolic response, a previously unimaginable possibility has now become one of the most critical unanswered questions in cHL. The long-term durable remission rate when PD-1 blockade is used as monotherapy in later lines of therapy is low.8-10 Will the combination of PD-1 blockade with moderate dose, but not high-dose chemotherapy, be sufficient for cure? How much chemotherapy is enough? As consolidative autoSCT was not performed for patients treated with T-GemOx, the results presented by Ding and colleagues are an incremental but important step in exploring long-term outcomes of anti-PD1-based salvage therapy without autoSCT. However, a relatively short median follow-up, that is shorter than the planned two years of tislelizumab maintenance, means that these results can only be taken as an early provocative signal of a potential pathway towards cure without autoSCT in R/R cHL. Other efforts in this area are ongoing; there is a cohort of the pembro-GVD study evaluating ongoing pembrolizumab maintenance without planned autoSCT (clinicaltrials.gov: NCT03618550), and brentuximab vedotin and nivolumab are being studied in the second-line setting specifically for patients who are ineligible for or who refuse autoSCT (clinicaltrials.gov: NCT04561206). Further optimization of therapy in R/R cHL

References

1. Ding K, Liu H, Ma J, et al. Tislelizumab with gemcitabine and oxaliplatin in patients with relapsed or refractory classic Hodgkin lymphoma: a multicenter phase II trial. Haematologica. 2023;108(8):2146-2154.

2. Moskowitz AJ, Shah G, Schöder H, et al. Phase II trial of pembrolizumab plus gemcitabine, vinorelbine, and liposomal doxorubicin as second-line therapy for relapsed or refractory classical Hodgkin lymphoma. J Clin Oncol. 2021;39(28):3109-3117.

3. Mei MG, Lee HJ, Palmer JM, et al. Response-adapted anti-PD-1based salvage therapy for Hodgkin lymphoma with nivolumab alone or in combination with ICE. Blood. 2022;139(25):3605-3616.

4. Song Y, Gao Q, Zhang H, et al. Treatment of relapsed or refractory classical Hodgkin lymphoma with the anti-PD-1, tislelizumab: results of a phase 2, single-arm, multicenter study. Leukemia. 2020;34(2):533-542.

5. Younes A, Santoro A, Shipp M, et al. Nivolumab for classical Hodgkin's lymphoma after failure of both autologous stem-cell transplantation and brentuximab vedotin: a multicentre, multicohort, single-arm phase 2 trial. Lancet Oncol. 2016;17(9):1283-1294.

6. Armand P, Shipp MA, Ribrag V, et al. Programmed death-1 blockade with pembrolizumab in patients with classical Hodgkin lymphoma after brentuximab vedotin failure. J Clin

will likely be bolstered by incorporation of more quantitative measures of disease burden such as tumor metabolic volume and circulating tumor DNA, both of which have shown significant promise in refining prognostication in cHL and may enable response-adapted approaches.11,12 In spite of the above limitations, Ding and colleagues have conducted an important trial which continues to build upon the remarkable story of combining PD-1 blockade and chemotherapy while potentially helping to usher in curative therapy for R/R cHL without autoSCT.

Disclosures

AFH has received research funding from BMS, Merck, Genentech Inc./F. Hoffmann-La Roche Ltd., Gilead Sciences, Seattle Genetics, AstraZeneca and ADC Therapeutics, and consultancy fees from BMS, Merck, Genentech Inc./F. Hoffmann-La Roche Ltd., Kite Pharma/Gilead, Seattle Genetics, Karyopharm, Takeda, Tubulis, AstraZeneca, Genmab, Regeneron, Pfizer, Abbvie, Adicet Bio and Caribou. MM has received research funding from BMS, Incyte and Morphosys, consultancy fees from Novartis, Seattle Genetics, CTI, Janssen, EUSA, and has been a member of the Speakers’ Bureau for Morphosys and for Seattle Genetics.

Contributions

MGM and AFH wrote and edited the manuscript.

Oncol. 2016;34(31):3733-3739.

7. Galluzzi L, Buqué A, Kepp O, Zitvogel L, Kroemer G. Immunological effects of conventional chemotherapy and targeted anticancer agents. Cancer Cell. 2015;28(6):690-714.

8. Chen R, Zinzani PL, Lee HJ, et al. Pembrolizumab in relapsed or refractory Hodgkin lymphoma: 2-year follow-up of KEYNOTE087. Blood. 2019;134(14):1144-1153.

9. Armand P, Kuruvilla J, Michot J-M, et al. KEYNOTE-013 4-year follow-up of pembrolizumab in classical Hodgkin lymphoma after brentuximab vedotin failure. Blood Adv. 2020;4(12):2617-2622.

10. Armand P, Engert A, Younes A, et al. Nivolumab for relapsed/refractory classic Hodgkin lymphoma after failure of autologous hematopoietic cell transplantation: extended follow-up of the multicohort single-arm phase II CheckMate 205 trial. J Clin Oncol. 2018;36(14):1428-1439.

11. Yhim H-Y, Eshet Y, Metser U, et al. Risk stratification for relapsed/refractory classical Hodgkin lymphoma integrating pretransplant Deauville score and residual metabolic tumor volume. Am J Hematol. 2022;97(5):583-591.

12. Kurtz DM, Soo J, Co Ting Keh L, et al. Enhanced detection of minimal residual disease by targeted sequencing of phased variants in circulating tumor DNA. Nat Biotechnol. 2021;39(12):1537-1547.

The lymphoma microenvironment comes of age in the R-CHOP era?

Alan Cooper and Justin Kline

Department of Medicine, Section of Hematology/Oncology, University of Chicago, Chicago, IL, USA

Correspondence: J. Kline

jkline@medicine.bsd.uchicago.edu

Received: February 23, 2023.

Accepted: March 7, 2023.

Early view: March 16, 2023.

https://doi.org/10.3324/haematol.2022.282656

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

The lymphoma microenvironment (LME) in diffuse large B-cell lymphoma (DLBCL) remains poorly characterized. While it has been increasingly established that DLBCL tumors vary in their degree of immune cell infiltration, how this relates to specific underlying genetic alterations, oncogenic pathways, and prognosis is less well understood. As chimeric antigen receptor (CAR) T-cell therapy and bispecific T-cell engagers (BiTE) therapy gain or are expected to gain approval for use in relapsed/refractory DLBCL, it is critical for those in the field to work towards an improved understanding of the DLBCL LME, its impact on treatment outcomes, and how it might be leveraged for successful therapeutic intervention. In this issue of Haematologica, Song et al.1 analyzed the LME of 57 newly-diagnosed, de novo DLBCL cases through multispectral immunofluorescence (mIF) and enumerated several immune cell types, including CD4+ T cells, CD8+ T cells, natural killer (NK) cells, and macrophages. The presence of these cell types was then correlated with R-CHOP therapy efficacy. While proportions of NK cells and macrophages in the LME were not correlated with survival outcomes, DLBCL cases harboring low T-cell proportions had significantly inferior survival than those with a high proportion of T cells. This striking survival difference was accentuated when considering proportions of CD4+PD-1+ T cells and CD8+PD-1+ T cells, suggesting a role for these putative exhausted cells in contributing to therapy resistance in DLBCL. These findings were validated with CIBERSORTx immune cell deconvolution in an independent R-CHOP-treated DLBCL cohort for which gene expression profiling was available. Importantly this overall trend for “cold” DLBCL cases is consistent with other mIF data2 and a recent large-scale transcriptomic study from Kotlov et al., 3 where DLBCL cases with a “depleted” LME were associated with significantly worse survival to rituximabbased chemoimmunotherapy.

The authors next correlated the proportions of immune cell populations in the LME with protein expression of B2M, HLA-I, and HLA-II. Interestingly, DLBCL cases with

decreased expression of these proteins contained fewer T cells, consistent with previous observations in DLBCL.4 Genetic alterations in B2M, as well as CD58, FAS, and TNFRSF14, were also collectively associated with lower proportions of T cells. Notably, these alterations are thought to contribute to distinct mechanisms of immune evasion; inactivation of B2M and CD58 together enable evasion from both CD8+ T cells and NK cells,5 whereas inactivation of FAS and TNFRSF14 is thought to enable evasion from T-follicular helper cell-mediated deletion during the germinal center reaction.6 Still, it remains unclear why these “cold” DLBCL tumors acquire genetic alterations to evade immune destruction. It may be that these lymphomas were “hot” early in their development after which the acquisition of additional genetic alterations (including those studied here) was sufficient to mediate immune cell exclusion as the disease progressed. In spite of this, while these mutations are interesting, they fail to elucidate the oncogenic pathways that define “hot” and “cold” lymphomas. For example, gain-of-function mutations in EZH2 and a transcriptional signature of doublehit lymphoma (DHITsig) have both been associated with “cold” microenvironments in the germinal center B-cell (GCB) subtype of DLBCL.4,7 However, neither of these two features significantly correlated with differences in the proportion of T cells in the current study, suggesting that there may be other oncogenic pathways associated with “cold” DLBCL tumors that have yet to be clarified. These new pathways, in turn, may explain their poorer response to chemoimmunotherapy.

This study makes clear that T-cell-enriched DLBCL cases are associated with improved survival outcomes following R-CHOP therapy, which the authors hypothesize could be related to chemotherapy-induced immunogenic cell death and a re-energized anti-lymphoma immune response. However, the authors’ suggestion to add immunotherapies to the R-CHOP backbone in the hopes of prolonging survival for “hot” DLBCL cases or perhaps improving outcomes for “cold” DLBCL cases should be met with

caution. A recent phase Ib/II trial testing the combination of the anti-PD-L1 antibody, atezolizumab, with R-CHOP reported that the regimen did not lead to a noticeably higher complete response rate and introduced additional immune-related adverse events.8 Results from ongoing trials testing the addition of an anti-CD20 BiTE to R-CHOP, such as those incorporating epcoritamab (clinicaltrials.gov identifier: NCT04663347) and glofitamab (clinicaltrials.gov identifier: NCT03467373), however, may show otherwise. While the work of Song and colleagues is limited by its small sample size and narrow analysis of immune cell subsets, it represents an important analysis of the prognostic value of the LME in treatment-naïve DLBCL. While

References

1. Song JY, Nwangwu M, He T-F, et al. Low T-cell proportion in the tumor microenvironment is associated with immune escape and poor survival in diffuse large B-cell lymphoma. Haematologica. 2023;108(8):2167-2177.

2. Xu-Monette ZY, Xiao M, Au Q, et al. Immune profiling and quantitative analysis decipher the clinical role of immunecheckpoint expression in the tumor immune microenvironment of DLBCL. Cancer Immunol Res. 2019;7(4):644-657.

3. Kotlov N, Bagaev A, Revuelta MV, et al. Clinical and biological subtypes of B-cell lymphoma revealed by microenvironmental signatures. Cancer Discov. 2021;11(6):1468-1489.

4. Ennishi D, Takata K, Béguelin W, et al. Molecular and genetic characterization of MHC deficiency identifies EZH2 as therapeutic target for enhancing immune recognition. Cancer Discov.

it is clear that there is a dichotomy between “hot” and “cold” DLBCL cases, the extent to which T cells in the LME drive this difference in survival remains poorly defined. Future studies must focus on better understanding the molecular correlates of these phenotypes and should strive to develop therapies that realize the unmet need of this at-risk “cold” DLBCL population.

Disclosures

No conflicts of interest to disclose.

Contributions

Both authors contributed equally.

2019;9(4):546-563.

5. Challa-Malladi M, Lieu YK, Califano O, et al. Combined genetic inactivation of β2 microglobulin and CD58 reveals frequent escape from immune recognition in diffuse large B-cell lymphoma. Cancer Cell. 2011;20(6):728-740.

6. Razzaghi R, Agarwal S, Kotlov N, et al. Compromised counterselection by FAS creates an aggressive subtype of germinal center lymphoma. J Exp Med. 2021;218(3):e20201173.

7. Ennishi D, Jiang A, Boyle M, et al. Double-hit gene expression signature defines a distinct subgroup of germinal center B-cell-like diffuse large B-cell lymphoma. J Clin Oncol. 2019;37(3):190-201.

8. Younes A, Burke JM, Cheson BD, et al. Safety and efficacy of atezolizumab with rituximab and CHOP in previously untreated diffuse large B-cell lymphoma. Blood Adv. 2023;7(8):1488-1495.

CAR T-cell therapy for triple-class exposed relapsed/refractory multiple myeloma

Michele Cavo

The treatment landscape for multiple myeloma (MM) has expanded progressively during the past two decades to include multiple proteasome inhibitors (PI), immunomodulators (IMiD), and monoclonal antibodies (MoAb). These agents are the pillars of modern MM therapy for both newly-diagnosed patients and those with relapsed/refractory disease. When given simultaneously, these agents allowed enhanced rates and depth of durable responses to be achieved, even in more advanced phases of the disease. Not surprisingly, these benefits led to significant improvements in progression-free survival (PFS) and overall survival (OS).1 Nevertheless, most patients eventually relapse and become progressively refractory to the main classes of agents during successive lines of therapy. For many years, no standard of care treatment was established in real-world clinical practice for patients exposed or refractory to at least an IMiD, a PI and a MoAb, referred to as triple-class exposed or triple-class refractory, who are candidates to receive T-cell redirecting therapies. Results from the retrospective MAMMOTH study demonstrated the poor outcomes for these therapeutically challenging patients, highlighting the need for more effective treatments with novel mechanisms of action.2 CAR Tcell therapies are likely to provide significant benefits in this setting, based on deep and durable disease control as shown in exploratory phase II clinical trials.3,4 However, the absence of a control arm in the KarMMa and CARTITUDE1 studies, which led to approval of idecabtagene vicleucel (ide-cel)3 and cilcabtagene autoleucel (cilta-cel),4 respectively, by regulatory agencies, raised the need for indirect evidence on the relative effectiveness of these novel therapies compared with real-life treatments. For this purpose, indirect treatment comparisons are possible by creating an external control arm from either real-world or clinical trial data sources. To avoid clinical outcome estimates that are biased by imbalances in baseline prognostic characteristics of non-randomized cohorts of patients, statistical methods that account for such confounding biases should be applied.

Correspondence: M. Cavo michele.cavo@unibo.it

Received: March 10, 2023.

Accepted: March 23, 2023.

Early view: April 6, 2023.

https://doi.org/10.3324/haematol.2022.282587

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

In the current issue of Haematologica, Mateos et al.5 report the results of a study aimed at retrospectively comparing the effectiveness of cilta-cel in the context of the single-arm of the CARTITUDE-1 study with real-world data extracted from the LocoMMotion study,6 the fi rst prospective, non-interventional, real-life study of tripleclass exposed MM patients. Overall, 113 patients were enrolled in CARTITUDE-1, and 97 of these were infused with cilta-cel after a mean of 52 days from the date of apheresis, while the remaining 16 patients discontinued the study after apheresis. The LocoMMotion study involved a total of 248 patients from European countries and the US who were triple-class exposed, the majority of them also triple-class refractory, after a median of four prior lines of therapy. Based on physicians’ choice, these patients received 92 treatments, each of them unique to the individual patient, a finding which reflects the lack of a standard of care therapy in this setting.

Matched-adjusted comparisons of individual patient data from CARTITUDE-1 and LocoMMotion studies were performed using inverse probability weighting methods to estimate the average treatment effect in the respective cohorts of patients. For the purposes of the study, two analyses were performed. The first of these involved the so-called infused/aligned cohorts and was aimed at comparing individual patient data from the set of 97 patients who were treated with cilta-cel with the aligned cohort of 170 patients from LocoMMotion who were progression-free and alive 52 days after start of treatment. This time period corresponded to the average time during which patients were required to be progression-free and alive in order to receive cilta-cel infusion in CARTITUDE-1, and was chosen to align the LocoMMotion cohort with the set of patients from CARTITUDE-1. The observed rates of response, including complete response or higher, at least very good partial response and partial response, were significantly higher in the cilta-cel-treated group (82.5%, 94.8%, and 97.9%, respectively) compared to the real-world treated group (0.6%, 17.6%, and 42.9%, respectively). Adjusted

IRCCS Azienda Ospedaliero-Universitaria di Bologna, Istituto di Ematologia “Seràgnoli” and Dipartimento di Scienze Mediche e Chirurgiche, Università di Bologna, Bologna, Italy

comparisons between the two cohorts showed that patients treated with cilta-cel were 5.7 times (95% CI: 3.258.08; P<0.0001) more likely to achieve at least a very good partial response than patients treated in the real-world clinical practice. The observed median PFS for this latter group was 4.3 months, while it was not reached in the cilta-cel-treated group. Adjusted and unadjusted hazard ratios (HR) for the set of patients who received CAR T-cell therapy compared to conventionally treated patients in real-world clinical practice were 0.15 (95% CI: 0.08-0.29; P<0.0001) and 0.19 (95% CI: 0.12-0.29; P<0.0001), respectively. Following adjusted comparison for OS, a reduced risk of death by 80% (HR 0.20, 95% CI: 0.09-0.41; P<0.0001) favored patients treated with cilta-cel versus the real-world treated group (who had an observed median OS of 11.3 months), a finding which supported the results of the unadjusted comparison between these groups. The second analysis included the overall cohorts of 113 and 248 patients enrolled on the CARTITUDE-1 and LocoMMotion studies, respectively, and substantially confirmed the results reported above. Overall, the magnitude of incremental improvements over time in patients’ quality of life measured by means of two different questionnaires was considerably higher for those who were alive and progression-free in the cilta-cel group versus patients in the realworld group. In particular, the difference in improvement versus baseline favoring CAR T-cell therapy was 13.4 at week 52 (P=0.0081) and increased to up to 30.8 (P<0.0001) when the analysis included death as an additional factor regarding patients’ health status. Patients infused with cilta-cel experienced more adverse events compared to the LocoMMotion study, although in this latter group the incidence was likely to be underestimated.6

The study by Mateos et al. supports the meaningful improvements offered by cilta-cel compared to physicians’ choice of therapy in triple-class exposed patients with relapsed/refractory MM. Results are consistent with similar analyses of cilta-cel versus other external cohorts7-9

References

1. Puertas B, González-Calle V, Sobejano-Fuertes E, et al. Novel agents as main drivers for continued improvement in survival in multiple myeloma. Cancers (Basel). 2023;15(5):1558.

2. Gandhi UH, Cornell RF, Lakshman A, et al. Outcomes of patients with multiple myeloma refractory to CD38-targeted monoclonal antibody therapy. Leukemia. 2019;33(9):2266-2275.

3. Munshi NC, Anderson LD Jr, Shah N, et al. Idecabtagene vicleucel in relapsed and refractory multiple myeloma. N Engl J Med. 2021;384(8):705-716.

4. Berdeja JG, Madduri D, Usmani SZ, et al. Ciltacabtagene autoleucel, a B-cell maturation antigen-directed chimeric antigen receptor T-cell therapy in patients with relapsed or refractory multiple myeloma (CARTITUDE-1): a phase 1b/2 openlabel study. Lancet. 2021;398(10297):314-324.

5. Mateos MV, Weisel K, Martin T, et al. Adjusted comparison of

and contribute to the growing body of evidence highlighting the potential of CAR T-cell therapy as a novel treatment strategy to address the high unmet need of this hard-to-treat set of patients. Although the prospective design of the LocoMMotion study and its alignment with CARTITUDE-1 for most of the eligibility criteria and clinical outcome measures allowed a robust comparison of ciltacel versus conventional therapies, the potential for confounding bias related to missing or unobserved patients' characteristics cannot be ruled out. Data on cytogenetics were not available in approximately one-third of patients enrolled into the LocoMMotion study.6 This finding precluded the possibility of making any adjustment for cytogenetic risk in the main analysis and represents a limitation of the study. In addition, two other prognostic variables of interest (i.e., prior history of stem cell transplantation and race) were not considered in the base case scenario since their inclusion had a negative impact on the balance between study populations. Another study limitation is represented by the limited number of patients treated with belantamab mafodotin and selinexor, which are currently approved for the management of triple-class refractory and penta-refractory MM (i.e., refractory to the two ImiD lenalidomide and pomalidomide, the two PI bortezomib and car fi lzomib, and the anti-CD38 MoAb daratumumab). Although this finding did not allow any evaluation of the relative effectiveness of cilta-cel versus these newer agents, results from such a comparison were reported elsewhere.9 CARTITUDE-4, a phase III randomized study comparing cilta-cel with standard of care regimens for lenalidomide-refractory patients after 1-3 prior lines of treatment including a PI and an IMiD, will more precisely inform clinical decision-making as to the effi cacy and safety of cilta-cel in a less heavily pretreated setting of MM patients.10

Disclosures

No conflicts of interest to disclose.

outcomes between patients from CARTITUDE-1 versus multiple myeloma patients with prior exposure to PI, IMiD and anti-CD38 antibody from the prospective, multinational LocoMMotion study of real-world clinical practice. Haematologica. 2023;108(8):2192-2204.

6. Mateos MV, Weisel K, De Stefano V, et al. LocoMMotion: a prospective, noninterventional, multinational study of real-life current standards of care in patients with relapsed and/or refractory multiple myeloma. Leukemia. 2022;36(5):1371-1376.

7. Costa LJ, Lin Y, Cornell RF, et al. Comparison of cilta-cel, an anti-BCMA CAR-T cell therapy, versus conventional treatment in patients with relapsed/refractory multiple myeloma. Clin Lymphoma Myeloma Leuk. 2022;22(5):326-335.

8. Martin T, Krishnan A, Yong K, et al. Comparative effectiveness of ciltacabtagene autoleucel in CARTITUDE-1 versus physician's

choice of therapy in the Flatiron Health multiple myeloma cohort registry for the treatment of patients with relapsed or refractory multiple myeloma. EJHaem. 2021;3(1):97-108.

9. Weisel K, Krishnan A, Schecter JM, et al. Matching-adjusted indirect treatment comparison to assess the comparative eficacy of ciltacabtagene autoleucel in CARTITUDE-1 versus belantamab mafodotin in DREAMM-2, selinexordexamethasone in STORM Part 2, and melphalan flufenamide-dexamethasone in HORIZON for the treatment of patients with triple-class exposed relapsed or refractory

multiple myeloma. Clin Lymphoma Myeloma Leuk. 2022;22(9):690-701.

10. A study comparing JNJ-68284528, a CAR-T therapy directed against B-cell maturation antigen (BCMA), versus pomalidomide, bortezomib and dexamethasone (PVd) or daratumumab, pomalidomide and dexamethasone (DPd) in participants with relapsed and lenalidomide-refractory multiple myeloma (CARTITUDE-4). ClinicalTrials.gov. Updated January 18, 2023. https://clinicaltrials.gov/ct2/show/NCT04181827 Accessed January 27, 2023.

Nandrolone decanoate: new therapeutic option for telomeropathies?

Camilla Frieri

AORN San Giuseppe Moscati, Avellino, Italy

In the May issue of Haematologica, Diego V. Clé and coworkers 1 presented clinical trial data about the safety and activity of nandrolone decanoate in the treatment of telomeropathies (clinicaltrials.gov NCT02055456). Telomeropathies are a heterogeneous group of pathologies due to the various pathogenic germline variants in genes encoding products involved in telomere maintenance. They are characterized by multidistrict clinical impairment; these are mainly hematologic, respiratory, hepatic and cutaneous-mucosal.2,3 Hematologic disorders such as bone marrow (BM) failure occur most often in young individuals and are often associated with organ disorders (hepatic or pulmonary) and early cancer. In adulthood, idiopathic pulmonary fibrosis is the most common symptom of telomeropathies; a significant proportion of these patients have hematologic abnormalities (macrocytosis and thrombocytopenia in particular), and a certain number will develop hematologic disease during the course of pulmonary follow-up.4 In the case of severe organ failure, such as aplastic anemia, lung and liver dysfunction, the only potentially curative treatment is transplantation.5

From a hematologic point of view, patients with BM failure, high-risk myelodysplastic syndromes or leukemias should undergo allogeneic stem cell transplantation from an HLA-matched non-mutated intrafamilial donor or an HLAmatched unrelated donor. However, BM transplantation (BMT) does not correct the genetic deficiency in the extrahematopoietic cells, and exposes the organs to infectious, toxic and immunological complications.

The prognosis of BMT in cases of telomeropathies is highly variable, often due to multi-organ impairment. In a recent review of the literature by Barbaro et al., 6 overall survival in a cohort of 109 patients was estimated at 57% and 23% at 5 and 10 years, respectively, due to liver and lung complications, and graft dysfunction. To reduce organ toxicity, there is a growing trend toward alkylator- and radiationfree protocols; recently, the European Group for Blood and Marrow Transplantation recommended using the FCC

Correspondence: C. Frieri camillafrieri@gmail.com

Received: February 2, 2023. Accepted: February 10, 2023. Early view: February 23, 2023.

https://doi.org/10.3324/haematol.2022.282540

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

regimen of fludarabine, cyclophosphamide, and alemtuzumab.7,8

Androgens, such as danazol, oxymetholone, and nandrolone are the main therapeutic alternative to allogeneic transplantation in cases of severe hematologic damage. In the only published prospective study, Townsley et al.9 observed a hematologic response in 79% of patients at 3 months (24 patients were evaluable) and 83% of patients at 24 months (12 patients were evaluable) treated with danazol. This study also suggests a benefit in terms of a reduction in forced vital capacity loss in patients with respiratory impairment. The main side-effects reported were elevation of transaminases in 41% of cases, severe hepatic dysfunction, headaches, muscular cramps, and weight gain. In a recent in vitro study, Vieri et al.10 did not observe any significant differences in the efficacy of danazol, oxymetholone and nandrolone to improve telomerase activity. The choice of the compound should be based on the patient’s individual co-morbidities, e.g., preexisting liver disease and expected side-effects. In the context of this scientific background, Clé and coworkers presented a phase I/II single-center prospective trial with the aim of evaluating the reduction in telomere attrition over time compared to known rates of telomere erosion in normal individuals, and in those who carry a mutation in the telomerase genes, treated with nandrolone decanoate. Hematologic response, pulmonary function, incidence of clonal evolution, safety, and survival were all assessed as secondary endpoints. The sample consisted of 17 patients with a median age of 36 years (range, 4-59 years) with age-adjusted mean telomere length under the 1st percentile and/or identified germline pathogenic variants in telomere-biology genes associated with at least one cytopenia and/or radiologic diagnosis of interstitial lung disease (ILD). All patients were diagnosed with BM failure, seven patients were also diagnosed with ILD, and four patients also had liver involvement. Five patients had skin features of dyskeratosis congenita.

The patients received 5 mg/kg of intramuscular nandrolone decanoate every 15 days for two years.

Of the 17 patients enrolled, 13 were evaluable for the primary end point at 12 months and ten at 24 months. Consistent telomere elongation, evaluated by flow-FISH, was achieved by 77% (10/13) of patients at 12 months and by all evaluable patients (10/10) at 24 months; the average increase in intelomere length was 0.87 kb (95%CI: 0.20-1.55 kb; P=0.01) at 12 months and 0.49 kb (95%CI: 0.24-1.23 kb; P=0.18) at 24 months.

Hematologic response, assessed as improvement in cytopenia or transfusion independence or 50% reduction from baseline, was achieved in 50% of patients with BM failure at 12 months, and in 63% at 24 months. The best response was observed in hemoglobin levels, while no significant differences were observed in neutrophil or platelet values.

In patients diagnosed with ILD at baseline, there was no significant change in pulmonary function as evaluated by clinical, spirometric and radiological parameters: 2/7 died of respiratory failure during nandrolone treatment, a mild improvement in lung function was observed in 3/7, while lung function remained stable in 2/7. One of the most intriguing findings was the comprehensive analysis of the effect of nandrolone on clonal hematopoiesis; clones carrying mutations in genes associated with myeloid neoplasms remained stable or decreased.

References

1. Clé DV, Catto LFB, Gutierrez-Rodrigues F, et al. Effects of nandrolone decanoate on telomere length and clinical outcome in patients with telomeropathies: a prospective trial. Haematologica. 2023;108(5)1301-1313.

2. Calado RT, Young NS. Telomere diseases. N Engl J Med. 2009;361(24):2353-2365.

3. Carvalho VS, Gomes WR, Calado RT. Recent advances in understanding telomere diseases. Fac Rev. 2022;11:31.

4. Stella GM, Balestro E, Lacedonia D, Baraldo S. Telomeropathies: an emerging spectrum of disorders with important implications for patients with interstitial lung disease. Minerva Med. 2016;107(1 Suppl 1):9-14.

5. Vieri M, Brümmendorf TH, Beier F. Treatment of telomeropathies. Best Pract Res Clin Haematol. 2021;34(2):101282.

6. Barbaro P, Vedi A. Survival following haematopoietic stem cell transplant in patients with dyskeratosis congenita – a systematic review of the literature. Biol Blood Marrow

These are interesting data and lay the groundwork for expanding our knowledge about the role of androgens in telomeropathies. However, some issues remain, mainly regarding safety. Firstly, from a practical point of view, the use of an intramuscular formulation could raise some concerns, especially in patients with low platelet concentrations. This formulation was chosen to bypass hepatic metabolism and reduce toxicity, an event, however, that has not been confirmed (elevated transaminases in 88% of cases vs. 41% in historical reports).9 Secondly, two deaths due to intracranial hemorrhages were observed which were not reported from previous trials or in retrospective registry reports. Formally, a correlation with the drug cannot be ruled out; however, it should be stressed that these patients had very low platelet concentrations (<10x109/L).

Overall, this is the second prospective study in the field of telomeropathies, rare diseases for which it is often not easy to find possible study candidates. The study confirms previous observations and adds new data on lung fibrosis, side-effects, safety, and clonal evolution during treatment. Rare diseases such as telomeropathies remain under-explored with few therapeutic alternatives, making additional information especially useful.

Disclosures

No conflicts of interest to disclose.

Transplant. 2016;22(7):1152-1158.

7. Dietz AC, Mehta PA, Vlachos A, et al. Current knowledge and priorities for future research in late effects after hematopoietic cell transplantation for inherited bone marrow failure syndromes: consensus statement from the Second Pediatric Blood and Marrow Transplant Consortium International Conference on Late Effects after Pediatric Hematopoietic Cell Transplantation. Biol Blood Marrow Transplant. 2017;23(5):726-735.

8. Peffault de Latour R. Transplantation for bone marrow failure: current issues. Hematology Am Soc Hematol Educ Program. 2016;2016(1):90-98.

9. Townsley DM, Dumitriu B, Young NS. Danazol treatment for telomere diseases. N Engl J Med. 2016;375(11):1095-1096.

10. Vieri M, Kirschner M, Tometten M, et al. Comparable effects of the androgen derivatives danazol, oxymetholone and nandrolone on telomerase activity in human primary hematopoietic cells from patients with dyskeratosis congenita. Int J Mol Sci. 2020;21(19):7196.

Deep learning applications in visual data for benign and malignant hematologic conditions: a systematic review and visual glossary

Andrew Srisuwananukorn,1 Mohamed E Salama2 and Alexander T. Pearson3

1Tisch Cancer Institute, Icahn School of Medicine at Mount Sinai, New York, NY; 2Sonic Healthcare USA, Austin, TX and 3Section of Hematology/Oncology, Department of Medicine, The University of Chicago, Chicago, IL, USA

Abstract

Correspondence: A.T. Pearson apearson5@medicine.bsd.uchicago.edu

Received: June 3, 2022.

Accepted: January 18, 2023.

Early view: January 26, 2023.

https://doi.org/10.3324/haematol.2021.280209

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Deep learning (DL) is a subdomain of artificial intelligence algorithms capable of automatically evaluating subtle graphical features to make highly accurate predictions, which was recently popularized in multiple imaging-related tasks. Because of its capabilities to analyze medical imaging such as radiology scans and digitized pathology specimens, DL has significant clinical potential as a diagnostic or prognostic tool. Coupled with rapidly increasing quantities of digital medical data, numerous novel research questions and clinical applications of DL within medicine have already been explored. Similarly, DL research and applications within hematology are rapidly emerging, although these are still largely in their infancy. Given the exponential rise of DL research for hematologic conditions, it is essential for the practising hematologist to be familiar with the broad concepts and pitfalls related to these new computational techniques. This narrative review provides a visual glossary for key deep learning principles, as well as a systematic review of published investigations within malignant and non-malignant hematologic conditions, organized by the different phases of clinical care. In order to assist the unfamiliar reader, this review highlights key portions of current literature and summarizes important considerations for the critical understanding of deep learning development and implementations in clinical practice.

Introduction

Recent advances in large-scale data storage, availability, and computational power have led to significant interest in the development of new techniques for “big data” analysis. Rapidly evolving artificial intelligence (AI) algorithms aim to efficiently utilize vast amounts of information with minimal human interaction to address tasks that automate or improve upon human-level assessment. Artificial intelligence takes many forms and includes domains of deep learning (DL), convolutional neural networks (CNN), and other related techniques that are capable of processing imaging data quickly and automatically. Research divisions within commercially successful technology companies have popularized DL models for vision-related tasks, such as facial recognition, image segmentation, object detection, and many other examples that are currently being integrated into daily life.

Within the medical field, visual assessment of digitized clini-

cal imaging and biospecimens by physicians is critical in numerous phases of clinical care for patients. As a result, early investigations that employ clinical DL using histology slides or radiological images within medicine have produced promising results, including diagnosis detection,1 clinical subtyping,2 cancer mutation prediction,3,4 and survival.5 Recognizing the clinical importance of these algorithms, the US Food and Drug Administration has approved a number of novel AI and DL products.6

However, DL algorithms exploring malignant and non-malignant hematologic conditions are still scarce. With digitization tools generating larger biospecimen image databases7,8 and researchers becoming increasingly familiar with DL techniques, examples of applications in hematology are growing exponentially.9-14 As such, it is inevitable for hematologists to be familiar with the broad concepts, applications, and limitations of clinical DL.

In this structured narrative review, we aim to describe the general concepts, provide a visual glossary for key terms

within image-based DL, and conduct a systematic review to provide an up-to-date assessment of the application of image-based DL in benign and malignant hematology across various phases of patient care.

Neural networks and deep learning

The concept of “deep learning” is poorly defined, imprecise, and often used interchangeably with terms such as “machine learning” and “artificial intelligence.” Traditionally, “artificial intelligence” is the use of automated systems to perform a particular task. “Machine learning” represents a subset of AI in which rules are not explicitly predetermined, but are acquired by training and optimizing parameters based on observed data. Machine learning workflows traditionally separate data into training, validation, and external testing cohorts for model assessment. Examples of machine learning that are probably the most familiar include linear regression, logistic regression, or Cox proportional hazards models. “Deep learning” is a recently-popularized subset of machine learning utilizing a specialized neural-

network architecture undergoing millions of arithmetic operations (Figure 1).15,16 DL architectures are loosely modeled after the complex neural connections of the human brain.17

Although the term “deep learning” is derived from “deep convolutional neural networks” and has gained interest particularly in clinical research, the strict definition has become increasingly ambiguous and may not completely represent modern state-of-the-art techniques. The field of DL and the list of essential glossary terms are rapidly changing, but in keeping with contemporary clinical manuscripts, this review will use the term “deep learning” to mean “deep convolutional neural networks and other contemporary techniques related to computer vision”. There are also non-image-based neural networks and image-based machine learning architectures without neural networks; however, both are beyond the scope of this current review.

Image preprocessing

A standard workflow 3,18,19 for DL research typically requires preprocessing input images, which can expedite DL training time or improve performance. Preprocessing steps are typically dependent on the modality of the

Figure 1. Exemplifying differences in “Artificial Intelligence”, “Machine Learning”, and “Deep Learning” with regards to anemia. “Artificial Intelligence” (AI) involves automation of tasks, and can be an explicitly programmed rule to categorize based on the level of a laboratory value. “Machine Learning” (ML) methods, such as linear or logistic regression, derive associations from given training data. More complex image-based “Deep Learning” (DL) models utilize complex architectures termed “Neural Networks” to associate subtle features associated with particular outcomes of interest by using input training data, similar to other machine learning frameworks. In this figure, the clinical condition of anemia based on hemoglobin (Hb) values is used as an example for the above computational frameworks. DL methods may extract features in non-traditional images, such as fundoscopic exams, to derive clinically meaningful categorizations.

Haematologica | 108 August 2023 1994

image type. While radiological images may be input either whole or with particular Regions of Interest (ROI) segmented, histopathology slides are typically tessellated into smaller tiles representing tissue or segmented cells of interest prior to inclusion into the model. Normalization of pathology images may reduce artifacts specific to a clinical site or particular scanning device, but there is no current standard normalization process. Data augmentation may be performed with random image rotations, vertical or horizontal flips, and simulated compression artifacts to increase the size of the training set and broaden generalizability. In addition to using images alone, researchers can include other data modalities such as clinical information with multi-modal models to supplement image inputs in attempts to improve model performance.

General neural network structure

In a simplified viewpoint of neural network structures (Figure 2), the input image is transformed at various intermediate states, termed “nodes,” with each node representing a different graphical feature of the image. As the image is passed from node to node, the connection between each node involves mathematical transformations to represent more complex features in later nodes. Each node can be connected to multiple subsequent nodes simultaneously, and the group of nodes with similar numbers of sequential connections from the input image represent a layer of intermediate nodes. Shallow and Deep neural networks refer to the number of node layers within a particular architecture, but there is no strict definition to differentiate the two. In addition, nodes may not necessarily connect to the nodes in the immediately subsequent layer, but may connect by “skip connections” to nodes in later layers. The penultimate layer of nodes, each representing only a single numerical value, is termed the Logit Layer, the values of which are then normalized between the range of 0 and 1 to give the final probabilities for the outcomes of interest. Common outcomes of interest and examples include object detection, segmentation, classification, regression, survival analysis, and detail optimization (Figure 3).

Information propagation and parameter training

To develop a neural network model, the input image is represented numerically by each pixel. The numerical information is propagated though intermediate nodes and layers towards the direction of the output layer. The connections between nodes are mathematically represented by either non-linear operations or matrix multiplication and addition with potentially millions of trainable parameters, whose values are updated while optimizing the end outcome. Upon initial model evaluation, the sequential movement of information from input image towards the outcome of interest is deemed “forward propagation” or “forward pass”

(Figure 2A-F). To complement an initial prediction, the user defines a particular loss function to quantitatively describe the incorrectness of the model’s prediction from available ground truth. Using an additional user-defined optimizer algorithm, the trainable parameters are iteratively adjusted to decrease the loss value in subsequent forward passes. This framework of optimizing parameters in earlier layers using information from the predicted outcome is deemed “back propagation.” During training, forward and back propagation are repeated for a defined number of repetitions, or epochs, but training can also be stopped if other defined optimal conditions are met. Given the need for at least 109 calculations per forward pass, parallel computing often requires specific hardware such as Graphics Processing Units (GPU) to expedite necessary matrix operations to be finished within reasonable timeframes.

Convolutions

At the time of writing, the most popular type of DL architecture is the convolutional neural network (CNN). The prototypical CNN algorithm assesses a smaller grid-like portion of each input image prior to propagation towards the next layer (Figure 2C). CNN utilize the convolution operation between layers, which involves matrix multiplication across overlapping sub-sections of the input image to produce a lower-dimensional output representation.

Pre-trained networks and transfer learning

Initially, CNN trained to perform object detection required millions of manually-annotated images, training for days or weeks on industry-grade computational equipment.20 After training is complete, CNN have traditionally been understood to learn “low-level” general features such as lines, edges, and shapes in earlier layers of the network, but more complex “high-level” features such as faces, patterns, and spatial distributions are learned in subsequent layers that are more closely associated with the evaluated outcome.21

In clinical research, it is rare for clinicians to have the resources to develop new CNN architectures with initially random parameters; such a feat requires large-scale databases with expert-level annotations and access to industry-grade supercomputers. Researchers have taken advantage of the learned features along progressive layers by using models previously trained on large databases for non-clinical tasks, but repurposing the final few layers to predict specific clinically-relevant outcomes. The concept of transfer learning involves utilizing a pre-trained network such as those already trained on the ImageNet database of over 1 million general images,22 initializing the model with the parameters that learned “low-level” features from images unrelated to the application of interest, and

allowing the model to retrain and modify parameters in the last few layers to learn “higher-level” features on images for specific patient-related tasks. By utilizing transfer learning, the minimum required dataset and computational power is significantly less than fully training a network from completely random parameters.23

Specific deep learning architectures in clinical research

While DL is a framework of neural networks for outcome prediction, each specific model architecture incorporates drastically different complexities with regards to number of layers, connections between layers, functions, and many other highly-engineered features. In fact, newer contempor-

Figure 2. Brief representation of the structure and training phases of deep convolutional neural networks. Collectively, A-H represent “forward propagation” and G-H represent “back propagation.” (A) Images are first passed into the network to predict an outcome of interest. In this example, 4x4 pixelated images of written numbers are used to train a network to predict the numeric value of the image. (B) Pixels are initially converted into numeric values based on pixel intensity. (C) Smaller subsections of the input images are transformed with the convolutional operation, which involves matrix multiplication and addition with trainable parameters. (D) As information is passed into subsequent layers, the image undergoes non-linear transformations, such as the Rectified Linear Unit function that allow the model to represent non-linear relationships within the data. (E) Within this figure, intermediate layers are restructured to a layer of single numerical values in the Logit layer. (F) After propagation through the pre-defined number of convolutional layers, the final activation function normalizes the Logit layer into a distribution of probabilities across the space of available outcomes. The value with the highest probability is deemed the model’s prediction. (A-F) The framework outlined as information is passed from image input to model prediction is termed “forward pass” or “forward propagation.” (G) After the first pass of the model’s predictions, a loss function specific to the outcome data type is calculated to quantitatively assess the level of error produced by the initial prediction. The loss function is chosen before training by the user. Common examples of loss functions are “cross entropy” for categorical outcomes and “mean square error” for regression outcomes. (H) Optimization algorithms iteratively alter the trainable parameters within each of the previous convolutional layers based on the defined loss function. The direction and magnitude of parameter adjustments is calculated by either maximizing or minimizing the loss function in future forward passes, as chosen by the user for the outcome of interest. (G and H) The framework outlined for automatically adjusting earlier parameters to optimize model performance is termed “back propagation.” The process of forward (A-F) and back (G-H) propagation is repeated until a prespecified set of conditions is fulfilled, typically leading to more accurate predictions.

Haematologica | 108 August 2023

Figure 3. Examples of outcome tasks and explainability methods in Deep Learning. In this example, the initial input image is a promyelocyte with visible Auer Rods as seen on a peripheral blood smear, a possible pathognomonic finding for acute promyelocytic leukemia. Output tasks can include localization of white blood cells (Detection), creation of a region of interest around the nucleus (Segmentation), disease prediction (Classification/Regression), or increasing the visual quality of the input image (Detail Optimization). Explainability methods are necessary to ensure biological feasibility. In exploratory analyses, parameters within the intermediate layers can be directly visualized (Feature Maps), heatmaps can be generated to highlight specific areas associated with the outcome (Saliency or Attention Maps), synthetic images can be generated from noise to represent an outcome of interest (generative adversarial networks), or cluster analyses can be performed with dimensionality reduction techniques.

ary models lack any convolutional layers, and infer local and global image features by other methods.24 Thus far, the predominant architecture for hematology-specific questions tend to be from a class of CNN known as Residual Neural Networks (ResNets), which utilize skip connections. Most specific ResNet architectures, such as Inception, EfficientNets, MobileNets, and other various ResNet models are open-source and widely available.25 Certain model architectures are engineered to provide an output that is an additional image; these model structures are needed for dimensionality reduction, bounding-box detection, segmentation, and noise reduction tasks. One specific architecture, Autoencoders, are networks that pass an input image through an intermediate lower-dimensional representation, followed by upsizing to a higher-dimensional space to recreate the input image.26 Theoretically, the lowerdimensional intermediate representation still retains features of the original image which may be clinically or biologically relevant. Similar architectures such as U-Net require additional training data, such as object ROI or low/high-quality image pairs, to accomplish tasks such as image segmentation or digital optimization.

An additional relevant DL framework utilized is Multiple Instance Learning (MIL)27,28 and its attention-based derivatives,29,30 including the Clustering-constrained Attention Multiple Instance Learning (CLAM).31 The main distinction in MIL frameworks is the prediction for data subsets and not for single instances. Specifically, input images are separated into smaller subsets. The entirety of the subset is predicted “positive” if at least one image in the subset is predicted “positive”. As an example of MIL in histopathology, a biopsy whole-slide image would be predicted “cancerous” when one

extracted tile is predicted as such.1 This framework may be particularly helpful when single annotations are provided across an entire image, or “weak supervision”, and not necessarily labels for each specific segmented ROI. In addition, Attention, or a numeric weight, can be assigned to each image tile to produce weighted predictions, as well as provide explainable heatmaps. Using Attention, CLAM was developed to increase the speed of MIL and reduce the noise from irrelevant image tiles.

Vision Transformers (ViT) are a novel technique that do not utilize the convolution operator.24 The entire image is separated into a grid of sub-images that are analyzed in parallel along with the relative location of each sub-image. With this method, global relationships across the entire image may be learned by the model as opposed to only local features that are seen by the previously-described CNN. Currently, most architectures for hematology-specific questions utilize ResNet architectures, with just a few examples also incorporating MIL. However, the emergence of ViT and CLAM frameworks are part of a changing landscape of implemented DL architectures. In general, the choice of model architecture is somewhat informed by expected outcome task, but it is still largely empiric. However, there are broad advantages and disadvantages for each of the previously-mentioned frameworks. With weak supervision, MIL tends to require significantly larger amounts of training data than ResNets.1 CNN and ViT perform equally well at the scale of currently available clinical datasets. However, ViT are superior to CNN for larger scale datasets and are more computationally efficient with significantly fewer parameters.24 There are numerous methods to attempt to explain the inner mechanisms of

standard CNN,32 but similar methods to “open the black box” of ViT are currently under development.33

Explainability

While “explainability” in DL research is loosely defined, in this review, “explainability” refers to the efforts in describing DL models and predictions in humanly-understandable concepts.34

Although DL may empirically exhibit a high performance, DL is often criticized for its highly complex mechanisms and is often thought of as a “black box.” In multiple examples,

Figure 4. Literature search. (A) Search terms to extract relevant manuscripts related to deep learning in malignant and non-malignant hematology. Articles were queried in PubMed using one “Deep Learning” term in addition to one “Hematology” term. (B) PRISMA diagram of “Deep Learning in Hematology” survey. Initially 2,708 articles were found from a PubMed query. After initial review of abstracts and article titles, 237 reports were deemed eligible for further review of full manuscripts. Finally, 65 articles were included for the current narrative review. Justification for exclusion are provided.

seemingly high-performing models often utilize artifact or contextually irrelevant features for its predictions, as the artifactual features may be unintentionally over-represented in certain imaging subgroups.35 Multiple methods are under development to explain and validate biologically reasonable predictions. As such, explainability is increasingly important in clinical AI development and in developing physicians’ trust of DL.36

To give just a few examples, unsupervised data dimensionality reduction methods such as principal component analysis (PCA), t-distributed stochastic neighbor embedding

(t-SNE), and Uniform Manifold Approximation and Projection (UMAP) are statistical techniques used to group visually similar input images into clusters, which may overlap with relevant outcomes. These methods are also popularized in non-imaging data such as single cell molecular and cytometry time-of-flight analyses. Feature maps are direct visual representations of the intermediate trained parameters. Plotting Attention scores or using Saliency map methods such as Grad-CAM or Smooth-Grad can overlay heat-maps upon the input image to highlight relevant visual cues associated with the outcome of interest.35 For example, the heatmap explainability methods of a peripheral blood smear image may highlight pathognomonic Auer Rods for the accurate diagnosis of acute promyelocytic leukemia (Figure 3). More complex methods such as Generative Adversarial Networks are architectures trained to generate synthetic images, which can create representations of a particular class or outcome.37

Metrics

Common performance metrics for the evaluation of DL classification models include Area Under the Receiver Operator Curve (AUROC), sensitivity, specificity, and accuracy. The AUROC represent the tradeoff between true and false positive rates for a binary model along a range of possible threshold values. AUROC values nearing 1.0 represent a model with perfect discriminatory power, and values tending towards 0.5 perform no better than random chance. For segmentation tasks, the Sørensen-Dice similarity co-

efficient (Dice) represents the overlap between the predicted area of interest with the ground truth, where a Dice coefficient of 1.0 represents ideal predictive overlap. Other segmentation metrics include the similarly defined Jaccard index, also known as Intersection over Union (IoU).

Literature review for clinical application of deep learning in hematologic conditions

A Boolean query was submitted to PubMed to extract articles created between January 1, 1990, and August 1, 2022. Search terms included both a “deep learning” and a “hematology” specific term (Figure 4A). The query resulted in 2,708 initial articles. Further refinement by manual review by one author excluded a large number of articles (Figure 4B), resulting in 65 manuscripts. General trends and findings of the resulting articles are described in the context of how DL has been utilized to enhance phases of clinical care within various hematologic conditions, including task automation, detail optimization, disease detection, differential diagnosis, disease classification, risk prediction, complication assessment, therapy response, and survival prediction (Figure 5). General considerations for critical appraisal of the following manuscripts include performance metrics, use of external or prospective validation cohorts, use of explainability

Figure 5. Deep Learning applications within 65 malignant and non-malignant hematology manuscripts. Applications are divided into separate phases of clinical care, including task automation, detail optimization, disease detection, differential diagnosis, disease classification, risk prediction, complication assessment, therapy response, and survival/relapse prediction. Image domains include radiological, pathological, and other atypical image types such as electrocardiograms or funduscopic exams. Specific image modalities are detailed in Tables 1-4. CBC: complete blood count; MDS: myelodysplastic syndromes; MPN: myeloproliferative neoplasms; RBC: red blood cell; VTE: venous thromboembolism; WBC: white blood cell.

Haematologica | 108 August 2023

Summary

acute lymphoblastic leukemia; FAB: French-American-British classi fi ca tion; APL: acute promyelocytic leukemia; FL: follicular lymphoma; DLBCL: diffuse large B-cell lymphoma; MCL: mantle cell lymphoma; ECG: electrocardiogram; PBS: peripheral blood smear; IFC: imaging fl o w cytometry; BMA: bone marrow aspirate; LNB: lymph node biopsy; PET-CT: positron emission tomography / computed tomography; BMB: bone marrow biopsy; CT: computed tomography; AUROC: Area Under Receiver Operato r Curve; UMAP: Uniform Manifold Approximation and Projection; t-SNE: t-distributed Stochastic Neighbor Embedding; PCA: Principal Component Analysis; N: number.

Table 2. Summary and performance metrics of deep learning applications in hematology for disease

Table 3. Summary and performance metrics of deep learning applications in hematology for

MDS: myelodysplastic syndrome; AML: acute myeloid leukemia; ALL: acute lymphoblastic leukemia; CML: chronic myeloid

blood cell; DLBCL: diffuse large B-cell lymphoma; BL: Burkitt’s lymphoma; SLL: small lymphocytic lymphoma; NHL: non-Hodgkin lymphoma; FL: follicular lymphoma; PCNSL: primary central nervous system lymphoma; GBM: glioblastoma multiforme; RBC: red blood cell; AA: aplastic anemia; AB: French-American-British classi fi ca tion; PBS: peripheral blood smear;

BMA: bone marrow aspirate; LNB: lymph node biopsy; FNA: fi ne needle aspir ate; MRI: magnetic resonance imaging; IFC: imaging fl ow cytometry; AUROC: Area Under Receiver Operator Curve; t-SNE: t-distributed stochastic neighbor embedding; UMAP: uniform manifold approximation and projection; N: number.

Table 4. Summary and performance metrics of deep learning applications in hematology

assessment, therapy response, and survival/relapse prediction. MDS: myelodysplastic syndrome; PE: pulmonary embolus;

mantle cell lymphoma; SCD: sickle cell disease; ALL: acute lymphoblastic leukemia; ENKTL: extranodal natural killer/T-cell lymphoma; BMB: bone marr ow biopsy; CT: computed tomography; BMA: bone marrow aspirate; UWF-FP: ultra-wide fi eld color fundus photographs; IFC: imaging fl ow cytometry; PET-CT: positron emission tomography / computed tomography; AUROC: Area

Under Receiver Operator Curve; Dice: Sør ensen-Dice similarity coef fi cient; UMAP: uniform manifold approximation and projection; t-SNE: t-distributed stochastic neighbor embedding; N: number.

methods, and comparison with human expert performance (Tables 1-4).

Task automation

Routine clinical workflows in pathology and radiology may involve repetitive actions. Automation models can be developed to increase efficiency and decrease physician burden for tasks such as counting cell types in peripheral blood smears or contouring the borders of suspicious lesions on imaging. For pathology workflows, DL models trained to contour white blood cell (WBC) borders in peripheral blood smears were highly effective with near perfect Dice co-efficients in multiple cohorts.38 Automatic detection of cells can be put through downstream analyses and provide an automated cell count, for which DL-based methods achieve high accuracy.39

In addition, chromosomal analyses are standard for diagnosis and prognostication for multiple hematologic malignancies. Manual segmentation and rotation of digital karyograms is time-consuming, but automated models can significantly expedite throughput.40 In radiology workflows, contouring suspicious lesions or organs can help characterize downstream parameters such as volume, width, and avidity. Hypermetabolic lesions on PET/CT have been localized with DL algorithms for multiple adult and pediatric lymphomas or multiple myeloma lesions.41,42 Segmentation metrics were reportedly high, with Dice coefficient 0.860.98 among various lymphomatous conditions.43-45 For other conditions, the automated volume calculation of particular regions of interest have been explored in myeloproliferative neoplasms (MPN) for spleen volume,46 as well as clot burden quantification for new pulmonary emboli.47

Detail optimization

For expert diagnosticians, image quality is critical for the identification of disease. Using U-Net architectures, DL-enhanced images may improve user readability and potentially reduce the amount of toxic contrast material given to patients. Enhancement of peripheral blood images to assess red blood cell (RBC) aberrations have yielded promising results. For sickle cell disease, mobile-device photos of peripheral blood have been digitally upscaled to match laboratory microscope quality; upon further validation, the upscaled images retained relevant visual cues with nearperfect classification.48 However, similar attempts to detect malaria RBC inclusion were less successful, noting that CNN-based enhancement of peripheral blood images was insufficient to resolve parasites that were not already easily distinguishable at low resolution.49 Multiple optimization efforts in radiology have investigated whether DL can improve image quality from lower-contrast images, which may help spare patients from nephrotoxic or radioactive risks. For both positron emission tomography/magnetic resonance imaging (PET-MRI) in lymphomatous conditions and com-

puted tomography (CT) scans in multiple myeloma, authors have concluded that reduced contrast volumes may be feasible while still maintaining diagnostic quality.50, 51

Disease detection

In clinical practice, a common initial diagnostic step for hematologic disorders is the analysis of peripheral blood to observe morphologic abnormalities of RBC, WBC, and platelets. The detection of structural RBC aberrations can identify certain infectious diseases and hemoglobinopathies. In endemic areas of malaria, the Plasmodium parasites are often identified by light microscopy as RBC inclusions. Multiple DL initiatives report high accuracy and good model performance for the diagnosis of malaria from peripheral blood in both cross-validated and external cohorts.52-56 Other RBC aberrations, such as hemoglobin H inclusions in α-thalassemia, can be detected by DL with appropriate peripheral blood staining protocols.57 With regards to transfusion medicine needs, the quality and degradation of RBC products prior to transfusion can also be determined with DL methods. Using explainability techniques, Doan et al. explored their proposed autoencoder network trained on RBC images to identify novel features associated with poor storage quality RBC products.58 For certain disorders, the detection of aberrant WBC morphologies from peripheral blood is paramount. DL algorithms consistently detect dysplastic neutrophils pathognomonic for myelodysplastic syndrome (MDS),59 as well as other white blood precursors to aid in the diagnosis of MPN,60 acute promyelocytic leukemia (APL),12 or acute lymphoblastic leukemia (ALL).61,62 Many DL models for WBC detection have performed with high accuracy and AUROC upon internal validation strategies. If translated into clinical practice, DL models for peripheral blood assessment may expedite critical diagnoses which necessitate emergent therapy, such as APL. Particularly for myeloid malignancies, bone marrow assessment is usually needed to establish a diagnosis. DL can detect particular cellular morphologies of neutrophils, megakaryocytes, promyelocytes, and plasma cells associated with MDS,63 MPN,64 APL,13 and multiple myeloma,65 respectively. Similarly for lymphoid malignancies, assessing lymph node architectures can aid the diagnosis of various lymphomas, such as diffuse large B-cell lymphoma (DLBCL)66 or follicular lymphoma (FL).67 DL models developed by Li et al. maintained high accuracy for the diagnosis of DLBCL from lymph biopsies across four separate institutional cohorts.66 Furthermore, Syrykh et al. utilized the clinical challenge of differentiating follicular lymphoma from follicular hyperplasia to develop a novel DL method quantifying prediction uncertainty, which is not often reported in DL studies. With their uncertainty method, the authors report higher classification capabilities when only considering the newly categorized low-uncertainty images.67

In addition to pathologic analysis, clinical guidelines commonly suggest radiologic assessment for the initial workup of suspected malignancy or thrombosis. Using PET/CT images, DL models exhibit high classification of the hypermetabolic lesions for DLBCL diagnosis.68 However, similar attempts using PET/CT images of mantle cell lymphoma (MCL) patients are challenged with tradeoffs between sensitivity and false positive rates for diagnosis in external cohorts.69 For select non-malignant conditions, multiple studies explored DL for the expedited and more affordable diagnosis of pulmonary emboli (PE) and deep vein thromboses (DVT), for which a diagnosis may require immediate intervention.70-72 Huang et al. integrated clinical data in conjunction with CT scans to improve their DL model for PE detection. The authors report that multi-modal models exhibit higher classification performance than image-only DL models.71 In addition, automated detection of common thrombotic conditions may reduce the financial burden, with cost analyses revealing positive financial benefit to health care systems.72

Finally, a particularly novel use of DL is the prediction of disease from imaging modalities beyond standard pathologic or radiologic domains. Multiple studies have shown that anemia can be detected with high accuracy utilizing DL on atypical modalities such as electrocardiograms (ECG)73 or funduscopic examinations.74 Both authors have implemented explainability methods to reveal features associated with anemia, such as QRS complexes in ECG or optic disk aberrations in funduscopic images. Thus, screening for anemia may offer a low-cost benefit for patients already undergoing these common examinations.

Differential diagnosis

Various hematologic conditions share similar features and presentations, posing challenges in providing a definitive diagnosis in clinical scenarios where radiologic findings may be non-specific and pathological morphologies may be subtle. Differentiating among possible diagnoses is a common clinical task, and various approaches of DL have been explored as a potential means to increase objectivity towards a true diagnosis. For example, Li et al. used transfer learning to pre-train their model with images of common household objects, such as bananas, rings, and pears to learn the analogous morphologies of similarlyshaped RBC inclusions of Toxoplasma, Plasmodium, and Babesia. 75

Interestingly, DL models have been shown to better extract subtle features for disease differentiation than can be assessed by humans. Cytopenias can be a common presentation for either MDS or aplastic anemia (AA) patients. Though either diagnosis typically requires bone marrow biopsy assessment, Kimura et al. trained a DL model on peripheral blood images to accurately differentiate between the two conditions.76 Newly diagnosed leukemia patients

commonly present with blasts in peripheral blood. The categorization of blasts into either myeloid or lymphoid lineages requires identifying cell-surface markers by flow cytometry; thus, visualization of blasts is not usually sufficient for classification. Similarly, lymphoma histology share visual commonalities and require immunohistochemical staining of cell-surface markers on biopsy specimens. To address these classification challenges, DL algorithms reportedly differentiate between acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL) utilizing only peripheral blood or bone marrow images,9,77,78 and similarly among various non-Hodgkin lymphomas (NHL) utilizing standard hematoxylin and eosin (H&E) lymph node biopsy images.79-82

For patients with malignant brain lesions found on MRI imaging, clinicians may be tasked to differentiate between primary central nervous system lymphoma (PCNSL) and glioblastoma multiforme (GBM).83 DL models for this revealed seemingly high initial performance but with a significant reduction to an AUROC of nearly 0.5 in external cohorts.83 The problem of generalizing results highlights the continued need for critical appraisal of any newly-developed DL model across patient populations.

Disease classification

The classification of blood cells in standard peripheral blood smear review is a ubiquitous task useful in a broad array of diseases. The differential of WBC is necessary to stratify the likelihood of the malignant and non-malignant causes of WBC abnormalities. Numerous studies developed DL models as a single cell WBC classifier. Across the studies, performance remained robust, with the majority of studies achieving accuracies above 90% and explainability techniques highlighting sensitive cellular features.11,84-86 However, validation upon external cohorts, which commonly reveal a lower performance,11 is still needed prior to deployment in clinical practice. In addition to WBC classification, the categorization of RBC morphologies is useful within various anemias,87 including sickle cell disease.88 To explore platelet abnormalities, Zhou et al. developed a highly accurate DL model predicting the identity of agonists causing platelet aggregation using imaging flow cytometry.89

Specific sub-classification of diagnoses is often necessary to guide prognostication, counseling, and therapeutic considerations. In numerous non-hematologic applications, previous DL models can accurately further categorize various cancers into genetic and clinical subtypes,4 which has led to similar explorations within leukemic and lymphomatous conditions. For leukemic classifications, ALL bone marrow images can be separated into the historically relevant French-American-British (FAB) classifications.90 Furthermore, genomic subtypes may be accurately identified by DL models; Eckardt et al. identified NPM1 mutations among newly diagnosed AML patients and characterized

novel cellular morphological features that had not previously been reported.14 Broader DL efforts to identify each clinically relevant molecular or cytogenetic abnormality have been attempted for MDS sub-classifications.91 For lymphoma, Swiderska-Chadaj et al. developed a DL model predicting MYC gene rearrangements in DLBCL patients using lymph node biopsy images. Though MYC rearrangement is typically assessed with ancillary fluorescent in situ hybridization, the DL model using only H&E images maintained high accuracy upon external cohorts.92

Advanced stages of patient care

There are currently few examples of DL for the assistance of later stages of patient care, including risk prediction, complication assessment, therapy response, and survival prediction. For such tasks, the disease processes and image modalities are heterogenous. Risk has been assessed with CT images or digitalized bone marrow biopsies (BMB) for DLBCL outcomes. DL models predict the transformation of low-grade lymphomas to high-grade DLBCL using BMB images,93 and, furthermore, known clinical risk factors such as sarcopenia can be extracted and quantified in CT images of DLBCL patients.94 Risk in thrombotic conditions can be characterized automatically using DL classification of right ventricular strain in chest imaging for PE workup.95 Cai et al. assessed complications of sickle cell disease by detecting sea fan neovascularization in funduscopic images, which is a vision-threatening complication warranting prophylactic management.96 Doan et al. evaluated therapy response in ALL patients by using DL methods to detect residual lymphoblasts after receiving induction chemotherapy.97 Finally, DL models for relapse prediction using baseline imaging have been developed for extranodal natural killer/T-cell lymphoma98 and mantle cell lymphoma.99 However, further evaluations upon external cohorts are needed for these advanced stage tasks.

Conclusions

The use of deep learning in hematologic conditions has attracted significant interest in recent years. As noted above, researchers have utilized multiple data structures including radiologic images, pathology specimens, clinical data, and atypical imaging such as funduscopic examinations to perform a variety of clinically relevant tasks. Most Authors reported high model performance for disease diagnosis, segmentation, and subtyping. Other studies explored tasks beyond human capabilities such as genomic inference and prognostication from imaging analysis alone. Few studies have used hematologic conditions as a means to implement state-of-the-art architectures to improve the field of DL in general. Compared to other clinical domains, DL in hematology is still in its infancy, so it is not widely used in clinical

practice. As such, the intention of this review is to introduce broad concepts to hematology clinicians to assist in the evaluation and understanding of future DL implementations, as well as to provide an overview of the clinical uses currently being explored throughout patient care.

The fact that it is still early days for DL in hematology may be due to a lack of appropriate algorithm design, data availability, computational resources, and insufficient diseasespecific expertise involved in DL development.100 To the best of our knowledge, there are still no large clinically-annotated multi-modal public datasets for many hematologic conditions. In addition, critical morphological information in hematopathology may only be available at higher magnification levels, surpassing the limits of standard pathology scanners. Although these structural barriers continue to compromise the development of DL in hematology, rapid technological advances continue, and interest for DL within the academic community is growing.101

Though promising, the methods and conclusions from the numerous studies are heterogenous and challenging to compare. As yet, there is no standardized approach in DL research, reporting, or implementation. In the present overview, the majority of publications were evaluated by internal validation strategies, with the minority evaluated on external institution cohorts. Explaining model predictions were not ubiquitous, and few DL models were compared directly against human evaluation. Major government initiatives currently aim to standardize DL protocol design,102 and, despite the variance in outcome reporting in DL analyses, the SPIRIT-AI, STARD-AI, and CONSORT-AI initiatives aim to standardize future clinical trial design and reporting of artificial intelligence interventions.103-105

The research and results of DL analyses must be interpreted cautiously, as a number of practical and ethical issues have arisen in other domains of machine learning. CNN are prone to “memorize” the training set; thus, the initial high performance may fail to be carried forward on new previously unseen data. For this reason, it is imperative to evaluate DL models on external cohorts from separate institutions. If training data are acquired from multiple institutions, care must be given to correct for known “batch effects,” as DL models may infer site-specific artifact signatures not related to the underlying disease biology.106 Similarly, researchers should investigate explainability and error analysis to ensure that the models rely on scientifically reasonable features and ignore irrelevant factors. In addition, uncertainty in model predictions are rarely reported but are arguably necessary for clinical implementation of DL algorithms.

In this review, the majority of DL applications are aimed towards earlier phases of clinical care, such as automation and disease detection. DL in lymphoma resulted in the plurality of exploratory analyses, likely due to the importance of both radiologic and pathologic findings in the care of lym-

phoma patients. Though explored in a myriad of malignant and non-malignant conditions, notably lacking are DL applications in stem cell transplantation and many other nonmalignant processes where morphological assessment is paramount, such as thrombotic microangiopathies. Future work is needed to address large scale applications of DL in hematology. As a hematopathologist typically assesses histology specimens at different magnification levels, customized architectures to implement multi-scale image analysis should be explored. DL in solid oncology is widely used, in part due to the publicly available digital biopsy specimens provided by The Cancer Genome Atlas,107 of which there is no analogous database for hematologic conditions. In addition, the combination of multi-modal data structures that incorporate images in concert with flow cytometry, molecular analyses, cytogenetics, or other clinical factors may provide additional relevant features to improve DL models. While numerous considerations remain before large-scale implementation of DL is feasible, the development of new models and applications in hematology is rapidly increasing, and it is imperative for clinicians to be aware of the opportunities that DL may provide.

Disclosures

ATP reports support via grants from NIH/NCI U01-CA243075, grants from NIH/NIDCR R56-DE030958, grants from Horizon 2021-SC1-BHC, grants from DoD Breakthrough Cancer Research program BC211095, and grants from SU2C (Stand Up

References

1. Campanella G, Hanna MG, Geneslaw L, et al. Clinical-grade computational pathology using weakly supervised deep learning on whole slide images. Nat Med. 2019;25(8):1301-1309.

2. Kather JN, Pearson AT, Halama N, et al. Deep learning can predict microsatellite instability directly from histology in gastrointestinal cancer. Nat Med. 2019;25(7):1054-1056.

3. Coudray N, Ocampo PS, Sakellaropoulos T, et al. Classification and mutation prediction from non-small cell lung cancer histopathology images using deep learning. Nat Med. 2018;24(10):1559-1567.

4. Kather JN, Heij LR, Grabsch HI, et al. Pan-cancer image-based detection of clinically actionable genetic alterations. Nat Cancer. 2020;1(8):789-799.

5. Saillard C, Schmauch B, Laifa O, et al. Predicting survival after hepatocellular carcinoma resection using deep learning on histological slides. Hepatology. 2020;72(6):2000-2013.

6. Artificial intelligence and machine learning (AI/ML)-enabled medical devices. https://www.fda.gov/medicaldevices/software-medical-device-samd/artificial-intelligence-an d-machine-learning-aiml-enabled-medical-devices#resources. Accessed November 3, 2021.

7. Matek C, Krappe S, Münzenmayer C, Haferlach T, Marr C. An expert-annotated dataset of bone marrow cytology in hematologic malignancies. The Cancer Imaging Archive; 2021.

8. Matek C, Schwarz S, Marr C, Spiekermann K. A single-cell morphological dataset of leukocytes from AML patients and

to Cancer) Fanconi Anemia Research Fund – Farrah Fawcett Foundation Head and Neck Cancer Research Team Grant during the conduct of the study. ATP reports grants from Abbvie via the UChicago – Abbvie Joint Steering Committee Grant, and grants from Kura Oncology. ATP reports personal fees from Prelude Therapeutics Advisory Board, from Elevar Advisory Board, from Ayala Advisory Board, personal fees from Abbvie Advisory Board, and from Privo Advisory Board outside of submitted work. MES is an employee of Sonic Healthcare. AS has no conflicts of interest to disclose.

Contributions

All authors conceived the manuscript. AS wrote the manuscript. All authors approved the final version.

Funding

ATP reports this study has been funded in whole or in part with Federal funding by the NCI-DOE Collaboration established by the U.S. Department of Energy (DOE) and the National Cancer Institute (NCI) of the National Institutes of Health, Cancer Moonshot Task Order N. 75N91019F00134 and under Frederick National Laboratory for Cancer Research Contract 75N91019D00024. This work was performed under the auspices of the U.S. Department of Energy by Argonne National Laboratory under Contract DE-AC02-06-CH11357.

Data-sharing statement

No applicable.

non-malignant controls. The Cancer Imaging Archive; 2019.

9. Schouten JPE, Matek C, Jacobs LFP, Buck MC, Bosnacki D, Marr C. Tens of images can suffice to train neural networks for malignant leukocyte detection. Sci Rep. 2021;11(1):7995.

10. Matek C, Schwarz S, Spiekermann K, Marr C. Human-level recognition of blast cells in acute myeloid leukaemia with convolutional neural networks. Nature Machine Intelligence. 2019;1(11):538-544.

11. Matek C, Krappe S, Munzenmayer C, Haferlach T, Marr C. Highly accurate differentiation of bone marrow cell morphologies using deep neural networks on a large image data set. Blood. 2021;138(20):1917-1927.

12. Sidhom JW, Siddarthan IJ, Lai BS, et al. Deep learning for diagnosis of acute promyelocytic leukemia via recognition of genomically imprinted morphologic features. NPJ Precis Oncol. 2021;5(1):38.

13. Eckardt JN, Schmittmann T, Riechert S, et al. Deep learning identifies acute promyelocytic leukemia in bone marrow smears. BMC Cancer. 2022;22(1):201.

14. Eckardt JN, Middeke JM, Riechert S, et al. Deep learning detects acute myeloid leukemia and predicts NPM1 mutation status from bone marrow smears. Leukemia. 2022;36(1):111-118.

15. Chollet F. Deep learning with Python. 1st ed. USA: Manning Publications Co.; 2017.

16. Murphy KP. Probabilistic machine learning: an introduction. The MIT Press; 2022.

17. Liu Y, Chen PC, Krause J, Peng L. How to read articles that use machine learning: Users' Guides to the medical literature. JAMA. 2019;322(18):1806-1816.

18. Bankhead P, Loughrey MB, Fernandez JA, et al. QuPath: open source software for digital pathology image analysis. Sci Rep. 2017;7(1):16878.

19. Dolezal J, Kochanny S, Howard F. Slideflow: a unified deep learning pipeline for digital histology: Zenodo; 2022.

20. Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions. In: 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR); June 7-12 2015; Boston, MA, USA. IEEE; c2015. p. 1-9.

21. Olah C, Satyanarayan A, Johnson I, et al. The building blocks of interpretability. Distill. 2018;3(3).

22. Deng J, Dong W, Socher R, Li L, Kai L, Li F-F. ImageNet: a largescale hierarchical image database. In: 2009 IEEE Conference on Computer Vision and Pattern Recognition; 20-25 June 2009; Miami, FL, USA. IEEE; c2009. p. 248-255.

23. Riasatian A, Babaie M, Maleki D, et al. Fine-tuning and training of densenet for histopathology image representation using TCGA diagnostic slides. Med Image Anal. 2021;70:102032.

24. Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16x16 words: transformers for image recognition at scale. arXiv. 3 Jun 2021. doi: 10.48550/arXiv.2010.11929 [preprint, not peer reviewed].

25. Bianco S, Cadene R, Celona L, Napoletano P. Benchmark analysis of representative deep neural network architectures. IEEE access. 2018;6:64270-64277.

26. Hinton GE, Salakhutdinov RR. Reducing the dimensionality of data with neural networks. Science. 2006;313(5786):504-507.

27. Combalia M, Vilaplana V. Monte-Carlo sampling applied to multiple instance learning for histological image classification. In: Stoyanov D, Taylor Z, Carneiro G, et al., eds. Deep learning in medical image analysis and multimodal learning for clinical decision support; 20 Sep 2018; Granada, Spain. Springer International Publishing; c2018. p. 274-281.

28. Dietterich TG, Lathrop RH, Lozano-Pérez T. Solving the multiple instance problem with axis-parallel rectangles. Artificial Intelligence. 1997;89(1-2):31-71.

29. Ilse M, Tomczak J, Welling M. Attention-based deep multiple instance learning. In: Proceedings of the 35th International Conference on Machine Learning; 10-15 July 2018; Stockholm, Sweden. PMLR; c2018. p. 2127-2136.

30. Sadafi A, Makhro A, Bogdanova A, et al. Attention based multiple instance learning for classification of blood cell disorders. In: Medical image computing and computer assisted intervention –MICCAI 2020; 4-8 Oct 2020; Lima, Peru. Springer International Publishing; c2020. p. 246-256.

31. Lu MY, Williamson DFK, Chen TY, Chen RJ, Barbieri M, Mahmood F. Data-efficient and weakly supervised computational pathology on whole-slide images. Nat Biomed Eng. 2021;5(6):555-570.

32. Hooker S, Erhan D, Kindermans P-J, Kim B. A benchmark for interpretability methods in deep neural networks. In: 33rd Conference on Neural Information Processing Systems (NeurIPS 2019); Vancouver, Canada. c2019. p. 9734-9745.

33. Raghu M, Unterthiner T, Kornblith S, Zhang C, Dosovitskiy A. Do vision transformers see like convolutional neural networks? In: Advances in neural information processing systems; c2021. p. 12116-12128.

34. Ghassemi M, Oakden-Rayner L, Beam AL. The false hope of current approaches to explainable artificial intelligence in health care. Lancet Digit Health. 2021;3(11):e745-e750.

35. Smilkov D, Thorat N, Kim B, Viégas F, Wattenberg M.

SmoothGrad: removing noise by adding noise. arXiv. 2017 Jun 12. doi: 10.48550/arXiv.1706.03825. [preprint, not peer reviewed].

36. Rudin C. Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. Nat Mach Intell. 2019;1(5):206-215.

37. Krause J, Grabsch HI, Kloor M, et al. Deep learning detects genetic alterations in cancer histology generated by adversarial networks. J Pathol. 2021;254(1):70-79.

38. Fan H, Zhang F, Xi L, Li Z, Liu G, Xu Y. LeukocyteMask: an automated localization and segmentation method for leukocyte in blood smear images using deep neural networks. J Biophotonics. 2019;12(7):e201800488.

39. Alam MM, Islam MT. Machine learning approach of automatic identification and counting of blood cells. Healthc Technol Lett. 2019;6(4):103-108.

40. Vajen B, Hanselmann S, Lutterloh F, et al. Classification of fluorescent R-Band metaphase chromosomes using a convolutional neural network is precise and fast in generating karyograms of hematologic neoplastic cells. Cancer Genet. 2022;260-261:23-29.

41. Jemaa S, Fredrickson J, Carano RAD, Nielsen T, de Crespigny A, Bengtsson T. Tumor segmentation and feature extraction from whole-body FDG-PET/CT using cascaded 2D and 3D convolutional neural networks. J Digit Imaging. 2020;33(4):888-894.

42. Xu L, Tetteh G, Lipkova J, et al. Automated whole-body bone lesion detection for multiple myeloma on (68)Ga-Pentixafor PET/CT imaging using deep learning methods. Contrast Media Mol Imaging. 2018;2018:2391925.

43. Weisman AJ, Kieler MW, Perlman SB, et al. Convolutional neural networks for automated PET/CT detection of diseased lymph node burden in patients with lymphoma. Radiol Artif Intell. 2020;2(5):e200016.

44. Weisman AJ, Kim J, Lee I, et al. Automated quantification of baseline imaging PET metrics on FDG PET/CT images of pediatric Hodgkin lymphoma patients. EJNMMI Phys. 2020;7(1):76.

45. Sadik M, Lopez-Urdaneta J, Ulen J, et al. Artificial intelligence could alert for focal skeleton/bone marrow uptake in Hodgkin's lymphoma patients staged with FDG-PET/CT. Sci Rep. 2021;11(1):10382.

46. Yang Y, Tang Y, Gao R, et al. Validation and estimation of spleen volume via computer-assisted segmentation on clinically acquired CT scans. J Med Imaging (Bellingham). 2021;8(1):014004.

47. Liu W, Liu M, Guo X, et al. Evaluation of acute pulmonary embolism and clot burden on CTPA with deep learning. Eur Radiol. 2020;30(6):3567-3575.

48. de Haan K, Ceylan Koydemir H, Rivenson Y, et al. Automated screening of sickle cells using a smartphone-based microscope and deep learning. NPJ Digit Med. 2020;3(1):76.

49. Shaw M, Claveau R, Manescu P, et al. Optical mesoscopy, machine learning, and computational microscopy enable high information content diagnostic imaging of blood films. J Pathol. 2021;255(1):62-71.

50. Huber N, Anderson T, Missert A, et al. Clinical evaluation of a phantom-based deep convolutional neural network for wholebody-low-dose and ultra-low-dose CT skeletal surveys. Skeletal Radiol. 2022;51(1):145-151.

51. Theruvath AJ, Siedek F, Yerneni K, et al. Validation of deep learning-based augmentation for reduced (18)F-FDG dose for PET/MRI in children and young adults with lymphoma. Radiol Artif Intell. 2021;3(6):e200232.

52. Rajaraman S, Antani SK, Poostchi M, et al. Pre-trained

convolutional neural networks as feature extractors toward improved malaria parasite detection in thin blood smear images. PeerJ. 2018;6:e4568.

53. Rajaraman S, Silamut K, Hossain MA, et al. Understanding the learned behavior of customized convolutional neural networks toward malaria parasite detection in thin blood smear images. J Med Imaging (Bellingham). 2018;5(3):034501.

54. Kuo PC, Cheng HY, Chen PF, et al. Assessment of expert-level automated detection of plasmodium falciparum in digitized thin blood smear images. JAMA Netw Open. 2020;3(2):e200206.

55. Manescu P, Shaw MJ, Elmi M, et al. Expert-level automated malaria diagnosis on routine blood films with deep neural networks. Am J Hematol. 2020;95(8):883-891.

56. Li S, Du Z, Meng X, Zhang Y. Multi-stage malaria parasite recognition by deep learning. Gigascience. 2021;10(6):giab040.

57. Lee SY, Chen CME, Lim EYP, et al. Image analysis using machine learning for automated detection of hemoglobin H inclusions in blood smears - a method for morphologic detection of rare cells. J Pathol Inform. 2021;12:18.

58. Doan M, Sebastian JA, Caicedo JC, et al. Objective assessment of stored blood quality by deep learning. Proc Natl Acad Sci U S A. 2020;117(35):21381-21390.

59. Acevedo A, Merino A, Boldu L, Molina A, Alferez S, Rodellar J. A new convolutional neural network predictive model for the automatic recognition of hypogranulated neutrophils in myelodysplastic syndromes. Comput Biol Med. 2021;134:104479.

60. Kimura K, Ai T, Horiuchi Y, et al. Automated diagnostic support system with deep learning algorithms for distinction of Philadelphia chromosome-negative myeloproliferative neoplasms using peripheral blood specimen. Sci Rep. 2021;11(1):3367.

61. Sahlol AT, Kollmannsberger P, Ewees AA. Efficient classification of white blood cell leukemia with improved swarm optimization of deep features. Sci Rep. 2020;10(1):2536.

62. Shafique S, Tehsin S. Acute lymphoblastic leukemia detection and classification of its subtypes using pretrained deep convolutional neural networks. Technol Cancer Res Treat. 2018;17:1533033818802789.

63. Mori J, Kaji S, Kawai H, et al. Assessment of dysplasia in bone marrow smear with convolutional neural network. Sci Rep. 2020;10(1):14734.

64. Sirinukunwattana K, Aberdeen A, Theissen H, et al. Artificial intelligence-based morphological fingerprinting of megakaryocytes: a new tool for assessing disease in MPN patients. Blood Adv. 2020;4(14):3284-3294.

65. Gehlot S, Gupta A, Gupta R. A CNN-based unified framework utilizing projection loss in unison with label noise handling for multiple myeloma cancer diagnosis. Med Image Anal. 2021;72:102099.

66. Li D, Bledsoe JR, Zeng Y, et al. A deep learning diagnostic platform for diffuse large B-cell lymphoma with high accuracy across multiple hospitals. Nat Commun. 2020;11(1):6004.

67. Syrykh C, Abreu A, Amara N, et al. Accurate diagnosis of lymphoma on whole-slide histopathology images using deep learning. NPJ Digit Med. 2020;3(1):63.

68. Sibille L, Seifert R, Avramovic N, et al. (18)F-FDG PET/CT uptake classification in lymphoma and lung cancer by using deep convolutional neural networks. Radiology. 2020;294(2):445-452.

69. Zhou Z, Jain P, Lu Y, et al. Computer-aided detection of mantle cell lymphoma on (18)F-FDG PET/CT using a deep learning convolutional neural network. Am J Nucl Med Mol Imaging. 2021;11(4):260-270.

70. Huang SC, Kothari T, Banerjee I, et al. PENet-a scalable deeplearning model for automated diagnosis of pulmonary embolism

using volumetric CT imaging. NPJ Digit Med. 2020;3(1):61.

71. Huang SC, Pareek A, Zamanian R, Banerjee I, Lungren MP. Multimodal fusion with deep neural networks for leveraging CT imaging and electronic health record: a case-study in pulmonary embolism detection. Sci Rep. 2020;10(1):22147.

72. Kainz B, Heinrich MP, Makropoulos A, et al. Non-invasive diagnosis of deep vein thrombosis from ultrasound imaging with machine learning. NPJ Digit Med. 2021;4(1):137.

73. Kwon JM, Cho Y, Jeon KH, et al. A deep learning algorithm to detect anaemia with ECGs: a retrospective, multicentre study. Lancet Digit Health. 2020;2(7):e358-e367.

74. Mitani A, Huang A, Venugopalan S, et al. Detection of anaemia from retinal fundus images via deep learning. Nat Biomed Eng. 2020;4(1):18-27.

75. Li S, Yang Q, Jiang H, Cortes-Vecino JA, Zhang Y. Parasitologistlevel classification of apicomplexan parasites and host cell with deep cycle transfer learning (DCTL). Bioinformatics. 2020;36(16):4498-4505.

76. Kimura K, Tabe Y, Ai T, et al. A novel automated image analysis system using deep convolutional neural networks can assist to differentiate MDS and AA. Sci Rep. 2019;9(1):13385.

77. Ahmed N, Yigit A, Isik Z, Alpkocak A. Identification of leukemia subtypes from microscopic images using convolutional neural network. Diagnostics (Basel). 2019;9(3):104.

78. Huang F, Guang P, Li F, Liu X, Zhang W, Huang W. AML, ALL, and CML classification and diagnosis based on bone marrow cell morphology combined with convolutional neural network: a STARD compliant diagnosis research. Medicine (Baltimore). 2020;99(45):e23154.

79. Achi HE, Belousova T, Chen L, et al. Automated diagnosis of lymphoma with digital pathology images using deep learning. Ann Clin Lab Sci. 2019;49(2):153-160.

80. Guan Q, Wan X, Lu H, et al. Deep convolutional neural network Inception-v3 model for differential diagnosing of lymph node in cytological images: a pilot study. Ann Transl Med. 2019;7(14):307.

81. Mohlman JS, Leventhal SD, Hansen T, Kohan J, Pascucci V, Salama ME. Improving augmented human intelligence to distinguish Burkitt lymphoma from diffuse large B-cell lymphoma cases. Am J Clin Pathol. 2020;153(6):743-759.

82. Miyoshi H, Sato K, Kabeya Y, et al. Deep learning shows the capability of high-level computer-aided diagnosis in malignant lymphoma. Lab Invest. 2020;100(10):1300-1310.

83. Yun J, Park JE, Lee H, Ham S, Kim N, Kim HS. Radiomic features and multilayer perceptron network classifier: a robust MRI classification strategy for distinguishing glioblastoma from primary central nervous system lymphoma. Sci Rep. 2019;9(1):5746.

84. Zhao J, Zhang M, Zhou Z, Chu J, Cao F. Automatic detection and classification of leukocytes using convolutional neural networks. Med Biol Eng Comput. 2017;55(8):1287-1301.

85. Lippeveld M, Knill C, Ladlow E, et al. Classification of human white blood cells using machine learning for stain-free imaging flow cytometry. Cytometry A. 2020;97(3):308-319.

86. Wu YY, Huang TC, Ye RH, et al. A hematologist-level deep learning algorithm (BMSNet) for assessing the morphologies of single nuclear balls in bone marrow smears: algorithm development. JMIR Med Inform. 2020;8(4):e15963.

87. Durant TJS, Olson EM, Schulz WL, Torres R. Very deep convolutional neural networks for morphologic classification of erythrocytes. Clin Chem. 2017;63(12):1847-1855.

88. Xu M, Papageorgiou DP, Abidi SZ, Dao M, Zhao H, Karniadakis GE. A deep convolutional neural network for classification of red blood cells in sickle cell anemia. PLoS Comput Biol. 2017;13(10):e1005746.

89. Zhou Y, Yasumoto A, Lei C, et al. Intelligent classification of

platelet aggregates by agonist type. Elife. 2020;9:e52779.

90. Rehman A, Abbas N, Saba T, Rahman SIU, Mehmood Z, Kolivand H. Classification of acute lymphoblastic leukemia using deep learning. Microsc Res Tech. 2018;81(11):1310-1317.

91. Bruck OE, Lallukka-Bruck SE, Hohtari HR, et al. Machine learning of bone marrow histopathology identifies genetic and clinical determinants in patients with MDS. Blood Cancer Discov. 2021;2(3):238-249.

92. Swiderska-Chadaj Z, Hebeda KM, van den Brand M, Litjens G. Artificial intelligence to detect MYC translocation in slides of diffuse large B-cell lymphoma. Virchows Arch. 2021;479(3):617-621.

93. Irshaid L, Bleiberg J, Weinberger E, et al. Histopathologic and machine deep learning criteria to predict lymphoma transformation in bone marrow biopsies. Arch Pathol Lab Med. 2022;146(2):182-193.

94. Jullien M, Tessoulin B, Ghesquieres H, et al. Deep-learning assessed muscular hypodensity independently predicts mortality in DLBCL patients younger than 60 years. Cancers (Basel). 2021;13(18):4503.

95. Cahan N, Marom EM, Soffer S, et al. Weakly supervised attention model for RV strain classification from volumetric CTPA scans. Comput Methods Programs Biomed. 2022;220:106815.

96. Cai S, Parker F, Urias MG, Goldberg MF, Hager GD, Scott AW. Deep learning detection of sea fan neovascularization from ultra-widefield color fundus photographs of patients with sickle cell hemoglobinopathy. JAMA Ophthalmol. 2021;139(2):206-213.

97. Doan M, Case M, Masic D, et al. Label-free leukemia monitoring by computer vision. Cytometry A. 2020;97(4):407-414.

98. Guo R, Hu X, Song H, et al. Weakly supervised deep learning for determining the prognostic value of (18)F-FDG PET/CT in extranodal natural killer/T cell lymphoma, nasal type. Eur J Nucl

Med Mol Imaging. 2021;48(10):3151-3161.

99. Lisson CS, Lisson CG, Mezger MF, et al. Deep neural networks and machine learning radiomics modelling for prediction of relapse in mantle cell lymphoma. Cancers (Basel). 2022;14(8):2008.

100. Kochanny SE, Pearson AT. Academics as leaders in the cancer artificial intelligence revolution. Cancer. 2021;127(5):664-671.

101. Radakovich N, Nagy M, Nazha A. Machine learning in haematological malignancies. Lancet Haematol. 2020;7(7):e541-e550.

102. New NCI-DOE collaboration project, IMPROVE, seeks deep learning model approaches. https://datascience.cancer.gov/news-events/news/new-nci-doecollaboration-project-improve-seeks-deep-learning-model-app roaches. Accessed May 21, 2022.

103. Cruz Rivera S, Liu X, Chan AW, et al. Guidelines for clinical trial protocols for interventions involving artificial intelligence: the SPIRIT-AI extension. Nat Med. 2020;26(9):1351-1363.

104. Liu X, Cruz Rivera S, Moher D, et al. Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension. Nat Med. 2020;26(9):1364-1374.

105. Sounderajah V, Ashrafian H, Golub RM, et al. Developing a reporting guideline for artificial intelligence-centred diagnostic test accuracy studies: the STARD-AI protocol. BMJ Open. 2021;11(6):e047709.

106. Howard FM, Dolezal J, Kochanny S, et al. The impact of sitespecific digital histology signatures on deep learning model accuracy and bias. Nat Commun. 2021;12(1):4423.

107. Cancer Genome Atlas Research N, Weinstein JN, Collisson EA, et al. The Cancer Genome Atlas Pan-Cancer analysis project. Nat Genet. 2013;45(10):1113-1120.

Strategies to optimize chimeric antigen receptor T-cell therapy in hematologic malignancies: Chinese experience

Wei Sun,1 Ai-Bin Liang,2 He Huang3 and Xiao-Jun Huang1

1Peking University People's Hospital, Peking University Institute of Hematology, National Clinical Research Center for Hematologic Disease, Beijing Key Laboratory of Hematopoietic Stem Cell Transplantation, Peking University, Beijing; 2Department of Hematology, Tongji Hospital, Tongji University School of Medicine, Shanghai and 3Bone Marrow Transplantation Center, Institute of Hematology, First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China.

Abstract

Correspondence: X-J. Huang huangxiaojun@bjmu.edu.cn

Received: October 26, 2022.

Accepted: February 7, 2023.

Early view: February 16, 2023.

https://doi.org/10.3324/haematol.2022.282316

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Chimeric antigen receptor (CAR) T-cell therapy has emerged as a promising form of adoptive T-cell immunotherapy for selected hematologic malignancies including leukemia, lymphoma and multiple myeloma. China has become the country with the largest number of registered CAR T-cell trials. Despite the remarkable clinical outcomes achieved with CAR Tcell therapy, challenges such as disease relapse, the process of manufacturing the CAR T cells and safety have limited the therapeutic efficacy of CAR T cells in hematologic malignancies. In this period of innovation, several clinical trials have reported the design of CAR directed at new targets in hematologic malignancies. In this review, we comprehensively summarize the contemporary landscape and clinical development of CAR T-cell therapy in China. In addition, we present strategies for further improving the clinical utility of CAR T-cell therapy, such as increasing the efficacy and response duration, in hematologic malignancies.

Introduction

Over the past few decades, treatment strategies for hematologic malignancies have made tremendous headway. However, the morbidity and mortality rates attributed to these malignancies remain substantial.1 Advances in molecular genetics have paved the way for further in-depth understanding of the interaction between the immune system and cancer cells and revealed the great potential of T cells for use in immunotherapy of hematologic malignancies.1 At this breakthrough juncture, multiple iterations of adoptive cell therapies have been designed to overcome immune evasion mechanisms in cancer by directly targeting cancer cells and activating specific immune responses to tumors.2 One such adoptive T-cell-centered immunotherapy, which has been successfully translated from bench to bedside, is genetically engineered chimeric antigen receptors (CAR) that can recognize cancer-associated antigens, leading to T-cell activation, proliferation and memory.2 CAR T-cell therapies engineered against different tumor antigens have shown astonishing efficacy and durable clinical responses in many types of malignancies, especially hematopoietic ones such as acute lymphoblastic leukemia (ALL), large B-cell lymphoma (LBCL) and multiple myeloma (MM),

and have revolutionized the therapeutic landscape of cancer immunotherapy. Moreover, given their great potential for continuous optimization, CAR T-cell therapies are attractive replacements of conventional therapies (chemotherapy, radiation therapy, stem cell transplantation) as new targets continue to emerge. Approval of two CAR T-cell products, tisagenlecleucel (Kymriah) for the treatment of Bcell ALL (B-ALL) in pediatric and young adult patients (aged ≤25 years)3 and axicabtagene ciloleucel (Yescarta) for the treatment of LBCL4 in adult patients by the USA Food and Drug Administration (FDA) in 2017 was a milestone in the history of cancer research. In 2022, ciltacabtagene autoleucel (Carvykti) was approved by the FDA for the treatment of patients with MM, becoming the sixth approved CAR Tcell therapy.

Due to the remarkable success of CAR T-cell therapy, the number of clinical trials on this treatment has increased rapidly across the globe, with USA and China being the major forces contributing to about 33% of all such trials. In China, T-cell immunotherapy has been widely used for treating cancer, and several studies have revealed the remarkable antitumor effects of CAR T cells. These findings inspired researchers from China to implement further domestic CAR T-cell trials and clinical trials on precision im-

munotherapy (CAR T-cell therapy) have expanded rapidly in the country, which has taken the place of the USA as the nation with the most CAR T-cell studies (~444 as of 2021) and nowadays plays a paramount role in developing innovative strategies for CAR T-cell therapy. In this review, we comprehensively describe the current status of CAR T-cell trials in China. In addition, we present an updated overview of CAR T-cell therapeutic options in hematologic malignancies, as well as strategies to improve the efficacy and safety of CAR T-cell therapy, which will have a tremendous impact on the field of T-cell immunotherapy.

Approval of chimeric antigen receptor T-cell therapies in China

Axicabtagene ciloleucel was the first commercially available autologous CD19-directed CAR T-cell therapy approved in China for the treatment of adult patients with relapsed or refractory (r/r) LBCL, including diffuse large B-cell lymphoma (DLBCL), after two or more lines of systemic therapy. The approval was based on the results of a single-arm, open-label, multicenter bridging trial (FKC876-2018-001)

ChiCTR1800019661, in which 79.2% of patients achieved a response after a single infusion of axicabtagene ciloleucel.5 Relmacabtagene autoleucel (Carteyva) was the second approved CD19-targeting CAR construct for the treatment of LBCL after at least two prior lines of systemic therapy.6 The approval was based on the findings of the RELIANCE study, in which patients with r/r LBCL who failed at least two lines of therapy were treated with relmacabtagene autoleucel. The overall response rate (ORR) was 75.9%, with a complete response (CR) rate of 51.7% and a 12-month overall survival rate of 76.8% (as of June 17, 2020, the data cutoff).7 Table 1 summarizes the CAR T-cell therapies approved by the Chinese National Medical Products Administration (NMPA) or FDA, with data updated to March 2022.

Overview and characteristics of the clinical development of chimeric antigen receptor T-cell therapy in China

We retrieved clinical trials on CAR T-cell therapy from the ClinicalTrials.gov website using the keywords “CAR T” or “CAR-T” or “chimeric antigen receptor T cell” or “chimeric

Ciltacabtagene autoleucel (Carvykti®)

Axicabtagene ciloleucel (Yescarta®)

Relmacabtagene autoleucel (Carteyva®)

Legend Bio and Janssen

Kite Biotechnology

ID: identity; FDA: Food and Drug Administration; NMPA: National Medical Products Administration; r/r: relapsed/refractory; ALL: acute lymphoblastic leukemia; DLBCL: diffuse large B-cell lymphoma; FL: follicular lymphoma; MCL: mantle cell lymphoma; B-ALL: B-cell acute lymphoblastic leukemia; LBCL: large B-cell lymphoma; MM: multiple myeloma.

Table 1. Chimeric antigen receptor T-cell products approved by the American Food and Drug Administration or the Chinese National Medical Products Administration.

antigen receptor”, followed by manual verification to exclude the non-CAR T-cell therapy trials. As of December 2022, there were 458 trials from China reported and/or registered at ClinicalTrials.gov. The majority involved investigations on hematologic malignancies (73%, n=337) followed by solid tumors (24%, n=111) (Figure 1A). Since research on CAR T-cell therapy is mainly distributed in China and the USA, we also compared the clinical trials in the two countries. While the percentages of trials were comparable between China and the USA, China has a greater number of trials registered than the USA across all tumor types (Figure 1A). In comparison to Chinese CAR T-cell therapy clinical trials in hematologic malignancies, which started in 2012, such trials started in 2009 in the USA. The number of trials in hematologic malignancies was comparable in the two countries between 2012 and 2015, and then in 2016 China surpassed the USA in number of CAR T-cell clinical studies in hematologic malignancies and is still leading (Figure 1B). In terms of the distribution of the phase of the trials, the same trend was seen in China and the USA with the highest percentage of trials being phase I trials and the lowest percentage being phase III trials. As

regard to the sample size of the clinical studies on CAR Tcell therapy in hematologic malignancies, this varied between 10-30 patients in China with only a few studies recruiting approximately 100 patients. In contrast, a higher proportion of trials in the USA had a sample size of >30 patients (Figure 1C, D).

CD19 and B-cell maturation antigen (BCMA) are the most common antigens being targeted in China, with a total of 127 and 36 trials, respectively, in hematologic malignancies. A similar trend was observed in trials in hematologic malignancies in the USA (Figure 2A, B). While CD7 (17 trials) has been more investigated in Chinese CAR T-cell trials in hematologic malignancies, CD4 (2 trials) and CD33 (3 trials) have been studied more frequently in trials in the USA (Figure 2A, B). Compared with the USA, China has a higher proportion of trials involving the use of multi-target (≥2) CAR T cells. Most of the multi-target CAR T-cell clinical trials in hematologic malignancies in China have involved the combination of CD19 and CD22 (25 trials) followed by CD19 combined with CD20 (17 trials). The multi-agent trials in the USA have also most commonly involved the same targets (Figure 2C, D). Meanwhile, Chi-

Figure 1. Overview of the clinical development of chimeric antigen receptor T-cell therapy and trials in hematologic malignancies. (A) Proportions of chimeric antigen receptor (CAR) T-cell clinical trials in hematologic and solid tumors in China and the USA. (B) Number of CAR T-cell clinical trials in hematologic malignancies in China and the USA from 2009-2021. (C, D) The percentages of CAR T-cell trials in hematologic malignancies by number of patients enrolled in the USA (C) and in China (D)

nese researchers have more frequently explored CAR Tcell therapy with BCMA combined with another target. Concomitant treatments used along with CAR T-cell therapy were somewhat similar between the USA and China with the major differences being the use of chidamide/decitabine (2 trials) and dasatinib (1 trial) in leukemia/lymphoma and MM trials in China versus PI3K δ / γ inhibitors (1 trial) and bi-specific antibodies (1 trial) used in leukemia/lymphoma trials and immunomodulatory imide drugs, γ -secretase inhibitors and antibody-drug conjugates used in MM in trials in the USA (Figure 3A-D).

Strategies to improve the efficacy of chimeric antigen receptor T-cell therapy: Chinese experience

In China, although CAR T-cell therapy has entered the therapeutic arsenal for hematologic malignancies and it has demonstrated efficacy in leukemia, lymphoma and MM, approximately 30-60% of cases relapse after this therapy.8 In addition, CAR T-cell therapy has several shortcomings such as difficulties in identifying ideal target tumor antigens, inhibition and resistance, antigen escape, decreased persistence and expansion of CAR T cells, sus-

Figure 2. Antigen targets of chimeric antigen receptor T-cell therapy in hematologic malignancies. (A, B) Top ten targets of chimeric antigen receptor (CAR) T-cell therapy in hematologic malignancies in the USA (A) and in China (B). (C, D) Multi-targets used in CAR T cells in hematologic malignancies in the USA (C) and in China (D). BCMA: B-cell maturation antigen; ND: no data.

ceptibility of CAR T cells to an immunosuppressive microenvironment, limited efficacy during rescue therapy, and life-threatening toxicities. These problems associated with CAR T cells still exist and pose enormous challenges, because they undermine the prospective efficacy and durability of CAR T cells.9 Several strategies are currently under investigation to address these problems. In this section, we summarize the current status of clinical development of CAR T-cell therapy in leukemia, lymphoma and MM in China (Tables 2-7).10-58 Furthermore, we detail a series of promising strategies to optimize the curative effect of CAR T-cell therapy (Figure 4).

Acute lymphoblastic leukemia

ALL is a hematologic malignancy that originates from malignant precursor B or T lymphocytes with a morbidity rate of 0.69 per 100,000 persons in China.9 With conventional chemotherapy and standardized intensive therapies, many patients still suffer from r/r disease, with a relapse rate of

15-20% in pediatric B-ALL and 50% in adult B-ALL. Indeed, r/r disease still remains a major obstacle in the therapy of ALL.

In China, the first CAR T-cell therapy for B-ALL targeting CD19 was reported in 2013, further studies have developed rapidly with the greatest efficacy of CD19-targeted CAR Tcell therapy demonstrated in r/r B-ALL. Although it is hoped that CAR T cells targeting CD19 will provide an additional CR for most r/r patients, durable remissions are difficult to achieve due to subsequent relapses.59 Because resistance and relapse are intractable issues that preclude further development of CAR T-cell therapy in ALL, strategies to improve the efficacy of CAR T cells and repeated treatment after recurrence need to be considered. Toxicities and safety events may also prevent patients from benefiting from CAR T-cell therapy.

Autologous CAR T-cell therapy has been susceptible to failure as a consequence of limited quantity (in patients receiving lymphodepletion and/or chemotherapy) and poor

Figure 3. Different combination treatments used together with chimeric antigen receptor T cells. (A, B) Different combination treatments used together with chimeric antigen receptor (CAR) T cells in leukemia trials in the USA (A) and in China (B). (C, D) Different combination treatments used together with chimeric CAR T cells in multiple myeloma trials in the USA (C) and in China (D). PD-1: programmed cell death protein 1; PD-L1: programmed death ligand 1; mAb: monoclonal antibody; PI3K: phosophoinositide 3-kinase; IL-2: interleukin-2; BTK: Bruton tyrosine kinase; ADC: antibody-drug conjugate; IMiD: immunomodulatory drug.

Ref.

Key fi ndings

CR or CRi was achieved in 24 of 30 evaluable r/r B-ALL patients in whom previous CD19 CAR T-cell therapy had failed.

Phase

Following bi-specific CD19/CD22 CAR T-cell therapy, all 6 r/r B-ALL patients experienced MRD-negative CR.

I

-

Trial ID

ChiCTR-OIC- 17013523

NCT03185494

CAR T-cell/object

Seventeen of 20 pediatric patients with r/r B-ALL remained in remission at the cutoff date, resulting in a leukemia-free survival rate of 79.5% at 12 months.

I

Combination strategy of sequential CD19 and CD22 CAR T-cell therapy significantly improved the long-term survival (OS and EFS rates were 88.5% and 67.5%, respectively) in B-ALL patients who relapsed after transplantation.

I

Both CD28 and 4-1BB CAR T cells produced responses, although they differed for response pattern (peak reaction time, reaction lasting time and reaction degree), adverse events, cytokine secretion and level of immunesuppressive factors.

The performance of 4-1BB CAR T cells was superior to that of CD28 CAR T cells in suppressing CD19 + B-ALL.

Among the 14 r/r ALL patients naïve to previous CAR T-cell therapy, 13 achieved CR or CRi on day 30, whereas 1 of the 3 patients who failed a second murine CAR T-cell infusion achieved CR after hCART19s infusion.

Humanized selective CAR T cells were infused into 5 patients who had relapsed after receiving murine CAR T-cell treatments and 4 patients achieved molecular CR.

90% of the 20 r/r B-ALL patients treated with infusions of CNCT19 cells reached CR or CRi within 28 days.

19

20

Strategy

CD22 CAR T-cell

ChiCTR-OIB- 17013670

ChiCTRONC- 17013648

I/II

-

I

I

Within 1 month after CAR T-cell infusions, 10 patients achieved CR and 9 achieved molecular CR.

II

I

Bi-specific CAR T cells targeting CD19 and CD22

Sequential CD19/CD22 CAR T cells

Combination of CD19 and CD22 CAR T cells

NCT02349698

CD19 CAR T cells containing CD28 or 4-1BB

NCT03173417

4-1BBor CD28-based CD19 CAR T cells

NCT02782351

Humanized CD19-targeted CAR T cells

ChiCTR1800014761

Humanized selective CD19 CAR T cells

NCT02975687

CD19 CAR T cells binding to different CD19 epitopes

NCT03919240

NCT03614858

NCT03896854

Combination of decitabine and CAR T cells in r/r acute leukemia with TP53 alterations

NCT04227015

CRISPR-edited universal offthe-shelf CD19/CD22 dual-targeted CAR T cells (CTA101)

Alternative targets

Multiple targets

CAR structure

Combination therapy

Universal CAR T cells

CAR: chimeric antigen receptor; ID: identity; Ref: reference; CR: complete remission; CRi: complete remission with incomplete blood count recovery; r/r: relapsed/refractory; B-ALL: B-cell acute lymphoblastic leukemia; MRD: minimal residual disease; OS: overall survival;

EFS: event-free survival; hCART19s: humaniz ed CD19-speci fi c CAR T cells; GvHD: graftversushost disease.

Haematologica | 108 August 2023 2016 REVIEW ARTICLE - CAR T-cell therapy in China W. Sun et al.

Monitor/ biomarker

Real-time monitoring of Th1/Th2 cytokine pattern using cytometric bead array technology

Virus reactivation Efficacy and safety of CAR T-cell therapy in patients with concomitant HBV infection

CRS management Using corticosteroids instead of tocilizumab as the first-line agent to manage CRS

- - CD19 CAR T-cell therapy can be safely administered to patients with relapsed and refractory leukemia under "real-time" monitoring of a simple 6- Th1/Th2 cytokine pattern.

NCT02822326 - CAR T cells from HBV-positive patients exerted potent antileukemia effects without inducing HBV reactivation under close monitoring of HBV DNA and liver function.

ChiCTR-OIC-17013623, ChiCTR-ONC-17013648, ChiCTR-OIC-17013523

- Corticosteroids did not influence the efficacy and kinetics of CAR T cells for B-ALL.

CAR: chimeric antigen receptor; ID: identity; Ref: reference; Th: T helper cell; HBV: hepatitis B virus; CRS: cytokine release syndrome; B-ALL: B-cell acute lymphoblastic leukemia.

quality (due to apheresis); therefore, allogeneic CAR T-cell therapy has become an attractive replacement. Nevertheless, allogeneic CAR T-cell therapy has its own challenges, such as graft-versus-host disease (GvHD) and graft rejection. Multiple studies have indicated that the failure of allogeneic CAR T-cell expansion related to GvHD and graft rejection in r/r ALL patients receiving allogeneic CD19-directed CAR T-cell therapy before or after allogeneic hematopoietic stem cell transplantation (HSCT) can be avoided or minimized. Zhang et al. presented the safety and efficacy of donor-derived anti-CD19 CAR T cells in 43 subjects with B-ALL relapsing after allotransplants: approximately 79% (n=34) patients achieved a CR. Two subjects had grade ≤2 acute GvHD.60 On the other hand, Jin et al. described the first-in-human use of HLA-matched allogeneic CAR T cells (CD19-directed) before allogeneic HSCT: 75% (3/4) patients achieved a CR and no GvHD was observed.61 Recent advances in allogeneic CAR T cells have focused on off-the-shelf products called universal CAR T cells. Huang et al. developed a CRISPR-edited universal off-theshelf CD19/CD22 dual-targeted CAR T-cell product for the therapy of r/r ALL patients and documented that 83.3% (5/6) patients achieved minimal residual disease-negative CR with manageable adverse events.20

So far, the research on CAR T-cell therapy has been mainly focused on B-ALL, and there are relatively few studies on T-ALL. CD7 is highly expressed on the surface of T-ALL/Tcell lymphoblastic lymphoma T cells and is considered a viable CAR T-cell therapeutic target. In a single-center, phase I trial, Pan et al. administered anti-CD7 CAR T cells, manufactured from either previous stem-cell transplantation donors or new donors, to patients with r/r T-ALL, in whom the CR rate was 90% and adverse events were reversible.62 Lu et al. described a novel approach using patient- or donor-derived “naturally selected” CD7-targeted CAR T cells (NS7CAR) without additional CD7 gene ablation

or protein expression blockade. In their first-in-human, phase I trial (clinicaltrials gov. Identifier: NCT04572308), 20 patients with r/r T-ALL and T-cell lymphoblastic lymphoma were treated with NS7CAR. Nineteen patients achieved minimal residual disease-negative CR in the bone marrow by day 28, and five of nine patients achieved extramedullary CR.63 These results indicate that CD7 CAR T-cell therapy is a safe and highly effective treatment for T-ALL. More patients and longer follow-up are needed for validation.

It should be noted that 10-30% of B-ALL patients relapse because of antigen escape (antigen-negative relapse). However, another common cause is the loss of CAR T cells, leading to antigen-positive relapse. In antigen-positive relapse, the components of CAR constructs (costimulatory domains and scFv) can influence the potency and persistence of CAR T cells. Several studies identified the therapeutic potential of anti-CD19 CAR containing either CD28 or 4-1BB co-stimulatory signaling in ALL, and the results hinted that 4-1BB-based CAR T cells have greater efficacy (as a result of stronger persistence) than CD28-based CAR T cells. Chen et al. recently initiated a trial with third-generation CD19 CAR T cells, combining 41BB and CD28 signaling domains, in the treatment of adults with r/r B-ALL: the results are awaited.64 In addition, numerous studies have shown the therapeutic potential of humanized scFv CD19-targeted CAR T-cell therapy in B-ALL patients with no response or who relapsed after prior murine CD19 CAR T-cell therapy. Furthermore, CD19 CAR T cells with scFv capable of binding to different CD19 epitopes may provide an alternative for patients who undergo CD19-positive relapse. Wang et al described a new CD19 CAR T-cell with a scFv that interacts with an epitope of the human CD19 antigen that is distinct from that recognized by the current FMC63 clone. This approach may be an alternative choice for some patients,

Table 3. Chimeric antigen receptor T-cell therapy in leukemia: overview of current strategies to enhance safety.

18F-FDG PET/CT in predicting efficacy and adverse effects of CAR T-cell therapy

- - r/r NHL patients with higher baseline tumor burdens were found to have significantly increased CRS incidence and cytokine levels. The metabolic parameters including standardized uptake value, metabolic tumor volume and total lesion glycolysis were closely related to OS, PFS, and CRS in r/r NHL patients treated with CAR T cells.

Virus reactivation Risk of HBV reactivation after CD19-CAR Tcell therapy

Humanized anti-CD19 CAR T-cell therapy in patients with chronic and resolved HBV infection

Infection Incidence and risk factors associated with infection after CAR T-cell therapy

NCT02537977 - r/r DLBCL patients with chronic HBV infection who receive CD19 CAR-T cell therapy are at risk of HBV reactivation, especially HBeAg-positive patients. Adequate antiviral prophylaxis is essential to prevent HBV reactivation in these patients.

ChiCTR1800019622, ChiCTR1800018059

- Among 15 patients with resolved HBV infection, 2 received antiviral prophylaxis, and the other 13 did not experience HBV reactivation without antiviral prophylaxis. One patient with resolved HBV infection experienced HBV reactivation 6 months after human CAR T-cell therapy and sequential allogeneic HSCT.

ChiCTR-ORN16008948, ChiCTROIC-17011310, ChiCTR1800015575

- The incidence of infection was 15.8% in the NHL cohort. An absolute neutrophil count <500 cells/mm3 before CTI and infection during prior treatment were independent risk factors associated with a significantly increased infection density within 28 days after CTI. Similarly, corticosteroid treatment during CRS was an independent risk factor during days 29-180 after CTI.

CAR: chimeric antigen receptor; ID: identity; Ref: reference; 18F-FDG: 18 fluoro-deoxyglucose; PET: positron emission tomography; CT: computed tomography; r/r: relapsed/refractory; NHL: non-Hodgkin lymphoma; CRS: cytokine release syndrome; OS: overall survival; PFS: progression-free survival; HBV: hepatitis B virus; DLBCL: diffuse large B-cell lymphoma; HSCT: hematopoietic stem cell transplantation; CTI: CAR T-cell infusion.

especially those with CD19-positive relapse from CAR Tcell therapy based on the FMC63 clone.18 The mechanism of developing an antigen-negative response is multifactorial in origin. As a consequence, the development of strategies to overcome antigen-negative relapse is complex. Recent studies have suggested that alternative targets (CD22, CD38)/combinations of multiple targets11 might benefit ALL patients with antigen-negative response. However, Huang’s group reported a retrospective comparison study of single CD19- and bi-specifi c CD19/CD22-targeted CAR T-cell therapy in patients with r/r ALL and suggested that the CR rate to the bi-specific treatment was comparable to the CR rate to the monospecific treatment and did not reduce the recurrence rate in r/r ALL.65 With respect to the diverse factors affecting the efficacy of CAR T cells, including the characteristics of the patients, the manufacturing process, and the infusion process of bi-specific products, more prospective studies are warranted in order to demonstrate that bispecific CAR T cells could be an option to overcome antigen escape and delay the time of recurrence.

Given that severe adverse events may occur during CAR T-cell therapy, in particular those related to cytokine release, the management of these adverse events is very im-

portant. Tong et al. reported on the use of corticosteroids instead of tocilizumab as the first-line agent to manage cytokine release syndrome (CRS), and described that, even at high doses, corticosteroids did not undermine the efficacy of the CAR T cells, with regard to either proliferation or duration.23 In addition, there was an exploratory attempt in one case to manage CRS, following the use of shRNAIL6-modified CAR T cells, with suppression of IL6 gene expression in the CAR T cells.66

Subjects with viral infection are usually excluded from clinical trials on CAR T cells, as elimination of B cells by anti-CD19 CAR T cells may lead to the reactivation of hepatitis B virus (HBV) and related hepatitis in the case of HBV infection. However, reports from Wen et al. and Li et al. indicated that HBV infection may not be an absolute contraindication to CAR T-cell therapy for r/r ALL patients if effective antiviral drugs are administered properly.22,46

Acute myeloid leukemia

Acute myeloid leukemia (AML) is an aggressive heterogeneous malignant disease of hematopoietic stem and progenitor cells and affects the blood and bone marrow. A wide variety of therapeutic strategies, including chemotherapy, immunotherapy and targeted therapy, has been

Table 5. Chimeric antigen receptor T-cell therapy in lymphoma: overview of current strategies to enhance safety.

Multi-target A bi-specific CAR T-cell therapy targeting BCMA and CD38 (BM38 CAR T cells)

ChiCTR1800018143 I

Sixteen patients, including 10 (62.5%) with genetic abnormalities and 5 (31.25%) with extramedullary lesions, received BM38 CAR T cells. Fourteen (87.5%) patients achieved an overall response with 8 (50%) sCR, 2 (12.5%) VGPR and 4 (25.00%) PR. Fourteen (87.5%) achieved bone marrow MRD-negative status. PFS rate at 9 months was 75%.

CD38 and BCMA bi-specific CAR T cells

Anti-CD19 and anti-BCMA CAR T-cell sequential infusion followed by lenalidomide maintenance after ASCT

ChiCTR1900026286 - Of 16 evaluable patients, 14 (87.5%) responded to the treatment, including 13 with sCR and 1 with PR, while 2 patients did not respond. At a median follow-up of 11.5 months, of the 13 patients who achieved sCR, 76.9% (10/13) did not relapse or progress during follow-up.

- Ten high-risk newly diagnosed MM patients received treatment. The ORR was 100%: the best response being sCR (90%), and 10% had a CR.

Among 17 cases, the ORR was 88.2%, with 13 achieving sCR and 2 reaching VGPR, while 1 patient did not respond.

The ORR was 100%, with 72.2% of the patients achieving CR or sCR. Patients who relapse after prior murine BCMA CAR T-cell therapy may still benefit from CT103A.

Humanized anti-BCMA CAR T cells

BCMA-targeted fourth-generation CAR T cells secreting IL-7 and CCL19

ChiCTR1800017051 - Seven r/r MM patients with extramedullary disease and 13 without extramedullary disease received humanized anti-BCMA CAR T-cell therapy. The ORR in all the r/r MM patients was 80% (16/20). The CRS and ICANS grades were much higher in patients with extramedullary disease.

NCT03778346 -

Preliminary results showed that 1 of 2 patients achieved CR and the other patient achieved VGPR of an extramedullary recurrence.

CAR: chimeric antigen receptor; ID: identity; Ref: reference; BCMA: B-cell maturation antigen; sCR: stringent complete remission; VGPR: very good partial remission; PR: partial remission; MRD: minimal residual disease; MM: multiple myeloma; CR: complete remission; ORR: overall response rate; r/r: relapsed/refractory; CRS: cytokine release syndrome; ICANS: immune effector cell-associated neurotoxicity syndrome.

developed for AML. The prognosis and survival outcomes in AML patients after standard chemotherapy remain poor with estimated 5-year survival rates of 40-55% and 10-15% in patients <60 and >60 years old, respectively, making it imperative to develop new, targeted immunotherapies.67 Selective elimination of cancerous cells is of the utmost importance in AML patients, because many myeloid antigens are also expressed on healthy hematopoietic stem and progenitor cells, leading to the destruction of the bone marrow and other toxic effects if

non-selective agents are used. It is therefore crucial to find a suitable target for CAR T-cell treatment in AML. CD33 is known to be expressed highly in most AML patients thereby making it a potential target for the treatment of AML. In 2014 the first Chinese clinical study on autologous CD33-targeted CAR T-cell therapy in r/r AML patients was reported: a remarkable decrease in blasts in bone marrow was observed within 2 weeks after starting therapy and the adverse events were manageable.68 Building on this, numerous tumor antigens, including

Table 6. Chimeric antigen receptor T-cell therapy in multiple myeloma: overview of current strategies to enhance efficacy.

Strategy CAR T cells/object Trial ID Phase Key findings Ref. Monitor ALC prior to LD affects outcomes in MM patients treated with CAR T cells

- - A better deep response rate was observed in patients with a high pre-LD ALC than in patients with a low pre-LD ALC (76% vs. 41%; P=0.002). Patients with a low pre-LD ALC had significantly inferior OS (median 15.4 months vs. not reached) and PFS (median PFS 8.4 vs. 27.3 months) compared with those with a high pre-LD ALC.

55

Virus reactivation BCMA CAR T cells in MM patients co-infected with chronic HBV

-

- Among 8 patients with MM who had resolved HBV infection, 2 given prophylactic anti-HBV drugs did not exhibit HBV reactivation. Further, 5/6 patients who did not receive prophylactic antiviral drugs, did not exhibit HBV reactivation, while 1 showed recurrence of HBsAg without detection of HBV DNA or damage to liver function. The ORR was 100%, and PFS at 12 months was 88.89%.

56

Immune reconstitution Humoral immune reconstitution after anti-BCMA CAR T cells

Renal impairment Anti-mBCMA and/or anti-hsCD19 CAR T cells

ChiCTR-OIC17011272

ChiCTR-OIC17011272

- Anti-BCMA CAR T cells caused 7 months of aplasia of normal bone marrow plasma cells and a longer period of hypogammaglobulinemia, suggesting a profound and lasting humoral deficiency.

- Combined anti-mBCMA with anti-hsCD19 CAR T cells or single anti-mBCMA CAR T-cell therapy is effective and well-tolerated in r/r MM patients with renal impairment and can restore renal function at high response rate.

57

58

CAR: chimeric antigen receptor; ID: identity; Ref: reference; ALC: absolute lymphocyte count; LD: lymphodepletion; MM: multiple myeloma; OS: overall survival; PFS: progression-free survival; HBV: hepatitis B virus; ORR: overall response rate; BCMA: B-cell maturation antigen; mBCMA: murine BCMA; r/r: relapsed/refractory; hsCD19: humanized CD19-specific.

CD38, CLL1, and CD123, have been explored as potential target antigens for AML treatment. Recently, CD38-targeted CAR T-cell therapy was tested as a new option in AML patients who relapsed following allogeneic HSCT. Qingya et al. conducted a prospective study to evaluate the efficacy and safety of CD38-targeted CAR T cells in such patients and reported that 4 weeks of infusion of CD38 CAR T cells led to CR in four of six (66.7%) patients, with median overall survival and leukemia-free survival times of 7.9 and 6.4 months, respectively. Furthermore, adverse events were clinically manageable in all six patients.69

CLL1 is highly expressed on AML stem cells, monocytes and blast cells but not on normal hematopoietic stem cells, thereby making it an actionable target in AML. Zhang et al. described that autologous anti-CLL1 CAR Tcell therapy in four children with r/r AML was efficacious; three of the children achieved CR and minimal residual disease negativity. Moreover, adverse events were lowgrade and manageable in all the patients.70 In addition, a recent comparative study by Kunlin et al. evidenced similar efficacy/safety profiles of 4-1BB and CD28/CD27equipped CLL1-based CAR T cells in the treatment of children with r/r AML, with ORR of 67% and 75% in the two groups, respectively.71

Lymphoma

Lymphomas are systemic malignancies originating from lymphocytes. These heterogeneous lymphoid neoplasms can be classified into Hodgkin lymphomas and non-Hodgkin lymphomas (NHL). The incidence rates of Hodgkin lymphoma and NHL in China are ~0.46 and 4.29 cases per 100,000 persons, respectively.72 In this review, we mainly discuss the development of CAR T-cell therapy in NHL because of the higher incidence of this form of lymphoma. The first-line treatment for NHL is chemoimmunotherapy with or without radiation. However, ~20-30% of patients eventually develop resistance, and the outcome of such patients is not entirely satisfactory, thereby warranting new approaches. CD19 is the most explored target of CAR T-cell therapy in lymphoma, and research has focused on both murine and fully human binding domains. Several CD19-targeted clinical studies have documented ORR ranging from 50% to 100% and CR rates from 20% to 66.7%.7 Despite the significant efficacy of CD19 CAR T-cell therapy in NHL, 20-30% of cases relapse after this treatment because of antigen loss. Furthermore, given the heterogeneity of NHL, CD19 is not universally expressed on all lymphoma cells. The search for other targets is, therefore, very important. A robust pipeline of different targets for treating NHL, including B7-H3, Igβ, CD79b, CD30, BAFF,

Table 7. Chimeric antigen receptor T-cell in multiple myeloma: overview of current strategies to enhance safety.

Figure 4. Chimeric antigen receptor T-cell therapy. (A) Structure and process of chimeric antigen receptor (CAR) T-cell therapy. (B) Obstacles to CAR T-cell therapy in hematologic malignancies and strategies to overcome them. HBV: Hepatitis B virus.

CTLA4, CD20 and CD70, is currently being explored and both pre-clinical and clinical studies are underway. Considering the significant efficacy of CD20 monoclonal antibody in NHL, CD20 was selected as one target and it is in the early exploratory stage. In 2014, Yao Wang et al. reported prolonged tumor regression following the use of CD20 CAR T cells in patients with DLBCL, and three of five evaluable patients with bulky tumor burden attained 3 to 6 months of tumor regression.73

Multi-target CAR T-cell therapy is an optimal strategy to overcome the immune escape of tumor cells. At present, the majority of studies have used CD19 in combination with other targets to construct multi-target CAR T cells, which simultaneously express multiple targets on the surface of the T cells. Lymphoma patients with high-risk factors, such as extra-nodal involvement, high-risk cytogenetics and limited response after salvage treatment are more prone to disease progression and may possibly benefit more from multi-target CAR T-cell therapy. Chen et al. reported that ten of 14 patients with r/r aggressive B-cell lymphoma with extra-nodal involvement who received sequential anti-CD22/anti-CD19 CAR T cells achieved objective responses, and seven of 14 achieved CR.74 Jia et al. described that CD19/22 CAR T-cell cocktail therapy improved the long-term outcome of patients with r/r double-hit lymphoma.27

CAR T-cell therapy is a personalized immunotherapy and there are now a few potential therapeutic targets for the CAR T cells in lymphoma. As a result, the targets of the CAR T cells can be selected according to the patient's own characteristics. Cheng et al. reported a multi-CAR T-cell regimen for r/r B-cell lymphoma based on the patients’ specific tumor antigen profile. The choice of CAR T-cell targets was determined by immunostaining tumor biopsies for CD19, CD22, CD30, GD2, and PSMA. Three of four patients achieved CR, and all of them have been in remission for >1 year.75

The mechanisms underlying relapse after CD19-targeted therapy are multifactorial and still poorly elucidated. A possible way to improve the efficacy of CAR T-cell therapy is to combine it with other treatment options. Cuicui et al. found that intensive debulking chemotherapy improved both short-term and long-term efficacy of anti-CD19 CAR T-cell therapy in r/r DLBCL with high tumor bulk.76 Changju et al. reported that radiotherapy before CAR T-cell therapy in r/r DLBCL patients with high tumor burden produced a higher ORR (100%) and less severe CRS and neurotoxicity.77 In addition to traditional treatment, combined targeted therapy (BTK inhibitor/PD-1 blocker) and immunotherapy are also hot subjects for combination regimens. Another strategy is to optimize the structure of the CAR T cells themselves. Inhibitory signals that CAR T cells encounter in the tumor microenvironment are often reported to impair the efficacy of CAR T-cell therapy. Xiaoqian et al

evaluated the efficacy of a novel dominant-negative PD1-armored anti-CD19 CAR T cells in nine NHL patients and found an ORR of 77.8% (n=7/9) and a CR rate of 55.6% (n=5/9). In addition, the CAR T cells expanded after infusion and continued to be detectable at >12 months in patients with ongoing CR.33 Similarly, Wenbin Qian et al. illustrated the efficacy of novel CD19-specific CAR T cells that express a PD-1/CD28 chimeric switch-receptor (CD19PD-1/CD28-CAR) in r/r PD-L1-positive B-cell lymphoma and DLBCL patients who had relapsed after different CD19-directed CAR T-cell therapies.35 There are also studies comparing the effect of different co-stimulatory domains on CAR T-cell efficacy and the control of CAR Tcell expansion, and apoptosis through suicide switches.36 Although autologous HSCT is the standard-of-care treatment for r/r lymphoma, studies are now suggesting that the clinical outcomes after CAR T-cell therapy are superior to those produced by autologous HSCT. The next step forward could be to combine CAR T cells and autologous HSCT. Indeed, Wang et al. described that the combination of autologous HSCT and anti-CD19 CAR T-cell therapy was beneficial in r/r DLBCL patients (n=14) who had a median progression-free survival of 14.82 months and an overall survival rate of 64.29%.78

Despite the aforementioned valuable options, several obstacles, such as the quality and quantity of T cells in intensive treatments, have limited the availability of autologous CAR T cells and their clinical usage. Recent studies have indicated the feasibility of using allogeneic universal CAR T cells in r/r lymphoma. Guo et al. reported two successful cases of treatment using CRISPR/Cas9 genome-edited universal CAR T cells negative for T-cell receptor and human leukocyte antigen class I molecules in patients with r/r lymphoma.43

CAR T-cell therapy is associated with unique adverse events, so appropriate methodology must be established to predict the occurrence and severity of such events.

Jiasheng et al. conducted a retrospective study and showed that NHL patients with greater baseline disease burden were susceptible to more severe CRS, whereas patients with mild and moderate CRS (grade 0-2) had significantly lower metabolic tumor volume and total lesion glycolysis than those with severe CRS (grade 3/4).79

HBV reactivation is a well-recognized complication in lymphoma patients with concomitant viral infection. Wei et al. conducted a post-hoc analysis of two prospective clinical trials involving the use of CNCT19 CAR T cells (autologous second-generation anti-CD19 CAR T cells with 4-1BB as a co-stimulatory domain) in B-cell lymphoma patients and reported that anti-CD19 CAR T-cell therapy could be safely administered in B-cell lymphoma patients with concomitant HBV infection. However, antiviral prophylaxis was suggested for the patients treated with CNCT19 cells.80

Multiple myeloma

MM is characterized by uncontrolled proliferation of clonal plasma cells in bone marrow and accounts for 10% of blood cancers. It is associated with a mortality rate of 0.67 per 100,000 persons in China.81 BCMA-targeted CAR T-cell therapy has achieved great success in MM, and two products have been approved by the FDA for MM patients who have received ≥4 lines of therapy, namely idecabtagene vicleucel and ciltacabtagene autoleucel (also named LCARB38M). The efficacy of different BCMA CAR T-cell therapies varies and there are other differences between the products. Pivotal studies demonstrated ORR and CR rates of 73% and 33%, respectively, for idecabtagene vicleucel82 and 97% and 67%, respectively, for ciltacabtagene autoleucel.83 Head-to-head comparisons of randomized controlled clinical studies are therefore now warranted. The long-term outcome of patients treated with idecabtagene vicleucel and ciltacabtagene autoleucel also varies considerably.

In addition to differences in study design and patients’ characteristics, the inherent characteristics of different BCMA CAR T cells might affect their efficacy. Here we summarize the relevant studies on the modification and optimization of BCMA CAR T cells in China. The variable heavy chain domain of heavy-chain-only antibodies is the variable fragment of heavy-chain antibodies of camelidae and, like conventional antibodies, it is functional in antigen binding. It is a small, stable and single domain structure with high affinity and specificity comparable to those of single chain variable fragments (scFv). Lu et al. evaluated the efficacy of a single variable heavy chain domain of heavy-chain antibody-directed BCMA CAR T cells in r/r MM patients (n=34) and reported an ORR of 88.2% and stringent CR rate of 55.9%; the median progression-free survival was 12.1 months.84 Wan-Hong Zhao et al. conducted a phase I study of LCAR-B38M, which is a dual epitopebinding CAR T- cell therapy directed against two distinct BCMA epitopes and documented an ORR of 88% (39/57 patients) and CR rate of 68% in r/r MM patients.85 The long-term follow-up (median 19 months) results, presented at the 61st American Society of Hematology Annual Meeting in 2019, included a median progression-free survival of 20 months.86

Other strategies to optimize CAR T cells include humanization and arming. Duan et al . constructed BCMA-targeted fourth-generation CAR T cells expressing IL-7 and CCL19 for the purpose of enhancing the cells’ expansion, differentiation, migration and cytotoxicity and demonstrated their efficacy in r/r MM patients. The preliminary results showed that one of two patients achieved a CR, and the other patient had a very good partial response of an extramedullary recurrence.54

Since the efficacy of CAR T cells targeting BCMA has been validated, CAR T cells targeting other antigens have been

used in combinations with those targeting BCMA. One of the most common targets combinations is BCMA and CD19. B-lymphocyte antigen CD19, which is expressed by B cells prior to terminal differentiation into plasma cells, is associated with enhancement of myeloma tumorpropagating and drug-resistance properties. Zhiling Yan et al. conducted a phase II trial to evaluate the efficacy of a combination of humanized anti-CD19 (1×106 cells/kg) and anti-BCMA CAR T cells (1×106 cells/kg) in r/r MM patients (n=22) and reported an ORR of 95% including nine (43%) stringent CR, three (14%) CR, five (24%) very good partial responses, and three (14%) partial responses.7 Using an alternative strategy, Lingzhi Yan et al. tested sequential CD19 and BCMA-specific CAR T-cell treatments in r/r MM. The patients received one dose of a CD19 CAR T-cell infusion on day 0 and thereafter a split-dose of BCMA CAR Tcell infusions over 2 days. The ORR was 90% (5 partial responses and 4 stringent CR).88 Other targets combined with BCMA include CD3849 and CS1, and both have been studied in MM, although these novel dual-targeted CAR T cells are mostly in preclinical development.

Renal impairment is a common complication of MM, but immunomodulatory agents and other treatments have been shown to be effective in patients with varying degrees of renal impairment. Shao-long et al. reported the efficacy of anti-BCMA CAR T-cell therapy in r/r MM patients with impaired renal function, with a median progression-free survival of 181 days and overall survival of 238 days, and further suggested that CAR T-cell therapy could be beneficial to renal function in r/r MM.89

Reactivation of HBV infection while undergoing anticancer therapy is an unwanted event in patients with chronic or resolved HBV infection. However, Han et al. described that BCMA CAR T-cell therapy could be administered safely and no HBV reactivation was observed among the nine r/r MM patients with resolved HBV infection.56

Ongoing challenges with chimeric antigen receptor T-cell therapy and future directions

As discussed, CAR T cells have become a major source of cellular immunotherapy for hematologic malignancies. In China, a number of CAR T-cell products are poised to launch a new therapeutic era. The main CAR T-cell trials are in the field of B-cell malignancies, such as lymphoma, leukemia and myeloma. Admittedly, CAR T-cell therapy is a complex process, and challenges occur throughout all parts of the exploratory work, including patient’s recruitment and enrollment, the manufacturing process, delivery, the gap period between leukapheresis and infusion, in addition to the enormous cost, and so on. Despite higher ORR, relapse and resistance have been barriers limiting

the clinical feasibility of this therapy. It is to be hoped that innovative engineering can circumvent these barriers.90,91 Another challenge that remains to be addressed adequately is the management of CRS and immune effector cell-associated neurotoxicity syndrome (ICANS), which are the most common toxicities related to CAR T-cell therapy.92 Chinese researchers have actively explored a variety of cytokine inhibitors, based on drug accessibility, for the management of CRS. These include etanercept (a tumor necrosis factor- α inhibitor),93 tocilizumab (an anti-IL-6 monoclonal antibody), and ruxolitinib (a JAK 1 and JAK2 inhibitor).94 Vascular endothelial activation has been shown to contribute to the development of CRS and ICANS after CAR T-cell therapy. Therefore, blockade of tumor necrosis factor-α and interleukin-1β is also being investigated as a potential therapeutic target for the treatment of CAR T-cell therapy-associated CRS and ICANS.95 Furthermore, Lu et al. have described a role for the poreforming protein gasdermin E (GSDME) in release of proinflammatory cytokines during tumor cell pyroptosis leading to CRS and hence blockade of this pathway could be another potential strategy for the management of CRS.96 The incidence of ICANS following CAR T-cell therapy appears to be significantly lower in the Chinese population than in the US population. For myeloma, the incidence of ICANS in BCMA CAR T-cell-treated Chinese r/r MM patients was 2.1%,97 compared with 17% reported in the US population.83 For lymphoma, neurotoxicity was reported in 87% of patients in the ZUMA-1 study conducted in the USA and Israel and the incidence of grade ≥3 adverse events was 31%.98 In contrast, in the Chinese bridging study of axicabtagene ciloleucel, neurological toxicity occurred in 42% of patients and grade ≥3 adverse events were reported in 8% of patients. Understanding the mechanisms underlying these differences in ICANS between Chinese and American populations after CAR T-cell therapy would help in the development of better treatments and facilitate the prevention of these adverse events.

In China, the competition towards the commercial development of CAR T-cell therapy has intensified. However, consensus and guidelines regarding the targets and discrepancies between the efficacy of CAR T-cell products are challenging and urgently needed. Furthermore, the variable distribution of cell doses in the clinical trials conducted so far might reflect an insufficient exploration of cellular potency and pharmacodynamic characteristics of

References

1. Yan W, Liu Z, Liu J, Xia Y, Hu K, Yu J. Application of chimeric antigen receptor T cells in the treatment of hematological malignancies. Biomed Res Int. 2020;2020:4241864.

2. Boyiadzis MM, Dhodapkar MV, Brentjens RJ, et al. Chimeric antigen receptor (CAR) T therapies for the treatment of

CAR T cells. Therefore, uni fied systematic management and operational guidance need to be implemented across hospitals/clinical study centers in order to promote the CAR T-cell industry.

Conclusions

CAR T-cell therapy is developing rapidly due to continuous scientific breakthroughs from CAR T cells targeting CD19 and BCMA, providing another pathway to improve the prognosis and quality of life of patients with hematologic malignancies. In contrast, CAR T-cell therapy has less impact on solid tumors. Admittedly, non-negligible issues, such as high cost, the time-consuming production process, inherent risks from manufacturing failures, immune-related adverse events, the problems of r/r disease, and inability to infiltrate solid tumor tissues, remain to be resolved, and are currently posing limits to the treatment of certain hematologic malignancies. It is now essential to develop products with acceptable cost and safety in order to extend the benefits of CAR T cells to a larger population.

Disclosures

No conflicts of interest to disclose.

Contributions

X-JH designed the review and wrote the manuscript. WS wrote the manuscript. AB-L and HH discussed and revised the manuscript. All authors gave final approval of the manuscript.

Acknowledgments

The authors acknowledge medical writing support provided by Dr Amit Bhat (PhD) from Indegene (Bangalore, India).

Funding

This work was supported by the National Key Research and Development Program of China (N. 2022YFA1103300), Major Program of the National Natural Science Foundation of China (N. 82293630), and Key Program of the National Natural Science Foundation of China (N. 81930004).

Data-sharing statement

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

hematologic malignancies: clinical perspective and significance. J Immunother Cancer. 2018;6(1):137.

3 FDA approves tisagenlecleucel for B-cell ALL and tocilizumab for cytokine release syndrome. https://www.fda.gov/drugs/resourcesinformation-approved-drugs/fda-approves-tisagenlecleucel-b-cel

l-all-and-tocilizumab-cytokine-release-syndrome. Accessed September 7, 2017.

4. FDA approves axicabtagene ciloleucel for large B-cell lymphoma. https://www.fda.gov/drugs/resources-informationapproved-drugs/fda-approves-axicabtagene-ciloleucel-large-bcell-lymphoma. Accessed October 5, 2017.

5. An open-label clinical study for evaluating the safety and efficacy of FKC876 in the treatment of relapsed and refractory aggressive non-Hodgkin’s lymphoma (NHL). https://adis.springer.com/trials/700315011. Accessed November 3, 2022.

6. Cao X, Li W, Yu Y, et al. China enters CAR-T cell therapy era. Innovation (Camb). 2022;3(1):100197.

7. Ying Z, Yang H, Guo Y, et al. Relmacabtagene autoleucel (relmacel) CD19 CAR-T therapy for adults with heavily pretreated relapsed/refractory large B-cell lymphoma in China. Cancer Med. 2021;10(3):999-1011.

8. Mostafa Kamel Y. CAR-T therapy, the end of a chapter or the beginning of a new one? Cancers (Basel). 2021;13(4):853.

9. Cancer Stat Facts: Leukemia — Acute Lymphocytic Leukemia (ALL). SEER. https://seer.cancer.gov/statfacts/html/alyl.html. Accessed September 9, 2022.

10. Pan J, Niu Q, Deng B, et al. CD22 CAR T-cell therapy in refractory or relapsed B acute lymphoblastic leukemia. Leukemia. 2019;33(12):2854-2866.

11. Dai H, Wu Z, Jia H, et al. Bispecific CAR-T cells targeting both CD19 and CD22 for therapy of adults with relapsed or refractory B cell acute lymphoblastic leukemia. J Hematol Oncol. 2020;13(1):30.

12. Pan J, Zuo S, Deng B, et al. Sequential CD19-22 CAR T therapy induces sustained remission in children with r/r B-ALL. Blood. 2020;135(5):387-391.

13. Liu S, Deng B, Yin Z, et al. Combination of CD19 and CD22 CAR-T cell therapy in relapsed B-cell acute lymphoblastic leukemia after allogeneic transplantation. Am J Hematol. 2021;96(6):671-679.

14. Li S, Zhang J, Wang M, et al. Treatment of acute lymphoblastic leukaemia with the second generation of CD19 CAR-T containing either CD28 or 4-1BB. Br J Haematol. 2018;181(3):360-371.

15. Zhao X, Yang J, Zhang X, et al. Efficacy and safety of CD28- or 41BB-based CD19 CAR-T cells in B cell acute lymphoblastic leukemia. Mol Ther Oncolytics. 2020;18:272-281.

16. Cao J, Wang G, Cheng H, et al. Potent anti-leukemia activities of humanized CD19-targeted chimeric antigen receptor T (CAR-T) cells in patients with relapsed/refractory acute lymphoblastic leukemia. Am J Hematol. 2018;93(7):851-858.

17. Zhao Y, Liu Z, Wang X, et al. Treatment with humanized selective CD19CAR-T cells shows efficacy in highly treated B-ALL patients who have relapsed after receiving murine-based CD19 CAR-T therapies. Clin Cancer Res. 2019;25(18):5595-5607.

18. Gu R, Liu F, Zou D, et al. Efficacy and safety of CD19 CAR T constructed with a new anti-CD19 chimeric antigen receptor in relapsed or refractory acute lymphoblastic leukemia. J Hematol Oncol. 2020;13(1):122.

19. Qu C, Song Y, Yin J, et al. Decitabine may improve CAR-T efficacy in refractory/relapsed acute leukemia patients carrying TP53 alterations. Bone Marrow Transplant. 2021;56(7):1710-1713.

20. Hu Y, Zhou Y, Zhang M, et al. CRISPR/Cas9-engineered universal CD19/CD22 dual-targeted CAR-T cell therapy for relapsed/refractory B-cell acute lymphoblastic leukemia. Clin Cancer Res. 2021;27(10):2764-2772.

21. Shen D, Song H, Xu X, et al. Chimeric antigen receptor T cell therapy can be administered safely under the real-time monitoring of Th1/Th2 cytokine pattern using the cytometric

bead array technology for relapsed and refractory acute lymphoblastic leukemia in children. Pediatr Hematol Oncol. 2020;37(4):288-299.

22. Lai P, Chen X, Qin L, et al. The efficacy and safety of CAR-T cell therapy in patients with refractory ALL and concomitant HBV infection. Leukemia. 2020;34(10):2790-2793.

23. Liu S, Deng B, Yin Z, et al. Corticosteroids do not influence the efficacy and kinetics of CAR-T cells for B-cell acute lymphoblastic leukemia. Blood Cancer J. 2020;10(2):15.

24. Wang D, Zeng C, Xu B, et al. Anti-CD30 chimeric antigen receptor T cell therapy for relapsed/refractory CD30(+) lymphoma patients. Blood Cancer J. 2020;10(1):8.

25. Zhu H, Deng H, Mu J, Lyu C, Jiang Y, Deng Q. Anti-CD22 CAR-T cell therapy as a salvage treatment in B cell malignancies refractory or relapsed after anti-CD19 CAR-T therapy. Onco Targets Ther. 2021;14:4023-4037.

26. Sang W, Shi M, Yang J, et al. Phase II trial of co-administration of CD19- and CD20-targeted chimeric antigen receptor T cells for relapsed and refractory diffuse large B cell lymphoma. Cancer Med. 2020;9(16):5827-5838.

27. Wei J, Mao Z, Wang N, et al. Long-term outcomes of relapsed/refractory double-hit lymphoma (r/r DHL) treated with CD19/22 CAR T-cell cocktail therapy. Clin Transl Med. 2020;10(5):e176.

28. Tong C, Zhang Y, Liu Y, et al. Optimized tandem CD19/CD20 CARengineered T cells in refractory/relapsed B-cell lymphoma. Blood. 2020;136(14):1632-1644.

29. Zhou X, Ge T, Li T, et al. CAR19/22 T cell therapy in adult refractory Burkitt's lymphoma. Cancer Immunol Immunother. 2021;70(8):2379-2384.

30. Liu Y, Deng B, Hu B, et al. Sequential different B-cell antigentargeted CAR T-cell therapy for pediatric refractory/relapsed Burkitt lymphoma. Blood Adv. 2022;6(3):717-730.

31. Wei G, Zhang Y, Zhao H, et al. CD19/CD22 dual-targeted CAR Tcell therapy for relapsed/refractory aggressive B-cell lymphoma: a safety and efficacy study. Cancer Immunol Res. 2021;9(9):1061-1070.

32. Zhang Y, Li J, Lou X, et al. A prospective investigation of bispecific CD19/22 CAR T cell therapy in patients with relapsed or refractory B cell non-Hodgkin lymphoma. Front Oncol. 2021;11:664421.

33. Liu X, Zhang Y, Li K, et al. A novel dominant-negative PD-1 armored anti-CD19 CAR T cell is safe and effective against refractory/relapsed B cell lymphoma. Transl Oncol. 2021;14(7):101085.

34. Liang Y, Liu H, Lu Z, et al. CD19 CAR-T expressing PD-1/CD28 chimeric switch receptor as a salvage therapy for DLBCL patients treated with different CD19-directed CAR T-cell therapies. J Hematol Oncol. 2021;14(1):26.

35. Liu H, Lei W, Zhang C, et al. CD19-specific CAR T cells that express a PD-1/CD28 chimeric switch-receptor are effective in patients with PD-L1-positive B-cell lymphoma. Clin Cancer Res. 2021;27(2):473-484.

36. Ying Z, He T, Wang X, et al. Parallel comparison of 4-1BB or CD28 co-stimulated CD19-targeted CAR-T cells for B cell nonHodgkin's lymphoma. Mol Ther Oncolytics. 2019;15:60-68.

37. Zhou X, Tu S, Wang C, et al. Phase I trial of fourth-generation anti-CD19 chimeric antigen receptor T cells against relapsed or refractory B cell non-Hodgkin lymphomas. Front Immunol. 2020;11:564099.

38. Li S, Xue L, Wang M, et al. Decitabine enhances cytotoxic effect of T cells with an anti-CD19 chimeric antigen receptor in treatment of lymphoma. Onco Targets Ther. 2019;12:5627-5638.

39. Cao Y, Lu W, Sun R, et al. Anti-CD19 chimeric antigen receptor T

cells in combination with nivolumab are safe and effective against relapsed/refractory B-cell non-Hodgkin lymphoma. Front Oncol. 2019;9:767.

40. Liu M, Deng H, Mu J, et al. Ibrutinib improves the efficacy of anti-CD19-CAR T-cell therapy in patients with refractory nonHodgkin lymphoma. Cancer Sci. 2021;112(7):2642-2651.

41. Cao Y, Xiao Y, Wang N, et al. CD19/CD22 chimeric antigen receptor T cell cocktail therapy following autologous transplantation in patients with relapsed/refractory aggressive B cell lymphomas. Transplant Cell Ther. 2021;27(11):910.e1-910.e11.

42. Wu J, Meng F, Cao Y, et al. Sequential CD19/22 CAR T-cell immunotherapy following autologous stem cell transplantation for central nervous system lymphoma. Blood Cancer J. 2021;11(7):131.

43. Guo Y, Tong C, Su L, et al. CRISPR/Cas9 genome-edited universal CAR T cells in patients with relapsed/refractory lymphoma. Blood Adv. 2022;6(8):2695-2699.

44. Hong R, Tan Su Yin E, Wang L, et al. Tumor burden measured by 18F-FDG PET/CT in predicting efficacy and adverse effects of chimeric antigen receptor T-cell therapy in non-Hodgkin lymphoma. Front Oncol. 2021;11:713577.

45. Yang C, Xie M, Zhang K, et al. Risk of HBV reactivation post CD19-CAR-T cell therapy in DLBCL patients with concomitant chronic HBV infection. Leukemia. 2020;34(11):3055-3059.

46. Cui R, Lyu C, Li Q, et al. Humanized anti-CD19 chimeric antigen receptor-T cell therapy is safe and effective in lymphoma and leukemia patients with chronic and resolved hepatitis B virus infection. Hematol Oncol. 2021;39(1):75-86.

47. Zhu F, Wei G, Liu Y, et al. Incidence and risk factors associated with infection after chimeric antigen receptor T cell therapy for relapsed/refractory B-cell malignancies. Cell Transplant. 2021;30:9636897211025503.

48. Li C, Mei H, Hu Y, et al. A bispecific CAR-T cell therapy targeting BCMA and CD38 for relapsed/refractory multiple myeloma: updated results from a phase 1 dose-climbing trial. Blood. 2019;134(Suppl_1):930.

49. Tang Y, Yin H, Zhao X, et al. High efficacy and safety of CD38 and BCMA bispecific CAR-T in relapsed or refractory multiple myeloma. J Exp Clin Cancer Res. 2022;41(1):2.

50. Shi X, Yan L, Shang J, et al. Anti-CD19 and anti-BCMA CAR T cell therapy followed by lenalidomide maintenance after autologous stem-cell transplantation for high-risk newly diagnosed multiple myeloma. Am J Hematol. 2022;97(5):537-547.

51. Xu J, Chen LJ, Yang SS, et al. Exploratory trial of a biepitopic CAR T-targeting B cell maturation antigen in relapsed/refractory multiple myeloma. Proc Natl Acad Sci U S A. 2019;116(19):9543-9551.

52. Wang D, Wang J, Hu G, et al. A phase 1 study of a novel fully human BCMA-targeting CAR (CT103A) in patients with relapsed/refractory multiple myeloma. Blood. 2021;137(21):2890-2901.

53. Deng H, Liu M, Yuan T, et al. Efficacy of humanized anti-BCMA CAR T cell therapy in relapsed/refractory multiple myeloma patients with and without extramedullary disease. Front Immunol. 2021;12:720571.

54. Duan D, Wang K, Wei C, et al. The BCMA-targeted fourthgeneration CAR-T cells secreting IL-7 and CCL19 for therapy of refractory/recurrent multiple myeloma. Front Immunol. 2021;12:609421.

55. Liu Y, Chen W, Yu M, et al. Absolute lymphocyte count prior to lymphodepletion impacts outcomes in multiple myeloma patients treated with chimeric antigen receptor T cells. Transplant Cell Ther. 2022;28(2):118.e1-118.e5.

56. Han L, Zhou J, Zhou K, et al. Safety and efficacy of CAR-T cell targeting BCMA in patients with multiple myeloma coinfected with chronic hepatitis B virus. J Immunother Cancer. 2020;8(2):e000927.

57. Wang Y, Li C, Xia J, et al. Humoral immune reconstitution after anti-BCMA CAR T-cell therapy in relapsed/refractory multiple myeloma. Blood Adv. 2021;5(23):5290-5299.

58. Li H, Yin L, Wang Y, et al. Safety and efficacy of chimeric antigen receptor T-cell therapy in relapsed/refractory multiple myeloma with renal impairment. Bone Marrow Transplant. 2020;55(11):2215-2218.

59. Li X, Chen W. Mechanisms of failure of chimeric antigen receptor T-cell therapy. Curr Opin Hematol. 2019;26(6):427-433.

60. Zhang C, Wang XQ, Zhang RL, et al. Donor-derived CD19 CAR-T cell therapy of relapse of CD19-positive B-ALL post allotransplant. Leukemia. 2021;35(6):1563-1570.

61. Jin X, Cao Y, Wang L, et al. HLA-matched and HLAhaploidentical allogeneic CD19-directed chimeric antigen receptor T-cell infusions are feasible in relapsed or refractory B-cell acute lymphoblastic leukemia before hematopoietic stem cell transplantation. Leukemia. 2020;34(3):909-913.

62. Pan J, Tan Y, Wang G, et al. Donor-derived CD7 chimeric antigen receptor T cells for T-cell acute lymphoblastic leukemia: firstin-human, phase I trial. J Clin Oncol. 2021;39(30):3340-3351.

63. Lu P, Liu Y, Yang J, et al. Naturally selected CD7 CAR-T therapy without genetic manipulations for T-ALL/LBL: first-in-human phase 1 clinical trial. Blood. 2022;140(4):321-334.

64 Tang XY, Sun Y, Zhang A, et al. Third-generation CD28/4-1BB chimeric antigen receptor T cells for chemotherapy relapsed or refractory acute lymphoblastic leukaemia: a non-randomised, open-label phase I trial protocol. BMJ Open. 2016;6(12):e013904.

65. Wang Y, Yang Y, Hong R, et al. A retrospective comparison of CD19 single and CD19/CD22 bispecific targeted chimeric antigen receptor T cell therapy in patients with relapsed/refractory acute lymphoblastic leukemia. Blood Cancer J. 2020;10(10):105.

66. Liu ZF, Chen LY, Wang J, et al. Successful treatment of acute B lymphoblastic leukemia relapse in the skin and testicle by antiCD19 CAR-T with IL-6 knocking down: a case report. Biomark Res. 2020;8:12.

67. Kantarjian H, Kadia T, DiNardo C, et al. Acute myeloid leukemia: current progress and future directions. Blood Cancer J. 2021;11(2):41.

68. Wang QS, Wang Y, Lv HY, et al. Treatment of CD33-directed chimeric antigen receptor-modified T cells in one patient with relapsed and refractory acute myeloid leukemia. Mol Ther. 2015;23(1):184-191.

69. Cui Q, Qian C, Xu N, et al. CD38-directed CAR-T cell therapy: a novel immunotherapy strategy for relapsed acute myeloid leukemia after allogeneic hematopoietic stem cell transplantation. J Hematol Oncol. 2021;14(1):82.

70. Zhang H, Wang P, Li Z, He Y, Gan W, Jiang H. DLBCL Clin Cancer Res. 2021;27(13):3549-3555.

71. Pei K, Xu H, Wang PF, et al. A comparison study of anti-CLL1 CART cells equipped with different co-stimulatory domains in the treatment of children with refractory/relapsed acute myeloid leukemia. Blood. 2021;138(Suppl 1):824.

72. Liu W, Liu J, Song Y, et al. Burden of lymphoma in China, 20062016: an analysis of the Global Burden of Disease Study 2016. J Hematol Oncol. 2019;12(1):115.

73. Wang Y, Zhang WY, Han QW, et al. Effective response and delayed toxicities of refractory advanced diffuse large B-cell lymphoma treated by CD20-directed chimeric antigen receptormodified T cells. Clin Immunol. 2014;155(2):160-175.

74. Zeng C, Cheng J, Li T, et al. Efficacy and toxicity for CD22/CD19 chimeric antigen receptor T-cell therapy in patients with relapsed/refractory aggressive B-cell lymphoma involving the gastrointestinal tract. Cytotherapy. 2020;22(3):166-171.

75. Jiao C, Zvonkov E, Lai X, et al. 4SCAR2.0: a multi-CAR-T therapy regimen for the treatment of relapsed/refractory B cell lymphomas. Blood Cancer J. 2021;11(3):59.

76. Lyu C, Cui R, Wang J, et al. Intensive debulking chemotherapy improves the short-term and long-term efficacy of anti-CD19CAR-T in refractory/relapsed DLBCL with high tumor bulk. Front Oncol. 2021;11:706087.

77. Qu C, Ping N, Kang L, et al. Radiation priming chimeric antigen receptor T-cell therapy in relapsed/refractory diffuse large B-cell lymphoma with high tumor burden. J Immunother. 2020;43(1):32-37.

78. Wang T, Gao L, Wang Y, et al. Hematopoietic stem cell transplantation and chimeric antigen receptor T cell for relapsed or refractory diffuse large B-cell lymphoma. Immunotherapy. 2020;12(13):997-1006.

79. Wang J, Hu Y, Yang S, et al. Role of fluorodeoxyglucose positron emission tomography/computed tomography in predicting the adverse effects of chimeric antigen receptor T cell therapy in patients with non-Hodgkin lymphoma. Biol Blood Marrow Transplant. 2019;25(6):1092-1098.

80. Liu W, Huang W, Wang M, et al. Risk of hepatitis B reactivation is controllable in patients with B-cell lymphoma receiving anti-CD19 CAR T cell therapy. Br J Haematol. 2020;191(1):126-129.

81. Liu J, Liu W, Mi L, et al. Incidence and mortality of multiple myeloma in China, 2006-2016: an analysis of the Global Burden of Disease Study 2016. J Hematol Oncol. 2019;12(1):136.

82. Munshi NC, Anderson LD Jr, Shah N, et al. Idecabtagene vicleucel in relapsed and refractory multiple myeloma. N Engl J Med. 2021;384(8):705-716.

83. Berdeja JG, Madduri D, Usmani SZ, et al. Ciltacabtagene autoleucel, a B-cell maturation antigen-directed chimeric antigen receptor T-cell therapy in patients with relapsed or refractory multiple myeloma (CARTITUDE-1): a phase 1b/2 open-label study. Lancet. 2021;398(10297):314-324.

84. Han L, Zhang JS, Zhou J, et al. Single VHH-directed BCMA CAR-T cells cause remission of relapsed/refractory multiple myeloma. Leukemia. 2021;35(10):3002-3006.

85. Zhao WH, Liu J, Wang BY, et al. A phase 1, open-label study of LCAR-B38M, a chimeric antigen receptor T cell therapy directed against B cell maturation antigen, in patients with relapsed or refractory multiple myeloma. J Hematol Oncol. 2018;11(1):141.

86. Wang BY, Zhao WH, Liu J, et al. Long-term follow-up of a phase 1, first-in-human open-label study of LCAR-B38M, a structurally differentiated chimeric antigen receptor T (CAR-T) cell therapy targeting B-cell maturation antigen (BCMA), in patients (pts) with relapsed/refractory multiple myeloma (RRMM). Blood. 2019;134(Suppl_1):579.

87. Yan Z, Cao J, Cheng H, et al. A combination of humanised antiCD19 and anti-BCMA CAR T cells in patients with relapsed or refractory multiple myeloma: a single-arm, phase 2 trial. Lancet Haematol. 2019;6(10):e521-e529.

88. Yan L, Qu S, Shang J, et al. Sequential CD19 and BCMA-specific CAR T-cell treatment elicits sustained remission of relapsed and/or refractory myeloma. Cancer Med. 2021;10(2):563-574.

89. He SL, Cheng YH, Wang D, et al. Anti-BCMA CAR-T cell therapy in relapsed or refractory multiple myeloma patients with impaired renal function. Curr Med Sci. 2021;41(3):474-481.

90. Zhang C, He J, Liu L, et al. CD19-directed fast CART therapy for relapsed/refractory acute lymphoblastic leukemia: from bench to bedside. Blood. 2019;134(Suppl_1):1340.

91. Su CT, Ye JC. Emerging therapies for relapsed/refractory multiple myeloma: CAR-T and beyond. J Hematol Oncol. 2021;14(1):115.

92. Xiao X, Huang S, Chen S, et al. Mechanisms of cytokine release syndrome and neurotoxicity of CAR T-cell therapy and associated prevention and management strategies. J Exp Clin Cancer Res. 2021;40(1):367.

93. Zhang L, Wang S, Xu J, et al. Etanercept as a new therapeutic option for cytokine release syndrome following chimeric antigen receptor T cell therapy. Exp Hematol Oncol. 2021;10(1):16.

94. Pan J, Deng B, Ling Z, et al. Ruxolitinib mitigates steroidrefractory CRS during CAR T therapy. J Cell Mol Med. 2021;25(2):1089-1099.

95. Chen Y, Li R, Shang S, et al. Therapeutic potential of TNFalpha and IL1beta blockade for CRS/ICANS in CAR-T therapy via ameliorating endothelial activation. Front Immunol. 2021;12:623610.

96. Lu L, Zhang Y, Tan X, et al. Emerging mechanisms of pyroptosis and its therapeutic strategy in cancer. Cell Death Discov. 2022;8(1):338.

97. Martin T, Usmani SZ, Berdeja JG, et al. Ciltacabtagene autoleucel, an anti-B-cell maturation antigen chimeric antigen receptor Tcell therapy, for relapsed/refractory multiple myeloma: CARTITUDE-1 2-year follow-up. J Clin Oncol. 2023;41(6):1265-1274.

98. Locke FL, Ghobadi A, Jacobson CA, et al. Long-term safety and activity of axicabtagene ciloleucel in refractory large B-cell lymphoma (ZUMA-1): a single-arm, multicentre, phase 1-2 trial. Lancet Oncol. 2019;20(1):31-42.

RNA helicase DHX15 exemplifies a unique dependency in acute leukemia

Hao Guo,1,2,3* Jin Xu,1,2* Peiqi Xing,4* Qilong Li,1,2 Donghai Wang,1,2 Chao Tang,5 Bruno Palhais,6 Juliette Roels,6 Jiaxu Liu,1,2 Sa Pan,1,2 Jinyan Huang,7 Zhaoqi Liu,4 Ping Zhu,5 Tom Taghon,8 Guoliang Qing,2 Pieter Van Vlierberghe6 and Hudan Liu1,2

1Department of Hematology, Zhongnan Hospital of Wuhan University, Wuhan University, Wuhan, China; 2Frontier Science Center for Immunology and Metabolism, Medical Research Institute, Wuhan University, Wuhan, China; 3Department of Hematology, The Affiliated Cancer Hospital of Zhengzhou University, Henan Cancer Hospital, Zhengzhou, China; 4CAS Key Laboratory of Genomic and Precision Medicine, Beijing Institute of Genomics, Chinese Academy of Sciences and China National Center for Bioinformation, Beijing, China; 5State Key Laboratory of Experimental Hematology, National Clinical Research Center for Blood Diseases, Institute of Hematology and Blood Diseases Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Tianjin, China; 6Department of Biomolecular Medicine, Ghent University, Ghent, Belgium; 7Bio-Med Big Data Center, the First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China and 8Department of Diagnostic Sciences, Ghent University, Ghent, Belgium

*HG, JX and PX contributed equally as first authors.

Abstract

Correspondence: H. Liu

hudanliu@whu.edu.cn

P. Van Vlierberghe pieter.vanvlierberghe@ugent.be

Received: September 7, 2022.

Accepted: February 22, 2023.

Early view: March 2, 2023.

https://doi.org/10.3324/haematol.2022.282066

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

RNA-binding proteins (RBP) have emerged as essential regulators that control gene expression and modulate multiple cancer traits. T-cell acute lymphoblastic leukemia (T-ALL) is an aggressive hematologic malignancy derived from transformation of T-cell progenitors that normally undergo discrete steps of differentiation in the thymus. The implications of essential RBP during T-cell neoplastic transformation remain largely unclear. Systematic evaluation of RBP identifies RNA helicase DHX15, which facilitates the disassembly of the spliceosome and release of lariat introns, as a T-ALL dependency factor. Functional analysis using multiple murine T-ALL models demonstrates the essential importance of DHX15 in tumor cell survival and leukemogenesis. Moreover, single-cell transcriptomics reveals that DHX15 depletion in T-cell progenitors hinders burst proliferation during the transition from doublenegative to double-positive cells (CD4 - CD8 - to CD4 + CD8 + ). Mechanistically, abrogation of DHX15 perturbs RNA splicing and leads to diminished levels of SLC7A6 and SLC38A5 transcripts due to intron retention, thereby suppressing glutamine import and mTORC1 activity. We further propose a DHX15 signature modulator drug ciclopirox and demonstrate that it has prominent anti-T-ALL efficacy. Collectively, our data highlight the functional contribution of DHX15 to leukemogenesis through regulation of established oncogenic pathways. These findings also suggest a promising therapeutic approach, i.e., splicing perturbation by targeting spliceosome disassembly, may achieve considerable anti-tumor efficacy.

Introduction

Normal T-cell development is a strictly regulated process in which hematopoietic progenitor cells migrate from the bone marrow to the thymus and differentiate from early Tcell progenitors into mature T cells.1 During this maturation process, accumulation of multiple oncogenic lesions can drive immature thymocytes into uncontrolled clonal expansion and cause T-cell acute lymphoblastic leukemia (TALL),2 which accounts for 10-15% of pediatric and 20-25% of adult cases of ALL.3 Although the use of intensified cyto-

toxic drugs significantly improves the clinical prognosis in pediatric patients, the outcomes of adult patients remain poor.4 More than 50% of adult patients with T-ALL do not achieve a complete remission or relapse after consolidated chemotherapy, highlighting the need for improved therapeutic interventions.5

A large body of data indicates that abnormal RNA-binding protein (RBP) expression/activity is a driving force of leukemia progression and an attribute of aggressive forms of disease.6 Frequent somatic mutations in RBP that serve as splicing factors have been found in patients with mye-

loid leukemia, resulting in pathologically altered splicing which promotes the initiation and/or maintenance of leukemia.7 Targeting strategies against crucial splicing factors represents an excellent therapeutic opportunity to improve outcomes of leukemia treatments.8 However, such mutations have not yet been found in T-ALL, and the RBP dependency in T-ALL remains to be understood.

DEAH-box helicase 15 ( DHX15 ) encodes an outstanding member of the DEAD/H-box RNA helicase family, characterized by the conserved DEAH (Asp-Glu-Ala-His) motif in the helicase domain.9 According to early studies on its yeast ortholog Prp43, DHX15 has cellular functions in the last step of splicing by facilitating the release of lariat introns and disassembly of the spliceosome.10-12 Cryo-electron microscopy structure analysis has shown that DHX15 is a key subunit in the intron lariat spliceosome complex in both yeast and humans, providing mechanistic insight into how DHX15 engages in spliceosome disassembly.13,14 Accumulating evidence suggests a prominent role for DHX15 in human cancers,15,16 although whether and how aberrant DHX15 expression leads to deregulated RNA splicing and contributes to tumorigenesis remains largely unclear. We here identify DHX15, transcriptionally activated by oncogenic MYB, as a dependency factor in the pathogenesis of T-ALL. Intriguingly, specific ablation of Dhx15 in the T-cell lineage resulted in a blockade of burst proliferation during the transition from the double-negative (DN) to double-positive (DP) stage, from which T-ALL arises. DHX15 is therefore a crucial RBP during T-cell development, which is hijacked by transformed T cells to confer neoplastic phenotypes. Mechanistically, aberrant expression of DHX15 leads to cancer-specific misspliced events associated with mTORC1 activation. Our data thus highlight the importance of the MYB-DHX15mTORC1 axis in T-cell leukemogenesis, and the possibility of targeting DHX15 for therapeutic purposes.

Methods

Cell cultures

Human T-ALL cell lines KOPTK1, CUTLL1 and MOLT-3 were kindly provided by Dr Warren Pear (University of Pennsylvania). 293T, CCRF-CEM and Jurkat cell lines were purchased from American Type Culture Collection (ATCC). Human primary specimens were obtained with informed consent from Zhongnan Hospital of Wuhan University. Cell culture details are provided in the Online Supplementary Material .

Mice

Dhx15 fl/fl mice were obtained from Biocytogen. Lck -Cre mice (JAX: 003802) and Mx1 -Cre mice (JAX: 003556)

were kindly provided by Dr Jinyong Wang and Dr Haojian Zhang, respectively. NOD.Cg-Prkdcscid Il2rgtm1Vst/Vst (NPG) mice were from Vitalstar. All animal experiments were performed under animal ethical regulations and the study protocol was approved by the Institutional Animal Care and Use Committee of Wuhan University.

NOTCH1-induced T-cell acute lymphoblastic leukemia mouse model

The NOTCH1-induced T-ALL model was established as previously decribed.17 Briefly, bone marrow cells were isolated from 8-week-old donor mice. Lineage negative (Lin–) cells were enriched using a Lineage Cell Depletion Kit (Miltenyi Biotec), and pre-stimulated for 24 h in Dulbecco modified Eagle medium (Gibco) containing 20% fetal bovine serum (Gibco), 1% penicillin/streptomycin (Hyclone), 20 ng/mL murine FLT3-L, 20 ng/mL murine thrombopoietin and 100 ng/mL murine stem cell factor (PeproTech). Cells were then transduced with MigR1-ICN1 retroviruses and centrifuged in the presence of 6 g/mL polybrene. A second round of transduction was carried out on the next day. One million cells were then transplanted into sub-lethally (5.5 Gy) irradiated 8-week-old C57BL/6 female mice by tail vein injection.

High-throughput sequencing

For single-cell RNA sequencing, single cells were captured in droplet emulsions using a GemCode Single-Cell Instrument (10xGenomics). Libraries were constructed using Chromium Single-Cell 3 Library and Gel Bead Kit V3.1 (10xGenomics), and sequenced using Illumina NovaSeq 6000. We demultiplexed the sequencing reads into single cells according to unique barcode sequences and aligned them to the mm10 reference using Cell Ranger (version 6.0.2). We merged the single-cell expression data from Dhx15 fl/fl and Dhx15 knockout (KO) thymocytes and eliminated potential doublets using Scanpy (version 1.7.1).18 Uniform manifold approximation and projection was used to visualize the distribution of cells in the projection of the significant principal components and cells were clustered using the Leiden algorithm.19 Trajectory analysis was performed using Monocle 3. Cell cycle states (G1, S, G2/M) were annotated by scoring gene sets with Scanpy. Details on bulk RNA sequencing and RNA immunoprecipitation sequencing (RIP-Seq) are provided in the Online Supplementary Material.

Statistical analysis

Data were analyzed using GraphPad Prism 8. Statistical significance was calculated by unpaired t tests between two groups or by one-way analysis of variance for multiple groups. Log-rank analysis was used to evaluate differences in Kaplan-Meier survival curves. Differences were considered statistically significant when P <0.05.

Results

Aberrant expression of RNA helicase DHX15 in human Tcell acute lymphoblastic leukemia

We compared the expression profiles between 21 normal thymocyte and 57 T-ALL samples,20 and revealed metabolism of RNA as the most significant enriched term across all differentially expressed genes (Figure 1A, B). Since RBP are key players in RNA metabolism, we then focused on analyzing the expression of 490 well-defined RBPs containing canonical RNA-binding motifs21,22 in two pairs of independent datasets of normal thymocyte and T-ALL samples20,23,24 and searched for upregulated RBP in leukemia. There were eight overlapping genes commonly elevated in T-ALL with a false discovery rate less than 0.01 (Figure 1C, Online Supplementary Figure S1A). Based on results from a previously performed CRISPR-Cas9 screen of RBP in Jurkat cells,22 we identified DHX15 as the most highly dependent factor among the eight candidates (Figure 1C). Gene expression studies, based on multiple human leukemia datasets, 20,23-26 verified that DHX15 expression was significantly higher in T-ALL patient samples than in normal thymocytes or bone marrow cells (Figure 1D, Online Supplementary Figure S1B). We further confirmed that DHX15 mRNA and protein were more abundant in seven human T-ALL cell lines relative to normal human thymocytes (Figure 1E, F). Consistently, elevated DHX15 protein expression was also observed in primary human T-ALL cells (Figure 1F, Online Supplementary Table S2). Of note, increased DHX15 expression was not associated with either specific T-ALL subgroups or NOTCH1 mutational status ( Online Supplementary Figure S1C, D). We next analyzed the Cancer Cell Line Encyclopedia27 and found the highest expression of DHX15 in TALL among human cancer cell lines from 38 tumor types (Online Supplementary Figure S1E). Taken together, these data demonstrate a global increase of DHX15 expression in T-ALL.

MYB directly activates DHX15 transcription in T-cell acute lymphoblastic leukemia

To understand the molecular mechanism underlying DHX15 upregulation in T-ALL, we tested the dependency of DHX15 expression on a panel of prominent transcription factors previously identified in T-ALL. Notably, only short hairpin RNA (shRNA)-mediated MYB knockdown significantly reduced DHX15 expression at mRNA and protein levels (Figure 2A, B, Online Supplementary Figure S2). Levels of cleaved caspase-3 were comparable in these samples, ruling out nonspecific DHX15 loss due to cell death upon MYB knockdown (Figure 2B). We analyzed previously reported chromatin immunoprecipitation sequencing data and found a strong binding signal of MYB in the DHX15 locus,28 along with histone H3 lysine 27 acetylation (H3K27ac), a

marker of active transcription (Figure 2C). We next performed conventional chromatin immunoprecipitation assays and validated a significant increase in MYB recruitment to the DHX15 promoter region (Figure 2D). We revealed three putative MYB responsive elements (YAACG/TG)29 proximal to the transcription start site (-436 bp, +137 bp, +143 bp) and constructed the promoter sequence in a luciferase reporter vector. The DHX15 reporter was strongly activated by enforced MYB expression whereas the reporter with mutant MYB binding sites was barely induced (Figure 2E). Furthermore, gene expression profiling of 562 primary T-ALL samples30-33 revealed a significant correlation between DHX15 and MYB mRNA levels (R=0.55, P<2.2x10-16) (Figure 2F). Collectively, this evidence demonstrates that oncogenic MYB specifically binds to DHX15 for direct transcriptional activation in T-ALL.

Essential role of DHX15 in sustaining T-cell acute lymphoblastic leukemia cell survival in vitro and in vivo

We next determined the role of DHX15 in T-ALL cell growth and survival using two shRNA. Both shRNA markedly depleted DHX15 expression at the mRNA and protein levels in human T-ALL KOPTK1 and CUTLL1 cells (Figure 3A), resulting in substantial growth inhibition (Figure 3B) and robust apoptosis (Online Supplementary Figure S3A, B). Cell cycle analysis showed a decreased G2/M and an increased sub-G1 phase distribution in cells devoid of DHX15 expression (Online Supplementary Figure S3C). To rule out the possibility of an off-target effect of shRNA, we reconstituted DHX15 expression by ectopically expressing its coding region resistant to shRNA targeting the 3 untranslated region (sh1) (Online Supplementary Figure S3D). Exogenous DHX15 rescued cell growth and viability to a great extent (Online Supplementary Figure S3E). Notably, while T-ALL cells showed a strict dependency on DHX15 expression, DHX15 loss led to relatively mild growth defects in normal CD34+ hematopoietic stem and progenitor cells (HSPC) (Online Supplementary Figure S3F-H).

To examine the in vivo significance of DHX15 in T-ALL, we established human xenografts using MOLT-3 cells co-expressing green fluorescent protein (GFP) as a surrogate marker and firefly luciferase to visualize leukemia cell expansion in viv o. Control or DHX15 -depleted MOLT-3 cells were intravenously injected into NPG mice ( Online Supplementary Figure S4A, B ). As shown in Figure 3C and Online Supplementary Figure S4C , we observed a marked delay in leukemia progression in mice receiving DHX15 -deficient cells, as determined by bioluminescent imaging. When control mice became moribund and exhibited severe splenomegaly, mice bearing DHX15 -depleted cells showed spleens of normal size (Figure 3D). Moreover, flow cytometry analysis of spleen and bone marrow cells from xenografted mice revealed much lower percentages of GFP + infiltrating leukemia

cells with DHX15 depletion (Figure 3E), which correlated with significantly prolonged survival (Figure 3F). These observations support the indispensable role of DHX15 in T-ALL progression in vivo .

DHX15 depletion inhibits NOTCH1-driven T-cell leukemogenesis

We then assessed the requirement of DHX15 in T-ALL initiation, and generated a conditional Dhx15 KO mouse

Figure 1. Elevated DHX15 expression in T-cell acute lymphoblastic leukemia. (A) Volcano plots showing the differentially expressed genes (fold change >1.3, false discovery rate <0.01) between T-cell acute lymphoblastic leukemia (T-ALL) patients’ samples (N=57, GSE33469) and normal thymocyte samples (n=21, GSE33470). (B) Bar graph showing enriched gene ontology terms across differentially expressed genes in (A). (C) Heatmap presentation of eight commonly elevated RNA-binding proteins (RBP) in 57 T-ALL patients’ samples (GSE33469) as compared to 21 normal thymocyte samples (GSE33470) from dataset 1 (left). Bar plots showing dependency scores of the eight genes from the RBP CRISPR-Cas9 screen in Jurkat cells (right).22 (D) DHX15 mRNA expression presented in dataset 1 (left) and dataset 2 (60 T-ALL patients’ samples and 20 normal thymocyte samples from GSE110637 and GSE151081) (right). The distributions of DHX15 mRNA expression are presented in box-and-whisker plots with the median value (line). Boxes and whiskers depict 25-75 and 10-90 percentiles respectively. Values outside of this range are plotted as individual points. ***P<0.001. (E) Analysis of DHX15 mRNA in normal human thymocytes and T-ALL cell lines by quantitative polymerase chain reaction. Data shown represent the means (± standard deviation), ***P<0.001. (F) Immunoblots of DHX15 in normal human thymocytes, T-ALL cell lines (upper) and primary patients’ T-ALL cells (bottom), as indicated. ACTIN served as a loading control. FDR: false discovery rate; DEG: differentially expressed genes

moter (Lck-Cre) (Online Supplementary Figure S5A). Deletion of exon 2 of the Dhx15 gene results in a premature stop codon and truncated RNA is subject to nonsense mediated decay (Online Supplementary Figure S5A). We purified CD4–CD8– double-negative (DN2-3 and DN4) and

Haematologica | 108 August 2023 2033 ARTICLE - The MYB-DHX15-mTORC1 axis promotes T-cell leukemia H. Guo et al.

CD4+CD8+ double-positive (DP) thymocytes from the LckCre Dhx15fl/fl strain (hereafter termed Dhx15 KO) and confirmed the knockout efficiency in these subsets (Online Supplementary Figure S5B, C).

To generate the T-ALL mouse model, lineage negative (Lin–) bone marrow cells from either Dhx15fl/fl or Dhx15 KO mice were isolated and transduced with the MSCV-IRESGFP retroviral vector expressing intracellular NOTCH1

Figure 3. DHX15 is required to maintain T-cell acute lymphoblastic leukemia cell survival in vitro and in vivo. (A) KOPTK1 and CUTLL1 cells were transduced with lentiviruses expressing short hairpin (sh)RNA targeting control (Ctrl) or DHX15 (sh1: targeting 3 UTR and sh2: targeting coding region). DHX15 mRNA and protein were analyzed by quantitative polymerase chain reaction (top) and immunoblot (bottom). (B) Live KOPTK1 and CUTLL1 cells were counted at the indicated time points and cell growth was plotted as shown. (C) Human T-cell acute lymphoblastic leukemia MOLT-3 cells, expressing both luciferase and green fluorescence protein (GFP) markers, were infected with shRNA targeting DHX15 mRNA (sh1) or control. Two million GFP+ cells were injected into irradiated NPG mice, followed by in vivo bioimaging to assess leukemia progression. Representative images of leukemia burden assessed by bioimaging in xenografted mice. Time points after engraftment are shown on the top. (D) Representative spleen images are shown (left) with spleen weights plotted (right) at the 28th day after transplantation. (E) Representative flow cytometry images of GFP+ spleen and bone marrow cells from mice shown in (D) (left). GFP+ percentages were plotted and are shown on the right (N=5 per group). (F) Kaplan-Meier survival curves of MOLT-3 xenografts (N=5 per group). **P<0.01; ***P<0.001. BM: bone marrow.

(ICN1), then transplanted into irradiated recipient mice (Figure 4A). Mice transplanted with ICN1-expressing Dhx15fl/fl cells showed an early increase in the frequency of GFP+ leukemia cells in peripheral blood which developed into overt CD4+CD8+ T-ALL within 6 weeks (Figure 4B, C, Online Supplementary Figure S5D), concomitant with accumulation of lymphoblasts in the peripheral blood (Figure 4D). In contrast, mice transplanted with Dhx15 KO progenitors showed a marginal increase of lymphoblastic leukemia cells over time (Figure 4B-D). Whereas all the mice transplanted with Dhx15fl/fl cells developed severe splenomegaly from leukemic infiltration within 8 weeks, enlarged spleens were not detected in mice transplanted with Dhx15 KO cells (Figure 4E). Histological examination of the spleens by hematoxylin-eosin and proliferating cell nuclear antigen staining confirmed much less leukemia infiltration as a result of Dhx15 gene inactivation (Figure 4F). While all mice in the control cohort succumbed to T-ALL within 70 days, mice bearing Dhx15 KO cells exhibited significantly prolonged survival (Figure 4G). Of note, almost half of the mice transplanted with Dhx15 KO cells were viable and remained diseasefree over 12 months after transplantation (Figure 4G), suggesting that inactivation of Dhx15 significantly suppresses leukemia onset and reduces the penetrance of ICN1-induced T-ALL. We hardly detected Dhx15 protein expression in leukemia cells derived from Dhx15 KO donors (Online Supplementary Figure S5E), indicating that these cells were not escapers from the Cre-induced inactivation and that other mechanisms may be involved in leukemogenesis despite Dhx15 KO. Moreover, knockout of Dhx15 barely affected HSPC infiltration into the bone marrow and spleen at the early stage (Online Supplementary Figure S5F, G), ruling out the possibility that the phenotype of delayed leukemogenesis was due to defective homing or engraftment of HSPC. These findings underscore the essential importance of DHX15 in T-ALL initiation. To test the effect of Dhx15 deletion on T-ALL maintenance, we also generated Mx1-Cre Dhx15fl/fl mice for the NOTCH1induced T-ALL model, which allows inducible Dhx15 KO following pIpC (poly I:C) treatment (Online Supplementary Figure S6A, B). As shown in Online Supplementary Figure S6C, D, pIpC injection significantly impaired GFP+ leukemia cell expansion in peripheral blood, bone marrow and spleen. Taken together, these data support the notion that DHX15 is required for both T-ALL initiation and maintenance.

DHX15 is required for normal T-cell development

Anatomical analysis of Dhx15 KO mice revealed a striking reduction in thymus size and cellularity (Figure 5A, Online Supplementary Figure S7A), which was much more pronounced in homozygotes than in heterozygotes (Online Supplementary Figure S7B), without a perceptible abnor-

mality in spleen size (Online Supplementary Figure S7C). Immunohistological staining of Ki67 and cleaved caspase 3 demonstrated that decreased thymic cellularity in Dhx15 KO mice resulted mainly from impaired cell proliferation but not enhanced cell death (Online Supplementary Figure S7D). As an expected outcome, mature CD3+, CD4+ and CD8+ T cells declined dramatically in the peripheral blood and spleen of Dhx15 KO animals (Online Supplementary Figure S7E, F).

We attempted to determine at which developmental stage Dhx15 has a major impact. Flow cytometry analysis of CD4 and CD8 expression revealed substantial reductions in the proportion and cellularity of DP thymocytes in Dhx15 KO mice, concomitant with accumulation of DN progenitors (Figure 5B, Online Supplementary Figure S7G). We further investigated the DN populations and observed increased DN3-4 cells in Dhx15 KO mice compared to those in Dhx15fl/fl controls (Online Supplementary Figure S7H). Again, the altered cellularity was not associated with enhanced apoptotic cell death (Online Supplementary Figure S7I).

To further clarify the heterogeneity of thymocyte subsets and the abnormal gene expression profiles in the absence of Dhx15, we performed single-cell RNA sequencing using the 10x Genomics platform on isolated single cells from thymus with or without Dhx15 expression (Online Supplementary Figure S8A). A total of 9,434 and 8,484 live cells were obtained from Dhx15fl/fl and Dhx15 KO thymus, respectively. Cells were clustered into nine annotated T-cell subgroups based on previously reported marker gene expression (Online Supplementary Figure S8B, C):34,35 DN cells, DN-to-DP transitional cells (DN/DPtrans), DP cells undergoing rearrangement (DPre1-4), DP cells under selection (DPsel1-2), and CD4/CD8 single-positive (SP) cells (Figure 5C). Consistent with flow cytometry analysis, cell type composition and pseudotime analysis showed substantial reductions of DP and SP populations, and accumulation of DN progenitors in Dhx15 KO thymus (Figure 5D, Online Supplementary Figure S8D). We next compared the gene expression profile of each cluster between two groups, and found the most striking differential expression in the DN-to-DP transitional cluster (Online Supplementary Figure S8E). Gene ontology analysis revealed that the majority of downregulated genes in Dhx15 KO cells were involved in cell cycle (Figure 5E). Cell cycle distribution and scoring analysis validated a significant impaired G2/M phase in Dhx15 KO cells (Figure 5F), consistent with the proliferation defect upon Dhx15 inactivation. Taken together, these results demonstrate that DHX15 is a key regulator of progenitor T-cell proliferation.

DHX15 loss alters RNA splicing and impairs mTORC1 activation

We next sought to understand the mechanism underlying the phenotypic consequences resulting from DHX15 de-

Figure 4. DHX15 deficiency impedes T-cell leukemogenesis in vivo. (A) Graphical illustration of the mouse model of NOTCH1-induced T-cell acute lymphoblastic leukemia. Lin– bone marrow cells from Dhx15fl/fl or Dhx15 knockout (KO) mice were transduced with MigR1-ICN1 retroviruses and transplanted into irradiated recipient mice, followed by assessment of leukemia dissemination. (B) Percentages of green fluorescent protein (GFP)+ cells in peripheral blood at the indicated times after transplantation (N=8 per group). (C) Representative flow cytometry analysis of CD4+CD8+ double-positive leukemia cells 6 weeks after transplantation. (D) Representative peripheral blood smear examined by Wright-Giemsa staining. Scale bar, 25 μm. (E) Spleen images (top) and flow cytometry analysis of GFP+ leukemia cells (bottom) at the indicated times after transplantation. (F) Representative images of spleen histology stained with hematoxylin and eosin (top) and immunohistochemical staining of proliferating cell nuclear antigen (bottom). Scale bar, 50 μm. (G) Kaplan-Meier survival curves for recipients of ICN1-transduced Lin– bone marrow cells from Dhx15fl/fl or Dhx15 KO donor mice (N=10 per group). ***P<0.001. BM: bone marrow; HE: hematoxylin and eosin; PCNA: proliferating cell nuclear antigen.

and retained introns. We detected widespread splicing changes in DHX15-depleted cells with skipped exons being the most frequently affected alternative splicing event. We also observed a distinct transcriptomic landscape which may result from altered RNA splicing (Figure 6A, Online Supplementary Figure S9A). By overlapping the targets undergoing differential splicing and expression from two independent shRNA we were able to identify 466 DHX15responsive genes whose expression changes were con-

Haematologica | 108 August 2023 2037 ARTICLE - The MYB-DHX15-mTORC1 axis promotes T-cell leukemia H. Guo et al.

Figure 6. DHX15 deficiency alters RNA splicing and leads to mTORC1 inactivation. (A) Scatterplot showing significant changes in alternative splicing (|ΔPSI|≥0.1) in CUTLL1 cells expressing DHX15 shRNA (sh1) or control (Ctrl). Counts of differentially spliced events in five types of alternative splicing are also shown (left). Heatmap showing differentially expressed genes ( P <0.05) upon DHX15 depletion (sh1 and sh2) in CUTLL1 cells (right). (B) Venn diagram showing the overlapping genes with alternative splicing and differential expression ( P <0.05, fold change >1.5) in (A). (C) Bubble graph showing gene ontology enrichment of 466 overlapping genes (B) analyzed by Metascape ( www.metascape.org ); the enriched terms were illustrated and aligned by P values. (D) DHX15 was depleted by shRNA (sh1 and sh2) in CUTLL1 and KOPTK1 cells. Phosphorylation of S6K (threonine 389) and 4EBP1 (threonine 37/46) was analyzed by immunoblot. (E) RNA-sequencing coverage plot of SLC7A6 and SLC38A5 in CUTLL1 cells with or without DHX15 depletion, overlaid with anti-DHX15 RNA immunoprecipitation sequencing tracks. Gray bars represent indicated exons. (F) DHX15 was depleted by shRNA (sh1) in CUTLL1 cells. SLC7A6 and SLC38A5 mRNA and protein were analyzed by quantitative polymerase chain reaction (left) and immunoblot (right). Data shown represent the means (± standard deviation), *** P <0.001. PSI: percentage splice in; SE: skipped exons; MXE: mutually exclusive exons; RI: retained introns; A3SS: alternative 3 splice sites; A5SS: alternative 5 splice sites; DEG: differentially expressed genes; AS: alternative splicing; RIP: RNA immunoprecipitation.

comitant with altered splicing (Figure 6B). Enrichment analysis of these genes revealed prominent downregulation of the mTORC1 pathway (Figure 6C), known to be required for cell cycle control37 as well as T-ALL progression and Tcell development.38 Moreover, gene set enrichment analysis (GSEA) showed the enrichment of mTORC1 signature genes in both CUTLL1 cells and DN/DP transitional cluster cells subjected to DHX15 ablation (Online Supplementary Figure S9B). In support of this notion, DHX15 deficiency significantly attenuated both S6K phosphorylation at threonine 389 and 4EBP1 phosphorylation at threonine 37/46, two classical markers of mTORC1 activation,39 in CUTLL1 and KOPTK1 cells (Figure 6D), corroborating an essential role of DHX15 in sustaining mTORC1 activity.

Splicing analysis revealed retained introns in the mRNA encoding solute carrier family amino acid transporters SLC7A6 and SLC38A5, which import glutamine and activate mTORC1.40,41 Furthermore, genome-wide DHX15-RNA interaction analyzed by RIP-Seq showed that DHX15 was capable of binding to these retained introns (Figure 6E), suggesting that efficient splicing of these genes is mediated by direct association of DHX15 to the mRNA transcripts. These intron-retained transcripts containing premature stop codons were subjected to nonsense mediated decay, leading to decreased mRNA and protein expression of SLC7A6 and SLC38A5 (Figure 6F, Online Supplementary Figure S9C). We also confirmed that knockdown of each transporter resulted in attenuation of mTORC1 signaling (Online Supplementary Figure S9D). To verify that mTORC1 is a major downstream effector in response to DHX15 activity, we knocked down TSC2, the negative regulator of mTORC1 signaling,42,43 and achieved constitutive activation in KOPTK1 cells. Sustained mTORC1 activation reversed the phenotype of massive cell death induced by DHX15 ablation (Online Supplementary Figure S9E). Activation of mTORC1 by the small molecule agonist 3BDO44 yielded the similar result of rescuing cell death (Online Supplementary Figure S9F). These studies demonstrate that the dependency of DHX15 in T-ALL is, at least in part, attributable to pathogenic splicing of genes involved in mTORC1 activation.

Identification of ciclopirox as a DHX15 modulator drug

Given the pro-leukemogenic role of DHX15 in multiple TALL models, we surmised that targeting DHX15 in leukemic lymphoblasts could be therapeutically beneficial. However, small molecules with DHX15 inhibitory potential are unavailable. We thus considered a drug repurposing strategy by interrogating the next-generation Connectivity Map, a large-scale compendium of gene expression readouts derived from tumor cells treated with bioactive small molecules.45 We intended to search for compounds with gene expression signatures resembling that induced by DHX15 depletion.

As shown in Figure 7A and Online Supplementary Figure S10A , this analysis identified ciclopirox, a broad-spectrum antifungal agent,46 as a top hit as a DHX15 inhibitory modulator. It is notable that this analysis also revealed rapamycin, the mTORC1 inhibitor, which mimics the effect induced by DHX15 loss. Intriguingly, we observed profound DHX15 loss in CUTLL1 and KOPTK1 cells treated with ciclopirox in a dose-dependent manner (Figure 7B, Online Supplementary Figure S10B). As mRNA and protein stability of DHX15 remained unaltered after ciclopirox exposure, it is most likely that ciclopirox downregulates DHX15 at the transcriptional level (Online Supplementary Figure S10C, D). We next compared the gene expression profiling of CUTLL1 cells treated with ciclopirox and cells subjected to DHX15 depletion, and found a large number of overlapping genes that are clustered with analogous expression changes (Figure 7C, D). As further support, GSEA of differentially expressed genes upon ciclopirox treatment showed significant enrichment in the gene expression signature elicited by DHX15 loss (Figure 7E). Vice versa, GSEA of differentially upregulated and downregulated genes upon DHX15 depletion showed a highly significant enrichment in the gene expression signature responsive to ciclopirox exposure (Online Supplementary Figure S10E). Similar to DHX15 ablation, administration of ciclopirox inhibited SLC38A5 and SLC7A6 expression as well as S6K and 4EBP1 phosphorylation (Figure 7F). These results argue for a convergent effect of ciclopirox treatment and DHX15 deficiency in T-ALL. Following these findings, we next assessed the effects of ciclopirox treatment on a panel of human T-ALL cells as well as normal bone marrow cells. As shown in Figure 7G, administration of ciclopirox elicited a strong anti-leukemic effect on TALL cells, whereas it had a minimal effect on normal bone marrow cells. More importantly, enforced DHX15 expression significantly reversed the cytotoxic effect induced by ciclopirox, suggesting that DHX15 is a crucial downstream sensing molecule in response to ciclopirox. Our data therefore suggest a targeting approach of modulating splicing catalysis by repressing DHX15 in TALL.

Discussion

Eukaryotic cells can fine-tune gene expression through a variety of mechanisms, in which many co- and post-transcriptional processes are coordinated by RBP.47,48 Given the pivotal role of these proteins in gene regulatory events, alterations of RBP expression and activity underlie many pathological conditions, including cancer.49 In this study, we show that the RNA helicase DHX15, transcriptionally activated by MYB, is required for T-cell leukemogenesis. Disruption of DHX15 resulted in mTORC1 inhibition via al-

tered splicing and downregulation of glutamine transporters. We here delineate a MYB-DHX15-mTORC1 axis in T-ALL that engages in leukemogenesis, and also present a DHX15 signature modulator drug, ciclopirox, which shows strong anti-leukemic effects.

As an RBP with helicase activity, accumulating evidence suggests the pivotal role of DHX15 in RNA splicing.9 We indeed observed alterations of RNA splicing and gene expression in response to DHX15 loss in T-ALL. Differential splicing and expression identify mTORC1 as a prominent

Figure 7. The DHX15 signature modulator drug ciclopirox shows antileukemic effects. (A) Connectivity Map analysis identified ciclopirox (CPX) as the top hit with signatures overlapping those induced by DHX15 depletion in CUTLL1 cells (left). A gene set consisting of the top 150 upregulated and downregulated genes (P<0.01) was used as input (www.broad.mit.edu/cmap). The structure of CPX is shown on the right. (B) Immunoblots of DHX15 in CUTLL1 and KOPTK1 cells treated with CPX for 12 h. (C) Venn diagram showing the overlap between DHX15-regulated and CPX-responsive genes in CUTLL1 cells (P<0.05, fold change >1.3). (D) Density scatter plot showing the correlation of 3,445 overlapping genes indicated in (C) between DHX15 depletion and CPX exposure. (E) Gene set enrichment analysis of CPX-responsive genes in the expression signature induced by DHX15 loss (P<0.001, fold change >1.4; top upregulated and downregulated 300 genes). (F) CUTLL1 and KOPTK1 cells were treated with CPX (5 or 10 μM) for 12 h; SLC7A6, SLC38A5, p-S6K T389 and p-4EBP1 T37/46 were analyzed by immunoblot. (G) Analysis of cell viability using the CCK8 Cell Proliferation Assay Kit in T-cell acute lymphoblastic leukemia cells and normal bone marrow cells subjected to CPX treatment for 12 h at various concentrations (left). KOPTK1 cells were infected with lentiviruses expressing ectopic DHX15 (DHX15 OE) or vector control (Ctrl), then subjected to CPX treatment for 12 h at the indicated concentrations. Cell viability was assessed by annexin V-propidium iodide staining and normalized to the untreated control (right). Data shown represent the means (± standard deviation), ***P<0.001. FC: fold change; NES: normalized enrichment score.

DHX15 responsive signaling. The phenotypic consequences of DHX15 depletion resemble those of mTORC1 inactivation in T-cell lineage that burst proliferation during DN-to-DP transition is profoundly blocked.38 Our findings thus link DHX15 to mTORC1 signaling as an important mechanism underlying T-cell maturation and T-ALL pathogenesis. Nevertheless, the precise mechanism of how DHX15 dysregulation affects alternative splicing in T-ALL remains to be fully defined.

In contrast to somatic spliceosomal mutations frequently found in myeloid leukemia, such genetic abnormalities are apparently lacking in T-ALL. We previously showed that enhanced SHQ1 expression driven by activating NOTCH1 mutations modulates snRNA peusdouridylation and regulates RNA splicing,50 suggesting that lymphoid malignancies may exploit distinct mechanisms for splicing regulation instead of splicing factor somatic mutations. In support of this notion, we here provide further evidence that MYB, overactivated by translocations, duplications and deregulation in T-ALL,51,52 induces the expression of DHX15, which modulates RNA splicing to enhance the oncogenic mTORC1 pathway. As such, aberrant expression of splicing factors could be a key mechanism underlying oncogenic splicing events and neoplastic phenotypes. Recent reports provide the proof of principle that deubiquitinase USP7, which harbors recurrent genetic mutations in T-ALL, contributes to SRSF6 and SF3B1 stabilization.22,53 It is therefore reasonable to speculate that T-ALL-specific genetic alterations, exemplified by those found in NOTCH1, MYB and USP7, govern aberrant expression of splicing factors/modulators. The consequent disordered splicing programs meet the needs for neoplastic transformation, with no necessity of additional spliceosomal mutations.

The widespread splicing changes in human cancers produce pro-tumorigenic isoforms, at the expense of reducing the fidelity of normal splicing function, thus conferring a specific vulnerability to splicing inhibitors. This unique feature propels the development of splicing modulators for the treatment of cancer. Although a number of chemical compounds that modulate splicing catalysis have been described to date, nearly all of these drugs inhibit early spliceosome assembly or SR protein phosphorylation.7 We here propose clinically approved ciclopirox as a DHX15 signature modulator, thus providing actionable insights toward targeting the late stage of spliceosome disassembly. Supposing that not all introns require the same level of splicing fidelity/complexity, targeting spliceosome disassembly may be less toxic and have more specific effects than inhibiting the early stage of splicing during which the majority of introns are probably affected to some extent. Although the mechanism of action of ciclopirox in regu-

lating DHX15 remains to be determined, these findings suggest novel means of splicing perturbation with clinically available drugs.

Disclosures

No conflicts of interest to disclose.

Contributions

HL and PVV conceived and designed the study. HL, GQ and PVV supervised the study. HL wrote the manuscript. HG and JX performed most of the experiments. PX and ZL conducted the RNA splicing analysis. QL and SP provided technical support. DW analyzed the RNA-sequencing data and performed the Connectivity MAP screen. JL performed RNA immunoprecipitation sequencing and conducted the data analysis. JR, BP, JH, TT and PVV helped to analyze gene expression profiles of primary T-cell and T-acute lymphoblastic leukemia samples. CT and PZ helped to analyze the single-cell RNA sequencing data.

Acknowledgments

We thank members of Liu’s , Van Vlierberghe’s and Qing’s laboratories for helpful suggestions and the Core Facility of the Medical Research Institute at Wuhan University for flow sorting and histological analysis. We are grateful to Dr Jinyong Wang (Guangzhou Institute of Biomedicine and Health, CAS, China) and Dr Haojian Zhang (Wuhan University, China) for providing the Lck-Cre and Mx1-Cre mouse strains. We gratefully acknowledge the Therapeutically Applicable Research to Generate Effective Treatment (TARGET)/Children's Oncology Group (COG), St. Jude Children’s Research Hospital and Japan Adult Leukemia Study Group (JALSG) for making their invaluable data available.

Funding

This study was supported by grants from the National Key R&D Program of China (2022YFA1103200), National Natural Science Foundation of China (82161138024, 82025003 and 82011530151 to HL, 81830084 to GQ, 32170565 to ZL), Hubei Provincial Natural Science Fund for Creative Research Groups (2021CFA003 to HL), the Research Foundation Flanders (G0E6222N to PVV) and CAS Hundred Talents Program to ZL.

Data-sharing statement

Raw sequencing data have been deposited in the NCBI Gene Expression Omnibus (GEO) database under accession number GSE208746. Other data supporting the findings of this study are available from the corresponding author, Hudan Liu.

References

1. Yang Q, Jeremiah Bell J, Bhandoola A. T-cell lineage determination. Immunol Rev. 2010;238(1):12-22.

2. Van Vlierberghe P, Ferrando A. The molecular basis of T cell acute lymphoblastic leukemia. J Clin Invest. 2012;122(10):3398-3406.

3. Pui CH, Robison LL, Look AT. Acute lymphoblastic leukaemia. Lancet. 2008;371(9617):1030-1043.

4. Litzow MR, Ferrando AA. How I treat T-cell acute lymphoblastic leukemia in adults. Blood. 2015;126(7):833-841.

5. Cordo V, van der Zwet JCG, Canté-Barrett K, Pieters R, Meijerink JPP. T-cell acute lymphoblastic leukemia: a roadmap to targeted therapies. Blood Cancer Discov. 2021;2(1):19-31.

6. Elcheva IA, Spiegelman VS. Targeting RNA-binding proteins in acute and chronic leukemia. Leukemia. 2021;35(2):360-376.

7. Chen S, Benbarche S, Abdel-Wahab O. Splicing factor mutations in hematologic malignancies. Blood. 2021;138(8):599-612.

8. Lee SC, Abdel-Wahab O. Therapeutic targeting of splicing in cancer. Nat Med. 2016;22(9):976-986.

9. Bohnsack KE, Kanwal N, Bohnsack MT. Prp43/DHX15 exemplify RNA helicase multifunctionality in the gene expression network. Nucleic Acids Res. 2022;50(16):9012-9022.

10. Arenas JE, Abelson JN. Prp43: an RNA helicase-like factor involved in spliceosome disassembly. Proc Natl Acad Sci U S A. 1997;94(22):11798-11802.

11. Tsai RT, Fu RH, Yeh FL, et al. Spliceosome disassembly catalyzed by Prp43 and its associated components Ntr1 and Ntr2. Genes Dev. 2005;19(24):2991-3003.

12. Martin A, Schneider S, Schwer B. Prp43 is an essential RNAdependent ATPase required for release of lariat-intron from the spliceosome. J Biol Chem. 2002;277(20):17743-17750.

13. Wan R, Yan C, Bai R, Lei J, Shi Y. Structure of an intron lariat spliceosome from Saccharomyces cerevisiae. Cell. 2017;171(1):120-132.

14. Zhang X, Zhan X, Yan C, et al. Structures of the human spliceosomes before and after release of the ligated exon. Cell Res. 2019;29(4):274-285.

15. Ito S, Koso H, Sakamoto K, Watanabe S. RNA helicase DHX15 acts as a tumour suppressor in glioma. Br J Cancer. 2017;117(9):1349-1359.

16. Jing Y, Nguyen MM, Wang D, et al. DHX15 promotes prostate cancer progression by stimulating Siah2-mediated ubiquitination of androgen receptor. Oncogene. 2018;37(5):638-650.

17. Li X, Gounari F, Protopopov A, Khazaie K, von Boehmer H. Oncogenesis of T-ALL and nonmalignant consequences of overexpressing intracellular NOTCH1. J Exp Med. 2008;205(12):2851-2861.

18. Wolf FA, Angerer P, Theis FJ. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol. 2018;19(1):15.

19. Traag VA, Waltman L, van Eck NJ. From Louvain to Leiden: guaranteeing well-connected communities. Sci Rep. 2019;9(1):5233.

20. Van Vlierberghe P, Ambesi-Impiombato A, Perez-Garcia A, et al. ETV6 mutations in early immature human T cell leukemias. J Exp Med. 2011;208(13):2571-2579.

21. Wang E, Lu SX, Pastore A, et al. Targeting an RNA-binding protein network in acute myeloid leukemia. Cancer Cell. 2019;35(3):369-384.

22. Zhou Y, Han C, Wang E, et al. Posttranslational regulation of the exon skipping machinery controls aberrant splicing in leukemia. Cancer Discov. 2020;10(9):1388-1409.

23. Verboom K, Van Loocke W, Volders PJ, et al. A comprehensive inventory of TLX1 controlled long non-coding RNAs in T-cell

acute lymphoblastic leukemia through polyA+ and total RNA sequencing. Haematologica. 2018;103(12):e585-e589.

24. Roels J, Kuchmiy A, De Decker M, et al. Distinct and temporaryrestricted epigenetic mechanisms regulate human αβ and γδ T cell development. Nat Immunol. 2020;21(10):1280-1292.

25. Homminga I, Pieters R, Langerak AW, et al. Integrated transcript and genome analyses reveal NKX2-1 and MEF2C as potential oncogenes in T cell acute lymphoblastic leukemia. Cancer Cell. 2011;19(4):484-497.

26. Haferlach T, Kohlmann A, Wieczorek L, et al. Clinical utility of microarray-based gene expression profiling in the diagnosis and subclassification of leukemia: report from the International Microarray Innovations in Leukemia Study Group. J Clin Oncol. 2010;28(15):2529-2537.

27. Barretina J, Caponigro G, Stransky N, et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature. 2012;483(7391):603-607.

28. Mansour MR, Abraham BJ, Anders L, et al. Oncogene regulation. An oncogenic super-enhancer formed through somatic mutation of a noncoding intergenic element. Science. 2014;346(6215):1373-1377.

29. Biedenkapp H, Borgmeyer U, Sippel AE, Klempnauer KH. Viral myb oncogene encodes a sequence-specific DNA-binding activity. Nature. 1988;335(6193):835-837.

30. Gocho Y, Liu J, Hu J, et al. Network-based systems pharmacology reveals heterogeneity in LCK and BCL2 signaling and therapeutic sensitivity of T-cell acute lymphoblastic leukemia. Nat Cancer. 2021;2(3):284-299.

31. Yang L, Chen F, Zhu H, et al. 3D genome alterations associated with dysregulated HOXA13 expression in high-risk T-lineage acute lymphoblastic leukemia. Nat Commun. 2021;12(1):3708.

32. Zhu H, Dong B, Zhang Y, et al. Integrated genomic analyses identify high-risk factors and actionable targets in T-cell acute lymphoblastic leukemia. Blood Sci. 2022;4(1):16-28.

33. Seki M, Kimura S, Isobe T, et al. Recurrent SPI1 (PU.1) fusions in high-risk pediatric T cell acute lymphoblastic leukemia. Nat Genet. 2017;49(8):1274-1281.

34. Li Y, Li K, Zhu L, et al. Development of double-positive thymocytes at single-cell resolution. Genome Med. 2021;13(1):49.

35. Zhou W, Yui MA, Williams BA, et al. Single-cell analysis reveals regulatory gene expression dynamics leading to lineage commitment in early T cell development. Cell Syst. 2019;9(4):321-337.

36. Zhang J, Huang J, Xu K, et al. DHX15 is involved in SUGP1mediated RNA missplicing by mutant SF3B1 in cancer. Proc Natl Acad Sci U S A. 2022;119(49):e2216712119.

37. Fingar DC, Richardson CJ, Tee AR, Cheatham L, Tsou C, Blenis J. mTOR controls cell cycle progression through its cell growth effectors S6K1 and 4E-BP1/eukaryotic translation initiation factor 4E. Mol Cell Biol. 2004;24(1):200-216.

38. Hoshii T, Kasada A, Hatakeyama T, et al. Loss of mTOR complex 1 induces developmental blockage in early T-lymphopoiesis and eradicates T-cell acute lymphoblastic leukemia cells. Proc Natl Acad Sci U S A. 2014;111(10):3805-3810.

39. Ma XM, Blenis J. Molecular mechanisms of mTOR-mediated translational control. Nat Rev Mol Cell Biol. 2009;10(5):307-318.

40. Jewell JL, Kim YC, Russell RC, et al. Metabolism. Differential regulation of mTORC1 by leucine and glutamine. Science. 2015;347(6218):194-198.

41. Durán RV, Oppliger W, Robitaille AM, et al. Glutaminolysis activates Rag-mTORC1 signaling. Mol Cell. 2012;47(3):349-358.

ARTICLE - The MYB-DHX15-mTORC1 axis promotes T-cell leukemia

42. Garami A, Zwartkruis FJ, Nobukuni T, et al. Insulin activation of Rheb, a mediator of mTOR/S6K/4E-BP signaling, is inhibited by TSC1 and 2. Mol Cell. 2003;11(6):1457-1466.

43. Inoki K, Li Y, Xu T, Guan KL. Rheb GTPase is a direct target of TSC2 GAP activity and regulates mTOR signaling. Genes Dev. 2003;17(15):1829-1834.

44. Ge D, Han L, Huang S, et al. Identification of a novel MTOR activator and discovery of a competing endogenous RNA regulating autophagy in vascular endothelial cells. Autophagy. 2014;10(6):957-971.

45. Subramanian A, Narayan R, Corsello SM, et al. A next generation connectivity map: L1000 platform and the first 1,000,000 profiles. Cell. 2017;171(6):1437-1452.

46. Sehgal VN. Ciclopirox: a new topical pyrodonium antimycotic agent. A double-blind study in superficial dermatomycoses. Br J Dermatol. 1976;95(1):83-88.

47. Glisovic T, Bachorik JL, Yong J, Dreyfuss G. RNA-binding proteins and post-transcriptional gene regulation. FEBS Lett. 2008;582(14):1977-1986.

48. Xiao R, Chen JY, Liang Z, et al. Pervasive chromatin-RNA binding protein interactions enable RNA-based regulation of transcription. Cell. 2019;178(1):107-121.

49. Pereira B, Billaud M, Almeida R. RNA-binding proteins in cancer: old players and new actors. Trends Cancer. 2017;3(7):506-528.

50. Su H, Hu J, Huang L, et al. SHQ1 regulation of RNA splicing is required for T-lymphoblastic leukemia cell survival. Nat Commun. 2018;9(1):4281.

51. Clappier E, Cuccuini W, Kalota A, et al. The C-MYB locus is involved in chromosomal translocation and genomic duplications in human T-cell acute leukemia (T-ALL), the translocation defining a new T-ALL subtype in very young children. Blood. 2007;110(4):1251-1261.

52. Lahortiga I, De Keersmaecker K, Van Vlierberghe P, et al. Duplication of the MYB oncogene in T cell acute lymphoblastic leukemia. Nat Genet. 2007;39(5):593-595.

53. Han C, Khodadadi-Jamayran A, Lorch AH, et al. SF3B1 homeostasis is critical for survival and therapeutic response in T cell leukemia. Sci Adv. 2022;8(3):eabj8357.

Comprehensive molecular and clinical characterization of NUP98 fusions in pediatric acute myeloid leukemia

Eline J. M. Bertrums,1,2,3* Jenny L. Smith,4* Lauren Harmon,5* Rhonda E. Ries,4 Yi-Cheng J. Wang,6,7 Todd A. Alonzo,6,7 Andrew J. Menssen,8 Karen M. Chisholm,9 Amanda R. Leonti,4 Katherine Tarlock,4,10 Fabiana Ostronoff,11 Era L. Pogosova-Agadjanyan,4 Gertjan J. L. Kaspers,1,12,13 Henrik Hasle,14 Michael Dworzak,15,16 Christiane Walter,17 Nora Mühlegger,15 Cristina Morerio,18 Laura Pardo,8 Betsy Hirsch,19 Susana Raimondi,19 Todd M. Cooper,10 Richard Aplenc,20 Alan S. Gamis,21 Edward A. Kolb,22 Jason E. Farrar,23 Derek Stirewalt,4 Xiaotu Ma,24 Tim I. Shaw,24 Scott N. Furlan,4 Lisa Eidenschink Brodersen,8 Michael R. Loken,8 Marry M. van den Heuvel-Eibrink,1,25

C. Michel Zwaan,1,2,13 Timothy J. Triche Jr.,5,6,26 Bianca F. Goemans1#and Soheil Meshinchi4,7,10#

1Princess Máxima Center for Pediatric Oncology, Utrecht, the Netherlands; 2Department of Pediatric Oncology/Hematology, Erasmus Medical Center – Sophia Children’s Hospital, Rotterdam, the Netherlands; 3Oncode Institute, Utrecht, the Netherlands; 4Fred Hutchinson Cancer Research Center, Clinical Research Division, Seattle, WA, USA; 5Department of Epigenetics, Van Andel Institute, Grand Rapids, MI, USA; 6Department of Translational Genomics, University of Southern California, Los Angeles, CA, USA; 7Children's Oncology Group, Monrovia, CA, USA; 8 Hematologics Inc., Seattle, WA, USA; 9Department of Laboratories, Seattle Children’s Hospital, Seattle, WA, USA; 10Division of Hematology and Oncology, Seattle Children's Hospital, Seattle, WA, USA; 11Intermountain Blood and Marrow Transplant and Acute Leukemia Program, Intermountain Healthcare, Salt Lake City, UT, USA; 12Emma Children’s Hospital, Amsterdam UMC, Vrije Universiteit Amsterdam, Pediatric Oncology, Amsterdam the Netherlands; 13Dutch Childhood Oncology Group, Den Haag, the Netherlands; 14Department of Pediatrics, Aarhus University Hospital, Aarhus, Denmark; 15Children's Cancer Research Institute, Medical University of Vienna, Vienna, Austria; 16St. Anna Kinderspital, Department of Pediatrics, Medical University of Vienna, Vienna, Austria; 17Department of Pediatric Hematology and Oncology, University Hospital Essen, Essen, Germany; 18Laboratory of Human Genetics, IRCCS Istituto Giannina Gaslini, Genoa, Italy; 19Department of Pathology, St. Jude Children’s Research Hospital, Memphis, TN, USA; 20Division of Oncology and Center for Childhood Cancer Research, Children's Hospital of Philadelphia, Philadelphia, PA, USA; 21Division of Hematology/Oncology, Children's Mercy Kansas City, Kansas City, MO, USA; 22Nemours Alfred I. duPont Hospital for Children, Wilmington, DE, USA; 23Arkansas Children’s Research Institute and Department of Pediatrics, Hematology/Oncology Section, Department of Pediatrics, University of Arkansas for Medical Sciences, Little Rock, AR, USA; 24Computational Biology Department, St. Jude Children’s Research Hospital, Memphis, TN, USA; 25Utrecht University, Utrecht, the Netherlands and 26Department of Pediatrics, Michigan State University College of Human Medicine, Grand Rapids, MI, USA

*EJMB, JLS and LH contributed equally as first authors #BFG and SM contributed equally as last authors.

Abstract

Correspondence: Eline J. M. Bertrums

e.j.m.bertrums@prinsesmaximacentrum.nl

S. Meshinchi smeshinc@fredhutch.org

Received: June 27, 2022.

Accepted: February 14, 2023.

Early view: February 23, 2023.

https://doi.org/10.3324/haematol.2022.281653

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

NUP98 fusions comprise a family of rare recurrent alterations in AML, associated with adverse outcomes. In order to define the underlying biology and clinical implications of this family of fusions, we performed comprehensive transcriptome, epigenome, and immunophenotypic profiling of 2,235 children and young adults with AML and identified 160 NUP98 rearrangements (7.2%), including 108 NUP98-NSD1 (4.8%), 32 NUP98-KDM5A (1.4%) and 20 NUP98-X cases (0.9%) with 13 different fusion partners. Fusion partners defined disease characteristics and biology; patients with NUP98-NSD1 or NUP98-KDM5A had distinct immunophenotypic, transcriptomic, and epigenomic profiles. Unlike the two most prevalent NUP98 fusions, NUP98-X variants are typically not cryptic. Furthermore, NUP98-X cases are associated with WT1 mutations, and have epigenomic profiles that resemble either NUP98-NSD1 or NUP98-KDM5A. Cooperating FLT3-ITD and WT1 mutations define NUP98-NSD1, and chromosome 13 aberrations are highly enriched in NUP98-KDM5A. Importantly, we demonstrate that NUP98 fusions portend dismal overall survival, with the noteworthy exception of patients bearing abnormal chromosome 13 (clinicaltrials gov. Identifiers: NCT00002798, NCT00070174, NCT00372593, NCT01371981).

Introduction

Acute myeloid leukemia (AML) accounts for 15-20% of all pediatric leukemias and is a very heterogeneous disease.1,2 Besides early response to induction treatment assessed by morphology and flow cytometry-based measurable residual disease (MRD), cytogenetic and molecular aberrations are the most important prognostic factors that guide risk group stratification.1,3 Although survival rates of pediatric AML (pAML) patients have improved significantly, over the last decade these have reached a plateau, with long-term survival rates around 70-80%.3,4 A third of all pAML patients relapse, and their outcome is poor.3 In addition, treatment-related toxicity and mortality make intensification of treatment challenging.2,3 Thus, the identification of prognostic subgroups for risk group and treatment stratification is of utmost value to improve treatment and outcomes of specifically high-risk subtypes.2 Due to the very low prevalence of some subgroups, studies to identify these cases can be challenging and therefore require international collaboration.

NUP98 (chromosome 11p15) encodes a nucleoporin protein, which is part of the nuclear pore complex.5 NUP98 was first shown to be fused to HOXA9 in t(7;11) FAB (French-American-British classification) M2 and M4 AML in 1996.6 In the last 20 years, over 30 different partner genes in AML and therapy-related myelodysplastic syndrome have been described.7-10 NUP98 fusion proteins involve the N-terminal portion of NUP98 and the C-terminal portion of the fusion partner.5 These fusion partners consist of homeodomain proteins, which are transcription factors, and non-homeodomain proteins, which are thought to play a role in transcriptional or epigenetic regulation.5 In pAML patients, NUP98 translocations with KDM5A and NSD1 have been most frequently described.11,12 These patients are now notorious for inferior outcome compared to non-NUP98-translocated patients and are treated as high-risk patients in most current treatment protocols.9,13 However, NUP98 translocations with other partners, here called NUP98-X, are rare, and their prognostic relevance is unknown; consequently, there is a necessity to define the optimal risk stratification and treatment strategy for these patients. Here, we present the molecular and clinical characteristics of NUP98-translocated pAML patients within four consecutive Children’s Oncology Group (COG) trials and an International Berlin-FrankfurtMünster AML study group (I-BFM AML SG) collaboration. We aim to define the clinical relevance for all NUP98 translocations with cooperating mutations and copy number variants.

Methods

Patient samples

Patients enrolled in the COG trials CCG-2961, AAML03P1,

AAML0531 and AAML1031 were eligible for this study. Details of these studies have been previously described.14-17 In total, 3,493 patients were included in these studies, of which 2,235 were eligible for inclusion due to availability of comprehensive NUP98 fusion, molecular, and clinical data (Online Supplementary Figure 1; Online Supplementary Tables S1 and S2). For the remaining patients, these data were unavailable. In addition, we sent out an I-BFM AML SG proposal to include pediatric AML patients with a NUP98-X translocation from other study groups. Consent, in accordance with the Declaration of Helsinki, was obtained from all study participants. The Fred Hutchinson Cancer Research Center Institutional Review Board and the COG Myeloid Biology Committee approved and oversaw the conduct of this study. Adult AML patients from the Beat AML study, The Cancer Genome Atlas AML (TCGA LAML), and Southwestern Oncology Group (SWOG) AML studies were included as comparators for NUP98 fusion analysis and details were reported accordingly in references.18-23

Screening of NUP98 fusions

The NUP98 fusions were detected by either karyotype or combined fusion detection algorithms STAR-fusion v1.8.1, TransAbyss v1.4.10, and CICERO v0.1.824-26 completed on RNA sequencing (RNA-seq). For differences in detection methods per COG trial, see the Online Supplementary Appendix. The majority (94%) of NUP98-translocated patients had RNA-seq evidence of their fusion. STAR-fusion was run using default parameters with the premade GRCh37 resource library with Gencode v19 annotations (https://data.broadinstitute.org/Trinity/CTAT_RESOURCE_LIB/ ). The TransAbyss software was executed with the GRCh37lite reference genome with the following parameters included: fusion breakpoint reads ≥1, flanking pairs and spanning reads ≥2 counts. CICERO fusion detection was performed with default parameters with GRCh37-lite. Fusions detected computationally were verified using Fusion Inspector v.1.8.1 (Broad Institute, Cambridge, MA) and visualized on IGV27-30 and BAMBINO.31 Beat AML (n=440) and SWOG AML (n=206) transcriptome sequence reads were analyzed using STAR-fusion v1.8.1 with the same reference resource library and parameters as above.24 TCGA LAML (n=179) RNA-seq fusion data were downloaded from supplementary materials.19

Statistical methods

Data were current as of March 31, 2019. The Kaplan-Meier method was used to estimate overall survival (OS, defined as time from study entry to death) and event-free survival (EFS, time from study entry until failure to achieve complete remission [CR] during induction, relapse, or death). Relapse risk (RR) was calculated by cumulative incidence methods defined as time from the end of induction I for patients in CR to relapse or death, where deaths without a relapse were

considered competing events. Patients who withdrew from therapy due to relapse, persistent central nervous system (CNS) disease, or refractory disease with >20% bone marrow blasts by the end of induction I were defined as induction I failures. MRD was defined at the end of course one using flow cytometry with a cut-off of 0.1% detection of disease. The I-BFM patients were excluded from survival analyses due to variation in study groups and treatment protocols.

Results

Clinical characteristics

Between 1995 and 2017, 3,493 AML patients were treated on consecutive COG trials CCG-2961, AAML03P1, AAML0531, and AAML1031, of which 2,235 were eligible for comprehensive outcome (see Methods) and cytomolecular association analyses. Within this cohort, 160 patients (7.2%) with a NUP98 translocation were identified (Figure 1A); the remaining 2,075 patients were included as a reference cohort. In addition, six NUP98-X cases were included via the I-BFM AML SG, demonstrating that while rare, NUP98-X cases are present in multiple cohorts of patients. However, to prevent bias due to confounding variables, such as differences in study groups, fusion identification methods and treatment protocols, these patients were excluded from further analyses. Characteristics of all NUP98-X patients are depicted in the

Online Supplementary Table S3

The most common NUP98 translocations were NUP98-NSD1 (n=108) and NUP98-KDM5A (n=32; Online Supplementary Figure S2A). Furthermore, we identified 20 patients with 13 different NUP98 translocation partners, including HOXA9 (n=4), HOXD13 (n=3), PHF15 (n=2), PHF23 (n=2) and single cases of BPTF, BRWD3, DDX10, HMGB3, HOXA13, KAT7, PRRX1, SET, and TOP1 (Figure 1B). Interestingly, contrasting the cryptic NUP98-NSD1 and NUP98-KDM5A fusions, an overwhelming majority of NUP98-X fusions were detectable as karyotypic variants with 17 of 20 (85%) having gross alterations by gbanding cytogenetics involving chr11p15.

Initial comparison of the NUP98 fusion cohort to the reference patients demonstrated a significant sex bias in NUP98 cases, with 61.3% being male versus 38.8% female (P=0.012) (Figure 1C). In particular, the NUP98-NSD1 cohort contained 64.8% male versus 35.2% female patients. Additionally, NUP98 fusions were enriched in children aged 3-10 years old (35.6%; P=0.005). Clinical characteristics for NUP98-translocated subgroups are summarized in Figure 1C-E and the Online Supplementary Table S1. In NUP98-NSD1-translocated patients, white blood cells and blast cell counts were both significantly higher, while in NUP98-KDM5A patients a reverse trend was seen. Classification by conventional cytomolecular stratification schemas, as previously described,17 revealed that 39% of NUP98-NSD1 patients had been classified as standard-risk (SR) and 61% as high-risk (HR). In

contrast, most NUP98-KDM5A patients (97%) and NUP98-X patients (95%) were classified as SR.

Comparison of NUP98 translocations with age at diagnosis based on fusion partners (Figure 1C) showed that NUP98NSD1 cases had a median age of 10.2 years (reference cohort 10.0; P=0.228), whereas NUP98-KDM5A cases had a median age of 2.7 years (P<0.001). NUP98-X patients showed a median age of 7.9 years (P=0.30) with a bi-modal distribution; 40% of the patients were under 3 years and 50% over 5 years with no patients over 18 years old (Online Supplementary Figure S2B, C). Almost all NUP98-X patients with homeobox fusion partners (n=9) were over 3 years old (8/9), with one exception (NUP98-HOXD13; P=0.025).

From Beat AML, TCGA LAML and SWOG, 825 adult AML cases were screened for NUP98 fusions by RNA-seq fusion detection algorithms. Zero NUP98-KDM5A, 11 (1.3%) NUP98-NSD1, and two (0.2%) NUP98-X fusions were identified (NUP98TOP1 and NUP98-RAP1GDS1). These results demonstrate that NUP98 rearrangements are less common, but still present, in older adult AML patients (13/825, 1.6%) compared to pediatric and young adult AML patients (160/2,235, 7.2%; P<0.001; Online Supplementary Figure S2D)9.

Implications of variation of fusion junction

Analysis of NUP98 fusion breakpoints by RNA-seq revealed a high diversity of NUP98 exon junctions. Nearly 85% of NUP98 fusions had a breakpoint junction in exon 12 (39.7%) or 13 (44.9%), while the remaining breakpoints occurred in various positions from exon 11 to exon 29 (Online Supplementary Figure S2E). Exon junctions correlated with the fusion partner, and NUP98-NSD1 fusions primarily had exon 12 (52.9%) and 13 (43.23%) junctions (Figure 1F). However, exon 14 breakpoints were almost uniformly restricted to NUP98-KDM5A compared to other NUP98 fusions (P<0.001). NUP98-X cases showed a larger variability in NUP98 exon breakpoints (Online Supplementary Table S4; Online Supplementary Figure S3). NUP98 homeobox gene fusions were enriched in exon 12 breakpoints (6/9, 66%), while PHF15 (n=2), PHF23 (n=2), and TOP1 partners had exon 13 breakpoints. Besides the commonly included nucleoporin FG repeat domains of the NUP98 protein, a minority of cases (n=3) included a larger portion of the protein with nucleoporin autopeptidase or, additionally, the Nup96 domains (Figure 2A).

Immunophenotypes

NUP98 fusions were previously reported to be associated with erythroid and megakaryocytic phenotypes.10,32 Upon morphology, we identified that only NUP98-KDM5A fusions were more often associated with the FAB M6/M7 category compared to the reference cohort (46.9% vs. 5.5%; P<0.001). Additionally, the immunophenotype of NUP98 fusions was examined using multidimensional flow cytometry.33 NUP98-NSD1 patients expressed early progenitor

Figure 1. Clinical characteristics of patients with and without NUP98 translocations. (A) Oncoprint depicting the major drivers of pediatric acute myeloid leukemia (AML) patients. (B) Circos plot depicting commonly co-occurring mutations and cytogenetic abnormalities in NUP98-translocated pediatric AML patients. (C) Pie charts depicting the sex divisions of patients in the NUP98-translocated AML subgroups. (D) Circos plot representing different fusion partner genes of NUP98-X translocations in pediatric AML patients. (E) Age distribution of AML patients. (F) Barchart (polar axis) illustrating the prevalence of NUP98 exon junctions in NUP98translocated AML. The Figure legend is ordered by decreasing NUP98 exon prevalence in the NUP98 fusion-positive cohort.

markers such as CD34 and CD117 (Figure 2B; Online Supplementary Figure S4). Patients harboring NUP98-NSD1 and FLT3 internal tandem duplication (-ITD) retained the immature markers but also showed evidence of monocytic maturation compared to NUP98-NSD1 without FLT3-ITD, as demonstrated by expression of CD11b (84%), CD36 (55%) and CD64 (71%) (Online Supplementary Table S6). Nevertheless, NUP98-NSD1 associated phenotypes were not as specific, or consistent, as seen in NUP98-KDM5A. The NUP98-KDM5A immunophenotype corresponded to megakaryocytic maturation, with at least partial expression of CD36, and absence of pluripotent markers CD34 and CD123. Notably, NUP98-KDM5A patients showed clusters that associated with co-occurrence of abnormal chromosome 13q (Figure 2C), demonstrating that these subsets of patients display a unique immunophenotype. Finally, NUP98-X fusions lacked consistent immunophenotype, aside from the majority expressing markers of early progenitors.

Cooperating karyotypic and molecular variants

Diagnostic specimens were evaluated for common translocations, chromosomal aberrations and common mutations, namely FLT3-ITD, WT1, NPM1, CEBPA, KIT and CBL mutations (Figures 1A and 2C). We confirmed the wellknown enrichment of FLT3-ITD (74%) and WT1 mutations (42%) in the NUP98-NSD1 cohort.5 Almost half of NUP98NSD1 patients with FLT3-ITD also harbored a WT1 mutation, indicating triple positivity for adverse outcome variants in AML.34 NUP98-NSD1 patients also had a significant association with trisomy 8 (18.8%) compared to the reference cohort (5.3%; P<0.001). NUP98-KDM5A displayed a paucity of cooperating mutations. NUP98-X patients showed a higher prevalence of WT1 mutations compared to the reference cohort (25% vs. 9.6%; P=0.039), associated with a higher age at diagnosis (median age 16.3 vs. 2.3 years; P=0.032). We identified a notably high correlation of NUP98-KDM5A with chromosome 13 (chr13) structural variants, including del(13q), monosomy 13 and chr13 translocations. Abnormal chr13 (NUP98-KDM5A/13abn) was identified in 19 NUP98KDM5A patients (63.3% vs. 2.3% in the reference cohort; P<0.001). NUP98-KDM5A/13abn were significantly younger than NUP98-KDM5A/13normal patients (median age 1.8 vs 9.6 years; P<0.001). Thirteen NUP98-KDM5A patients (43.3%) patients harbored del(13q) versus one NUP98-X patient (NUP98-SET). Monosomy 13 (2/32) and translocation 13 (4/32) occurred less frequently in NUP98-KDM5A and were not found in NUP98-NSD1 or NUP98-X (Online Supplementary Table S5).

The majority of del(13q) in NUP98-translocated cases (92%) began at band 13q12; the deletions ranged from 8 Mb to 59.5 Mb to the entire chromosome. The minimal commonly deleted segment was del(13)(q14.2q14.3), containing the RB1 tumor suppressor gene (Online Supplementary Figure S5).

RB1 loss has been previously reported in patients with NUP98-KDM5A;10 however, we demonstrated a much larger region of copy number alterations, including numerous additional genes. Of NUP98-KDM5A/13abn patients, 84% (16/19) had NUP98 exon 13 breakpoints, suggesting that specific NUP98 exon breakpoints in the fusion transcript may be linked to the presence of additional cytogenetic abnormalities. Finally, all ten acute megakaryoblastic leukemia (AMKL; FAB M7) NUP98-KDM5A cases with karyotype data available had chr13 alterations compared to three AMKL cases without NUP98 fusions (P<0.001).

Gene expression profiling

Unsupervised hierarchal clustering of gene expression in NUP98-translocated patients and a reference cohort of known fusions including KMT2A, CBFB-MYH11, RUNX1RUNX1T1 and DEK-NUP214, as well as 84 healthy controls (n=988), revealed that the majority of NUP98-NSD1 (n=104), NUP98-KDM5A (n=32), and the reference cohort cluster by fusion identity, while no uniform clustering of NUP98-X was observed (n=20) (Figure 3A). In order to further understand transcriptional similarities and differences between the diverse NUP98 fusions, uniform manifold approximation and projection (UMAP) was completed on the NUP98-translocated patients’ gene expression data (n=156). The Leiden algorithm35 identified five transcriptional clusters. NUP98-NSD1 patients clustered together, clearly separated from the majority of NUP98-KDM5A patients (Figure 3B). NUP98-X patients were dispersed, clustering more closely with NUP98-NSD1 cases. The largest proportion of NUP98-X clustering together included seven homeobox and both PHF23 partners, suggesting transcriptional similarities between NUP98-HOX fusions (cluster C3). The next cluster most associated with NUP98-X (cluster C5) included the majority of non-AMKL NUP98KDM5A cases.

UMAP revealed segregation based on an AMKL and agebased signature (cluster C4), which embodied 78.6% of AMKL NUP98-translocated patients, all 3 years old or younger. The cluster primarily contained NUP98-KDM5A (22/32) cases and was enriched in NUP98 exon 13 breakpoints. NUP98-X patients in C4 included single cases of NUP98-SET with del(13q), NUP98-BPTF with AMKL morphology, and NUP98-DDX10. Additionally, C4 included all NUP98-KDM5A/13abn cases (Online Supplementary Figure S6A), separating NUP98-KDM5A with and without chr13 abnormalities. Conversely, in a separate UMAP including heterogenous pAML fusions (N=1,482), this abn13-based clustering was not observed for non-NUP98-translocated subtypes (Online Supplementary Figure S6B).

Differential expression analysis compared NUP98-X directly to NUP98-NSD1 and NUP98-KDM5A individually. Expression of MECOM and PRDM16, known prognostic markers in adult and pediatric AML,36 effectively separated

Figure 2. Cytogenetics of NUP98-translocated pediatric acute myeloid leukemia. (A) Locations of breakpoints across the NUP98 gene for all NUP98-translocated acute myeloid leukemia (AML). (B) Oncoprint depicting additional copy number variations (CNV) and mutations in NUP98-translocated patients. (C) Heatmap depicting the presence and absence of flow-cytometry immunophenotype markers in NUP98-translocated AML groups. NUP98-HOX-like fusions include fusion partners HOXA9, HOXA13, HOXD13, and PRRX1. NUP98-Reader-like fusions include fusion partners BPTF, BRWD3, DDX10, HMGB3, KAT7, PHF15, SET, and TOP1

Figure 3. Expression pattern of pediatric acute myeloid leukemia with various NUP98 translocations. (A) Unsupervised hierarchical clustering by gene expression including heterogenous pediatric acute myeloid leukemia (AML) subtypes, NUP98-translocated subgroups, normal healthy bone marrows (NBM) and CD34+ peripheral blood cells (CD34 PB). Annotation bars show AML subtype and co-occurring mutations. (B) Uniform manifold approximation and projection (UMAP) of gene expression, followed by Leiden clustering, for NUP98-translocated pediatric AML samples identifies five different transcriptional clusters. NUP98 fusions are indicated in different colors: NUP98-KDM5A in purple, NUP98-NSD1 in blue, and NUP98-X in green. (C) Expression of MECOM and PRDM16 genes in different subgroups of NUP98-translocated pediatric leukemia. Same identification colors for NUP98 fusions as in (B) are used. (D) Expression of stemness marker genes in all NUP98-translocated samples. Top bars represent French-American-British (FAB) classification and age category.

the NUP98 subgroups (Figure 3C). Interestingly, about two-thirds of NUP98-KDM5A highly expressed both genes, while NUP98-X and NUP98-NSD1 almost exclusively overexpressed one or the other (Figure 3D). NUP98-KDM5A patients with low MECOM and low PRDM16 expression almost uniformly lacked chr13 alterations. Additionally, NUP98-KDM5A/13abn patients had reduced expression of genes in the involved area, including RB1 (P<0.001), DLEU7 and SPRYD7 (Online Supplementary Figure S6C). We attempted to identify transcriptional signatures that might be shared between all NUP98-translocated cases and performed differential expression analysis comparing each group (NUP98-X, NUP98-NSD1, and NUP98-KDM5A) independently to the reference cohort (n=1,326) (Figure 4A). This analysis confirmed high inter-patient variability of NUP98-X fusions (Online Supplementary Figure S7).

Gene expression profiling revealed 27 differentially expressed genes (DEG) exclusively shared between NUP98X and NUP98-NSD1, including upregulation of DNMT3B, MYCN, and PBX3 (Figure 4B). Within NUP98-KDM5A, a bimodal expression pattern of DNMT3B and MYCN was related to chr13 alterations, where cases lacking chr13 aberrations had decreased expression. NUP98-X and NUP98-KDM5A exclusively shared 26 dysregulated genes, including overexpression of MLLT3, IRX3, and CD79a

The NUP98-translocated cohort had 38 DEG in common, including upregulation of numerous HOX genes. Among these 38 genes, 15 were also dysregulated in NUP98translocated cohorts compared to healthy bone marrow samples. This minimal set of 15 genes strongly implicated dysregulation at the HOX loci; these targets include HOXA (chr7p15), HOXB (chr17q21), hsa-mir-10a (chr17q21), and CACNG4 (chr17q24) transcripts (Figure 4C). NUP98-X cases expressed HOXA/B genes regardless of their fusion partner, and 60% (12/20) expressed both HOXA/B while the remaining third primarily overexpressed the HOXA cluster (Online Supplementary Figure S8). Overexpression of HOX genes and hsa-mir-10a was previously reported in NUP98KDM5A and NUP98-NSD1 and is now shown to be a common feature of NUP98 translocations.9,13

Single-sample gene-set enrichment analysis (ssGSEA) addresses the inherent variability within diverse NUP98 fusions and was performed to investigate alterations in the expression of down-stream targets of hsa-mir-10a and HOX transcription factors.37 NUP98-translocated subgroups had significantly lower enrichment scores of miR-10a-3p and miR-10a-5p/miR-10b-5p target genes, an indication of negative regulation, compared to normal bone marrow samples (P<0.001). Investigation of HOX transcription factor (TF) pathways by ssGSEA, revealed enrichment in HOXB8 molecular interactions (adj.P<0.008). The HOXB8 pathway included well known HOX transcriptional co-factors MEIS1, MEIS2, PBX1, PBX3, PBX3, 38 and the proto-oncogene RAF1. NUP98-X and NUP98-KDM5A exhibited a positive enrich-

ment of HOXA9 interacting partners (P<0.001; Figure 4D). Additionally, we employed RCIS-Target to identify TF motifs enriched in the overexpressed genes (fold-change >2.0) for each NUP98-translocated cohort (Online Supplementary Table S7). This revealed a transcriptional network in NUP98KDM5A with GATA1 and GATA2 both highly upregulated compared to the reference cohort, and their downstream target genes concomitantly overexpressed, with concurrent downregulation of ERG, which is known to have an inverse relationship with GATA expression.39

DNA methylation profiling

We analyzed DNA methylation data from 334,934 CpG probes. We then performed dimensionality reduction using non-negative matrix factorization (NMF) and used UMAP to determine how the variation in DNA methylation associates with NUP98 fusion groups and normal bone marrow (NBM). We found that NUP98 fusion groups cluster together (Figure 5A). Specifically, the HOX-activating fusions (NSD1, HOX, and PRRX1) form a unique cluster, and also the fusion partners with reader-like functions (BPTF, BRWD3, DDX10, HMGB3, KAT7, PHF15, PHF23, SET, and TOP1) cluster together. The reader-like fusions also cluster more closely to NBM. By performing unsupervised clustering of the NMF factors that associate with each group, we found that the NUP98-HOX-like group clusters distinctly from the NUP98readers and NBM, further illustrating that NUP98 fusions differ in methylation profiles (Figure 5B).

We further analyzed the NMF factors that associate significantly with the NUP98-HOX, NUP98-Reader, and NUP98-Reader plus abn13 groups (Figure 5C). In order to identify the defining characteristics within each of these factors, we performed enrichment analyses against chromatin states, histone marks, and transcription factor binding sites (Figure 5D). The NUP98-HOX group enrichments in NMF 3 indicate that these fusions lead to Polycombmediated hypermethylation at actively transcribed genes, evidenced by H2AK119ub, H3K23me2, H3K36me2/3, and H3K27me3 enrichment at binding motifs for RYBP (a subunit of Polycomb repressor complex 1). This likely occurs because H3K36me2/3 increases throughout the HOXA/B clusters and at HOX targets, while H3K27me3 peaks disappear as hyperactive NSD1 displaces PRC1/2 from the HOX clusters. This may result in reducing expressing potential and arresting cellular differentiation, which often coincides with loss of imprinting, as evidenced by dual enrichment for H3K27me3 and H3K36me3, as well as transcription factor binding site enrichment for ZFP57, the master regulator of genomic imprinting control regions. The enrichment of NMF 6 suggests that NUP98-Reader fusions likely lead to localization of transcriptional condensates at already highly expressed developmental genes, leading to an enrichment of DNA hypermethylation in transcribed exons.

The enrichment of NMF 8 suggests that abnormal chr13 cases, all of which are NUP98-Reader fusions, show additional hypermethylation of actively expressed gene bodies, evidenced by enrichment for H3K36me3 and H2BK120ub, along with loss of imprinting (though far less pronounced than in NUP98-HOX fusions), which is evidenced by an enrichment for ZFP57 binding sites.

Clinical outcome and prognostic relevance

We evaluated the impact of NUP98 translocations on response to initial induction therapy. Overall, the morphologic CR rate after course one for the NUP98 fusion cohort was 50% versus 78% for the reference cohort (P<0.001). NUP98NSD1 patients had a significantly lower CR rate of 38% (P<0.001) compared the reference cohort, while NUP98-

Figure 4. Differential expression of all NUP98-translocated pediatric acute myeloid leukemia patients. (A) Schematic of differential expression analyses completed for NUP98-translocated samples. The overlap of differentially expressed genes (DEG) identified in each NUP98 cohort is represented in the Venn diagram. (B) DEG between NUP98-translocated AML groups compared to the reference cohort and normal bone marrow (NBM) were identified. Subsets of dysregulated genes were commonly identified in both NUP98-X and NUP98-NSD1 (upper panel) or were identified as shared between NUP98-X and NUP98-KDM5A (lower panel). (C) Commonly DEG found in all three NUP98-translocated pediatric acute myeloid leukemia (AML) subgroups. (D) Mean expression (Z-score transformed) of HOXA9 and HOXB8 interacting partners in NUP98-X translocated AML. The darker shades of red indicate higher expression in the NUP98-X cohort. CPM: counts per million.

Figure 5. DNA methylation of pediatric acute myeloid leukemia patients with NUP98 translocations.

(A) Uniform manifold approximation and projection (UMAP) of DNA methylation data in NUP98 -translocated acute myeloid leukemia (AML) subgroups compared to the reference cohort and normal bone marrow (NBM).

(B) Heatmap of non-negative matrix factorizations (NMF) of DNA methylation data. The NMF factors are those that were signi fi can tly associated with NUP98 translocation AML subgroups.

(C) NMF factor associations of DNA methylation with NUP98-HOX -like fusions ( NUP98-NSD1 , NUP98-HOX , and NUP98-PRRX1 ) and NUP98 -Reader-like fusions ( NUP98KDM5A , NUP98BPTF, NUP98-BRWD3, NUP98-DDX10, NUP98-KAT7, NUP98PHF15, NUP98-SET, and NUP98-TOP1 ) with or without a co-occuring abnormal chromosome 3 (chr3).

(D) NMF factor enrichments of DNA methylation for chromatin states, chromatin marks, and transcription factor binding sites. F actor 3 is enriched in NUP98-HOXlike fusions, factor 6 is enriched in NUP98 -Reader-like fusions, and factor 8 is enriched in NUP98 -Reader-like fusions with an abnormal chr13. FDR: false discovery rate; OR: odds ratio; ReprPC: repressed PolyComb; TssAFlank: fl anking active TSS, BivFlank: fl anking bivalent; TSS/Enh, EnhBiv: bivalent enhancer; TxWk: weak transcription;

Tx: Strong transcription; Enh: enhancers; EnhG: genic enhancers.

Figure 6. Survival of pediatric acute myeloid leukemia patients with NUP98 translocations. Kaplan Meier estimates of (A) overall survival (OS) and (B) relapse risk (RR) of pediatric NUP98-translocated acute myeloid leukemia (AML) patients with different translocation partners compared to a reference cohort without NUP98 fusions. OS of (C) NUP98-NSD1 and (D) NUP98-X, when divided by NUP98 fusion exon breakpoint. Outcome was also examined for (E) OS and (F) event-free survival (EFS) of NUP98KDM5A subgroups by chromosome 13 (chr13) status (monosomy 13, del(13q), translocation 13). Abn3: abnormal chr 3.

KDM5A and NUP98-X had CR rates of 81% (P=0.729) and 65% (P=0.176), respectively. NUP98-NSD1 and NUP98-KDM5A patients had significantly higher evidence of MRD (73%; P<0.001, and 52%; P=0.005, respectively), while this was similar to the reference cohort in NUP98-X (22% vs. 27%; P=0.793). The 5-year OS for the NUP98 fusion cohort was 35% versus 64% for the reference group (P<0.001). NUP98-NSD1 patients had inferior OS (36% vs. 64%, P<0.001) and event-free survival (EFS) (17% vs. 47%; P<0.001) compared to the reference (Online Supplementary Table S1; Figure 6A, B). Similarly, adverse outcomes for NUP98-KDM5A were observed for OS (30%; P<0.001) and EFS (25%; P=0.01). NUP98-NSD1 and NUP98-KDM5A cases showed a significantly higher 5year relapse risk (RR) of 64% (P=0.001) and 68% (P=0.010) respectively, compared to the reference cohort (42%) (Figure 6B). NUP98-X displayed a similar inferior OS (35%; P=0.009); however, EFS (35%; P=0.333) and RR (69%; P=0.071) differences did not reach significant difference. Response to treatment in NUP98-translocated subgroups, examined by disease-free survival (DFS) estimates 5 years after induction one, was lower compared to the reference cohort (27% vs. 52%; P<0.001). This held true for all subsets; NUP98-NSD1 (28%; P<0.001), NUP98-KDM5A (28%; P=0.012) and NUP98-X (23%; P=0.044) (Online Supplementary Figure S9A).

Multivariable cox regression analyses were performed to adjust for cytomolecular risk groups, white blood cells, and different NUP98-translocated subgroups (Online Supplementary Table S8). After correction, significantly inferior OS (hazard ratio [HR]=1.463; 95% confidence interval [CI]: 1.1-1.94; P=0.009), EFS (HR=2.032; 95% CI: 1.59-2.59; P<0.001) and RR (HR=1.743; 95% CI: 1.1-2.76; P=0.018) were observed in NUP98NSD1 patients compared to the reference group. Also, NUP98-KDM5A (HR=1.825; 95% CI: 1.13-2.96; P=0.015) and NUP98-X patients (HR=1.75; 95% CI: 1.01-3.04; P=0.046) showed poor OS, without significant differences in EFS and RR.

We examined outcomes corresponding to fusion exon junctions (Figure 6C, D; Online Supplementary Figure 9A, B). There were no significant differences in outcome for NUP98-NSD1 or NUP98-X by exon junction, though a trend toward improved outcomes was observed for NUP98-X exon 13 breakpoints. NUP98-KDM5A patients with exon 13 junctions (n=19) had an OS of 51% compared to 0% for exon 14 breakpoints (n=12; P=0.011) with corresponding EFS (40% vs. 0%, respectively; P=0.174). Due to high concurrence of chr13 alterations with exon 13 junctions, a similar trend was observed in NUP98-KDM5A/13abn compared to NUP98-KDM5A/13normal patients (EFS 45% vs. 0%; P=0.052). NUP98-KDM5A patients had a worse prognosis compared to the reference cohort without NUP98 fusions regardless of chr13 alterations; however, the presence of chr13 alterations within the NUP98KDM5A group was associated with increased OS and EFS (Figure 6E, F).

Discussion

NUP98-translocated pAML has emerged as a distinct but heterogeneous group, and a comprehensive study defining varied fusion partners, phenotypes, transcript subclasses and outcomes was still lacking. Incorporation of genome, transcriptome, methylation, and clinical data from several large pediatric and adult AML studies provided deep insight into this family of fusions. Our study demonstrates that the underlying biology of NUP98-translocated AML is defined by the fusion partner. Furthermore, although fusions involving NSD1 and KDM5A are cryptic, an overwhelming majority of NUP98-X fusions can be identified by conventional karyotype, facilitating identification at diagnosis.

Importantly, we identified a significant overlap of cooperating lesions including mutations (FLT3, WT1) and karyotypic alterations (trisomy 8, del13q). We confirmed prior observation of substantial enrichment of FLT3-ITD in NUP98-NSD1 patients (80%). This extreme prevalence, and NUP98-NSD1 preceding FLT3-ITD, suggest a causal relationship; this intriguing hypothesis is being studied in our laboratory.

Recently, exon usage and fusion junctions were shown to have clinical and biological implications; patients with a CBFB-MYH11 fusion with the common exon 5/33 breakpoint had significantly inferior EFS than those with less common fusion junctions.40 We here demonstrated that patients with NUP98-KDM5A with exon 13 involvement had a more favorable prognosis. However, the strong association of exon 13 usage with chr13 alterations in NUP98-KDM5A patients makes it difficult to discern which of these factors is underlying this outcome difference (Figure 6; Online Supplementary Figure S9). The difference that we discovered in prognosis may suggest that NUP98-KDM5A cases with or without exon 13 breakpoints and chr13 abnormalities could be divided into different subgroups. The observation that NUP98KDM5A/abn13 patients have a more favorable prognosis potentially affects treatment stratification of these patients in future. Furthermore, these findings provide a rationale that future studies must go beyond simple defining the presence or absence of a fusion and investigate specific exon usage, inclusion/exclusion of critical functional domains, and functionality of the oncoprotein.

Transcriptome profiling further defined functional classifications of NUP98 fusions. Expression of PRDM16 and MECOM could clearly segregate NUP98-translocated subsets. PRDM16 and MECOM encode H3K9-mono methyltransferases that are important in the maintenance of heterochromatin integrity and are selectively expressed in hematopoietic stem cells (HSC)36 and linked to oncogenic transformation;41 their deregulation could play a role in leukemogenesis of NUP98 translocations. Gene expression profiling also revealed distinct expression networks defined by translocation partner and cooperating mutations/alterations. Not only did NUP98KDM5A patients cluster based on abn13, bimodal expression

of DNMT3B, MYCN, MECOM, and PRDM16 was associated with abn13. Interestingly, where PRDM16 is a poor prognostic factor in AML,36 high expression was associated with NUP98KDM5A/abn13, which had a better prognosis in our cohort. This association may indicate different molecular pathways underlying leukemogenesis within NUP98-KDM5A, where NUP98-KDM5A/13abn may have more immature HSC-like features.

Regardless of fusion partner, NUP98 translocations shared overexpression of HOXA/B genes. The translocation partners KDM5A and PHF23 contain PHD protein domains, which function in histone methylation and nucleosome remodelling.25 The HOX cluster was shown to be in a locked, transcriptionally active position due to the H3K4me3-binding PHDdomain when fused to NUP9825 and this may be extendable to the leukemogenic ability of PHF15 and BPTF, which retain their PHD finger. Translationally, upregulation of the HOXA cluster indicates a potential therapeutic role of menin-inhibitors in NUP98-translocated AML, as has been recently shown in mice42 and in vitro studies of primary pAML samples.43

Chromatin modifiers, such as NSD1 and KDM5A, are frequently the targets of oncogenic fusions in pediatric disease.44 DNA methylation profiling suggests that rare and diverse NUP98-X fusions share one of the two key mechanisms to promote leukemogenesis: either by activating the HOX genes and their targets and promoting loss of genomic imprinting (like NUP98-NSD1), or by directing transcriptional machinery to developmentally inappropriate targets (as seen in NUP98-KDM5A fusions and chromatin reader fusions). The mutational, structural, transcriptional, and epigenomic signatures of these two major groups of NUP98 fusion partners are so starkly distinct that one cannot help but speculate that each group should be treated as a separate subtype of AML, where both common and rare partners are likely to respond to similar treatments, whether repurposed (disulfiram for chromatin reader fusions) or novel (CDK9 inhibitors for HOX fusions).

Overall, NUP98 fusions constitute a highly refractory class of AML, which justifies reclassification of NUP98 fusions, regardless of fusion partner, as a high-risk subtype in future trials. Further research may focus on NUP98 fusion cases with aberrations of chr13, typically co-occurring with NUP98 exon 13 breakpoints and a distinctive immunophenotype, whose outcomes are relatively favorable given standard of care induction and/or transplantation. The balance of NUP98 fusions, with or without characteristic co-occurring mutations, remains an urgent, unmet therapeutic need. The immunophenotype, transcriptome, and epigenome of HOX-activating (versus chromatin-reader) fusion partners may provide important leads towards more effective ther-

apies, while their signatures may permit rapid discontinuation of ineffective therapies in this high-risk group of patients.

Disclosures

AJM, LEB and LP are employees/paid consultants for Hematologics Inc.. MRL is an employee/paid consultant for and holds ownership interest in Hematologics. All other authors have no conflicts of interest to disclose.

Contributions

BFG and SM supervised the study. HH, MD, CW, NM and CM included additional patient data. JLS, LH, YJW, AJM, TJT, XM, TIS, RER, ARL, and ELP processed and analyzed the data. EJMB and JLS drafted the manuscript. All authors edited and approved the manuscript.

Acknowledgments

The authors wish to gratefully acknowledge the important contributions of the late Dr. Stephen H. Petersdorf to SWOG and to the study S0106.

Funding

This work was supported by the following NIH/NCI/NCTN grant awards: RO1CA190661, R01CA160872, R01AI171984, U10CA180888, U10CA180819, and U24CA196175, U10CA180886, U10CA180899, St. Baldricks Foundation, the Rally Foundation, and the Michelle Lunn Hope Foundation. This content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Data-sharing statement

RNA-sequencing and DNA methylation array data on primary patient samples, as well as associated clinical/outcome data, are deposited in Genomic Data Commons (GDC, https://portal.gdc.cancer.gov/) and the Target Data Matrix (https://ocg.cancer.gov/programs/target/data-matrix) under project ID "TARGET-AML". Access to protected files hosted on the Sequence Read Archive (SRA), such as raw sequencing data in bam or fastq format, are available through dbGaP TARGET: Acute Myeloid Leukemia study (accession: phs000465.v20.p8). Additional DNA methylation data are hosted on the Gene Expression Omnibus (GEO) under accessions GSE190931 and GSE124413. The Beat AML Study controlled access RNA-sequencing data were downloaded from the Genomic Data Commons (GDC) portal and are available through the Functional Genomic Landscape of Acute Myeloid Leukemia study on dbGaP (accession: phs001657.v1.p1). TCGA LAML RNA-sequencing fusion data were accessed from the GDC Data Portal (https://gdc.cancer.gov/about-data/publications/laml_2012).

References

1. Balgobind BV, Hollink IH, Arentsen-Peters ST, et al. Integrative analysis of type-I and type-II aberrations underscores the genetic heterogeneity of pediatric acute myeloid leukemia. Haematologica. 2011;96(10):1478-1487.

2. Zwaan CM, Kolb EA, Reinhardt D, et al. Collaborative efforts driving progress in pediatric acute myeloid leukemia. J Clin Oncol. 2015;33(27):2949-2962.

3. de Rooij JD, Zwaan CM, van den Heuvel-Eibrink M. Pediatric AML: from biology to clinical management. J Clin Med. 2015;4(1):127-149.

4. Creutzig U, Zimmermann M, Ritter J, et al. Treatment strategies and long-term results in paediatric patients treated in four consecutive AML-BFM trials. Leukemia. 2005;19(12):2030-2042.

5. Michmerhuizen NL, Klco JM, Mullighan CG. Mechanistic insights and potential therapeutic targets for NUP98-rearranged hematologic malignancies. Blood. 2020;136(20):2275-2289.

6. Nakamura T, Largaespada DA, Lee MP, et al. Fusion of the nucleoporin gene NUP98 to HOXA9 by the chromosome translocation t(7;11)(p15;p15) in human myeloid leukaemia. Nat Genet. 1996;12(2):154-158.

7. Gough SM, Slape CI, Aplan PD. NUP98 gene fusions and hematopoietic malignancies: common themes and new biologic insights. Blood. 2011;118(24):6247-6257.

8. Struski S, Lagarde S, Bories P, et al. NUP98 is rearranged in 3.8% of pediatric AML forming a clinical and molecular homogenous group with a poor prognosis. Leukemia. 2017;31(3):565-572.

9. Hollink IH, van den Heuvel-Eibrink MM, Arentsen-Peters ST, et al. NUP98/NSD1 characterizes a novel poor prognostic group in acute myeloid leukemia with a distinct HOX gene expression pattern. Blood. 2011;118(13):3645-3656.

10. de Rooij JD, Branstetter C, Ma J, et al. Pediatric non-Down syndrome acute megakaryoblastic leukemia is characterized by distinct genomic subsets with varying outcomes. Nat Genet. 2017;49(3):451-456.

11. Jaju RJ, Fidler C, Haas OA, et al. A novel gene, NSD1, is fused to NUP98 in the t(5;11)(q35;p15.5) in de novo childhood acute myeloid leukemia. Blood. 2001;98(4):1264-1267.

12. de Rooij JD, Hollink IH, Arentsen-Peters ST, et al. NUP98/JARID1A is a novel recurrent abnormality in pediatric acute megakaryoblastic leukemia with a distinct HOX gene expression pattern. Leukemia. 2013;27(12):2280-2288.

13. Noort S, Wander P, Alonzo TA, et al. The clinical and biological characteristics of NUP98-KDM5A in pediatric acute myeloid leukemia. Haematologica. 2020;106(2):630-634.

14. Lange BJ, Smith FO, Feusner J, et al. Outcomes in CCG-2961, a children's oncology group phase 3 trial for untreated pediatric acute myeloid leukemia: a report from the children's oncology group. Blood. 2008;111(3):1044-1053.

15. Cooper TM, Franklin J, Gerbing RB, et al. AAML03P1, a pilot study of the safety of gemtuzumab ozogamicin in combination with chemotherapy for newly diagnosed childhood acute myeloid leukemia: a report from the Children's Oncology Group. Cancer. 2012;118(3):761-769.

16. Aplenc R, Meshinchi S, Sung L, et al. Bortezomib with standard chemotherapy for children with acute myeloid leukemia does not improve treatment outcomes: a report from the Children's Oncology Group. Haematologica. 2020;105(7):1879-1886.

17. Pollard JA, Loken M, Gerbing RB, et al. CD33 expression and its association with gemtuzumab ozogamicin response: results from the randomized phase III Children's Oncology Group Trial AAML0531. J Clin Oncol. 2016;34(7):747-755.

18. Tyner JW, Tognon CE, Bottomly D, et al. Functional genomic

landscape of acute myeloid leukaemia. Nature. 2018;562(7728):526-531.

19. Ley TJ, Miller C, Ding L, et al. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med. 2013;368(22):2059-2074.

20. Anderson JE, Kopecky KJ, Willman CL, et al. Outcome after induction chemotherapy for older patients with acute myeloid leukemia is not improved with mitoxantrone and etoposide compared to cytarabine and daunorubicin: a Southwest Oncology Group study. Blood. 2002;100(12):3869-3876.

21. Petersdorf SH, Rankin C, Head DR, et al. Phase II evaluation of an intensified induction therapy with standard daunomycin and cytarabine followed by high dose cytarabine for adults with previously untreated acute myeloid leukemia: a Southwest Oncology Group study (SWOG-9500). Am J Hematol. 2007;82(12):1056-1062.

22. Godwin JE, Kopecky KJ, Head DR, et al. A double-blind placebocontrolled trial of granulocyte colony-stimulating factor in elderly patients with previously untreated acute myeloid leukemia: a Southwest Oncology Group study (9031). Blood. 1998;91(10):36073615.

23. List AF, Kopecky KJ, Willman CL, et al. Benefit of cyclosporine modulation of drug resistance in patients with poor-risk acute myeloid leukemia: a Southwest Oncology Group study. Blood. 2001;98(12):3212-3220.

24. Haas BJ, Dobin A, Li B, et al. Accuracy assessment of fusion transcript detection via read-mapping and de novo fusion transcript assembly-based methods. Genome Biol. 2019;20(1):213.

25. Robertson G, Schein J, Chiu R, et al. De novo assembly and analysis of RNA-seq data. Nat Methods. 2010;7(11):909-912.

26. Tian L, Li Y, Edmonson MN, et al. CICERO: a versatile method for detecting complex and diverse driver fusions using cancer RNA sequencing data. Genome Biol. 2020;21(1):126.

27. Robinson JT, Thorvaldsdóttir H, Winckler W, et al. Integrative genomics viewer. Nat Biotechnol. 2011;29(1):24-26.

28. Thorvaldsdóttir H, Robinson JT, Mesirov JP. Integrative Genomics Viewer (IGV): high-performance genomics data visualization and exploration. Brief Bioinform. 2013;14(2):178-192.

29. Robinson JT, Thorvaldsdóttir H, Wenger AM, Zehir A, Mesirov JP. Variant review with the integrative genomics viewer. Cancer Res. 2017;77(21):e31-e34.

30. Robinson JT, Thorvaldsdóttir H, Turner D, Mesirov JP. igv.js: an embeddable JavaScript implementation of the Integrative Genomics Viewer (IGV). Bioinformatics. 2023;39(1):btac830.

31. Edmonson MN, Zhang J, Yan C, et al. Bambino: a variant detector and alignment viewer for next-generation sequencing data in the SAM/BAM format. Bioinformatics. 2011;27(6):865-866.

32. Chisholm KM, Heerema-McKenney AE, Choi JK, et al. Acute erythroid leukemia is enriched in NUP98 fusions: a report from the Children's Oncology Group. Blood Adv. 2020;4(23):6000-6008.

33. Eidenschink Brodersen L, Alonzo TA, Menssen AJ, et al. A recurrent immunophenotype at diagnosis independently identifies high-risk pediatric acute myeloid leukemia: a report from Children's Oncology Group. Leukemia. 2016;30(10):2077-2080.

34. Ostronoff F, Othus M, Gerbing RB, et al. NUP98/NSD1 and FLT3/ITD coexpression is more prevalent in younger AML patients and leads to induction failure: a COG and SWOG report. Blood. 2014;124(15):2400-2407.

35. Traag VA, Waltman L, van Eck NJ. From Louvain to Leiden: guaranteeing well-connected communities. Sci Rep. 2019;9(1):5233.

36. Shiba N, Ohki K, Kobayashi T, et al. High PRDM16 expression identifies a prognostic subgroup of pediatric acute myeloid leukaemia correlated to FLT3-ITD, KMT2A-PTD, and NUP98-NSD1: the results of the Japanese Paediatric Leukaemia/Lymphoma Study Group AML-05 trial. Br J Haematol. 2016;172(4):581-591.

37. Barbie DA, Tamayo P, Boehm JS, et al. Systematic RNA interference reveals that oncogenic KRAS-driven cancers require TBK1. Nature. 2009;462(7269):108-112.

38. Dard A, Reboulet J, Jia Y, et al. Human HOX proteins use diverse and context-dependent motifs to interact with TALE class cofactors. Cell Rep. 2018;22(11):3058-3071.

39. Thirant C, Ignacimouttou C, Lopez CK, et al. ETO2-GLIS2 hijacks transcriptional complexes to drive cellular identity and selfrenewal in pediatric acute megakaryoblastic leukemia. Cancer Cell. 2017;31(3):452-465.

40. Huang BJ, Smith JL, Wang YC, et al. CBFB-MYH11 fusion transcripts

distinguish acute myeloid leukemias with distinct molecular landscapes and outcomes. Blood Adv. 2021;5(23):4963-4968.

41. Ivanochko D, Halabelian L, Henderson E, et al. Direct interaction between the PRDM3 and PRDM16 tumor suppressors and the NuRD chromatin remodeling complex. Nucleic Acids Res. 2019;47(3):1225-1238.

42. Heikamp EB, Henrich JA, Perner F, et al. The Menin-MLL1 interaction is a molecular dependency in NUP98-rearranged AML. Blood. 2022;139(6):294-906.

43. Rasouli M, Szoltysek K, Cameron R, et al. NUP98/NSD1-positive AML is addicted to functional Menin-MLL interaction. EHA Library. 06/09/21;325136;EP382.

44. Bolouri H, Farrar JE, Triche T, Jr., et al. The molecular landscape of pediatric acute myeloid leukemia reveals recurrent structural alterations and age-specific mutational interactions. Nat Med. 2018;24(1):103-112.

Impact of trisomy 19 on outcome according to genetic makeup in patients with acute myeloid leukemia

Sabine Kayser,1,2,3 David Martínez-Cuadrón,4,5 Rebeca Rodriguez-Veiga,4 Mathias Hänel,6 Mar Tormo,7 Kerstin Schäfer-Eckart,8 Carmen Botella,9 Friedrich Stölzel,10 Teresa Bernal del Castillo,11 Ulrich Keller,12 Carlos Rodriguez-Medina,13 Gerhard Held,14 Maria-Luz Amigo,15 Christoph Schliemann,16 Mercedes Colorado,17 Martin Kaufmann,18 Manuel Barrios Garcia,19 Stefan W. Krause,20 Martin Görner,21 Edgar Jost,22 Björn Steffen,23 Sven Zukunft,10 Uwe Platzbecker,3 Anthony D. Ho,24 Claudia D. Baldus,25 Hubert Serve,23 Carsten Müller-Tidow,24 Christian Thiede,10 Martin Bornhäuser,10 Pau Montesinos,4,5 Christoph Röllig10 and Richard F. Schlenk,2,24,26

1Institute of Transfusion Medicine and Immunology, Medical Faculty Mannheim, Heidelberg University, German Red Cross Blood Service Baden-Württemberg-Hessen, Mannheim, Germany; 2NCT Trial Center, National Center of Tumor Diseases, German Cancer Research Center (DKFZ), Heidelberg, Germany; 3Medical Clinic and Policlinic I, Hematology and Cellular Therapy, University Hospital Leipzig, Leipzig, Germany; 4Hematology Department, Hospital Universitari i Politècnic, La Fe, València, Spain; 5CIBERONC, Instituto Carlos III, Madrid, Spain; 6Klinikum Chemnitz, Chemnitz, Germany; 7Hematology Department, Hospital Clínico Universitario, INCLIVA Research Institute, University of Valencia, Valencia, Spain; 8Hospital Nord, Nurnberg, Germany; 9Hospital General, Alicante, Spain; 10Department of Medicine I, University Hospital Carl-Gustav-Carus, Dresden, Germany, Department of Stem Cell Transplantation and Cellular Immunotherapy, Department of Internal Medicine II, University Hospital Schleswig-Holstein, Kiel University, Kiel, Germany;

11Hospital Central de Asturias, Oviedo, Spain; 12Department of Hematology, Oncology and Cancer Immunology, Charité-University Medical Center, Campus Benjamin Franklin, Berlin, Germany; 13Hematology Department, Hospital Universitario de Gran Canaria Doctor Negrín, Las Palmas de Gran Canaria, Spain; 14Westpfalz Klinikum, Kaiserslautern, Germany; 15Hospital General Universitario Morales Meseguer, Murcia, Spain; 16University Hospital Muenster, Muenster, Germany; 17Hospital Universitario Marqués de Valdecilla, Santander, Spain; 18Robert Bosch Hospital Stuttgart, Stuttgart, Germany; 19Department of Hematology, Hospital Regional Universitario de Málaga, Málaga, Spain; 20Department of Internal Medicine 5, Hematology/Oncology, University Hospital of Erlangen, Erlangen, Germany; 21Klinik für Hämatologie, Onkologie und Palliativmedizin, Klinikum Bielefeld Mitte, Bielefeld, Germany; 22Department of Hematology, Oncology, Hemostaseology, and Stem Cell Transplantation, Faculty of Medicine, University Hospital RWTH Aachen, Aachen, Germany; 23Department of Internal Medicine II, University Hospital of Frankfurt Main, Frankfurt Main, Germany; 24Department of Internal Medicine V, Heidelberg University Hospital, Heidelberg, Germany; 25Department of Internal Medicine II, University Hospital of Kiel, Kiel, Germany and 26Department of Medical Oncology, National Center for Tumor Diseases (NCT), Heidelberg University Hospital, Heidelberg, Germany

Abstract

Correspondence: S. Kayser

sabine.kayser@medma.uni-heidelberg.de

Received: October 28, 2022.

Accepted: February 14, 2023.

Early view: February 23, 2023.

https://doi.org/10.3324/haematol.2022.282127

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

We retrospectively studied 97 acute myeloid leukemia patients with trisomy 19 (median age at diagnosis 57 years; range, 1783 years) treated between 2001 and 2019 within two multicenter study groups. Trisomy 19 occurred alone in ten (10.5%) patients, with additional abnormalities being present in non-complex karyotypes in eight (8%) patients and in complex karyotypes in 79 (82%) patients. Altogether, karyotypes characterized by trisomies only were present in 27 (28%) patients. Data on response and outcome of intensively treated patients were available for 92 cases. The median follow-up was 6.4 years (95% confidence interval [95% CI]: 2.9-9.0 years). The complete remission (CR) rate after induction therapy was 52% (48 patients); the early death rate was 10% (n=9). Notably, patients with trisomy 19 as the sole abnormality had a CR rate of 89%. Allogeneic hematopoietic stem cell transplantation (allo-HCT) was performed in 34 (35%) patients (CR, n=19; active disease, n=15). Five-year relapse-free and overall survival rates were 26% (95% CI: 16-43%) and 20% (95% CI: 13-31%), respectively.

Overall survival rates were significantly higher in patients with trisomy 19 as the sole abnormality or within karyotypes characterized by trisomies only (P=0.05). An Andersen-Gill model including allo-HCT as a time-dependent covariable on overall survival revealed that trisomy 19 as the sole abnormality or within karyotypes characterized by trisomies only was a favorable factor (hazard ratio [HR]=0.47; P=0.021); higher age at diagnosis had an adverse impact (10 years difference; HR=1.29; P=0.002), whereas allo-HCT did not have a beneficial impact (odds ratio=1.45; P=0.21). In our cohort, patients with trisomy 19 as the sole abnormality or within karyotypes characterized by trisomies only had a high CR rate and better clinical outcome. Haematologica

Introduction

Trisomy 19 is a recurrent but very rare cytogenetic abnormality reported in patients with acute myeloid leukemia (AML).1 In a large analysis of 5,876 younger adult AML patients treated in United Kingdom Medical Research Council (MRC) trials, only 58 (1%) harbored a trisomy 19.1 The prognostic significance of this abnormality in AML patients is not clear. Informed clinical decision-making in situations in which cytogenetic analysis shows rare cytogenetic abnormalities has been hampered by a lack of consensus regarding the likely outcome of such patients. According to National Comprehensive Cancer Network guidelines2 as well as European LeukemiaNet recommendations3 AML patients with trisomy 19 in the absence of other abnormalities would be assigned to the intermediate-risk group. However, this risk group comprises a rather large, heterogeneous set of abnormalities, leaving the impact of trisomy 19 on outcome unclear. Apart from the benefit of achieving greater consensus in cytogenetic classification, establishing the outcome associated with rare cytogenetic abnormalities is important, particularly given the results of a meta-analysis suggesting a relapse risk in excess of 35% can provide a useful working threshold to identify patients in whom allogeneic hematopoietic stem cell transplantation (allo-HCT) may confer a survival benefit.4

Outcome seems to be poor as compared to that of patients with normal cytogenetics with a 10-year overall survival (OS) rate of 12% versus 38% ( P <0.001) and 10year cumulative incidence of relapse rate of 74% versus 49% ( P <0.001).1 Allo-HCT may improve survival if performed early in first CR. However, neither prospective clinical nor larger retrospective cohort studies are available to support these results.

Methods

Patients and treatment

Information on 97 adult patients with trisomy 19 AML diagnosed between 2001 and 2019 (2001-2010, n=40; after 2010, n=57) was collected within a large, multicenter international cohort (Study Alliance Leukemia [SAL], n=53; Programa Español de Tratamientos en Hematología [PETHEMA], n=44). Detailed case report forms, including information on baseline characteristics, chemotherapy, allo-HCT, response, and survival, were collected from all participating centers. Inclusion criteria were adult AML patients with trisomy 19 and all patients who fulfilled these criteria were included by the participating groups/institutions. The diagnosis of AML was based on French-American-British Cooperative Group criteria, 5 and, after 2003, on revised International Working Group

criteria.6 Chromosome banding was performed using standard techniques, and karyotypes were described according to the International System for Human Cytogenetic Nomenclature.7 A complex karyotype was defined according to the 2017 European LeukemiaNet classification.3 FLT3 mutation screening for internal tandem duplications (ITD) and point mutations within the tyrosine kinase domain (TKD) was carried out at each institution as previously described. 8,9 Data collection and analysis were approved by the institutional review boards of the participating centers.

Treatment

Ninety-three (96%) of the 97 patients received intensive induction treatment either within clinical trials (n=22) or according to local institutional standards (n=71). Treatment protocols for patients treated within the SAL (n=53) included AML60+ (n=2),10 AML96 (n=9)11, and AML2003 (n=11).12 Additionally, 31 patients were included within the prospective SAL registry (NCT03188874). All patients from PETHEMA (n=44) were included within the PETHEMA AML registry (NCT02607059).13

Induction therapy of the 71 patients treated according to local institutional standards consisted of the anthracycline/cytarabine based “7+3” regimen (n=50) or comparable intensive treatment (n=21).

Four (4%) of the 97 patients were treated non-intensively. Of those, one received low-dose cytarabine, one decitabine, one venetoclax in combination with azacitidine14 and one patient volasertib or placebo in combination with low-dose cytarabine.15 Response was assessed according to International Working Group recommendations.6 All clinical studies were approved by the institutional review boards of the participating centers. All patients provided written informed consent to participation in one of the treatment trials or to therapy according to local standards.

Statistical analyses

Survival endpoints including OS and relapse-free survival were defined according to the revised recommendations of the International Working Group.6 Comparisons of patients’ characteristics were performed with the KruskalWallis rank sum test for continuous variables and Fisher exact test for categorical variables. To identify prognostic variables with respect to response to induction therapy a logistic regression model was used. Variables included white blood cell count, age, gender, complex karyotype, as well as trisomy 19 as the sole abnormality or trisomy 19 in combination with trisomies only (i.e., trisomy 19 and one additional trisomy as well as trisomy 19 and additional trisomies only within a complex karyotype). The median follow-up time was computed using the reverse Kaplan-Meier estimate.16 The Kaplan-Meier method was

used to estimate the distributions of relapse-free survival and OS.17 Confidence interval (CI) estimations for survival curves were based on the cumulative hazard function using the Greenwood formula for variance estimation. Log-rank tests were employed to compare survival curves between groups. The effect of allo-HCT (including all transplanted patients) on OS as a time-dependent intervening event was tested in a multivariable Andersen-Gill model.18 Variables included in the model were trisomy 19 as the sole abnormality or trisomy 19 in combination with trisomies only, age with a difference of 10 years as well as allo-HCT. All statistical analyses were performed with the statistical software environment R, version 4.2.1, using the R packages rms, version 6.3-0, and survival, version 3.4-0.19

Results

Study cohort

Overall demographic and clinical data were collected from 97 patients diagnosed with AML and trisomy 19 between 2001 and 2019. The median age was 57 years (range, 17-83 years) and 35 patients (36%) were female. The patients’ baseline characteristics are summarized in Table 1.

Cytogenetic and molecular analyses

Cytogenetic analysis revealed trisomy 19 as the sole abnormality in ten (10.5%) patients, with additional abnormalities in a non-complex karyotype in eight (8%) patients, and in a complex karyotype in 79 (82%) patients. In patients with trisomy 19, the most frequent additional abnormality was trisomy 8 (n=46, 47%). Of those, most had a complex karyotype (n=42, 91%). Trisomy 19 and one additional trisomy occurred in six (6%) patients (Table 2). In addition, trisomy 19 and additional trisomies only within a complex karyotype occurred in 11 (11.5%) patients. Altogether, karyotypes characterized by trisomy 19 as the sole abnormality or trisomy 19 and additional trisomies only were found in 27 (28%) patients (Figure 1).

There was only one case of gene fusion (inversion 16) being present besides trisomy 19. A total of 65 patients (67%) underwent mutation testing for NPM1 and FLT3-ITD. Of those, three (5%) and one (1.5%) harbored NPM1 and FLT3ITD mutations, respectively. None of 12 patients with available data had a FLT3-TKD mutation. Four (8%) of 51 analyzed patients were CEBPA double-mutated (Table 1).

Response to induction therapy

Four patients were treated with non-intensive therapies because of higher age (median, 66.9 years; range, 55-70.6 years) or comorbidities. Of those, only one achieved a CR after one cycle of venetoclax/azacitidine treatment. The patient received a second cycle of venetoclax/azacitidine and

went on to allo-HCT. Unfortunately, he relapsed 77 days later and succumbed to his disease 73 days later. Cytogenetically, the patient showed a trisomy 19 and additional trisomies only within a complex karyotype. Molecularly, NPM1 and FLT3-ITD were unmutated. All three other patients treated with non-intensive therapy did not respond and died within 1.5 years (median, 11.6 months; range, 2.8-18 months). Two of them had trisomy 19 within a complex karyotype (consisting of other abnormalities than trisomies only) and one patient had trisomy 19 as the sole abnormality. Data on response to intensive induction therapy were available for 92 patients (data were missing for 1 patient). CR after induction therapy was achieved by 48 (52%). Early death occurred in nine (10%) patients. Notably, patients with trisomy 19 as the sole abnormality had a CR rate of 89% (n=8/9); the patients with trisomy 19 and one additional trisomy as well as trisomy 19 and additional trisomies only had a CR rate of 73% (n=11/15) as compared to 43% (n=29/67) in patients with trisomy 19 within a

Results may not add up to 100 due to rounding. *Available for 65 (67%) patients; **available for 51 (53%) patients. AML: acute myeloid leukemia; s-AML: AML after previous myelodysplastic syndrome/myeloproliferative neoplasm; t-AML: therapy-related AML; NPM1: nucleophosmin 1; FLT3: fms-related tyrosine kinase 3; ITD: internal tandem duplication; CEBPA: CCAAT enhancer binding protein A; ELN: European LeukemiaNet; WBC: white blood cell; BM: bone marrow.

Table 1. Baseline characteristics of patients with acute myeloid leukemia and trisomy 19.

complex karyotype (consisting of other abnormalities than trisomies only). A logistic regression model with limited backward selection on response to induction therapy revealed trisomy 19 as the sole abnormality or within a karyotype characterized by trisomies only as favorable factors (odds ratio [OR]=5.55; P=0.005), whereas higher age at diagnosis had an adverse impact (10 years difference; OR=0.58; P=0.002).

Further therapy including intensive consolidation and allogeneic hematopoietic stem cell transplantation

Twenty-nine (60%) of 48 intensively treated patients in first CR received intensive chemotherapy consolidation consisting of high-dose cytarabine as a single agent or combined with additional cytostatic agents. Nineteen (40%) patients proceeded to allo-HCT in first CR with ten of the transplanted patients receiving consolidation chemotherapy prior to transplantation. Patients proceeding to allo-HCT were significantly younger (P=0.001) than patients receiving intensive consolidation with chemotherapy, whereas there was no difference in median white blood cell count (P=0.89) and European LeukemiaNet risk classification (P=0.37) between the two groups.

Among patients consolidated with chemotherapy, relapses occurred in 15 and three experienced treatment-related mortality after consolidation. In patients consolidated with allo-HCT in first CR, five patients relapsed and most of them died shortly thereafter (median, 1.5 months; range, 0.3-14.9 months); additionally, five patients died of transplant-related causes.

Characteristics of patients undergoing allogeneic hematopoietic stem cell transplantation

Allo-HCT was performed overall in 34 (35%) patients, of

N Karyotype

1 48,XY,+8,+19[19]/46,XY[1]

2 47,XX,+8[1]/48,XX,+8,+19[1]/46,XX[5] (+8 and +19 confirmed by FISH)

3 48,XY,+8,+19[3]/46,XY[22]

4 48,XX,+19,+21[2]/46,XX[23]

5 48,XY,+8,+19

6 48,XY,+19,+21

7 52,XY,+4,+4,+9,+10,+19,+22[6]/46,XY[18]

8 47,XY,+6[1]/48,XY,+4,+6[15]/48,XY,+6,+19[3]/46,XY[6] 9 47,XY,+19[1]/48,sl,+16[6]/49,sdl,+8[17]/46,XY[1] (+19 confirmed by FISH) 10 47,XY,+8[16]/58,XY,+Y,+2,+8,+8,+8,+8,+13,+15,+15,+19, +20,+22[4]/46,XY[2]

11 48,XY,+14,+19[3]/48;inc[1]/69,XXY[1]/46,XY[3] 12 54,XXY,+8,+10,+11,+12,+19,+21,+22[3]/46,XY[17]

49,XY,+8,+19,+20[4]/46,XY[16]

14 56-59,XY,+X,+1,+1,+6,+11,+12,+13,+14, +15,+19,+21,+21,+22[20]

15 49,XX,+15,+19,+21

16

98,XY,+X,+Y,+Y,+1,+1,+3,+3,+4,+4,+5,+5,+6,+6,+7,+7,+8,+8,+9, +10,+10,+11,+12,+12,+13,+14,+14,+15,+15,+15,+16,+17,+17, +18,+18,+18,+19,+19,+20,+20,+21,+21,+22,+22, +22[40]/46,XY[21]

17 52,XY,+X,+8,+13,+14,+17,+19[17]/46,XY[3]

Numbers 1-6: trisomy 19 and one additional trisomy. Numbers 7-17: trisomy 19 and additional trisomies only within a complex karyotype. FISH: fluorescence in-situ hybridization.

Figure 1. Cytogenetic abnormalities of 97 patients with trisomy 19 showing the categorization of cytogenetic abnormalities used for further analysis.

Table 2. Karyotypes characterized by trisomy 19 and additional trisomies only.

whom 19 patients were transplanted in first CR. Fifteen patients underwent allo-HCT with active disease. Fourteen patients received myeloablative conditioning and 18 patients reduced-intensity conditioning (missing data, n=2). The type of donor was matched related in 12 cases and matched unrelated in 22 patients.

Relapse-free and overall survival

The median follow-up of the cohort was 6.4 years (95% CI: 2.9-9.0 years). The median survival of non-intensively treated patients (n=4) was 0.8 years and none of these patients sur-

vived beyond 1.5 years. In intensively treated patients the 5year relapse-free survival and OS rates were 26% (95% CI: 16-43%) and 20% (95% CI: 13-31%), respectively.

OS rates were significantly higher in patients with trisomy 19 as the sole abnormality or within a karyotype characterized by trisomies only as compared to all other abnormalities) (P=0.05) (Figure 2). An Andersen-Gill model including allo-HCT whenever performed as a time-dependent covariable revealed that trisomy 19 as the sole abnormality or within a karyotype characterized by trisomies only (hazard ratio [HR]=0.47; P=0.021) was a favorable factor, whereas age with a difference of 10 years (HR=1.29; P=0.002) had a negative impact. Allo-HCT (OR=1.45; P=0.21) did not have a significant impact.

The influence of allo-HCT, assessed in univariable analysis as a time-dependent co-variable, as post remission therapy on OS is also illustrated by a Simon Makuch plot (P=0.60) (Figure 3). There was a trend towards better OS measured from the date of allo-HCT if patients proceeded to allo-HCT in first CR (n=19) rather than being transplanted with active disease (n=15; P=0.10) (Figure 4). In patients achieving first CR, 5-year relapse-free survival was 35% (95% CI: 18-70%) for those patients proceeding to allo-HCT (n=19) as compared to 15% (95% CI: 3-80%) in those who received consolidation chemotherapy.

Discussion

The focus of our study was to characterize adult AML patients with trisomy 19 in an international, multicenter cohort study and compare outcomes according to treatment

Figure 2. Kaplan-Meier plot of overall survival in intensively treated patients according to cytogenetic abnormality. Tris: trisomy; abn: abnormalities.

Figure 3. Simon Makuch plot of overall survival illustrating the impact of allogeneic hematopoietic stem cell transplantation evaluated as a time-dependent event. Allo-HCT: allogeneic hematopoietic stem cell transplantation.

Figure 4. Overall survival after allogeneic hematopoietic stem cell transplantation according to remission status. Allo-HCT: allogeneic hematopoietic stem cell transplantation. CR1: first complete remission.

strategies, with a specific focus on the impact of allo-HCT as compared to conventional chemotherapy on survival.

Trisomy 19 seems to be a very rare abnormality,1 and further large cohort studies on this abnormality are not available. Thus, its influence on outcome remains unclear. We here present the largest cohort of 97 patients with trisomy 19 to date, of whom 93 were treated intensively.

Trisomy 19 is frequently associated with complex karyotypes and, interestingly, with karyotypes characterized by trisomies only, but seems to be very infrequent in patients with gene fusions or disease-defining mutations.

Contrary to published data by the MRC reporting a high CR rate of 81% after intensive induction therapy,1 we observed a high CR rate only in patients with trisomy 19 as the sole abnormality or within a karyotype characterized by trisomies only (that is, trisomy 19 and one additional trisomy as well as trisomy 19 and additional trisomies only within a complex karyotype). Currently, the prognostic impact of a complex karyotype, defined by trisomies only, also called a hyperdiploid karyotype, has not been addressed conclusively.20-22 The prognosis of AML patients with trisomies only, particularly of those with a hyperdiploid karyotype, seems to be poor as compared to that of patients with intermediate-risk cytogenetics.22 However, in our analysis patients with trisomy 19 as their sole abnormality or within a karyotype characterized by trisomies only, including a hyperdiploid karyotype, had a high CR rate and better clinical outcome.

Nevertheless, most patients relapsed, which seems not to be largely improved by allo-HCT, suggesting that other treatment approaches are needed to prolong survival. The only factor associated with prolonged survival was trisomy 19 as the sole abnormality or within a karyotype characterized by trisomies only.

Treatment with venetoclax in combination with hypomethylating agents seems to be promising, particularly in patients with NPM1- or IDH-mutated AML.23,24 Responses do also occur in patients with adverse genetic features, such as high-risk cytogenetic abnormalities as well as TP53 mutations.23,24 However, responses in such patients have been mostly short-lived and not durable. Whether other combinations, such as venetoclax/azacitidine and magrolimab, will lead to higher rates of remission and prolong survival is currently being evaluated within a phase III randomized trial (NCT05079230). In addition, venetoclax in combination with standard intensive chemotherapy is now being studied as frontline therapy in younger and older patients with AML (e.g., trials NCT03709758 and NCT04628026). Preliminary data suggest a very high overall response rate of 100% (n=10), with 75% (n=6/8) of the patients achieving minimal residual disease-negative remissions assessed using multiparameter flow cytometry.25

An isolated trisomy 19 can be detected in myeloid disorders,26 and there seems to be an association with mega-

karyoblastic leukemia (M7).27,28 In our cohort, we detected secondary chromosomal abnormalities in 90%, most frequently trisomy 19 within a complex karyotype or trisomy 19 in combination with trisomy 8. Regarding the molecular makeup, the frequencies of mutations in NPM1, FLT3-ITD or CEBPA as well as disease-defining fusion genes were very low, suggesting that other cooperating mutations may play a role in the pathogenesis. To date, however, the pathogenic role of trisomy 19 per se in leukemogenesis is still unclear.

Of note, allo-HCT, performed in first CR or with active disease did not result in improved outcome. This is in line with our recent findings in a cohort of AML patients characterized by trisomy 4.29 Nonetheless, this is in contrast to previous reports30-32 focusing on other trisomies such as +8, +11, +13 and +21 and may indicate that the outcome of patients with trisomies needs to be evaluated individually. However, we would like to emphasize that retrospectively collected data have serious limitations since the factors for allocating patients to allo-HCT, such as comorbidities, individual assessment of the treating physician, choice of conditioning, and availability of a donor, remain unknown and these need to be taken into account when evaluating the value of allo-HCT in our series. Furthermore, the presence of minimal residual disease, as detected by polymerase chain reaction, flow cytometry, or next-generation sequencing, was shown to be a predictive factor for outcome.33-36 Thus, minimal residual disease status is being increasingly used to allocate patients for transplantation.35 However, in our cohort of patients data on minimal residual disease status were not available. Thus, we could not separate those patients who went on to allo-HCT in first CR according to their minimal residual disease status.

In conclusion, patients with trisomy 19 are very heterogeneous, in particular with respect to genetic abnormalities. In our cohort, patients with trisomy 19 as their sole abnormality or within a karyotype characterized by trisomies only had a high CR rate and better clinical outcome. In the total cohort, it seems that allo-HCT may not improve OS; nevertheless, shortcomings of retrospective cohort studies need to be taken into account.

Disclosures

No conflicts of interest to disclose.

Contributions

SK and RFS were responsible for the concept of this paper, contributed to the literature search and data collection, analyzed and interpreted data, and wrote the manuscript. DM-C, RR-V, MH, MT, KS-E, CB, FS, TBdC, UK, CR-M, GH, MLA, CS, MC, MK, MBG, SWK, MG, EJ, BS, SZ, UP, ADH, CDB, HS, CM-T, MB, PM, and CR contributed patients and critically revised the manuscript. CTh performed research and

critically revised the manuscript. All authors reviewed and approved the final manuscript.

Funding

For the publication fee we acknowledge financial support from the Deutsche Forschungsgemeinschaft within the

References

1. Grimwade D, Hills RK, Moorman AV, et al. Refinement of cytogenetic classification in acute myeloid leukemia: determination of prognostic significance of rare recurring chromosomal abnormalities among 5876 younger adult patients treated in the United Kingdom Medical Research Council trials. Blood. 2010;116(3):354-365.

2. O'Donnell MR, Tallman MS, Abboud CN, et al. Acute myeloid leukemia, version 3.2017, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw. 2017;15(7):926-957.

3. Döhner H, Estey E, Grimwade D, et al. Diagnosis and management of AML in adults: 2017 ELN recommendations from an international expert panel. Blood. 2017;129(4):424-447.

4. Cornelissen JJ, van Putten WLJ, Verdonck LF, et al. Results of a HOVON/SAKK donor versus no-donor analysis of myeloablative HLA-identical sibling stem cell transplantation in first remission acute myeloid leukemia in young and middle-aged adults: benefits for whom? Blood. 2007;109(9):3658-3666.

5. Bennett JM, Catovsky D, Daniel MT, et al. Proposed revised criteria for the classification of acute myeloid leukemia. A report of the French-American-British Cooperative Group. Ann Intern Med. 1985;103(4):620-625.

6. Cheson BD, Bennett JM, Kopecky KJ, et al. Revised recommendations of the International Working Group for Diagnosis, Standardization of Response Criteria, Treatment Outcomes, and Reporting Standards for Therapeutic Trials in Acute Myeloid Leukemia. J Clin Oncol. 2003;21(24):4642-4649.

7. Mitelman F: ISCN: An International System for Human Cytogenetic Nomenclature. Basel, Switzerland: S. Karger; 1995.

8. Yokota S, Kiyoi H, Nakao M, et al. Internal tandem duplication of the FLT3 gene is preferentially seen in acute myeloid leukemia and myelodysplastic syndrome among various hematological malignancies. A study on a large series of patients and cell lines. Leukemia. 1997;11(10):1605-1609.

9. Thiede C, Steudel C, Mohr B, et al. Analysis of FLT3-activating mutations in 979 patients with acute myelogenous leukemia: association with FAB subtypes and identification of subgroups with poor prognosis. Blood. 2002;99(12):4326-4335.

10. Röllig C, Kramer M, Gabrecht M, et al. Intermediate-dose cytarabine plus mitoxantrone versus standard-dose cytarabine plus daunorubicin for acute myeloid leukemia in elderly patients. Ann Oncol. 2018;29(4):973-978.

11. Röllig C, Thiede C, Gramatzki M, et al. A novel prognostic model in elderly patients with acute myeloid leukemia: results of 909 patients entered into the prospective AML96 trial. Blood. 2010;116(6):971-997.

12. Schaich M, Parmentier S, Kramer M, et al. High-dose cytarabine consolidation with or without additional amsacrine and mitoxantrone in acute myeloid leukemia: results of the prospective randomized AML2003 trial. J Clin Oncol. 2013;31(17):2094-2102.

13. Paiva B, Vidriales MB, Sempere A, et al. Impact of measurable residual disease by decentralized flow cytometry: a PETHEMA

funding program “Open Access Publikationskosten“, as well as from Heidelberg University.

Data-sharing statement

Questions regarding data sharing should be addressed to the corresponding author.

real-world study in 1076 patients with acute myeloid leukemia. Leukemia. 2021;35(8):2358-2370.

14. DiNardo CD, Pratz K, Pullarkat V, et al. Venetoclax combined with decitabine or azacitidine in treatment-naive, elderly patients with acute myeloid leukemia. Blood. 2019;133(1):7-17.

15. Döhner H, Symeonidis A, Deeren D, et al. Adjunctive volasertib in patients with acute myeloid leukemia not eligible for standard induction therapy: a randomized, phase 3 trial. Hemasphere. 2021;5(8):e617.

16. Schemper M, Smith TL. A note on quantifying follow-up in studies of failure time. Control Clin Trials. 1996;17(4):343-346.

17. Kaplan E, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53(282):457-481.

18. Andersen P, Gill RD. Cox’s regression model for counting processes: a large sample study. Ann Stat. 1982;10(4):1100-1120.

19. R Development Core Team. R: a Language and Environment for Statistical Computing. R Foundation for Statistical Computing. Vienna, Austria, 2014. https://www.R-project.org/ Accessed 2023 March 15.

20. Chilton L, Hills RK, Harrison CJ, et al. Hyperdiploidy with 49-65 chromosomes represents a heterogeneous cytogenetic subgroup of acute myeloid leukemia with differential outcome. Leukemia. 2014;28(2):321-328.

21. Luquet I, Lai JL, Barin C, et al. Hyperdiploid karyotypes in acute myeloid leukemia define a novel entity: a study of 38 patients from the Groupe Francophone de Cytogenetique Hematologique (GFCH). Leukemia. 2008;22(1):132-137.

22. Stölzel F, Mohr B, Kramer M, et al. Karyotype complexity and prognosis in acute myeloid leukemia. Blood Cancer J. 2016;6(1):e386.

23. DiNardo CD, Jonas BA, Pullarkat V, et al. Azacitidine and venetoclax in previously untreated acute myeloid leukemia. N Engl J Med. 2020;383(7):617-629.

24. DiNardo CD, Tiong IS, Quaglieri A, et al. Molecular patterns of response and treatment failure after frontline venetoclax combinations in older patients with AML. Blood. 2020;135(11):791-803.

25. Stone RM, DeAngelo DJ, Letai AG, et al. Maximal tolerated dose of the BCL-2 inhibitor venetoclax in combination with daunorubicin/cytarabine induction in previously untreated adults with acute myeloid leukemia (AML). Blood. 2020;136(Suppl 1):40-41.

26. Jung SI, Cho HS, Lee CH, et al. Two cases of trisomy 19 as a sole chromosomal abnormality in myeloid disorders. Korean J Lab Med. 2008;28(3):174-178.

27. Nimer SD, MacGrogan D, Jhanwar S, Alvarez S. Chromosome 19 abnormalities are commonly seen in AML, M7. Blood. 2002;100(10):3838.

28. Alvarez S, MacGrogan D, Calasanz MJ, Nimer SD, Jhanwar SC. Frequent gain of chromosome 19 in megakaryoblastic leukemias detected by comparative genomic hybridization. Genes Chromosom Cancer. 2001;32(3):285-293.

29. Kayser S, Martínez-Cuadrón D, Hanoun M, et al. Characteristics and outcome of patients with acute myeloid leukemia and trisomy 4. Haematologica. 2023;108(1):34-41.

30. Farag SS, Archer KJ, Mrózek K, et al. Isolated trisomy of chromosomes 8, 11, 13 and 21 is an adverse prognostic factor in adults with de novo acute myeloid leukemia: results from Cancer and Leukemia Group B 8461. Int J Oncol. 2002;21(5):10411051.

31. Schaich M, Schlenk RF, Al-Ali HK, et al. Prognosis of acute myeloid leukemia patients up to 60 years of age exhibiting trisomy 8 within a non-complex karyotype: individual patient data-based meta-analysis of the German Acute Myeloid Leukemia Intergroup. Haematologica. 2007;92(6):763-770.

32. Chevallier P, Labopin M, Nagler A, et al. Outcome after allogeneic transplantation for adult acute myeloid leukemia patients exhibiting isolated or associated trisomy 8

chromosomal abnormality: a survey on behalf of the ALWP of the EBMT Bone Marrow Transplant. 2009;44(9):589-594.

33. Ngai LL, Kelder A, Janssen JJWM, Ossenkoppele GJ, Cloos J. MRD tailored therapy in AML: what we have learned so far. Front Oncol. 2021;10:603636.

34. Heuser M, Freeman SD, Ossenkoppele GJ, et al. 2021 update on MRD in acute myeloid leukemia: a consensus document from the European LeukemiaNet MRD Working Party. Blood. 2021;138(26):2753-2767.

35. Craddock C, Raghavan M. Which patients with acute myeloid leukemia in CR1 can be spared an allogeneic transplant? Curr Opin Hematol. 2019;26(2):58-64.

36. Short NJ, Zhou S, Fu C, et al. Association of measurable residual disease with survival outcomes in patients with acute myeloid leukemia: a systematic review and meta-analysis. JAMA Oncol. 2020;6(12):1890-1899.

Hemophagocytic lymphohistiocytosis and disseminated intravascular coagulation are underestimated, but fatal adverse events in chimeric antigen receptor T-cell therapy

Zhiqiang Song,1* Dingyuan Tu,2* Gusheng Tang,1* Na Liu,1 Zongguang Tai,3 Jianmin Yang1 and Yang Wang1

1Department of Hematology, Institute of Hematology, Shanghai Changhai Hospital, Naval Medical University; 2Department of Cardiology, Shanghai Changhai Hospital, Naval Medical University and 3Shanghai Skin Disease Hospital, School of Medicine, Tongji University, Shanghai 200443, China

*ZS, DT, and GT contributed equally as first authors.

Abstract

Correspondence: Zongguang Tai taizongguang@126.com

Jianmin Yang chyangjianmin@163.com

Yang Wang yang060124@126.com

Received: May 25, 2022.

Accepted: February 3, 2023.

Early view: February 16, 2023.

https://doi.org/10.3324/haematol.2022.281455

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Hematotoxicity is the most common long-term adverse event (AE) after chimeric antigen receptor T-cell (CAR T) therapy. However, patients who receive CAR T therapy in pivotal clinical trials are subjected to restrictive selection criteria, and this means that rare but fatal toxicities are underestimated. Here, we systematically analyzed CAR T-associated hematologic AE using the US Food and Drug Administration Adverse Event Reporting System (FAERS) between January 2017 and December 2021. Disproportionality analyses were performed using reporting odds ratios (ROR) and information component (IC); the lower limit of the ROR and IC 95% confidence interval (CI) (ROR025 and IC025) exceeding one and zero was considered significant, respectively. Among the 105,087,611 reports in FAERS, 5,112 CAR T-related hematotoxicity reports were identified. We found 23 significant over-reporting hematologic AE (ROR025 >1) compared to the full database, of which hemophagocytic lymphohistiocytosis (HLH; n=136 [2.7%], ROR025 = 21.06), coagulopathy (n=128 [2.5%], ROR025 = 10.43), bone marrow failure (n=112 [2.2%], ROR025 = 4.88), disseminated intravascular coagulation (DIC; n=99 [1.9%], ROR025 = 9.64), and B-cell aplasia (n=98 [1.9%], ROR025 = 118.16, all IC025 > 0) were highly under-reported AE in clinical trials. Importantly, HLH and DIC led to mortality rates of 69.9% and 59.6%, respectively. Lastly, hematotoxicity-related mortality was 41.43%, and 22 death-related hematologic AE were identified using LASSO regression analysis. These findings could help clinicians in the early detection of those rarely reported but lethal hematologic AE, thus reducing the risk of severe toxicities for CAR T recipients.

Introduction

CD19 chimeric antigen receptor T-cell (CAR T) therapy is a promising treatment for patients with relapsed or refractory (r/r) B-cell acute lymphoblastic leukemia (B-ALL) and large B-cell lymphoma.1,2 The CD19 CAR T products axicabtagene-ciloleucel (axicabtagene) and tisagenlecleucel have gained wide popularity due to the unprecedented treatment efficacy of hematologic malignancies. For patients with r/r diffuse large B-cell lymphoma, traditional chemotherapy resulted in a poor complete response rate

(CR, 7%) and overall survival (OS) of 2 years (20%),3 while the rate of CR and OS at 2 years exceeded 50% after axicabtagene therapy.4 Similarly, tisagenlecleucel also contributed to a remarkable CR rate (90%) and 2-year OS rate (73%) in patients with r/r B-ALL.5

However, the administration of CD19 CAR T products increases the risk of serious and fatal adverse events (AE). Cytokine release syndrome (CRS) is the most common adverse event caused by rapid activation of the immune system.6 The incidence of CRS varies from 35% to 93%, depending on the CAR T product used, the tumor burden,

and the CRS grading criteria.7 Furthermore, hematotoxicity has become the most common long-term AE after CAR T-cell infusion with severe outcomes.8 The incidence of serious anemia (grade 3) and thrombocytopenia and neutropenia after three weeks of CAR T transfusion has been reported to range from 5% to 17%, 21% to 29%, and 30% to 38%, respectively.9-12 Meanwhile, 16% patients had prolonged cytopenia for 22 months and had to receive transfusions or growth factor support after CD19 CAR Tcell therapy.13 Importantly, prolonged neutropenia increases the risk of infectious complications, which are the most common cause of non-relapse mortality (NRM).2 It is critical to comprehensively understand CAR T-associated hematologic AE, and a systematic safety profile should be evaluated before treating patients, especially given the rapid increase in the use of this revolutionary therapy. Meanwhile, it is helpful to detect and prevent some rare but lethal AE, such as hemophagocytic lymphohistiocytosis (HLH) and disseminated intravascular coagulation (DIC). However, clinical trials are subject to restrictive inclusion criteria and thus include highly selected patients,4,14 leading to the frequency of hematologic AE, especially some rare AE, being underestimated. Therefore, post-marketing AE reports could help identify the safety profile of new therapies by revealing real-world data and detecting rare but lethal toxicities.

The aim of this study was to extensively evaluate hematologic AE associated with CAR T using the US Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS) and to alert clinicians earlier to those rarely reported but severe hematologic toxicities, thus improving the management of CAR T-cell recipients and reducing the risk of lethal hematotoxicity.

Methods

Data sources and study design

This retrospective post-marketing CAR T-cell therapy safety analysis was performed on the basis of data extracted from FAERS between 1 January 2017 and 31 December 2021. In this study, only the adverse reports, including CAR T-associated trade or generic names, such as “AXICABTAGENE,” “CILOLEUCEL,” “AXICABTAGEN-CILOLEUCEL,” “AXI-CEL,” “KTE C19,” “TISAGENLECLEUCEL,” “CTL019,” “KYMRIAH,” “TISA-CEL” were selected for analysis (see Online Supplementary Table S1 for details). In particular, all AE in the FAERS database were coded and standardized with reference to the level of preferred term (PT) according to the FDA’s Medical Dictionary for Regulatory Activities (MedDRA). Therefore, the hematologic toxicities were coded according to the PT of MedDRA. All datasets in this study can be accessed at: https://www.fda.gov/regulatory-information/freedom-information.

Statistical analysis

Disproportionality analysis methods, including the reporting odds ratio (ROR) and the information component (IC), were used to detect potential hematologic AE in total or specific CAR T products.15,16 A statistical shrinkage transformation was performed to maintain more robust results,17 and the relative statistical formula is as follows:

nexp = (ndrug * nevent)/ntotal

ROR = (nobe + 0.5)/(nexp + 0.5)

IC = log2 ((nobe + 0.5)/(nexp + 0.5)) nexp and nobe are the numbers of expected and observed drug-adverse event records, respectively. ndrug is the number of records of the relevant drugs. Similarly, nevent represents the number of associated AE records. ntotal is the number of drug AE records. In particular, signal detection was only conducted for drug-AE pairs with at least three case records. A statistically significant signal was identified if the lower limit value of the ROR 95% confidence interval (ROR025) exceeded one or the lower bound value of the 95% confidence interval for the IC (IC025) was greater than zero,15,16 which means target AE or combinations were reported more frequently in the targeted drugs than in the control group.

In addition to classic signal detection at the PT level, signal analysis was also performed in AE of clinical interest based on HLT. Furthermore, we conducted a signal comparison of several variables, including products (tisagenlecleucel vs axicabtagene), sex (male vs. female), and age (age <65 vs 65 years). A log-rank test was performed to compare timeto-onset differences in various groups, such as two CAR T products, the top ten frequently reported PT, and six types of HLT. Furthermore, the proportion of deaths was calculated for different PT and HLT. We also performed a LASSO regression analysis to select statistically significant PT associated with death. Finally, the UpSet plot, Venn diagram, and charts were used to explore the potential overlap between different PT, HLT, and CRS, and provide relative indications for prompt treatment. All statistical analyzes were performed using SAS software (version 9.4, SAS Institute Inc., Cary, NC, USA) and R software (version 4.1.2).

Results

Descriptive analysis

The study design is illustrated in Figure 1. Among the 105,087,611 reports in the FAERS database between 1 January 2017 and 31 December 2021, 43,830 toxicity reports were associated with CAR T therapy (axicabtagene, n=14,464 [33.0%]; tisagenlecleucel, n=29,366 [67.0%]) (Figure 1). Among them, 5,112 reports were hematologic AE, of which 1,494 (29.2%) reports were associated with axicabtagene and 3,618 (70.8%) with tisagenlecleucel. The overall reporting rate for hematologic AE was lower in patients

receiving axicabtagene than in those receiving tisagenlecleucel (10.3% vs. 12.3%, respectively). According to (MedDRA,18 23 and 36 PT of hematologic AE are associated with axicabtagene and tisagenlecleucel, respectively. In general, reported hematologic AE consistently increased from 2017 to 2020 in CAR T recipients, with a slight decrease in 2021 (Table 1). Hematotoxicity-related mortality was 41.43% (n=2,118), of which the fatality rates were 32.46% (n=485) and 45.14% (n=1,633) after treatment with axicabtagene and tisagenlecleucel, respectively. Among various countries, the United States reported the highest number of hematologic AE (70.54%). Most of the patients who received CAR T therapy were younger than 65 years in both the axicabtagene (n = 729, 66.8%) and tisagenlecleucel (n=2,589, 84.9%) groups, but younger patients (<65 years) had a higher reporting frequency of hematologic AE compared to older patients (≥65 years, ROR=1.33, 95% confidence interval [CI:] 1.23-1.45) (Online Supplementary Table S2).

CAR T-therapy-associated hematologic

adverse events

In general, the top ten reported hematologic AE after CAR T therapy were cytopenia (n=517, 10.1%), neutropenia (n=463, 9.1%), thrombocytopenia (n=318, 6.2%), febrile neutropenia (n=273, 5.3%), pancytopenia (n=225, 4.4%), hemophagocytic lymphohistiocytosis (HLH, n=136, 2.7%), anemia (n=129, 2.5%), coagulopathy (n=128, 2.5%), bone marrow failure (n=112, 2.2%), and disseminated intravascular coagulation (DIC, n=99, 1.9%) (Table 2 and Online Supplementary Figure S1). Furthermore, we identified 23 over-reported hematologic AE (ROR025 > 1) (Table 2). For axicabtagene, neutropenia (n=221) was the most frequently over-reported hematologic AE, corresponding to ROR025 = 6.47 and IC025 = 2.64. Cytopenia (n=427) was the most reported hematologic AE after tisagenlecleucel (ROR025 = 66.05, IC025 = 5.90). The signal values of the IC025 and ROR025 distribution of the top ten reported AE are presented in Online Supplementary Figure S1. The strongest signal values of axicabtagene and tisagenlecleucel-associated hematologic AE were HLH (ROR025 = 22.77, IC025 = 4.42) and B-cell aplasia (ROR025 = 104.20, IC025 = 6.65), respectively.

To better understand hematologic AE associated with CAR T-cell therapy, hematotoxicity PT were divided into six subgroups based on MedDRA HLT, including bone marrow depression, coagulopathies and bleeding diatheses, hematologic disorders, hemolyses, spleen disorders, and other events (Online Supplementary Table S3). Bone marrow depression, coagulopathies and bleeding diatheses, and hematologic disorders were significantly correlated with axicabtagene and tisagenlecleucel (ROR025 > 1 and IC025 > 0) (Online Supplementary Table S4). However, spleen disorders and hemolyses were significantly associated only with tisagenlecleucel and axicabtagene, respectively. Hematologic disorders (ROR025 = 11.51, IC025 = 3.45), coagulopathies

and bleeding diatheses (ROR025 = 7.82, IC025 = 2.91) had the strongest signal values in AE associated with axicabtagene and tisagenlecleucel, respectively.

The differences between axicabtagene and tisagenlecleucel in hematologic adverse events

Therapy with axicabtagene and tisagenlecleucel resulted in 21 concurrent hematologic AE (Table 2). However, hematotoxicity and hemolytic anemia were only associated with axicabtagene (AE with <3 AE were not analyzed). In addition, only 15 hematologic AE were correlated with tisagen-

Figure 1. Study flowchart. From 1 January 2017 to 31 December 2021, a total of 105,087,611 adverse events were reported and 43,830 toxicity reports were associated with CAR T therapy. Among them, 5,112 reports were hematologic adverse events, and this study was performed among 38 preferred terms (PT) with more than 3 reports, of which 23 PT were related to axicabtagene and 36 related to tisagenlecleucel. CAR T: chimeric antigen receptor T cell.

lecleucel. The concurrent hematologic AE of these two CAR T products were further investigated to determine the differences between the treatment with axicabtagene and tisagenlecleucel. The reported frequency of tisagenlecleucel was significantly higher than axicabtagene for cytopenia (ROR=2.35, 95% CI: 1.87-2.96), coagulopathy (ROR=3.97, 95% CI: 2.28-6.92), and B-cell aplasia (ROR=2.51, 95% CI: 1.474.28) (Figure 2A). In contrast, the notably lower reported AE were neutropenia, febrile neutropenia, pancytopenia, HLH, bone marrow failure, hyperbilirubinemia, and neutropenic sepsis (ROR < 1) after tisagenlecleucel therapy. When comparing the hematologic AE of these two CAR T products based on HLT, we found that tisagenlecleucel had a higher reported frequency in coagulopathies and bleeding diatheses (ROR=2.20, 95% CI: 1.62-2.98) and spleen disorders (ROR=6.47, 95% CI: 1.54-27.13), but lower in bone marrow depression (ROR=0.83, 95% CI: 0.76-0.90), hematologic disorders (ROR=0.61, 95% CI: 0.45-0.84), and hemolyses (ROR=0.53, 95% CI: 0.32-0.90) compared to axicabtagene (Figure 2B). Furthermore, there was a significant difference in time from CAR T infusion to the

onset of hematologic AE between the axicabtagene and tisagenlecleucel groups (P<0.0001) (Online Supplementary Figure 2A). The time of onset of hematologic AE is faster and shorter for axicabtagene than for tisagenlecleucel. Similarly, significant differences were also observed among the top ten reported AE and six different HLT (Online Supplementary Figure S2B and C). Of note, most hematologic AE occurred within ten days following CAR T infusion.

Clinical characteristics of hematologic adverse events

To better understand the clinical characteristics of hematologic AE, we further explored the correlations between CRS, the most common toxicity after CAR T therapy, and the top ten most frequently reported hematologic AE (Figure 3A). Considerable overlap was observed between various hematologic AE and CRS, and most of the patients experienced more than one hematologic AE. The overlap between the hematotoxicity of the two CAR T products and CRS is shown in Figure 3B. Concurrent CRS and hematotoxicity were reported in 54.0% and 48.8% of axi-

interquartile range; kgs: kilograms.

Table 1. Baseline characteristics of patients with tisagenlecleucel- and axicabtagene-associated hematologic adverse events.

cabtagene and tisagenlecleucel recipients, respectively. Importantly, the rate of hematologic AE independent of CRS was approximately 49.5% (Figure 3B). The relationship between the six different HLT and CRS was further investigated, and correlation analyzes revealed that all hematologic AE were positively related to CRS (Figure 3C). Moreover, bone marrow depression, coagulopathies and bleeding diatheses, and hematologic disorders were

strongly associated with CRS, while a weaker relationship was observed in spleen disorders and hemolyses. To help clinicians detect highly lethal hematologic AE, we further calculated the fatality rates of different AE (number of death reports/AE reports) after CAR T therapy (Figure 4A and Online Supplementary Table S5). The results showed that 14 AE had a mortality rate of over 50% in tisagenlecleucel recipients, and only two AE after axi-

HLH: hemophagocytic lymphohistiocytosis; DIC: disseminated intravascular coagulation; CAR T: chimeric antigen receptor T cell; ROR: reporting odds ratio; IC: information component. Blank means not applicable.

Table 2. Signal detection of tisagenlecleucel- and axicabtagene- associated hematologic adverse events.

cabtagene therapy. However, eight AE had fewer than ten death reports among the over 14 AE reported after tisagenlecleucel therapy (Online Supplementary Table S5). Importantly, the fatality rates of the five highly underestimated hematologic AE, HLH, coagulopathy, bone marrow failure, DIC and B-cell aplasia were 69.9%, 47.7%, 27.7%, 59.6%, and 17.1%, respectively (Figure 5). Notably, death rates were higher after axicabtagene than after tisagenlecleucel in HLH (74.60% vs. 65.75%), DIC (66.67% vs. 57.33%), and coagulopathy (50% vs. 47.37%). Nevertheless, axicabtagene had a

lower death rate than tisagenlecleucel in patients with bone marrow failure (25.9% vs. 29.6%). An analysis in terms of HLT showed a higher proportion of death in spleen disorders, hemolyses, coagulopathies and bleeding diatheses, and bone marrow depression after tisagenlecleucel therapy compared to axicabtagene (Figure 4B and Online Supplementary Table S6). However, the death rate of hematologic disorders was reported to be lower in patients treated with tisagenlecleucel than in those treated with axicabtagene. Finally, LASSO regression analysis was performed to identify

Figure 2. Signal comparison between tisagenlecleucel and axicabtagene (tisagenlecleucel vs. axicabtagene) in different adverse events. (A) Signal comparison between tisagenlecleucel and axicabtagene (tisagenlecleucel vs. axicabtagene) in different preferred terms (PT). (B) Signal comparison between tisagenlecleucel and axicabtagene (tisagenlecleucel vs. axicabtagene) in different high level terms (HLT). DIC: disseminated intravascular coagulation; HLH: hemophagocytic lymphohistiocytosis.

hematologic AE that were strongly associated with death (Online Supplementary Figure S3). Twenty-two hematologic AE were identified to be closely related to death in CAR T recipients, including cytopenia, DIC, febrile neutropenia, HLH, splenic hemorrhage, and thrombocytopenia (Table 3).

Discussion

CAR T therapy has revolutionized the treatment of patients with r/r hematologic malignancies due to its high clinical efficacy compared to traditional chemotherapy. There has

Figure 3. Correlation between cytokine release syndrome and various hematologic adverse events. (A) Overlap between cytokine release syndrome (CRS) and top 10 CAR T-associated hematologic adverse events. (B) Overlap between CRS, axicabtagene- and tisagenlecle-associated hematologic adverse events. (C) Overlap between CRS and 6 high level terms (HLT) of hematologic adverse events. CAR T: chimeric antigen receptor T cell; DIC: disseminated intravascular coagulation; HLH: hemophagocytic lymphohistiocytosis; BMD: bone marrow depression; CBD: coagulopathies and bleeding diatheses; SD: spleen disorders; HD: hematologic disorders; OE: other events.

been a consistent increase in the use of CAR T therapy in recent years, and the number of CAR T therapies had increased by 24% in 2022 compared to 2021.19 With more and more people are now eligible for CAR T therapy, it is essential to comprehensively determine treatment-related AE and differences among various CAR T products to optimize this revolutionary therapy. Hematotoxicity is the most common long-term toxicity in CAR T recipients, with severe outcomes. However, clinical trials have never been sufficiently powered to detect hematologic AE because the patients enrolled in the study had been highly selected. In this study, we extensively investigated CAR T-associated hematotoxicity using FAERS, which is instrumental in the early detection and prevention of some rare but fatal AE. To the best of our knowledge, this is the largest post-marketing study of CAR T-associated hematologic AE. Of the 43,830 CAR T-associated AE reported, 5,112 (11.7%) were hematologic AE. The analysis revealed that younger patients (age <65 years) were more likely to experience hematologic AE (age <65 vs. ≥65 years, ROR025 = 1.23) (Online Supplementary Table S2), according to a meta-analysis of hematologic toxicity in CAR T recipients.20 This may be associated with a more potent immune response after CAR T infusion in younger patients. In general, the incidence of hematologic toxicity was significantly higher after infusion with tisagenlecleucel than axicabtagene (Online Supplementary Table S2). Tisagenlecleucel was approved for the treatment of patients with r/r DLBCL and r/r B-ALL, and axicabtagene only for r/r DLBCL. Therefore, the rate of cytopenia-related hematotoxicity could be higher with tisagenlecleucel than with axicabtagene. In addition, this finding may also be associated with the different costimulatory domains of these two CAR T products, CD28 and 4-1BB for axicabtagene and tisagenlecleucel, respectively.4 Co-stimulation of 4-1BB mediates longer persistence of CAR T cells than CD2821 and it is reasonable to observe more reports of hematologic AE in tisagenlecleucel, especially long-term and delayed hematologic AE, such as B-cell aplasia.

Our analysis identified the possible top ten significant hematologic AE after CAR T therapy: cytopenia, neutropenia, thrombocytopenia, febrile neutropenia, pancytopenia, HLH, coagulopathy, bone marrow failure, DIC and B-cell aplasia. (Table 2). According to pivotal clinical trials,4,14,22-24 a high number of reports of cytopenia, neutropenia, thrombocytopenia, febrile neutropenia, and pancytopenia were also observed. However, there are few reports of HLH, coagulopathy, bone marrow failure, DIC, and B-cell aplasia, suggesting that these over-reported hematologic AE were largely underestimated in clinical trials. The reason may be associated with strict inclusion criteria and the careful selection of patients in these clinical trials.4,14 Moreover, HLH and DIC had substantial fatality rates of 69.9% and 59.6%, respectively. In particular, the incidence

of HLH was significantly lower in patients receiving tisagenlecleucel than in those receiving axicabtagene, and the death rates of HLH and DIC were higher after infusion of axicabtagene than tisagenlecleucel (Figures 2A, 4A). According to the reported studies by ZUMA-1 and JULIET,4,24 one patient died from HLH out of 119 patients receiving axicabtagene, while there were no deaths related to HLH out of 167 patients receiving tisagenlecleucel. HLH and DIC are rapidly progressing life-threatening hematologic AE that need to be monitored, and clinicians must stay on the alert for these rare but fatal hematologic AE in CAR T recipients, especially those receiving axicabtagene. Furthermore, we identified a strong correlation between HLH, DIC, and CRS (Figure 3C), in accordance with previously published studies.25-27 Therefore, tisagenlecleucel could reduce the risk of HLH and DIC in patients who are more likely to develop serious CRS, such as a high tumor burden prior to the infusion of CAR T.28

DIC: disseminated intravascular coagulation; HLH: hemophagocytic lymphohistiocytosis. S3 and S6 are co-efficients of PT selected from 3-fold and 6-fold cross-validation LASSO regression, respectively.

Table 3. Death-associated hematologic adverse events selected from K-fold cross-validation LASSO regression.

In addition to HLH and DIC, coagulopathy is also an underestimated but severe hematologic toxicity related to CAR T in clinical trials, always causing bleeding and thrombosis in CAR T recipients.29 Our findings suggest the fatality rate of coagulopathy is 47.7% (Figure 5). Tisagenlecleucel has been reported to have a higher rate of coagulopathy than axicabtagene (Table 2 and Figure 2A), which is consistent with published research in which 56.6% and 15.7% of patients suffered from coagulopathy after tisagenlecleucel and axicabtagene, respectively.29,30 B-cell aplasia, an 'offtarget' event, was also reported to be more common in patients receiving tisagenlecleucel than in those receiving axicabtagene (Figure 2A). This may be due to longer-lasting

co-stimulation of tisagenlecleucel, leading to longer persistence of CAR T, which are more likely to cause off-target AE. On the contrary, the frequency of bone marrow failure after infusion of tisagenlecleucel was lower than after axicabtagene. These findings revealed that we could optimize the choice of CAR T products prior to the therapy according to each patient's basic profile, thus reducing the risk of severe toxicities. Furthermore, early detection of these under-reported but lethal hematologic AE is essential because the patients we treated were not always subjected to strict selection criteria, unlike those on clinical trials. Neutropenia was the most frequently reported hematologic AE for axicabtagene (Table 2), which is consistent

Figure 4. Mortality rates of various hematologic adverse events following axicabtagene and tisagenlecleucel therapy. (A) Proportion of death in different preferred terms (PT) of hematologic adverse events. (B) Proportion of death in different high level terms (HLT) of hematologic adverse events. DIC: disseminated intravascular coagulation; HLH: hemophagocytic lymphohistiocytosis.

with previous clinical trials.23,31 In a multicenter phase II clinical trial of patients with refractory large B-cell lymphoma treated with axicabtagene, neutropenia was also the most common serious AE (grade ≥ 3).23 In particular, neutropenia can predispose patients to infection, which is the most common cause of NRM.2,32 Therefore, neutropenia must be prevented in patients receiving axicabtagene therapy by, for example, ensuring sufficient blood preparation prior to CAR T infusion and correct administration of hematopoietic growth factors after CAR T infusion. In addition, baseline cytopenia and high levels of C-reactive protein and ferritin prior to the therapy have been reported to be closely related to neutropenia after CAR T infusion.9 Therefore, correcting baseline cytopenia and reducing pretreatment levels of inflammatory factors can reduce the risk of neutropenia in CAR T recipients. In addition, early or prophylactic stimulation of hematopoietic stem cells may improve the outcomes of persistent neutropenia after CAR T infusion.33 For patients receiving tisagenlecleucel therapy, the highest reported frequency of hematologic AE was cytopenia, which is consistent with a phase II clinical trial.14

Febrile neutropenia was reported to have a significantly lower frequency in tisagenlecleucel recipients compared to axicabtagene, corresponding to ROR=0.64 (95% CI: 0.500.81) (Figure 2A). This finding is in line with the pivotal data reported from the JULIET and ZUMA-1 studies with 17% and 36% febrile neutropenia,4,24 respectively. Similarly, a lower reported frequency was observed in pancytopenia and hyperbilirubinemia after tisagenlecleucel therapy. Importantly, pancytopenia is a hallmark of hyperactive immune diseases, such as HLH,25 which also had a lower reported frequency in tisagenlecleucel than in axicabtagene. This finding further confirmed that tisagenlecleucel may be more suitable for patients at high risk for HLH,

such as those with a genetic profile associated with HLH.34 The time of onset of hematologic AE was shorter after axicabtagene infusion than after tisagenlecleucel infusion (Online Supplementary Figure S2). This may be due to the faster expansion of CAR T for axicabtagene than for tisagenlecleucel.35 Most patients suffered from hematologic AE within ten days after CAR T infusion, and a small proportion of patients experienced delayed hematotoxicity at 30 days or longer, which is consistent with previous studies.4,36 CRS is the most common AE, according to published data and real-world studies.2,14,37 The overlap between CRS, the top ten reported hematologic AE and two CAR T products was further analyzed to explore their relationship (Figure 3). Most patients had concurrent hematologic AE and CRS, which may be due to the detrimental effects of high concentrations of various cytokines on stem cells. For example, high levels of interleukin-6 and tumor necrosis factor alpha have been reported to damage the bone marrow microenvironment,38,39 and interferon-γ can impair hematopoietic stem cell proliferation.40 Therefore, early administration of tocilizumab to treat CRS may reduce the risk of hematologic AE. The frequency of concurrent CRS and hematotoxicity was higher in patients receiving axicabtagene than in those receiving tisagenlecleucel (54.0% vs. 48.8%). This may be associated with quicker expansion of CAR T and stronger hyperactive immune response in axicabtagene therapy, given their different co-stimulatory domains.35 It should be noted that 49.5% of patients reported hematotoxicity independently of CRS, suggesting that other off-target AE were associated with hematotoxicity in CAR T therapy.

Various hematologic AE were further divided into six subgroups based on HLT: bone marrow depression, coagulopathies and bleeding diatheses, hematologic disorders, hemolyses, spleen disorders, and other events (Online

Figure 5. Number of reports and mortality rates of 5 highly underestimated hematologic adverse events in clinical trials. These include B-cell aplasia, disseminated intravascular coagulation (DIC), bone marrow failure, coagulopathy, and hemophagocytic lymphohistiocytosis (HLH).

Supplementary Table S3). The results suggested that bone marrow depression, coagulopathies and bleeding diatheses, and hematologic disorders were strongly correlated with CRS (Figure 3C). According to previous findings, CRS manifestations include cytopenia, fever, coagulopathy, HLH, DIC, and end-organ dysfunction.41,42 Furthermore, hypofibrinogenemia coagulopathy has been reported to correlate with severe CRS in tisagenlecleucel therapy.43,44 In contrast, a weak correlation was observed between CRS and hemolyses and spleen disorders. In particular, the strongest connection was observed between bone marrow depression and CRS, and the probable reason could be that CRS was the risk factor for bone marrow depression.10,36

In terms of death reports, the mortality rates of HLH, DIC and coagulopathy were higher in patients treated with axicabtagene than in those treated with tisagenlecleucel (Figure 4A). Although the pathophysiology of HLH and DIC associated with CAR T remains unclear, numerous studies have postulated that severe CRS plays a vital role in triggering the occurrence of HLH and DIC.41,45 Previous clinical trials have found that the rates of severe CRS (grade ≥3) after treatment with axicabtagene and tisagenlecleucel were 7-9% and 1-5%, respectively.2,21,46 This may be the reason why higher frequency and mortality rates of HLH and DIC were observed after axicabtagene compared to tisagenlecleucel. Coagulopathy is strongly correlated with CRS.47 Therefore, vigilant monitoring of HLH, DIC, and coagulopathy is needed in patients with severe CRS. Moreover, there was a 100% death rate for hemolysis, neutropenic colitis, neutropenic sepsis, normocytic anemia, purpura, splenic hemorrhage, thrombotic microangiopathy, and white blood cell disorder after infusion of tisagenlecleucel (Figure 4A and Online Supplementary Table S5). However, there are few reports of these hematologic AE and further research is warranted.

In the LASSO regression analysis, 22 hematologic AE were associated with death in patients receiving CAR T therapy, such as cytopenia, HLH, and DIC (Table 3). In particular, all of the top ten reported hematologic AE were related to death, except coagulopathy. Combined with the results of the analysis of the fatality rate (Figure 4A), we found that HLH and DIC were life-threatening hematologic AE, with death rates exceeding 50% in both axicabtagene and tisagenlecleucel.

In summary, this post-marketing report analysis comprehensively revealed CAR T-related hematologic AE and compared the differences between axicabtagene and tisagenlecleucel. In general, HLH and DIC were largely under-reported, but fatal hematologic AE in CAR T recipients. These findings were instrumental in optimizing the choice of CAR T products according to the pre-treatment patient profile and in preventing rarely reported hematologic AE, reducing the risk of lethal toxicities, and improv-

ing the prognosis of CAR T recipients.

This study has some limitations. FAERS is the largest global repository of post-marketing drug AE reports, encompassing more than 28 million AE reports. It plays a vital role in post-marketing surveillance to characterize drugassociated AE, and provides a reliable foundation for developing ideas, generating hypotheses, and constructing study designs. However, the number of patients treated with axicabtagene and tisagenlecleucel has not been reported in FAERS. Therefore, the incidence of CAR T-associated hematologic AE remains unknown.

FAERS is a voluntary reporting system with the limitation that data on patients, disease, treatment, and outcomes are fragmented and unsystematic. Consequently, information was missing for some variables, such as age and sex. Meanwhile, there are some overlapping toxicities between cytopenia and anemia, thrombocytopenia, and neutropenia. Importantly, the treatment responses and prognosis of patients following CAR T therapy is unclear. In contrast to clinical trials, the comparison between axicabtagene and tisagenlecleucel is limited by the potential unbalanced characteristics of the patient population. However, clinical trials could not fully reflect AE due to restrictive inclusion criteria and careful patient selection, leading to some rare hematologic AE being underestimated. Therefore, a realworld study is essential to comprehensively identify hematologic AE related to CAR T cells, especially in identifying rare but life-threatening AE.

Disclosures

No conflicts of interest to disclose.

Contributions

YW, JY and ZT are responsible for study concept, methodology and software. ZS, DT and GT are responsible for data curation, and prepared and wrote the original draft. ZS, DT and GT are responsible for visualization and investigation of data. YW supervised the study. ZS, DT and NL are responsible for software and data validation. YW, JY and ZT wrote, reviewed and edited the paper.

Acknowledgments

We would like to thank Dr. Hu for reviewing our statistical methods and checking the analysis, and Editage (www.editage.jp) for English Language editing services.

Funding

Funding was received from the National Natural Science Foundation of China (NSFC) (81770209, 82270202) and Shanghai 2021 “Action Plan of Technological Innovation” Biomedical Science and Technology Support Special Project (21S11906100) (to JY). Funding was received from the Youth Start-up Foundation of the First Affiliated Hospital of Naval Medical University (2020QNB03, 2022QN067) (to YW).

Data-sharing statement

This study was performed on the basis of data from Food and Drug Administration Adverse Event Reporting System

References

1. Shah NN, Lee DW, Yates B, et al. Long-term follow-up of CD19CAR T-cell therapy in children and young adults with B-ALL. J Clin Oncol. 2021;39(15):1650-1659.

2. Nastoupil LJ, Jain MD, Feng L, et al. Standard-of-care axicabtagene ciloleucel for relapsed or refractory large B-cell lymphoma: results from the US Lymphoma CAR T Consortium. J Clin Oncol. 2020;38(27):3119-3128.

3. Crump M, Neelapu SS, Farooq U, et al. Outcomes in refractory diffuse large B-cell lymphoma: results from the international SCHOLAR-1 study. Blood. 2017;130(16):1800-1808.

4. Locke FL, Ghobadi A, Jacobson CA, et al. Long-term safety and activity of axicabtagene ciloleucel in refractory large B-cell lymphoma (ZUMA-1): a single-arm, multicentre, phase 1-2 trial. Lancet Oncol. 2019;20(1):31-42.

5. Frey NV, Shaw PA, Hexner EO, et al. Optimizing chimeric antigen receptor T-cell therapy for adults with acute lymphoblastic leukemia. J Clin Oncol. 2020;38(5):415-422.

6. Frey N, Porter D. Cytokine release syndrome with chimeric antigen receptor T cell therapy. Biol Blood Marrow Transplant. 2019;25(4):e123-e127.

7. Hirayama AV, Turtle CJ. Toxicities of CD19 CAR-T cell immunotherapy. Am J Hematol. 2019;94(S1):S42-S49.

8. Wudhikarn K, Pennisi M, Garcia-Recio M, et al. DLBCL patients treated with CD19 CAR T cells experience a high burden of organ toxicities but low nonrelapse mortality. Blood Adv. 2020;4(13):3024-3033.

9. Rejeski K, Perez A, Sesques P, et al. CAR-HEMATOTOX: a model for CAR T-cell-related hematologic toxicity in relapsed/refractory large B-cell lymphoma. Blood. 2021;138(24):2499-2513.

10. Fried S, Avigdor A, Bielorai B, et al. Early and late hematologic toxicity following CD19 CAR-T cells. Bone Marrow Transplant. 2019;54(10):1643-1650.

11. Nahas GR, Komanduri KV, Pereira D, et al. Incidence and risk factors associated with a syndrome of persistent cytopenias after CAR-T cell therapy (PCTT). Leuk Lymphoma. 2020;61(4):940-943.

12. Sesques P, Ferrant E, Safar V, et al. Commercial anti-CD19 CAR T cell therapy for patients with relapsed/refractory aggressive B cell lymphoma in a European center. Am J Hematol. 2020;95(11):1324-1333.

13. Cordeiro A, Bezerra ED, Hirayama AV, et al. Late events after treatment with CD19-targeted chimeric antigen receptor modified T cells. Biol Blood Marrow Transplant. 2020;26(1):26-33.

14. Schuster SJ, Bishop MR, Tam CS, et al. Tisagenlecleucel in adult relapsed or refractory diffuse large B-cell lymphoma. N Engl J Med. 2019;380(1):45-56.

15. Ang PS, Chen Z, Chan CL, et al. Data mining spontaneous adverse drug event reports for safety signals in Singapore - a comparison of three different disproportionality measures. Expert Opin Drug Saf. 2016;15(5):583-590.

16. Hou Y, Ye X, Wu G, et al. A comparison of disproportionality analysis methods in national adverse drug reaction databases of China. Expert Opin Drug Saf. 2014;13(7):853-857.

17. Norén GN, Hopstadius J, Bate A. Shrinkage observed-to-

(FAERS), a publicly available and anonymized database (https://www.fda.gov/regulatory-information/freedom-information).

expected ratios for robust and transparent large-scale pattern discovery. Stat Methods Med Res. 2013;22(1):57-69.

18. Brown EG, Wood L, Wood S. The medical dictionary for regulatory activities (MedDRA). Drug Saf. 1999;20(2):109-117.

19. Saez-Ibañez AR, Upadhaya S, Partridge T, et al. Landscape of cancer cell therapies: trends and real-world data. Nat Rev Drug Discov. 2022;21(9):631-632.

20. Luo W, Li C, Zhang Y, et al. Adverse effects in hematologic malignancies treated with chimeric antigen receptor (CAR) T cell therapy: a systematic review and meta-analysis. BMC Cancer. 2022;22(1):98.

21. Westin JR, Kersten MJ, Salles G, et al. Efficacy and safety of CD19-directed CAR-T cell therapies in patients with relapsed/refractory aggressive B-cell lymphomas: observations from the JULIET, ZUMA-1, and TRANSCEND trials. Am J Hematol. 2021;96(10):1295-1312.

22. Jacobson CA, Chavez JC, Sehgal AR, et al. Axicabtagene ciloleucel in relapsed or refractory indolent non-Hodgkin lymphoma (ZUMA-5): a single-arm, multicentre, phase 2 trial. Lancet Oncol. 2022;23(1):91-103.

23. Neelapu SS, Locke FL, Bartlett NL, et al. Axicabtagene ciloleucel CAR T-cell therapy in refractory large B-cell lymphoma. N Engl J Med. 2017;377(26):2531-2544.

24. Schuster SJ, Tam CS, Borchmann P, et al. Long-term clinical outcomes of tisagenlecleucel in patients with relapsed or refractory aggressive B-cell lymphomas (JULIET): a multicentre, open-label, single-arm, phase 2 study. Lancet Oncol. 2021;22(10):1403-1415.

25. Sandler RD, Tattersall RS, Schoemans H, et al. Diagnosis and management of secondary HLH/MAS following HSCT and CAR-T cell therapy in adults; a review of the literature and a survey of practice within EBMT centres on behalf of the Autoimmune Diseases Working Party (ADWP) and Transplant Complications Working Party (TCWP). Front Immunol. 2020;11:524.

26. Cutini I, Puccini B, Fabbri A, et al. Late haemophagocytic lymphohistiocytosis in a patient treated with axicabtagene ciloleucel. Transpl Immunol. 2022;75:101719.

27. Wang Y, Qi K, Cheng H, et al. Coagulation disorders after chimeric antigen receptor T cell therapy: analysis of 100 patients with relapsed and refractory hematologic malignancies. Biol Blood Marrow Transplant. 2020;26(5):865-875.

28. Yan Z, Zhang H, Cao J, et al. Characteristics and risk factors of cytokine release syndrome in chimeric antigen receptor T cell treatment. Front Immunol. 2021;12:611366.

29. Johnsrud A, Craig J, Baird J, et al. Incidence and risk factors associated with bleeding and thrombosis following chimeric antigen receptor T-cell therapy. Blood Adv. 2021;5(21):4465-4475.

30. Jiang H, Liu L, Guo T, et al. Improving the safety of CAR-T cell therapy by controlling CRS-related coagulopathy. Ann Hematol. 2019;98(7):1721-1732.

31. Strati P, Varma A, Adkins S, et al. Hematopoietic recovery and immune reconstitution after axicabtagene ciloleucel in patients with large B-cell lymphoma. Haematologica.

2021;106(10):2667-2672.

32. Haidar G, Dorritie K, Farah R, et al. Invasive mold infections after chimeric antigen receptor-modified T-cell therapy: a case series, review of the literature, and implications for prophylaxis. Clin Infect Dis. 2020;71(3):672-676.

33. Gagelmann N, Wulf GG, Duell J, et al. Hematopoietic stem cell boost for persistent neutropenia after CAR-T cell therapy: a GLA/DRST study. Blood Adv. 2023;7(4):555-559.

34. Tesi B, Lagerstedt-Robinson K, Chiang SCC, et al. Targeted highthroughput sequencing for genetic diagnostics of hemophagocytic lymphohistiocytosis. Genome Med. 2015;7:130.

35. Larson RC, Maus MV. Recent advances and discoveries in the mechanisms and functions of CAR T cells. Nat Rev Cancer. 2021;21(3):145-161.

36. Jain T, Knezevic A, Pennisi M, et al. Hematopoietic recovery in patients receiving chimeric antigen receptor T-cell therapy for hematologic malignancies. Blood Adv. 2020;4(15):3776-3787.

37. Goldman A, Maor E, Bomze D, et al. Adverse cardiovascular and pulmonary events associated with chimeric antigen receptor Tcell therapy. J Am Coll Cardiol. 2021;78(18):1800-1813.

38. Harmer D, Falank C, Reagan MR. Interleukin-6 interweaves the bone marrow microenvironment, bone loss, and multiple myeloma. Front Endocrinol (Lausanne). 2018;9:788.

39. Tie R, Li H, Cai S, et al. Interleukin-6 signaling regulates hematopoietic stem cell emergence. Exp Mol Med. 2019;51(10):1-12.

40. de Bruin AM, Demirel Ö, Hooibrink B, et al. Interferon-γ impairs proliferation of hematopoietic stem cells in mice. Blood. 2013;121(18):3578-3585.

41. Brudno JN, Kochenderfer JN. Recent advances in CAR T-cell toxicity: mechanisms, manifestations and management. Blood Rev. 2019;34:45-55.

42. Gust J, Hay KA, Hanafi L-A, et al. Endothelial activation and blood-brain barrier disruption in neurotoxicity after adoptive immunotherapy with CD19 CAR-T cells. Cancer Discov. 2017;7(12):1404-1419.

43. Grupp SA, Kalos M, Barrett D, et al. Chimeric antigen receptormodified T cells for acute lymphoid leukemia. N Engl J Med. 2013;368(16):1509-1518.

44. Maude SL, Frey N, Shaw PA, et al. Chimeric antigen receptor T cells for sustained remissions in leukemia. N Engl J Med. 2014;371(16):1507-1517.

45. Freyer CW, Porter DL. Cytokine release syndrome and neurotoxicity following CAR T-cell therapy for hematologic malignancies. J Allergy Clin Immunol. 2020;146(5):940-948.

46. Pasquini MC, Hu Z-H, Curran K, et al. Real-world evidence of tisagenlecleucel for pediatric acute lymphoblastic leukemia and non-Hodgkin lymphoma. Blood Adv. 2020;4(21):5414-5424.

47. Buechner J, Grupp SA, Hiramatsu H, et al. Practical guidelines for monitoring and management of coagulopathy following tisagenlecleucel CAR T-cell therapy. Blood Adv. 2021;5(2):593-601.

Automated production of specific T cells for treatment of refractory viral infections after allogeneic stem cell transplantation

Amadeus T. Heinz,1* Friso G.J. Calkoen,2* Alexander Derbich,1 Lea Miltner,1 Christian Seitz,1 Michaela Doering,1 Christiane Braun,1 Daniel Atar,1 Michael Schumm,1 Florian Heubach,1 AnneMarie Arendt,1 Ansgar Schulz,3 Friedhelm R. Schuster,4 Roland Meisel,4 Brigitte Strahm,5 Juergen Finke,6 Beatrice Heineking,7 Susanne Stetter,8 Gerda Silling,9 Daniel Stachel,10 Bernd Gruhn,11 Klaus-Michael Debatin,3 Juergen Foell,12 Johannes H. Schulte,13 Wilhelm Woessmann,14 Christine Mauz-Koerholz,14,15 Johanna Tischer,16 Tobias Feuchtinger,17 Rupert Handgretinger18 and Peter Lang1

1Department of Pediatric Hematology and Oncology, University Children’s Hospital Tübingen, Tübingen, Germany; 2Princess Máxima Center, Utrecht, the Netherlands; 3Department of Pediatrics, University Medical Center Ulm, Ulm, Germany; 4Division of Pediatric Stem Cell Therapy, Department of Pediatric Oncology, Hematology and Clinical Immunology, Medical Faculty, Heinrich-Heine-University, Dusseldorf, Germany; 5Department of Pediatrics and Adolescent Medicine, Division of Pediatric Hematology and Oncology, Medical Center, Faculty of Medicine, University of Freiburg, Freiburg, Germany; 6Department of Hematology and Oncology, University Hospital Freiburg, Freiburg, Germany; 7Department of Pediatric Cardiology and Intensive Care Medicine, Ludwig-Maximilians-University, Munich, Germany; 8Department of Medicine III, University Hospital Regensburg, Regensburg, Germany; 9Department of Hematology, Oncology, Hemostaseology and Stem Cell Transplantation, RWTH Aachen University, Aachen, Germany; 10Hospital for Pediatric and Adolescent Medicine, University of Erlangen, Erlangen, Germany; 11Department of Pediatrics, Jena University Hospital, Jena, Germany; 12Department of Hematology and Oncology, University

Children’s Hospital Regensburg, Regensburg, Germany; 13Department of Pediatric Oncology & Hematology, Charité University Medicine, Berlin, Germany; 14Department of Pediatric Hematology and Oncology, Justus-Liebig-University, Giessen, Germany; 15Medical Faculty of the Martin-Luther-University of Halle-Wittenberg, Halle, Germany; 16Department of Medicine III, Ludwig-Maximilians-University, Munich, Germany; 17Department of Pediatric Hematology, Oncology, Hemostaseology and Stem Cell Transplantation, Dr. von Hauner University Children’s Hospital, Munich, Germany and 18Abu Dhabi Stem Cells Center, Rowdhat, Abu Dhabi, UAE

*ATH and FGJC contributed equally as first authors.

Abstract

Correspondence: Amadeus T. Heinz amadeus.heinz@med.uni-tuebingen.de

Received: August 25, 2022.

Accepted: February 3, 2023.

Early view: February 16, 2023.

https://doi.org/10.3324/haematol.2022.281996

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Therapy-resistant viral reactivations contribute significantly to mortality after hematopoietic stem cell transplantation. Adoptive cellular therapy with virus-specific T cells (VST) has shown efficacy in various single-center trials. However, the scalability of this therapy is hampered by laborious production methods. In this study we describe the in-house production of VST in a closed system (CliniMACS Prodigy® system, Miltenyi Biotec). In addition, we report the efficacy in 26 patients with viral disease following hematopoietic stem cell transplantation in a retrospective analysis (adenovirus, n=7; cytomegalovirus, n=8; Epstein-Barr virus, n=4; multi-viral, n=7). The production of VST was successful in 100% of cases. The safety profile of VST therapy was favorable (n=2 grade 3 and n=1 grade 4 adverse events; all three were reversible). A response was seen in 20 of 26 patients (77%). Responding patients had a significantly better overall survival than patients who did not respond (P<0.001). Virus-specific symptoms were reduced or resolved in 47% of patients. The overall survival of the whole cohort was 28% after 6 months. This study shows the feasibility of automated VST production and safety of application. The scalability of the CliniMACS Prodigy® device increases the accessibility of VST treatment.

Introduction

Hematopoietic stem cell transplantation has shown great efficacy in various malignant and non-malignant diseases.1 A significant drawback of this therapy is opportunistic infections due to immunosuppression and prolonged T-cell aplasia, which impose relevant transplant-related morbidity and mortality.2-4 Delayed immune reconstitution can occur when using matched unrelated donors and especially in haplo-identical stem cell transplants because of extensive in vivo or ex vivo T-cell depletion, performed to prevent graft-versus-host disease (GvHD).5

Patients should therefore be routinely monitored after hematopoietic stem cell transplantation for reactivation of a latent infection with Herpesviridae (human cytomegalovirus [CMV] and Epstein-Barr virus [EBV]) or infection with human adenovirus (ADV), because of the high prevalence and morbidity of these viral infections.4,6-10 Antiviral therapy, either therapeutic or preemptive, against ADV and CMV is mainly based on antiviral agents such as cidofovir, foscarnet, ganciclovir11,12 and letermovir,13 but therapy failure due to viral resistance or toxic side effects is repeatedly observed.14-17 Recovery of cellular immunity is essential for eradication of the viral infections.

A current extension to virostatic agents is to transplant antiviral immunity via adoptive transfer of virus-specific T cells (VST) against ADV, CMV or EBV.18-21 Even multi-specific T cells have been evaluated.22 VST have shown promising antiviral efficacy as well as establishment of long-term immunity. Unfortunately, the extensive clean-room procedures for ex vivo culturing of VST are restricting the use of cellular therapies to specialized centers and it takes several weeks until the cells are ready to be used.3

A possible approach to these limitation is an automated closed system of VST production using the interferon (IFN)

cytokine capture system (CCS), as described by Kim et al and Kállay et al. 3,23 This method of production is based on the presentation of viral antigens to donor lymphocytes and subsequent magnetic separation of VST reacting to antigen stimulation with IFN-γ expression in a fully automated way, using the CliniMACS Prodigy® system from Miltenyi Biotec (Bergisch-Gladbach, Germany).

Following a retrospective analysis of real-world data, we report the results of 31 VST preparations and the experience of 12 treating centers after application of our VST with regard to safety and response to treatment in 26 patients.

Methods

Virus-specific T cells

Unstimulated apheresis products containing at least 1x109 T cells from the stem cell donor or, in the case of ineligibility, a third-party donor (n=5), were collected. Thirdparty donors were haploidentical family donors except one unrelated third-party donor who was tested on two HLA alleles with 50% matching. VST were isolated using the IFN-γ CCS (Miltenyi Biotec) on a CliniMACS Prodigy® (Figure 1). Within this device, cells were stimulated with viral peptides (ADV: MACS GMP PepTivator AdV5 Hexon; CMV: MACS GMP PepTivator HCMV pp65 or EBV: MACS GMP PepTivator Select, all from Miltenyi Biotec) for 4 h. For multi-speci fi c VST, the appropriate antigens were combined. Apheresis products were labeled with the CliniMACS CCS Catchmatrix Reagent, capturing the secreted IFN-γ on the surface of activated T cells. Labeled cells were separated using CliniMACS IFN-γ Enrichment Reagent, consisting of IFN- γ - speci fi c antibody-conjugated superparamagnetic particles. The fi nal product

Figure 1. Schematic of the production process for virus-specific T cells using the CliniMACS Prodigy® system. VST: virus-specific T cells; IFN: interferon.

was ready for infusion within 2 days. Release criteria for infusion were: (i) sterility; (ii) absence of endotoxins; (iii) purity (>20% CD3+ T cells); and (iv) cell viability (>60% of CD3+ cells). A maximum of 25x103 and 50x103 T cells/kg were infused per donation from haploidentical donors and matched donors, respectively.

Patients

Between January 2015 and December 2017, VST preparations for 42 patients were produced at the University Children’s Hospital in Tübingen. Patients were referred from Tübingen and the University Clinics in Berlin, Regensburg, Giessen, Marburg, München, Düsseldorf, Aachen, Ulm, Jena, Freiburg and Erlangen.

There were no fixed inclusion criteria, as this publication reports real-world data outside clinical trials. Usually, VST were given at increasing viral load despite anti-viral pharmacotherapy after allogeneic stem cell transplantation. Patients were excluded if they had active acute GvHD grade ≥2 or were receiving strong immunosuppressive drugs (prednisolone ≥1.0 mg/kg/day or equivalent).

Data collection and evaluation

Blood samples for viral polymerase chain reaction analysis were collected according to standard-of-care protocols in the different centers. Side effects, response and survival data as well as concomitant, prior or subsequent antiviral treatment were reported by the participating centers using standardized case report forms. Safety was assessed using Common Terminology Criteria for Adverse Events, version 3.

The data analysis was conducted with the formal approval of the institutional review board (University Tübingen, approval number 044/2017BO2) and in accordance with the Declaration of Helsinki, as revised in 2008. The requirement for informed consent was waived by the ethics committee due to the retrospective character of this analysis.

Definition of terms and statistics

Response was defined as reduction in viral load by ≥1 log. In the case of multi-viral infection, reduction of the viral load by ≥1 log in at least one infection was defined as response. Response kinetics were differentiated into straight response (reduction of viral load by ≥1 log within 8 weeks with no subsequent increase of viral load), transient response (reduction of viral load by ≥1 log within 8 weeks with subsequent re-increase of viral load by ≥ 1 log), and no response, as previously described. 24 Viral control was defined as a viral load below 1,000 copies per microliter, while final clearance was defined as no evidence of viral load. Relapse was defined as an increase of viral load ≥1 log after former viral control. Statistics were calculated using IBM SPSS 26 (Armonk,

NY, USA) and GraphPad 8 (San Diego, CA, USA). For overall survival, time from infusion of VST to death due to any cause or last follow-up was calculated. For comparison of overall survival levels, the log-rank test was used. Results were considered statistically significant at P values <0.05.

Results

Patients

Out of 42 patients for whom VST were produced, 32 received the T cells (7 products were conserved for later use in patients with a stable viral load and/or absence of urgent clinical symptoms at the time the VST product had been manufactured, 2 patients succumbed to disease prior to administration of the product, and in 1 patient the disease was cured prior to administration of the product). Six patients had to be excluded from the retrospective analysis of this cohort because of missing data (n=5) or death due to the patient’s underlying disease 1 day after administration of the product (n=1). In total, 26 patients were eligible for analysis; 31 preparations were manufactured, as VST were produced repetitively for two patients. The detailed characteristics of the eligible patients are presented in Table 1 and Online Supplementary Figure S1 Their median age was 13 years, with six adults also analyzed in this study. Leukemia was the most common indication for transplantation (n=11), followed by hematologic diseases plus immunodeficiency (n=10). Haploidentical donors were used more frequently than HLAidentical donors (14 vs. 12 patients). One patient received a reduced intensity conditioning regimen due to Fanconi anemia. All HLA-identical donors (11 matched unrelated donors, 1 matched sibling donor) were matched for 10/10 HLA alleles with the patients.

CMV infection was the most common reason for VST treatment (n=15), followed by ADV (n=11) and EBV (n=7), adding up to 33 viral reactivations in 26 patients (7 patients had multiple reactivations). Most patients received intensive antiviral treatment regimens with ≥2 antiviral drugs before therapy with VST was initiated.

Virus-specific T-cell production

In all 31 eligible procedures the product characteristics ful fi lled the predefi ned criteria for infusion of at least 1x103 VST (range, 59.4x103 2,948x103). A minimum of 1x109 cells were harvested during apheresis. The proportion of VST was determined by fl ow cytometry analysis as exemplified in Figure 2. The CD3+ cells in the apheresis product contained 0.02-0.94% IFN-γ+ cells after stimulation. After isolation of the IFN- γ + fraction, the mean number of VST was 773x103 with a mean recovery rate over 81%, defined as the proportion of IFN-γ+ cells in the

final cell product related to the count in the apheresis product. These products contained 76% VST on average. CMV- and EBV-specific T cells were more abundant in the apheresis product than were ADV-specific T cells (mean: 0.25% CMV; 0.33% EBV; 0.13% ADV), resulting in a higher VST yield after separation (mean: 560x103 CMV; 420x103 EBV; 150x103 ADV). The purity was also higher for those products (75% CMV; 84% EBV; 68% ADV) but with a recovery of 93%, cell loss was lowest for ADV VST products (85% CMV; 52% EBV). Enriched T cells were more likely to be CD4+ than CD8+ for ADV (78% CD4+; 18% CD8+), whereas EBV-VST were mainly CD8+ T cells (12% CD4+; 80% CD8+). For CMV the two subgroups were equally represented (41% CD4+; 52% CD8+). The highest yield of VST was observed when generating multi-specific T cells using a combination of peptides (880x103 ADV/CMV; 2,611x103 CMV/EBV). Per donation, 1x103-50x103 T cells/kg were used. The summarized production data are shown in Table 2.

Safety

With regard to the safety assessment of VST infusion, infusion-related symptoms were reported for 23 patients (no data for 3 patients). In total, nine adverse events attributed to VST infusion were reported in a total of three patients (1 with a HLA-identical donor, 2 with haploidentical donors, none in patients receiving third party-derived VST). Six of nine of the adverse events were grade 1 or 2, as presented in Online Supplementary Table S1. All infusion-related severe adverse events abated.

Grade 4 acute GvHD occurred in two out of 23 evaluable patients after transplantation, with complete control of GvHD symptoms prior to the administration of VST. These patients did not experience a relapse of GvHD after VST infusion.

One patient developed grade 2 acute GvHD of the skin after transplantation, also with control of the symptoms prior to VST infusion. After infusion of ADV-specific T cells in this patient, grade 3 diarrhea and grade 1 nausea occurred after infusion of VST. Rectoscopic biopsy showed diffuse T-cell infiltration. This patient was considered to have grade 3 GvHD and received immunosuppression, with complete relief of gastrointestinal symptoms. The ADV infection was cured, and the patient is alive with chronic GvHD of the skin. In one patient, the occurrence of grade 4 diarrhea and grade 3 nausea after infusion of ADV-specific T cells was reported. This patient had no history of GvHD prior to VST infusion. T-cell expansion was measured in this patient, showing simultaneous expansion of VST at the onset of symptoms. Thus, these gastrointestinal symptoms were considered as direct infiltration of ADV-specific T cells into the gastrointestinal tract, not GvHD, and the patient was not treated with immunosuppression. No rectoscopic biopsy was taken. The gastrointestinal symptoms resolved spontaneously, and the ADV infection was cleared.

Response to VST treatment was achieved by 18 of the 26 patients within 4 weeks (73%). After 8 weeks, two more patients responded, making a total of 20/26 responding patients (77%) and six non-responders.

MDS: myelodysplastic syndrome; BMF: bone marrow failure; EBV: Epstein-Barr virus; PTLD: post-transplant lymphoproliferative syndrome; TBI: total body irradiation; GvHD: acute graft-versus-host disease; VST: virus-specific T cells.

Table 1. Characteristics of the 26 patients who received virusspecific T cells.

The median time for a 1 log reduction in plasma viral load was 17.5 days (range, 6-41 days). Figure 3 shows the response rates within 8 weeks with regard to ADV-, CMV-, EBV- and multi-specific VST. The response rate to VST against CMV was 100%, whereas response rates were lower for VST against ADV (71%), multi-specific targets (71%) and EBV (50%).

A detailed investigation of response kinetics (Figure 4A) revealed that 13 of the 20 responders had a straight response (Figure 4B), leading directly to sustained viral control until the end of follow-up. Seven of the 13 straight responders had final clearance of the viral load after 8 weeks, whereas the viral load was cleared in 9/13 at the end of the follow-up. Four of the straight responders died

Figure 2. General gating strategy for determining the T-cell frequency and the proportion of virus-specific T cells in the cell product by flow cytometry before and after interferon-γ enrichment. Cells were gated on lymphocytes and single cells according to their morphological characteristics in forward scatter and sideward scatter. Cell viability was assessed by exclusion of cells stained by 7-aminoactinomycin D. CD45+CD3+ T-cell frequency was determined by antibody staining and further classified into CD4+ or CD8+ T cells. Cells that bound interferon-γ on their surface during the interferon-γ secretion phase were considered to be virus-specific T cells. SSC: side scatter; FSC: forward scatter; 7-aminoactinomycin D; IFN: interferon.

from non-virus-related causes shortly after responding to treatment, before achieving viral clearance.

Four of 19 responders showed a transient response (Figure 4C) with a secondary increase of viral load after the initial response, leading to delayed viral control. Two of these patients relapsed after achieving viral control, which diminished afterwards without requiring further antiviral treatment. Both patients were alive at the end of followup. Viral clearance was achieved in three of these four patients at the end of follow-up; the fourth patient did not completely clear the virus, but maintained a state of viral control (<1000 copies/μL).

The remaining three of 20 responders showed transient responses not leading to viral control (Figure 4D). One died due to the viral disease (n=1, ADV), while the deaths of the other two were potentially correlated with viral disease (n=1 ADV, n=1 EBV).

In total, final clearance of viral load was achieved after a median of 55 days (range, 13-350 days) and was observed in 12 of the 20 responding patients at the end of followup. Viral clearance was observed after a median of 42 days in patients with a straight response and 86 days after transient response with viral control (Figure 5).

Sixty-five percent of patients (15 out of 23 reported; 3 without data) had virus-specific symptoms at the onset of treatment, with nausea (n=7), pneumonia (n=7), diar-

responding to treatment with virus-specific T cells within 8 weeks, distinguishing between T cells against adenovirus, cytomegalovirus, Epstein-Barr virus or multi-specific targets. ADV: adenovirus; CMV: cytomegalovirus; EBV: Epstein-Barr virus.

aPurity: proportion of CD3+ IFN-γ+ T cells to total cell count in the final cell product. bRecovery: proportion of IFN-γ+ cells in the final cell product related to the count in the apheresis product. ADV: adenovirus; CMV: cytomegalovirus; EBV: Epstein-Barr virus; IFN-γ: interferon gamma.

Table 2. Product characteristics of 23 preparations of virus-specific T cells (data missing for 8 cases).

Figure 3. Proportions of patients

rhea (n=6) and hepatitis (n=3) being the most prevalent, as listed in Online Supplementary Table S2. Resolution of all symptoms occurred in 47% of patients (7 out of 15 patients with reported symptoms) within 13 weeks. For the remaining patients symptoms persisted (n=3) or worsened (n=5) until the end of follow-up.

A correlation analysis between VST dose and response revealed a correlation coefficient of r=0.16 and Pearson P=0.4, so no dose-dependent effect of VST on response was observed.

The response rate to third-party VST (80%, 4/5 patients) was effectively the same as that to VST acquired from the stem cell donor (76%, 16/21 patients) with no statistically significant difference in survival (log-rank test: P=0.63). Patients transplanted from an HLA-identical donor had a higher response rate (91%, 11/12 patients) than patients with a haploidentical stem cell donor (64%, 9/14 patients), but this difference was not statistically significant (χ2-test, P=0.1). There was no difference in survival (log-rank, P=0.4).

Concomitant antiviral treatment and immunosuppression

At the time point of VST infusion, six of the seven patients suffering from ADV infection were being treated with cidofovir (n=4), brincidofovir (n=1) or an unknown drug (n=1); the seventh patient started treatment with brincidofovir 4 weeks after VST infusion despite a prior response to the VST. All eight patients with CMV infection received concomitant treatment with valaciclovir/ganciclovir (n=5) or foscavir

Figure 4. Response kinetics, as previously defined, after infusion of virus-speci fic T cells. (A) Proportion of patients showing straight and transient responses with or without viral control or no response. (B-D) Changes in viral load in the 20 responding patients at different time points, with viral load measured by copies per µL at the time of infusion (day 0), lowest value within the first 8 weeks, highest value after 8 weeks, and value at the last follow-up in patients who had a straight response (B), a transient response with viral control (C) and a transient response without viral control. The definitions of the different responses are those described by Feucht et al. 24 LFU: last follow-up.

(n=3). All patients with EBV infection received rituximab during VST treatment, and one patient was also given foscavir. Patients suffering from multi-virus infection (n=7) were treated against one (n=3) or both (n=4) infections with concomitant antiviral medications.

Immunosuppression at the time of VST infusion was applied in 2/13 evaluable patients after haploidentical transplantation because of GvHD: in one case the immunosuppression was achieved with a calcineurin inhibitor (CNI) plus imatinib, in

Figure 5. Days until achievement of viral clearance depending on response kinetics (straight response vs. transient response with viral control).

the other case a CNI + prednisolone <1 mg/kg/day was used. After HLA-identical transplantation, 6/11 evaluable patients had ongoing immunosuppression with a CNI (n=4) or a CNI plus mycophenolate mofetil (n=1) due to early VST treatment within the first 100 days after transplantation or with CNI plus prednisolone <1 mg/kg/day (n=1) due to GvHD. There was no statistically significant difference in terms of response to VST treatment between patients receiving concomitant immunosuppression or no immunosuppression (88% vs. 69% response rate; χ2 , P=0.3); likewise there was no statistically significant difference in survival (log-rank, P=0.2).

Outcome

The median follow-up time was 70 days (range, 4-440 days). Eight patients died within the first 30 days, with the deaths being potentially related to the viral infection in half of these patients. The 6-month overall survival (Figure 6A) of the entire cohort was 28% (95% confidence interval [95% CI]: ±19%). Patients showing a response (Figure 6B) had a 6-month overall survival of 37% (95% CI: ±12%) with 3/12 patients dying due to or with their viral infection, as shown in Online Supplementary Table S3, whereas all six non-responders died within 100 days (P<0.001). Patients

with a straight or transient response with viral control showed a 6-month overall survival of 45% (95% CI: ±27%), whereas all patients with a transient response without viral control died after a median of 75 days and non-responders after a median of 26 days (Figure 6C). There was no difference in outcome between patients with ADV, CMV, EBV or multi-viral infection (P=0.63) (Figure 6D). Of the seven patients in the multi-viral group, four had short follow-up because of early deaths (2 virus-related, 2 not virus-related) within <30 days after VST administration.

Discussion

The aim of this retrospective analysis was to evaluate the feasibility of VST production with the CliniMACS Prodigy® system as well as the safety and efficacy of VST against different viral targets, based on real-world data.

Concerning feasibility, VST production using the automated IFN-γ CCS CliniMACS Prodigy® was reliable. VST against CMV, ADV, and EBV, as well as multi-specific VST, could be produced with sufficient cell numbers for 100% of the patients. The final cell product was ready for infusion within 2 days,

Figure 6. Overall survival of the investigated cohort. (A) Overall survival (OS) of the entire cohort. (B) OS depending on response to treatment with virus-specific T cells. (C) OS depending on response kinetics. (D) OS depending on type of infection. CMV: cytomegalovirus; ADV: adenovirus; EBV: Epstein-Barr virus; w/: with; w/o: without.

which is a significant reduction of production time compared with former ex vivo culture methods taking 2-12 weeks to be completed.25,26 In our experience, the fully automated CliniMACS Prodigy® system shortened hands-on time from 14 hours to 2-4 hours during regular working time, which reduces infrastructure requirements and the burden on the Good Manufacturing Practice team. These manufacturing times are in line with those reported by Priesner et al., 27 who compared CliniMACS Prodigy®-based production with manufacturing with the CliniMACS Plus®. In that study three healthy donors were used to produce CMV-specific T cells by both methods. The recovery rate was comparable, but the purity was higher using the CliniMACS Prodigy® (purity range on Prodigy®: 79.2-96.4% vs. 19.2-81.1% on the Plus®). The comparable pre-clinical study by Kim et al 3 extensively investigated the product characteristics of five production runs of CMV-specific T cells from healthy donor leukapheresis products. The final VST yield was lower than in our cohort, ranging from 2.7-470x103 IFN-γ+ T cells, despite a comparable percentage of CMV-specific T cells in the leukapheresis products. Kallay et al. 23 reported on the clinical use of CMV-specific T cells produced with the CliniMACS Prodigy® in a pediatric cohort and described a purity range of 26.5-94.4% for CD4+IFN-γ+ and 29.9-98.7% CD8+IFN-γ+ cells or 39.0-94.4% and 53.8-98.7%, respectively, when two outliers were excluded. These results are in line with our data for CMV-specific T cells as well as with those reported by Priesner et al 27

Regarding the safety profile of VST treatment, no major safety concerns arise from our data, in accordance with earlier studies evaluating the use of VST.20,23,26,28,29 One patient with GvHD grade 2 of the skin before VST treatment had a worsening to GvHD grade 3 with nausea and diarrhea after treatment with ADV-specific T cells, with histological proof of lymphocytic invasion into the gastrointestinal tract. Another patient treated in a different center, received ADVspecific T cells and developed the same symptoms. However, this patient was not considered to have developed GvHD, and showed spontaneous abatement of symptoms without the use of immunosuppression. So, it remains ambiguous whether these symptoms were a sign of GvHD, an invasion of ADV-specific T cells clearing the infection, or a combination of both. A comparable case of cystitis with lymphocytic invasion of BKV-reactive T cells after clearance of BKV using VST has been described.30

The very low risk of inducing de novo GvHD after T-cell infusion in our cohort might be correlated with the high purity of VST administered. In our opinion, it is of utmost importance to allow only minimal numbers of contaminating, unspecific T cells. Such cells can be potentially alloreactive and therefore cause GvHD, so we respected the limits of 25x103 IFN-γ-negative T cells for haploidentical transplants and 100x103/kg for HLA-identical transplants in the manufactured products, which were defined as risk thresholds for

GvHD according to our in-house experience. Looking at the efficacy of VST treatment, we can report responses in 77% (50-100%) of all patients, which is consistent with the reported response rate of 74% (62-85% depending on the virus targeted) in the pooled meta-analysis by Käuferle et al. 31 Furthermore, we can confirm the highly predictive value of response to VST treatment for survival and the observation of different response kinetics, as already shown:24 a large group of straight responders with fast, direct achievement of viral control had a relatively favorable outcome compared to transient responders. Transient response was correlated with delayed viral control in some and uncontrolled viral replication and death in other cases. With transient response not achieving viral control, at least median survival could be prolonged when compared with that of patients with no response. It does, however, remain an interesting question whether the response in these patients was achieved due to initial clonal expansion of VST, followed by secondary T-cell exhaustion, or whether these patients never had sufficient VST expansion, and the response was only caused by concomitant virostatic treatment. Unfortunately, we cannot answer this question through this retrospective dataset.

Two patients showed a relapse of their viral disease, but both achieved slow viral clearance without further antiviral treatment. This suggests a long-term efficacy of VST treatment.

Of course, the validity of our data is impaired by the co-administration of antiviral drugs. This prevented the effects of VST and virostatic treatment from being separated clearly, especially as no parallel measurement of clonal expansion of VST was performed during the study.

Unfortunately, overall survival was still poor with a 28% survival rate after 6 months in the whole cohort and 9/18 deaths potentially correlated with the viral infection, including the deaths of patients with unspecific multi-organ failure. Mortality rates of 82% for disseminated ADV viremia32 and almost 100% for CMV pneumonia33 were reported for historical cohorts. Pre-emptive virostatic treatment greatly improved survival rates for both post-transplant ADV and CMV reactivation,10,34 but a direct comparison of our results to these datasets is not feasible, as our cohort mainly suffered from refractory viral infections.

A main issue of VST treatment are early deaths <30 days which contributed to four of the virus-related fatalities. This underlies the importance of beginning treatment early, as the prognosis deteriorates quickly through uncontrolled viral replication.24 An automated IFN-γ CCS might be an important step in the direction of speeding up treatment, but regulatory barriers still limit this production method to specialized centers. Therefore, additional measures such as the implementation of donor databases35,36 may be needed to further broaden availability of cellular-based antiviral therapy. Another promising approach might be the pre-emptive use

of third-party VST at the first signs of CMV/EBV reactivation, as performed by Wei Jiang et al., 37 which led to an impressive response rate of 94% and in comparison with historical controls, a lower percentage of patients receiving third-line antiviral therapy. Although these results cannot be directly compared to our data, as refractory patients were included in our cohort, we can confirm that third-party VST seem to produce equal response rates compared to those of donorderived VST. Xu-Ying Pei et al. made the same observation.38 CMV-directed VST as prophylactic therapy has also been reported to be safe in matched unrelated donor recipients.39 The efficacy of this approach will be examined in a subsequent study; however, this method has previously been reported to be effective in preventing EBV-driven lymphoproliferative disease.40

Lastly, experimental approaches such as vaccination of the donor with viral peptides to boost the cell count of VST before apheresis, or PD-L1 inhibition in the patient to inhibit T-cell exhaustion41 may be attractive steps for further investigation, and controlled clinical trials, such as the TRACE study (NCT04832607), which is currently evaluating the feasibility and efficacy of decentralized VST production in different European countries, will be of great importance to prove the effectiveness of VST treatment in a randomized setting.

Disclosures

No

References

1. Sobocinski KA, Horowitz MM, Rowlings PA, et al. Bone marrow transplantation—1994: a report from the International Bone Marrow Transplant Registry and the North American Autologous Bone Marrow Transplant Registry. J Hematother. 1994;3(2):95-102.

2. Dykewicz CA. Summary of the guidelines for preventing opportunistic infections among hematopoietic stem cell transplant recipients. Clin Infect Dis. 2001;33(2):139-144.

3. Kim N, Nam YS, Im KI, et al. Robust production of cytomegalovirus pp65-specific T cells using a fully automated IFN-γ cytokine capture system. Transfus Med Hemother. 2018;45(1):13-22.

4. Ross SA, Novak Z, Pati S, Boppana SB. Overview of the diagnosis of cytomegalovirus infection. Infect Disord Drug Targets. 2011;11(5):466-474.

5. Leen AM, Tripic T, Rooney CM. Challenges of T cell therapies for virus-associated diseases after hematopoietic stem cell transplantation. Expert Opin Biol Ther. 2010;10(3):337-351.

6. Allen UD, Preiksaitis JK. Epstein-Barr virus and posttransplant lymphoproliferative disorder in solid organ transplantation. Am J Transplant. 2013;13(Suppl 4):107-120.

7. Chakrabarti S, Mautner V, Osman H, et al. Adenovirus infections following allogeneic stem cell transplantation: incidence and outcome in relation to graft manipulation, immunosuppression, and immune recovery. Blood. 2002;100(5):1619-1627.

8. Lucas KG, Burton RL, Zimmerman SE, et al. Semiquantitative Epstein-Barr virus (EBV) polymerase chain reaction for the

Contributions

AD, FGJC and LM acquired the data, which ATH, FGJC, AD and LM analyzed. CB, DA, MS , FH and AMA produced the virus-specific T cells. CS, MD, AS, FRS, RM, BS, JFi, BH, SS, GS, DS, BG, KMD, JFo, JHS, WW, CMK, JT, and TF reported data on the treatment and clinical outcome of treated patients. ATH and FGJC wrote the manuscript. RH and PL supervised the study. All authors contributed substantially to the manuscript.

Acknowledgments

The authors would like to thank SM Heinz for revising this manuscript for linguistic correctness.

Funding

This work was supported by grants from the excellence cluster iFIT (EXC 2180) [Gefördert durch die Deutsche Forschungsgemeinschaft (DFG) im Rahmen der Exzellenzstrategie des Bundes und der Länder – WXC 2180 – 390900677], from the Dieter Schwarz Stiftung Neckarsulm, and from the Reinhold-Beitlich Stiftung Tuebingen to PL; from the Foerderverein and from the Stiftung fuer krebskranke Kinder Tuebingen e.V. to PL and to ATH.

Data-sharing statement

The data that support the findings of this study are available from the corresponding author, ATH, upon reasonable request.

determination of patients at risk for EBV-induced lymphoproliferative disease after stem cell transplantation. Blood. 1998;91(10):3654-3661.

9. Sukdolak C, Tischer S, Dieks D, et al. CMV-, EBV- and ADVspecific T cell immunity: screening and monitoring of potential third-party donors to improve post-transplantation outcome. Biol Blood Marrow Transplant. 2013;19(10):1480-1492.

10. Travi G, Pergam SA. Cytomegalovirus pneumonia in hematopoietic stem cell recipients. J Intensive Care Med. 2014;29(4):200-212.

11. Cho SY, Lee DG, Kim HJ. Cytomegalovirus infections after hematopoietic stem cell transplantation: current status and future immunotherapy. Int J Mol Sci. 2019;20(11):2666.

12. Yusuf U, Hale GA, Carr J, et al. Cidofovir for the treatment of adenoviral infection in pediatric hematopoietic stem cell transplant patients. Transplantation. 2006;81(10):1398-1404.

13. Marty FM, Ljungman P, Chemaly RF, et al. Letermovir prophylaxis for cytomegalovirus in hematopoietic-cell transplantation. N Engl J Med. 2017;377(25):2433-2444.

14. Goodrich JM, Mori M, Gleaves CA, et al. Early treatment with ganciclovir to prevent cytomegalovirus disease after allogeneic bone marrow transplantation. N Engl J Med. 1991;325(23):1601-1607.

15. Kinchington PR, Araullo-Cruz T, Vergnes JP, Yates K, Gordon YJ. Sequence changes in the human adenovirus type 5 DNA polymerase associated with resistance to the broad spectrum

antiviral cidofovir. Antiviral Res. 2002;56(1):73-84.

16. Lindemans CA, Leen AM, Boelens JJ. How I treat adenovirus in hematopoietic stem cell transplant recipients. Blood. 2010;116(25):5476-5485.

17. Smith IL, Cherrington JM, Jiles RE, Fuller MD, Freeman WR, Spector SA. High-level resistance of cytomegalovirus to ganciclovir is associated with alterations in both the UL97 and DNA polymerase genes. J Infect Dis. 1997;176(1):69-77.

18. Feucht J, Joachim L, Lang P, Feuchtinger T. Adoptive T-cell transfer for refractory viral infections with cytomegalovirus, Epstein-Barr virus or adenovirus after allogeneic stem cell transplantation. Klin Padiatr. 2013;225(3):164-169.

19. Feuchtinger T, Matthes-Martin S, Richard C, et al. Safe adoptive transfer of virus-specific T-cell immunity for the treatment of systemic adenovirus infection after allogeneic stem cell transplantation. Br J Haematol. 2006;134(1):64-76.

20. Feuchtinger T, Opherk K, Bethge WA, et al. Adoptive transfer of pp65-specific T cells for the treatment of chemorefractory cytomegalovirus disease or reactivation after haploidentical and matched unrelated stem cell transplantation. Blood. 2010;116(20):4360-4367.

21. Saglio F, Hanley PJ, Bollard CM. The time is now: moving toward virus-specific T cells after allogeneic hematopoietic stem cell transplantation as the standard of care. Cytotherapy. 2014;16(2):149-159.

22. Gerdemann U, Christin AS, Vera JF, et al. Nucleofection of DCs to generate multivirus-specific T cells for prevention or treatment of viral infections in the immunocompromised host. Mol Ther. 2009;17(9):1616-1625.

23. Kállay K, Kassa C, Réti M, et al. Early experience with CliniMACS Prodigy CCS (IFN-gamma) system in selection of virus-specific T cells from third-party donors for pediatric patients with severe viral infections after hematopoietic stem cell transplantation. J Immunother. 2018;41(3):158-163.

24. Feucht J, Opherk K, Lang P, et al. Adoptive T-cell therapy with hexon-specific Th1 cells as a treatment of refractory adenovirus infection after HSCT. Blood. 2015;125(12):1986-1994.

25. Gerdemann U, Vera JF, Rooney CM, Leen AM. Generation of multivirus-specific T cells to prevent/treat viral infections after allogeneic hematopoietic stem cell transplant. J Vis Exp. 2011;(51):2736

26. Riddell SR, Watanabe KS, Goodrich JM, Li CR, Agha ME, Greenberg PD. Restoration of viral immunity in immunodeficient humans by the adoptive transfer of T cell clones. Science. 1992;257(5067):238-241.

27. Priesner C, Esser R, Tischer S, et al. Comparative analysis of clinical-scale IFN-γ-positive T-cell enrichment using partially and fully integrated platforms. Front Immunol. 2016;7:393.

28. Icheva V, Kayser S, Wolff D, et al. Adoptive transfer of Epstein-Barr

virus (EBV) nuclear antigen 1-specific T cells as treatment for EBV reactivation and lymphoproliferative disorders after allogeneic stem-cell transplantation. J Clin Oncol. 2013;31(1):39-48.

29. Pei XY, Zhao XY, Chang YJ, et al. Cytomegalovirus-specific T-cell transfer for refractory cytomegalovirus infection after haploidentical stem cell transplantation: the quantitative and qualitative immune recovery for cytomegalovirus. J Infect Dis. 2017;216(8):945-956.

30. Papadopoulou A, Gerdemann U, Katari UL, et al. Activity of broad-spectrum T cells as treatment for AdV, EBV, CMV, BKV, and HHV6 infections after HSCT. Sci Transl Med. 2014;6(242):242ra283.

31. Kaeuferle T, Krauss R, Blaeschke F, Willier S, Feuchtinger T. Strategies of adoptive T -cell transfer to treat refractory viral infections post allogeneic stem cell transplantation. J Hematol Oncol. 2019;12(1):13.

32. Lion T. Adenovirus infections in immunocompetent and immunocompromised patients. Clin Microbiol Rev. 2014;27(3):441-462.

33. Ljungman P. Cytomegalovirus pneumonia: presentation, diagnosis, and treatment. Semin Respir Infect. 1995;10(4):209-215.

34. Kampmann B, Cubitt D, Walls T, et al. Improved outcome for children with disseminated adenoviral infection following allogeneic stem cell transplantation. Br J Haematol. 2005;130(4):595-603.

35. Haque T, Wilkie GM, Jones MM, et al. Allogeneic cytotoxic T-cell therapy for EBV-positive posttransplantation lymphoproliferative disease: results of a phase 2 multicenter clinical trial. Blood. 2007;110(4):1123-1131.

36. Withers B, Clancy L, Burgess J, et al. Establishment and operation of a third-party virus-specific T cell bank within an allogeneic stem cell transplant program. Biol Blood Marrow Transplant. 2018;24(12):2433-2442.

37. Jiang W, Clancy LE, Avdic S, et al. Third-party CMV- and EBVspecific T-cells for first viral reactivation after allogeneic stem cell transplant. Blood Adv. 2022;6(17):4949-4966.

38. Pei X-Y, Liu X-F, Zhao X-Y, et al. Comparable anti-CMV responses of transplant donor and third-party CMV-specific T cells for treatment of CMV infection after allogeneic stem cell transplantation. Cell Mol Immunol. 2022;19(4):482-491.

39. Rubinstein JD, Lutzko C, Leemhuis T, et al. Scheduled administration of virus-specific T cells for viral prophylaxis after pediatric allogeneic stem cell transplant. Blood Adv. 2022;6(9):2897-2907.

40. Heslop HE. How I treat EBV lymphoproliferation. Blood. 2009;114(19):4002-4008.

41. Hashimoto M, Kamphorst AO, Im SJ, et al. CD8 T cell exhaustion in chronic infection and cancer: opportunities for interventions. Ann Rev Med. 2018;69(1):301-318.

GIMEMA phase II LLC1518 – VERITAS study

Francesca R. Mauro,1 Irene Della Starza,1,2 Monica Messina,2 Gianluigi Reda,3 Livio Trentin,4

Marta Coscia,5 Paolo Sportoletti,6 Lorella Orsucci,7 Valentina Arena,2 Gloria Margiotta

Casaluci,8 Roberto Marasca,9 Roberta Murru,10 Luca Laurenti,11 Fiorella Ilariucci,12 Caterina

Stelitano,13 Donato Mannina,14 Massimo Massaia,15 Gian Matteo Rigolin,16 Lydia Scarfò,17 Monia

Marchetti,18 Luciano Levato,19 Monica Tani,20 Annalisa Arcari,21 Gerardo Musuraca,22 Marina

Deodato,23 Piero Galieni,24 Valeria Belsito Patrizi,25 Daniela Gottardi,26 Anna Marina Liberati,27

Annamaria Giordano,28 Maria Chiara Molinari,1 Daniela Pietrasanta,18 Veronica Mattiello,3

Andrea Visentin,4 Candida Vitale,5 Francesco Albano,28 Antonino Neri,3 Lucia Anna De Novi,1

Maria Stefania De Propris,1 Mauro Nanni,1 Ilaria Del Giudice,1 Anna Guarini,1 Paola Fazi,2 Marco Vignetti,2 Alfonso Piciocchi,2 Antonio Cuneo16, # and Robin Foà1, #

1Hematology, Department of Translational and Precision Medicine, Sapienza University, Rome; 2GIMEMA Foundation, Rome; 3Hematology Department, Foundation IRCCS Ca’ Granda Ospedale Maggiore Policlinico of Milan, Milan; 4Hematology and Clinical Immunology Unit, Department of Medicine, University of Padua, Padua; 5Division of Hematology, A.O.U. Città della Salute e della Scienza di Torino and Department of Molecular Biotechnology and Health Sciences, University of Turin, Turin; 6Department of Medicine and Surgery, Institute of Hematology, Centro di Ricerca Emato Oncologica (CREO), University of Perugia, Perugia; 7Department of Hematology, Azienda Ospedaliera Universitaria Città della Salute e della Scienza di Torino, Turin; 8Division of Hematology, Department of Translational Medicine, Università del Piemonte Orientale and AOU Maggiore della Carità, Novara; 9Hematology Unit, Department of Medical and Surgical Sciences, University of Modena and Reggio Emilia, Modena; 10Hematology and Stem Cell Transplantation Unit, Ospedale A. Businco, ARNAS "G. Brotzu", Cagliari; 11Fondazione Policlinico Universitario A Gemelli, IRCCS, Rome; 12Hematology Unit, Azienda USL-IRCCS di Reggio Emilia, Reggio Emilia; 13Department of Hematology, Azienda Ospedaliera Bianchi Melacrino Morelli, Reggio Calabria; 14Division of Hematology, Azienda Ospedaliera Papardo, Messina; 15Division of Hematology, Santa Croce e Carle Hospital, Cuneo; 16Hematology Section, St. Anna University Hospital, Ferrara; 17Strategic Research Program on CLL, IRCCS Ospedale San Raffaele and Università Vita-Salute San Raffaele, Milan; 18Hematology and Transplant Unit, Azienda Ospedaliera SS Antonio e Biagio e Cesare Arrigo, University of Eastern Piedmont, Alessandria; 19Department of Hematology, Pugliese Ciaccio Hospital, Catanzaro; 20Division of Hematology, Santa Maria delle Croci Hospital, Ravenna; 21Division of Hematology, Guglielmo da Saliceto Hospital, Piacenza; 22Istituto Scientifico Romagnoli per lo Studio e la Cura dei Tumori-IRST, Meldola; 23ASST Grande Ospedale Metropolitano Niguarda, Milan; 24Hematology, Mazzoni Hospital, Ascoli Piceno; 25Hematology Department, Umberto I Hospital, Nocera Inferiore; 26A.O. Ordine Mauriziano di Torino, Turin; 27Università degli Studi di Perugia, Azienda Ospedaliera Santa Maria di Terni, Terni and 28University of Bari “Aldo Moro,” Hematology and Stem Cell Transplantation Unit, Department of Emergency and Organ Transplantation, Bari, Italy.

#AC and RF contributed equally as last authors.

Abstract

Correspondence: F.R. Mauro mauro@bce.uniroma1.it

Received: September 20, 2022.

Accepted: January 4, 2023.

Early view: January 12, 2023. https://doi.org/10.3324/haematol.2022.282116

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

The GIMEMA phase II LLC1518 VERITAS trial investigated the efficacy and safety of front-line, fixed-duration venetoclax and rituximab (VenR) in combination in young (≤65 years), fit patients with chronic lymphocytic leukemia and unmutated IGHV and/or TP53 disruption. Treatment consisted of the venetoclax ramp-up, six monthly courses of the VenR combination,

Haematologica | 108 August 2023 2091

High rate of durable responses with undetectable minimal residual disease with front-line venetoclax and rituximab in young, fit patients with chronic lymphocytic leukemia and an adverse biological profile: results of the

followed by six monthly courses of venetoclax as a single agent. A centralized assessment of minimal residual disease (MRD) was performed by allele-specific oligonucleotide polymerase chain reaction assay on the peripheral blood and bone marrow at the end of treatment (EOT) and during the follow-up. The primary endpoint was the complete remission rate at the EOT. Seventy-five patients were enrolled; the median age was 54 years (range, 38-65), 96% had unmutated IGHV, 12% had TP53 disruption, and 4% had mutated IGHV with TP53 disruption. The overall response rate at the EOT was 94.7%, with a complete remission rate of 76%. MRD was undetectable in the peripheral blood of 69.3% of patients and in the bone marrow of 58.7% of patients. The 12-month MRD-free survival in the 52 patients with undetectable MRD in the peripheral blood at the EOT was 73.1%. After a median follow-up of 20.8 months, no cases of disease progression were observed. Three patients had died, two due to COVID-19 and one due to tumor lysis syndrome. The first report of the VERITAS study shows that front-line VenR was associated with a high rate of complete remissions and durable response with undetectable MRD in young patients with chronic lymphocytic leukemia and unfavorable genetic characteristics. ClinicalTrials.gov identifier: NCT03455517.

Introduction

Chronic lymphocytic leukemia (CLL) is the most frequent leukemia in western countries and affects predominantly elderly subjects, with a median age of 72 years at presentation.1 Patients under 65 are less frequently diagnosed but more likely to have CLL as a cause of mortality than the elderly population.2,3

In recent years, relevant advances in our understanding of the biology of CLL have led to the development of targeted agents, namely the Bruton tyrosine kinase (BTK) inhibitors and the B-cell lymphoma 2 (BCL2) inhibitor venetoclax. The excellent therapeutic activity of these agents has radically changed the treatment approach for CLL, partially overcoming the unfavorable prognostic impact of adverse biological characteristics, including the unmutated configuration of the variable portion of the immunoglobin gene heavy chain (IGHV) gene and TP53 disruption (deletion and/or mutation of the TP53 gene). Continuous treatment with ibrutinib has demonstrated efficacy regardless of high-risk biological features and superiority over chemoimmunotherapy in relapsed/refractory CLL and previously untreated patients.4-12 Recent studies have shown similar efficacy with a better toxicity profile of the covalent BTK inhibitors acalabrutinib13-15 and zanubrutinib.16-18 The effectiveness of a non-covalent BTK inhibitor, pirtobrutinib, in patients resistant to ibrutinib due to a BTK mutation has also been described.19 Venetoclax, a selective oral BCL2 inhibitor, restores activation of CLL apoptosis.20 In several studies that included relapsed/refractory and treatment-naïve patients with CLL, fixed-duration treatment with venetoclax in combination with an anti-CD20 monoclonal antibody led to responses with undetectable minimal residual disease (MRD) in a large proportion of patients, including those with adverse genetic aberrations.21-26

The updated results of the randomized CLL14 study showed a superior 5-year progression-free survival in unfit patients treated front-line with venetoclax and obinutuzumab fixed-duration therapy compared to those who received chlorambucil and obinutuzumab (62.6% vs. 27.0%,

respectively).21,23 The randomized Murano trial for patients with relapsed/refractory CLL demonstrated a significant improvement in progression-free survival and overall survival with venetoclax and rituximab (VenR) as compared to chemoimmunotherapy.24-26 In addition, a high rate of deep responses with undetectable MRD was recorded with VenR, which was associated with a highly favorable impact on progression-free survival. Moreover, the safety profile of fixed-duration VenR was favorable, and severe rituximab-related infusion reactions did not occur. In addition, late adverse events, a relevant issue when treating younger patients with CLL and a long-life expectancy, were not observed.

Although BTK inhibitors are effective agents, fixed-duration therapy with venetoclax, capable of inducing profound and durable responses followed by a therapy-free period, is more appealing than continuous therapy, particularly for younger patients.

Based on the efficacy of fixed-duration VenR in the setting of patients with relapsed/refractory CLL, including those with unmutated IGHV and TP53 disruption, the Gruppo Italiano Malattie EMatologiche dell'Adulto (GIMEMA) investigated the efficacy and safety of a front-line VenR regimen in young (≤65 years), fit patients with CLL and an unfavorable biological profile. Here we report the first results of the GIMEMA phase II, single-arm, multicenter LLC1518 VERITAS study in 75 previously untreated, young patients with CLL and an unmutated IGHV profile and/or a TP53 disruption.

Methods Patients

The VERITAS study included previously untreated patients with CLL requiring treatment according to the International Workshop on CLL (iwCLL) criteria.27 The study was approved by the ethics committee of La Spaienza University (date of approval 07/06/2018; approval file number CE 497/18; reference 5049).

Patients were required to be ≤65 years, have a cumulative

illness rating scale (CIRS) score ≤6,28 have a creatinine clearance ≥30 mL/min, an unmutated IGHV gene and, or a TP53 disruption (17p deletion and/or TP53 mutation).2931 The IGHV profile and TP53 status were assessed centrally at the Hematology Center of the Sapienza University of Rome. The cytogenetic profile was investigated by fluorescence in situ hybridization at four reference laboratories (Rome, Ferrara, Bari, Milan).

Treatment

Study treatment consisted of a venetoclax ramp-up and six monthly courses of the VenR combination, followed by six monthly courses of venetoclax given as a single agent. During the ramp-up phase, patients were given venetoclax according to a 5-week escalation schedule with a gradual increase in the dose from 20 mg/day to 400 mg/day.22 Once the 5 weeks of the ramp-up phase had been completed, the following six cycles of VenR started on day 1 of cycle 1. Rituximab was administered on day 1 of each cycle. Venetoclax was continued at the dose of 400 mg/day in combination with rituximab at the dose of 375 mg/m2 on day 1 of cycle 1 (month 1) and at the dose of 500 mg/m2 on day 1 of cycles 2-6 (months 2-6). After the end of the combination therapy (EOCT), patients continued venetoclax monotherapy until day 28 of cycle 13, or unacceptable toxicity or disease progression. The risk of tumor lysis syndrome was assessed according to the presence of bulky lymphadenopathy (diameter ≥5 cm) and the peripheral absolute lymphocyte count (≥25×10⁹/L).32 Patients received prophylaxis against tumor lysis syndrome with urate-reducing agents and oral or intravenous hydration. Tumor lysis syndrome events were classified according to Howard's criteria.32 Adverse events were graded according to the Common Terminology Criteria for Adverse Events (CTCAE) version 5.33

Response

The response was assessed according to the iwCLL guidelines27 at the end of combination therapy (EOCT, month 7) and the end of treatment (EOT, month 15). Response assessment included clinical examination, peripheral blood evaluation, bone marrow aspirate and biopsy, and computed tomography scan. A centralized MRD assessment was carried out at the Hematology Center in Rome on peripheral blood and bone marrow cells by allele-specific oligonucleotide polymerase chain reaction (ASO-PCR) assay as previously reported.34,35 MRD was categorized as undetectable with a cut-off of <1 cell in 10,000 leukocytes. During the follow-up, MRD was monitored every 6 months.

Study endpoints

The primary endpoint of this study was the complete remission rate at the EOT. The secondary endpoints included the overall response rate, the rate of responses

M: male; F: female; ECOG: Eastern Cooperative Oncology Group; CIRS: Cumulative Illness Rating Scale; Hb: hemoglobin; LDH: lactate dehydrogenase; TLS: tumor lysis syndrome; IGHV: immunoglobulin heavy chain variable region gene; Del: deletion; Tris: trisomy; FISH: fluorescence in situ hybridization; TP53 gene: tumor protein p53 gene.

with undetectable MRD at the EOT, progression-free survival, and overall survival. Further secondary endpoints were the time to MRD conversion from undetectable to detectable, the time from the re-emergence of detectable leukemic cells to clinical progression of disease, and the time to a new CLL treatment.

Details on supportive treatment, statistical analysis, and ethics are reported in the Online Supplementary Material

Results

Patients

Between October 2018 and May 2020, 75 young patients with CLL and an unfavorable biological profile requiring front-line therapy from 28 Italian centers were included in this study and formed the intention-to-treat population

Table 1. Patients’ baseline characteristics.

assessed for treatment response and safety. The patients’ disposition is illustrated in Online Supplementary Figure S1. The patients’ baseline clinical and biological characteristics are summarized in Table 1. Their median age was 54 years (range, 38-65). Thirty-eight percent of the patients had advanced III-IV Rai stage disease; 41% had an increased level of β2 microglobulin, and 25% had bulky lymphadenopathy. The risk of tumor lysis syndrome was high at baseline in 44% of patients. Seventy-two patients (96%) had an unmutated IGHV gene profile, with a TP53 disruption in six (TP53 mutation, n=5; TP53 mutation and deletion, n=1), while three patients (4%) were IGHV-mutated and carried a TP53 disruption (TP53 mutation and deletion, n=3). The median CIRS score was 1 (range, 0-6), with nine (12%) patients having a CIRS score >3.

Response to treatment

Response at the end of the combination therapy

Seventy-two patients (96%) achieved a response at the EOCT (month 7). Responses included a complete response with or without blood count recovery in 41 patients (complete response, 52%; complete response with incomplete blood count recovery, 2.7%), and partial response in 31 (41.3%) (Figure 1). Three patients discontinued treatment because of an adverse event and were censored as treatment failures. The ASO-PCR assay demonstrated undetectable MRD at the EOCT in the peripheral blood and bone marrow of 70.7% and 46.7% of patients, respectively (Figure 2). The proportion of patients in complete remission with no measurable MRD by ASO-PCR in the peripheral blood and bone marrow was 78% and 61%, respectively (Figure 2). In patients who achieved a partial response, MRD could not be detected in the peripheral blood and bone marrow of 68% and 32% patients, respectively.

Response at the end of treatment

At the EOT (month 15), after a further 6 months of treatment with venetoclax as a single agent, the overall response rate was 94.7%, and the complete response rate increased from 54.2% to 76% (57 patients) (Figure 1). A partial response was recorded in 14 patients (18.7%) who showed residual lymph nodes (median longitudinal lymph node diameter, 1.95 cm; range, 1.5-4.5 cm). Two patients discontinued treatment because of an adverse event and were censored as treatment failures. A significantly lower complete response rate at the EOT was observed in older patients (P=0.032) and those with higher CIRS scores (P=0.009) (Online Supplementary Table S1). However, the only factor that retained a borderline statistical signifi-

Figure 1. Responses at the end of combination therapy and end of treatment according to International Working Group Chronic Lymphocytic Leukemia criteria. EOCT: end of combination therapy; EOT: end of treatment; ORR: overall response rate; CR: complete response; CRI: complete response with incomplete blood count recovery; PR: partial response.

Figure 2. Rates of responses with undetectable minimal residual disease (10-4) in the peripheral blood and bone marrow by allele-specific oligonucleotide polymerase chain reaction at the end of combination therapy and end of treatment. EOCT: end of combination therapy; EOT: end of treatment; PB: peripheral blood; BM: bone marrow.

cance in multivariate analysis was the CIRS score (P=0.054). A response with undetectable MRD by ASO-PCR was recorded in 69.3% of patients when examining peripheral blood and in 58.7% when bone marrow was tested (Figure 2). Six of the nine patients with TP53 disruption achieved a response with undetectable MRD in the peripheral blood and bone marrow. We analyzed the impact of the patients’ baseline characteristics and iwCLL-defined response measured at the EOCT on the probability of achieving undetectable MRD in the peripheral blood and bone marrow at the EOT. While no factors showed a significant impact on the rate of responses with undetectable MRD in the peripheral blood, the only factor associated with a higher probability of achieving undetectable MRD was a cut-off level of CD38 expression <30% in the bone marrow (Figure 3). MRD was monitored during the follow-up in 52 patients with a response and undetectable MRD in the peripheral blood at the EOT. MRD remained undetectable in the pe-

ripheral blood in 38 (73%) patients, 13 (25%) converted to detectable MRD, and one patient died from an adverse event. The 12-month MRD-free survival was 73.1% (95% confidence interval [95% CI]: 62-86.2) (Figure 4). There was no significant difference in the proportion of patients with complete or partial response and undetectable MRD in the bone marrow who lost the response at month 21 (undetectable MRD at month 21: partial response, 0/10 vs. complete response, 4/33).

Survival

After a median follow-up of 20.8 months (range, 0.2-36.5), no patient showed clinical progression, and three patients had died from adverse events. The 24-month overall survival was 96% (95% CI: 91.6-100) (Figure 5).

Safety

The grade ≥3 adverse reactions are described in Online

Figure 3. Impact of baseline factors and complete response measured at the end of combination therapy on responses with undetectable minimal residual disease in the bone marrow at the end of treatment. CR: complete response; TP53 gene: tumor protein p53 gene; Del: deletion; Tris: trisomy; IGHV: immunoglobulin heavy chain variable region gene; TLS: tumor lysis syndrome; LR: low risk; IR: intermediate risk; HR: high risk; LDH: lactate dehydrogenase; Hb: hemoglobin; CIRS: Cumulative Illness Rating Scale; ECOG: Eastern Cooperative Oncology Group; OR: odds ratio; 95% CI: 95% confidence interval.

Supplementary Table S2. Thirty-four patients (45.3%) experienced at least one grade ≥3 adverse event. Granulocytopenia was recorded in 28 patients (37.3%), and 26 (35%) received granulocyte colony-stimulating factors. Grade ≥3 infections were observed in nine patients (12%), including five patients (6.7%) who developed coronavirus disease 2019 (COVID-19) at the time of the first SARS-CoV2 pandemic when vaccination was unavailable. The nine patients who developed grade ≥3 infections (COVID-19 in 5/9 cases) were not characterized by increased risk factors for severe infections such as older age, high CIRS score, increased risk factor for tumor lysis syndrome, low creatinine clearance, or low granulocyte count at baseline. A fungal infection was reported in two patients. One patient showed clinical signs suggestive of sinusitis of sus-

pected, but not documented, fungal etiology. The other patient with steroid-controlled hemolytic anemia developed an Aspergillus pulmonary infection which was successfully treated with voriconazole.

A transient increase in liver enzymes was reported in three patients (4%). Thirty-three patients (44%) were at high risk of tumor lysis syndrome. Two patients had a creatinine clearance <60 mL/min at baseline but neither of them developed tumor lysis syndrome. Despite hospitalization, intravenous hydration, and the administration of anti-uric agents, one patient at increased risk of tumor lysis syndrome developed a grade 5 syndrome during the ramp-up phase. This patient with severe osteoporosis suffered from severe pain due to a vertebral fracture and used self-administered fentanyl patches for analgesic

diseasefree survival. uMRD: undetectable minimal residual disease; EOT: end of treatment.

Figure 4. Undetectable minimal residual

Figure 5. Overall survival of the whole cohort of 75 patients enrolled in the study. OS: overall survival.

purposes (more details on this clinical case are reported in the Online Supplementary File). One patient was diagnosed with follicular lymphoma 24 months after the start of treatment. No cases of Richter syndrome or non-hematologic cancers were recorded. Three patients died due to an adverse event, one with a clinical tumor lysis syndrome, and two with COVID-19.

Discussion

The first analysis of the VERITAS study showed that 94.7% of previously untreated, young patients with CLL and an adverse biological profile achieved a response with the VenR fixed-duration treatment. Moreover, no evidence of residual disease was detected in the peripheral blood of 69.3% of patients and in the bone marrow of 58.7%. After a median follow-up of 20.8 months, no cases of disease progression had occurred. These data confirm, in previously untreated patients, the efficacy of the VenR combination described in patients with relapsed/refractory CLL in the Murano trial.24-26

The primary endpoint of this study, the complete response rate at the EOT, was met with a 76% complete response rate, which compares favorably with that reported in fit patients treated with fluradabine, cyclophosphamide and rituximab (FCR) chemoimmunotherapy in the CLL10 trial (40%), the ECOG1912 study (30.3%),9,36 and also in the CLL14 trial in unfit patients treated with venetoclax plus obinutuzumab (49.5%).21-23 High complete response rates were also described with ibrutinib and venetoclax in the Flair trial (59.6%)38 and the MRD and fixed-duration cohorts of the Captivate trial (46% and 52.2%, respectively).37-39

Our study included young (≤65 years), fit patients with a CIRS score of ≤6. The presence of comorbidities, even with a CIRS score <6, was associated with a lower complete response rate. Interestingly, in a real-world study, higher CIRS scores were also associated with an adverse impact on the outcome of patients with CLL who received ibrutinib.40

The follow-up of this study, 20.8 months, is relatively short, and progression-free survival data are therefore premature. A valid surrogate of the efficacy of VenR is represented by the rate of patients with undetectable MRD in the peripheral blood, as determined by ASO-PCR in the CLL14 trial.21 The 69.3% rate of responses with undetectable MRD in peripheral blood recorded in our study compares favorably with the rates of undetectable MRD observed with FCR in the CLL10 and ECOG1912 trials (49% and 59.2%, respectively).9,36 In the CLL13 trial, which included patients with a more favorable genetic profile, a similar schedule produced responses with undetectable MRD in 57% of cases.41

Higher rates of responses with undetectable MRD in the peripheral blood were found in the CLL14 and CLL13 trials with the venetoclax and obinutuzumab combination (76% and 86.5%, respectively).21,41 Obinutuzumab, a more potent CD20 monoclonal antibody with a greater capacity for direct killing of B cells and a glyco-engineered Fc-fragment for improved effector-cell recruitment, has shown an advantage over rituximab as a partner of venetoclax. Although in the CLL13 trial infusion-related reactions associated with obinutuzumab were more severe than those seen with rituximab, patients treated with venetoclax and obinutuzumab showed a higher rate of responses with undetectable MRD and more prolonged progressionfree survival compared to those treated with VenR.41 In the Glow trial, which included elderly/unfit patients with CLL, the venetoclax plus ibrutinib combination resulted in 54.7% of patients having responses with undetectable MRD in the peripheral blood,42 while higher rates were observed in the Flair trial (71.3%),37 and in the MRD and fixed-duration cohorts of the Captivate study (75% and 77%, respectively).38,39 In the CLL13 trial, the triplet combination of venetoclax, ibrutinib, and obinutuzumab was associated with the highest rate of responses with undetectable MRD in the peripheral blood (92.2%).41 A comparison between the rates of undetectable MRD in the different studies is hampered by the technique used to measure residual disease, which was ASO-PCR in some studies,21,42 like ours, and flow cytometry in others.38-39;41 Although cross-trial comparisons must be interpreted with caution, it is important to underline that in our study, 96% of patients had an unmutated IGHV gene profile, whereas the proportion of patients with unmutated IGHV ranged between 43.5% and 60.5% in the studies mentioned above.

Longer follow-up of this and other studies may show whether patients with these same unfavorable genetic characteristics can benefit from different and more prolonged venetoclax-based treatments.

CD38 positivity, recorded in 51% of patients, emerged as the only factor with an unfavorable impact on the rate of undetectable MRD in the bone marrow. CD38, a multifunctional surface transmembrane glycoprotein,43 is associated with an IGHV unmutated status, advanced-stage disease, poor response to chemotherapy, shorter time to first treatment, and survival.44-46 To the best of our knowledge, the prognostic impact of CD38 expression has not yet been tested in patients treated with venetoclax. In a study by Sargent et al., 47 a significant inverse relationship was observed in vitro between the proportion of CD38positive cells and the level of BCL2 expression. Based on this finding, we speculate that CD38-negative patients could express higher levels of the anti-apoptotic BCL2 protein, resulting in a more pronounced activity of venetoclax.

Due to the number of patients included in this trial, the predictive value of novel mutations occurring in a minority of patients was not analyzed.

The re-emergence of MRD after FCR treatment is more rapid in patients with unmutated IGHV than in those with mutated IGHV.48 In our study, which included mainly patients with unmutated IGHV, 73% of patients who achieved a response with undetectable MRD maintained the status of undetectable MRD in the peripheral blood at 12 months after the EOT. Despite the unfavorable genetic characteristics of our study cohort, the MRD-free survival in our analysis is in line with that observed in unfit patients treated with venetoclax and obinutuzumab in the CLL14 trial.23

VenR treatment was well tolerated. Notably, no grade ≥3 infusion reactions to rituximab were recorded. The most frequent adverse event was granulocytopenia, which was easily manageable with granulocyte growth factors. Unfortunately, our study was carried out during the outbreak of the SARS-CoV-2 pandemic before vaccines were introduced, and five of the nine grade ≥3 infections were due to SARS-CoV-2 infection. About half of the patients in this study had a high risk of tumor lysis syndrome. However, only one case of fatal clinical tumor lysis syndrome was observed in a patient who underwent the ramp-up phase of the treatment regimen while receiving a drug, for analgesic purposes, which may have interfered with the metabolism of venetoclax. One patient discontinued therapy due to the diagnosis of indolent lymphoma, while no cases of Richter transformation or second malignancies were observed.

In conclusion, this first report of the VERITAS study shows that the VenR combination as front-line treatment is easily manageable, well-tolerated, and associated with high rates of complete responses and durable responses with undetectable MRD in younger patients with CLL and unfavorable genetic characteristics.

Disclosures

FRM reports advisory board and lecture fees from Abbvie, Janssen, AstraZeneca, BeiGene; and research funding from AbbVie and Takeda. MMar has received consulting fees from GILEAD. GMR has received payments or honoraria for lectures, presentations, speakers' bureau, manuscript writing or educational events from Astrazeneca, GILEAD, Janssen, Abbvie; has received support for attending meetings and/or travel from Janssen. IDG has received payments or honoraria for lectures, presentations, speakers' bureau, manuscript writing or educational events from Astrazeneca; has received support for attending meetings and/or travel from Roche, Janssen and Takeda; has sat on a Data Safety Monitoring Board or Advisory Board of Janssen; has held a leadership or fiduciary role in another board, society, committee or advocacy group, paid or unpaid of Fondazione Ita-

liana Linfomi (FIL), Commissione studi biologici. RMu has received payments or honoraria for lectures, presentations, speakers' bureau, manuscript writing or educational events from Astrazeneca and Janssen; and has received support for attending meetings and/or travel from Abbvie, Astrazeneca and Janssen. RMa has received consulting fees from Abbvie and Jannsen; has receiv ed payments or honoraria for lectures, presentations, speakers' bureau, manuscript writing or educational events from Abbvie, Janssen and Astrazeneca; and has received support for attending meetings and/or travel from Abbvie and Astrazeneca. MT participated in an Advisory Board for Abbvie. PS has received grants or contracts from GILEAD, Novartis and Abbvie. AA has received payments or honoraria for lectures, presentations, speakers' bureau,manuscript writing or educational events from Janssen, Abbvie and Servier; and has received support for attending meetings and/or travel from Takeda and Janssen. AV has received support for attending meetings and/or travel from Abbvie, Janssen, and Beigene; and has sat on a Data Safety Monitoring Board or Advisory Board for Janssen, Abbvie, and Beigene. MV plays a consultant or advisory role or has participated in speakers' bureau for Amgen, Novartis, Mattioli srl, IQVIA and Dephaforum S.r.l. The other authors do not have any conflicts of interest to disclose.

Contributions

FRM, AC and RF developed the concept and design of the study, interpreted data and wrote the manuscript. MMes, VA, PF, MV, and AP managed the data collection and assembly, performed the statistical analysis and interpreted data. IDS, FA, AN, LDN, MSDP, MN, and IDG performed biological and molecular studies and analyzed and interpreted data. GR, LT, MC, PS, LO, GMC, RMa, RMu, LLa, FI, CS, DM, MMas, GMR, LS, MMar, LLe, MT, AA, GM, MD, PG, VBP, DG, AML, AG, MCM, DP, VM, AV, and CV managed patients and participated in the collection of clinical data. All authors critically revised the manuscript and reviewed and approved the final version.

Acknowledgments

The authors thank all the patients included in this study and their families.

Funding

AbbVie Srl contributed with an unconditioned grant to GIMEMA and provided the study drug. The study was partly supported by Associazione Italiana Ricerca sul Cancro (AIRC), Special 5x1000 Program Metastases (21198), Milan (Italy) to RF. Work at the AC Unit was supported by BEAT LEUKEMIA Onlus.

Data-sharing statement

The data presented in this study are available on request. Details on sharing criteria and processes for requesting access to data can be obtained from a.piciocchi@gimema.it.

References

1. The Surveillance Epidemiology and End Results (SEER) Program of the National Cancer Institute. Cancer Stat Facts: LeukemiaChronic Lymphocytic Leukemia (CLL). https://seer.cancer.gov/statfacts/html/clyl.html; 2021.

2. Shanafelt TD, Rabe KG, Kay NE, et al. Age at diagnosis and the utility of prognostic testing in patients with chronic lymphocytic leukemia. Cancer. 2010;116(20):4777-4787.

3. Mauro FR, Foa R, Giannarelli D, et al. Clinical characteristics and outcome of young chronic lymphocytic leukemia patients: a single institution study of 204 cases. Blood. 1999;94(2):448-454.

4. Burger JA, Tedeschi A, Barr PM, et al. Ibrutinib as initial therapy for patients with chronic lymphocytic leukemia. N Engl J Med. 2015;373(25):2425-2437.

5. Burger JA, Barr PM, Robak T, et al. Long-term efficacy and safety of first-line ibrutinib treatment for patients with CLL/SLL: 5 years of follow-up from the phase 3 RESONATE-2 study. Leukemia. 2020;34(3):787-798.

6. Woyach JA, Ruppert AS, Heerema NA, et al. Ibrutinib regimens versus chemoimmunotherapy in older patients with untreated CLL. N Engl J Med. 2018;379(26):2517-2528.

7. Woyach JA, Ruppert AS, Heerema NA, et al. Long-term results of Alliance AO41202 show continued advantage of ibrutinibbased regimens compared with bendamustine plus rituximab (BR) chemoimmunotherapy. Blood. 2021;138(Suppl 1):639.

8. Moreno C, Greil R, Demirkan F, et al. First-line treatment of chronic lymphocytic leukemia with ibrutinib plus obinutuzumab versus chlorambucil plus obinutuzumab: final analysis of the randomized, phase III iLLUMINATE trial. Haematologica. 2022;107(9):2108-2120.

9. Shanafelt TD, Wang XV, Kay NE, et al. Ibrutinib–rituximab or chemoimmunotherapy for chronic lymphocytic leukemia. N Engl J Med. 2019;381(5):432-443.

10. Shanafelt TD, Wang XV, Hanson CA, et al. Long-term outcomes for ibrutinib-rituximab and chemoimmunotherapy in CLL: updated results of the E1912 trial. Blood. 2022;140(2):112-120.

11. Byrd JC, Brown JR, O’Brien S, et al. Ibrutinib versus ofatumumab in previously treated chronic lymphoid leukemia. N Engl J Med. 2014;371(3):213 223.

12. Munir T, Brown JR, O’Brien S, et al. Final analysis from RESONATE: up to six years of follow up on ibrutinib in patients with previously treated chronic lymphocytic leukemia or small lymphocytic lymphoma. Am J Hematol. 2019;94(12):1353 1363.

13. Ghia P, Pluta A, Wach M, et al. ASCEND: phase III, randomized trial of acalabrutinib versus idelalisib plus rituximab or bendamustine plus rituximab in relapsed or refractory chronic lymphocytic leukemia. J Clin Oncol. 2020;38(25):2849 2861.

14. Sharman JP, Egyed M, Jurczak W, et al. Acalabrutinib with or without obinutuzumab versus chlorambucil and obinutuzumab for treatment-naive chronic lymphocytic leukaemia (ELEVATE TN): a randomised, controlled, phase 3 trial. Lancet. 2020;395(10232):1278 1291.

15. Byrd JC, Hillmen P, Ghia P, et al. First results of a head to head trial of acalabrutinib versus ibrutinib in previously treated chronic lymphocytic leukemia. J Clin Oncol. 2021;39(15_suppl):7500.

16. Tam CS, Brown JR, Kahl BS, et al. Zanubrutinib versus bendamustine and rituximab in untreated chronic lymphocytic leukaemia and small lymphocytic lymphoma (SEQUOIA): a randomised, controlled, phase 3 trial. Lancet Oncol. 2022;23(8):1031-1043.

17. Tedeschi A, Ferrant E, Flinn IW, et al. Zanubrutinib in

combination with venetoclax for patients with treatment-naïve (TN) chronic lymphocytic leukemia (CLL) or small lymphocytic lymphoma (SLL) with del(17p): early results from arm D of the SEQUOIA (BGB-3111-304) trial. Blood. 2021;138(Suppl 1):67.

18. Hillmen P, Eichhorst B, Brown JR, et al. First interim analysis of ALPINE study: results of a phase 3 randomized study of zanubrutinib vs ibrutinib in patients with relapsed/refractory chronic lymphocytic leukemia/small lymphocytic lymphoma. Blood. 2021;138(Suppl 1):67.

19. Mato AR, Shah NN, Jurczak W, et al. Pirtobrutinib in relapsed or refractory B-cell malignancies (BRUIN): a phase 1/2 study. Lancet. 2021;397(10277):892-901.

20. Anderson MA, Deng J, Seymour JF, et al. The BCL2 selective inhibitor venetoclax induces rapid onset apoptosis of CLL cells in patients via a TP53-independent mechanism. Blood. 2016;127(25):3215-3224.

21. Fischer K, Al-Sawaf O, Bahlo J, et al. Venetoclax and obinutuzumab in patients with CLL and coexisting conditions. N Engl J Med. 2019;380(23):2225-2236.

22. Al-Sawaf O, Zhang C, Tandon M, et al. Venetoclax plus obinutuzumab versus chlorambucil plus obinutuzumab for previously untreated chronic lymphocytic leukaemia (CLL14): follow-up results from a multicentre, open-label, randomised, phase 3 trial. Lancet Oncol. 2020;21(9):1188-1200.

23. Al-Sawaf O, Zhang C, Robrecht S, et al. Venetoclaxobinutuzumab for previously untreated chronic lymphocytic leukemia: 4-year follow-up analysis of the randomized CLL14 study. HemaSphere. 2022;6(Suppl 3):49-50 (abstract S148).

24. Seymour JF, Kipps TJ, Eichhorst B, et al. Venetoclax–rituximab in relapsed or refractory chronic lymphocytic leukemia. N Engl J Med. 2018;378(12):1107 1120.

25. Kater AP, Kipps TJ, Eichhorst B, et al. Five-year analysis of Murano study demonstrates enduring undetectable minimal residual disease (uMRD) in a subset of relapsed/refractory chronic lymphocytic leukemia (R/R CLL) patients (Pts) following fixed-duration venetoclax-rituximab (VenR) therapy (Tx). Blood. 2020;136(Supp 1):19-21 (abstract 642).

26. Seymour JF, Kipps TJ, Eichhorst BF, et al. Enduring undetectable MRD and updated outcomes in relapsed/refractory CLL after fixed-duration venetoclaxrituximab. Blood. 2022;25;140(8):839-850.

27. Hallek M, Cheson BD, Catovsky D, et al. iwCLL guidelines for diagnosis, indications for treatment, response assessment, and supportive management of CLL. Blood. 2018;131(25):2745-2760.

28. Salvi F, Miller MD, Grilli A, et al. A manual of guidelines to score the modified Cumulative Illness Rating Scale and its validation in acute hospitalized elderly patients. J Am Geriatr Soc. 2008;56(10):1926-1931.

29. Rosenquist R, Ghia P, Hadzidimitriou A, et al. Immunoglobulin gene sequence analysis in chronic lymphocytic leukemia: updated ERIC recommendations. Leukemia. 2017;31(7):1477-1481.

30. Döhner H, Stilgenbauer S, Benner A, et al. Genomic aberrations and survival in chronic lymphocytic leukemia. N Engl J Med. 2000;343(26):1910-1916.

31. Malcikova J, Tausch E, Rossi D, et al. ERIC recommendations for TP53 mutation analysis in chronic lymphocytic leukemiaupdate on methodological approaches and results interpretation. Leukemia. 2018;32(5):1070-1080.

32. Howard SC, Jones DP, Pui C-H, et al. Managing tumor lysis syndrome in the era of novel cancer therapies. J Adv Pract Oncol.

2017;8(7):705-720.

33. Common Terminology Criteria for Adverse Events (CTCAE) Version 5.0. November 2017.

https://ctep.cancer.gov/protocoldevelopment/electronic_applica tions/docs/ctcae_v5_quick_reference_5x7.pdf

34. Rawstron AC, Fazi C, Agathangelidis A, et al. A complementary role of multiparameter flow cytometry and high-throughput sequencing for minimal residual disease detection in chronic lymphocytic leukemia: an European Research Initiative on CLL study. Leukemia. 2016;30(4):929-936.

35. Raponi S, Della Starza I, De Propris MS, et al. Minimal residual disease monitoring in chronic lymphocytic leukaemia patients. A comparative analysis of flow cytometry and ASO IgH RQ-PCR. Br J Haematol. 2014;166(3):360-368.

36. Eichhorst B, Fink AM, Bahlo J, et al; German CLL Study Group (GCLLSG). First-line chemoimmunotherapy with bendamustine and rituximab versus fludarabine, cyclophosphamide, and rituximab in patients with advanced chronic lymphocytic leukaemia (CLL10): an international, open-label, randomised, phase 3, non-inferiority trial. Lancet Oncol. 2016;17(7):928-942.

37. Hillmen P, Pitchford A, Bloor A, et al. The combination of ibrutinib plus venetoclax results in a high rate of MRD negativity in previously untreated CLL: the results of the planned interim analysis of the phase III NCRI FLAIR trial. HemaSphere. 2022;6(Suppl 3):46-47 (abstract S145).

38. Wierda WG, Allan JN, Siddiqi T, et al. Ibrutinib plus venetoclax for first-line treatment of chronic lymphocytic leukemia: primary analysis results from the minimal residual disease cohort of the randomized phase II CAPTIVATE study. J Clin Oncol. 2021;39(34):3853-3865.

39. Tam CS, Allan JN, Siddiqi T, et al. Fixed-duration ibrutinib plus venetoclax for first-line treatment of CLL: primary analysis of the CAPTIVATE FD cohort. Blood. 2022;139(22):3278-3289.

40. Tedeschi A, Frustaci AM, Mauro FR, et al. Do age, fitness, and

concomitant medications influence management and outcomes of patients with CLL treated with ibrutinib? Blood Adv. 2021;5(24):5490-5500.

41. Eichhorst B, Niemann C, Kater A, et al. Time-limited venetoclaxobinutuzumab +/- ibrutinib is superior to chemoimmunotherapy in front-line chronic lymphocytic leukemia (CLL): PFS co-primary endpoint of the randomized phase 3 GAIA/CLL13 trial. HemaSphere. 2022;6(Suppl 3):abstract LB2365.

42. Munir T, Moreno C, Owen C, et al. First prospective data on minimal residual disease (MRD) outcomes after fixed-duration ibrutinib plus venetoclax (Ibr+Ven) versus chlorambucil plus obinutuzumab (Clb+O) for first-line treatment of CLL in elderly or unfit patients: the GLOW study. Blood. 2021;138(Suppl 1):70.

43. Malavasi F, Funaro A, Roggero S, Horenstein A, Calosso L, Mehta K. Human CD38: a glycoprotein in search of a function. Immunol Today. 1994;15(3):95-97.

44. Damle RN, Wasil T, Fais F, et al. Ig V gene mutation status and CD38 expression as novel prognostic indicators in chronic lymphocytic leukemia. Blood. 1999;94(6):1840-1847.

45. 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.

46. Damle RN, Temburni S, Calissano C, et al. CD38 expression labels an activated subset within chronic lymphocytic leukemia clones enriched in proliferating B cells. Blood. 2007;110(9):3352–3359.

47. Sargent RL, Craig FE, Swerdlow SH. Comparison of Bcl-2, CD38 and ZAP-70 expression in chronic lymphocytic leukemia. Int J Clin Exp Pathol. 2009;2(6):5745-5782.

48. Thompson PA, Peterson CB, Strati P, et al. Serial minimal residual disease (MRD) monitoring during first-line FCR treatment for CLL may direct individualized therapeutic strategies. Leukemia. 2018;32(11):2388-2398.

The immunomodulatory molecule TIGIT is expressed by chronic lymphocytic leukemia cells and contributes to anergy

Correspondence: S. Deaglio silvia.deaglio@unito.it

Received: October 10, 2022.

1Laboratory of Functional Genomics, Department of Medical Sciences, University of Turin, Turin, Italy; 2School of Cancer and Pharmaceutical Sciences, King’s College London, London, UK; 3Division of Hematology, Department of Translational Medicine, University of Eastern Piedmont, Novara, Italy; 4Laboratory of Experimental Hematology, Institute of Oncology Research, Faculty of Biomedical Sciences, Università della Svizzera Italiana, Lugano, Switzerland; 5Department of Molecular Biotechnology and Health Sciences, University of Turin and Division of Hematology, A.O.U. Città della Salute e della Scienza di Torino, Turin, Italy; 6Hematology Unit, IRCCS Fondazione Policlinico Gemelli, Catholic University of "Sacred Heart", Rome, Italy; 7Hematology, P.O. "S. Luca", ASL Salerno, Salerno, Italy and 8Department of Hematology, Weill Cornell Medicine, New York, NY, USA.

Abstract

Accepted: January 11, 2023. Early view: January 19, 2023. https://doi.org/10.3324/haematol.2022.282177

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

T-cell immunoreceptor with Ig and ITIM domains (TIGIT) is an inhibitory checkpoint receptor that negatively regulates Tcell responses. CD226 competes with TIGIT for binding to the CD155 ligand, delivering a positive signal to the T cell. Here we studied the expression of TIGIT and CD226 in a cohort of 115 patients with chronic lymphocytic leukemia (CLL) and report expression of TIGIT and CD226 by leukemic cells. By devising a TIGIT/CD226 ratio, we showed that CLL cells favoring TIGIT over CD226 are typical of a more indolent disease, while those favoring CD226 are characterized by a shorter time to first treatment and shorter progression-free survival after first treatment. TIGIT expression was inversely correlated to the B-cell receptor (BCR) signaling capacity, as determined by studying BTK phosphorylation, cell proliferation and interleukin-10 production. In CLL cells treated with ibrutinib, in which surface IgM and BCR signaling capacity are temporarily increased, TIGIT expression was downmodulated, in line with data indicating transient recovery from anergy. Lastly, cells from patients with Richter syndrome were characterized by high levels of CD226, with low to undetectable TIGIT, in keeping with their high proliferative drive. Together, these data suggest that TIGIT contributes to CLL anergy by downregulating BCR signaling, identifying novel and actionable molecular circuits regulating anergy and modulating CLL cell functions.

Introduction

Chronic lymphocytic leukemia (CLL) is a B-cell malignancy characterized by clinical and molecular heterogeneity.1 The leukemic niche is critical to CLL development and progression in that it provides signals that influence CLL cell behavior,2,3 among which those channeled through the B-cell receptor (BCR) regulate key biological programs such as proliferation, metabolic adaptation and chemokine/cytokine secretion.4 BCR signaling capacity varies according to somatic hypermutation of the variable region of the immunoglobulin heavy chain region gene (IGHV). CLL samples harboring unmutated (UM) IGHV have stronger signaling ca-

pacity compared to mutated (M) cases that display a more anergic phenotype.5

CLL typically shows remarkable perturbations of both the innate and the adaptive immune responses, which are already evident from early stages of the disease and become severe in advanced/relapsed or therapy-resistant cases.6-10

Notably, leukemic cells play a pivotal role in shaping the microenvironment towards tolerance through multiple mechanisms.11 For example, circulating CLL cells share phenotypic features of regulatory B cells and secrete interleukin (IL)-10, which in turn affects T-cell responses.12

Interestingly, it was observed that IL-10 production is enhanced in anergic IGHV-M compared to IGHV-UM CLL

Francesca Arruga,1 Marta Rubin,1 Despoina Papazoglou,2 Andrea Iannello,1 Nikolaos Ioannou,2 Riccardo Moia,3 Davide Rossi,4 Gianluca Gaidano,3 Marta Coscia,5 Luca Laurenti,6 Giovanni D’Arena,7 John N. Allan,8 Richard R. Furman,8 Tiziana Vaisitti,1 Alan G. Ramsay2 and Silvia Deaglio1

cases, which are characterized by a more aggressive clinical course.13

The hypothesis behind this work is that the immunomodulatory molecule T-cell immunoreceptor with Ig and ITIM domains (TIGIT) can contribute to promote B-cell anergy and to shape the environment towards tolerance. TIGIT is an inhibitory receptor expressed on T, NK and NKT cells, sharing structural and mechanistic similarities with PD-1 and CTLA-4.14 The cytoplasmic tail contains an immunoglobulin tail tyrosine (ITT)-like phosphorylation motif and an ITIM domain, like PD-1, through which TIGIT recruits the phosphatase SHIP1 to inhibit downstream activation of NFκB, PI3K and MAPK pathways.15 TIGIT has a competing receptor, CD226/DNAM-1 (DNAX accessory molecule-1), resulting in opposite signaling outcomes upon binding to the same set of ligands, similar to what has been described for the CTLA-4/CD28 pair.16 The TIGIT/CD226 ligands belong to the nectin-family member poliovirus receptor (PVR), the best known of which is CD155. Signaling triggered upon CD155 binding to CD226 potentiates T-cell receptor (TCR) signaling and CD8+ T-cell cytotoxicity against tumor cells (positive signaling).17 In contrast, concomitant TIGIT expression on the cell surface prevents CD226 activation either by sequestering CD155 or by impeding CD226 homodimerization and phosphorylation, resulting in negative signaling.18 Whether TIGIT triggers a full inhibitory cascade or functions by preventing the CD226-mediated positive co-stimulatory signal remains unclear. Since the two receptors are co-expressed on the cell surface, a TIGIT/CD226 ratio is often preferred to highlight the imbalance towards a positive or a negative signaling outcome.19 Even though few data are available regarding TIGIT expression in the B-cell compartment, a recent paper has described the molecule on the surface of normal human memory B cells, where it directly suppresses T-cell responses.20

In CLL patients, TIGIT expression was shown to be progressively increased in the CD4+ T-cell compartment, reaching the highest levels in advanced stages of the disease. Functionally, TIGIT+/CD4+ T lymphocytes sustain CLL cell viability more efficiently than the TIGIT- counterpart and TIGIT inhibition interferes with the production of prosurvival cytokines by CD4+ T cells.21 However, no information is available on TIGIT expression in the leukemic cell compartment. The present study was undertaken to investigate expression of the TIGIT/CD226/CD155 axis in CLL, with a specific focus on leukemic B cells, and to explore the role of this axis in BCR activation.

Methods

Sample cohort

Peripheral blood samples from CLL patients were obtained

after informed consent, in accordance with Institutional Guidelines and the Declaration of Helsinki. The study was approved by the Institutional Review Board of each recruiting center. We examined a retrospective cohort of 115 CLL samples and 11 buffy coats from age- and sex-matched healthy subjects. The patients’ characteristics are summarized in Online Supplementary Table S1. Serial samples collected before treatment initiation and after 2 and 24 weeks of ibrutinib treatment were obtained from 14 additional patients (Online Supplementary Table S2). Clinical and molecular characteristics of CLL samples used in histological studies on lymph node biopsies are reported in Online Supplementary Table S3.

Xenografts derived from patients with Richter syndrome (RS-PDX) were obtained as described previously.22,23 Where indicated, primary CLL cells were cultured in RPMI 10% fetal calf serum in the presence or absence of ibrutinib used at 1 and 5 μM for 48 h.

Antibodies and reagents

A list of the antibodies used in flow cytometry and of the specific reagents used in functional assays is provided in Online Supplementary Table S4.

Flow cytometry

Surface expression of TIGIT, CD226 and CD155 was evaluated by flow cytometry on CLL peripheral blood mononuclear cells performing multiparametric staining to identify B or T lymphocytes and monocytes (Online Supplementary Table S5). Samples were acquired with a FACSCelesta cytometer (BD Biosciences) and data were analyzed with FlowJo v10 software (FlowJo).

Modulation of the TIGIT/CD226/CD155 axis

Modulation of the signaling triggered by CD155 binding either TIGIT or CD226 was performed both in the short term, to evaluate its inference on αIgM-mediated pBTK induction, and in the long term, to investigate the impact on CpG/IL-15-induced CLL proliferation. For the short-term experiments, cells were pre-treated in ice for 1 h with 5 µg/mL recombinant human (rh)TIGIT-Fc or with αTIGIT or αCD226 blocking monoclonal antibodies (5 µg/106 cells) for 30 min and then with rhCD155-Fc (5 µg/mL) for 1 h, before αIgM stimulation. For the long-term experiments, to prevent internalization, αTIGIT and αCD226 blocking monoclonal antibodies were coated onto magnetic beads and rhTIGIT-Fc and rhCD155-Fc chimeras were immobilized onto a cell culture plate. Briefly, 10x106 Dynabeads Sheep anti-Mouse IgG (Invitrogen, Thermofisher) were washed twice with phosphate-buffered saline, 0.1% bovine serum albumin and then coated with 1.5 μg of either antibody, by incubating overnight at 4°C on a rotating wheel, following the manufacturer’s instructions. Coated beads were used to treat CLL cells by pre-mixing them at a 2:1 bead:cell ratio

immediately before plating the cells in a 96-well plate. In parallel, 96-well plates were coated overnight at 4°C with 1 μg/well of rhTIGIT-Fc or rhCD155-Fc chimeras.

Statistical analyses

Statistical analyses were performed with GraphPad v7 (GraphPad Software Inc, La Jolla, CA, USA). A Mann-Whitney or Wilcoxon matched-pairs signed rank test was used to determine statistical significance. Contingency tests were performing using the Fisher test.

Methods for RNA extraction and quantitative real-time polymerase reaction, confocal microscopy, the phosflow assay, CpG/IL-15 stimulation and IL-10 production are entirely described in the Online Supplementary Material.

Results

TIGIT axis expression in peripheral blood mononuclear cells patients with chronic lymphocytic leukemia

Peripheral blood mononuclear cell preparations from 115 patients with a confirmed diagnosis of CLL (Online Supplementary Table S1) were tested for expression of TIGIT, CD226 and of the CD155 ligand and compared to preparations from a small cohort (n=11) of sex-matched healthy subjects of a comparable age. To this aim, we set up a multiparametric staining protocol for flow cytometry to analyze expression of the three molecules simultaneously on B and T lymphocytes and monocytes (Online Supplementary Table S5; Online Supplementary Figures S1 and S2). Leukemic B lymphocytes variably expressed TIGIT on the cell surface, whereas normal CD19+ B cells were uniformly negative (mean levels of expression were 21.22±21.97% of TIGIT+ cells in CLL samples vs. 0.97±0.47% in healthy subjects). CD226 was also expressed at significantly higher levels in CLL samples than in normal B cells (mean 24±12.9% in CLL vs. 15.6±5.2% in healthy subjects) (Figure 1A).

Histological analyses of lymph nodes confirmed low germinal center B-cell-associated TIGIT expression in reactive lymph node samples, while CLL lymph nodes showed higher levels of expression. Similar results were obtained when examining the co-staining of CD226 with CD20+ B cells (Figure 1B).

To complete the picture, we evaluated TIGIT axis expression on T lymphocytes and monocytes from CLL patients. The results indicate that TIGIT was variably expressed on CD4+ T lymphocytes with a significant increase in advanced stages (Rai II-IV) and IGHV-UM CLL, in line with previous data.21 In addition, CD8+ T cells were highly TIGIT+, marking an exhausted phenotype (Online Supplementary Figures S3A and S4A). Accordingly, CD226 expression on CD8+ T lymphocytes decreased with advanced CLL stages, concurrent with the acquisition of

further exhaustion (Online Supplementary Figures S3B and S4B). Lastly, monocytes showed the highest CD155 levels, suggesting that this cell lineage provides the ligand to engage either TIGIT or CD226. In line with a picture of progressive immune cell dysfunction, CD155 expression on monocytes decreased in Rai II-IV CLL patients (Online Supplementary Figures S3C and S5A-C).

We then correlated TIGIT expression on leukemic B cells with clinical and molecular features of the disease. We found that samples bearing markers of indolent disease or good prognosis (including Rai stage 0-I, normal karyotype or deletion 13 and IGHV-M genes) expressed TIGIT at significantly higher levels than the counterparts (Rai stage IIIV, trisomy 12, deletion 11 or deletion 17, and IGHV-UM genes) (Figure 1C). Lower TIGIT levels were also observed in NOTCH1-mutated cases, although this finding did not reach statistical significance. No differences were observed according to CD38 or CD49d levels (Online Supplementary Figure S6A).

CD226 had an opposite trend of expression, being associated with features of more aggressive disease, such as absence of somatic hypermutations in the IGHV genes, presence of NOTCH1 mutation and surface expression of CD38 and CD49, suggesting higher CD226 expression in CLL subsets that have a greater BCR signaling capacity compared to their counterparts (Figure 1C, Online Supplementary Figure S6A).

CD155 was generally present at low levels in CLL cells, without significant differences across disease subsets (Online Supplementary Figure S7).

Definition of an operational TIGIT:CD226 ratio

Considering that TIGIT and CD226 have opposing roles on the cell surface and compete for binding to the same ligand, we determined a TIGIT:CD226 ratio, based on percentage of expression. A TIGIT:CD226 ratio ≥1 indicates prevalence of TIGIT-expressing cells and consequently a predominance of negative signaling, whereas a TIGIT:CD226 ratio <1 indicates prevalence of CD226-expressing cells and therefore positive signaling. In line with this reasoning, CLL samples with a TIGIT:CD226 ratio ≥1 were enriched in the good prognosis subsets, while samples with a ratio <1 were more frequent in the presence of adverse prognostic markers (e.g., advanced stages, IGHV-UM, unfavorable cytogenetics, NOTCH1 mutations), confirming the validity of this approach (Figure 2A, Online Supplementary Figure S6B). Interestingly, prevalence of CD226 signaling, as defined by a TIGIT:CD226 ratio <1, correlates with significantly earlier time to first treatment and shorter progression-free survival after first-line therapy (Figure 2B).

TIGIT expression is associated with chronic lymphocytic leukemia anergy

To investigate the interplay between TIGIT and BCR sig-

Figure 1. TIGIT and CD226 are deregulated in chronic lymphocytic leukemia and differentially expressed among subsets of patients. (A) Percentages of TIGIT+ and CD226+ cells in 115 chronic lymphocytic leukemia (CLL) samples and 11 age- and sex-matched healthy donors. Statistical analysis: Student t test. (B) Representative multispectral immunofluorescence confocal images of non-malignant reactive (n=4) or CLL (n=6) lymph node formalin-fixed paraffin-embedded biopsy tissues for TIGIT or CD226 (red) expression in the lymph node microenvironment (CD20, white). Original magnification, x20, scale bar: 50 μm. (C) From left to right: percentages of TIGIT+ and CD226+ cells in samples stratified according to Rai stage and cytogenetic profile (top panels); percentages of TIGIT+ and CD226+ cells in samples stratified according to IGHV mutational status and to the presence of NOTCH1 mutations (bottom panels). Statistical analyses: Student t test. HD: healthy donor; RLN: reactive lymph node; CLL LN: CLL lymph node; M: mutated; UM: unmutated; WT: wildtype.

Figure 2. Ratio between TIGIT+ and CD226+ cells in chronic lymphocytic leukemia samples. (A) We calculated a ratio between the percentage of TIGIT+ and CD226+ cells in our cohort of chronic lymphocytic leukemia samples. A ratio ≥1 indicates prevalence of TIGIT+ cells and predominant negative signaling; a ratio <1 indicates prevalence of CD226+ cells and predominant positive signaling. For each clinical or molecular marker (Rai stage, IGHV mutational status, cytogenetics, NOTCH1 mutations) there is a dot plot showing ratio values for each sample (left) and a contingency plot indicating the enrichment of samples with a ratio ≥1 or <1 in either prognostic category. The dashed line at y=1 indicates the threshold discriminating between negative signaling (TIGIT:CD226 ratio ≥1, prevalence of TIGIT) and positive signaling (TIGIT:CD226 ratio <1, prevalence of CD226). Statistical analysis: Student t test. (B) Kaplan-Meier curves comparing the time to first treatment and progression-free survival (before and after treatment, respectively) of CLL patients divided according to TIGIT:CD226 ratio. Statistical analysis: Mantel-Cox test. M: mutated; UM: unmutated; WT: wildtype.

naling capacity, we selected a homogeneous subset of samples carrying IGHV-UM, with a normal fluorescence in situ hybridization profile or deletion 13q as a sole abnor-

mality and without NOTCH1 mutation, to avoid experimental biases.24 We found an inverse correlation between TIGIT expression and BCR signaling capacity, evaluated

analyzing baseline surface IgM levels and BTK phosphorylation in response to IgM crosslinking as read outs (Figure 3A, left panels). Moreover, when splitting CLL samples according to TIGIT:CD226 ratio, we confirmed that samples with a ratio <1 (TIGIT-) had significantly higher surface IgM levels and stronger phospho-BTK induction than cases with a ratio ≥1 (TIGIT+) (Figure 3A, right panels).

We next examined whether high TIGIT levels were associated with weaker responses of CLL cells to activation/proliferation signals, such as CpG/IL-15, analyzed in the same CLL subset used to test BCR signaling. In line with the results obtained in the BCR signaling studies, when dividing samples according to TIGIT:CD226 ratio, samples with prevalence of TIGIT showed a lower proliferative response to CpG/IL-15 compared to samples with a ratio <1 (Figure 3B, upper panel).

Analysis of TIGIT and CD226 expression in these cells, after 6 days of exposure to CpG/IL-15, showed a marked upregulation of CD226, with a concomitant slight downmodulation of TIGIT (Online Supplementary Figure S8A). Unstimulated cells showed a high level of spontaneous apoptosis; in the remaining live cells, CD226 expression was downregulated compared to the expression levels before starting in vitro culture (Figure 3B, lower panel). Modulation of CD226 and TIGIT levels could explain, at

least in part, the observation that TIGIT+ CLL samples showed a productive proliferative response, albeit weaker than TIGIT- samples. Accordingly, the TIGIT:CD226 ratio in stimulated cells was <1 in all samples, in line with a prevalence of “positive” signaling (Online Supplementary Figure S8B).

In line with these findings, we found that CLL lymph node biopsies with higher TIGIT expression showed lower CD226 levels (Figure 3C). These tissues samples had a significantly lower expression of the proliferation marker Ki67, when compared to samples showing low TIGIT and high CD226. Consistently, we found that CD226+ CLL cells were mainly associated with Ki67 expression regardless of TIGIT levels, which was significant in TIGITlow, CD226high lymph nodes as revealed by Ki67/TIGIT versus Ki67/CD226 co-localization image analysis (Figure 3C).

TIGIT+ chronic lymphocytic leukemia cells produce more interleukin-10

Considering that IGHV-M anergic CLL cells produce and secrete more IL-10 than IGHV-UM reactive cells,13 we investigated whether high TIGIT expression correlated with IL-10 production. We found that samples with a TIGIT:CD226 ratio ≥1 had significantly higher IL10 mRNA levels, both in the IGHV-UM and the IGHV-M CLL groups

(A) Top panels. Inverse correlation between surface IgM levels and the percentage of TIGIT-expressing cells in chronic lymphocytic leukemia (CLL) samples harboring unmutated IGHV and with a normal karyotype or deletion 13 as the sole abnormality (left), and surface IgM levels in CLL samples divided according to TIGIT:CD226 ratio (right). Bottom panels. Inverse correlation between the induction of BTK phosphorylation (pBTK) upon B-cell receptor stimulation and the TIGIT:CD226 ratio in CLL samples (left), and fold changes of αIgM-mediated pBTK induction in CLL samples divided according to TIGIT:CD226 ratio (right). (B) Ki67 staining of TIGIT+ and TIGITCLL samples (top) and flow cytometry analysis of surface CD226 upregulation (bottom) in response to CpG/IL-15 culture. (C) Representative multispectral immunofluorescence and three-dimensional volume rendered confocal images of TIGIThigh/CD226low (n=3) or TIGITlow/CD226high (n=3) CLL lymph node formalin-fixed paraffin-embedded biopsy tissues stained for CD20 (blue), Ki67 (magenta) and TIGIT or CD226 (green). Original magnification, x20, scale bars of the larger image: 100 μm, of the lower image: 50 μm. The images on the right represent a magnification of the top image, as indicated by the arrow. Quantification of the co-localization of Ki67 and TIGIT or Ki67 and CD226 limited to CD20+ cells from CLL lymph node tissues. Graphs relative to the quantification in TIGIThigh/CD226low lymph nodes (top), TIGITlow/CD226high lymph nodes (middle) or all the lymph node samples together (bottom) are shown. Statistical analyses: Student t test. MFI: mean fluorescent intensity; FC/NS: fold-change compared to unstimulated sample; LN: lymph nodes.

(Figure 4A). Similar results were obtained when measuring IL-10 production by flow cytometry, with TIGIT+ CLL cells staining more positive than TIGIT- cells for IL-10 after 5 h stimulation with phorbol 12-myristate 13-acetate and ionomycin, even in IGHV-UM cases (Figure 4B, C). The observation that IGHV-M cells showed higher IL10 expression and production, at the mRNA and at the protein levels, re-

spectively, both in the TIGIT+ and in the TIGIT- subsets, suggests that TIGIT is associated with different IL-10 profiles but also that other regulatory mechanisms exist.

TIGIT axis expression during disease follow-up

We next studied modulation of surface TIGIT, CD226 and CD155 over time, focusing specifically on the effects ex-

Figure 3. High surface TIGIT expression is associated with chronic lymphocytic leukemia cell anergy.

erted by BTK inhibitors, since they are known to modulate BCR signaling. To this aim, we took advantage of a cohort of 14 samples collected systematically before ibrutinib initiation, after 2 weeks and after 24 weeks of treatment. The characterization of these samples is reported in Online Supplementary Table S2 . Previous studies have shown that BTK blockade is followed by upregulation of surface IgM levels, evident already after 1 week on therapy and maintained for at least 3 months.25 This apparently paradoxical behavior was also observed in our cohort, in which sIgM levels were increased after 2 weeks of treatment and remained higher than the baseline at 24 weeks (Figure 5A). In line with heightened BCR signaling

activity, surface TIGIT, which was expressed before therapy initiation by all samples, invariably decreased upon ibrutinib treatment, starting within the first 2 weeks of treatment and reaching minimal levels at the 24-week time-point (Figure 5B). TIGIT mRNA levels showed the same behavior ( Online Supplementary Figure S9A ). In contrast, mRNA levels of CD226 increased markedly (Online Supplementary Figure S9A), while surface levels were minimally, but signi fi cantly decreased (Figure 5C, left panel), raising the question of whether the molecule can reach the cell surface. However, given the relative changes of TIGIT and CD226, their ratio dropped to <1 in all samples examined, including those with a ratio ≥1 be-

Figure 4. TIGIT+ cells produce more IL-10. (A) Quantitative real-time polymerase chain reaction analysis of IL10 baseline expression of chronic lymphocytic leukemia (CLL) samples divided according to IGHV mutational status and TIGIT surface levels. (B) FACS analysis of IL-10 intracellular staining after 5 h stimulation with phorbol 12-myristate 13-acetate (50 ng/mL) and 1 μM ionomycin in CLL samples divided according to IGHV mutational status and TIGIT surface levels. (C) Representative flow cytometry plots of IL-10 production in unstimulated and stimulated CLL cells. Statistical analyses: Student t test. RE: relative expression; IL-10: interleukin-10: B2M: β2-microglobulin; UM: unmutated: M: mutated; NS: unstimulated; PMA/Iono: stimulated with phorbol 12-myristate 13-acetate/ionomycin.

fore treatment, suggesting a switch towards “positive” CD226-mediated signaling (Figure 5C, right panel). TIGIT downregulation appeared specific to B cells, as we found that ibrutinib treatment did not alter its expression on CD4+ or CD8+ T-cell subsets (Online Supplementary Figure S9C, D). In contrast, CD155 surface expression decreased minimally along the patients’ follow-up (Online Supplementary Figure S9B).

This response to ibrutinib was also documented in vitro, by exposing primary CLL cells from untreated patients to ibrutinib (1 μM and 5 μM for 48 h). In these cells, we observed increased sIgM levels, with a concomitant decrease of the TIGIT:CD226 ratio (Figure 5D, Online Supplementary Figure S9E). Previous investigators have reported that upon BTK inhibition, the BCR retains the capability to mobilize Ca2+ in response to antibody ligation, as well as the capability of tyrosine phosphorylating SYK and ERK1/2.26 In our samples, despite inhibition of tyrosine phosphorylation of BTK, a marked increase in intracellular Ca2+ mobilization upon BCR ligation was observed, comparing untreated versus ibrutinib-treated primary CLL cells. The same cells showed prominent tyrosine phosphorylation of SYK and ERK1/2, clearly indicating that the BCR pathway is bypassing the signaling block imposed by ibrutinib (Figure 5E).

TIGIT and CD226 are expressed on Richter syndrome samples

Stemming from these observations, we evaluated expression of TIGIT and CD226 in cases of Richter syndrome (RS), a rare but often fatal complication of CLL characterized by transformation of the leukemia into an aggressive lymphoma.27-29 To this purpose, we exploited RNA-sequencing analysis performed on primary formalin-fixed paraffin-embedded RS lymph nodes and compared it to that of CLL samples and matched healthy subjects, from previously published datasets (EGA accession numbers EGAD00001004046 and EGAD00001000258).30,31 RS samples showed lower TIGIT expression compared to CLL samples, in line with our observation of TIGIT marking a more indolent disease. Accordingly, CD226 expression in RS cells was higher than in either cells from healthy donors or CLL samples (Figure 6A). Results were substantiated by using RS-patient-derived xenograft models recently established in our laboratory.22,23 Both quantitative polymerase reaction performed on four RS-PDX at different passages and flow cytometry analyses confirmed the RNA-sequencing results, with TIGIT being expressed at lower levels than in CLL cases and CD226 showing the highest expression (Figure 6B, C). Interestingly, all these four models show a highly active BCR signaling pathway.32

Figure 5. TIGIT and CD226 expression during the follow-up of chronic lymphocytic leukemia patients treated with ibrutinib. (A) Surface IgM levels before and during ibrutinib therapy. (B) Surface expression of TIGIT before treatment initiation, and after 2 and 24 weeks of ibrutinib therapy (C) (Left panel) Surface expression of CD226 and (Right panel) TIGIT:CD226 ratio before treatment initiation, and after 2 and 24 weeks of ibrutinib therapy. The dashed line at y=1 indicates the threshold discriminating between negative signaling (TIGIT:CD226 ratio ≥1, prevalence of TIGIT) and positive signaling (TIGIT:CD226 ratio <1, prevalence of CD226). Statistical analyses (A-C): one-way analysis of variance. (D) (Left panel) Surface IgM levels of chronic lymphocytic leukemia (CLL) cells treated in vitro with 1 and 5 μM ibrutinib for 48 h. (D) (Right panel) TIGIT:CD226 ratio measured in primary CLL cells in the presence or in the absence of ibrutinib used at 1 and 5 μM for 48 h. The dashed line at y=1 indicates the threshold discriminating between negative and positive signaling. (E) Phospho-BTK MFI levels in response to anti-IgM ligation in primary CLL cells left untreated or exposed to 1 μM ibrutinib for 48 h (left panel). In the same cells, intracellular Ca2+ levels were monitored by flow cytometry (middle panel) and SYK and ERK1/2 phosphorylation by western blot (right panel). (D, E) Statistical analyses: Student t test. MFI: mean fluorescence intensity; UnTX: before treatment; PreTX: before treatment; w: weeks; NT: untreated; FC/NS: fold change compared to unstimulated samples; Ibr: ibrutinib.

CD155 transcript levels were barely detectable in RS samples, although the molecule was expressed on the cell surface (Online Supplementary Figure S10).

Modulation of TIGIT and CD226 signaling

To understand the functional role of TIGIT and CD226 in altering BCR signaling capacity and cell proliferation, we selectively interfered with either receptor/ligand interaction, taking advantage of specifi c rhFc chimeras or blocking monoclonal antibodies. Read-outs were BTK phosphorylation in response to α IgM-mediated BCR crosslinking or Ki67 staining following CpG/IL-15 culture, respectively. A schematic representation of the experiments and reagents is depicted in Figure 7A. Specifically, we used: (i) a rhTIGIT-Fc chimera that sequesters CD155

expressed on the cell surface and prevents it from binding either TIGIT or CD226, blocking downstream signaling of the prevalent receptor in that CLL population; (ii) a rhCD155-Fc chimera that works as an artificial ligand and can bind and activate both TIGIT and CD226, depending on which receptor is prevalent; and (iii) the rhCD155-Fc chimera in combination with blocking monoclonal antibodies directed against either TIGIT or CD226 (αTIGIT or αCD226) in order to discriminate between the activation of one receptor and the other.

Experiments were carried out in the same IGHV-UM CLL cases with a normal fluorescence in situ hybridization profile or deletion 13 used in previous experiments. Pretreatment of TIGIT- CLL cells with rhTIGIT-Fc affects mostly CD226 signaling, which is more expressed than

TIGIT in these samples. We found that the BCR signaling capacity upon receptor engagement in the presence of rhTIGIT-Fc was significantly downregulated in TIGITsamples compared to when α IgM was given alone, and phospho-BTK induction was similar to that of TIGIT+ samples. In contrast, pre-treatment of TIGIT+ CLL cells with rhTIGIT-Fc mostly prevents CD155 binding to TIGIT, likely abrogating its negative regulation of the BCR. Accordingly, in these samples, α IgM-induced BTK phosphorylation was significantly increased in the presence of rhTIGIT-Fc, and was more similar to that of TIGIT- samples. Furthermore, stimulation of the BCR in the presence of rhCD155-Fc and of α TIGIT blocking antibody enhanced BTK phosphorylation in TIGIT+ samples, while this combination had no effects on TIGIT- samples, which already had maximal phospho-BTK induction. Lastly, pre-treatment of CLL cells with rhCD155-Fc and αCD226 blocking antibody downmodulated the BCR signaling capacity of TIGIT- samples, in which CD226 is prevalent and could help to improve BCR responses, to levels comparable to those of TIGIT+ CLL samples, whose phospho-BTK induction was unaffected in the presence of agents preventing CD226 signaling (Figure 7B).

Similar results were obtained when examining the proliferative response to CpG/IL-15 stimulation in the same conditions (Figure 7C).

These results suggest that CD155 binding to TIGIT triggers an inhibitory signaling that decreases responsiveness of CLL cells to the antigen, and that, if CD155-TIGIT inter-

actions are interrupted, we can boost CLL cell responses. On the other hand, CD155 binding to CD226 exerts an opposite “positive” effect on intracellular signaling and interrupting this axis might induce CLL cell anergy.

Discussion

This work shows that the immunomodulatory molecule TIGIT is expressed by CLL cells where it is a marker of anergy. TIGIT was identified nearly a decade ago and shown to be part of an axis including CD226 and CD155 which shares similarities with other checkpoint inhibitors.33 In the current view, TIGIT and CD226 are expressed by T cells and can negatively (TIGIT) or positively (CD226) affect TCR signaling, once engaged by the common CD155 ligand. Coherent with this view, T cells expressing high levels of TIGIT are reported in different cancers where they define a subset of exhausted and dysfunctional T lymphocytes.21,34-36 In CLL, high TIGIT expression is found on T lymphocytes from patients with advanced disease, co-expressing exhaustion markers21 (Online Supplementary Figure S6).

The recent finding of TIGIT expression on normal memory B cells, where it directly contributes to suppress T-cell responses,20 prompted us to extend these observations to CLL cells. Here, we show for the first time that circulating and resident leukemic B cells express TIGIT and CD226, at variance with the normal CD19+ subset. CD155, in

B

Figure 6. TIGIT and CD226 expression in Richter syndrome. (A) Values of TIGIT and CD226 from RNA-sequencing experiments carried out in normal B cells, chronic lymphocytic leukemia (CLL) samples and primary formalin-fixed paraffin-embedded lymph nodes from Richter syndrome (RS) samples. (B) Quantitative real-time polymerase chain reaction analysis of TIGIT and CD226 expression in our cohort of healthy subjects, CLL samples and RS-patient-derived xenograft (PDX) models at different passages. (C) Flow cytometry analysis of TIGIT and CD226 surface expression in healthy subjects, CLL and RS-PDX. Statistical analyses: Student t test. TPM: transcripts per million; HD: healthy donors; RE: relative expression; B2M: β2-microglobulin.

Figure 7. Modulation of TIGIT and CD226 interactions with CD155. (A) Schematic representation of the mechanisms of action of rhTIGIT-Fc and rhCD155-Fc chimeras and of αTIGIT and αCD226 blocking monoclonal antibodies: (i) CD155 can bind either TIGIT or CD226, triggering opposite signaling outcomes; (ii) rhTIGIT-Fc chimera prevents CD155 binding and can therefore inhibit both TIGIT and CD226 signaling, thus affecting the signaling through the prevalent receptor on the cell surface. rhCD155-Fc chimera works as an artificial ligand and can bind to either receptor; therefore, (iii) when giving it in combination with the α TIGIT blocking monoclonal antibody it is possible to induce signaling through CD226; while (iv) in combination with α CD226 monoclonal antibody, signaling through TIGIT is preserved. (B) Flow cytometry analysis of pBTK induction in response to αIgM-mediated B-cell receptor crosslinking in the presence of modulators of TIGIT and CD226 activity: top panels show plots of two representative TIGIT+ and TIGIT- samples, bottom panel shows cumulative results of pBTK induction. (C) Cumulative results of Ki67 staining upon CpG/IL-15 culture in the presence of modulators of TIGIT and CD226 activity. Statistical analyses: Student t test. MFI: mean fluorescent intensity.

contrast, was mostly expressed on the monocyte compartment. When dividing our cohort of 115 patients according to specific prognostic markers, we observed that high TIGIT was associated with features of indolent disease while high CD226 was more frequent in subsets characterized by elevated BCR signaling capacity (e.g., IGHV-UM, NOTCH1-M or CD38+ and CD49d+ cases). Since TIGIT and CD226 are concomitantly present on the cell surface, we devised a ratio between the two markers: a ratio ≥1 indicates predominance of TIGIT and hence of inhibitory effects (negative signaling), while a ratio in favor of CD226 prompts for a co-stimulatory effect (positive signaling). Accordingly, aggressive cases of CLL were enriched with samples showing a ratio in favor of CD226. It is therefore likely that this axis might modulate signaling of CLL cells, similarly to what is observed in T lymphocytes.

To determine a possible role for TIGIT in CLL homeostasis, we first explored the effects on BCR signaling capacity, specifically focusing on IGHV-UM samples, selected to harbor high or low surface TIGIT. We found an inverse correlation between TIGIT expression and baseline sIgM levels or the induction of BTK phosphorylation in response to receptor engagement, suggesting that surface TIGIT is associated with a more anergic CLL behavior. In a cohort of samples collected systematically after ibrutinib therapy, we observed a sharp decrease in surface TIGIT following treatment. This is in line with the recent observation that leukemic cells, released from the lymph nodes by ibrutinib, upregulate sIgM and SYK because they no longer receive persistent antigen stimulation, as if they turned less anergic despite downstream inhibition of BTK.25 The mechanism behind TIGIT downregulation remains to be determined. Speculatively, it could rely on the inhibition of transcription factors downstream to the BCR signaling, including NFATC1, 37 FOXP1 38 and NFKB, 39 which have putative binding sites on the TIGIT promoter (not shown, prediction made using the CiiiDER online tool at http://www.ciiider.org/).40

TIGIT downregulation with concomitant surface IgM upregulation were also confirmed by in vitro exposure of primary CLL cells to ibrutinib. While BTK tyrosine phosphorylation was invariably inhibited in these cells, BCR ligation was followed by SYK and ERK1/2 tyrosine phosphorylation and mobilization of intracellular Ca2+ to levels higher than those observed in untreated cells, indicating recovery from anergy. This behavior was previously attributed to the interruption of chronic antigen stimulation due to release of CLL cells exposed to ibrutinib from the lymph node niche, at the same time making them more dependent on BCR engagement and consequently more susceptible to apoptosis if the ligand is not present, as is the case for peripheral circulation.25

A formal demonstration of the effects of TIGIT and CD226

on BCR signaling capacity comes from experiments in which interactions of these receptors with the CD155 ligand were interrupted using specific recombinant chimeras and monoclonal antibodies. rhTIGIT-Fc chimera sequesters CD155 and prevents its binding to either receptor, thus affecting the signaling through the prevalent receptor on the cell surface (Figure 7Aii). Therefore, in TIGIT- samples, blocking CD155 mostly affects signaling through CD226, removing its positive contribution to BCR signaling and, consistently, we observed a reduced αIgMinduced BTK phosphorylation. In contrast, in TIGIT+ samples, TIGIT signaling is affected by CD155 sequestering, removing its inhibitory effect and increasing BTK phosphorylation. Comparable results were achieved when selectively activating CD226 or TIGIT by providing chimeric CD155 ligand together with a monoclonal antibody blocking the unintended receptor (Figure 7Aiii-iv). Using the same experimental set-up, we examined CLL cell proliferation in response to CpG/IL-15 stimulation and observed signi fi cant differences between TIGIT + and TIGIT- samples, with the latter showing a proliferative advantage over the former. Again, when interrupting TIGIT/CD226 interactions with CD155 we could modulate responses to CpG/IL-15 with different outcomes in TIGIT+ and TIGIT- samples.

Lastly, when analyzing IL-10 secretion in our sample cohort, we found that TIGIT+ cases of CLL produce more IL10 than do TIGIT- ones, both in the IGHV -M and in the IGHV-UM subsets. This finding is in line with previous observations that TIGIT+ normal memory B cells suppress T-cell responses more efficiently than the TIGIT- counterpart, possibly via IL-10,20 and also with existing literature showing that IL-10 production is enhanced in more anergic CLL and is associated with a less aggressive clinical phenotype.13,41

Our results indicate that TIGIT and CD226 are aberrantly expressed on leukemic B cells, and this is the first time that deregulation of this axis has been described on tumor cells and not only in the T-cell compartment. In addition, this work provides evidence of an association between TIGIT expression and an anergic phenotype of the CLL cell. The mechanism behind TIGIT upregulation in CLL is still not understood. However, a recent paper reported a signature of aberrantly expressed immune regulatory molecules, including TIGIT, in CLL cells with a peculiar methylation pattern compared to that of healthy B lymphocytes.42

The translational implications of these results remain to be determined. Since therapeutic anti-TIGIT antibodies are in clinical trials for cancer patients, it would be tempting to determine whether, in CLL patients, they may revert anergy, increasing BCR signaling capacity, and therefore making leukemic cells more susceptible to targeted inhibitors. Further research will tell us more about

this immunoregulatory pathway and its possible clinical implications.

Disclosures

No conflicts of interest to disclose.

Contributions

FA designed the study, performed experiments, analyzed and interpreted data and together with SD wrote the paper. MR, DP and AI performed experiments. RM, DR, GG, MC, LL, GD’A, JNA and RRF provided patients’ samples and relevant clinical information and contributed to data interpretation. DP, NI and AGR performed confocal microscopy experiments and analyses on lymph node tissue biopsies collected at King’s College (London, UK). TV discussed results and contributed to data interpretation. SD designed the study, interpreted data and together with FA wrote the paper.

References

1. Chiorazzi N, Rai KR, Ferrarini M. Chronic lymphocytic leukemia. N Engl J Med. 2005;352(8):804-815.

2. Puente XS, Jares P, Campo E. Chronic lymphocytic leukemia and mantle cell lymphoma: crossroads of genetic and microenvironment interactions. Blood. 2018;131(21):2283-2296.

3. Herishanu Y, Perez-Galan P, Liu D, et al. The lymph node microenvironment promotes B-cell receptor signaling, NFkappaB activation, and tumor proliferation in chronic lymphocytic leukemia. Blood. 2011;117(2):563-574.

4. Stevenson FK, Krysov S, Davies AJ, Steele AJ, Packham G. Bcell receptor signaling in chronic lymphocytic leukemia. Blood. 2011;118(16):4313-4320.

5. Packham G, Krysov S, Allen A, et al. The outcome of B-cell receptor signaling in chronic lymphocytic leukemia: proliferation or anergy. Haematologica. 2014;99(7):1138-1148.

6. Forconi F, Moss P. Perturbation of the normal immune system in patients with CLL. Blood. 2015;126(5):573-581.

7. Peters FS, Strefford JC, Eldering E, Kater AP. T-cell dysfunction in chronic lymphocytic leukemia from an epigenetic perspective. Haematologica. 2021;106(5):1234-1243.

8. Vlachonikola E, Stamatopoulos K, Chatzidimitriou A. T cells in chronic lymphocytic leukemia: a two-edged sword. Front Immunol. 2020;11:612244.

9. Fiorcari S, Maffei R, Atene CG, Potenza L, Luppi M, Marasca R. Nurse-like cells and chronic lymphocytic leukemia B cells: a mutualistic crosstalk inside tissue microenvironments. Cells. 2021;10(2):217.

10. Griggio V, Perutelli F, Salvetti C, et al. Immune dysfunctions and immune-based therapeutic interventions in chronic lymphocytic leukemia. Front Immunol. 2020;11:594556.

11. Arruga F, Gyau BB, Iannello A, Vitale N, Vaisitti T, Deaglio S. Immune response dysfunction in chronic lymphocytic leukemia: dissecting molecular mechanisms and microenvironmental conditions. Int J Mol Sci. 2020;21(5):1825.

12. DiLillo DJ, Weinberg JB, Yoshizaki A, et al. Chronic lymphocytic leukemia and regulatory B cells share IL-10 competence and immunosuppressive function. Leukemia. 2013;27(1):170-182.

13. Drennan S, D'Avola A, Gao Y, et al. IL-10 production by CLL cells is enhanced in the anergic IGHV mutated subset and associates

Acknowledgments

The authors thank the Nikon Imaging Facility at King’s College London.

Funding

This work was supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC IG-23095 to SD; My First AIRC grant MFAG-23107 to TV; AIRC 5x1000 #21198 to GG), by the ITN INTEGRATA program (grant agreement 813284 to SD), by the Italian Ministry of Health (GR-2016-02364298 to TV), by the Ministry of Education, Univ ersity and Research-MIUR and “Progetto Strategico di Eccellenza Dipartimentale” (D15D18000410001 to SD as part of the Department of Medical Sciences, University of Turin).

Data-sharing statement

The authors adhere to the policy of data sharing.

with reduced DNA methylation of the IL10 locus. Leukemia. 2017;31(8):1686-1694.

14. Yu X, Harden K, Gonzalez LC, et al. The surface protein TIGIT suppresses T cell activation by promoting the generation of mature immunoregulatory dendritic cells. Nat Immunol. 2009;10(1):48-57.

15. Liu S, Zhang H, Li M, et al. Recruitment of Grb2 and SHIP1 by the ITT-like motif of TIGIT suppresses granule polarization and cytotoxicity of NK cells. Cell Death Differ. 2013;20(3):456-464.

16. Le Mercier I, Lines JL, Noelle RJ. Beyond CTLA-4 and PD-1, the generation Z of negative checkpoint regulators. Front Immunol. 2015;6:418.

17. Chan CJ, Andrews DM, McLaughlin NM, et al. DNAM-1/CD155 interactions promote cytokine and NK cell-mediated suppression of poorly immunogenic melanoma metastases. J Immunol. 2010;184(2):902-911.

18. Johnston RJ, Comps-Agrar L, Hackney J, et al. The immunoreceptor TIGIT regulates antitumor and antiviral CD8(+) T cell effector function. Cancer Cell. 2014;26(6):923-937.

19. Deng C, Li W, Fei Y, et al. Imbalance of the CD226/TIGIT immune checkpoint is involved in the pathogenesis of primary biliary cholangitis. Front Immunol. 2020;11:1619.

20. Hasan MM, Nair SS, O'Leary JG, et al. Implication of TIGIT(+) human memory B cells in immune regulation. Nat Commun. 2021;12(1):1534.

21. Catakovic K, Gassner FJ, Ratswohl C, et al. TIGIT expressing CD4+ T cells represent a tumor-supportive T cell subset in chronic lymphocytic leukemia. Oncoimmunology. 2017;7(1):e1371399.

22. Vaisitti T, Braggio E, Allan JN, et al. Novel Richter syndrome xenograft models to study genetic architecture, biology, and therapy responses. Cancer Res. 2018;78(13):3413-3420.

23. Vaisitti T, Arruga F, Vitale N, et al. ROR1 targeting with the antibody-drug conjugate VLS-101 is effective in Richter syndrome patient-derived xenograft mouse models. Blood. 2021;137(24):3365-3377.

24. Arruga F, Bracciama V, Vitale N, et al. Bidirectional linkage between the B-cell receptor and NOTCH1 in chronic lymphocytic leukemia and in Richter's syndrome: therapeutic

implications. Leukemia. 2020;34(2):462-477.

25. Drennan S, Chiodin G, D'Avola A, et al. Ibrutinib therapy releases leukemic surface IgM from antigen drive in chronic lymphocytic leukemia patients. Clin Cancer Res. 2019;25(8):2503-2512.

26. Chiodin G, Drennan S, Martino EA, et al. High surface IgM levels associate with shorter response to ibrutinib and BTK bypass in patients with CLL. Blood Adv. 2022;6(18):5494-5504.

27. Tadmor T, Levy I. Richter transformation in chronic lymphocytic leukemia: update in the era of novel agents. Cancers (Basel). 2021;13(20):5141.

28. Condoluci A, Rossi D. Richter syndrome. Curr Oncol Rep. 2021;23(3):26.

29. Rossi D, Spina V, Gaidano G. Biology and treatment of Richter syndrome. Blood. 2018;131(25):2761-2772.

30. Beekman R, Chapaprieta V, Russinol N, et al. The reference epigenome and regulatory chromatin landscape of chronic lymphocytic leukemia. Nat Med. 2018;24(6):868-880.

31. Ferreira PG, Jares P, Rico D, et al. Transcriptome characterization by RNA sequencing identifies a major molecular and clinical subdivision in chronic lymphocytic leukemia. Genome Res. 2014;24(2):212-226.

32. Iannello A, Vitale N, Coma S, et al. Synergistic efficacy of the dual PI3K-delta/gamma inhibitor duvelisib with the Bcl-2 inhibitor venetoclax in Richter syndrome PDX models. Blood. 2021;137(24):3378-3389.

33. Ge Z, Peppelenbosch MP, Sprengers D, Kwekkeboom J. TIGIT, the next step towards successful combination immune checkpoint therapy in cancer. Front Immunol. 2021;12:699895.

34. Freed-Pastor WA, Lambert LJ, Ely ZA, et al. The CD155/TIGIT axis promotes and maintains immune evasion in neoantigenexpressing pancreatic cancer. Cancer Cell.

2021;39(10):1342-1360.e14.

35. Chauvin JM, Pagliano O, Fourcade J, et al. TIGIT and PD-1 impair tumor antigen-specific CD8(+) T cells in melanoma patients. J Clin Invest. 2015;125(5):2046-2058.

36. Shao Q, Wang L, Yuan M, Jin X, Chen Z, Wu C. TIGIT induces (CD3+) T cell dysfunction in colorectal cancer by inhibiting glucose metabolism. Front Immunol. 2021;12:688961.

37. Wolf C, Garding A, Filarsky K, et al. NFATC1 activation by DNA hypomethylation in chronic lymphocytic leukemia correlates with clinical staging and can be inhibited by ibrutinib. Int J Cancer. 2018;142(2):322-333.

38. Cerna K, Oppelt J, Chochola V, et al. MicroRNA miR-34a downregulates FOXP1 during DNA damage response to limit BCR signalling in chronic lymphocytic leukaemia B cells. Leukemia. 2019;33(2):403-414.

39. Rozovski U, Harris DM, Li P, et al. Activation of the B-cell receptor successively activates NF-kappaB and STAT3 in chronic lymphocytic leukemia cells. Int J Cancer. 2017;141(10):2076-2081.

40. Gearing LJ, Cumming HE, Chapman R, et al. CiiiDER: a tool for predicting and analysing transcription factor binding sites. PLoS One. 2019;14(9):e0215495.

41. Hanna BS, Llao-Cid L, Iskar M, et al. Interleukin-10 receptor signaling promotes the maintenance of a PD-1(int) TCF-1(+) CD8(+) T cell population that sustains anti-tumor immunity. Immunity. 2021;54(12):2825-2841.e10.

42. Wierzbinska JA, Toth R, Ishaque N, et al. Methylome-based cellof-origin modeling (Methyl-COOM) identifies aberrant expression of immune regulatory molecules in CLL. Genome Med. 2020;12(1):29.

The E3 ligase TRIM31 regulates hematopoietic stem cell homeostasis and MLL-AF9 leukemia

Kai Zhang,1,* Dingdong Liu,1,* Yafei Li,1 Zhencan Shi,2 Jun Guo,2 Chengjiang Gao,3 Hu Wang,1 Zhenyu Ju1 and Daojun Diao1

1Key Laboratory of Regenerative Medicine of Ministry of Education, Institute of Aging and Regenerative Medicine, Jinan University, Guangzhou; 2Department of Cardiology, the First Affiliated Hospital of Jinan University, Guangzhou and 3Department of Immunology and Key Laboratory of Infection, Immunity of Shandong Province, Shandong University School of Basic Medical Sciences, Jinan, China.

*KZ and DL contributed equally as first authors.

Abstract

Correspondence: D. Diao diaodaojun@hotmail.com

Z. Ju zhenyuju@163.com

Received: August 18, 2022.

Accepted: December 28, 2022.

Early view: January 12, 2023.

https://doi.org/10.3324/haematol.2022.281955

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Hematopoietic stem cells (HSC) are kept in a quiescent state to maintain their self-renewal capacity. Proper regulation of cyclin-dependent kinases (CDK) and cyclin proteins is critical for the maintenance of HSC homeostasis. Here, we found that the E3 ligase, TRIM31, regulates HSC homeostasis and leukemia through the accumulation of CDK8. TRIM31 deficiency promotes hematopoietic stem and progenitor cell proliferation and long-term HSC exhaustion. Serial competitive transplantation assays showed that TRIM31-deficient HSC exhibit impaired reconstitution ability. TRIM31 loss led to a lower rate of survival of mice under conditions of stress (5-fluorouracil administration), which was correlated with a lower number of hematopoietic stem and progenitor cells. In a murine model of acute myeloid leukemia, the initiation of leukemia was significantly accelerated upon TRIM31 deletion. Mechanistically, we found that ubiquitin-mediated degradation of CDK8 was impaired by TRIM31 deletion, which further induced transcriptional expression of PBX1 and cyclin D1. Taken together, these findings reveal the function of TRIM31 in the regulation of HSC homeostasis and leukemia initiation, and indicate the physiological importance of TRIM31 in the early stage of the development of leukemia.

Introduction

Under steady conditions, hematopoietic stem cells (HSC) are in a quiescent state to prevent exhaustion of the cells and to restrict the occurrence of replication-associated mutations.1 An uncontrolled self-renewal program disrupts stem cell maintenance, and over-proliferation of stem cells often leads to HSC exhaustion and leukemia.2 Quality control of the cell cycle is achieved through cell-intrinsic regulators, including transcription factors, signal transducers, cell-cycle inhibitors, surface receptors, and cellextrinsic factors, such as the bone marrow niche and cytokines. Notably, cell-intrinsic regulators can be manipulated by the ubiquitin-proteasome degradation system at the protein level.

Ubiquitin E3 ligases play a critical role in the hematopoietic system. Loss of E3 ligases (including c-Cbl, Itch and SCFFbxw7) may affect HSC quiescence and lead to

stem cell expansion, myeloid proliferative disorders and acute myeloid leukemia (AML).3-5 As E3 ligases, TRIM proteins are widely involved in various cell processes, such as cell proliferation, differentiation, development and apoptosis. Some of them, notably TRIM19 and TRIM33, are vital for the maintenance and function of HSC.6,7 Other members of the TRIM family, such as TRIM13 and TRIM24, have also been reported to be essential in various hematologic malignancies.8,9 TRIM31, another TRIM family protein, is crucial for intracellular signaling, innate immunity, autophagy and carcinogenesis. TRIM31 regulates the immune response through MAVS and SYK, suppresses NLRP3-induced inflammasome activation and promotes autophagy in intestinal cells.10-13 TRIM31 not only contributes significantly to cerebral ischemic injury by promoting degradation of TIGAR,14 but also contributes to hypertensive nephropathy by promoting degradation of MAP3K7.15 Through targeting Rhbdf2 in mouse hepato-

cytes, Trim31 alleviates non-alcoholic fatty liver disease.16 In cases of aggressive AML, the mixed lineage leukemia 1 protein (MLL) has frequently been found to be fused with a partner (e.g., AF9).17 Various investigations of MLL-AF9induced AML have studied transcriptional regulators (e.g., HoxA9) and epigenetic modulators (e.g., Dot1L);18,19 however, the study of post-translational regulation in MLLAF9-mediated leukemogenesis remains scarce. Cyclin-dependent kinase 8 (CDK8) is a cell-intrinsic regulator that functions conservatively in transcription, as a part of the CDK8-mediator complex.20 The CDK8-mediator complex functions through releasing RNA polymerase II (RNAPII) from a paused state to start transcription.21 The CDK8-mediator complex also functions as a tethering module in enhancement of gene transcription regulated by noncoding RNA.22 During the innate immune response and inflammation, the mediator-associated kinase CDK8 plays a role as a negative regulator of interleukin-10.23 CDK8 has also been reported to act as an oncogene in both colon cancer24,25 and melanoma.26 In addition, the growth of AML cells can be inhibited by repressing CDK8 activity with small-molecule drugs.27

In this study, we found that TRIM31 deletion leads to HSC proliferation and functional decline, while accelerating the initiation of leukemia. Mechanistically, TRIM31 functions as an E3 ligase of CDK8, and its deletion causes blockage of ubiquitin-mediated CDK8 degradation. The accumulation of CDK8 leads to enhanced expression of PBX1 and cyclin D1. Potentially, an increase in E3 ligase activity of TRIM31 could be used as an anti-leukemia target because TRIM31 deletion promotes the initiation and development of MLLAF9-related AML.

Methods

Mice

TRIM31+/- mice12 were a kind gift from Professor Chengjiang Gao of Shandong University and were produced by microinjecting transcription activator-like effector nuclease (TALEN) mRNA into fertilized eggs of mice with a C57BL/6 background. The TRIM31-/- mice were genotyped by sequencing polymerase chain reaction (PCR) fragments (250 bp) in the TALEN-targeting region, which was amplified from isolated genomic DNA from the mouse tail using the following primers: forward 5’-GGCCTTGGATTTCTGTACTTTCACATC-3’ and reverse 5’-TGGGCCTGAACGTATTCTTATTCACAG-3’. Wildtype (WT) and TRIM31-/- mice aged 8-12 weeks were used in the experiments, except for those of marked age. The recipient mice, which were used in the competitive transplantation assays, were either CD45.1 mice or CD45.1/CD45.2 mice with a C57BL/6 background. The Animal Care and Ethics Committee at Jinan University approved all animal experiments in our study.

Flow cytometry and cell sorting

Bone marrow cells were freshly isolated from mice and incubated in a lineage cocktail of antibodies targeting CD4 (1:100, RM4-5), CD8 (1:100, 53-6.7), Ter-119 (1:100, TER-119), CD11b (1:150, M1/70), Gr-1 (1:150, RB6-8C5) and B220 (1:100, RA3-6B2) for 30 min. The cells were then washed and incubated in an antibody mix containing antibodies against CD34 (1:100, RAM34), and CD48 (1:200, HM48-1), CD45.2 (1:100, 104), IL-7R (1:100, A7R34), Flt3 (1:100, A2F10), CD150 (1:100, TC15-12F12.2), CD45.1 (1:100, A20), Sca1 (1:100, E13161.7), c-Kit (1:100, ACK2), and CD16/32 (1:100, 93) and streptavidin. All antibodies were monoclonal and purchased from BD Biosciences. For cell sorting, the bone marrow cells were enriched with anti-antigen-presenting cell microbeads (Miltenyi Biotec) and then stained with antibodies for surface markers. Cell analysis and data acquisition were performed using an LSR Fortessa (BD Biosciences) cell analyzer, and cell sorting was carried out using an Aria 3 cell sorter (BD Biosciences). The data were analyzed using FlowJo software.

Cytokine stimulation of hematopoietic stem cells

HSC were sorted from the bone marrow of WT mice via flow cytometry, and then plated in SFEM medium supplemented with stem cell factor (10 ng/mL; Pepro Tech), thrombopoietin (10 ng/mL; Pepro Tech), interleukin-3 (10 ng/mL; Pepro Tech) and 100 U/mL penicillin/streptomycin. Cells were collected at 0 h, 24 h and 48 h time-points for the experiment.

MLL-AF9-mediated leukemia transformation assay

c-Kit+ cells were enriched from the bone marrow of WT and TRIM31-/- mice and then cultured in Iscove-modified Dulbecco medium containing 10% fetal bovine serum, 50 ng/mL stem cell factor, 10 ng/mL interleukin-3 and 10 ng/mL interleukin-6 overnight to stimulate cell proliferation. The next day, cells were transduced with retrovirus encoding MLL-AF9. After 72 h, cells were harvested and GFP+ cells were sorted by an Aria 3 Sorter. GFP+ cells (5x104) were transplanted together with 5x105 bone marrow cells into lethally irradiated recipients. Mice were monitored for MLL-AF9 AML development. Five thousand GFP+ cells were sorted and plated in colony-forming units assay using MethoCult M3434 (STEMCELL Technologies) media for 10 days before colony counting.

Results

TRIM31 deletion impairs hematopoietic stem and progenitor cell homeostasis

E3 ligase regulates the cell cycle efficiently through precise control of cell cycle factors (e.g., CDK and cyclin proteins) at the protein level. An increasing number of studies

have shown that the E3 ligase function of TRIM proteins plays an essential role in tumorigenesis;28 however, its role in the hematopoietic system and leukemia initiation has rarely been studied. Real-time PCR revealed that the mRNA level of TRIM31 is reduced significantly after stimulation of HSC proliferation and differentiation (Online Supplementary Figure S1A), which indicates high expression of TRIM31 may play a potential role in the cell cycle control of HSC. To determine the function of TRIM31 in regulating HSC, TRIM31 knockout mice were created12 (Online Supplementary Figure S1B, C) and analyzed by using fluorescence activated cell sorting (FACS). In the

hematopoietic system, the cellularity of TRIM31 -/- bone marrow and Lin- cells remained similar to that of WT mice (Online Supplementary Figures S1D-F); however, the number of LK (Lin-Sca-1-c-Kit+) progenitor cells decreased significantly (Online Supplementary Figure S1G, H). In the bone marrow of TRIM31-/- mice, FACS analysis revealed higher numbers of common lymphoid progenitors (CLP) and lower numbers of common myeloid progenitor (CMP), but no variation in the number of granulocyte/monocyte progenitors (GMP) (Figure 1A, B; Online Supplementary Figure S1G, I-K). A more than 2-fold increase was also shown in both the absolute number and frequency of LSK (Lin-

Figure 1. TRIM31 deficiency induces hematopoietic stem cell expansion and long-term hematopoietic stem cell exhaustion. (A) FACS plots of granulocyte/monocyte progenitor (GMP; CD34 + CD16/32 + LKS - ), common myeloid progenitor (CMP; CD34+CD16/32-LKS-) and megakaryocyte/erythroid progenitor (MEP; CD34-CD16/32-LKS-) populations in wildtype (WT) and TRIM31 -/- mice (n=5-6 per group) with the frequencies indicated. (B) The numbers of GMP, MEP and CMP in WT and TRIM31 -/- mice (n=5-6 per group). (C) Representative FACS plots of LSK cells (Lin-Sca-1+c-Kit+), long-term hematopoietic stem cells (LT-HSC; CD34 - Flt3 - LSK), short-term hematopoietic stem cells (ST-HSC; CD34+Flt3-LSK) and multipotent progenitors (MPP; CD34+Flt3+LSK) with the frequencies indicated. (D, E) The absolute numbers (D) and percentages (E) of LSK cells (among Lin - cells) from WT and TRIM31 -/- mice (n=5-6 per group). (F) The numbers of LT-HSC, ST-HSC, and MPP in WT and TRIM31-/- mice (n=5-6 per group). All results are presented as the mean ± standard deviation. * P <0.05; ** P <0.01; *** P <0.001; **** P <0.0001. NS=not significant. Eight- to 12-week-old WT and TRIM31-/- mice were used in the experiments.

Sca-1+c-Kit+) cells in TRIM31-/- bone marrow (Figure 1C-E). To further characterize the composition of the LSK population, CD34 and Flt3 markers were used to distinguish long-term HSC (LT-HSC; CD34-Flt3- LSK), short-term HSC (ST-HSC; CD34+Flt3- LSK), and multipotent progenitors (MPP; CD34+Flt3+ LSK). In TRIM31-/- bone marrow, the numbers of ST-HSC and MPP increased dramatically, while the absolute number of LT-HSC decreased significantly (Figure 1C, F), which is probably attributable to stem cell exhaustion.

In bone marrow, TRIM31 depletion resulted in a significant decrease of B lymphocyte number (Online Supplementary Figure S2A, B), while the difference in peripheral blood was minimal (Online Supplementary Figure S2C). During erythrocyte development, the percentage of E3 (TER119+CD71mid) cells increased dramatically in TRIM31-/- mice (Online Supplementary Figure S2D, E). In addition, the spleen was hypertrophic in TRIM31-/- mice (Online Supplementary Figure S3A, B), and the hypertrophic spleen contained a higher percentage of LSK cells (Online Supplementary Figure S3C), indicating that TRIM31 deletion may cause malfunction of the spleen. Consistently, the composition of B cells decreased while the proportion of myeloid cells increased in TRIM31-/- mouse spleens (Online Supplementary Figure S3D). Overall, TRIM31 deletion impaired HSC homeostasis, causing harm to the cellularity of LT-HSC, ST-HSC, MPP and LSK cells, which further resulted in abnormal regulation of progenitors, such as CMP and CLP.

Loss of TRIM31 impairs hematopoietic stem cell function of reconstitution capacity

To evaluate the self-renewal and differentiation capacity of LT-HSC in vivo, we performed a competitive LT-HSC transplantation experiment. As donor cells, 300 LT-HSC were sorted from WT and TRIM31-/- mice and transplanted into lethally irradiated recipient mice together with 3x105 bone marrow competitors (Online Supplementary Figure S3E). The percentage of donor-derived cells (chimerism) in peripheral blood was analyzed every 4 weeks. The chimerism of TRIM31-/- cells in peripheral blood was much lower than that of WT cells (Figure 2A). After 16 weeks, bone marrow analysis of chimeric mice showed that donor-derived TRIM31-/- LSK cells were also greatly reduced in comparison to donor-derived WT cells (Figure 2B). Further analysis revealed that the chimerism of TRIM31-/- cells was significantly decreased at the LT-HSC, ST-HSC and MPP (MPP2 and MPP3) levels (Figure 2B). For consecutive transplantation, 1x106 chimeric bone marrow cells were re-transplanted into secondary recipient mice (Online Supplementary Figure S3E). As in the first round of transplantation, the percentage of donor-derived TRIM31-/- cells in peripheral blood dropped further, while that of donor-derived WT cells remained nearly the same. The peripheral blood chimerism of TRIM31-/--derived cells

was almost zero at 28 weeks after transplantation, while that of WT-derived cells remained approximately 50% (Figure 2A). Bone marrow analysis after the second transplantation showed that LSK chimerism was dramatically decreased in TRIM31-/--derived cells in comparison to WTderived cells (Figure 2C). Furthermore, chimerism analysis of LT-HSC, ST-HSC and MPP showed dramatic reductions in TRIM31-/--derived cells (Figure 2C). The chimerism in T lymphocytes, B lymphocytes and myeloid cells in peripheral blood was also clearly decreased following both the first and second transplants (Figure 2D-F).

To confirm the results of LT-HSC transplantation, a competitive LSK transplantation experiment was performed using 4,000 LSK cells sorted from WT and TRIM31-/- mice, together with 1x106 competitive cells (Online Supplementary Figure S4A). As for the LT-HSC transplantation, peripheral blood chimerism was analyzed every 4 weeks, and the percentage of donor-derived TRIM31-/- cells in peripheral blood was significantly decreased in comparison to that of donor-derived WT cells (Online Supplementary Figure S4B). Bone marrow analysis also indicated that chimerism of donor-derived LSK cells was significantly reduced in TRIM31-/--derived cells (Online Supplementary Figure S4C). This reduction in chimerism was simultaneously observed at the LT-HSC, ST-HSC and MPP levels in TRIM31-/--derived cells (Online Supplementary Figure S4D, E). Additionally, T lymphocytes, B lymphocytes, and myeloid cells all displayed drastic reductions in chimerism (Online Supplementary Figure S4F-H). To exclude the effect of TRIM31 deletion in the stem cell niche, 4,000 WT LSK cells were sorted and transplanted into WT and TRIM31-/--recipient mice. No statistically significant difference of donor-derived LSK chimerism was shown between WT and TRIM31/--recipient mice (Online Supplementary Figure S4I, J). In summary, after serial competitive transplantation, TRIM31 deletion significantly reduced peripheral blood chimerism and bone marrow chimerism at the levels of LSK cells, LTHSC, ST-HSC, and MPP, indicating that TRIM31 deletion impairs the reconstitution ability of HSC.

TRIM31 deficiency promotes the proliferation of hematopoietic stem cells and reduces their selfrenewal capacity under stress

To ascertain the underlying reasons for hematopoietic stem and progenitor cell alterations in TRIM31-/- mice, a BrdU incorporation assay was performed. TRIM31 deletion induced a significant increase of BrdU-positive cells both at the LSK cell and LT-HSC levels, indicating that TRIM31 deletion promotes proliferation of HSC (Figure 3A, B). This indication was also supported by the decrease of Ki67negative staining in TRIM31-/- cells (Online Supplementary Figure S5A-D). Conditions of stress, such as 5-fluorouracil (5-FU) treatment, could cause apoptosis of cycling HSC, whereas quiescent HSC might remain viable.29 To detect

the stress effect of TRIM31 deficiency on HSC, we challenged TRIM31-/- mice with sequential 5-FU administration. TRIM31-/- mice challenged with sequential 5-FU treatment died much earlier than WT controls (Figure 3C), demonstrating that TRIM31-null HSC were more activated.

5-FU treatment was also used to quantify cell self-renewal capacity. Analysis of the results at 6 days after 5FU treatment revealed that TRIM31-/- mice had considerably fewer LSK cells than WT mice. Subsequent analysis demonstrated that the numbers of LT-HSC, ST-

Figure 2. Loss of TRIM31 impairs long-term hematopoietic stem cell function. (A) The percentage of donor-derived peripheral blood cells at the indicated time-points in the serial competitive transplantation assay is shown (1st, for the first competitive transplantation, 4 recipient mice per group; 2nd, for the second competitive transplantation, 6 recipient mice per group). (B) Percentages of donorderived LSK (Lin-Sca-1+c-Kit+) cells, long-term hematopoietic stem cells (LT-HSC; CD48-CD150+Flt3-LSK), short-term hematopoietic stem cells (ST-HSC; CD48-CD150-Flt3-LSK) and multipotent progenitors (MPP; CD48+Flt3-LSK) 16 weeks after the first transplantation are shown (n=4 per group). (C) Percentages of donor-derived LSK cells, LT-HSC, ST-HSC and MPP cells 12 weeks after secondary transplantation are shown (n=6 per group). (D-F) Two-round serial transplantation was conducted using 300 purified LT-HSC cells along with 3x105 fresh competitors each time. Chimerism of T, B and myeloid cells in peripheral blood is shown at the indicated time-points after transplantation. All results are presented as the mean ± standard deviation. **P<0.01; ***P<0.001; ****P<0.0001. Eight- to 12week-old wildtype and TRIM31-/- mice were used in the experiments. WT: wildtype; PB: peripheral blood; BM: bone marrow.

HSC and MPP after 5-FU administration were all substantially reduced in TRIM31-/- mice compared to the numbers in WT mice (Figure 3D, E). To further confirm the defect caused by TRIM31 deletion, a single-colony formation assay was used to quantify the ability of HSC to prolifer-

ate and differentiate in vitro. In TRIM31-/- LT-HSC, the number of large colonies was significantly decreased, indicating that the self-renewal ability of LT-HSC was impaired after long-term activation (Figure 3F). In short, TRIM31 deletion induces HSC proliferation in vivo and

Figure 3. TRIM31 deletion leads to enhanced hematopoietic stem cell proliferation and reduced self-renewal capacity under stress. (A) Representative FACS plots showing the proliferation analysis in long-term hematopoietic stem cells (LT-HSC; CD48CD150+Flt3- LSK) and LSK (Lin-Sca-1+c-Kit+) cells. The percentages of BrdU-positive cells are shown. (B) The percentages of BrdUpositive cells among LT-HSC and LSK cells from wildtype (WT) and TRIM31-/- mice after long-term BrdU labeling (10 days) (n=4-5 per group). (C) Survival curves of WT and TRIM31-/- mice following sequential 5-fluoruracil (5FU) treatment (n=13 per group). 5FU was injected into mice once a week for a total of two injections. (D) Representative FACS plots of LSK, LT-HSC, short-term hematopoietic stem cells (ST-HSC; CD48-CD150-Flt3- LSK) and multipotent progenitors (MPP2; CD48+CD150+Flt3- LSK and MPP3; CD48+CD150-Flt3- LSK) 6 days after 5FU treatment. (E) The absolute numbers of LT-HSC, ST-HSC, MPP2 and MPP3 cells in WT and TRIM31-/- mice after 5FU treatment (n=4 per group). (F) Representative images of large, intermediate and small colonies are shown. The percentages of colonies formed after 14 days of culture of single LT-HSC sorted from WT and TRIM31-/- mice are shown (n=3 per group). All results are presented as the mean ± standard deviation. *P<0.05; **P<0.01; ****P<0.0001. Eight- to 12-week-old WT and TRIM31-/- mice were used in the experiments.

leads to a functional decline of self-renewal ability under a condition of stress.

TRIM31 deletion accelerates leukemia initiation in a model of MLL-AF9-induced acute myeloid leukemia

To evaluate whether the accelerated proliferation of TRIM31-/- hematopoietic cells is related to leukemia, a database analysis was carried out and revealed that low expression of TRIM31 was associated with poor AML survival (Figure 4A, B; Online Supplementary Figure S5E). Among patients with various types of AML, TRIM31 expression was relatively lower in those with MLL rearrangements (Online Supplementary Figure S5F). Therefore, a murine AML model driven by the human oncogene MLL-AF9 was used to investigate the role of TRIM31 in leukemia initiation and development.30 c-Kit+ cells were enriched from bone marrow of WT and TRIM31-/- mice, and then transduced with MLL-AF9 retrovirus (Online Supplementary Figure S6A). In vitro, a surrogate functional examination using methylcellulose medium revealed that TRIM31-null MLL-AF9 primary transformed cells had significantly increased colony-forming capability (Online Supplementary Figure S6B). During primary plating and second plating, the numbers of colonies and total cells were markedly higher for MLL-AF9-converted TRIM31-/- cells than for WT cells (Online Supplementary Figure S6C, D). To assess leukemogenesis in vivo, MLL-AF9 retrovirus-transduced WT and TRIM31-/- cells were sorted and transplanted into lethally irradiated recipient mice after which the development of AML was monitored (Online Supplementary Figure S6A). The frequency of L-GMP (Lin-Sca1-c-Kit+CD34+CD16/32+) cells was reported as leukemia-initiating cells in the MLLAF9-induced leukemia.31 A higher frequency of L-GMP was detected in the bone marrow of TRIM31-/- donors than in WT donors (Figure 4C). In peripheral blood, the frequency of GFP+ leukemic cells was significantly higher in TRIM31/- donors than in WT ones (Figure 4D). Meanwhile, recipient mice that received MLL-AF9-transduced TRIM31 -/donor cells had a significantly shorter overall survival than WT mice (Figure 4E). Moreover, after the L-GMP cells from leukemia mice had been sorted and transplanted, the disparity in overall survival of mice receiving grafts from WT or TRIM31-/- donors increased dramatically (Figure 4F). Taking into consideration the uniform expression, copies of MLL-AF9 fusion gene and mean fluorescent intensity of GFP in both WT and TRIM31-/- donor cells (Online Supplementary Figure S6E-J), these data collectively support that TRIM31 deletion accelerated the initiation and development of MLL-AF9-induced AML.

TRIM31 functions as an E3 ligase of CDK8 and regulates its ubiquitin-mediated degradation

The selected TRIM31 interacting protein candidates after mass spectrometry analysis are summarized.

Through MAVS and NLRP3, TRIM31 plays a critical role in macrophages and intestinal cells.15-17 In order to identify the underlying mechanism and the functional substrate of TRIM31 in hematopoietic cells, we performed mass spectrometry (MS) on WT cells, with TRIM31-/- cells used as the negative control (Online Supplementary Figure S7A). Several TRIM31 interacting protein candidates in WT cells were selected after MS (Online Supplementary Table S1). Interestingly, FACS analysis showed that the median fluorescence intensity of CDK8 was dramatically increased in LT-HSC, LSK cells and Lin- cells from TRIM31-/- mice, whereas no difference in the mean fluorescence intensity of EIF5A and Cdkn2a was detected between WT and TRIM31-/- mice (Figure 5A; Online Supplementary Figure S7B, C). The increased CDK8 level in cells from TRIM31-/mice was verified by western blot (Figure 5B). Moreover, the cell cycle regulator cyclin D1 was strongly upregulated in TRIM31-/- cells (Figure 5B), and this upregulation was accompanied by an increase in its interaction partner CDK6 (Online Supplementary Figure S7D), which may be due to cell proliferation. To verify whether CDK8 is a direct ubiquitin substrate of TRIM31, we performed a co-immunoprecipitation assay of CDK8 antibody using WT and TRIM31-/- cells, which were treated with a proteasomal inhibitor (MG132) before cell lysis. The level of ubiquitination of CDK8, but not of CDK6, was reduced dramatically in TRIM31-/- cells (Figure 5C; Online Supplementary Figure S7E), indicating that CDK8 is a ubiquitin substrate of the E3 ligase, TRIM31. Furthermore, the successful pulling down of TRIM31 and CDK8 together confirmed a direct protein interaction between them (Figure 5D). Surprisingly, CDK8 could pull down and co-localize with cyclin D1 (Figure 5D; Online Supplementary Figure S7F), suggesting that cyclin D1 might function together with CDK8 in hematopoietic cells. Importantly, upregulation of CDK8 and cyclin D1 was observed in TRIM31-/- mouse spleens (Figure 5E), and the level of CDK8 protein was also enhanced in TRIM31-/- leukemia and L-GMP cells (Figure 5F). In brief, CDK8 accumulation in hematopoietic and leukemia cells was caused by deletion of its E3 ligase TRIM31 and eventually led to enhanced expression of the cell cycle regulator cyclin D1

Genetic knockdown of CDK8 or cyclin D1 rescues the defects in TRIM31-/- hematopoietic stem cells

To confirm that the functional decline in stem cells in TRIM31-/- mice was attributable to upregulation of CDK8, an experiment of inhibition of CDK8 was performed.

LY2857785 is a type I reversible and competitive ATP kinase inhibitor of both CDK8 and CDK9.32 After the administration of LY2857785 to WT and TRIM31-/- mice for 2 weeks, LT-HSC were sorted and transplanted together with competitors into recipient mice (Online Supplementary Figure S8A). Eight to 10 weeks later, FACS analysis

Figure 4. TRIM31 deletion accelerated leukemia initiation in an MLL-AF9-induced mouse model. (A) Kaplan-Meier survival curve analysis of The Cancer Genome Atlas dataset (https://xenabrowser.net/) comparing the acute myeloid leukemia (AML) patients with high (n=33) versus low (n=99) TRIM31 expression. (B) Kaplan-Meier survival curve analysis of the GSE12417 dataset comparing AML patients with high (n=39) versus low (n=40) TRIM31 expression. (C) Representative FACS plots showing the percentages of L-GMP cells (Lin-Sca1-c-Kit+CD34+CD16/32+) in the bone marrow (BM) of wildtype (WT) and TRIM31-/- recipient mice (n=4-6 per group). (D) Representative FACS plots showing the percentages of GFP+ cells and myeloid cells in the peripheral blood (PB) of WT and TRIM31-/- recipient mice (n=4-6 per group). (E) Survival curves of mice transplanted with MLL-AF9 WT or TRIM31-/- leukemic cells (n=5 per group). (F) Survival curves of mice transplanted with 5,000 WT or TRIM31-/- L-GMP cells (n=5 per group). All results are presented as the mean ± standard deviation. *P<0.05; **P<0.01. Eight- to 12-week-old WT and TRIM31-/- mice were used in the experiment. GDC: Genomic Data Commons; TCGA: The Cancer Genome Atlas; SSC: side scatter; FSC: forward scatter; GFP: green fluorescent protein.

showed that the treatment of TRIM31-/- mice with a CDK8 inhibitor caused enhancement of peripheral blood LSK and LT-HSC chimerism, while no difference was shown in the control group (Figure 6A, B; Online Supplementary Figure S8B-D). To inhibit CDK8 more specifically, LT-HSC from TRIM31-/- mice were infected with lentivirus containing

GFP-labeled shRNA against CDK8 and transplanted into recipient mice. Twelve, 16 and 20 weeks later, peripheral blood analysis showed that the percentage of donor-derived cells was dramatically increased in the CDK8 knockdown group (Figure 6C). A similar phenomenon was also found in donor-derived LSK cells after bone marrow

Figure 5. Accumulation of CDK8 and upregulation of cyclin D1 were caused by TRIM31 deletion. (A) Representative FACS plots of CDK8 mean fluorescence intensity (MFI) in Lin– cells from wildtype (WT) and TRIM31-/- mice. The MFI of CDK8 in long-term hematopoietic stem cells (LT-HSC; CD48-CD150+Flt3- LSK), LSK cells (Lin-Sca-1+c-Kit+) and Lin- cells from WT and TRIM31-/- mice is shown (n=5 per group). (B) Immunoblot (IB) of cyclin D1 and CDK8 in the whole cell lysate (WCL) of Lin- cells from WT and TRIM31-/- mice. α-tubulin was used as a loading control (n=3 per group). IB of CDK8 in the WCL of LSK cells from WT and TRIM31-/- mice. β-actin was used as a loading control (n=3 per group). (C) Ubiquitination assay of CDK8 in the whole cell lysate of Lin- cells from WT and TRIM31-/- mice treated with a proteasomal inhibitor (MG132). IB against ubiquitin was performed with the immunoprecipitation (IP) assay precipitate of anti-CDK8 and CDK8 from WCL used as a loading control. (D) IP of anti-CDK8 and anti-TRIM31 was performed on the WCL of Lin- cells from WT mice; IP of anti-IgG was used as a control. IB against TRIM31, cyclin D1 and CDK8 was performed with the IP assay precipitate and the WCL of Lin- cells, which was used as a loading control. (E) IB of cyclin D1 and CDK8 in WCL of spleens from WT and TRIM31-/- mice. β-actin was used as a loading control (n=2 per group). (F) IB of CDK8 in the WCL of c-Kit+ cells from recipient mice transplanted with MLL-AF9 WT and TRIM31-/- cells. β-actin was used as a loading control (n=3-4 per group). IB of CDK8 in the WCL of L-GMP cells from recipient mice transplanted with MLL-AF9 WT and TRIM31-/- cells. β-actin was used as a loading control (n=3 per group). All results are presented as the mean ± standard deviation. **P<0.01; ***P<0.001. Eight- to 12-weekold WT and TRIM31-/- mice were used in the experiments. Ub: ubiquitin.

analysis (Figure 6D; Online Supplementary Figure S8E-G). The above results demonstrate that knockdown of CDK8 in TRIM31-/- LT-HSC could lead to enhanced stem cell function, which might be due to the prevention of LT-HSC exhaustion (Figure 6E, F). In addition, knockdown of cyclin D1 in TRIM31-/- LT-HSC also resulted in an increased percentage of donor-derived cells (Online Supplementary Figure S9A-F), which was similar to the effects of overexpression of TRIM31 (Online Supplementary Figure S10A-C), showing that TRIM31 and cyclin D1 may function in the same pathway (Online Supplementary Figure S9G). Meanwhile, overexpression of TRIM31 caused reduced colony formation in WT MLL-AF9 leukemia cells (Online Supplementary Figure S10D). In short, TRIM31 functions through the CDK8/cyclin D1 pathway, which also inhibits the development of MLL-AF9, since both pharmaceutical inhibition of CDK8 and genetic knockdown of CDK8 and cyclin D1 rescued the functional defects in TRIM31-/- HSC.

CDK8 regulates the expression of cyclin D1 through transcriptional factor PBX1

The CDK8-mediator is a large macromolecular complex comprising four modules, the head, middle, tail, and kinase modules, which consists of CDK8 and three other factors: Med12, Med13 and cyclin C.33 CDK8, MED12 and RNAPII were found to bind the promotor region of the transcriptional factor PBX1 after analysis of chromatin immunoprecipitation (ChIP) sequencing data (GEO dataset: GSE128242) (Figure 7A).34 PBX1 is a proto-oncogene in the hematopoietic system, and its loss results in exhaustion of LT-HSC and defects in HSC maintenance and function.35 Meanwhile, the upregulation of PBX1 promotes cell proliferation.36 Through ChIP-quantitative PCR, CDK8 was verified to be enriched at the promoter region of PBX1 (Figure 7B). To determine the underlying mechanism of CDK8 regulation, quantitative PCR analysis of cyclin D1 and PBX1 was performed. Both mRNA expression levels of cyclin D1 and PBX1 were enhanced in TRIM31-/- hematopoietic and L-GMP cells (Figure 7C), which was verified at the protein level by western blot (Figure 7D). In order to further determine the relationship between CDK8, PBX1 and cyclin D1, CDK8 was knocked down and overexpressed in NIH/3T3 mouse fibroblasts. The transcriptional expression of PBX1 and the mRNA expression of cyclin D1 were regulated through the mRNA level of CDK8 (Figure 7E). Similarly, the downregulation of PBX1 caused by knockdown of CDK8 was also shown in hematopoietic cells (Figure 7F). Moreover, the changes in PBX1 directly regulated the expression of cyclin D1, indicating that PBX1 is located upstream of cyclin D1 (Online Supplementary Figure S11A). Functionally, knockdown of PBX1 in TRIM31-/- LT-HSC resulted in an increased percentage of donor-derived cells (Online Supplementary Figure S11B-F), and knockdown of TRIM31 resulted in increased expression of PBX1 and cyclin

D1 in THP1 cells (Online Supplementary Figure S11G), implying that TRIM31 and PBX1 may act in the same pathway. In concordance with our quantitative PCR data, the expression of PBX1 and cyclin D1 in primary AML patients was correlated with CDK8 expression (Online Supplementary Figure S12A, B). Furthermore, cyclin D1 was significantly increased in patients with higher PBX1 expression (Online Supplementary Figure S12C). In general, CDK8 regulates the expression of cyclin D1 through the transcriptional factor PBX1.

Discussion

Ubiquitin E3 ligases are vital regulators of the hematopoietic system, and their absence dramatically impairs HSC maintenance and function. For example, deletion of Huwe1, a ubiquitin ligase, leads to increased proliferation and stem cell exhaustion via upregulation of N-myc expression.37 Conditional deletion of SCFFbxw7, another ubiquitin E3 ligase, causes loss of HSC quiescence and stem cell exhaustion via enhancement of c-Myc.3 In the present study, we found that the E3 ligase TRIM31 could regulate HSC homeostasis and function. TRIM31 deletion caused enhanced proliferation of LT-HSC and further led to their exhaustion. Furthermore, HSC function was significantly impaired in TRIM31-/- mice. In a serial competitive transplantation assay and 5-FU challenge, TRIM31-/- HSC showed dramatically reduced self-renewal capacity. TRIM31 was reported to be an E3 ligase in the immune system11,12 and was also found to play a role in cancer cells.38,39 Here, we discovered a novel role for TRIM31, which is crucial for the hematopoietic system. Prospectively, TRIM31 could be targeted to regulate HSC maintenance and function.

In human AML, MLL is mutated by translocation in about 4% of cases and MLL fusion partners include AF9, ENL, and AF4. 30 Sustaining proliferative signaling is a hallmark of cancer development; and deletion of TRIM31 promotes breast cancer progression through ubiquitination of p53.40 Based on our findings of cell proliferation induced by TRIM31 deletion, we explored the function of TRIM31 in the initiation of AML. TRIM31 may promote leukemia progression through Wnt/β-catenin signaling in AML cell lines.41 However, by using a murine model of MLL-AF9 in leukemia initiation and development, we demonstrated that TRIM31 depletion increased the proportion of leukemia-initiating cells and accelerated the kinetics of leukemia development. These effects might be due to the functional mechanisms of TRIM31 varying at different disease stages. Consistent with our results, a database analysis in AML patients also showed that higher expression of TRIM31 correlated with significantly better survival. These data indicate that TRIM31 functions as a suppressor in AML initiation and devel-

opment. Consolidating all the results of the TRIM31-/- HSC analysis, TRIM31-mediated control of HSC proliferation is integral to suppress the vulnerability of HSC to leukemic transformation and disease initiation in AML. Members of the CDK8 module are vital in various cancers and developmental diseases.42,43 Loss of the CDK8 module member MED12 causes rapid bone marrow failure and

acute lethality, showing that MED12 is essential for HSC homeostasis.44 The kinase activity of CDK8 was reported to be inhibited by small molecules;45 however, the mechanism of its upstream regulation remains unknown. In this study, we found that CDK8 was regulated by the E3 ligase TRIM31 through ubiquitin-mediated protein degradation. Upregulation of CDK8 in TRIM31-/- hematopoietic and

Figure 6. Inhibition and knockdown of CDK8 rescues TRIM31 -/- hematopoietic stem cell function. (A) Percentage of donorderived peripheral blood (PB) cells at the indicated time-points after competitive transplantation in long-term hematopoietic stem cells (LT-HSC: CD48 - CD150 + Flt3 - LSK) sorted from wildtype (WT) and TRIM31 -/- mice treated with a CDK8 inhibitor (LY2857785) (n=4 per group). (B) Percentage of donor-derived LSK (Lin-Sca-1+c-Kit+) cells and LT-HSC 10 weeks after LT-HSC competitive transplantation. (C) The percentage of GFP + cells among donor-derived PB cells after transplantation with shCDK8-1 and -2 virus-infected TRIM31 -/- LT-HSC (n=4 per group). (D) The percentage of GFP+ cells in donor-derived LSK cells after transplantation with shCDK8-1 and -2 virus infected TRIM31 -/- LT-HSC (n=4 per group). (E) The percentage of G0 phase (Ki67-negative) cells from mice 12 weeks after transplantation with WT and TRIM31-/- LT-HSC infected with shCDK8-1 lentivirus (n=4-5 per group). (F) The knockdown efficiency of shCDK8-1 in GFP + LSK cells after transplantation with shCDK8-1 virus-infected TRIM31 -/- LT-HSC (n=5-6 mixed per group). All results are presented as the mean ± standard deviation. * P <0.05; ** P <0.01; *** P <0.001. Eight- to 12-week-old WT and TRIM31 -/- mice were used in the experiment.

leukemia cells leads to enhanced stem cell proliferation, indicating that CDK8 may function as a regulator of cell cycle-related pathways in the hematopoietic system. Therefore, TRIM31 might be used as a pharmaceutical target in CDK8-enhanced tumors.

CDK8 is reported to be involved in several signaling path-

ways. In colon cancer, β-catenin hyperactivity promotes cell proliferation and drives tumor transformation which is inhibited by repression of CDK8 expression.24 Through the downstream target Myc, CDK8 functions in regulating the pluripotent state of embryonic stem cells.46 Here, in hematopoietic cells, we found that PBX1 acts as a direct down-

Figure 7. CDK8 regulates the expression of PBX1 and cyclin D1. (A) The binding sites of CDK8, MED12 and RNAPII in the region of the PBX1 gene are shown. Raw data (from GEO dataset: GSE128242) were analyzed. (B) Real-time polymerase chain reaction (PCR) analysis of CDK8 chromatin immunoprecipitated DNA. Two pairs of primers were designed based on the PBX1 promoter region (2,000 bp upstream of the start point). For each primer pair, amplifications of chromatin before immunoprecipitation and chromatin immunoprecipitated with preimmune serum were performed as input and negative controls, respectively. The value of bound DNA relative to input is shown as a percentage (n=3 per group). (C) The mRNA levels of PBX1 and cyclin D1 in LSK (Lin-Sca-1+c-Kit+) cells from WT and TRIM31-/- mice were measured via real-time PCR. The relative expression was normalized to β-actin expression for statistical analysis (n=3 per group). The mRNA levels of CDK8, PBX1 and cyclin D1 in L-GMP cells from recipient mice transplanted with MLLAF9 WT and TRIM31-/- cells were measured via real-time PCR. The relative expression was normalized to β-actin expression for statistical analysis (n=3 per group). (D) Immunoblot of PBX1 in the whole cell lysate of Lin- cells from WT and TRIM31-/- mice. β-actin was used as a loading control (n=3 per group). (E) The relative mRNA expression of CDK8, PBX1 and cyclin D1 in 3T3NIH cells with and without CDK8 knockdown (CDK8-shRNA1, CDK8-shRNA 2) or overexpression (CDK8-OE) was measured via real-time PCR. The relative expression was normalized to β-actin expression for statistical analysis (n=3 per group). (F) The relative mRNA expression of CDK8 and PBX1 in control and CDK8 knockdown (GFP+) Lin- cells in vivo, which were isolated from mice 8 weeks after transplantation with TRIM31-/- long-term hematopoietic stem cells (LT-HSC: CD48-CD150+Flt3- LSK) infected with lentivirus. The relative expression was normalized to β-actin expression for statistical analysis (n=3 per group). All results are presented as the mean ± standard deviation. *P<0.05; **P<0.01; ***P<0.001; ns=not significant. Eight- to 12-week-old WT and TRIM31-/- mice were used in the experiments.

stream target of CDK8, which binds to the promoter region of PBX1; the mRNA expression levels of PBX1 and cyclin D1 were upregulated in TRIM31-/- mice. CDK8 regulates transcription factors through phosphorylation and proteasomal degradation, suggesting that CDK8 may function as a general regulator of transcription factors.47 CDK8 acts as a key mediator of BCR-ABL1-driven leukemia through transcriptional changes of the mTOR signaling pathway.48 In this study, we found that CDK8 regulates cyclin D1 through the transcription factor PBX1. Meanwhile, the association of CDK8 with mediators is reversible and can dramatically alter the structure and function of the mediators.49 In our context, the accumulation of CDK8 in TRIM31-/- hematopoietic and leukemia cells may open the structure of the transcriptional complex and lead to modification of chromosome elements such as histones, further promoting transcription of downstream genes.

PBX1, a proto-oncogene in childhood leukemia, is a typical homeodomain transcription factor belonging to the TALE family. PBX1 is very importantor the hematopoietic system and its downstream targets could be TGF β and JAK2/Stat3.35,36 In TRIM31-/- mice, cyclin D1 upregulation was induced by enhanced expression of PBX1 (Online Supplementary Figure S11G), which may occur through the Jak2/Stat3 pathway.36 In addition, cyclin D1 may function together with CDK8 in the hematopoietic system of TRIM31-/- mice, although the underlying mechanism needs to be further investigated. In conclusion, the accumulation of CDK8 caused by TRIM31 deletion induces transcription of PBX1, which further leads to upregulation of cyclin D1 (Online Supplementary Figure S11H). Enhancement of

References

1. Wilson A, Trumpp A. Bone-marrow haematopoietic-stem-cell niches. Nat Rev Immunol. 2006;6(5):93-106.

2. Li J. Quiescence regulators for hematopoietic stem cell. Exp Hematol. 2011;39(5):511-520.

3. Thompson BJ, Jankovic V, Gao J, et al. Control of hematopoietic stem cell quiescence by the E3 ubiquitin ligase Fbw7. J Exp Med. 2008; 205(6):1395-1408.

4. Rathinam C, Matesic LE, Flavell RA. The E3 ligase Itch is a negative regulator of the homeostasis and function of hematopoietic stem cells. Nat Immunol. 2011;12(5):399-407.

5. Rathinam C, Thien CB, Flavell RA, et al. Myeloid leukemia development in c-Cbl RING finger mutant mice is dependent on FLT3 signaling. Cancer Cell. 2010;18(4):341-352.

6. Bai X, Kim J, Yang Z, et al. TIF1gamma controls erythroid cell fate by regulating transcription elongation. Cell. 2010;142(1):133-143.

7. Ito K, Bernardi R, Morotti A, et al. PML targeting eradicates quiescent leukaemia-initiating cells. Nature. 2008;453(7198):1072-1078.

8. Gatt ME, Takada K, Mani M, et al. TRIM13 (RFP2) downregulation decreases tumour cell growth in multiple myeloma through inhibition of NF kappa B pathway and proteasome activity. Br J Haematol. 2013;162(2):210-220.

CDK8 and PBX1 target genes is observed in human leukemia;50 therefore, using TRIM31 as an anti-leukemia target is promising.

Disclosures

No conflicts of interest to disclose.

Contributions

DD and ZJ conceived and designed the experiments. KZ, DL, YL and ZS performed the experiments. KZ and DL analyzed the data. DD, KZ, DL and YL wrote the paper. CG provided TRIM31-/- mice and JG and HW provided valuable suggestions.

Acknowledgments

We thank Dr. H. Cheng for his kind gift of the MSCV-MLLAF9-IRES-GFP encoding plasmid.

Funding

This work was supported by the National Natural Science Foundation of China (grants #81970096 and 91849128 to DD) and the Natural Science Foundation of Guangdong Province, China (#2020A1515010453 to DD); the National Natural Science Foundation of China (#92049304, 82030039), the National Key R&D Program of China (2021YFA1100103) and the Program for Guangdong Introducing Innovative and Entrepreneurial Teams (2017ZT07S347) to ZJ.

Data-sharing statement

All plasmids and cell lines generated in this study are available from the authors upon request.

9. Gandini D, De Angeli C, Aguiari G, et al. Preferential expression of the transcription coactivator HTIF1alpha gene in acute myeloid leukemia and MDS-related AML. Leukemia. 2002;16(2):886-893.

10. Ra EA, Lee TA, Won Kim S, et al. TRIM31 promotes Atg5/Atg7independent autophagy in intestinal cells. Nat Commun. 2016;24(7):ncomms11726.

11. Song H, Liu B, Huai W, et al. The E3 ubiquitin ligase TRIM31 attenuates NLRP3 inflammasome activation by promoting proteasomal degradation of NLRP3. Nat Commun. 2016;8(7):ncomms13727.

12. Liu B, Zhang M, Chu H, et al. The ubiquitin E3 ligase TRIM31 promotes aggregation and activation of the signaling adaptor MAVS through Lys63-linked polyubiquitination. Nat Immunol. 2017;18(2):214-224.

13. Wang X, Zhang H, Shao Z, et al. TRIM31 facilitates K27-linked polyubiquitination of SYK to regulate antifungal immunity. Signal Transduct Target Ther. 2021;6(1):298.

14. Zeng S, Zhao Z, Zheng S, et al. The E3 ubiquitin ligase TRIM31 is involved in cerebral ischemic injury by promoting degradation of TIGAR. Redox Biol. 2021;45:102058.

15. Zhang J, Cao L, Wang X, et al. The E3 ubiquitin ligase TRIM31

plays a critical role in hypertensive nephropathy by promoting proteasomal degradation of MAP3K7 in the TGF-β1 signaling pathway. Cell Death Differ. 2022;29(3):556-567.

16. Xu M, Tan J, Dong W, et al. The E3 ubiquitin-protein ligase Trim31 alleviates non-alcoholic fatty liver disease by targeting Rhbdf2 in mouse hepatocytes. Nat Commun. 2022;13(1):1052.

17. Marschalek R. Mechanisms of leukemogenesis by MLL fusion proteins. Br J Haematol. 2011;152(2):141-154.

18. Smith LL, Yeung J, Zeisig BB, et al. Functional crosstalk between Bmi1 and MLL/Hoxa9 axis in establishment of normal hematopoietic and leukemic stem cells. Cell Stem Cell. 2011;8(6):649-662.

19. Nguyen AT, Taranova O, He J, et al. DOT1L, the H3K79 methyltransferase, is required for MLL-AF9-mediated leukemogenesis. Blood. 2011;117(25):6912-6922.

20. Taatjes DJ. The human mediator complex: a versatile, genomewide regulator of transcription. Trends Biochem Sci. 2010;35(6):315-322.

21. Galbraith MD, Allen MA, Bensard CL, et al. HIF1A employs CDK8mediator to stimulate RNAPII elongation in response to hypoxia. Cell. 2013;153(6):1327-1339.

22. Lai F, Orom UA, Cesaroni M, et al. Activating RNAs associate with Mediator to enhance chromatin architecture and transcription. Nature. 2013;494(7438):497-501.

23. Johannessen L, Sundberg TB, O'Connell DJ, et al. Smallmolecule studies identify CDK8 as a regulator of IL-10 in myeloid cells. Nat Chem Biol. 2017;13(10):1102-1108.

24. Firestein R, Bass AJ, Kim SY, et al. CDK8 is a colorectal cancer oncogene that regulates beta-catenin activity. Nature. 2008;455(7212):547-551.

25. Starr TK, Allaei R, Silverstein KA, et al. A transposon-based genetic screen in mice identifies genes altered in colorectal cancer. Science. 2009;323(5922):1747-1750.

26. Kapoor A, Goldberg MS, Cumberland LK, et al. The histone variant macroH2A suppresses melanoma progression through regulation of CDK8. Nature. 2010;468(7327):1105-1109.

27. Pelish HE, Liau BB, Nitulescu II, et al. Mediator kinase inhibition further activates super-enhancer-associated genes in AML. Nature. 2015;526(7572):273-276

28. Hatakeyama S. TRIM proteins and cancer. Nat Rev Cancer Cell. 2011;11(11):792-804.

29. Cheng T, Rodrigues N, Shen H, et al. Hematopoietic stem cell quiescence maintained by p21cip1/waf1. Science. 2000;287(5459):1804-1808.

30. Krivtsov AV, Twomey D, Feng Z, et al. Transformation from committed progenitor to leukaemia stem cell initiated by MLLAF9. Nature. 2006;442(17):818-822.

31. Huang D, Sun G, Hao X, et al. ANGPTL2-containing small extracellular vesicles from vascular endothelial cells accelerate leukemia progression. J Clin Invest. 2021;131(1):e138986.

32. Yin T, Lallena MJ, Kreklau EL, et al. A novel CDK9 inhibitor shows potent antitumor efficacy in preclinical hematologic tumor models. Mol Cancer Ther. 2014;13(6):1442-1456.

33. Tsai K, Tomomori-Sato C, Sato S, Conaway RC, Conaway J, Asturias F. Subunit architecture and functional modular rearrangements of the transcriptional mediator complex. Cell.

2014;158(2):1430-1444.

34. Moyo MB, Parker JB, Chakravarti D. Altered chromatin landscape and enhancer engagement underlie transcriptional dysregulation in MED12 mutant uterine leiomyomas. Nat Commun. 2020;11(1):1019.

35. Ficara F, Murphy MJ, Lin M, et al. Pbx1 regulates self-renewal of long-term hematopoietic stem cells by maintaining their quiescence. Cell Stem Cell. 2008;2(5):484-496.

36. Wei X, Yu L, Yi L. PBX1 promotes the cell proliferation via JAK2/STAT3 signaling in clear cell renal carcinoma. Biochem Biophys Res Commun. 2018;500(3):650-657.

37. King B, Boccalatte F, Moran-Crusio K, et al. The ubiquitin ligase Huwe1 regulates the maintenance and lymphoid commitment of hematopoietic stem cells. Nat Immunol. 2016;17(11):1312-1321.

38. Sugiura T, Miyamoto K. Characterization of TRIM31, upregulated in gastric adenocarcinoma, as a novel RBCC protein. J Cell Biochem. 2008;105(4):1081-1091.

39. Yu C, Chen S, Guo Y, et al. Oncogenic TRIM31 confers gemcitabine resistance in pancreatic cancer via activating the NF-κB signaling pathway. Theranostics. 2018;8(12):3224-3236.

40. Hanahan D, Robert AW. Hallmarks of cancer: the next generation. Cell. 2011;144(5):646-674.

41. Xiao Y, Deng T, Ming X, Xu J. TRIM31 promotes acute myeloid leukemia progression and sensitivity to daunorubicin through the Wnt/β-catenin signaling. Biosci Rep. 2020;40(4):BSR20194334.

42. Barbieri CE, Baca SC, Lawrence MS, et al. Exome sequencing identifies recurrent SPOP, FOXA1 and MED12 mutations in prostate cancer. Nat Genet. 2012;44(6):685-689.

43. Mäkinen N, Mehine M, Tolvanen J, et al. MED12, the mediator complex subunit 12 gene, is mutated at high frequency in uterine leiomyomas. Science. 2011;334(6053):252-255.

44. Aranda-Orgilles B, Saldaña-Meyer R, Wang E, et al. MED12 regulates HSC-specific enhancers independently of mediator kinase activity to control hematopoiesis. Cell Stem Cell. 2016;19(6):784-799.

45. Dale T, Clarke PA, Esdar C, et al. A selective chemical probe for exploring the role of CDK8 and CDK19 in human disease. Nat Chem Biol. 2015;11(12):973-980.

46. Adler AS, McCleland ML, Truong T, et al. CDK8 maintains tumor dedifferentiation and embryonic stem cell pluripotency. Cancer Res. 2012;72(8):2129-2139.

47. Alarcón C, Zaromytidou AI, Xi Q, et al. Nuclear CDKs drive Smad transcriptional activation and turnover in BMP and TGF-β pathways. Cell. 2009;139(4):757-769.

48. Menzl I, Zhang T, Berger-Becvar A, et al. A kinase-independent role for CDK8 in BCR-ABL1+ leukemia. Nat Commun. 2019;10(1):4741.

49. Tsai KL, Sato S, Tomomori-Sato C, et al. A conserved MediatorCDK8 kinase module association regulates Mediator-RNA polymerase II interaction. Nature Struct Mol Biol. 2013;20(5):611-619.

50. Alsadeq A, Schewe DM. Acute lymphoblastic leukemia of the central nervous system: on the role of PBX1. Haematologica. 2017;102(4):611-613.

ANKRD26 is a new regulator of type I cytokine receptor signaling in normal and pathological hematopoiesis

Francesca Basso-Valentina,1,2* Alessandro Donada,1,2* Vladimir T. Manchev,1 Manuel Lisetto,1 Nathalie Balayn,1 Jean Edouard Martin,1,2 Delphine Muller,1 Cecilia Paola Marin Oyarzun,1 Hélène Duparc,1 Brahim Arkoun,1 Alessandro Cumin,1,3,4 Lionel Faivre,5 Nathalie Droin,1 Ida Biunno,6,7 Alessandro Pecci,8,9 Alessandra Balduini,10,11 Najet Debili,1 Iléana Antony-Debré,1 Caroline Marty,1 William Vainchenker,1 Isabelle Plo,1 Remi Favier1,12 and Hana Raslova1

1INSERM, UMR 1287, Gustave Roussy, Université Paris Saclay, Equipe Labellisée Ligue Nationale Contre le Cancer, Villejuif, France; 2Ecole Doctorale Hematopoïèse, Oncogénèse et Biothérapie, Université Paris Diderot, Université Sorbonne Paris Cité, Paris, France; 3Dipartimento di Scienze della Vita, University of Trieste, Trieste, Italy; 4University of Paris Diderot, Paris, France; 5Unité Thérapie Cellulaire, Hôpital St Louis, Assistance PubliqueHôpitaux de Paris, Paris, France; 6Integrated Systems Engineering, Milan, Italy; 7Institute for Genetic and Biomedical Research-CNR, Milan, Italy; 8Department of Internal Medicine, University of Pavia, Pavia, Italy; 9General Medicine 1, IRCCS Policlinico San Matteo Foundation, Pavia, Italy; 10Department of Molecular Medicine, University of Pavia, Pavia, Italy; 11Department of Biomedical Engineering, Tufts University, Medford, OR, USA and 12Centre de Référence des Pathologies Plaquettaires, Hôpital Armand Trousseau, Assistance Publique-Hôpitaux de Paris, Paris, France

*FB-V and AD contributed equally as first authors.

Abstract

Correspondence: H. Raslova

hana.raslova@gustaveroussy.fr

Received: September 5, 2022.

Accepted: February 7, 2023.

Early view: February 16, 2023.

https://doi.org/10.3324/haematol.2022.282049

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Sustained ANKRD26 expression associated with germline ANKRD26 mutations causes thrombocytopenia 2 (THC2), an inherited platelet disorder associated with a predisposition to leukemia. Some patients also present with erythrocytosis and/or leukocytosis. Using multiple human-relevant in vitro models (cell lines, primary patients’ cells and patient-derived induced pluripotent stem cells) we demonstrate for the first time that ANKRD26 is expressed during the early steps of erythroid, megakaryocyte and granulocyte differentiation, and is necessary for progenitor cell proliferation. As differentiation progresses, ANKRD26 expression is progressively silenced, to complete the cellular maturation of the three myeloid lineages. In primary cells, abnormal ANKRD26 expression in committed progenitors directly affects the proliferation/differentiation balance for the three cell types. We show that ANKRD26 interacts with and crucially modulates the activity of MPL, EPOR and G-CSFR, three homodimeric type I cytokine receptors that regulate blood cell production. Higher than normal levels of ANKRD26 prevent the receptor internalization that leads to increased signaling and cytokine hypersensitivity. These findings afford evidence how ANKRD26 overexpression or the absence of its silencing during differentiation is responsible for myeloid blood cell abnormalities in patients with THC2.

Introduction

ANKRD26 (ankyrin repeat domain containing 26) is the ancestor of a family of primate-specific genes termed POTE (Prostate-, Ovary, Testis-, and placenta-Expressed genes) with a gene expression restricted to a few normal tissues and a larger number of pathological tissues.1,2 It encodes for a protein of 192 kDa, containing both spectrin helices and ankyrin repeats, protein domains known to interact with cytoskeletal and signaling proteins.3-5 Germline mutations in the regulatory region of the gene encoding ANKRD26 are associated with thrombocytopenia 2 (THC2).6 All THC2 patients present with moderate thrombocytopenia and mild bleed-

ing, while a smaller number of patients present with erythrocytosis and/or leukocytosis.7 Importantly, 10% of THC2 patients develop myeloid malignancies, classifying THC2 as an inherited thrombocytopenia predisposing to leukemia.7-9 We have previously demonstrated that ANKRD26 is indeed involved in modulating thrombopoietin (TPO)-dependent signaling10 and its expression increases the intensity of mitogen activated protein kinase (MAPK)/extracellular signalregulated kinase (ERK)1/2 activation. Coupled with clinical observations, we made the hypothesis that ANKRD26 plays a broader role in hematopoiesis by modifying cytokine-mediated cell signaling.

Type I receptors, particularly the homodimeric ones, play

a fundamental role in myeloid blood cell production. Granulocyte colony-stimulating factor receptor (G-CSFR) activation drives neutrophil differentiation,11 erythropoietin (EPO)/erythropoietin receptor (EPOR) signaling is indispensable for red blood cell production,12 and TPO/TPO receptor (MPL) signaling is essential for megakaryocyte differentiation13 and hematopoietic stem cell quiescence.14 Loss- or gain-of-function mutations of the homodimeric receptors are associated with several inherited or acquired diseases,15-18 as are mutations in proteins directly modulating ligand/receptor signaling.19 These receptors are traditionally activated through ligand binding, and their cell surface density is tightly regulated to prevent aberrant activation.20 After ligand binding, Janus kinases (JAK) activate and phosphorylate the receptor cytoplasmic domain. This phosphorylation allows the recruitment of several signaling molecules such as the signal transducer and activator of transcription (STAT) proteins, which in turn are phosphorylated by JAK.21,22 Other pathways commonly activated by type I cytokine receptors which are also dependent on JAK activation are the phosphoinositide 3-kinase (PI3K)/AKT pathway and the ERK pathway, both of which are essential for the control of cell proliferation, cell survival and differentiation.23,24 To protect cells from over-stimulation, several mechanisms exist to attenuate signal transduction: recruitment of tyrosine phosphatases to inactivate JAK;15 synthesis of suppressor of cytokine signaling (SOCS) proteins that can inactivate JAK;25 receptor internalization and subsequent degradation through the proteasome and lysosome pathways;26,27 and, finally, the modification of ligand-receptor kinetic parameters.28 There is therefore strong interest in identifying other proteins able to modulate the intensity of these crucial signaling axes. Using different human-relevant models, we identified a crucial role for ANKRD26 in modulating TPO-, EPO- and G-CSF-dependent cell signaling, and thereby in the generation of the hematopoietic cells controlled by these cytokines.

Methods

Study approval

Blood samples from patients and healthy subjects were collected after informed written consent and obtained in accordance with the Declaration of Helsinki. The study was approved by the Comité de Protection des Personnes (CPP 2020T2-02) and by AP-HP, Hôpital Saint-Louis, Unité de Thérapie Cellulaire, CRB-Banque de Sang de Cordon, Paris, France (AC-2016-2759).

Animal experiments were performed in accordance with 2010/63/UE European legislation and decree N 2013-118 of French legislation and recorded under protocol number APAFIS# 2016-008-7175.

Primary cell culture

CD34+ cells were isolated from umbilical cord blood or peripheral blood by positive selection using an immunomagnetic bead cell-sorting system (AutoMacs; Miltenyi Biotec).

UT7 cell lines

Human UT7 megakaryoblastic cells expressing HA_MPL29 and HA_EPOR30 have been previously reported, and a human UT7 line expressing HA_G-CSFR was obtained by the transduction of UT7 parental cells with a retrovirus encoding HA_G-CSFR.

Generation and expansion of induced pluripotent stem cells

Patients’ CD34+ cells were expanded in serum-free medium containing EPO (1 U/mL), FLT3L (10 ng/mL), G-CSF (20 ng/mL), interleukin (IL)-3 (10 ng/mL), IL-6 (10 ng/mL), stem cell factor (SCF; 25 ng/mL), TPO (10 ng/mL) and GM-CSF (10 ng/mL) for 6 days and transduced with VSV-G pseudotyped retroviruses encoding for the OSKM combination (OCT4, SOX2, KLF4 and c-MYC). Colonies with an ES-like morphology were manually isolated, expanded for a reduced number of passages and frozen. The induced pluripotent stem cell (iPSC) control cell lines C1, C2 and C3 were already established and previously characterized.31-33

Clonogenic potential of primary cells in semi-solid culture Methylcellulose culture assay

Hematopoietic progenitors (CD34+CD43+ for iPSC and CD34+ for primary cells) were plated in human methylcellulose medium H4434 (Stemcell Technologies), containing recombinant human cytokines and scored for the presence of colonies 14 days after.

Fibrin clot assay

To assess colony-forming unit - megakaryocyte (CFU-MK) potential, CD34+GFP+ or CD34+Cherry+ sorted cells were seeded at 1.5x103 cells/mL in triplicate in fibrin clot medium, as previously described.32

Additional information

Other protocols are detailed in the Online Supplementary Materials and Methods and all the primers, antibodies and other reagents are listed in Online Supplementary Table S1.

Statistics

All data are shown as mean ± standard deviation unless specified differently. The statistical analyses were performed using PRISM software. Statistical significance was established using a Student t test, as specified in the legends, or one-way analysis of variance (ANOVA), followed by all pairwise multiple comparison procedures. Differences were considered statistically significant for P values <0.05.

Results

ANKRD26 regulates the early steps of megakaryopoiesis

We have previously shown that the overexpression of ANKRD26 in megakaryocytes of THC2 patients leads to thrombocytopenia in these patients as a result of deregulation of the TPO/MPL signaling axis10 at the late stages of megakaryocyte differentiation. Indeed, the abnormal presence of ANKRD26 in the terminal phases of megakaryocyte maturation and consequent stronger MAPK-pathway activation cause a strong reduction in the number of pro-

platelet-forming megakaryocytes in THC2 patients. We have transduced cord blood CD34+ progenitor cells with a lentiviral vector encoding ANKRD26 (Online Supplementary Figure S1A), in order to assess the impact of this protein on the early steps of megakaryopoiesis. ANKRD26 overexpression only slightly increased the number of megakaryocyte progenitors at 10 ng of TPO (Figure 1A) without affecting their proliferation (Figure 1B) but, as previously shown, completely prevented proplatelet formation (Figure 1C, D) without disturbing MPL expression level at the megakaryocyte cell surface (Online Supplementary Figure S1B).

Figure 1. ANKRD26 overexpression alters late but not early stages of megakaryopoiesis. (A, B) Primary CD34+ cells were transduced with empty lentivirus (EV) or lentivirus encoding ANKRD26_V5 (ANKRD26) and Cherry. CD34+-Cherry cells were sorted at day 2 after transduction and cultured in semi-solid medium (fibrin clot medium) in the presence of stem cell factor (SCF) and thrombopoietin (TPO). Colonies derived from megakaryocyte progenitors were assessed by anti-CD41 antibody labeling. (A) Plating efficiency of megakaryocyte progenitors (CFU-MK) was only slightly increased after ANKRD26 overexpression in the presence of 10 ng/mL TPO. (B) The proliferative rate of CFU-MK was not affected by increased ANKRD26 level. Proliferation was assessed according to the size of the colonies (three types of colonies were scored: <20 cells/colony, <50 cells/colony and >50 cells/colony), each colony corresponding to one progenitor. The averages of three independent experiments are shown as the mean ± standard error of mean. **P<0.01; paired t test, ns: non-significant. (C, D) Primary CD34+ cells were transduced with EV or lentivirus encoding ANKRD26_V5 (ANKRD26) and a gene resistant to hygromycin B. ANKRD26 overexpression in primary CD34+ cells cultured in the presence of SCF and TPO and hygromycin B prevented proplatelet formation, evaluated by inverted light microscopy at day 14 of culture. (C) Representative pictures of one experiment are shown. (D) The histogram represents the averages of five independent experiments as mean ± standard deviation. ***P<0.005; paired t test. MK: megakaryocytes.

ANKRD26 has been shown to be expressed more in CD34+ progenitor cells than in mature megakaryocytes, suggesting that its expression is necessary during the early stages of hematopoiesis10,34 (Online Supplementary Figure S2A, B). As overexpression had no major effect on CD34+ cells, we decided to investigate whether ANKRD26 downregulation at the CD34+ progenitor level could affect the number of megakaryocyte progenitors (CFU-MK). We transduced cord blood CD34+ hematopoietic progenitors with lentiviruses encoding short hairpin RNA (sh) against ANKRD26 (shANK_1

or shANK_2) (Online Supplementary Figure S1C) and cultured them in fibrin clot medium, in the presence of SCF and increasing doses of TPO. We observed a 50% decrease in CFU-MK plating efficiency (Figure 2A), as well as a reduction of their proliferative rate, as attested by the lower number of megakaryocytes per colony (Figure 2B). This result was confirmed in liquid culture in the presence of SCF and TPO (Online Supplementary Figure S1D, E). Interestingly, the reduction in ANKRD26 also led to a lower frequency of mature megakaryocytes (CD41+CD42+) at day 10 of culture

Figure 2. ANKRD26 is necessary for early but not late stages of megakaryopoiesis. (A-F) CD34+ cells were transduced with shSCR or shANK and GFP encoding lentiviruses, sorted 2 days after transduction and cultured in semi-solid medium (fibrin clot medium) in the presence of stem cell factor (SCF) and thrombopoietin (TPO) (A, B) or in liquid medium (C-F). (A, B) Colonies derived from megakaryocyte progenitors were assessed by anti-CD41 antibody. (A) Plating efficiency of megakaryocyte progenitors (CFU-MK) was decreased after inhibition of ANKRD26 (shANK), in the presence of different TPO doses. (B) The proliferation rate of megakaryocyte progenitors was decreased after ANKRD26 inhibition, as shown by the increase of colonies composed of less than 20 cells and the decrease of those with more than 50 cells. The averages of three independent experiments are shown as the mean ± standard error of mean. *P<0.05; **P<0.01; ns: non-significant; paired t test. (C) The inhibition of ANKRD26 led to a decreased frequency of megakaryocytes (CD41+CD42+ cells) at day 10 of culture. (D-F). In contrast, an increase in the ploidy level (D) and in the percentage of proplatelet-forming megakaryocytes was detected after ANKRD26 inhibition (E, F). (E) Representative pictures of one experiment are shown. (C, D, F) The histograms represent the averages of three (D), four (C) or six (F) independent experiments as the mean ± standard deviation, *P<0.05; ****P<0.001, paired t test. All the ANKRD26 inhibition experiments were performed at least twice with shANK_1 and once with shANK_2. MK: megakaryocytes.

(Figure 2C), but with a higher ploidy level (Figure 2D) and enhanced proplatelet formation capacity (Figure 2E, F). Together these results demonstrate that ANKRD26 is required for optimal proliferation and differentiation of megakaryocyte progenitors, but that its downregulation is required for thrombopoiesis.

ANKRD26 regulates the early steps of granulopoiesis

ANKRD26 expression decreases not only during normal

megakaryocyte differentiation but also during erythroid and granulocytic differentiation (Online Supplementary Figure S2A, B). To confirm that ANKRD26 deregulation could be responsible for the leukocytosis detected in some THC2 patients, we studied granulopoiesis in primary patients’ cells (Online Supplementary Table S2). Although ANKRD26 expression was not different in progenitors (CD34+ cells) isolated from the peripheral blood of THC2 patients, a significant increase in ANKRD26 levels was de-

Figure 3. ANKRD26 regulates early stages of granulopoiesis. Primary CD34+ cells obtained from peripheral blood of patients with thrombocytopenia 2 with different 5’ untranslated region mutations were induced to granulocytic differentiation in the presence of granulocyte colony-stimulating factor (G-CSF), interleukin-3 (IL-3) and stem cell factor (SCF). (A) ANKRD26 expression in patients’ CD34+ cells [P] was similar to that in control CD34+ cells [C] obtained from healthy individuals, but increased during in vitro granulocytic differentiation, with a peak at day 8 of culture. ANKRD26 transcript level was normalized to PPIA. Averages are shown for four (for CD34+ cells) and two (for granulocytic differentiation) independent experiments. (B) The number of patients’ myeloid progenitors (CFU-G) was significantly higher, compared to control progenitors, as assessed by a methylcellulose assay. The averages of five independent experiments are shown as mean ± standard error of mean. *P<0.05; t test with Mann-Whitney correction. (C) Proliferation assay performed in liquid culture supplemented with G-CSF, SCF and IL-3 showed a significant increase in cell number for patients’ samples at days 11 and 14 of culture. The cell count was normalized to day 0. Averages of three independent experiments are shown as mean ± standard error of mean, *P<0.05; **P<0.01; paired t test. (D) May-Grünwald Giemsa staining of samples from two patients and one control at day 15 of culture showed an increase in the proportion of immature cells (myeloblasts, promyelocytes and myelocytes) and a decrease in the proportion of more mature cells (metamyelocytes and polynuclear neutrophils). (E, F) Effect of ANKRD26 inhibition on the granulocytic lineage. CD34+ cells were transduced with lentiviruses encoding shSCR and shANK (shANK_1 or shANK_2, respectively). (E) CD34+-GFP cells were sorted 2 days after transduction and grown in semi-solid medium (methylcellulose) in the presence of 25 ng/mL SCF and different doses of G-CSF. Granulocytic progenitors (CFU-G) were enumerated at day 14 of culture. (F) CD34+-GFP cells were sorted 2 days after transduction and grown in liquid medium in the presence of 25 ng/mL SCF, 10 ng/mL IL-3 and 20 ng/mL G-CSF. Proliferation assays showed a significant decrease in shANK transduced cell number at days 7, 11, 14 and 18. The averages of three independent experiments are shown as mean ± standard error of mean, *P<0.05; **P<0.01; paired t test. PNN: polynuclear neutrophils.

tected through in vitro granulocyte differentiation (Figure 3A). To evaluate the biological effects of increased ANKRD26 expression, we assessed the colony-forming potential of patients’ CD34+ cells in semi-solid medium and their proliferative rate in the presence of SCF, IL-3 and GCSF.35 We detected a more than two-fold increase in granulocyte colony (CFU-G) numbers (Figure 3B), as well as an increased proliferative capacity (Figure 3C), compared to healthy controls. We performed May-Grünwald Giemsa staining at day 15 of culture for two patients, and detected an increased frequency of immature cells (myeloblasts, promyelocytes and myelocytes) compared to controls, while the percentages of metamyelocytes and polynuclear neutrophils were clearly decreased (Figure 3D, Online Supplementary Figure S3A). Nevertheless, the percentages of mature granulocytes for patients’ cells increased over time (Online Supplementary Figure S3B), suggesting a maturation delay due to the increased proliferation, and not a complete blockage. Overall, these results show that ANKRD26 overexpression leads to defective granulopoiesis in THC2 patients, with increased proliferation and delayed maturation. It should be noted that different subtypes of acute myeloid leukemia express higher ANKRD26 levels as compared to the corresponding committed progenitors (Online Supplementary Figure S2B) suggesting that ANKRD26 overexpression may play an important role in the proliferative phenotype of leukemic cells. To further confirm the role of ANKRD26 in the granulocytic lineage, we transduced the cord blood CD34+ progenitors with lentiviruses encoding shANK and cultured them in increasing doses of G-CSF. We observed a reduction in the number of granulocytic progenitor-derived colonies (CFU-G) in methylcellulose assay in the presence of 2 and 20 ng of G-CSF (Figure 3E), and a decreased proliferation rate in liquid culture (Figure 3F) with accelerated differentiation (Online Supplementary Figure S3C), and no effect on apoptosis (Online Supplementary Figure S3D).

Patient-derived induced pluripotent stem cells recapitulate the defective granulopoiesis

To overcome the limitations associated with the rarity of THC2 patients, we derived iPSC lines from two patients harboring two different mutations: c.-118A>C (ANK1) and c.-127delAT (ANK2) (Online Supplementary Table S3). The iPSC lines were generated from CD34+ cells isolated from peripheral blood via integrative reprogramming and were characterized for their phenotypic and functional pluripotency (Online Supplementary Figure S4). Three different clones were studied for each patient-derived iPSC. As a control, we used three different iPSC lines derived from healthy individuals.31-33

We used a serum-free, xeno-free differentiation protocol (Online Supplementary Figure S5). We isolated the hema-

topoietic progenitors after 14 days of culture (CD34 + CD43 + ), a time described to give a bias towards granulo-monocytic differentiation output.33 ANKRD26 expression in patients’ iPSC-derived CD34+CD43+ cells was three times higher than in controls (Figure 4A); this difference with respect to ANKRD26 expression in cord blood-derived early progenitors (CD34+ cells) can be explained by the fact that iPSC-derived CD34+CD43+ cells are developmentally different from adult CD34+ cells and are already committed to the myeloid lineage. We detected three times more CFU-G colonies in semi-solid medium compared to the numbers from healthy controls (Figure 4B). We also found an increased proliferative rate in liquid culture in the presence of SCF, IL-3 and G-CSF (Figure 4C). CD15 + granulocytic cells derived from CD34 + CD43 + patients’ cells expressed higher levels of ANKRD26 as compared to control CD15+ cells (Figure 4A).

Flow cytometry and analysis of May-Grünwald Giemsa staining revealed a delay in granulocytic differentiation at days 17-23 of culture ( Online Supplementary Figure S6A): myeloid cells derived from patients’ iPSC displayed less than 10% of mature CD11b + CD15 + CD14 - cells, compared to about 30% for the control cell lines at day 23 (Figure 4D, Online Supplementary Figure S6A ). Overall, these results clearly show defective granulopoiesis, similar to that observed for primary cells. To gain more insights into this defect, we investigated the transcriptomic profile of the patients’ granulocytic progenitors: iPSC-derived granulocytic progenitors (CD43+CD11b+CD14-) were sorted and profiled by RNA sequencing. We found 24 genes significantly upregulated and seven significantly downregulated ( P <1x10-5) in patients’ cells compared to controls (Figure 4E, Online Supplementary Table S4 ). Gene set enrichment analysis performed with two different databases (KGEA and GOBP) revealed a significant enrichment in the JAK/STAT signaling pathway in patient-derived iPSC progenitors (Figure 4F) with a tendency to increased granulocytic count and delayed myeloid development (Online Supplementary Figure S6B) confirming the results obtained with primary patients’ cells and with shANK in cord blood-derived progenitors.

Within the significantly upregulated genes, the CCNI2 gene (Cyclin I family member 2), a cyclin responsible for cyclin-dependent kinase 5 (CDK5) activity,36 caught our attention. CDK5 is generally not directly involved in cell cycle regulation and has been described to phosphorylate NOXA thereby promoting its cytosolic sequestration and suppression of apoptosis in leukemic cells.37 However, depletion of CCNI2 has been shown to inhibit cell cycle progression and proliferation.36 CCNI2 expression was almost completely absent from control iPSC-derived CD43+CD11b+CD14- cells. After validation of CCNI2 upregulation in patients’ iPSC-derived progenitors by quantitative

Figure 4. Increased ANKRD26 expression level enhances proliferation of granulocyte progenitors in a model of induced pluripotent stem cells, through an enhanced JAK/STAT pathway. CD34+CD43+ progenitors derived from induced pluripotent stem cells (iPSC) from patients [ANK] and controls [C] were differentiated into granulocytes in the presence of granulocyte colony-stimulating factor (G-CSF), interleukin-3 (IL-3) and stem cell factor (SCF). (A) ANKRD26 expression in CD34+CD43+ progenitors (n=4) and in CD11b+CD15+CD14- cells (n=6) derived from patients’ iPSC was increased when compared to that in controls. ANKRD26 transcript was normalized to PPIA. Results are shown as mean ± standard deviation, *P<0.05; **P<0.01; paired t test. (B) The number of patients’ granulocytic progenitors (CFU-G) derived from iPSC was significantly higher than the control, as assessed by methylcellulose assay. The averages of seven independent experiments are shown as mean ± standard deviation. **P<0.01; paired t test. (C) The proliferation rate was significantly increased at days 17, 20 and 22 of culture for patients’ cells as compared to controls. The averages of four independent experiments are shown as the mean ± standard error of mean. *P<0.05; paired t test. (D) CD11b+CD15+CD14- cell frequency analyzed at days 21-23 reflected a delay in maturation for patients’ cells. Averages of 14 experiments for controls and ten for patients’ cells are shown as the mean ± standard error of mean. ***P<0.005, unpaired t test. (E) An RNA-sequencing assay was performed on CD43+CD11b+CD14- progenitors sorted at day 16 of culture. Seven downregulated and 24 upregulated genes in patients’ cells (n=3 for each, ANK1 and ANK2) (P<1x10-5) were identified compared to controls (n=1 for C1 and C3, n=3 for C2). (F) Gene set enrichment analysis for the KEGG_JAK_STAT_signaling pathway and for the receptor signaling pathway via STAT (GO:0007259, GOBP_RECEPTOR_SIGNALING_PATHWAY_VIA_STAT) in ANK versus control iPSC. NES: normalized enrichment score; FDR: false discovery rate.

Figure 5. ANKRD26 regulates the early stage of erythropoiesis. (A, B) CD34+ cells were transduced with lentiviruses encoding shSCR or shANK and erythroid progenitors (BFU-E) were grown in semi-solid medium (methylcellulose) in the presence of 25 ng/mL stem cell factor (SCF) and different doses of erythropoietin (EPO) and enumerated at day 14 of culture. (A) ANKRD26 inhibition led to a significant decrease in BFU-E number. Averages of three independent experiments are shown as the mean ± standard deviation. *P<0.05; one-tailed t test with Mann-Whitney correction. (B) Representative pictures of BFU-E colonies showing that ANKRD26 inhibition led to a lack of hemoglobinization of BFU-E-derived colonies, both at 0.1 and 0.01 U/mL of EPO. (C-F) Transduced CD34+ cells (C-E) or primary patients’ CD34+ progenitors (F) were cultured in liquid medium in the presence of EPO (1 U/mL), SCF and interleukin-3 for 18 days. (C, D) Kinetics of erythroid differentiation assessed by fluorescence activated cell sorting (C) showed that ANKRD26 inhibition leads to a delay in differentiation (D). CD36-GPA- cells represent immature, CD36+GPA+ intermediate and CD36-GPA+ mature erythroid cells. Statistical analysis of different populations is shown as the average of three independent experiments (mean ± standard deviation, *P<0.05, ***P<0.005, 2-way analysis of variance with multiple comparisons). (E, F) ANKRD26 expression level affected proliferation of CD34+ cells grown in erythroid conditions. (E) Inhibition of ANKRD26 led to a significant decrease in the proliferation of transduced CD34+ cells cultured in erythroid conditions. The average of three independent experiments is shown as mean ± standard deviation, *P<0.05; paired t test. (F) Proliferation rate of primary CD34+ cells from patients with thrombocytopenia 2 cultured in erythroid conditions was significantly higher compared to controls. The average of three independent experiments is shown as mean ± standard deviation, *P<0.05; t test with MannWhitney correction. D: day; GPA: glycophorin A.

real-time polymerase chain reaction analysis (Online Supplementary Figure S6C), we transduced the CD34+CD43+ progenitors at day 14 of culture with a lentivirus encoding shCCNI2 (shCCNI2_1 and shCCNI2_2), sorted the GFP + cells 48 hours later and measured their proliferative rate in granulocytic conditions at days 19, 21 and 23. The inhibition of CCNI2 by shRNA significantly reduced the number of granulocytes in patient-derived iPSC (Online Supplementary Figure S6D). Although an in silico analysis of the CCNI2 promoter revealed the presence of two STAT5 binding sites (data not shown ), the implication of this in the observed phenotype needs to be investigated further.

ANKRD26 regulates the early steps of erythropoiesis

The role of ANKRD26 in the regulation of erythropoiesis was investigated in cord blood CD34 + cells. ANKRD26 downregulation led to a reduction in the plating efficiency of erythroid progenitors (BFU-E) in semi-solid assay, in the presence of SCF and increasing doses of EPO (Figure 5A). Moreover, colonies from shANK progenitors were paler, indicating a decrease in hemoglobin content at lower EPO doses (0.01 and 0.1 U/mL), a sign of impaired differentiation (Figure 5B). A delay of erythroid differentiation in the presence of lower ANKRD26 levels was confirmed in a kinetics assay in liquid medium, in the presence of EPO, SCF and IL-3 (Figure 5C, D). Finally, cord blood CD34+ cells transduced with shANK and cultured in erythroid conditions showed reduced proliferation compared to the control (Figure 5E). This is consistent with the increased proliferation observed for THC2 patient-derived cells cultured in the same conditions (Figure 5F). We measured the serum EPO levels in seven THC2 patients and detected normal values for three patients without erythrocytosis. In four patients with erythrocytosis, the serum EPO levels were slightly lower and either below or at the lower limit of normal values, suggesting that erythrocytosis was cell autonomous (Online Supplementary Table S5).

ANKRD26 regulates MPL, G-CSFR and EPO-mediated signaling

To further understand how ANKRD26 could modify the proliferation and differentiation of hematopoietic progenitors under TPO, G-CSF and EPO stimulation, we investigated whether it may interfere with the signaling of their cognate receptors. For these studies we used two cell lines, the murine pro-B Ba/F3 and the human erythro-MK UT7 cell lines which were rendered dependent on TPO, GCSF or EPO via retrovirus encoding MPL, G-CSFR or EPOR In the Ba/F3 cell line we overexpressed human ANKRD26 and in the UT7 cell line expressing a high level of human ANKRD26, we downregulated its expression by two different shRNA.

ANKRD26 levels did not affect MPL, G-SCFR and EPOR levels at the cell surface either at the steady state (Figure 6A , Online Supplementary Figures S7, S8A and S9A) or after overnight starvation (Figure 6B, Online Supplementary Figures S7, S8B, and S9B). We confirmed our previous observation10 that higher ANKRD26 levels induced stronger STAT5, ERK, AKT and also STAT3 phosphorylation at 10 ng/mL of TPO (Figure 6C, Online Supplementary Figure S10A). Moreover, we observed that a very low TPO dose (0.01 ng/mL) was still able to stimulate TPO/MPL signaling in cells expressing higher levels of ANKRD26 (Figure 6D). To determine whether ANKRD26 increases MPL activation and signaling by different ligands, we assessed the response to the TPO-mimetic eltrombopag which, in contrast to TPO, binds to the transmembrane region involving H499,38 inducing downstream signaling in a different conformation than that done by TPO.39 Higher levels of ANKRD26 led to the same hypersensitivity for eltrombopag as observed for TPO (Online Supplementary Figure S10B, C), suggesting that the main effect of ANKRD26 is to stabilize MPL surface expression without changing its conformation.

Using UT7 cells, we showed that the higher ANKRD26 levels led to stronger G-CSF-mediated STAT3 and AKT phosphorylation, with a similar tendency for ERK1/2 (On-

ARTICLE - ANKRD26 regulates cytokine-mediated signaling

Figure 6. ANKRD26 regulates MPL-mediated signaling. UT7 or Ba/F3 cell lines expressing MPL were transduced with lentiviruses harboring control scramble shRNA (shSCR), shANKRD26 (shANK), ANKRD26 cDNA or empty vector (EV). (A, B) Downregulation (A) or upregulation (B) of ANKRD26 expression level did not affect the expression of MPL measured with anti-MPL antibody. The receptor levels are presented as median fluorescence intensity at the cell surface. The averages of three independent experiments are shown as mean ± standard deviation (2 with shANK1_1 and 1 with shANK_2). **P<0.01; ****P<0.001; ns: non-significant, t test with Mann-Whitney correction. (C, D) One of at least three independent western blot (WB) analyses on signaling proteins in Ba/F3 and UT7 cells, at different times after stimulation with 10 ng/mL of thrombopoietin (TPO) (C), and different TPO doses (D) at 10 min. The histograms show quantification of the WB representing averages of three or four independent experiments as mean ± standard deviation. *P<0.05; **P<0.01; ***P<0.005; paired t test. (E, F) Number of UT7/MPL (E) and Ba/F3_MPL (F) cells measured at day 4 of culture, with three different doses of TPO are shown as mean ± standard error of mean of three independent experiments, *P<0.05, **P<0.01; ns: non-significant, t test with Mann-Whitney correction. APC: allophycocyanin; MFI: median fluorescence intensity, PE: phycoerythrin; NS: not stimulated.

line Supplementary Figure S8C). The overexpression of human ANKRD26 in the Ba/F3 cell line expressing G-CSFR did not affect STAT3 activation but led to enhanced ERK and AKT phosphorylation (Online Supplementary Figure S8D). Using UT7/G-CSFR cells, we also observed the activation of STAT3 at a lower G-CSF dose (0.2 ng/mL), in the presence of higher ANKRD26 levels (Online Supplementary Figure S8E).

Finally, we showed that higher ANKRD26 levels led to stronger and more sustained EPO/EPOR-mediated activation of ERK1/2, AKT and, to a lesser extent, of STAT5 in UT7/EPOR cells (Online Supplementary Figure S9C). An increased level of AKT phosphorylation and the same tendency for ERK1/2 were also detected in the Ba/F3 cell line overexpressing ANKRD26 and EPOR (Online Supplementary Figure S9D). Moreover, a significantly increased level of STAT5 phosphorylation was detected at 0.1 U/mL of EPO in the presence of higher ANKRD26 levels (Online Supplementary Figure S9E).

To analyze the biological consequence of the TPO, G-CSF and EPO hypersensitivity, we measured the proliferation rate of UT7/MPL, UT7/G-CSFR and UT7/EPOR cells expressing shSCR and shANK at different cytokine concentrations. We observed that at 0.1 ng/mL of TPO only the UT7/MPL cells with higher ANKRD26 levels were able to proliferate, after 4 days of culture. At higher doses of TPO (10 ng/mL), no difference in the proliferation rate was detected (Figure 6E), a finding that corroborates the idea of receptor saturation at higher doses. A similar result was obtained in the presence of eltrombopag (Online Supplementary Figure S10C). In line with these results, the overexpression of ANKRD26 in murine Ba/F3_MPL cells led to an increased proliferation rate. At 0.1 ng/mL of TPO, only cells overexpressing ANKRD26 were able to grow (Figure 6F).

In a similar manner, UT7/G-CSFR cells expressing higher levels of ANKRD26 were able to proliferate in the presence of 0.2 ng/mL of G-CSF, unlike cells expressing lower ANKRD26 levels. Using higher doses of G-CSF (2 ng/mL), ANKRD26 downregulation led to a decreased proliferation rate, suggesting a progressive saturation of the cell surface pool of available G-CSFR (Online Supplementary Figure S8F). Finally, UT7/EPOR cells expressing higher ANKRD26 levels proliferated more at low EPO concentration (0.1 U/mL) than

the control. This proliferation gap was absent at higher EPO doses (1 U/mL) (Online Supplementary Figure S9F). Overall, these results demonstrate that higher ANKRD26 levels lead to MPL, G-CSF and EPO hypersensitivity.

ANKRD26 interacts with the homodimeric type I receptors

We then explored whether this effect of ANKRD26 was a direct or indirect effect on receptor signaling. First we investigated whether ANKRD26 interacts with MPL, G-CSFR and EPOR respectively. To this end, HEK293T cells were transduced with retroviruses encoding, respectively, HA_MPL, HA_G-CSFR or HA_EPOR and with a lentivirus encoding ANKRD26_V5. The interaction of ANKRD26 with each of these three homodimeric receptors was demonstrated by co-immunoprecipitation assays (Figure 7A-C). As a proof of concept, the ANKRD26 and MPL interaction was confirmed by proximity ligation assays in the UT7 cell line (Online Supplementary Figure S11) and in γ2A cells, in which it was shown that JAK2 was dispensable for this interaction, although its presence slightly enhanced it (Figure 7D-F). Second, as ANK repeat-containing proteins have been shown to interact with different receptors and to participate in their internalization, one possible mechanism was that ANKRD26 affects receptor-mediated signaling by interfering with their internalization. To investigate this hypothesis, we used the murine Ba/F3 cell line overexpressing ANKRD26 and MPL, G-CSFR or EPOR, as this cell line is a better model for internalization studies than UT7 cells.40 After overnight starvation, ANKRD26 did not modify receptor expression at the cell surface (Figure 6B, Online Supplementary Figures S8B and S9B). The MPL-expressing Ba/F3 cells were then exposed to 50 ng/mL of TPO. In the absence of human ANKRD26, about 40% of MPL was internalized 15 min after TPO exposure and additional exposure only slightly increased the quantity of internalized MPL. In contrast, the presence of human ANKRD26 almost completely abrogated MPL internalization, even after 60 min of TPO exposure (Figure 8A). Similarly, the overexpression of human ANKRD26 in the BaF/3 cell line expressing G-CSFR or EPOR led to a defect in G-CSFR or EPOR internalization at 15 and 30 min of stimulation of starved cells with G-CSF and EPO, respectively (Figure 8B, C). These results demonstrate that receptor internalization is finely regulated by ANKRD26 levels.

Figure 7. ANKRD26 interacts with homodimeric type I receptors. (A-C) Co-immunoprecipitation assay performed in HEK293 cells showing the presence of ANKRD26 and MPL (A), ANKRD26 and G-CSFR (B), and ANKRD26 and EPOR (C) in the same protein complex. For each receptor, one of three independent experiments with similar results are shown. Input represents western blot analysis of cells expressing empty vectors or cells co-expressing ANKRD26_V5 and HA_MPL (A), ANKRD26_V5 and HA_GCSFR (B), or ANKRD26_V5 and HA_EPOR (C). The antibodies used were anti-V5 (for ANKRD26_V5), anti-MPL (for HA-MPL), and anti-HA (for HA-G-CSFR and HA-EPOR). (D-F) Proximity ligation assay for the ANKRD26 and MPL interaction. FLAG_ANKRD26 (ANK), HA_MPL (MPL) and JAK2 were overexpressed in γ2A cells (cells not expressing endogenous Jak2). Monoclonal anti-FLAG antibody was used for ANKRD26 and polyclonal anti-HA for MPL. (D) Representative pictures of the proximity ligation assay for the ANKRD26 and MPL interaction. The red staining represents the ANKRD26/MPL interaction, scale bar = 30 μM. (E) Data represent the mean of two independent experiments. (F) Data represent the number of dots per positive cell. At least 40 positive cells were analyzed for each condition. ***P<0.005, ****P<0.001, t test with Mann-Whitney correction. IP: immunoprecipitation, IB: immunoblot.

Discussion

In this work, we thoroughly demonstrate the role of ANKRD26 in the regulation of three myeloid lineages by modulating the activity of three type I cytokine receptors that are essential in normal hematopoiesis.

We confirmed a role for ANKRD26 in the regulation of the megakaryocyte lineage, in which its presence is crucial in the early steps of differentiation, but gets in the way of correct terminal maturation. Indeed, shRNA-mediated decrease of ANKRD26 expression in CD34+ progenitors greatly reduced plating efficiency and the proliferative potential of megakaryocyte progenitors. On the other hand, overexpression of ANKRD26 prevented correct proplatelet formation in megakaryocytes, which agrees with observations in THC2 patients.10

For the first time we show here that ANKRD26 also plays a crucial role in the granulocytic and erythroid lineages, the other two main myeloid lineages governed by type I homodimeric receptors, G-CSFR and EPOR, respectively. At the immature stage, hematopoietic stem and progenitor cells express basal levels of ANKRD26. Gene downregulation in these cells leads to decreased proliferation and clonogenic potential, for the megakaryocytic, granulocytic and erythroid lineages, suggesting an important role of this protein in committed progenitors. Physiologically, ANKRD26 is progressively silenced along the cellular differentiation and maturation of these lineages. Our results show that abnormal ANKRD26 expression in the gra-

nulocytic lineage of THC2 patients affects the normal process of cellular proliferation/differentiation, which could lead to the leukocytosis reported in some patients. Whether this mechanism could be responsible for the establishment of a fertile substrate prone to the acquisition of secondary mutations remains to be explored. Previously we have shown that the sustained ANKRD26 expression in the late phase of megakaryopoiesis leading to a defect in proplatelet formation was due to a deregulation of TPO/MPL signaling.10 Here we demonstrated that the modulation of ANKRD26 expression levels modifies TPO sensitivity. Interestingly, although ANKRD26 expression does not increase MPL level at the cell surface, its higher expression increases MPL signaling activity leading to enhanced proliferation. This observation prompted us to hypothesize that ANKRD26 could be part of a complex that regulates the internalization of the receptor. By different approaches, we collected evidence that ANKRD26 and MPL interact, although further studies will be needed to determine whether this is a direct or indirect process. Similarly we showed that both G-CSFR and EPOR interact with ANKRD26. Interestingly, thrombocytopenia rather than thrombocytosis is detected in THC2 patients. This observation could be explained by the fact that despite an amplification of megakaryocyte progenitors due to increased JAK2/STAT signaling at the early stage of megakaryopoiesis, the inactivation of the MAPK-pathway necessary for correct proplatelet formation is not achieved.10 In erythroid and granulocytic lineages, ANKRD26 modulates the proliferation rate but not late

Figure 8. ANKRD26 regulates the internalization of homodimeric type I receptors. (A-C) Ba/F3 cells overexpressing the three receptors (MPL, G-CSFR, and EPOR) and transduced with either an empty vector or ANKRD26 cDNA encoding lentivirus were used for internalization assays. Internalization of MPL was measured with anti-MPL antibody (A), and that of G-CSFR (B) and EPOR (C) with anti-HA antibody. (A) In the absence of ANKRD26 overexpression, almost 50% of cell surface MPL was internalized as soon as 15 min after the addition of thrombopoietin, while in the presence of ANKRD26, MPL was not internalized. Mean fluorescence intensity (MFI) is normalized to that of Ba/F3/HA_MPL cells expressing empty vector. (B) ANKRD26 overexpression inhibited GCSFR internalization at 15 and 30 min after stimulation of starved Ba/F3 cells with 20 ng/mL G-CSF. MFI is normalized to that of Ba/F3/HA_GCSFR cells expressing empty vector. (C) ANKRD26 overexpression significantly inhibited EPOR internalization at 15 and 30 min after starved Ba/F3 cell stimulation with 1 U/mL of EPO. MFI is normalized to that of Ba/F3/HA_EPOR cells expressing empty vector. The averages of three independent experiments are shown as mean ± standard deviation. *P<0.05; unpaired t test with Mann-Whitney correction.

stages of differentiation. Therefore, the severity of the phenotypes observed in THC2 patients is less pronounced and could be dependent on the mutation position that affects, to a variable extent, the binding of lineage-specific transcription factor(s) regulating ANKRD26 expression level in the granulocytic and erythroid lineages.

Leukocytosis and erythrocytosis are often a consequence of mutations that alter the correct signaling of G-CSFR and EPOR. CSF3R mutations are recurrent in chronic neutrophilic leukemia, which is characterized by excessive proliferation of the neutrophil lineage,41-43 while EPOR mutations are a hallmark of familial erythrocytosis.30,44 In both cases, the mutations may lead to the generation of truncated receptors that lack the specific domains that control the negative regulation of the signaling. We hypothesize that the persistent presence of ANKRD26 could act similarly to these mutations, by disturbing receptor internalization and shutdown. Several reports link ANK repeat-containing proteins to the interaction and regulation of signaling pathways and receptor internalization. Ankrd26, the mouse homolog of ANKRD26, is involved in the onset of obesity,45-47 via regulation of signaling pathways,48 and ANKRD26 was also described to interact with different proteins, including hyaluronan-mediated motility receptor (HMMR).49

How exactly ANKRD26 interferes with the regulation of type I cytokine receptor internalization and signaling is not completely clear. For example, ANKRD13A, 13B, and 13D were shown to bind to the ligand-activated EGFR through a ubiquitin-interacting motif (UIM), and direct rapid EGFR internalization, probably by connecting the receptor with the endocytic machinery via their ANK domain.50 However, ANKRD26 does not contain the UIM domain and, according to the recently reported thrombocytopenia caused by WACANKRD26 fusion, with a conserved C- but not N-terminal part containing ANK repeats,34 only the C-terminal domain with a coiled-coil region seems to be necessary for the interactions with MPL, EPOR and G-CSFR. Further studies are necessary to clarify whether these interactions are direct or not and to identify precise mechanisms by which ANKRD26 prevents internalization.

In conclusion, we demonstrate a novel central role for ANKRD26 as responsible for the fine-tuning of the physiology of at least three different receptors. Small changes in their activity induce notable anomalies in proliferation of megakaryocytes, erythrocytes and granulocytes, and differentiation of megakaryocytes which cause thrombocytopenia, erythrocytosis and leukocytosis, respectively.

Disclosures

No conflicts of interest to disclose.

Contributions

FB-V and AD designed and performed experiments, analyzed data, and contributed to drafting the manuscript. VTM, ML, NB, CPMO, HD, BA, DM, J-EM and AC performed experiments and analyzed data. ND provided plasmid constructs, supervised RNA-sequencing and discussed results. IB analyzed induced pluripotent stem cell lines and discussed results. AB, ND, IA-D, CM, IP and WV designed experiments, discussed results and contributed to editing the manuscript. LF, AP and RF provided samples from patients and healthy controls and discussed results. HR designed and supervised the work and wrote the paper. All the authors gave final approval of the manuscript.

Acknowledgments

The authors thank the patients for their participation in this study and Prof. MC Alessi who coordinates the Centre de Référence des Pathologies Plaquettaires (France). The authors also thank P Rameau, C Catelain, and T Manoliu from the Imaging and Cytometry Platform PFIC, UMS AMMICA (Gustave Roussy Villejuif, France) for their expertise in cytometry, MK Diop and T Dayris from the Bioinformatics Platform, UMS AMMICA (Gustave Roussy Villejuif, France) for RNA-sequencing analysis, M Breckler from the Genomic Platform UMS AMMICA (Gustave Roussy Villejuif, France) for RNA-sequencing performed thanks to the TA2016_P27_ALDO, S Peressini from Servizio Di Analisi Chimico-Cliniche, IRCCS Policlinico S. Matteo (Pavia, Italy) for helping with the erythropoietin measurements, and Unité de Thérapie Cellulaire, CRB-Banque de Sang de Cordon, Hôpital Saint-Louis, AP-HP (Paris, France) for cord blood samples.

Funding

FB-V, AD, VTM were supported by the Université Sorbonne Paris Cité/Université Paris Diderot and French Society of Hematology (FBV and VTM) and Ligue National Contre le Cancer (AD). This work was supported by a grant from the Ligue Nationale Contre le Cancer (Equipe Labellisée 2016 and 2019 to HR), and H2020-FETOPEN-1-2016-2017- SilkFusion.

Data-sharing statement

All data generated in this study are included in the article and its Online Supplementary File. The datasets used and analyzed during the current study are also available from the corresponding author on request. References

1. Lee Y, Ise T, Ha D, et al. Evolution and expression of chimeric POTE-actin genes in the human genome. Proc Natl Acad Sci U S

A. 2006;103(47):17885-17890.

2. Liu XF, Bera TK, Liu LJ, Pastan I. A primate-specific POTE-actin

fusion protein plays a role in apoptosis. Apoptosis. 2009;14(10):1237-1244.

3. Djinovic-Carugo K, Gautel M, Ylänne J, Young P. The spectrin repeat: a structural platform for cytoskeletal protein assemblies. FEBS Lett. 2002;513(1):119-123.

4. Mosavi LK, Cammett TJ, Desrosiers DC, Peng ZY. The ankyrin repeat as molecular architecture for protein recognition. Protein Sci. 2004;13(6):1435-1448.

5. Hahn Y, Bera TK, Pastan IH, Lee B. Duplication and extensive remodeling shaped POTE family genes encoding proteins containing ankyrin repeat and coiled coil domains. Gene. 2006;366(2):238-245.

6. Pippucci T, Savoia A, Perrotta S, et al. Mutations in the 5' UTR of ANKRD26, the ankirin repeat domain 26 gene, cause an autosomal-dominant form of inherited thrombocytopenia, THC2. Am J Hum Genet. 2011;88(1):115-120.

7. Noris P, Perrotta S, Seri M, et al. Mutations in ANKRD26 are responsible for a frequent form of inherited thrombocytopenia: analysis of 78 patients from 21 families. Blood. 2011;117(24):6673-6680.

8. Noris P, Favier R, Alessi MC, et al. ANKRD26-related thrombocytopenia and myeloid malignancies. Blood. 2013;122(11):1987-1989.

9. Balduini A, Raslova H, Di Buduo CA, et al. Clinic, pathogenic mechanisms and drug testing of two inherited thrombocytopenias, ANKRD26-related thrombocytopenia and MYH9-related diseases. Eur J Med Genet. 2018;61(11):715-722.

10. Bluteau D, Balduini A, Balayn N, et al. Thrombocytopeniaassociated mutations in the ANKRD26 regulatory region induce MAPK hyperactivation. J Clin Invest. 2014;124(2):580-591.

11. Touw IP, Palande K, Beekman R. Granulocyte colony-stimulating factor receptor signaling: implications for G-CSF responses and leukemic progression in severe congenital neutropenia. Hematol Oncol Clin North Am. 2013;27(1):61-73, viii.

12. Krantz SB. Erythropoietin. Blood. 1991;77(3):419-434.

13. Bluteau D, Lordier L, Di Stefano A, et al. Regulation of megakaryocyte maturation and platelet formation. J Thromb Haemost. 2009;7(Suppl 1):227-234.

14. Yoshihara H, Arai F, Hosokawa K, et al. Thrombopoietin/MPL signaling regulates hematopoietic stem cell quiescence and interaction with the osteoblastic niche. Cell Stem Cell. 2007;1(6):685-697.

15. Dong F, Qiu Y, Yi T, Touw IP, Larner AC. The carboxyl terminus of the granulocyte colony-stimulating factor receptor, truncated in patients with severe congenital neutropenia/acute myeloid leukemia, is required for SH2-containing phosphatase-1 suppression of Stat activation. J Immunol. 2001;167(11):6447-6452.

16. Dwivedi P, Greis KD. Granulocyte colony-stimulating factor receptor signaling in severe congenital neutropenia, chronic neutrophilic leukemia, and related malignancies. Exp Hematol. 2017;46:9-20.

17. de la Chapelle A, Sistonen P, Lehväslaiho H, Ikkala E, Juvonen E. Familial erythrocytosis genetically linked to erythropoietin receptor gene. Lancet. 1993;341(8837):82-84.

18. He X, Chen Z, Jiang Y, Qiu X, Zhao X. Different mutations of the human c-mpl gene indicate distinct haematopoietic diseases. J Hematol Oncol. 2013;6:11.

19. Vainchenker W, Delhommeau F, Constantinescu SN, Bernard OA. New mutations and pathogenesis of myeloproliferative neoplasms. Blood. 2011;118(7):1723-1735.

20. Wilmes S, Hafer M, Vuorio J, et al. Mechanism of homodimeric cytokine receptor activation and dysregulation by oncogenic mutations. Science. 2020;367(6478):643-652.

21. Quelle FW, Sato N, Witthuhn BA, et al. JAK2 associates with the

beta c chain of the receptor for granulocyte-macrophage colony-stimulating factor, and its activation requires the membrane-proximal region. Mol Cell Biol. 1994;14(7):4335-4341.

22. Miyakawa Y, Rojnuckarin P, Habib T, Kaushansky K. Thrombopoietin induces phosphoinositol 3-kinase activation through SHP2, Gab, and insulin receptor substrate proteins in BAF3 cells and primary murine megakaryocytes. J Biol Chem. 2001;276(4):2494-2502.

23. Chaturvedi P, Reddy MV, Reddy EP. Src kinases and not JAKs activate STATs during IL-3 induced myeloid cell proliferation. Oncogene. 1998;16(13):1749-1758.

24. Pratt JC, Weiss M, Sieff CA, Shoelson SE, Burakoff SJ, Ravichandran KS. Evidence for a physical association between the Shc-PTB domain and the beta c chain of the granulocytemacrophage colony-stimulating factor receptor. J Biol Chem. 1996;271(21):12137-12140.

25. Krebs DL, Hilton DJ. SOCS proteins: negative regulators of cytokine signaling. Stem Cells. 2001;19(5):378-387.

26. Ai J, Druhan LJ, Loveland MJ, Avalos BR. G-CSFR ubiquitination critically regulates myeloid cell survival and proliferation. PLoS One. 2008;3(10):e3422.

27. Saur SJ, Sangkhae V, Geddis AE, Kaushansky K, Hitchcock IS. Ubiquitination and degradation of the thrombopoietin receptor c-Mpl. Blood. 2010;115(6):1254-1263.

28. Kim AR, Ulirsch JC, Wilmes S, et al. Functional selectivity in cytokine signaling revealed through a pathogenic EPO mutation. Cell. 2017;168(6):1053-1064.

29. Favale F, Messaoudi K, Varghese LN, et al. An incomplete trafficking defect to the cell-surface leads to paradoxical thrombocytosis for human and murine MPL P106L. Blood. 2016;128(26):3146-3158.

30. Pasquier F, Marty C, Balligand T, et al. New pathogenic mechanisms induced by germline erythropoietin receptor mutations in primary erythrocytosis. Haematologica. 2018;103(4):575-586.

31. Secardin L, Gomez Limia C, da Silva-Benedito S, et al. Induced pluripotent stem cells enable disease modeling and drug screening in calreticulin del52 and ins5 myeloproliferative neoplasms. Hemasphere. 2021;5(7):e593.

32. Saliba J, Hamidi S, Lenglet G, et al. Heterozygous and homozygous JAK2(V617F) states modeled by induced pluripotent stem cells from myeloproliferative neoplasm patients. PLoS One. 2013;8(9):e74257.

33. Antony-Debré I, Manchev VT, Balayn N, et al. Level of RUNX1 activity is critical for leukemic predisposition but not for thrombocytopenia. Blood. 2015;125(6):930-940.

34. Wahlster L, Verboon JM, Ludwig LS, et al. Familial thrombocytopenia due to a complex structural variant resulting in a WAC-ANKRD26 fusion transcript. J Exp Med. 2021;218(6):e20210444.

35. Lachmann N, Ackermann M, Frenzel E, et al. Large-scale hematopoietic differentiation of human induced pluripotent stem cells provides granulocytes or macrophages for cell replacement therapies. Stem Cell Rep. 2015;4(2):282-296.

36. Liu C, Zhai X, Zhao B, Wang Y, Xu Z. Cyclin I-like (CCNI2) is a cyclin-dependent kinase 5 (CDK5) activator and is involved in cell cycle regulation. Sci Rep. 2017;7:40979.

37. Lowman XH, McDonnell MA, Kosloske A, et al. The proapoptotic function of Noxa in human leukemia cells is regulated by the kinase Cdk5 and by glucose. Mol Cell. 2010;40(5):823-833.

38. Erickson-Miller CL, Delorme E, Tian SS, et al. Preclinical activity of eltrombopag (SB-497115), an oral, nonpeptide thrombopoietin receptor agonist. Stem Cells. 2009;27(2):424-430.

39. Leroy E, Defour JP, Sato T, et al. His499 regulates dimerization and prevents oncogenic activation by asparagine mutations of the human thrombopoietin receptor. J Biol Chem. 2016;291(6):2974-2987.

40. Pecquet C, Diaconu CC, Staerk J, et al. Thrombopoietin receptor down-modulation by JAK2 V617F: restoration of receptor levels by inhibitors of pathologic JAK2 signaling and of proteasomes. Blood. 2012;119(20):4625-4635.

41. Pardanani A, Lasho TL, Laborde RR, et al. CSF3R T618I is a highly prevalent and specific mutation in chronic neutrophilic leukemia. Leukemia. 2013;27(9):1870-1873.

42. Maxson JE, Gotlib J, Pollyea DA, et al. Oncogenic CSF3R mutations in chronic neutrophilic leukemia and atypical CML. N Engl J Med. 2013;368(19):1781-1790.

43. Plo I, Zhang Y, Le Couédic JP, et al. An activating mutation in the CSF3R gene induces a hereditary chronic neutrophilia. J Exp Med. 2009;206(8):1701-1707.

44. Hong WJ, Gotlib J. Hereditary erythrocytosis, thrombocytosis and neutrophilia. Best Pract Res Clin Haematol. 2014;27(2):95-106.

45. Bera TK, Liu XF, Yamada M, et al. A model for obesity and gigantism due to disruption of the Ankrd26 gene. Proc Natl Acad Sci U S A. 2008;105(1):270-275.

46. Raciti GA, Bera TK, Gavrilova O, Pastan I. Partial inactivation of Ankrd26 causes diabetes with enhanced insulin responsiveness of adipose tissue in mice. Diabetologia. 2011;54(11):2911-2922.

47. Acs P, Bauer PO, Mayer B, et al. A novel form of ciliopathy underlies hyperphagia and obesity in Ankrd26 knockout mice. Brain Struct Funct. 2015;220(3):1511-1528.

48. Fei Z, Bera TK, Liu X, Xiang L, Pastan I. Ankrd26 gene disruption enhances adipogenesis of mouse embryonic fibroblasts. J Biol Chem. 2011;286(31):27761-27768.

49. Liu XF, Bera TK, Kahue C, et al. ANKRD26 and its interacting partners TRIO, GPS2, HMMR and DIPA regulate adipogenesis in 3T3-L1 cells. PLoS One. 2012;7(5):e38130.

50. Tanno H, Yamaguchi T, Goto E, Ishido S, Komada M. The Ankrd 13 family of UIM-bearing proteins regulates EGF receptor endocytosis from the plasma membrane. Mol Biol Cell. 2012;23(7):1343-1353.

Tislelizumab with gemcitabine and oxaliplatin in patients with relapsed or refractory classic Hodgkin lymphoma: a multicenter phase II trial

Kaiyang Ding,1* Hailing Liu,2* Jie Ma,3* Haiyan Yang,4 Lei Cao,2 Huihan Wang,5 Hongling Peng,6 Wei Shi,7 Xiaoli Zhao,2 Wei Wu,2 Huayuan Zhu,2 Jianyong Li2 and Lei Fan2

1Department of Hematology, Anhui Provincial Hospital, the First Affiliated Hospital of USTC, Hefei, Anhui; 2Department of Hematology, Jiangsu Province Hospital and Nanjing Medical University First Affiliated Hospital, Jiangsu Key Lab of Cancer Biomarkers, Prevention and Treatment, Collaborative Innovation Center for Personalized Cancer Medicine, Nanjing Medical University, Nanjing, Jiangsu; 3Department of Hematology, the First Affiliated Hospital of Zhengzhou University, Zhengzhou, Henan; 4Department of Lymphoma, Zhejiang Cancer Hospital, Cancer Hospital of the University of Chinese Academy of Sciences, Hangzhou, Zhejiang; 5Department of Hematology, Shengjing Hospital, China Medical University, Shenyang, Liaoning; 6Department of Hematology, the Second Xiangya Hospital, Central South University, Changsha, Hunan and 7Department of Hematology, the Friendship Hospital of Ili Kazakh Autonomous Prefecture, Yining, Xinjiang, China.

*KD, HL, and JM contributed equally as first authors.

Abstract

Correspondence:

L. Fan fanlei3014@126.com

J. Li lijianyonglm@126.com

Received: October 13, 2022.

Accepted: January 19, 2023.

Early view: January 26, 2023.

https://doi.org/10.3324/haematol.2022.282266

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Although classic Hodgkin lymphoma (cHL) is highly curable with current treatment paradigms, therapy fails in 10-25% of patients. This prospective multicenter phase II study attempted to investigate the efficacy and safety of the combination of tislelizumab with gemcitabine and oxaliplatin (T-GemOx) in relapsed or refractory cHL. Participants received six to eight courses of gemcitabine (1 g/m2 on day 1) and oxaliplatin (100 mg/m2 on day 1) combined with tislelizumab (200 mg on day 2) at 21-day intervals, followed by tislelizumab maintenance (every 2 months for 2 years). The main outcome measure was the best complete remission rate. As of August 2022, a total of 30 patients had been consecutively enrolled and given induction therapy. The best overall response rate and complete remission rate were 100% (95% confidence interval [CI]: 88.4-100%) and 96.7% (95% CI: 82.8-99.9%), respectively. The median duration of follow-up after initiation of T-GemOx was 15.8 months. The 12-month progression-free survival rate without autologous stem cell transplant was 96% (95% CI: 74.8-99.4%). There were 122 adverse events recorded, of which 93.4% were grade 1 or 2. Thrombocytopenia (10%) and anemia (6.7%) were the most common grade 3 or 4 adverse events. Overall, T-GemOx demonstrated promising antitumor activity with manageable toxicities as a salvage treatment for relapsed or refractory cHL. A longer follow-up duration is required to determine whether maintenance therapy with tislelizumab rather than transplantation can be curative following such a highly active regimen. This trial was registered with the Chinese Clinical Trials Registry (http://www.chictr.org.cn) on June 1, 2020, identifier ChiCTR2000033441.

Introduction

Classic Hodgkin lymphoma (cHL) is one of the most common lymphoid neoplasms in adolescents and young adults, affecting 2-3 people per 100,000 per year.1 Multiagent chemotherapy with or without radiation is frequently used as the first-line treatment for cHL, with 5-year survival rates of approximately 90% and 80% for early- and late-stage patients, respectively.2-4 However, despite the high cure rate of cHL, approximately 10–25% of patients will have refractory disease or relapse after achieving remission.1 Designing effective regimens for re-

lapsed or refractory cHL (R/R cHL) is essential yet challenging.

Immune dysfunction is increasingly recognized as a crucial factor underlying the R/R status in patients with cHL.

As a result of alterations in chromosome 9p24.1, Hodgkin Reed–Sternberg cells express high levels of programmed death ligand-1 (PD-L1) and PD-L2, which engage programmed death-1 (PD-1)-positive T cells and result in Tcell exhaustion, thereby enabling tumor cells to evade immune surveillance.5-7 The anti-PD-1 antibodies nivolumab and pembrolizumab have been reported to produce a striking overall response rate (ORR) in patients with R/R

cHL.8,9 However, the complete remission rate (CRR) was less than 30%, and most patients experienced recurrence or progression within 18 months.8-9 Tislelizumab, a new, fully humanized immunoglobulin G4 monoclonal anti-PD1 antibody, was designed to minimize binding to the Fcγ receptors on macrophages to avoid antibody-dependent cellular phagocytosis, a mechanism of T-cell clearance and potential resistance to anti-PD-1 therapy.10,11 Given its specific binding to the PD-1 CC'-loop and modification of the Fc fragment, tislelizumab has a higher affinity and slower dissociation than other anti-PD-1 antibodies.12-14 Although lacking head-to-head comparisons, the reported CRR of tislelizumab monotherapy is numerically higher (62.9%) than that of other PD-1 antibodies.8,9,15-17

Considering the promising outcomes of tislelizumab monotherapy, we designed a tislelizumab-based combinatorial regimen to maximize the response rates of patients with R/R cHL. Although the CRR of gemcitabine combined with oxaliplatin (GemOx) alone was only 37.5%, the two drugs can synergize with checkpoint inhibitors to enhance the immunogenic death of tumor cells and exhibit direct cytotoxic effects.18-22 Additionally, the GemOx regimen frequently presented no cross-resistance to firstline drugs and no dose-dependent toxicities. Therefore, we conducted an investigator-initiated, open-label, prospective, single-arm, phase II study to investigate the efficacy and safety of a combination of tislelizumab with GemOx (T-GemOx) in patients with R/R cHL.

Methods

Patients

This study, which started in August 2020, was conducted at seven medical centers in China. It was approved by the Chinese Ethics Committee for Registering Clinical Trials (ChiECRCT20200186) (Online Supplementary File S1). The inclusion criteria were as follows: (i) male and female patients diagnosed with cHL according to the criteria of the World Health Organization classification; (ii) at least one prior treatment regimen; (iii) biopsy-proven recurrence or progression; (iv) Eastern Cooperative Oncology Group performance status of 0-2; (v) at least one measurable lesion; (vi) a life expectancy of at least 3 months; and (vii) adequate organ function. There were no age restrictions. Online Supplementary Table S1 presents the exclusion criteria for participant selection. The trial was registered on the Chinese Clinical Trials Registry Platform (ChiCTR2000033441) and followed the principles of the Declaration of Helsinki. All participants or parental guardians provided written informed consent to the study and its publication.

Procedures

The study was divided into two phases, i.e., the induction and maintenance phases. During the induction phase, all subjects received six to eight courses of gemcitabine 1 g/m2 intravenously (IV) (day 1), oxaliplatin 100 mg/m2 IV (day 1), and tislelizumab 200 mg IV (day 2) at 21-day intervals. The number of induction cycles was primarily influenced by the depth of remission in the first four courses. For complete responders, the planned induction therapy was six courses. The investigator had the option of adding up to two more courses. For partial responders, eight courses of T-GemOx were recommended. Following the completion of induction therapy, responders (patients who achieved a complete or partial remission) were continued into the maintenance phase (tislelizumab 200 mg IV at 2-month intervals) until disease progression, unacceptable toxicity, the patients’ withdrawal, or the 2-year period of maintenance therapy had been completed. Concomitant therapies were administered for complications at the discretion of the treating physicians.

Study assessments

Tumor responses were categorized as positron emission tomography (PET) complete remission, partial remission, stable disease, and disease progression assessed by local investigators, according to the modi fi ed Lugano 2014 criteria. All patients underwent baseline PET scans before beginning the drug study, interim scans three to four courses into therapy, and restaging PET scans following completion of induction. During the maintenance phase, assessments were performed every 3 months in the first year and then every 6 months until 5 years of therapy. In the case of lack of effi cacy or the patient withdrawing consent, an evaluation of efficacy outcomes was done in advance. The primary objective of this study was to identify the best CRR, defined as the percentage of patients who responded to treatment with best response being a complete remission. Secondary endpoints included the best ORR (complete plus partial remission), progression-free survival (time from study entry to disease progression or death), and safety profile. Safety was assessed by the frequency of adverse events (AE), graded per Common Terminology Criteria for Adverse Events, version 4.0.

Statistical methods

Sample size calculations were performed using PASS software (version 15.0.5). The study primarily aimed to evaluate the optimal CRR during T-GemOx treatment. The best CRR of tislelizumab monotherapy was 63%, which is higher than that of salvage combination chemotherapy (approximately 50%). Based on this, we assumed that the best CRR of T-GemOx, which was at least 88%, would be considered promising. Assuming a power of 80%, an α

value of 0.025 (one-sided), and an attrition rate of 20%, at least 30 patients would need to be enrolled in the study. Response rates were calculated using the Clopper-Pearson method and presented in proportions with the corresponding 95% confidence interval (95% CI). Survival was analyzed using a Kaplan-Meier plot. Quantitative variables were summarized as median and range whereas qualitative variables were described as counts and percentages. All P values were two-sided, and P<0.05 was considered statistically significant. Results were processed using R statistical software (version 4.1.0).

Results

Patients’ characteristics

A total of 30 patients of the Chinese Han population were enrolled as of August 1, 2022. Figure 1 illustrates the patients’ recruitment. All 30 patients completed induction

therapy, and 26 patients received at least one maintenance dose of tislelizumab. Table 1 summarizes the patients’ characteristics at study entry. The male:female ratio was 0.67. The median age was 33.5 years (range, 1373 years), and no significant difference was observed between male and female patients (P=0.233). The predominant histological subtype was the nodular sclerosis type (70%). Advanced-stage disease was observed in 24 (80%) patients, three of whom exhibited bulky disease, with a mediastinal mass in two patients and retroperitoneal mass in one patient. All patients at the early stage belonged to an unfavorable-risk group (n=6, 20%) according to the German Hodgkin Study Group criteria. In patients with advanced-stage disease, seven (23.3%), seven (23.3%), and ten (33.3%) were in the low-intermediate, high-intermediate, and high-risk categories, respectively. Prior therapies are summarized in Table 2. All patients received doxorubicin, bleomycin, vinblastine, and dacarbazine as frontline therapy, of whom, ten (33.3%) attained

Figure 1. Flow diagram illustrating the enrollment of patients.

remissions before relapsing after a median time of response of 11.4 months. The remaining 20 patients (66.7%) exhibited primary refractory disease (no complete response to frontline therapy). During the subsequent treatment, five (16.7%) patients underwent autologous stem cell transplant (ASCT), two (6.7%) received brentuximab vedotin, four (13.3%) were treated with anti-PD-1 antibodies, and two (6.7%) received PD-L1 inhibitors. The ORR to the previous PD-1/PD-L1 inhibitors was 33.3% in our study; however, there were no complete responses. Ap-

proximately 30% of patients received three or more lines of prior therapies, and two-thirds were refractory to the most recent therapy.

Responses

The cutoff date for analysis was August 1, 2022. The duration and depth of responses are presented in a swimmer’s plot (Figure 2A). The median number of induction treatment cycles completed was eight (range, 6-8), with 43.3% (n=13) of patients receiving fewer than eight courses. During the treatment course, the confirmed best ORR and CRR were 100% (95% CI: 88.4-100%) and 96.7% (95% CI: 82.8-99.9%), respectively. At the end of induction, the ORR was 96.7% (95% CI: 82.8-99.9%), with 93.3% of patients achieving complete remission (Figure 2B). Figure 3 depicts responses after four cycles of T-GemOx in each prespecified subgroup. Significance tests were not conducted because of the small sample sizes of the subgroups.

*Bulky disease was defined as any single node/nodal mass ≥10 cm in diameter or a mediastinal mass ratio of 0.33. cHL: classic Hodgkin lymphoma; GHSG: German Hodgkin Lymphoma Study Group; IPS: International Prognostic Score; ECOG PS: Eastern Cooperative Oncology Group performance status.

*The denominator is the number of patients who received prior PD-1 and PD-L1 antibodies. ABVD: adriamycin, bleomycin, vinblastine, dacarbazine; BEACOPP: bleomycin, etoposide, adriamycin, cyclophosphamide, oncovin, procarbazine, prednisone; GDP: gemcitabine, dexamethasone, cisplatin; ICE: ifosfamide, carboplatin, etoposide; IGEV: ifosfamide, gemcitabine, etoposide; GVD: gemcitabine, vinorelbine, doxorubicin; ASCT: autologous stem cell transplantation; PD-1: programmed death-1; PD-L1: programmed death ligand-1.

Table 1. Patients’ characteristics at study entry.

Table 2. Prior treatment characteristics.

Two patients in the cohort exhibited disease progression following two and three cycles of tislelizumab maintenance, with remission durations of 6.8 and 10.4 months, respectively. Two other patients did not achieve complete remissions following induction therapy. These four patients had advanced-stage disease and belonged to the high-risk group. Although all of them had received one prior treatment regimen, three had primary refractory disease, and one had an early recurrence within a year.

Survival

The median duration of follow-up after initiation of TGemOx was 15.8 months (range, 6.3-23.8 months). The

Figure 2. Efficacy following treatment with tislelizumab plus gemcitabine and oxaliplatin in patients with relapsed or refractory classic Hodgkin lymphoma.

(A) Swimmer plot depicting individual patients as lines. (B) Therapeutic efficacy. NSCHL: nodular sclerosing classic Hodgkin lymphoma; LRCHL: lymphocyte-rich classic Hodgkin lymphoma; MCCHL: mixed cellularity classic Hodgkin lymphoma; LDCHL: lymphocyte-depleted classic Hodgkin lymphoma; ORR: overall response rate; CR: complete remission; PR: partial remission; SD: stable disease; PD: disease progression.

median progression-free survival was not reached, and the 12-month progression-free survival rate was 96% (95% CI: 74.8-99.4%) (Figure 4). Among all responders, 80% and 56.7% had response durations of ≥6 months and ≥12 months, respectively. All patients survived until the last follow-up; however, overall survival data were not mature at the time of data cutoff.

Tolerability and safety

A total of 122 AE were recorded (Table 3), of which 114 (93.4%) were grade 1 or grade 2. Anemia (43.3%), pyrexia (40.0%), and fatigue (33.3%) were the most common grade 1 or 2 AE. Thrombocytopenia (10%) and anemia (6.7%) were

Figure 3. Responses after four cycles of tislelizumab with gemcitabine and oxaliplatin in each prespecified subgroup. One patient was excluded from this analysis as it was not possible to assess tumor response following the fourth cycle of tislelizumab plus gemcitabine and oxaliplatin. PR: partial remission; CR: complete remission; CRR: complete remission rate; 95% CI: 95% confidence interval; cHL: classic Hodgkin lymphoma; GHSG: German Hodgkin Lymphoma Study Group; IPS: International Prognostic Score; ECOG PS: Eastern Cooperative Oncology Group performance status; PD-1: programmed death-1; PD-L1: programmed death ligand-1.

the most common grade 3 or 4 AE. Grades 1-2 AE were generally tolerated whereas grades 3-4 AE were resolved with supportive care. No discontinuation caused by death or an AE was recorded.

As indicated in Online Supplementary Table S2, common immune-related AE included thyroid dysfunction (hypothyroidism, 30%; hyperthyroidism, 6.7%), rash (13.3%), and elevated transaminases (10%). Most immune-related AE were grade 1 or 2; only one was grade 3 in severity (elevated transaminase), which was treated with systemic steroids (prednisone), thus, contributing to treatment delays. Additionally, one patient experienced grade 1 cardiac toxicity (ventricular arrhythmias of unknown origin), which was considered immune-related. No endomyocardial biopsy or intervention was performed because of the pa-

Figure 4. Kaplan-Meier curve for progression-free survival. GemOx: gemcitabine and oxaliplatin.

tient’s continued asymptomatic presentation. Three patients had grade 2 hypothyroidism, and all remained stable on levothyroxine.

Discussion

This trial demonstrated that additional GemOx before tislelizumab further improved response rates, with the best ORR and CRR being improved to 100% and 96.7%, respectively. Theoretically, the sequence of drug administration (i.e., administering tislelizumab after GemOx) is essential for enhancing the impact of tislelizumab. It has been reported that increased circulating myeloid-derived suppressor cells in cHL were associated with poor efficacy, early progression, and resistance to checkpoint inhibitors.19,20 Preclinical data have demonstrated that gemcitabine reduces the number of myeloid-derived suppressor cells, favoring the reprogramming of tumor-associated macrophages toward an immunostimulatory phenotype.21,23 It can also stimulate histocompatibility complex-1 expression on tumor cells to increase their antigenicity.24 Oxaliplatin promotes the activity of neutrophils and macrophages and the depletion of myeloid-derived suppressor cells.22,25 Thus, the application of

GemOx before tislelizumab may potentially contribute to improve tumor responses and reverse resistance. In practice, we demonstrated that patients who previously responded poorly to anti-PD-1 or anti-PD-L1 therapy achieved deep remissions after the administration of T-GemOx, suggesting that such a regimen may restore sensitivity to checkpoint inhibitors.

Similar to our work, other researchers have assessed combination regimens containing a PD-1 inhibitor to obtain better clinical efficacy.26-30 A phase II PET-adapted study compared the efficacy of nivolumab monotherapy and nivolumab in combination with ifosfamide, carboplatin, and etoposide (NICE) and observed notably higher response rates in the NICE group than in the single-agent group (ORR 93% vs. 81%; CRR 91% vs. 71%).26 Another phase II trial found that the CRR in patients naïve to PD-1 blockade was significantly higher in those treated with low-dose decitabine plus camrelizumab than in those treated with camrelizumab alone (71% vs. 32%).30 Additionally, the CRR with gemcitabine, vinorelbine, and liposomal doxorubicin (GVD) alone was approximately 50% whereas that with pembrolizumab plus GVD was 95%.28 Our study demonstrated similar response rates, implying that the benefit of PD-1 blockade is greater when it is combined with other therapies.

Table 3. Treatment-related adverse events.

From a safety standpoint, immunosuppressants such as cyclophosphamide and ifosfamide are not the best options as they may increase the risk of infections. Previously used medications, particularly anthracyclines and vincristine, may contribute to resistance and dose-dependent toxicities. Considering the aforementioned, we combined the GemOx regimen in this study. One of the expected challenges in developing combination therapies was increased toxicities, which might be resolved by reducing doses and/or increasing treatment intervals. GemOx alone was typically administered every 2 weeks, and tislelizumab monotherapy was typically administered every 3 weeks. To balance clinical benefit with toxicity, we did not change the dose of GemOx and extended its treatment interval appropriately to a 3week course to synchronize the combination regimen. In all, the data suggested that T-GemOx was associated with a manageable safety profile.

Recent data suggest that treatment with PD-1 inhibitors may sensitize patients with chemorefractory cHL to subsequent high-dose chemotherapy and ASCT.31 ASCT after PD-1blockade has produced considerably favorable outcomes in multiple trials, with 2-year progression-free survival rates of 72-88%.26,29 Nevertheless, these efficacy benefits must be considered in the context of relative safety profiles. One conference abstract from the 2022 European Hematology Association reported that the incidence of engraftment syndrome of ASCT following anti-PD-1 treatment was high (18.6%), which can cause fulminant immune-related AE (myocarditis and pneumonitis).32 It must be noted that these findings were preliminary, and further research is required to confirm the safety of ASCT following anti-PD-1 treatment in the context of possible immune-related AE.

Previous data on anti-PD-1 monotherapy suggested that a subgroup of patients achieving an excellent response to PD1 blockade remain disease-free for >3 years even after discontinuation of anti-PD-1 treatment and thus may be cured.33 The therapeutic potential of anti-PD-1 combination regimens is under active study. Considering the high efficacy of our treatment combination, a challenge faced in this study was whether all patients required ASCT consolidation. In other words, can we now provide a path toward reducing the need for ASCT in the relapsed/refractory setting with secondary complete response status? The half-life of tislelizumab was 26 days following repeat administration in population pharmacokinetic analyses.14 Thus, we are investigating a brief maintenance treatment of every 2 months for 2 years instead of transplantation. Our observed 12month progression-free survival rate was 96%, and further follow-up is required to assess long-term outcomes. Our study has certain limitations. First, all our subjects were Chinese, which may limit generalizability to other racial/ethnic groups. Second, the proportion of patients treated with brentuximab vedotin and ASCT in our study was relatively low compared to that in western countries,

which can be attributed to country-specific differences in treatment landscapes. Third, a relatively short follow-up period and the small number of events limited the generalization of the findings. Last, the heterogeneity of the number of cycles received by patients was another potential limitation, adding to the challenge of establishing best practices.

In conclusion, our study illustrated that T-GemOx is a highly efficacious, less toxic, and cost-effective therapy in R/R cHL. This regimen can be completely implemented in the outpatient setting in the future because of the simple dosing strategy and favorable safety profile, thereby shortening hospital stays. It should be emphasized that it is premature to assess the durability of responses with tislelizumab maintenance. Longer follow-up and prospective controlled studies are required to investigate whether this transplantfree strategy can replace traditional consolidation with ASCT and whether T-GemOx can be used as a bridging therapy for patients who still need transplantation.

Disclosures

No conflicts of interest to disclose.

Contributions

LF and JL conceptualized and designed the study, supervised the data analysis, and reviewed and revised the manuscript critically. KD, HL, and JM acquired data, conducted statistical analyses, and drafted the paper. HY, LC, HW, HP, WS, XZ, WW, and HZ acquired data, helped to analyze and interpret the data, and reviewed and revised the manuscript. All authors approved the submitted and final versions.

Acknowledgments

The authors would like to thank the patients and families who participated in this clinical trial.

Funding

This investigation was supported by grant 81720108002 to LF from the National Natural Science Foundation of China, China; grant Y-Roche2019/2-0090 to LF from the Beijing Xisike Clinical Oncology Research Foundation; grant yl2021ms04 to LF from the Science Foundation Project of Ili & Jiangsu Joint Institute of Health; grant 2018ZX09734-007 to JL from the National Science and Technology Major Project, China; grant 2020ZKZB01 to JL from the Translational Research Grant of NCRCH, China; grant 82100211 to LC from the National Natural Science Foundation of China, China; and grant KJ2021-4 to LC from the Science and Technology Development Fund Project of Pukou branch of Jiangsu People's Hospital, China.

Data-sharing statement

The data generated in this study are available upon request from the corresponding author.

References

1. Brice P, de Kerviler E, Friedberg JW. Classical Hodgkin lymphoma. Lancet. 2021;398(10310):1518-1527.

2. Shanbhag S, Ambinder RF. Hodgkin lymphoma: a review and update on recent progress. CA Cancer J Clin. 2018;68(2):116-132.

3. Engert A, Raemaekers J. Treatment of early-stage Hodgkin lymphoma. Semin Hematol. 2016;53(3):165-170.

4. Vassilakopoulos TP, Johnson PW. Treatment of advanced-stage Hodgkin lymphoma. Semin Hematol. 2016;53(3):171-179.

5. Roemer MG, Advani RH, Ligon AH, et al. PD-L1 and PD-L2 genetic alterations define classical Hodgkin lymphoma and predict outcome. J Clin Oncol. 2016;34(23):2690-2697.

6. Green MR, Monti S, Rodig SJ, et al. Integrative analysis reveals selective 9p24.1 amplification, increased PD-1 ligand expression, and further induction via JAK2 in nodular sclerosing Hodgkin lymphoma and primary mediastinal large B-cell lymphoma. Blood. 2010;116(17):3268-3277.

7. Carey CD, Gusenleitner D, Lipschitz M, et al. Topological analysis reveals a PD-L1-associated microenvironmental niche for ReedSternberg cells in Hodgkin lymphoma. Blood. 2017;130(22):2420-2430.

8. Armand P, Engert A, Younes A, et al. Nivolumab for relapsed/refractory classic Hodgkin lymphoma after failure of autologous hematopoietic cell transplantation: extended follow-up of the multicohort single-arm phase II CheckMate 205 trial. J Clin Oncol. 2018;36(14):1428-1439.

9. Chen R, Zinzani PL, Lee HJ, et al. Pembrolizumab in relapsed or refractory Hodgkin lymphoma: 2-year follow-up of KEYNOTE087. Blood. 2019;134(14):1144-1153.

10. Zhang T, Song X, Xu L, et al. The binding of an anti-PD-1 antibody to FcγRΙ has a profound impact on its biological functions. Cancer Immunol Immunother. 2018;67(7):1079-1090.

11. Dahan R, Sega E, Engelhardt J, Selby M, Korman AJ, Ravetch JV. FcγRs modulate the anti-tumor activity of antibodies targeting the PD-1/PD-L1 axis. Cancer Cell. 2015;28(4):543.

12. Hong Y, Feng Y, Sun H, et al. Tislelizumab uniquely binds to the CC' loop of PD-1 with slow-dissociated rate and complete PD-L1 blockage. FEBS Open Bio. 2021;11(3):782-792.

13. Lee SH, Lee HT, Lim H, Kim Y, Park UB, Heo Y-S. Crystal structure of PD-1 in complex with an antibody-drug tislelizumab used in tumor immune checkpoint therapy. Biochem Biophys Res Commun. 2020;527(1):226-231.

14. Lee A, Keam SJ. Tislelizumab: first approval. Drugs. 2020;80(6):617-624.

15. Song Y, Gao Q, Zhang H, et al. Treatment of relapsed or refractory classical Hodgkin lymphoma with the anti-PD-1, tislelizumab: results of a phase 2, single-arm, multicenter study. Leukemia. 2020;34(2):533-542.

16. Shi Y, Su H, Song Y, et al. Safety and activity of sintilimab in patients with relapsed or refractory classical Hodgkin lymphoma (ORIENT-1): a multicentre, single-arm, phase 2 trial. Lancet Haematol. 2019;6(1):e12-e19.

17. Song Y, Wu J, Chen X, et al. A single-arm, multicenter, phase II study of camrelizumab in relapsed or refractory classical Hodgkin lymphoma. Clin Cancer Res. 2019;25(24):7363-7369.

18. Gutierrez A, Rodriguez J, Martinez-Serra J, et al. Gemcitabine and oxaliplatinum: an effective regimen in patients with refractory and relapsing Hodgkin lymphoma. Onco Targets Ther. 2014;7:2093-2100.

19. Romano A, Parrinello NL, Vetro C, et al. Circulating myeloid-

derived suppressor cells correlate with clinical outcome in Hodgkin lymphoma patients treated up-front with a riskadapted strategy. Br J Haematol. 2015;168(5):689-700.

20. Law AMK, Valdes-Mora F, Gallego-Ortega D. Myeloid-derived suppressor cells as a therapeutic target for cancer. Cells. 2020;9(3):561.

21. Suzuki E, Kapoor V, Jassar AS, Kaiser LR, Albelda SM. Gemcitabine selectively eliminates splenic Gr-1+/CD11b+ myeloid suppressor cells in tumor-bearing animals and enhances antitumor immune activity. Clin Cancer Res. 2005;11(18):6713-6721.

22. Gonzalez-Aparicio M, Alzuguren P, Mauleon I, et al. Oxaliplatin in combination with liver-specific expression of interleukin 12 reduces the immunosuppressive microenvironment of tumours and eradicates metastatic colorectal cancer in mice. Gut. 2011;60(3):341-349.

23. Nowak AK, Lake RA, Marzo AL, et al. Induction of tumor cell apoptosis in vivo increases tumor antigen cross-presentation, cross-priming rather than cross-tolerizing host tumor-specific CD8 T cells. J Immunol. 2003;170(10):4905-4913.

24. Liu WM, Fowler DW, Smith P, Dalgleish AG. Pre-treatment with chemotherapy can enhance the antigenicity and immunogenicity of tumours by promoting adaptive immune responses. Br J Cancer. 2010;102(1):115-123.

25. Tesniere A, Schlemmer F, Boige V, et al. Immunogenic death of colon cancer cells treated with oxaliplatin. Oncogene. 2010;29(4):482-491.

26. Mei MG, Lee HJ, Palmer JM, et al. Response-adapted anti-PD-1based salvage therapy for Hodgkin lymphoma with nivolumab alone or in combination with ICE. Blood. 2022;139(25):3605-3616.

27. Advani RH, Moskowitz AJ, Bartlett NL, et al. Brentuximab vedotin in combination with nivolumab in relapsed or refractory Hodgkin lymphoma: 3-year study results. Blood. 2021;138(6):427-438.

28. Moskowitz AJ, Shah G, Schöder H, et al. Phase II trial of pembrolizumab plus gemcitabine, vinorelbine, and liposomal doxorubicin as second-line therapy for relapsed or refractory classical Hodgkin lymphoma. J Clin Oncol. 2021;39(28):3109-3117.

29. Bryan LJ, Casulo C, Allen P, et al. Pembrolizumab (PEM) added to ICE chemotherapy results in high complete metabolic response rates in relapsed/refractory classic Hodgkin lymphoma (cHL): a multi-institutional phase II trial. Blood. 2021;138(Suppl 1):229.

30. Nie J, Wang C, Liu Y, et al. Addition of low-dose decitabine to anti-PD-1 antibody camrelizumab in relapsed/refractory classical Hodgkin lymphoma. J Clin Oncol. 2019;37(17):1479-1489.

31. Merryman RW, Redd RA, Nishihori T, et al. Autologous stem cell transplantation after anti-PD-1 therapy for multiply relapsed or refractory Hodgkin lymphoma. Blood Adv. 2021;5(6):1648-1659.

32. Mochkin N, Sarzhevskiy V, Protopopova Y, et al. Feasibility of ASCT after anti-PD-1 therapy for R/R classical Hodgkin lymphoma. Hemasphere. 2022;6(S3):981-982.

33. Manson G, Brice P, Herbaux C, et al. Risk of relapse after antiPD1 discontinuation in patients with Hodgkin lymphoma. Eur J Nucl Med Mol Imaging. 2021;48(4):1144-1153.

PARP1 and POLD2 as prognostic biomarkers for multiple myeloma in autologous stem cell transplant

Melissa Thomas,1 Junan Li,2,3 Kevan King,1 Avinash K. Persaud,2 Ernest Duah,2 Zachary Vangundy,2 Craig C. Hofmeister,4 Jatinder K. Lamba,5 Aik Choon Tan,6 Brooke L. Fridley,6 Ming J. Poi2,3 and Nathan D. Seligson1,3,7

1Department of Pharmacotherapy and Translational Research, The University of Florida, Jacksonville, FL; 2Division of Pharmacy Practice and Science, College of Pharmacy, The Ohio State University, Columbus, OH; 3Comprehensive Cancer Center, The Ohio State University, Columbus, OH; 4Department of Hematology and Medical Oncology, Winship Cancer Institute of Emory University, Atlanta, GA; 5Department of Pharmacotherapy and Translational Research, The University of Florida, Gainesville, FL; 6Department of Biostatistics and Bioinformatics, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL and 7Center for Pharmacogenomics and Translational Research, Nemours Children’s Health, Jacksonville, FL, USA

Abstract

Correspondence: N.D. Seligson

nseligson@cop.ufl.edu

Received: November 9, 2022.

Accepted: February 23, 2023.

Early view: March 2, 2023.

https://doi.org/10.3324/haematol.2022.282399

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Multiple Myeloma (MM) is an incurable plasma cell malignancy often treated by autologous stem cell transplant (ASCT). Clinical response to ASCT has been associated with DNA repair efficiency. Here we interrogated the role of the base excision DNA repair (BER) pathway in MM response to ASCT. Across 450 clinical samples and six disease stages, expression levels of genes in the BER pathway were found to be highly upregulated during the development of MM. In a separate cohort of 559 patients with MM treated with ASCT, expression of BER pathway members MPG and PARP3 was positively associated with overall survival (OS) while expression of PARP1, POLD1, and POLD2 was negatively associated with OS. In a validation cohort of 356 patients with MM treated with ASCT, PARP1 and POLD2 findings were replicated. In patients with MM who never received ASCT (n=319), PARP1 and POLD2 were not associated with OS, suggesting that the prognostic effect of these genes may be treatment-dependent. In preclinical models of MM, synergy was observed in anti-tumor activity when poly (ADPribose) polymerase (PARP) inhibitors (olaparib, talazoparib) were used in combination with melphalan. The negative prognosis associated with PARP1 and POLD2 expression along with the apparent melphalan-sensitizing effect of PARP inhibition may suggest this pathway as a potential biomarker in patients with MM in the setting of ASCT. Further understanding of the role of the BER pathway in MM is vital to improve therapeutic strategies related to ASCT.

Introduction

Multiple Myeloma (MM) is an incurable plasma cell malignancy that accounts for 18% of all hematologic malignancies.1 Despite significant therapeutic advances for MM, autologous stem cell transplant (ASCT) following high-dose, single-agent melphalan conditioning remains a cornerstone of therapy.1 Although ASCT provides significant benefits to some patients with MM, it is not curative and is associated with significant heterogeneity in clinical benefit.2 The primary driver of response to ASCT is the depth of anti-myeloma activity of melphalan.3 Patients achieving deeper remission receive the greatest duration of clinical benefit following ASCT.4 While studies have historically noted an increased overall survival (OS) in patients with MM who receive ASCT, the recent DETERMINATION trial demonstrates only progression-free survival (PFS) benefit in patients with MM randomized to receive triplet therapy (lenalidomide, bortezomib, and dexametha-

sone followed by lenalidomide maintenance) with or without ASCT.5 This data proves an urgent need to identify biomarkers of de novo response and resistance to ASCT in order to improve patient selection for this therapeutic modality.

Melphalan acts by alkylating DNA and causing single-strand DNA breaks as well as other DNA lesions, primarily repaired through the base excision repair (BER) pathway.6 Poly (ADPribose) polymerases (PARP) are a family of enzymes that catalyze the transfer of ADP-ribose to target proteins (poly ADP-ribosylation) and are involved in nucleic acid metabolism, modulation of chromatin structure, DNA synthesis, and DNA repair.7 PARP are a pivotal component of the BER complex, which consists of DNA ligase III, DNA polymerase β, and the XRCC1 proteins, and contributes to BER response to single-strand DNA breaks. Key proteins involved in the BER pathway, including APEX1/2, XRCC1, PARP1, POLD2, have been associated with chemoresistance across many cancer types.8–11 The expression and activity of genes in the BER

pathway increase in response to the accumulation of DNA alkylating or damaging agents as well as radiation-induced DNA damage.12,13 It is therefore unsurprising that the activity of DNA repair through the BER pathway in MM has been associated with response to melphalan.14–16 While individual members of the BER pathway have been studied in the setting of MM and ASCT, comprehensive assessment of this pathway has yet to be conducted.

Because of the heterogeneity inherent to melphalan exposure and anti-myeloma activity, there remains a critical need to better understand biomarkers of melphalan resistance, both de novo and acquired, in patients with MM receiving ASCT.17,18 In this study, we comprehensively assessed the expression of the BER pathway across MM. We leveraged large transcriptomic datasets containing MM tumors to quantify the role of genes in the BER pathway in MM. Our results indicated that the expression of PARP1 and POLD2 were significantly associated with OS in patients with MM treated with ASCT followed by high-dose melphalan treatment. We then used both in vitro and in vivo MM models to test the impact of BER pathway attenuation, through PARP inhibition, on sensitivity to melphalan. Together, our results demonstrate that PARP1 and POLD2 may represent an ASCT-specific biomarker with potential for optimizing the therapeutic modality of melphalan-conditioned ASCT in patients with MM.

Methods

Publicly available datasets

Publicly available data were collected from the Gene Expression Omnibus (GEO), the Multiple Myeloma Research Foundation (MMRF), and the genomics of drug sensitivity in cancer database as approved by the University of Florida Institutional Review Board (#IRB202101136).18-23 Further details are available in the Online Supplementary Appendix.

Multiple myeloma cell lines, reagents and assays

MM cell lines MM1S and NCI-H929 were obtained from American Type Culture Collection (ATCC; Manassas, VA). These cell lines were regularly authenticated using short tandem repeat polymorphism (STRP) analysis as recommended by ATCC, were mycoplasma free, and used within 6 months of receipt from ATCC. All in vitro studies were conducted in at least triplicate and in at least three independent experiments. Therapeutic reagents included melphalan, olaparib, and talazoparib (Sigma-Aldrich). Cell viability was assayed using the MTT Cell Proliferation Assay Kit (Roche, Indianapolis, IN) following the manufacturer's directions. Cellular apoptosis were measured using the Attune NxT Flow Cytometer and the standard manufacture recommended protocol (ThermoFisher). Further details are available in the Online Supplementary Appendix

In vivo experiments

MM1S cells were injected in a phosphate-buffered saline/matrigel suspension of 100 μL in both right and left flanks of 60 nude mice (22 females and 38 males; The Jackson Laboratory, Bar Harbor, ME). Mice were monitored daily until tumors reached 200 mm2 then randomized to receive vehicle control, talazoparib alone, melphalan alone, or melphalan in combination.24,25 Mice were removed from the study if their weight was <80% of baseline for 2 days, or if total tumor size per mouse reached >1,600 mm2. All animal studies were conducted following the Ohio State University Institutional Animal Care and Use Committee (IACUC) approval. Further details are available in the Online Supplementary Appendix.

Statistical methods

All data were analyzed in Rv.4.1.1 (The R Project for Statistical Computing, https://www.r-project.org) or Graphpad Prismv.9.2.0 (GraphPad Software, San Diego, CA). Additional graphics were created with BioRender.com (BioRender, Toronto, Ontario). Members of the BER pathway were defined by the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database.19 Single-sample gene set enrichment analysis (ssGSEA) was conducted using the GSVAv1.40.1 package. Twogroup analysis of continuous variables was conducted using the Mann Whitney Wilcoxon test while analysis of continuous variables across multiple groups was conducted using the Kruskal-Wallis test. Heatmaps were constructed using the pheatmapv1.0.12 package. Uniform manifold approximation and projection for dimension reduction (UMAP) plots were generated using the package umapv0.2.7.0. Survival analysis was tested using Cox proportional-hazards regression for continuous variables and log-rank tests for categorical variables. Stepwise Cox proportional-hazards regression was conducted using the package My.stepwise.coxphv0.1.0. Survival graphs were created using the Kaplan–Meier estimator. Combination index was calculated using CompuSyn20 (ComboSyn, Inc, Paramus, NJ) and Combenefit21 (CRUK Cambridge Institute, Cambridge, UK) using the highest single-agent (HSA) model. Changes in tumor growth over time were tested by ANOVA with repeated measures. Unless otherwise stated, two-sided P values ≤0.05 were considered statistically significant. Adjustment of P values was conducted using false discovery rate. Further details are available in the Online Supplementary Appendix

Results

Genes in the BER pathway across multiple myeloma developmental stages

In order to assess the expression levels of genes in the BER pathway across the developmental stages of MM, we collected data from four publicly available datasets:

GSE13591, GSE23113, GSE6477, GSE5900 (Online Supplementary Table S1). The data coved six developmental stages of MM including: normal plasma cells, monoclonal gammopathy of undetermined significance (MGUS), smoldering multiple myeloma (sMM), MM, relapsed MM (rMM), and plasma cell leukemia (PCL). Studies were analyzed separately due to potential batch effect between studies. Using ssGSEA to measure pathway enrichment, the BER pathway gene set was consistently up-regulated across the development of MM (Figure 1A). In order to further test the difference in expression of genes in the BER pathway between MM and MGUS, we compared expression between disease stages for the three datasets with data available (Figure 1B-D). Genes in the BER pathway were generally more highly expressed in MM. Across all three datasets included in this analysis, APEX1, FEN1, POLD2, POLD3, and UNG were significantly up-regulated, while APEX2, MBD4, PARP1, PARP2, PCNA, POLB, and TDG were significantly up-regulated in MM in at least two datasets. The only gene to be consistently down-regulated in MM compared to MGUS was NEIL3. Using only the genes in the BER pathway, MM and MGUS clustered separately with MM demonstrating consistent deregulation expression of the BER pathway in MM (Online Supplementary Figure S1).

Expression of genes in the BER pathway correlates with overall survival in the training set

In order to determine the prognostic significance of genes in the BER pathway in MM, we collected gene expression and clinical outcomes data for 559 patients with newly diagnosed MM from the geo set GSE2658. All patients in the dataset received ASCT and either Total Therapy 2 (TT2) or Total Therapy 3 (TT3) treatment. Receipt of TT2 or TT3 in this dataset was not associated with OS in this dataset (P=0.95). Due to significant co-expression of genes in the BER pathway (Online Supplementary Figure S2), we used a stepwise Cox proportional-hazards regression to build a multivariable model for OS using genes in the BER pathway and a significance level for entry/stay of 0.1 (Online Supplementary Table S2). The resulting multivariable model included five genes that met our significance threshold (P<0.1): MPG, PARP1, PARP3, POLD1, and POLD2 Using median expression to categorize high and low expression samples, high expression of PARP1 (hazard ratio [HR]=1.76; 95% confidence interval [95%CI]: 1.19-2.61; P=0.005) and POLD2 (HR=1.47; 95% CI: 0.99-2.17; P=0.06) were associated with reduced OS (Figure 2A). High expression of PARP3 (HR=0.45; 95% CI: 0.30-0.67; P<0.0001) and MPG (HR=0.67; 95% CI: 0.45-0.99: P=0.04) were associated with improved OS. POLD1 was not associated with OS (HR=1.30; 95% CI: 0.88-1.93; P=0.19). In a multivariable Cox proportional-hazards regression model of PARP1, POLD2, PARP3, MPG, and POLD1 as categorical variables, using a median gene expression cut-off, PARP1,

POLD2, PARP3, and MPG remained significantly associated with OS (Figure 2B). The multivariate model was not changed when the total therapy cohort was included (P=0.95).

PARP1 and POLD2 gene expression negatively correlate with overall survival in the validation set

In order to validate the findings from GSE2658, we collected gene expression and clinical data for 356 patients with MM who had received ASCT and 319 patients with MM who had never received a transplant from the MMRF (Online Supplementary Figure S3). In patients with MM who had received ASCT, we used a median expression cut-off to categorize high- and low-expression samples. High expression of PARP1 (HR=2.15; 95% CI: 1.34-3.45; P=0.002) and POLD2 (HR=1.67; 95% CI: 1.04-2.68; P=0.03) was significantly associated with reduced OS (Figure 3A). PARP3 (HR=0.88; 95% CI: 0.55-1.41; P=0.58) and MPG (HR=0.79; 95% CI:0.49-1.27; P=0.34) were not associated with OS.

Given the availability of extensive clinical data in the MMRF, we conducted univariable and multivariable Cox proportional-hazards regression using both gene expression and clinical variables (Online Supplementary Table S3). In a multivariable clinical-genomic model including patient age, revised International Staging System (R-ISS), International Myeloma Working Group classification (IMWG), TP53 status, Eastern Cooperative Oncology Group (ECOG) performance status, pre-ASCT induction therapy, and post-ASCT maintenance therapy, PARP1 and POLD2 were independently associated with OS (Figure 3B). It is notable that in this cohort R-ISS staging was not statistically associated with OS, driven by lack of full staging data and a low number of death events in the patients with R-ISS staging data available. Future analysis should make use of full up-to-date staging systems in their analysis.

We then tested the prognostic value of PARP1, POLD2, PARP3, and MPG in patients with MM who had never received a transplant to test whether the association between expression of genes in the BER pathway and OS in MM was inherent to the disease biology or related to receipt of ASCT. For patients with MM who had never received a transplant, expression of PARP1, POLD2, PARP3, and MPG did not correlate with OS (Table 1). The expression levels of these four genes were not significantly different between patients with MM receiving ASCT and patients who had not received ASCT (Online Supplementary Figure S4A-D). It is important to note that the patients who received ASCT and did not receive ASCT cannot be directly compared as they represent populations selected by clinical features, making them inherently different. The findings here only underscore that the relationship between PARP1 and POLD2 are validated only for patients receiving ASCT.

A potential co-variate of these findings is the known prognostic effect of chromosome 1q duplications in MM.22 Specifically, amplification of the 1q21 locus has been associated with poor clinical outcomes.23 In the MMRF dataset, PARP1 (1q42.12) and POLD2 (7p13) were overexpressed in samples with gain of 1q21 in at least 20% of cells identified by sequential fluorescence in situ hybridization (seqFISH) (Online Supplementary Figure S4E). PARP1 and POLD2 remained weakly correlated in samples with (Spearman’s correlation coefficient: 0.04) or without (Spearman’s correlation coefficient: 0.03) gain of 1q21. In order to test the role of 1q gain

in MM, we correlated the expression of genes located on 1q21 hypothesized to play a role in the mechanism of gain of 1q21 related clinical outcomes24 (Online Supplementary Figure S4F), including CKS1B (1q21.3), IL6R (1q21.3), MCL1 (1q21.2), and BCL9 (1q21.2). In patients receiving ASCT, all 1q21 genes queried were increased in MM with gain of 1q21, while in univariable analysis gain of 1q21, CSK1B, and IL6R were statistically associated with reduced OS (Online Supplementary Figure S4G, H). Increased PARP1 expression were associated with poor OS in both 1q21 amplified and non-amplified disease (Online Supplementary Figure S4I).

Figure 1. Base excision repair gene expression is increased across mupliple myeloma development. (A) Across datasets, the base excision repair (BER) pathway, as measured by single sample gene set enrichment analysis (ssGSEA), was consistently upregulated across the development of multiple myeloma (MM). P values were calculated using two-sided Mann Whitney Wilcoxon tests. ssGSEA is an extension of GSEA, which calculates gene set enrichment scores for each sample. Each ssGSEA enrichment score represents the degree of gene set up- or down-regulated within a sample. (B, C) Comparing monoclonal gammopathy of undetermined significance (MGUS) and MM in GSE13591, GSE2113, and GSE6477 demonstrated significant upregulation of genes in the BER pathway in MM. (D) Significant gene expression differences between MGUS and MM for genes in the BER pathway. All genes included are statistically significantly different as measures by Mann Whitney Wilcoxon test. Plot represents the center as the mean, error bars as ± standard deviation. ns: not significant; *P≤ 0.05; **P<0.01; ***P<0.001; ****P<0.0001.

For patients who had never received a transplant, all 1q21 genes queried remained increased in MM with gain of 1q21; however, no 1q21 gene or 1q21 status were associated with OS (Online Supplementary Figure S4J, K). PARP1 was not

associated with OS regardless of 1q21 status (Online Supplementary Figure S4L). While the role of PARP1 expression cannot be dissected from 1q amplification in this dataset, this data suggests that 1q status and PARP1 expression

Figure 2. Base excision repair gene expression correlation with survival in training set. Gene expression and survival data for 559 pretreatment patients with multiple myeloma (MM) who received autologous stem cell transplant (ASCT) were collected from GSE2658. Stepwise Cox proportional-hazards regression on identified 5 genes correlated with overall survival (OS) as continuous variables. (A) Using a median expression to categorize high- and low-expression samples, high expression of PARP1 and POLD2 were associated with reduced OS. High expression of PARP3 and MPG were associated with improved OS. Using median expression, POLD1 was not associated with a change in OS. (B) In a multivariable Cox proportional-hazards regression model of PARP1, POLD2, PARP3, MPG, and POLD1 as categorical variables using a median gene expression, PARP1, POLD2, PARP3, and MPG were significantly associated with OS. Points represent hazard ratios (HR), error bars represent 95% confidence interval (95% CI).

Table 1. Biomarkers of overall survival in patients with multiple myeloma.

may influence OS following ASCT and warrants further study.

PARP1 and POLD2 co-expression define high-risk population specific to patients with multiple myeloma receiving autologous stem cell transplant

In order to determine the prognostic significance of PARP1 and POLD2 co-expression in patients with MM, we used a median gene expression cut-off to categorize high- and low-expression samples for each gene into four

PARP1/POLD2 categories: High/High, High/Low, Low/High, and Low/Low. High expression of both PARP1 and POLD2 was associated with poor OS in patients from GSE2658 (P=0.0001; Figure 4A) and MMRF (P =0.0006; Figure 4B) who had received ASCT. Multivariable analysis of PARP1 and POLD2 co-expression demonstrated significance for PARP1 in both sets while POLD2 was only statistically significant in the MMRF dataset. For patients with MM in the MMRF dataset who never received transplant, PARP1 and POLD2 co-expression was not associated with OS (P=0.72;

Figure 3. PARP1 and POLD2 expression correlate with survival in validation set. Gene expression, clinical, and survival data for 356 patients with multiple myeloma (MM) who received autologous stem cell transplant (ASCT) was collected from the Multiple Myeloma Research Foundation (MMRF). (A) Using a median expression to categorize high- and low-expression samples, high expression of PARP1 and POLD2 were associated with reduced overall survival (OS). High expression of PARP3 and MPG were associated with improved OS. (B) In a multivariable Cox proportional-hazards regression model of PARP1, POLD2, and significant clinical variables, PARP1 and POLD2 maintained statistical significant independent of clinical factors. Points represent hazard ratios (HR), error bars represent 95% confidence interval (95% CI).

Figure 4C). Across the three cohorts, PARP1 and POLD2 were similarly, but weakly co-expressed (Spearman’s correlation coefficients 0.19, 0.14, and 0.16).

PARP inhibition potentiates melphalan cytotoxicity in preclinical multiple myeloma models

Given the link between upregulation of PARP1 and POLD2 and poor OS in patients with MM receiving ASCT, we hypothesized that attenuation of the BER pathway using PARP Food and Drug Administration-approved inhibitors (olaparib, talazoparib) increase melphalan sensitivity in preclinical MM models (Online Supplementary Figure S5A). In order to focus on the effect of BER pathway attenuation in MM, we selected two MM cell lines with the wild-type TP53 gene, MM1S and NCI-H929, to evaluate the effect of BER pathway attenuation.25,26 mRNA expression of PARP1 and POLD2 in these cell lines is representative of low-expressing (NCI-H929) and high-expressing (MM1S) MM cell lines (Online Supplementary Figure S5B). Despite the difference in PARP1 and POLD2 ex-

pression between MM1S and NCI-H929, we identified no significant difference in melphalan half maximal inhibitory concentration (IC50s) (Online Supplementary Figure S5C). As expected, PARP inhibition had no intrinsic anti-MM effect and did not exhibit cytotoxicity in either cell line (Online Supplementary Figure S5D). The combination of PARP inhibitors with melphalan was highly synergistic as measured by cell viability (Figure 5A, B) and apoptotic assays (Figure 5C). We then tested the drug combination in a subcutaneous xenograft model using the MM1S cell line (Figure 6A). Mirroring melphalan administration in ASCT, mice assigned to receive melphalan received a single intravenous (IV) dose on day 0. Mice assigned to receive the PARP inhibitor talazoparib received two doses daily from day -3 to day +3. There was no difference in tumor size or mouse weight between treatment groups (vehicle control, talazoparib, melphalan, talazoparib + melphalan) from day -3 to day 0 (Figure 6B). Mice treated with melphalan + talazoparib demonstrated smaller tumor volume from day +3 until the end of the study (mean

Figure 4. PARP1 and POLD2 co-expression highly prognostic in the setting of autologous stem cell transplant. For patients with multiple myeloma (MM) who received autologous stem cell transplant (ASCT) (GSE2658 (A), Multiple Myeloma Research Foundation (MMRF) (B)) high expression of both PARP1 and POLD1 was associated with poor overall survival (OS). Multivariable analysis demonstrated significance for PARP1 in both sets while POLD1 was only trended towards significance in GSE2658. For patients with MM who never received transplant (C), PARP1 and POLD1 were not associated with OS. Points represent hazard ratios (HR), error bars represent 95% confidence interval (95% CI). ns: not significant; ****P<0.0001.

C

Figure 5. PARP inhibition potentiates melphalan-mediated cytotoxicity for in vitro multiple myeloma models. (A) Combination index for olaparib or talazoparib and melphalan for multiple myeloma (MM) cell lines calculated using CompuSyn. (B) Combination index for olaparib or talazoparib and melphalan for MM cell lines calculated using Combenifit. (C) Apoptosis, measured by flow cytometry, was increased with combination of talazoparib and melphalan for 2 of 4 cell lines. Plots represent the center as the mean, error bars as ± standard error of the mean. ns: not significant; *P≤0.05; to.

± standard error of the mean [SEM]; melphalan 313.6±25.1 mm2 vs. melphalan + talazoparib 185.4±16.8 mm2; P=0.0001; Figure 6C; Online Supplementary Figure S6A). The combination arm did experience severe adverse effects from their treatment with 40% of mice in this arm being removed from the study due to extreme weight loss (Figure 6D; Online Supplementary Figure S6B). It is notable that no other treatment arm experienced this effect and that the mice receiving melphalan + talazoparib with severe weight loss had no measurable disease at the time of their removal from the study.

Discussion

Despite significant therapeutic advances for MM, ASCT following high-dose, single-agent melphalan conditioning

remains a cornerstone of therapy.27 Understanding ASCT and melphalan resistance is therefore vitally important. DNA repair capability, particularly the BER pathway, has unsurprisingly been associated with response to melphalan and thus ASCT.28 Using a big-data, clinical-genomic approach, we investigated the role of the BER pathway in MM and its prognostic relevance. Similar to previous findings, we identified an increase in the expression of genes in the BER pathway across the development of MM.29–32 Prior studies demonstrated upregulation of specific members of the BER pathway such as APEX1, APEX2, and PARP1, across the development of MM.33–36 Our findings suggest that these results are not limited to single genes, but are demonstrable across a majority of genes in the BER pathway.

In order to test the prognostic value of genes in the BER

Figure 6. PARP inhibition potentiates melphalan-mediated cytotoxicity for in vivo multiple myeloma models. An in vivo subcutaneous xenograft model of using the MM1S cell line. (A) Mirroring melphalan administration in autologous stem cell transplant (ASCT), mice assigned to receive melphalan received a single intravenous dose on day 0. Mice assigned to receive the PARP inhibitor talazoparib received 2 doses daily from day -3 to day +3. Graphic created with BioRender.com. (B) Tumors in mice treated with vehicle control or talazoparib were not significantly different over the course of treatment (q=0.82, ANOVA with repeated measures). The combination of talazoparib with melphalan reduced tumor burden to a greater extent than melphalan alone (q<0.0001). (C) By day +3, melphalan + talazoparib had significantly reduced tumor size compared to melphalan alone (P=0.0001; Mann Whitney Wilcoxon test). (D) Mice were removed from the study due to excessive tumor growth, excessive weight loss or at the end of the study period (EOS). Graph represent the proportion of mice in each treatment arm and the reason they were removed from the study. The melphalan + talazoparib arm experienced severe adverse effects from treatment with 40% mice in this arm removed due to extreme weight loss. Plot represent the center as the mean, error bars as ± standard error of the mean. ns: not significant; ***P<0.001; ****P<0.0001.

pathway in MM, we collected data for patients with MM who had received ASCT from two large datasets. These datasets represent clinically homogenous (GSE2658) and heterogeneous (MMRF) treatment courses from patients in controlled clinical trials as well as in real-world data. We found that elevated expression of both PARP1 and POLD2 were consistently associated with poor survival outcomes, independent of other clinical factors.35,37,38 This confirms prior association of PARP1 expression as a negative prognostic biomarker in MM and extends the current knowledge of this biomarker as well as POLD2. 28 While the correlation between PARP1/POLD2 expression with ASCT-related OS is purely prognostic, it is notable that these findings do not extend to patients with MM who did not receive ASCT. This may suggest that our findings may be due to the relationship between PARP1/POLD2 expression and resistance to melphalan-based ASCT. The direct mechanism relating PARP1/POLD2 expression with clinical outcomes for patients receiving ASCT requires further study.

In order to test the potential for targeting the BER pathway during melphalan administration, we attenuated PARP1 pharmacologically using FDA-approved PARP inhibitors in combination with melphalan in preclinical models of MM. Prior studies by Patel et al. and Xiong et al. found that PARP1 inhibitors are highly synergistic with melphalan in some, but not all, cell line models of MM.28,39 Our data here validates these findings and extends them to additional cell lines and treatment strategies. The clinical applicability of these findings, however, are unclear.28,39 Combination of PARP inhibition with traditional cytotoxic chemotherapy has been tested in a large number of clinical trials; however, few have been determined to be successful due to high rates of dose-limiting myotoxicity. This effect can be seen in our in vivo studies with 40% of mice receiving both melphalan and talazoparib experiencing severe weight loss and failure to thrive. It is unclear if this adverse event is a complication of the intended myeloablation, or an unintended site of action such as gut toxicity. In the setting of ASCT however, myotoxicity is a primary goal of conditioning therapy with transplant rescue vital to patient survival. Mice in our study did not receive stem cell transplants, which would better mirror the clinical setting of ASCT. Further study is necessary to determine the effect on minimal residual disease and adverse event rates if combining melphalan with a PARP inhibitor. Our data suggests that PARP1 is vital to the efficacy of this standard treatment regimen. However, there are key limitations to the broader application of our findings. This study is limited by the retrospective nature of the computational analysis conducted. Because our study did not include other relevant disease endpoints such as rate of PFS, disease-free survival, or minimal residual disease

status, it is not possible to determine the predictive effect of PARP1 and POLD2 in patients with MM receiving ASCT. Given the results of the DETERMINATION trial, PFS will be a vital endpoint for future study. Further research is necessary to fully elucidate the mechanism of the BER pathway in MM. Despite the limitations of our analysis, the stability of the PARP1 and POLD2 signal identified over multiple large datasets and both microarray and RNA-sequencing methods suggests that the external validity of our findings may still be strong. Finally, it is clear that PARP1 and POLD2 alone do not explain the total biological or clinical heterogeneity present in MM. Further research is necessary to understand the role of the BER pathway in the greater molecular context of MM.40,41 There remains a significant need to develop predictive and prognostic biomarkers in MM. In this study, we found that elevated expression of genes in the BER pathway, specifically the PARP1 and POLD2 genes, correlated with poor survival in patients with MM who received ASCT. Furthermore, targeting of the BER pathway during melphalan therapy may be a potential clinical strategy to address melphalanresistance mediated via PARP1 and POLD2. Prospective clinical evaluation will be required to validate these findings.

Disclosures

No conflicts of interest to disclose.

Contributions

JL, MJP and NDS designed the research. MT, JL, KK, AKP, ED, ZV and NDS performed the research. MT, JL, ACT, BLF, MJP and NDS analyzed data. MT, JL and NDS wrote the manuscript. MT, JL, KK, AKP, ED, ZV, CCH, JKL, ACT, BLF, MJP and NDS critically reviewed the manuscript.

Acknowledgments

Parts of the data presented in this report were generated as part of the Multiple Myeloma Research Foundation Personalized Medicine Initiatives (https://research.themmrf.org and www.themmrf.org).

Funding

This study was supported by a Pelotonia IDEA award (46050-502048) (to CCH and MJP) and startup research grants from the College of Pharmacy, The Ohio State University (to MJP) and the College of Pharmacy, The University of Florida (to NDS).

Data-sharing statement

All publicly available data is noted within the manuscript with directions to obtain raw and processed data as appropriate. For all original data generated in this manuscript, please contact the corresponding author.

References

1. Kumar SK, Callander NS, Adekola K, et al. Multiple myeloma, version 3.2021, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw. 2020;18(12):1685-1717.

2. Hamed RA, Bazarbachi AH, Malard F, Harousseau JL, Mohty M. Current status of autologous stem cell transplantation for multiple myeloma. Blood Cancer J. 2019;9(4):44.

3. Kumar S, Paiva B, Anderson KC, et al. International Myeloma Working Group consensus criteria for response and minimal residual disease assessment in multiple myeloma. Lancet Oncol. 2016;17(8):e328-e346.

4. Gu J, Liu J, Chen M, Huang B, Li J. Longitudinal flow cytometry identified “minimal residual disease” (MRD) evolution patterns for predicting the prognosis of patients with transplant-eligible multiple myeloma. Biol Blood Marrow Transplant. 2018;24(12):2568-2574.

5. Richardson PG, Jacobus SJ, Weller EA, et al. Triplet therapy, transplantation, and maintenance until progression in myeloma. N Engl J Med. 2022;387(2):132-147.

6. van Kan M, Burns KE, Helsby NA. A systematic review of interindividual differences in the DNA repair processes involved in melphalan monoadduct repair in relation to treatment outcomes. Cancer Chemother Pharmacol. 2021;88(5):755-769.

7. Pascal JM. The comings and goings of PARP-1 in response to DNA damage. DNA Repair (Amst). 2018;71:177-182.

8. Yang J, Yang D, Cogdell D, et al. APEX1 gene amplification and its protein overexpression in osteosarcoma: correlation with recurrence, metastasis, and survival. Technol Cancer Res Treat. 2010;9(2):161-169.

9. Weaver DA, Crawford EL, Warner KA, Elkhairi F, Khuder SA, Willey JC. ABCC5, ERCC2, XPA and XRCC1 transcript abundance levels correlate with cisplatin chemoresistance in non-small cell lung cancer cell lines. Mol Cancer. 2005;4(1):18.

10. Higuchi F, Nagashima H, Ning J, Koerner MVA, Wakimoto H, Cahill DP. Restoration of temozolomide sensitivity by PARP inhibitors in mismatch repair deficient glioblastoma is independent of base excision repair. Clin Cancer Res. 2020;26(7):1690-1699.

11. Givechian KB, Garner C, Garban H, Rabizadeh S, Soon-Shiong P. CAD/POLD2 gene expression is associated with poor overall survival and chemoresistance in bladder urothelial carcinoma. Oncotarget. 2018;9(51):29743-29752.

12. Rahmanian S, Taleei R, Nikjoo H. Radiation induced base excision repair (BER): a mechanistic mathematical approach. DNA Repair (Amst). 2014;22:89-103.

13. Lee KJ, Piett CG, Andrews JF, Mann E, Nagel ZD, Gassman NR. Defective base excision repair in the response to DNA damaging agents in triple negative breast cancer. PLoS One. 2019;14(10):e0223725.

14. Persaud AK, Li J, Johnson JA, et al. XRCC1-mediated DNA repair is associated with progression-free survival of multiple myeloma patients after autologous stem cell transplant. Mol Carcinog. 2019;58(12):2327-2339.

15. Chen Q, Van der Sluis PC, Boulware D, Hazlehurst LA, Dalton WS. The FA/BRCA pathway is involved in melphalan-induced DNA interstrand cross-link repair and accounts for melphalan resistance in multiple myeloma cells. Blood. 2005;106(2):698-705.

16. Saitoh T, Oda T. DNA damage response in multiple myeloma: the role of the tumor microenvironment. Cancers (Basel). 2021;13(3):504.

17. Nath CE, Trotman J, Tiley C, et al. High melphalan exposure is

associated with improved overall survival in myeloma patients receiving high dose melphalan and autologous transplantation. Br J Clin Pharmacol. 2016;82(1):149-159.

18. Seligson ND, Warner JL, Dalton WS, et al. Recommendations for patient similarity classes: results of the AMIA 2019 workshop on defining patient similarity. Am Med Inform Assoc. 2020;27(11):1808-1812.

19. Kanehisa M, Sato Y, Kawashima M, Furumichi M, Tanabe M. KEGG as a reference resource for gene and protein annotation. Nucleic Acids Res. 2016;44(D1):D457-D462.

20. Zhang N, Fu JN, Chou TC. Synergistic combination of microtubule targeting anticancer fludelone with cytoprotective panaxytriol derived from panax ginseng against MX-1 cells in vitro: experimental design and data analysis using the combination index method. Am J Cancer Res. 2016; 6(1):97-104.

21. Di Veroli GY, Fornari C, Wang D, et al. Combenefit: an interactive platform for the analysis and visualization of drug combinations. Bioinformatics. 2016;32(18):2866-2668.

22. Schmidt TM, Barwick BG, Joseph N, et al. Gain of chromosome 1q is associated with early progression in multiple myeloma patients treated with lenalidomide, bortezomib, and dexamethasone. Blood Cancer J. 2019;9(12):94.

23. Weinhold N, Salwender HJ, Cairns DA, et al. Chromosome 1q21 abnormalities refine outcome prediction in patients with multiple myeloma - a meta-analysis of 2,596 trial patients. Haematologica. 2021;106(10):2754-2758.

24. Burroughs Garcìa J, Eufemiese RA, Storti P, et al. Role of 1q21 in multiple myeloma: from pathogenesis to possible therapeutic targets. Cells. 2021;10(6):1360.

25. Iorio F, Knijnenburg TA, Vis DJ, et al. A landscape of pharmacogenomic interactions in cancer. Cell. 2016;166(3):740-754.

26. Gaballa S, Saliba RM, Srour S, et al. Outcomes in patients with multiple myeloma with TP53 deletion after autologous hematopoietic stem cell transplant. Am J Hematol. 2016;91(10):E442-E447.

27. Koomen DC, Meads MB, Magaletti DM, et al. Metabolic changes are associated with melphalan resistance in multiple myeloma. J Proteome Res. 2021;20(6):3134-3149.

28. Patel PR, Senyuk V, Sweiss K, et al. PARP inhibition synergizes with melphalan but does not reverse resistance completely. Biol Blood Marrow Transplant. 2020;26(7):1273-1279.

29. Beksac M, Balli S, Akcora Yildiz D. Drug targeting of genomic instability in multiple myeloma. Front Genet. 2020;11:228.

30. Teoh PJ, An O, Chung TH, et al. Aberrant hyperediting of the myeloma transcriptome by ADAR1 confers oncogenicity and is a marker of poor prognosis. Blood. 2018;132(12):1304-1317.

31. Taiana E, Cantafio MEG, Favasuli VK, et al. Genomic instability in multiple myeloma: a “non coding RNA” perspective. Cancers (Basel). 2021;13(9):2127.

32. Gourzones-Dmitriev C, Kassambara A, Sahota S, et al. DNA repair pathways in human multiple myeloma: role in oncogenesis and potential targets for treatment. Cell Cycle. 2013;12(17):2760-2773.

33. Kumar S, Talluri S, Pal J, et al. Role of apurinic/apyrimidinic nucleases in the regulation of homologous recombination in myeloma: mechanisms and translational significance. Blood Cancer J. 2018;8(10):92.

34. Liao C, Talluri S, Kumar S, et al. Base excision repair and homologous recombination pathway intermediates drive genomic instability and evolution in Myeloma. Blood. 2020;136(Suppl 1):S27-28.

35. Neri P, Ren L, Gratton K, et al. Bortezomib-induced “BRCAness” sensitizes multiple myeloma cells to PARP inhibitors. Blood. 2011;118(24):6368-6379.

36. Shen HY, Tang HL, Zheng YH, Feng J, Dong BX, Chen XQ. The PARP1 inhibitor niraparib represses DNA Damage repair and synergizes with temozolomide for antimyeloma effects. J Oncol. 2022;2022:2800488.

37. Krokan HE, Bjoras M. Base excision repair. Cold Spring Harb Perspect Biol. 2013;5(4):a012583.

38. Alagpulinsa DA, Ayyadevara S, Yaccoby S, Shmookler Reis RJ. A cyclin-dependent kinase inhibitor, dinaciclib, impairs homologous recombination and sensitizes multiple myeloma cells to PARP inhibition. Mol Cancer Ther. 2016;15(2):241-250.

39. Xiong T, Wei H, Chen X, Xiao H. PJ34, a poly(ADP-ribose)

polymerase (PARP) inhibitor, reverses melphalan-resistance and inhibits repair of DNA double-strand breaks by targeting the FA/BRCA pathway in multidrug resistant multiple myeloma cell line RPMI8226/R. Int J Oncol. 2015;46(1):223-232.

40. Schjesvold FH, Dimopoulos MA, Delimpasi S, et al. Melflufen or pomalidomide plus dexamethasone for patients with multiple myeloma refractory to lenalidomide (OCEAN): a randomised, head-to-head, open-label, phase 3 study. Lancet Haematol. 2022;9(2):e98-e110.

41. Bashir Q, Thall PF, Milton DR, et al. Conditioning with busulfan plus melphalan versus melphalan alone before autologous haemopoietic cell transplantation for multiple myeloma: an open-label, randomised, phase 3 trial. Lancet Haematol. 2019;6(5):e266-e275.

Low T-cell proportion in the tumor microenvironment is associated with immune escape and poor survival in diffuse large B-cell lymphoma

Correspondence: Joo Y. Song josong@coh.org

Received: October 12, 2022.

1Department of Pathology, City of Hope Medical Center, Duarte, CA, USA; 2ImmunoOncology, Beckman Research Institute, City of Hope Medical Center, Duarte, CA, USA;

3Department of Pathology and Microbiology, University of Nebraska Medical Center, Omaha, NE, USA; 4Department of Hematology, CancerCare Manitoba, Manitoba, Canada; 5Department of Pathology, SUNY Upstate Medical University, Syracuse, NY, USA; 6British Columbia Research Center, Vancouver, British Columbia, Canada; 7Department of Hematology & Hematopoietic Cell Transplantation, City of Hope Medical Center, Duarte, CA, USA and 8Department of Pathology, University of Michigan, Ann Arbor, MI, USA

Abstract

Accepted: January 3, 2023.

Early view: January 12, 2023.

https://doi.org/10.3324/haematol.2022.282265

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

The tumor microenvironment (TME) is important in the pathogenesis and prognosis of lymphoma. Previous studies have demonstrated that features of the diffuse large B-cell lymphoma (DLBCL) TME can be associated with prognosis, but questions remain about the mechanisms underlying these TME features, and the interplay between tumor cells and the local TME. Therefore, we performed multispectral immunofluorescence (mIF) using two 6-color panels to interrogate the cellular proportions of T-cell subsets, macrophages, and natural killer cells in 57 cases of de novo DLBCL treated with R-CHOP chemotherapy. We found that very low CD3+ T-cell proportion and low CD4+PD1+ and CD8+PD1+ T cells have poor survival compared to those with a high T-cell proportion. Also, cases with concurrently low TIM3 and PD1 have a poor prognosis. This poor prognosis with low T-cell proportion was validated using immune deconvolution of gene expression profiling data from 351 cases of DLBCL and an additional cohort of 53 cases of DLBCL using routine immunohistochemistry. In addition, cases with loss of B2M, HLA I and/or HLA II protein expression on the tumor cells also had a low T-cell proportion, providing evidence that lack of these proteins allows for immune evasion. Overall, our results show that patients with DLBCL with a low T-cell proportion in the TME have a poor survival when treated with R-CHOP and exhibit mechanisms of immune escape.

Introduction

The tumor microenvironment (TME) plays an important role in the pathogenesis and outcome of lymphomas.1-3 In diffuse large B-cell lymphoma (DLBCL), the majority of the cellular content is usually tumor cells with a paucity of non-malignant cells in the TME. Malignant B cells typically efface the architecture and few cells remain except for scattered histiocytes, natural killer (NK) cells, stromal cells, and T cells.2 Immune cells in the TME may express inhibitory receptors such as programmed cell death protein 1 (PD1), lymphocyte-activation gene 3 (LAG3), and Tcell immunoglobulin and mucin-domain containing 3 (TIM3), which may inhibit anti-tumor immune surveillance and facilitate the proliferation and survival of neoplastic cells.4 Whereas a number of studies5-7 have comprehensively evaluated the genomic profile of DLBCL, and

others8-14 have studied the relationship between immune cell proportions in the TME (particularly of T-cell subsets) and outcomes in DLBCL, comprehensive studies of the intersection between the tumor microenvironment and genomics of DLBCL are limited. Therefore, we performed a comprehensive analysis of the TME in DLBCL using multispectral immunofluorescence in conjunction with genomic analysis in a clinically-annotated cohort of patients with DLBCL treated with R-CHOP.

Methods

We identified 57 diagnostic cases of de novo DLBCL with ample material for tissue microarray (TMA) construction (2 mm cores) who were treated with R-CHOP at the City of Hope Medical Center, Duarte. Patients with primary

Joo Y. Song,1 Mary Nwangwu,2 Ting-Fang He,2 Weiwei Zhang,3 Hany Meawad,1 Victoria Bedell,1 Joyce Murata-Collins,1 Pamela Skrabek,4 Michel R. Nasr,5 David Scott,6 James Godfrey,7 Peter Lee,2 Wing C. Chan,1 Dennis D. Weisenburger,1,3 Anamarija M. Perry8 and Alex F. Herrera7

mediastinal B-cell lymphoma, Epstein-Barr virus (EBV)-positive DLBCL, T-cell/histiocyte-rich large B-cell lymphoma, prior low-grade B-cell lymphoma, or an immunocompromised state (including human immunodeficiency virus infection) were excluded. This study was approved by the Institutional Review Boards at the City of Hope Medical Center and the University of Manitoba.

Immunohistochemistry and fluorescence in situ hybridization cytogenetic analysis on tissue microarrays

Cases were re-reviewed to confirm the diagnosis of DLBCL, not otherwise specified (NOS). Immunohistochemistry (IHC) was performed using formalin-fixed, paraffin-embedded tissue microarrays and antibodies to CD20, CD3, CD10, BCL6, MUM1, MYC, BCL2, B2M, HLA I, HLA II, and TIM3. FISH cytogenetic analysis for MYC, BCL2 and BCL6 gene rearrangements was performed.

Mutation analysis and copy number analysis

We used a custom targeted panel of 334 genes, which includes the most frequently mutated genes in B-cell lymphoma, and performed DNA sequencing on an Illumina HiSeq 2500 (Illumina, San Diego, CA, USA) as previously described.15 The Oncoscan Copy Number Variation (CNV) assay (ThermoFisher, Waltham, MA, USA) was performed.

Gene expression analysis for classification

We used extracted RNA on the nCounter platform (NanoString Technologies, Seattle, WA, USA) to determine the cell of origin (COO) using the Lymphoma/Leukemia Molecular Profiling Project (LLMPP) code set Lymph2Cx16 as well as the DLBCL90 double-hit gene expression (DHIT) signature17 (DHITsig) on the nSolver as previously described.15

Multispectral immunofluorescence

For multispectral immunofluorescence (mIF) analysis, staining was performed using the Opal 7 kit (PerkinElmer, Waltham, MA, USA). Two panels were used (Panel 1: CD3, CD8, CD4, PD1, PAX5, DAPI; Panel 2: PAX5, CD163, CD79a, PD-L1, CD56, DAPI), and the stains were scanned on the Vectra spectral imaging system and analyzed using the InForm software.

Validation cohorts

Immune cellular proportion determination using gene expression profiling and CIBERSORTx - We used the gene expression profiling (GEP) data from 351 cases in the study of Lenz et al. of de novo DLBCL: 184 germinal center phenotype (GCB) DLBCL, 167 activated B-cell phenotype (ABC) DLBCL treated with R-CHOP.18 We used the CIBERSORTx method19 to determine the relative cell proportions (e.g., total T cells, follicular helper T cells, regulatory T cells, NK cells, and macrophages) in this independent cohort to confirm the relative cellular proportions by GEP.

Immunohistochemical staining for T cells and digital scanning - To determine whether routine IHC would correlate with the findings by mIF, we used another independent cohort of 54 cases of de novo DLBCL (37 GCB DLBCL, 17 ABC DLBCL) treated with R-CHOP as well as the discovery cohort of DLBCL. Cases on a TMA were stained with CD3 and PD1, and analyzed using the QuPath v0.3.0 qualitative pathology and bioimage analysis software. For further methodological details, including statistical approaches correlating molecular and clinicopathological findings, please see the Online Supplementary Appendix.

Results

Of the 57 cases included in the original cohort, there were 42 cases of GCB DLBCL (74%), 12 cases of ABC DLBCL (21%), and 3 cases of unclassified DLBCL (5%) as determined by Nanostring. There were 5 cases that were double- or triple-hit lymphoma for MYC/BCL2 (n=4) or MYC/BCL2/BCL6 (n=1) by FISH analysis, and 8 cases that were DHITsig-positive (pos) and 32 were DHITsig-negative (neg) by DLBCL90 analysis. The LymphGen tool (https://llmpp.nih.gov/lymphgen/index.php)5 was also employed and the cases were defined as follows: 11 cases (19%) were EZB, 5 cases (9%) BN2, 2 cases (4%) ST2, 3 cases (5%) MCD, 3 (5%) A53, and 33 cases (58%) as other.

Tumor microenvironment evaluation by multispectral immunofluorescence

Among the 57 DLBCL cases evaluated in our discovery cohort, the median proportions of each cell type observed were: 17% for CD3+ T cells (quartiles 1-3, 7-27%), 5.2% for CD4+ T cells (quartiles 1-3, 1.6-8.2%), 10.1% for CD8+ T cells (quartiles 1-3, 2.7-15.3%), 51% for B cells (quartiles 1-3, 3672%), 1.8% for plasma cells (quartiles 1-3, 0.6-6.2%), 2% for macrophages (quartiles 1-3, 0.8-9%), 10% for CD8+ T cells (quartiles 1-3, 3-15%), 0.006% for NK cells (quartiles 1-3, 0.003-0.2%), and 7% for PD1+ T cells (quartiles 1-3, 3-14%). Using logistic regression analysis, five cellular proportions in the TME were associated with a poor overall survival (OS) and progression-free survival (PFS) (Online Supplementary Figure S1). These were: 1) low total T-cell proportion by CD3 (Figure 1A and B, Figure 2A and B; ROC cutoff, 6.95%); 2) low CD4+ PD1+ T-cell proportion (CD3+CD4+PD1+/total cells) (Figure 2C and D; ROC cutoff, 2.55%); 3) low CD8+ PD1+ T cells (CD8+PD1+/total cells) (Figures 1C and D, 2E and F; ROC cutoff, 1.2%); 4) high Bcell/tumor cell proportion (ROC cutoff, 74%); and 5) high plasma cell proportion (ROC cutoff, 3.5%). The 5-year OS for patients with low (15/57 cases) and high (42/57 cases) total CD3+ T-cell proportions was 48% (95% confidence interval [CI]: 21-71%) versus 86% (95%CI: 72-99%), respectively (P=0.0036) (Figure 2A); for low (13/57 cases) versus

high (44/57 cases) CD4+PD1+ T cells, 39% (95%CI: 11-65%) versus 87% (95%CI: 75-98%), respectively (P=0.0004) (Figure 2C); and for low (11/57 cases) versus high (46/57 cases) CD8+PD1+ T cells, 32% (95%CI: 1-62%) versus 85% (95%CI: 73-98%), respectively (P=0.0004) (Figure 2E). The PFS curves showed similar findings (Figure 2B, D and F). Total T-cell and T-cell subset proportions remained significantly associated with survival irrespective of the COO (Online Supplementary Figure S2). No other immune cell subtype was associated with outcome. The clinical characteristics of patients with low and high total CD3+ T-cell proportions are shown in Table 1. The clinical characteristics, COO and DHITsig frequencies were similar between the two groups. Multivariate analysis adjusted for age and International Prognostic Index (IPI) (low=IPI 1,2; high=IPI 3,4), demonstrated that low CD3+ T-cell proportion, low CD3+CD4+PD1+, and low CD3+CD8+PD1+ T cells, and low CD4+PD1+ T cells with low TIM3 were all associated with OS, particularly in the high IPI groups (Online Supplementary Figure S3).

High-grade B-cell lymphomas, particularly cases with a double-hit signature, have been shown to be low in T cells.17 Since DHITsig-pos cases have a poor prognosis with

ns: not significant (P>0.05); ABC: activated B-cell type diffuse large B-cell lymphoma (DLBCL); GCB: germinal center B-cell type DLBCL; advanced stage: stages III/IV; LDH: lactate dehydrogenase; IPI: International Prognostic Index; DHITsig-positive: double-hit gene expression signature-positive; N: number. Low and high T-cell content cutoff is 6.95%.

Figure 1. Multispectral immunofluorescence stains of diffuse large B-cell lymphoma performed on the Vectra system. (A) High T-cell proportion versus (B) low T-cell proportion. (C) An example of a case with high CD8+PD1+ T cells and (D) a comparison case with low CD8+PD1+ T cells.

Table 1. Clinical characteristics of patients with diffuse large B-cell lymphoma according to T-cell proportion.

standard therapy, we investigated whether DHITsig-pos cases exhibited a low T-cell proportion.15,17 However, we found that the DHITsig-pos cases (n=8) had a median CD3+ T-cell percentage of 19% compared to 16% for the DHITsig-neg cases (n=32) (P=0.95) (Online Supplementary Figure S4), although this may be due to the low number of DHITsig-pos cases analyzed.

We also evaluated the proportion of PD-L1 expression on the histiocytes and tumor cells, and we see a moderate positive correlation with total T-cell proportion and PD-L1positive histiocytes (r=0.40, 95%CI: 0.16 to 0.60; P=0.002) but no correlation was seen with PD-L1-positive tumor cells (r=0.17, 95%CI: -0.098 to 0.41; P=0.21). A low correlation was seen with CD4+PD1+ T cells and PD-L1-positive histiocytes (r=0.28, 95%CI: 0.023 to 0.51; P=0.033) and PD-L1-positive tumor cells (r=0.28, 95%CI: 0.023 to 0.51; P=0.034).

In addition, nearest neighbor analysis was performed to determine if a particular cell type in the TME (e.g., CD4+ or CD8+ T cells, histiocytes, NK cells) within close proximity to the malignant B cells correlated with survival. Just as a lower proportion of T cells correlated with poor survival, we found that low numbers of CD3+CD4+PD1+ T cells and CD3+CD8+PD1+ T cells from the tumor cells (50 pixels) were both associated with poor OS (P=0.0095 and P=0.0022, respectively) (Online Supplementary Figure S5).

Mutation and copy number analysis

We also evaluated the 57 cases by mutational and copy number analysis. We found that cases with a low total Tcell proportion had more frequent abnormalities of immune-related genes such as B2M, TNFRSF14, CD58, and FAS (12/15, 80%) compared to those with high total T-cell

Figure 2. Overall survival and progression-free survival of diffuse large B-cell lymphoma patients based on T-cell proportions. (A and B) Overall T-cell proportion (low/high cutoff 6.95%; median 17%). (C and D) CD4+PD1+ T cells (low/high cutoff 2.55%; median 1.2%). (E and F) CD8+PD1+ T cells (low/high cutoff 1.2%; median 4.5%).

proportion (15/42, 36%; P=0.006) (Figure 3). There was no correlation between PD-L1 copy number gain and T-cell proportion (P=0.49).

B2M, HLA I, HLA II, and TIM3 protein expression by immunohistochemistry

We found that the cases of DLBCL were positive for B2M, HLA I, and HLA II in 40%, 47%, and 33% by IHC, respectively (Figure 4). Loss of expression for both B2M and HLA I was frequently seen in DLBCL (50%, 29/57). Interestingly, cases that showed low T-cell proportion by mIF showed frequent loss of B2M, HLA I, and/or HLA II by IHC compared to cases that had a higher T-cell proportion (Online Supplementary Table S1). Cases that had loss of all three proteins (B2M, HLA I, and HLA II) were more frequent in the low T-cell proportion group compared to the high T-cell proportion group (73% vs . 26%; P =0.002). Median T-cell proportions were also lower in cases that were negative for B2M, HLA I, and HLA II (Online Supplementary Table S3, Online Supplementary Figure S6). We also saw a lower percentage of NK cells (CD56+) in cases with loss of HLA II compared to HLA II positive cases (0.04% vs. 0.18%; P=0.0007), but this was not seen with B2M or HLA I. Using a targeted mutational panel, we observed that cases with B2M IHC loss had a higher number of mutations compared to B2M-positive cases (median 12.5 mutations vs. 8 mutations; P=0.048), but this was not significant for HLA I/II. Among the total of 8 cases with EZH2, 5 cases lacked expression of B2M, HLA I and HLA II (63%), but there was no significant difference in median T-cell proportions between EZH2 mutated and EZH2 wildtype cases (10% vs . 17%, respectively; P =0.44). We also found that loss of expression of B2M or HLA I by IHC cor-

related with B2M mutations ( P =0.04 and P =0.013, respectively) but not with copy number loss ( P=0.45 and P=0.054, respectively).

The median percentage of cells expressing TIM3 was 2.2% (range, 0-65.5%) (Figure 3). We observed that cases with low T-cell content had a lower median proportion of cells with TIM3 expression (0.56%, 95%CI: 0.11-1.67%) compared to cases with high T-cell content (2.77%, 95%CI: 2.02-7.1%; P=0.0011). There was also a correlation between TIM3 expression and B2M (P=0.03) as well as HLA I ( P =0.013) expression, but not with HLA II expression (P=0.61). There was no association between TIM3 proportion and OS or PFS (ROC cutoff 2.6%; P=0.086 and P=0.46, respectively) (Figure 5A and B). But, when combining PD1 (CD4+ PD1 + T cells) proportion with TIM3 proportion, we observed that cases with both low TIM3 and low PD1 proportion were associated with poor OS and PFS compared to cases with a high proportion of TIM3 and/or PD1 (P<0.0001 and P=0.0002, respectively) (Figure 5C and D).

Validation of the tumor microenvironment findings with CIBERSORTx analysis

Using data from Lenz et al., 18 we analyzed the cellular proportions in 351 cases of de novo DLBCL (184 GCB DLBCL, 167 ABC DLBCL) treated with R-CHOP. We used the CIBERSORTx method19 to infer cell-type gene expression profiles and found that the signatures of cases with a low T-cell content and low PD1+ T cells were associated with a poor OS (P=0.012 and P=0.0005, respectively) (Figure 6A and B). Cases with a low expression of TIM3 (HAVCR2) and low PD1+ T cells were associated with a poor OS (P=0.0003) (Figure 6C). We also observed a trend within the COO subtypes (Figure 6D and E).

Figure 3. Mutational spectra of cases with low and high T-cell content. Total T cells: low (green), high (salmon). There are more abnormalities in genes related to immune surveillance in the cases with low T-cell content compared to cases with high T-cell content (80% vs. 36%; P=0.006). Cases with low T-cell content also showed loss of B2M, HLA I, and HLA II by IHC. COO: cell-oforigin; DHITsig: double hit signature; CN: copy number; abn: abnormal; UNC: unclassified; GCB: germinal center type; ABC: activated B-cell type; IHC: immunohistochemistry.

Validation of the tumor microenvironment findings using routine immunohistochemical stains and bioimage analysis

An independent cohort of 54 cases of DLBCL was also stained for CD3 and PD1 by IHC, scanned using digital whole slide imaging and analyzed using bioimage analysis software to determine CD3+ and PD1+ T-cell proportions (Figure 7A and B). We found that the total T-cell proportion was associated with poor OS and PFS (P=0.03 and P=0.045, respectively; ROC cutoff 4.26%) (Figure 7C and D), with similar trends for PD1 (ROC cutoff 18.6%) (Figure 7E and F). Staining of the original discovery cohort with chromogenic IHC of 49 cases of DLBCL showed good correlation between the CD3 proportion by mIF and by IHC (r=0.78, 95%CI: 0.63-0.87; P<0.0001) and still showed prognostic significance with OS (see Online Supplementary Figure S8).

Discussion

The TME plays a critical role in the pathobiology of solid tumors and lymphomas, particularly classical Hodgkin lymphoma3 where the TME cells outnumber the tumor cells. We found that “hot” or inflamed tumors by mIF, GEP, and IHC had good outcomes whereas “cold” tumors had poor outcome. These “cold” tumors had low proportions of T cells, particularly T cells with an exhausted phenotype expressing PD1 and TIM3. We also found genetic differences that were associated with these “hot” and “cold”

tumors. “Cold” tumors had more frequent abnormalities related to immune surveillance such as B2M and TNFRSF14, as well as loss of the surface proteins B2M, HLA I and II.

In our study, similar to others,14,20 we found that “hot” tumors, specifically those with increased T-cell proportion and expressing PD1 and TIM3, were associated with an improved prognosis. We sought to investigate the TME in DLBCL using a multiparameter approach incorporating mIF and routine IHC, in conjunction with genomic analysis, to elucidate the major cellular subsets in tissue sections and determine the mechanisms of immune escape. We have correlated our findings with the clinical outcome and have confirmed our findings using two independent validation cohorts that similarly demonstrated T-cell proportion to be associated with outcome. In our study, we found that “hot” DLBCL tumors with a high T-cell content, high CD4+PD1+ T cells, and high CD8+PD1+ T cells had better prognosis when treated with R-CHOP (Figure 2). We used standard IHC stains for CD3 and PD1 in an independent cohort of DLBCL treated with R-CHOP and found that CD3 proportion predicted for survival, with a trend for PD1 as well. In addition, we re-evaluated the original discovery cohort with chromogenic IHC and found good correlation between the CD3 mIF and IHC proportions with continued significance related to OS. Our study confirms the findings of others that T-cell proportion is associated with outcomes14,20 and is a proof of principle that routine IHC, in conjunction with bioimaging, may be a viable and lowcost option to determine T-cell proportions when more

Figure 4. Diffuse large B-cell lymphoma staining with B2M, HLA I, and HLA II immunohistochemistry. A case of diffuse large B-cell lymphoma (DLBCL) with expression of (A) B2M, (B) HLA I, and (C) HLA II by immunohistochemistry (IHC). (D-F) A separate case of DLBCL that is negative for (D) B2M, (E) HLA I, and (F) HLA II by IHC.

advanced technology such as mIF is not available. Combining TIM3 proportion with PD1 expression also showed that cases with both low PD1 and low TIM3 have a very poor OS and PFS, which was confirmed using deconvoluted GEP data19 from an independent cohort of DLBCL.18 In addition, cases with low T-cell content exhibited mechanisms of immune escape such as decreased expression of HLA proteins as well as genomic abnormalities related to immune surveillance.

TIM3 is an inhibitory receptor that is expressed on exhausted T cells, and co-blockade of TIM3 and PD1 has resulted in greater restoration of the T-cell response compared to PD1 alone in various cancers.21,22 TIM3 also marks the most dysfunctional subset among tumor-infiltrating CD8+PD1+ T cells.23 Roussel et al 20 recently found that tumor-infiltrating lymphocytes expressing PD1 and TIM3 are expanded in DLBCL, particularly ABC DLBCL, and exhibit a transcriptomic signature related to T-cell exhaustion. Their ABC DLBCL cases had a poor survival in cases with low expression of PD1 and TIM3 by gene expression profile.20 Similarly, our cases with low expression of PD1 and TIM3 were associated with poor OS (Figure 5),

and this was further confirmed with the GEP data from Lenz et al. 18 We also found that low/absent TIM3 expression with low CD4+PD1+ T cells was associated with the absence of B2M and HLA I proteins. It may be that cases with low PD1 and low TIM3 expression represent particularly “cold” tumors that utilize loss of class I HLA, thus escaping recognition by cytotoxic T cells.

In solid tumors, it is well known that increased tumor infiltrating lymphocytes (TIL) are associated with better survival.24 “Cold” tumors with a suppressive TME prevent the anti-tumor function of cytotoxic T cells. One suppressive mechanism used by the tumor cells is to down-regulate their HLA class I proteins due to genetic abnormalities (e.g., B2M or HLA I), thereby avoiding immune surveillance. There are also suppressive cytokines and other signals released by the tumor or the TME cells that cause the T cells to be in an ‘exhausted’ state. These exhausted T cells have high levels of inhibitory receptors such as PD1, LAG3, TIM-3, cytotoxic T lymphocyte antigen-4 (CTLA-4), band T lymphocyte attenuator (BTLA), and T-cell immunoglobulin and immunoreceptor tyrosine-based inhibitory motif domain (TIGIT). One possible reason why cases of DLBCL rich

Figure 5. Survival based on TIM3 and PD1 staining in diffuse large B-cell lymphoma. (A) Overall survival (OS) and (B) progressionfree survival (PFS) in relation to TIM3 proportion in the tumor microenvironment. Combining CD4+PD1+ T-cell proportion (PD1) with TIM3 identifies a group of patients with poor (C) OS and (D) PFS having low expression of these two markers compared to cases with high PD1 (PD1h) and/or high TIM3 (TIM3h).

in T cells, particularly high CD8+PD1+ T cells, respond to immunochemotherapy like R-CHOP may be that the therapy reduces the tumor burden, results in tissue damage, releases tumor antigens, and allows these T cells to recover their anti-tumor properties, similar to observations in patients with solid tumors.25 Moreover, R-CHOP therapy may re-invigorate the “hot”, inflamed tumors since anthracycline-based chemotherapy can induce immunogenic death of lymphoma cells.26 Based on our findings, it may be worthwhile studying immunotherapies or immunomodulating therapies prior to or in combination with chemotherapy specifically in patients who have tumors with low T-cell content in order to reinvigorate the TME and potentially allow for more robust responses to treatment. Likewise, “hot” or T-cell inflamed tumors may be predisposed to benefit from immunotherapies or combined chemotherapy with immunotherapy/immunomodulatory approaches.27,28 We sought to evaluate potential underlying genomic mechanisms for our observations regarding the DLBCL TME. We evaluated whether the cases that were high-

grade B-cell lymphomas (DHITsig-pos) were inadvertently selected for in the low T-cell group, as double-hit lymphoma are found to have a paucity of T cells.17 However, in our cases, we found no difference in the proportion of T cells between the DHITsig-pos and DHITsig-neg cases. This lack of association may be due to the low number of DHITsig-pos cases in the cohort. We did observe that cases of DLBCL with a low T-cell content had a higher frequency of genomic abnormalities related to immune surveillance genes such as (B2M, TNFRSF14, CD58, and FAS; 80% vs 36%) (Figure 3). These mutations likely resulted in a loss of the ability to present antigen by the tumor cells and their lack of recognition by T cells, resulting in fewer T cells homing into the TME. Recent studies29,30 have demonstrated that patients who progress following chimeric antigen receptor (CAR) T cells exhibit FAS mutations, suggesting that there may be common antigen-independent mechanisms of resistance to chemotherapy and immunotherapy approaches.

In our study, we found that many cases that had low T-

Figure 6. Validation of TIM3 and PD1 by gene expression profiling in an independent cohort of diffuse large B-cell lymphoma. Utilizing CIBERSORTx and the diffuse large B-cell lymphoma (DLBCL) dataset from Lenz et al., 18 we confirm that low total T-cell and low PD1+ T-cell proportions are associated with poor overall survival OS (A and B). Cases with low PD1+ T cells and TIM3 (HAVCR2 expression) had poor (OS) compared to cases with high TIM3 (TIM3h) and/or high PD1 T cells (PD1h) (C). A similar trend was also seen based on cell-of-origin for germinal center type (GCB) but not activated B-cell type (ABC) DLBCL (D and E).

Figure 7. Validation of total T-cell content and PD1 staining in an independent cohort of diffuse large B-cell lymphoma. Using single stain immunohistochemistry for CD3 and PD1, we scanned the slides and used bioimage software to objectively enumerate the positive cells: CD3 (red), tumor cells (blue). Low total CD3+ T-cell content (A) compared to high T-cell content (B) was associated with poor overall survival and progression-free survival (C and D) (ROC cutoff 4.26%), with similar trends for PD1 staining (E and F) (ROC cutoff 18.6%).

cell proportion and decreased CD4 and CD8 T cells also showed absence of B2M, HLA I, and/or HLA II proteins (Figure 4). A study by Rimsza et al. 31 similarly showed that cases of DLBCL with loss of HLA II protein also had exceedingly few CD8+ T cells. A recent study showed an association between B2M mutations and absence of HLA I expression in DLBCL, resulting in decreased presentation of neoantigens on the tumor surface and leading to evasion of immune surveillance.32 Similarly, we see in our study that cases with B2M mutations showed an association with loss of B2M and HLA I protein expression. Unlike a recent study,27 we did not find a correlation between PD-L1 copy number gains and T-cell proportion or association with inferior PFS. In that study, there was a high proportion of non-GCB DLBCL cases with high PD-L1 protein expression which correlates with PD-L1 alterations and inferior PFS. We may not have observed these findings due to the small number of cases of ABC DLBCL in our cohort, and we also did not evaluate for PD-L1 translocation in our study.

Ennishi et al. 33 described cases of GCB DLBCL with EZH2 mutations and showed lower expression of MHC II and a significantly lower number of TIL in the TME. There were 8 cases in our study with an EZH2 mutation, and 5 of these showed lack of B2M, HLA I, and HLA II protein expression (63%). Cases with mutations in EZH2 also showed lower Tcell content, but we had too few cases with EZH2 mutation to adequately evaluate this for statistical significance. Limitations to this study are that we used a TMA with representative cores rather than whole tissue sections. As a result, we cannot exclude the possibility that the cores may not completely represent the TME within the whole tumor. In addition, we validated our mIF findings using different approaches from mIF (GEP and routine IHC), but there were consistent findings across methodologies, which supports our hypothesis that “cold” tumors show a poor prognosis regardless of what method is used for analysis and that our findings are not method-dependent. It is unclear which method provides optimal evaluation of the T-cell proportion, but other studies have also shown correlation with mIF and deconvoluted GEP analysis,14,20 and we also saw good correlation between mIF and IHC in our discovery cohort.

References

1. Leivonen SK, Pollari M, Bruck O, et al. T-cell inflamed tumor microenvironment predicts favorable prognosis in primary testicular lymphoma. Haematologica. 2019;104(2):338-346.

2. Scott DW, Gascoyne RD. The tumour microenvironment in B cell lymphomas. Nat Rev Cancer. 2014;14(8):517-534.

3. Aoki T, Chong LC, Takata K, et al. Single-cell transcriptome analysis reveals disease defining T-cell subsets in the tumor microenvironment of classic Hodgkin lymphoma. Cancer Discov. 2020;10(3):406-421.

In conclusion, our findings demonstrate that analysis of the TME can play an important prognostic role in DLBCL, with “cold” tumors containing a low proportion of exhausted T cells with inhibitory molecules such as PD1 and TIM3 associated with a poor prognosis. We identified underlying genomic abnormalities associated with these low T-cell content DLBCL along with absence of HLA molecules (B2M, HLA I, and HLA II) on the tumor cells that allow for evasion of immune surveillance. These findings further our understanding of the role of the TME in DLBCL and the mechanisms that tumor cells utilize for immune escape.

Disclosures

No conflicts of interest to disclose.

Contributions

JYS and AFH are responsible for study conception and design. All authors collected and assembled the data. WZ, HM, VB, JC, PS, MRN, DS, PL, WCC, DDW and AMP performed data analysis. JYS prepared the first draft of the paper and all authors helped with revisions.

Acknowledgments

We thank Dr. George Wright and the members of the Lymphoma and Leukemia Profiling Project (LLMPP) for providing the gene expression profiling data. The content of this paper is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Funding

This research was supported by the City of Hope National Medical Center Department of Pathology and the Toni Stephenson Lymphoma Center. Research reported in this publication included work performed in the City of Hope Pathology Core and Integrated Genomics Core supported by the National Cancer Institute of the National Institutes of Health under grant number P30CA033572 and the City of Hope Lymphoma SPORE under grant number P50CA107399.

Data-sharing agreement

Original data are available from the corresponding author.

4. Chen BJ, Dashnamoorthy R, Galera P, et al. The immune checkpoint molecules PD-1, PD-L1, TIM-3 and LAG-3 in diffuse large B-cell lymphoma. Oncotarget. 2019;10(21):2030-2040.

5. Wright GW, Huang DW, Phelan JD, et al. A probabilistic classification tool for genetic subtypes of diffuse large B-cell lymphoma with therapeutic implications. Cancer Cell. 2020;37(4):551-568.e14.

6. Chapuy B, Stewart C, Dunford AJ, et al. Molecular subtypes of diffuse large B cell lymphoma are associated with distinct

pathogenic mechanisms and outcomes. Nat Med. 2018;24(5):679-690.

7. Lacy SE, Barrans SL, Beer PA, et al. Targeted sequencing in DLBCL, molecular subtypes, and outcomes: a Haematological Malignancy Research Network report. Blood. 2020;135(20):1759-1771.

8. Ansell SM, Stenson M, Habermann TM, Jelinek DF, Witzig TE. CD4+ T-cell immune response to large B-cell non-Hodgkin's lymphoma predicts patient outcome. J Clin Oncol. 2001;19(3):720-726.

9. Keane C, Gill D, Vari F, Cross D, Griffiths L, Gandhi M. CD4(+) tumor infiltrating lymphocytes are prognostic and independent of R-IPI in patients with DLBCL receiving R-CHOP chemoimmunotherapy. Am J Hematol. 2013;88(4):273-276.

10. Coutinho R, Clear AJ, Mazzola E, et al. Revisiting the immune microenvironment of diffuse large B-cell lymphoma using a tissue microarray and immunohistochemistry: robust semiautomated analysis reveals CD3 and FoxP3 as potential predictors of response to R-CHOP. Haematologica. 2015;100(3):363-369.

11. Shi Y, Deng L, Song Y, et al. CD3+/CD8+ T-cell density and tumoral PD-L1 predict survival irrespective of rituximab treatment in Chinese diffuse large B-cell lymphoma patients. Int J Hematol. 2018;108(3):254-266.

12. Li L, Sun R, Miao Y, et al. PD-1/PD-L1 expression and interaction by automated quantitative immunofluorescent analysis show adverse prognostic impact in patients with diffuse large B-cell lymphoma having T-cell infiltration: a study from the International DLBCL Consortium Program. Mod Pathol. 2019;32(6):741-754.

13. Xu-Monette ZY, Xiao M, Au Q, et al. Immune profiling and quantitative analysis decipher the clinical role of immunecheckpoint expression in the tumor immune microenvironment of DLBCL. Cancer Immunol Res. 2019;7(4):644-657.

14. Autio M, Leivonen SK, Bruck O, Karjalainen-Lindsberg ML, Pellinen T, Leppa S. Clinical impact of immune cells and their spatial interactions in diffuse large B-cell lymphoma microenvironment. Clin Cancer Res. 2022;28(4):781-792.

15. Song JY, Perry AM, Herrera AF, et al. Double-hit signature with TP53 abnormalities predicts poor survival in patients with germinal center type diffuse large B-cell lymphoma treated with R-CHOP. Clin Cancer Res. 2021;27(6):1671-1680.

16. Scott DW, Wright GW, Williams PM, et al. Determining cell-oforigin subtypes of diffuse large B-cell lymphoma using gene expression in formalin-fixed paraffin-embedded tissue. Blood. 2014;123(8):1214-1217.

17. Ennishi D, Jiang A, Boyle M, et al. Double-hit gene expression signature defines a distinct subgroup of germinal center B-celllike diffuse large B-cell lymphoma. J Clin Oncol. 2019;37(3):190-201.

18. Lenz G, Wright G, Dave SS, et al. Stromal gene signatures in large-B-cell lymphomas. N Engl J Med. 2008;359(22):2313-2323.

19. Newman AM, Steen CB, Liu CL, et al. Determining cell type abundance and expression from bulk tissues with digital cytometry. Nat Biotechnol. 2019;37(7):773-782.

20. Roussel M, Le KS, Granier C, et al. Functional characterization of PD1+TIM3+ tumor-infiltrating T cells in DLBCL and effects of PD1 or TIM3 blockade. Blood Adv. 2021;5(7):1816-1829.

21. Fourcade J, Sun Z, Pagliano O, et al. PD-1 and Tim-3 regulate the expansion of tumor antigen-specific CD8(+) T cells induced by melanoma vaccines. Cancer Res. 2014;74(4):1045-1055.

22. Sakuishi K, Apetoh L, Sullivan JM, Blazar BR, Kuchroo VK, Anderson AC. Targeting Tim-3 and PD-1 pathways to reverse T cell exhaustion and restore anti-tumor immunity. J Exp Med. 2010;207(10):2187-2194.

23. Wolf Y, Anderson AC, Kuchroo VK. TIM3 comes of age as an inhibitory receptor. Nat Rev Immunol. 2020;20(3):173-185.

24. Gooden MJ, de Bock GH, Leffers N, Daemen T, Nijman HW. The prognostic influence of tumour-infiltrating lymphocytes in cancer: a systematic review with meta-analysis. Br J Cancer. 2011;105(1):93-103.

25. Jiang W, He Y, He W, et al. Exhausted CD8+T cells in the tumor immune microenvironment: new pathways to therapy. Front Immunol. 2020;11:622509.

26. Fucikova J, Kralikova P, Fialova A, et al. Human tumor cells killed by anthracyclines induce a tumor-specific immune response. Cancer Res. 2011;71(14):4821-4833.

27. Godfrey J, Tumuluru S, Bao R, et al. PD-L1 gene alterations identify a subset of diffuse large B-cell lymphoma harboring a T-cell-inflamed phenotype. Blood. 2019;133(21):2279-2290.

28. Ansell SM, Minnema MC, Johnson P, et al. Nivolumab for relapsed/refractory diffuse large B-cell lymphoma in patients ineligible for or having failed autologous transplantation: a single-arm, phase II study. J Clin Oncol. 2019;37(6):481-489.

29. Jain MD, Ziccheddu B, Coughlin CA, et al. Whole-genome sequencing reveals complex genomic features underlying antiCD19 CAR T-cell treatment failures in lymphoma. Blood. 2022;140(5):491-503.

30. Upadhyay R, Boiarsky JA, Pantsulaia G, et al. A critical role for Fas-mediated off-target tumor killing in T-cell immunotherapy. Cancer Discov. 2021;11(3):599-613.

31. Rimsza LM, Roberts RA, Miller TP, et al. Loss of MHC class II gene and protein expression in diffuse large B-cell lymphoma is related to decreased tumor immunosurveillance and poor patient survival regardless of other prognostic factors: a followup study from the Leukemia and Lymphoma Molecular Profiling Project. Blood. 2004;103(11):4251-4258.

32. Fangazio M, Ladewig E, Gomez K, et al. Genetic mechanisms of HLA-I loss and immune escape in diffuse large B cell lymphoma. Proc Natl Acad Sci U S A. 2021;118(22):e2104504118.

33. Ennishi D, Takata K, Beguelin W, et al. Molecular and genetic characterization of MHC deficiency identifies EZH2 as therapeutic target for enhancing immune recognition. Cancer Discov. 2019;9(4):546-563.

Integrated genetic and clinical prognostic factors for aggressive adult T-cell leukemia/lymphoma

Takuro Kameda,1 Keisuke Kataoka,2,3+ Ayako Kamiunten,1 Michihiro Hidaka,4 Hiroaki Miyoshi,5

Nobuaki Nakano,6 Kisato Nosaka,7 Makoto Yoshimitsu,8 Jun-ichirou Yasunaga,7,9 Yasunori

Kogure,3 Kotaro Shide,1 Masaharu Miyahara,10 Takashi Sakamoto,11 Keiichi Akizuki,1 Tomonori

Hidaka,1 Yoko Kubuki,1 Junji Koya,3 Noriaki Kawano,12 Kiyoshi Yamashita,12 Hiroshi Kawano,1,3

Takanori Toyama,14 Kouichi Maeda,15 Kosuke Marutsuka,16 Yoshitaka Imaizumi,17 Koji Kato,18

Takeshi Sugio,18 Masahito Tokunaga,6 Yukie Tashiro,19 Akifumi Takaori-Kondo,11 Yasushi

Miyazaki,17 Koichi Akashi,18 Kenji Ishitsuka,8 Masao Matsuoka,7 Koichi Ohshima,5 Toshiki

Watanabe,20+ Akira Kitanaka,21 Atae Utsunomiya,6 Seishi Ogawa22+ and Kazuya Shimoda1+

1Division of Hematology, Diabetes, and Endocrinology, Department of Internal Medicine, Faculty of Medicine, University of Miyazaki, Miyazaki; 2Division of Hematology, Department of Medicine, Keio University School of Medicine, Tokyo; 3Division of Molecular Oncology, National Cancer Center Research Institute, Tokyo; 4National Hospital Organization Kumamoto Medical Center, Kumamoto; 5Department of Pathology, Kurume University School of Medicine, Kurume; 6Department of Hematology, Imamura General Hospital, Kagoshima; 7Department of Hematology, Rheumatology and Infectious Diseases, Faculty of Life Sciences, Kumamoto University of Medicine, Kumamoto; 8Department of Hematology and Rheumatology, Kagoshima University Hospital, Kagoshima; 9Institute for Frontier Life and Medical Sciences, Kyoto University, Kyoto; 10Department of Internal Medicine, Karatsu Red Cross Hospital, Saga; 11Department of Hematology and Oncology, Graduate School of Medicine, Kyoto University, Kyoto; 12Department of Internal Medicine, Miyazaki Prefectural Miyazaki Hospital, Miyazaki; 13Department of Internal Medicine, Koga General Hospital, Miyazaki; 14Department of Internal Medicine, Miyazaki Prefectural Nobeoka Hospital, Miyazaki; 15National Hospital Organization Miyakonojo Medical Center, Miyazaki; 16Department of Anatomic Pathology, Miyazaki Prefectural Miyazaki Hospital, Miyazaki; 17Department of Hematology, Atomic Bomb Disease and Hibakusha Medicine Unit, Atomic Bomb Disease Institute, Nagasaki University, Nagasaki; 18Department of Medicine and Biosystemic Science, Kyushu University Graduate School of Medical Sciences, Fukuoka; 19Department of Pathology, Imamura General Hospital, Kagoshima; 20Department of Practical Management of Medical Information, St Marianna University, Graduate School of Medicine, Tokyo; 21Department of Laboratory Medicine, Kawasaki Medical School, Kurashiki and 22Department of Pathology and Tumor Biology, Graduate School of Medicine, Kyoto University, Kyoto, Japan

+KK, SO and KS contributed equally to this work.

Abstract

Correspondence: K. Shimoda kshimoda@med.miyazaki-u.ac.jp

Received: June 6, 2022.

Accepted: February 7, 2023.

Early view: February 16, 2023.

https://doi.org/10.3324/haematol.2022.281510

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

The prognosis of aggressive adult T-cell leukemia/lymphoma (ATL) is poor, and allogeneic hematopoietic stem cell transplantation (allo-HSCT) is a curative treatment. In order to identify favorable prognostic patients after intensive chemotherapy, and who therefore might not require upfront allo-HSCT, we aimed to improve risk stratification of aggressive ATL patients aged <70 years. The clinical risk factors and genetic mutations were incorporated into risk modeling for overall survival (OS). We generated the m7-ATLPI, a clinicogenetic risk model for OS, that included the ATL prognostic index (PI) (ATL-PI) risk category, and non-silent mutations in seven genes, namely TP53, IRF4, RHOA, PRKCB, CARD11, CCR7, and GATA3. In the training cohort of 99 patients, the m7-ATLPI identified a low-, intermediate-, and highrisk group with 2-year OS of 100%, 43%, and 19%, respectively (hazard ratio [HR] =5.46; P<0.0001). The m7-ATLPI achieved superior risk stratification compared to the current ATL-PI (C-index 0.92 vs. 0.85, respectively). In the validation cohort of 84 patients, the m7-ATLPI defined low-, intermediate-, and high-risk groups with a 2-year OS of 81%, 30%, and 0%, respectively (HR=2.33; P=0.0094), and the model again outperformed the ATL-PI (C-index 0.72 vs. 0.70, respectively). The simplified m7-ATLPI, which is easier to use in clinical practice, achieved superior risk stratification compared to the ATLPI, as did the original m7-ATLPI; the simplified version was calculated by summing the following: high-risk ATL-PI category (+10), low-risk ATL-PI category ( 4), and non-silent mutations in TP53 (+4), IRF4 (+3), RHOA (+1), PRKCB (+1), CARD11 (+0.5), CCR7 ( 2), and GATA3 ( 3).

Introduction

Adult T-cell leukemia/lymphoma (ATL) is an aggressive peripheral T-cell neoplasm characterized by the clonal proliferation of human T-cell leukemia virus type 1 (HTLV1)-infected T cells.1 ATL is classi fi ed into four subtypes, namely the acute, lymphoma, chronic, and smoldering types.2 Acute, lymphoma, and chronic types with one or more unfavorable prognostic factors (high lactate dehydrogenase, high blood urea nitrogen, and low albumin) are defined as aggressive ATL.3 The current frontline treatments for aggressive ATL vary by geographic location.4 In the US and Europe, antiviral therapy using zidovudine and interferon α (IFN-α) is the standard treatment for leukemic-type ATL. In Europe, chemotherapy is the first-line therapy for lymphoma-type ATL. In Japan, the current standard therapy for aggressive ATL is combination chemotherapy,5,6 and Katsuya et al. and we reported that the median overall survival (OS) was less than 1 year for aggressive ATL.6,7

Allogeneic hematopoietic stem cell transplantation (alloHSCT) is a curative treatment option for ATL, and 30–40% of patients who underwent allo-HSCT exhibited longer survival.8-10 Since 10–20% of aggressive ATL patients survived longer even after chemotherapy alone, a method of predicting the optimal treatment for each patient is keenly anticipated.11

Several ATL prognostic models have been reported. All are based on clinical and basic laboratory findings. The acuteand lymphoma-type ATL prognostic index (ATL-PI), which is the most widely used prognostic model, includes stage, Eastern Cooperative Oncology Group performance status (ECOG PS), age, albumin, and soluble interleukin-2 receptor (sIL-2R) level,12 while the Japan Clinical Oncology Group prognostic index (JCOG-PI) includes corrected calcium (cCa) level and ECOG PS.11 The modified ATL-PI, which includes acute type, poor PS, high sIL-2R level, high cCa level, and high C-reactive protein (CRP) level, has been proposed for patients aged ≤70 years who have aggressive ATL and who are candidates for allo-HSCT.13

We previously reported the landscape of gene mutations in ATL.14-16 Among 50 significantly mutated genes in ATL, 13 of which were affected in more than 10% of patients, and many influenced T-cell receptor (TCR)–NF-κB signaling.14 In order to select aggressive ATL patients who achieve longer survival with intensive chemotherapy, we clarified the effect of gene mutations on their prognosis and developed a clinicogenetic prognostic model for aggressive ATL by integrating gene mutation status and clinical parameters.

Methods

Study design and participants

We conducted an analysis of gene mutations and clinical risk

factors in two independent cohorts of patients aged <70 years with aggressive ATL who had ATL cells or tumor biopsy specimens before starting chemotherapy. Details regarding diagnostic criteria, methods for collecting tumor DNA and clinical data, and methods for targeted capture sequencing are described in the Online Supplementary Appendix and Online Supplementary Table S1. The training cohort was derived from patients in University of Miyazaki Hospital and Imamura General Hospital (Online Supplementary Figure S1). Aggressive-type ATL samples were obtained from 43 patients aged <70 years who were diagnosed from 2014 to 2017 and who received standard chemotherapy. We previously reported the gene mutation status of 117 ATL patients who were diagnosed from 2003 to 2014 at these two institutions.17 Among them, 83 had aggressive-type ATL. In order to establish the training cohort, we first excluded 26 patients aged ≥70 years old, five patients with serious comorbidities (solid cancer or liver cirrhosis), and 11 patients who did not undergo intensive chemotherapy. This cohort ultimately included 41 patients with aggressive-type ATL who were aged <70 years and who received standard chemotherapy. We obtained DNA samples of formalin-fixed, paraffin-embedded lymph nodes from 20 patients in the training cohort who were diagnosed from 2000 to 2015. The validation cohort comprised two groups: 58 patients diagnosed from 1994 to 2017 at four institutions, whose mutational status was previously reported;17 and 52 patients from six institutions who were diagnosed from 2006 to 2017 and who underwent intensive chemotherapy. DNA samples were obtained from the second group (52 patients) for analysis in this study. This study was approved by the Research Ethics Committee of the Faculty of Medicine, University of Miyazaki, and those of other participating institutes in accordance with the Helsinki Declaration.

Statistical analysis

The detailed statistical analysis methods are described in the Online Supplementary Appendix. Briefly, we used elastic-net regression with a Cox proportional hazards model for feature selection.18 This elastic-net model was trained only on the training set, and was then applied to the independent validation set to obtain predictions. When developing the clinicogenetic risk models, patients who received mogamulizumab therapy or underwent allo-HSCT were censored on the day of mogamulizumab administration or allo-HSCT, respectively, to reduce the impact of these therapies on OS. A simplified PI, which is easier to use in clinical practice, was then developed based on the simplified β-coefficients of the original model.

Results

Patient characteristics

In this study of patients aged <70 years with aggressive

ATL who received intensive chemotherapy, 99 patients were included in a training cohort to develop the clinicogenetic risk model, and 84 different patients were included in the validation cohort ( Online Supplementary Figure S1 ). The characteristics of the training and validation cohorts are listed in Table 1.

In the training cohort, the median age was 61.0 years (interquartile range [IQR], 55.0–65.5) and 55 patients (56%) were male. Thirty-four patients (34%) received the VCAP–AMP–VECP regimen,5 and the remaining 65 patients (66%) received the CHOP-like regimen. Mogamulizumab was administered to 28 patients (28%), and 41 (41%) underwent allo-HSCT. The 2-year OS in the training cohort was 44.5% (95% confidence interval (CI): 31.1–63.5) (Online Supplementary Figure S2). According to the ATL-PI, 38, 50, and 11 patients were categorized into low, intermediate-, and high-risk groups, respectively, and their 2-year OS was 62%, 40%, and 21%, respectively.

The median number of mutated genes in targeted sequencing in each patient was five (IQR, 4–7) (Online Supplementary Figure S3; Online Supplementary Table S2). Thirteen genes had non-silent mutations in at least 15%

of patients (Figure 1A; Online Supplementary Figure S4). Overall, 77 (78%) ATL patients harbored one or more nonsilent mutations in genes coding for NF-κB activation molecules downstream of the TCR, namely PLCG1, PRKCB, CARD11, IRF4, VAV1, STAT3, and NFKB1A. Furthermore, 54 patients (55%) had non-silent mutations in one or both of CCR4 and CCR7. Additional mutations were observed in TP53, GATA3, RHOA, and FAS. The frequencies of all mutation types (transitions, transversions, and insertions/deletions) in the 13 genes are shown in the Online Supplementary Figure S5. A subset of mutations was clustered at known hotspots (Online Supplementary Figure S6). Mutations in PRKCB and IRF4 were highly clustered in the catalytic domain and in the DNA binding domain, respectively. In both CCR4 and CCR7, almost all mutations occurred in the cytoplasmic regions.

The effect of gene mutations on clinical parameters and survival

Components of the ATL-PI and other clinical parameters were associated with specific gene mutations in univariate analysis; however, none of these associations were

IQR: interquartile range; ECOG PS: Eastern Cooperative Oncology Group performance status; Alb: albumin; cCa: corrected calcium; CRP: Creactive protein; sIL-2R: soluble interleukin-2 receptor; ATL-PI: adult T-cell leukemia/lymphoma prognostic index, Ctx: chemotherapy; IQR: interquartile range; P: P value between the training and validation cohorts.

Table 1. Patient and disease characteristics.

Figure 1. The m7-ATLPI clinicogenetic risk model. (A) Mutation frequencies of recurrent mutant genes in adult Tcell leukemia/lymphoma (ATL) in training and validation cohorts. P values were determined by Fisher’s exact test without correction for multiple testing. Detailed mutation plots for both cohorts are shown in the Online Supplementary Table S2 and the Online Supplementary Figure S6. (B) The m7-adult T-cell leukemia/lymphoma prognostic index (ATLPI) is calculated as the sum of individual clinical and gene mutation predictor values weighted by their individual coefficients. (C) Kaplan-Meier overall survival (OS) curves for the training cohort by ATL-PI and by m7-ATLPI. (D) Kaplan-Meier OS curves for the validation cohort by ATL-PI and by m7-ATLPI. Numbers in parentheses show number of patients with events/number of patients per cohort. ATL-PI low/int/high: low-, intermediate-, or high-risk ATLPI; m7-ATLPI low/int/high: low-, intermediate-, or high-risk m7-ATLPI.

statistically significant after correction for multiple testing (Online Supplementary Table S3). Among them, VAV1 mutations were associated with albumin <3.5 mg/dL, cCa ≥11 mg/dL, and CRP >2.5 mg/dL, and subsequently with high or intermediate risk on the ATL-PI after correction for multiple testing.

Univariate analysis was performed to identify potentially significant correlations between shorter survival and mutations in each gene (Online Supplementary Figure S7). In our previous report,17 PLCG1 mutations were associated with shorter survival in aggressive ATL patients, but they had no influence in the training cohort. TP53 mutations alone were not associated with inferior OS (HR=1.92; P=0.077), but did show an association after adjustment for the ATL-PI (HR=2.11; P=0.044).

We next performed an exploratory correlation matrix analysis to determine whether populations with statistically cooccurring combinations of genetic mutations existed, and if so, whether these populations formed independent prognostic groups (Online Supplementary Figure S8). We identified one positive weak correlation between CCR7 and NOTCH1 mutations, with a false discovery rate (FDR) q<0.05, and this combination of mutations was observed in seven (7%) of 99 patients. However, this combination did not influence survival (Online Supplementary Figure S9).

Development of a clinicogenetic risk model for aggressive adult T-cell leukemia/lymphoma

Using 13 recurrent gene mutations with a mutation frequency of 15% or higher, along with the score of the clinically based ATL-PI model, we generated a clinicogenetic risk model for OS using elastic-net penalized Cox regression. This clinicogenetic model, which we termed the m7ATLPI, was calculated as the sum of predictor values weighted by elastic-net coefficients, and included the high-risk ATL-PI category (βelastic net = +1.1), low-risk ATL-PI category (-0.36), and non-silent mutations in seven genes: TP53 (+0.41), IRF4 (+0.26), RHOA (+0.11), PRKCB (+0.079), CARD11 (+0.056), CCR7 (-0.20), and GATA3 (-0.25) (Figure 1B; Online Supplementary Table S4). Coefficients of six other genes were effectively shrunken and set to zero. Cutoffs of -0.36 and 0.41 were adopted to define low-, intermediate-, and high-risk groups by the m7-ATLPI. In the training cohort of 99 patients, the m7-ATLPI identified a low-risk group (17 [17%]) with 2-year OS of 100% (95% CI: 100–100), an intermediate-risk group (65 [66%]) with 2year OS of 43% (95% CI: 27–70%), and a high-risk group (17 [17%]) with 2-year OS of 19% (95% CI: 7–52%) (HR=5.46; 95% CI: 2.73–10.9; P<0.0001) (Figure 1C).

External validation

The validation cohort consisted of 84 patients from ten institutions (Online Supplementary Figure S1). Compared with the training cohort, more patients in the validation cohort

C index is a generalization of the area under the receiver operating characteristic curve for survival data and quantifies prognostic discrimination. ATL-PI: adult T-cell leukemia/lymphoma prognostic index; CI: confidence interval.

had a poor PS (17% vs. 44%, respectively; P<0.001) and received the VCAP-AMP-VECP regimen as first-line chemotherapy (34% vs. 60%, respectively; P=0.001) (Table 1). The two cohorts were similar regarding the percentages of patients determined to be low-risk by the ATL-PI (38% vs. 36%, respectively; P=0.82). The median number of gene mutations was five (IQR, 3–7), which did not differ from that of the training cohort (Online Supplementary Figure S3). The mutational landscape is summarized in the Online Supplementary Figure S4 . The frequencies of gene mutations did not differ between the training and validation cohorts except for RHOA and POT1, which demonstrated lower mutation frequencies in the validation cohort (Figure 1A). With the same cutoffs as those used in the training cohort, the m7-ATLPI applied to the validation cohort defined a low-risk group (13 [15%]) with 2year OS of 81% (95% CI: 60–100), an intermediate-risk group (57 [68%]) with 2-year OS of 30% (95% CI: 17–55), and a high-risk group (14 [17%]) with 2-year OS of 0% (95% CI: not available [na] - na) (HR=2.33; 95% CI: 1.23–4.41; P=0.0094) (Figure 1D).

Performance assessment of the m7-adult T-cell leukemia/lymphoma prognostic index

In order to quantify which prognostic model, the ATL-PI or m7-ATLPI, better fit the actual outcomes, we calculated C-indices. The C-index for the m7-ATLPI was 0.92 in the training cohort, which was better than that of 0.85 for the ATL-PI (Table 2). A similar result was obtained in the validation cohort, with C-indices of 0.72 and 0.70 for the m7-ATLPI and ATL-PI, respectively. Receiver operating characteristic (ROC) performance metrics for the m7ATLPI and ATL-PI based on OS at different time points in both cohorts are outlined in Table 3. The sensitivity and negative predictive value (NPV) for 2-year OS were superior for the m7-ATLPI relative to the ATLPI; that is, 100% and 100% in the training cohort, and 94% and 86%

Table 2. Improved concordance index with the m7-ATLPI and simplified m7-ATLPI.

in the validation cohort, respectively. This tendency was consistent with results of 1-year OS. These findings indicate that the m7-ATLPI has a high probability of correctly identifying patients with a good prognosis.

Reclassification of risk categories by the m7-adult Tcell leukemia/lymphoma prognostic index

In both cohorts, the improved performance by the m7ATLPI resulted mainly from reclassifying a subset of patients defined as low risk by the ATL-PI into the intermediate-risk m7-ATLPI category. This applied to 22 (58%) of 38 patients in the training cohort and 17 (57%) of 30 patients in the validation cohort, and their median sur-

vival time was 1.28 years (Figure 2A; Figure 3A). ATL cells from these patients harbored mutations in PRKCB, CARD11, IRF4, RHOA, or TP53 (Figure 2B, C). Patients defined as low risk by the ATL-PI and without TP53, IRF4, RHOA, PRKCB, or CARD11 mutations, or those with PRKCB or CARD11 mutations in addition to mutations in a gene with favorable impact (CCR7 or GATA3), exhibited a favorable prognosis with a median survival time of 7.55 years. Similarly, a subset of patients with intermediate ATL-PI scores was reclassified into the high-risk m7-ATLPI category, and their median survival time was 0.46 years (6 [12%] of 50 patients in the training cohort and 2 [4.8%] of 42 patients in the validation cohort) (Figures 2A and 3A). One ATL-PI in-

Sensitivity, specificity, positive predictive value, and negative predictive value for overall survival at different time points calculated by timedependent receiver operating characteristic analysis. These metrics were calculated based on a binalized index (high/intermediate-risk vs low-risk). ATL-PI: adult T-cell leukemia/lymphoma prognostic index; CI: confidence interval.

Table 3. Improved sensitivity and negative predictive value of overall survival at different time points with the m7-ATLPI and simplified m7-ATLPI.

termediate-risk patient was reclassified into the low-risk m7-ATLPI category.

Accordingly, the m7-ATLPI could identify which low-risk ATL-PI patients had favorable versus poor OS (Figure 3A). In the entire cohort, 68 (37%) of 183 patients were classified as low-risk ATL-PI, and they were reclassified by the m7-ATLPI into low- (29 [43%] of 68) and intermediate-risk (39 [57%] of 68) groups with 2-year OS of 89% and 35%, respectively (P=0.0078).

The m7-ATLPI could also segregate low-risk patients defined by other ATL prognostic models based on clinical parameters (Figure 3B, C). Among 114 patients defined as moderate risk by the JCOG-PI, the 2-year OS of 30 (26%) patients who were reclassified as low risk by the m7-ATLPI was 90%, whereas the 2-year OS of 81 (71%) and three (2.6%) patients who were reclassified as intermediate and high risk by m7-ATLPI was 45% and 0%, respectively (P<0.0001). As for 45 patients categorized as low risk by the modified ATL-PI, the 2-year OS of 16 (36%) and 29

(64%) patients who were reclassified into the low- and intermediate-risk groups by the m7-ATLPI was 83% and 49%, respectively, although their survival curves did not differ significantly (P=0.18).

The mutations in the seven genes included in the m7ATLPI also contributed to reclassification of patients categorized as intermediate risk by the modified ATL-PI (Figure 3C). Patients categorized as intermediate risk by the modified ATL-PI could be divided into low-, intermediate-, and high-risk groups by the m7-ATLPI, with 2-year OS of 100%, 37%, and 0%, respectively (P=0.0044).

Copy number alteration

We previously reported that a high number of copy number alterartions (CNA) was a characteristic of aggressive-type ATL.17 Among the seven genes that affect ATL prognosis, IRF4, TP53, CARD11, and GATA3 frequently have CNA. Like somatic mutations, CNA in these driver genes may influence ATL prognosis. In this study, data concerning CNA in IRF4,

Figure 2. Reclassification of risk categories by m7-ATLPI. (A) Transitions in the distribution of risk categories following the change from the adult T-cell leukemia/lymphoma prognostic index (ATL-PI) to the m7-ATLPI, for both the training and validation cohorts. (B) m7-ATLPI scores for patients in the training and validation cohorts, along with clinical predictors (ATL-PI categories) and molecular predictors. Boxes indicate high- or low-risk ATL-PI categories, or mutations in the indicated genes, and the color codes indicate the coefficients of the individual m7-ATLPI predictors. The corresponding Kaplan-Meier overall survival (OS) curves for patients classified as low-risk by the ATL-PI and then reclassified by the m7-ATLPI are shown in Figure 3A. (C) Relative frequencies of molecular predictors by m7-ATLPI category in low-risk ATL-PI patients in the training and validation cohorts.

TP53, CARD11, and GATA3 were available from only 83 of the 183 total samples; specifically, CNA of these genes were detected in 15, 8, 9, and 16 samples, respectively (Online Supplementary Figure S10A). The eight patients with CNA in TP53 exhibited worse prognosis than patients without these CNA, and their risk categories as defined by the m7-ATLPI were high and intermediate in three and five patients, respectively (Online Supplementary Figure S10A, B). Six, one,

and one of these patients harbored PRKCB, CARD11, and TP53 mutations, respectively. CNA in IRF4, CARD11, and GATA3 had no effect on OS. In addition, CNA in 9p24 (PDL1), which had an adverse effect on OS in our previous report,17 had no effect in this cohort.

Overall, 28, 17, 25, and 23 patients had mutations and/or CNA in IRF4, TP53, CARD11, and GATA3, respectively. Among these 83 patients, the risk category defined by the m7-ATLPI

Figure 3. m7-ATLPI segregates patients with favorable prognosis from among low- or intermediate-risk patients defined by ATL prognostic models based on clinical parameters. (A) Kaplan-Meier overall survival (OS) curves stratified by the m7-adult T-cell leukemia/lymphoma prognostic index (m7-ATLPI) for patients in the entire cohort (training and validation cohorts) who were classified as low risk or intermediate risk by the ATL-PI. (B) Kaplan-Meier OS curves stratified by the m7-ATLPI for patients in the entire cohort who were classified as moderate risk by the Japan Clinical Oncology Group prognostic index. (C) Kaplan-Meier OS curves stratified by the m7-ATLPI for patients in the entire cohort who were classified as low risk or intermediate risk by the modified ATL-PI. Numbers in parentheses show the number of patients with events/number of patients per cohort.

changed for nine patients when CNA were considered as risk factors: from low to intermediate risk in one patient, from intermediate to low risk in three patients, and from intermediate to high risk in five patients. Nonetheless, adding CNA data to the m7-ATLPI did not enhance the riskstratification potential of the model compared with the use of mutation data alone (C-index: 0.88 vs. 0.88, respectively) (Online Supplementary Figure S10C).

Progression-free survival prediction by m7-adult T-cell leukemia/lymphoma prognostic index

The m7-ATLPI was developed to predict OS in patients with aggressive ATL who were treated with standard intensive chemotherapy. In clinical practice in Japan, allo-HSCT is performed shortly after the diagnosis of ATL in eligible patients.10 Sixty-seven patients in this study underwent alloHSCT at a median of 164 days (IQR, 121–220) after diagnosis.

Figure 4. The m7-ATLPI provides better risk stratification of progression-free survival than the ATL-PI. (A) Kaplan-Meier progression-free survival (PFS) curves by the adult T-cell leukemia/lymphoma prognostic index (ATL-PI) and m7-ATLPI for the entire cohort (n=183). (B) Kaplan-Meier PFS curves by the ATL-PI and m7-ATLPI for patients without allogenic hematopoietic stem cell transplantation (allo-HSCT) in the entire cohort (n=116). (C) Kaplan-Meier PFS curves by the ATL-PI and m7-ATLPI for patients with neither allo-HSCT nor mogamulizumab therapy in the entire cohort (n=85).

Figure 5. The simpli fi ed m7-ATLPI has comparable prognostic ability as the original m7-ATLPI. (A)

Scoring of the simpli fi ed m 7-adult T-cell leukemia/lymphoma prognostic index (m7-ATLPI). In order to create the simpli fi ed m 7-ATLPI, the coef fi cients of the ATL-PI categories and mutation status of the 7 genes were converted to numbers re fl ecting their original values. After summing these values, scores ≤− 4, 3.5 to 4.5, and ≥ 5 points were categorized as low-, intermediate-, and high-risk simpli fi ed m 7-ATLPI, respectively.

(B) Kaplan-Meier overall survival (OS) curves for the training cohort using the ATL-PI and simpli fi ed m 7-ATLPI. (C) Kaplan-Meier OS curves for the validation cohort using the ATL-PI and simpli fi ed m 7-ATLPI. Numbers in parentheses show the number of patients with

events/number of patients per cohort. ATL-PI low/int/high: low-, intermediate-, or high-risk ATL-PI; Simpli fi ed m 7-ATLPI low/int/high: low-, intermediate-, or high-risk simpli fi ed m7-ATLPI.

C

Low-risk patients defined by the m7-ATLPI were more likely to undergo allo-HSCT: 20 of 30 low-risk patients, 41 of 122 intermediate-risk patients, and six of 31 high-risk patients. The disease status at transplantation in the 20 low-risk, 41 intermediate-risk, and six high-risk patients was complete response (CR)/partial response (PR) in 13, 25, and six, respectively, and stable disease (SD)/progressive disease (PD) in seven, 16, and zero, respectively. Eight of 44 patients who underwent allo-HSCT in CR/PR died of regimen-related toxicities rather than of ATL progression. In order to accurately determine the clinical course of ATL treated with chemotherapy, we performed censoring at allo-HSCT. When patient survival was not censored by allo-HSCT or mogamulizumab therapy, the m7-ATLPI identified a low-risk group with 2-year OS of 57% (95% CI: 41–79), an intermediate-risk group with 2-year OS of 37% (95% CI: 28–47), and a high-risk group with 2-year OS 6.7% (95% CI: 1.8–26) (Online Supplementary Figure S11A). Analyses of patients who did not undergo allo-HSCT and those who were treated only with chemotherapy and not with mogamulizumab or alloHSCT also showed superior risk stratification compared to the current ATL-PI (C-index 0.78 vs. 0.73, and 0.80 vs. 0.75, respectively) (Online Supplementary Figure S11B, C).

Progression-free survival (PFS) might be suitable as a means of evaluating the sensitivity to chemotherapy in ATL. In order to analyze this question, patients who underwent allo-HSCT or mogamulizumab therapy in CR/PR were censored at the time of allo-HSCT or the administration of mogamulizumab, respectively (Figure 4A). In the low-risk group defined by the m7-ATLPI, the median PFS of 2.0 years was longer than that of 0.92 years in the corresponding group defined by the ATL-PI, and the m7-ATLPI achieved superior risk stratification compared to the current ATL-PI (C-index: 0.67 vs . 0.63, respectively). These results suggest that the m7-ATLPI predicts sensitivity to chemotherapy. When the analysis was restricted to patients without allo-HSCT (n=116) or to patients with neither allo-HSCT nor mogamulizumab therapy (n=85), the m7-ATLPI exhibited superior risk stratification compared to the ATL-PI (C-index 0.70 vs . 0.68, and 0.73 vs . 0.68, respectively) (Figure 4B, C).

The effect of gene mutations on patients treated with mogamulizumab or allo-hematopoietic stem cell transplantation

In addition to chemotherapy, widely used treatments for aggressive-type ATL include IFN-α, zidovudine, mogamulizumab, and brentuximab vedotin. Since no patients in our cohort received first-line therapy with any of these agents, it follows that m7-ATLPI could be applied only for patients who received conventional intensive chemotherapy. Even though mogamulizumab was not administered as first-line therapy, 49 patients were treated with mogamulizumab as second- or later-line therapy. Since

three and eight patients received mogamulizumab therapy in CR/PR status after first-line chemotherapy or alloHSCT, respectively, 38 patients received mogamulizumab therapy in chemotherapy-resistant status at any time during their clinical course. We tested whether determining the prognosis of the 38 patients treated with mogamulizumab in chemotherapy-resistant status was impacted by considering the mutations in the seven genes included in the m7-ATLPI ( Online Supplementary Figure S12A ). The m7-ATLPI could not determine which patients treated with mogamulizumab in chemotherapyresistant status had a favorable prognosis.

Sakamoto et al. reported that patients with CCR4 mutations who received mogamulizumab-based treatment and not allo-HSCT had a favorable prognosis.19 Of 49 patients who underwent mogamulizumab therapy, 27 had CCR4 mutations and 22 did not. Survival time after mogamulizumab therapy was comparable regardless of CCR4 mutation status: 0.53 and 0.31 years in patients with and without CCR4 mutations, respectively ( P =0.85) ( Online Supplementary Figure S13A). When the analysis was restricted to patients who did not undergo allo-HSCT, the survival time after mogamulizumab therapy was comparable between patients with and without CCR4 mutations ( Online Supplementary Figure S13B). In our study, CCR4 mutations had little effect on the prognosis of patients treated with mogamulizumab.

Among the entire cohort, 67 patients underwent alloHSCT. The survival curves of the three risk groups defined by the m7-ATLPI almost overlapped, indicating that the m7-ATLPI had no ability to predict the prognosis of ATL patients after allo-HSCT ( Online Supplementary Figure S12B).

Simplified m7-adult T-cell leukemia/lymphoma prognostic index

We simplified the original m7-ATLPI to make the instrument easier to use in clinical practice. We based this simpli fi ed version on the sums of the weights of the following variables obtained from the ATL-PI risk category and the mutational statuses of seven genes: high-risk ATL-PI category (+10), low-risk ATL-PI category ( 4), and non-silent mutations in TP53 (+4), IRF4 (+3), RHOA (+1), PRKCB (+1), CARD11 (+0.5), CCR7 ( 2), and GATA3 ( 3) Scores ≤−4, from 3.5 to 4.5, and ≥5 were used to define the low-, intermediate-, and high-risk categories, respectively (Figure 5). In the training cohort, one patient classified as intermediate risk and one classified as high risk according to the m7-ATLPI were reclassified as low risk and intermediate risk according to the simplified m7ATLPI, respectively. The simplified m7-ATLPI was then applied to the validation cohort. The classi fi cation in the validation cohort showed high concordance with that in the original m7-ATLPI (weighted κ, 1.00), and no patients

were reclassified by the simplified m7-ATLPI. The C-index for the simpli fi ed m7-ATLPI was 0.92 and 0.72 in the training and validation cohort, respectively, which was same as that of 0.92 and 0.72 for the original m7-ATLPI (Table 2). The sensitivity and NPV for 2-year OS with the simpli fi ed m7-ATLPI were the same as those with the original m7-ATLPI; that is, 100% and 100% in the training cohort, and 94% and 86% in the validation cohort, respectively (Table 3). This tendency was consistent with results of 1-year OS.

Discussion

By considering gene mutation status and clinical parameters in patients with aggressive ATL who received intensive chemotherapy, we established the m7-ATLPI prognostic model, which could select patients with longer survival from among low-risk ATL patients aged <70 years who received intensive chemotherapy. Indices that stratify favorable subsets and help avoid excessive treatment have been developed for malignant lymphoma.20 However, such indices have not been developed in ATL. The m7-ATLPI was based on the previously established ATL-PI, which considers clinical risk factors, and on the mutational status of seven genes, specifically TP53, IRF4, RHOA, PRKCB, CARD11, CCR7, and GATA3. This mutational information was used to identify patients with a favorable prognosis after intensive chemotherapy alone from among low-risk patients defi ned not only by the ATL-PI, but also by other ATL prognostic models such as the JCOG-PI.

We previously reported that genetic abnormalities affected clinical outcomes in ATL patients.17 In that report, the status of PRKCB mutation and 9p24 (PD-L1) amplification, age ≥70 years, and the JCOG-PI high-risk category defined by poor PS or elevated cCa level were associated with poor survival by multivariate analysis. These adverse prognostic factors could divide aggressive ATL patients into different risk category groups; however, the 2-year OS of patients with none of these adverse factors was still below 30%. In follicular lymphoma, the Follicular Lymphoma International Prognostic Index (FLIPI), which is based on clinical and basic laboratory data, has been widely used in risk models.21 The m7-FLIPI is a clinicogenetic risk model that includes the mutation status of seven genes in addition to FLIPI and PS, and it was found to be able to identify patients at highest risk of treatment failure. 22 In our previous report,17 the chemotherapy regimens varied, the clinical information was insufficient, and patients aged ≥70 years comprised 30% of the cohort. In order to identify patients with a favorable prognosis after standard intensive chemotherapy alone, and who therefore might not require upfront allo-HSCT, we

focused on patients aged <70 years who were candidates for allo-HSCT, and collected detailed clinical data to generate a clinicogenetic prognostic model based on the ATL-PI.

The m7-ATLPI includes the ATL-PI risk category as well as the presence or absence of non-silent mutations in TP53, IRF4, RHOA, PRKCB, CARD11, CCR7, and GATA3 . Of these mutations, the fi rst fi ve were associated with a poor prognosis, while the last two were correlated with a favorable prognosis. Our previous report identi fi ed only PRKCB mutations as being associated with a poor prognosis.17 Sakihama et al. also reported that PRKCB mutations were associated with a poor prognosis in patients with aggressive ATL.23 Using the m7-ATLPI, more than half of patients classified as low risk by the ATL-PI were reclassified as intermediate risk. This means that patients determined to be low risk based on clinical and basic laboratory data were divided into two risk groups by also considering gene mutation information. Excluding patients who also had CCR7 or GATA3 mutations, those with mutations in one or more of TP53, IRF4, RHOA, PRKCB, and CARD11 had an inferior prognosis. They demonstrated a relatively poor clinical course with a mean survival time of 1.28 years, which was similar to that in the intermediate-risk group defined by the ATL-PI. Patients classified as low risk by the ATL-PI and who did not have mutations in TP53, IRF4, RHOA, PRKCB , or CARD11 , or those with PRKCB or CARD11 mutations in addition to mutations in a gene with favorable impact (CCR7 or GATA3), exhibited a favorable prognosis with a mean survival time of 7.55 years. These patients may not require upfront allo-HSCT, since only 30–40% of such patients who underwent alloHSCT achieved longer survival.8-10

We previously reported that a high CNA count was a characteristic of aggressive-type ATL.17 In that report, only 9p24 (PD-L1) amplification had an adverse effect on OS in aggressive-type ATL. In the subset of our cohort in which CNA was examined, patients with CNA in TP53 exhibited poor OS. However, adding CNA data to the m7ATLPI did not enhance the risk stratification potential of the model compared with inclusion of mutation data alone. This might be because all patients with CNA in TP53 also harbored one or more mutations in PRKCB, CARD11, or TP53, and these mutations had a negative impact on OS.

Even though the m7-ATLPI was developed to predict OS in patients with aggressive ATL who are treated with standard intensive chemotherapy, it is also useful to predict PFS. The median PFS in the low-risk group defined by the m7-ATLPI was 2.0 years, which was longer than the 0.92 years in the corresponding group defined by the ATLPI, suggesting that the m7-ATLPI predicts sensitivity to chemotherapy. As noted above, IFN-α, zidovudine, mogamulizumab, and brentuximab vedotin are commonly used

to treat aggressive-type ATL. We were unable to analyze the effects of these drugs on OS and PFS in this study, because all patients were treated with only conventional chemotherapy as first-line therapy. This is a major limitation of this study, and future research should develop prognostic models that integrate IFN-α, zidovudine, mogamulizumab, or brentuximab vedotin treatment. In summary, the m7-ATLPI, which was established by integrating the mutational status of seven genes with the previously established ATL-PI, and its simplified version, namely the simplified m7-ATLPI, could divide ATL patients into three risk categories. Sixteen percent of ATL patients aged <70 years were categorized as low risk by the m7ATLPI, and exhibited a relatively favorable prognosis with a 2-year OS of more than 80% with standard intensive chemotherapy alone. The current standard of care for ATL, which is to perform upfront allo-HSCT, might be reconsidered in this group. On the other hand, patients classi fi ed by the m7-ATLPI as intermediate or high risk had shorter survival, and may benefi t from alternative treatments such as allo-HSCT.

Disclosures

KS has received consulting fees from Novartis Pharma, Takeda Pharmaceutical, and Bristol-Myers, all outside the submitted work, and has received research grants from Perseus Proteomics, Pharma Essentia Japan KK, AbbVie GK, Astellas Pharma, MSD, Chugai Pharmaceutical, Kyowa Kirin, Pfizer, Novartis Pharma, Otsuka Pharmaceutical, and Asahi Kasei Medical, all outside the submitted work. KK holds stock in Asahi Genomics, has a patent for genetic alterations as a biomarker in T-cell lymphomas, and has received research funding from Chordia Therapeutics outside the submitted work. MH has received honoraria from Chugai Pharm and Huya Japan. MY has received consulting fees from Takeda Pharmaceutical, and honoraria from Takeda Pharmaceutical, Sanofi, and Novartis Pharma, all outside the submitted work. YK has received personal fees from Takeda Pharmaceutical, outside the submitted work. YI has received honoraria from Kyowa Kirin, Celgene, Bristol-Myers Squibb, Eisai, Sanofi , Sumitomo Dainippon Pharma, SymBio Pharmaceuticals , Meiji Seika Pharma, Chugai Pharmaceutical, and Nippon Shinyaku, all outside the submitted work. KK has received consulting fees from AbbVie, AstraZeneca, Celgene, Chugai, Eisai, Janssen, Novartis, and Daiichi Sankyo; honoraria from Takeda, MSD, Kyowa-Kirin, Janssen, Celgene, Ono, Mundi, DainipponSumitomo, and Bristol-Myers Squibb; and research funding from Chugai, Takeda, Kyowa Kirin, AbbVie, Novartis, Eisai, Janssen, Celgene, Ono, Novartis, and Daiichi Sankyo, all outside the submitted work. YM has received research funding from Sumitomo-Dainippon, and honoraria from Astellas, Sumitomo-Dainippon, Chugai, Kyowa-Kirin, Abbvie, Novartis, Bristol-Myers Squibb, P fi zer, Janssen,

Eisai, Daiichi-Sankyo, Takeda, Sanofi, Janssen, and Nippon-Shinyaku, all outside the submitted work. KI has received research grants from Ono Pharmaceutical and Kyowa Kirin; consulting fees from Daiichi Sankyo; and honoraria from Chugai Pharmaceutical, Celgene, and Kyowa Kirin, all outside the submitted work. AU has received honoraria from Kyowa Kirin, Daiichi Sankyo, Bristol-Myers Squibb, and Meiji Seika Pharma; and consulting fees from JIMRO and Otsuka Medical Devices , all outside the submitted work. AT-K serves as an advisor for Megakaryon and receives research funding from Ono Pharmaceutical, DSK, and Cognano.

Contributions

TK, KKataoka, SO, AU, MH, AKitanaka and KShimoda were involved in the conception, design, or planning of the study. AU, NN, MT, YT, MMiyahara, KI, MY, KN, JY, MMatsuoka, YI, YM, KKato, TSugio, KAkashi, HM, KO, KMarutsuka, TSakamoto, AT-K, TW, AKitanaka, KY, NK, HK, KMaeda, TT, TH and YK provided resources and administrative support. AKamiunten, KAkizuki, KShide, TH and YK collected the data. TK, YK, JK, KShide, KKataoka, SO, and KShimoda accessed, analyzed, and verifi ed the data. TK, YK, JK, and KKataoka performed the statistical analyses. TK, YK, JK and KKataoka developed the software application. TK and KShimoda wrote the fi rst draft of the manuscript, with contributions from all authors. KKataoka, SO and AU critically reviewed or revised the manuscript for important intellectual content. All authors reviewed the interim drafts and the final version of the manuscript, and agree with its content and submission. KShimoda takes final responsibility for the decision to submit this manuscript for publication.

Acknowledgments

The authors would like to thank Y. Kakizoe, T. Kawabata, M. Fukuyama, M. Makino, Y. Aratake, Y. Arashi, and Y. Kurogi for their support with data curation, and M. Matsushita, T. Shinmori, and S. Saitou for their technical assistance.

Funding

This research was supported by grants JP19ck0106254, 20ck0106538h0001, and 21ck0106538h0002 to KS from the Japan Agency for Medical Research and Development.

Data-sharing statement

Somatic mutations detected by targeted capture sequencing are listed in the Online Supplementary Appendix. Deidenti fi ed participant data will be made available to researchers conditionally upon receipt of an approved study proposal along with evidence of approval of the proposal by an accredited ethics committee. Proposals should be made by email to the corresponding author. In order to gain access, a data access agreement needs to be signed.

References

1. Uchiyama T, Yodoi J, Sagawa K, et al. Adult T-cell leukemia: clinical and hematologic features of 16 cases. Blood. 1977;50(3):481-492.

2. Shimoyama M. Diagnostic criteria and classification of clinical subtypes of adult T-cell leukaemia-lymphoma. A report from the Lymphoma Study Group (1984-87). Br J Haematol. 1991;79(3):428-437.

3. Shimoyama M. Chemotherapy of ATL. K Takatsuki (Ed), Adult TCell Leukemia, Oxford University Press. Oxford. 1994:221-227.

4. Cook LB, Fuji S, Hermine O, et al. Revised Adult T-cell Leukemia-Lymphoma International Consensus Meeting Report. J Clin Oncol. 2019;37(8):677-687.

5. Tsukasaki K, Utsunomiya A, Fukuda H, et al. VCAP-AMP-VECP compared with biweekly CHOP for adult T-cell leukemialymphoma: Japan Clinical Oncology Group Study JCOG9801. J Clin Oncol. 2007;25(34):5458-5464.

6. Sekine M, Kameda T, Shide K, et al. Higher average chemotherapy dose intensity improves prognosis in patients with aggressive adult T-cell leukemia/lymphoma. Eur J Haematol. 2021;106(3):398-407.

7. Katsuya H, Ishitsuka K, Utsunomiya A, et al. Treatment and survival among 1594 patients with ATL. Blood. 2015;126(24):2570-2577.

8. Utsunomiya A, Miyazaki Y, Takatsuka Y, et al. Improved outcome of adult T cell leukemia/lymphoma with allogeneic hematopoietic stem cell transplantation. Bone Marrow Transplant. 2001;27(1):15-20.

9. Hishizawa M, Kanda J, Utsunomiya A, et al. Transplantation of allogeneic hematopoietic stem cells for adult T-cell leukemia: a nationwide retrospective study. Blood. 2010;116(8):1369-1376.

10. Fuji S, Fujiwara H, Nakano N, et al. Early application of related SCT might improve clinical outcome in adult T-cell leukemia/lymphoma. Bone Marrow Transplant. 2016;51(2):205-211.

11. Fukushima T, Nomura S, Shimoyama M, et al. Japan Clinical Oncology Group (JCOG) prognostic index and characterization of long-term survivors of aggressive adult T-cell leukaemialymphoma (JCOG0902A). Br J Haematol. 2014;166(5):739-748.

12. Katsuya H, Yamanaka T, Ishitsuka K, et al. Prognostic index for acute- and lymphoma-type adult T-cell leukemia/lymphoma. J Clin Oncol. 2012;30(14):1635-1640.

13. Fuji S, Yamaguchi T, Inoue Y, et al. Development of a modified prognostic index of patients with aggressive adult T-cell leukemia-lymphoma aged 70 years or younger: a possible riskadapted management strategies including allogeneic transplantation. Haematologica. 2017;102(7):1258-1265.

14. Kataoka K, Nagata Y, Kitanaka A, et al. Integrated molecular analysis of adult T cell leukemia/lymphoma. Nat Genet. 2015;47(11):1304-1315.

15. Kogure Y, Kameda T, Koya J, et al. Whole-genome landscape of adult T-cell leukemia/lymphoma. Blood. 2022;139(7):967-982.

16. Kataoka K, Shiraishi Y, Takeda Y, et al. Aberrant PD-L1 expression through 3'-UTR disruption in multiple cancers. Nature. 2016;534(7607):402-406.

17. Kataoka K, Iwanaga M, Yasunaga JI, et al. Prognostic relevance of integrated genetic profiling in adult T-cell leukemia/lymphoma. Blood. 2018;131(2):215-225.

18. Reddy A, Zhang J, Davis NS, et al. Genetic and functional drivers of diffuse large B cell lymphoma. Cell. 2017;171(2):481-494.

19. Sakamoto Y, Ishida T, Masaki A, et al. CCR4 mutations associated with superior outcome of adult T-cell leukemia/lymphoma under mogamulizumab treatment. Blood. 2018;132(7):758-761.

20. Persky DO, Li H, Stephens DM, et al. Positron emission tomography-directed therapy for patients with limited-stage diffuse large B-cell lymphoma: results of intergroup National Clinical Trials Network Study S1001. J Clin Oncol. 2020;38(26):3003-3011.

21. Relander T, Johnson NA, Farinha P, et al. Prognostic factors in follicular lymphoma. J Clin Oncol. 2010;28(17):2902-2913.

22. Pastore A, Jurinovic V, Kridel R, et al. Integration of gene mutations in risk prognostication for patients receiving firstline immunochemotherapy for follicular lymphoma: a retrospective analysis of a prospective clinical trial and validation in a population-based registry. Lancet Oncol. 2015;16(9):1111-1122.

23. Sakihama S, Morichika K, Saito R, et al. Genetic profile of adult T-cell leukemia/lymphoma in Okinawa: association with prognosis, ethnicity, and HTLV-1 strains. Cancer Sci. 2021;112(3):1300-1309.

Maria-Victoria Mateos,1* Katja Weisel,2* Thomas Martin,3 Jesús G. Berdeja,4 Andrzej Jakubowiak,5 A. Keith Stewart,6 Sundar Jagannath,7 Yi Lin,8 Joris Diels,9 Francesca Ghilotti,10 Pushpike Thilakarathne,9 Nolen J. Perualila,9 Jedelyn Cabrieto,9 Benjamin Haefliger,11 Nichola Erler-Yates,12 Clare Hague,13 Carolyn C. Jackson,14 Jordan M. Schecter,14 Vadim Strulev,15 Tonia Nesheiwat,16 Lida Pacaud,16 Hermann Einsele17 and Philippe Moreau18

1University Hospital of Salamanca/IBSAL, CIC, Salamanca, Spain; 2University Medical Center Hamburg-Eppendorf, Hamburg, Germany; 3UCSF Helen Diller Family Comprehensive Cancer Center, San Francisco, CA, USA; 4Sarah Cannon Research Institute, Nashville, TN, USA; 5University of Chicago, Chicago, IL, USA; 6University Health Network and the Princess Margaret Cancer Centre, Toronto, Ontario, Canada; 7Mount Sinai Medical Center, New York, NY, USA; 8Mayo Clinic, Rochester, MN, USA; 9Janssen Pharmaceutica NV, Beerse, Belgium; 10Janssen-Cilag SpA, Cologno Monzese, Italy; 11Cilag GmbH International, Zug, Switzerland; 12Janssen-Cilag GmbH, Neuss, Germany; 13Janssen-Cilag N.V., High Wycombe, UK; 14Janssen R&D, Raritan, NJ, USA; 15EMEA Medical Affairs, Janssen Pharmaceutica NV, Beerse, Belgium; 16Legend Biotech USA Inc., Piscataway, NJ, USA; 17Universitätsklinikum Würzburg, Medizinische Klinik und Poliklinik II, Würzburg, Germany and 18Clinical Hematology, University Hospital Hotel-Dieu, Nantes, France

*M-VM and KW contributed equally as first authors.

Abstract

Correspondence: M.V. Mateos mvmateos@usal.es

Received: December 14, 2021.

Accepted: December 13, 2022.

Early view: December 22, 2022.

https://doi.org/10.3324/haematol.2022.280482

Published under a CC BY license

Ciltacabtagene autoleucel (cilta-cel) is a chimeric antigen receptor T-cell therapy studied in patients with multiple myeloma exposed to three classes of treatment in the single-arm CARTITUDE-1 study. To assess the effectiveness of cilta-cel compared to real-world clinical practice (RWCP), we performed adjusted comparisons using individual patients’ data from CARTITUDE-1 and LocoMMotion, a prospective, multinational study of patients with multiple myeloma triple-class exposed of treatment. Comparisons were performed using inverse probability weighting. In CARTITUDE-1, 113 patients were enrolled, and 97 patients were infused with cilta-cel. In LocoMMotion, 248 patients were enrolled, and 170 patients were included in the comparisons versus infused patients. Ninety-two unique regimens were used in LocoMMotion, most frequently carfilzomib-dexamethasone (13.7%), pomalidomide-cyclophosphamide-dexamethasone (13.3%) and pomalidomidedexamethasone (11.3%). Adjusted comparisons showed that patients treated with cilta-cel were 3.12-fold more likely to respond to treatment than those managed by RWCP (response rate, 3.12, 95% confidence interval [95% CI]: 2.24-4.00), had their risk of progression or death reduced to by 85% (progression-free survival hazard ratio=0.15, 95% CI: 0.08-0.29), and a risk of death lowered by 80% (overall survival hazard ratio HR=0.20, 95% CI: 0.09-0.41). The incremental improvement in healthrelated quality of life from baseline for cilta-cel versus RWCP at week 52, as measured by EORTC QLQ-C30 Global Health Status, was 13.4 (95% CI: 3.5-23.6) and increased to 30.8 (95% CI: 21.8-39.8) when including death as additional information regarding patients’ health status. Patients treated with cilta-cel experienced more adverse events than those managed with RWCP (any grade: 100% vs. 83.5%). The results from this study demonstrate improved efficacy outcomes of cilta-cel versus RWCP and highlight its potential as a novel and effective treatment option for patients with multiple myeloma triple-class exposed of antimyeloma treatment. CARTITUDE-1 is registered with clinicaltrials gov. Identifier: NCT03548207. LocoMMotion is registered with clinicaltrials gov. Identifier: NCT04035226.

Adjusted comparison of outcomes between patients from CARTITUDE-1 versus multiple myeloma patients with prior exposure to proteasome inhibitors, immunomodulatory drugs and anti-CD38 antibody from the prospective, multinational LocoMMotion study of real-world clinical practice

Introduction

Multiple myeloma (MM) is a hematologic cancer in which clonal proliferation of malignant plasma cells occurs along with overproduction of myeloma protein (M-protein).1 MM is a highly heterogeneous cancer associated with a variable clinical course and significant clinical burden whose severity progresses over time.2-4 MM represents 1% of all cancers worldwide and nearly 10% of hematologic neoplasms.1 Approximately 50,000 patients in the European Union and USA are diagnosed with MM each year, while nearly 30,000 die during this same time frame.5

Therapies such as immunomodulatory agents, proteasome inhibitors and monoclonal antibodies have contributed to meaningful improvements in patients’ outcomes over the past decade.6-11 However, despite these therapeutic advances, MM remains an incurable disease.2,12 For MM patients previously exposed to proteasome inhibitors, immunomodulatory agents and anti-CD38 antibodies (“triple-class exposed”), there is currently no standard of care; patients’ outcomes are very poor, and include a median overall survival of 9.3 months.4,13 New, more efficacious treatment options are needed for these patients to extend their survival, halt disease progression and improve their quality of life.13-15 Chimeric antigen receptor T-cell (CAR-T) therapy is a novel approach to treatment that offers potential for long-term disease control in some hematologic cancers.16 Ciltacabtagene autoleucel (cilta-cel; JNJ-68284528) is an experimental CAR-T therapy that targets B-cell maturation antigen (BCMA).17 CARTITUDE-1 (clinicaltrials gov. Identifier: NCT03548207), an open-label, single-arm, clinical trial, investigated the safety and efficacy of cilta-cel in patients with triple-class exposed relapsed/refractory MM (RRMM).18, 19 CARTITUDE-1 was designed as a single-arm study because of the lack of clinical equipoise and the absence

of an established standard-of-care therapy for patients with triple-class exposed RRMM, which precluded the performance of a traditional randomized trial. In such situations, adjusted comparisons of trial outcomes compared to those observed in an external cohort of similar patients may provide valuable information on the benefits of ciltacel relative to treatments used in clinical practice, thereby creating an external control arm. LocoMMotion (clinicaltrials gov. Identifier: NCT04035226) was designed to be the first prospective, non-interventional, multinational study of therapies used in real-world clinical practice (RWCP) in tripleclass exposed patients and was designed such that clinical outcome measures and eligibility criteria were matched to those of CARTITUDE-1.20,21 In this study, individual patients’ data from CARTITUDE-1 and LocoMMotion were analyzed to compare the effectiveness of cilta-cel versus currently available real-world clinical practice (RWCP), therapies in patients with triple-class exposed RRMM.

Methods

A synopsis of the study methods is provided below. Detailed descriptions regarding data sources, design, outcomes and approach to analysis are provided in the Online

Supplementary Appendix S1

Individual patients’ data from the CARTITUDE-1 trial (clinicaltrials gov. Idenfifier: NCT03548207, data cut July 2021) and the LocoMMotion prospective, multinational, non-interventional cohort study (clinicaltrials gov. Idenfifier: NCT04035226, data cut May 2021) were used to conduct adjusted comparisons between the effects of cilta-cel and RWCP. In CARTITUDE-1, 113 patients were enrolled and underwent apheresis. Sixteen patients discontinued the study between apheresis and infusion with cilta-cel. Data

Figure 1. Selection of patients for the CARTITUDE-1 and LocoMMotion populations. Patients in CARTITUDE-1 were treated with ciltacabtagene autoleucel, patients in the LocoMMotion study were treated with real-world clinical practice.

from the set of 97 patients infused with cilta-cel in CARTITUDE-1 were compared with the data from the set of 170 patients from LocoMMotion, who were progressionfree 52 days after treatment initiation (Online Supplementary Appendix S2). These groups are referred to as the infused/aligned populations. Second, analyses were also performed involving the 113 patients enrolled in CARTITUDE-1, along with all 248 patients enrolled in LocoMMotion, referred to as the enrolled populations. In CARTITUDE-1, the index date was the date of apheresis for the enrolled population and the date of infusion for the infused population. The index date for the enrolled population from LocoMMotion was the date of treatment initiation, while the date of treatment initiation plus 52 days was used for the aligned population. Overall response rate (ORR), very good partial response or better (≥VGPR), complete response or better (≥CR), progression-free survival (PFS) and overall survival (OS) were compared between cilta-cel and RWCP. Two patient-related outcomes, the EuroQoL Group’s EQ visual analog scale (EQ VAS) and the European Organization for the Research and Treatment of Cancer Quality of Life Questionnaire (EORTC QLQ-C30) global health status (GHS) were measured over time and were compared without adjustment. The frequency and severity of adverse events were also compared.

Adjusted comparisons between cilta-cel and RWCP were performed using inverse probability weighting methods to estimate the average treatment effect in the treated population (IPW-ATT). The prognostic baseline characteristics to be adjusted for in the statistical analyses were based upon a review of the literature and consultations with clinical experts. The degree of imbalance between groups was assessed using standardized mean differences, with values >0.2 considered to reflect importance differences. Weighted logistic regression was used for response outcomes to estimate odds ratios (OR) with 95% confidence intervals (95% CI), transformed to responserate ratios (RR). Weighted Cox proportional hazards regression was used to estimate hazard ratios (HR) with corresponding 95% CI. Sensitivity analyses using overlap weighting22 (ATO) and multivariable regression analyses including the same prognostic variables as covariates in the models to estimate the relative treatment effects were performed. Given the wide range of treatment regimens used in RWCP, two sensitivity analyses were performed to explore the impact of this on the relative treatment effect. Subgroups of patients treated with novel therapies (immunomodulatory agents, proteasome inhibitors, monoclonal antibodies, selinexor and belantamab) and patients treated with a combination of three or more therapies were explored.

The sponsor together with the investigators designed the comparative study, were involved in the data analysis and

interpretation and the writing of the manuscript. CARTITUDE-1 and LocoMMotion, funded and conducted by the sponsor, were performed in accordance with the Declaration of Helsinki and International Conference on Harmonisation guidelines for Good Clinical Practice. An independent ethics committee/institutional review board at each study center approved the protocols.

Results

Patient populations

Two populations of patients were analyzed: the infused/aligned and enrolled cohorts. The infused cohort for CARTITUDE-1 consisted of 97 patients18 and its aligned population from LocoMMotion contained 170 patients. The enrolled cohort contained 113 patients from CARTITUDE-1 and 248 patients from LocoMMotion (Figure 1).

Treatment regimens received in real-world clinical practice

In total, 92 unique regimens were used in LocoMMotion. A full list of the treatments received at baseline is provided in Online Supplementary Appendix S3. Treatment regimens most commonly received at baseline by patients in the RWCP cohort were carfilzomib-dexamethasone (12.9%), pomalidomide-cyclophosphamide-dexamethasone (10.9%) and pomalidomide-dexamethasone (9.7%). The ten most frequently used regimens at baseline were given to 54.7% of all patients, and 43 patients (17.3%) received a regimen that was not received by any other patient in the sample. Selinexor, available only in the USA during study recruitment for LocoMMotion, was used twice. Belantamab mafodotin was approved in the USA and European Union for 3 months during the LocoMMotion recruitment period and was received by seven patients in their line of treatment of interest (in 5 cases as monotherapy, in 2 cases in combination regimens).

Prognostic value of baseline characteristics

The following baseline characteristics were considered a priori to be prognostic and adjusted for in the comparative analyses: refractory status, International Staging System (ISS), time to progression on last prior line of treatment, presence of extramedullary disease, number of prior lines of treatment, years since MM diagnosis, average duration of prior lines of treatment, age, sex, hemoglobin, lactate dehydrogenase level, creatinine clearance, Eastern Cooperative Oncology Group performance status (ECOG PS), MM type, history of stem cell transplant, race and cytogenetic risk. The prognostic value of these baseline characteristics was explored first in the LocoMMotion cohort (Online Supplementary Appendix S6). Refractory status, ISS stage, time to progression on last prior line of treatment, hemoglobin concentration, and ECOG PS were shown to be signifi-

cantly associated with outcomes. However, in a combined analysis of CARTITUDE-1 and LocoMMotion using a multivariable model, only ISS stage and ECOG PS retained a statistically significant influence on survival, as there are associations between baseline characteristics (Online Supplementary Appendix S6). The prognostic value of the

baseline characteristics for response rates and PFS were generally similar (Online Supplementary Appendix S6).

Comparative analysis of efficacy endpoints

Infused/aligned populations

Table 1 shows the baseline characteristics before and

Summary diagnostics

The pre-weighting and post-weighting distributions of demographics by intervention group are shown. Standardized mean differences >0.2 were considered to indicate differences between groups. Cilta-cel: ciltacabtagene autoleucel; IPW: inverse probability weighting; ATT: average treatment effect in the treated population; RWCP: real-world clinical practice; SMD: standardized mean difference; ISS: International Staging System; LOT: lines of therapy; LDH: lactate dehydrogenase; ECOG PS: Eastern Cooperative Oncology Group Performance Status; MM: multiple myeloma; IgG: immunoglobulin G.

Table 1. Demographic balance between groups before and after inverse probability weighting for the infused/aligned population.

after weighting for the base case including refractory status, ISS stage, time to progression on last prior line of treatment, presence of extramedullary disease, number of prior lines of treatment, years since MM diagnosis, average duration of prior lines of treatment, age, sex, hemoglobin concentration, lactate dehydrogenase level, creatinine

clearance, ECOG PS and MM type. The models including the extended variables are shown in sensitivity analyses (Online Supplementary Appendix S4, section 3.4.3 sensitivity analyses, Online Supplementary Appendix S7). Prior to reweighting of the infused/aligned populations, examination of standardized mean differences found imbalances

Figure 2. Unadjusted and adjusted Kaplan-Meier survival curves for the infused/aligned population. (A) Progression-free survival. (B) Overall survival. Blue lines represent the survival of patients treated with ciltacabtagene autoleucel, the solid red lines represent the unadjusted survival of patients treated with real-world clinical practice (RMCP) and the dotted orange lines are adjusted Kaplan-Meier curves following inverse probability weighting. Cilta-cel: ciltacabtagene autoleucel; ATT: average treatment effect in the treated population; HR: hazard ratio.

Observed and adjusted data comparing rates of clinical response in the infused/aligned population between patients treated with ciltacabtagene autoleucel or real-world clinical practice (RWCP). Adjusted comparisons account for the effects of refractory status, International Staging System stage, time to progression on prior line of treatment, presence of extramedullary disease, number of prior lines of treatment, years since diagnosis of multiple myeloma, average duration of prior lines of treatment, patients’ age and sex, hemoglobin at index date, lactate dehydrogenase at index date, creatinine clearance at index date, Eastern Cooperative Oncology Group Performance Status, and type of multiple myeloma. All comparisons favored ciltacabtagene autoleucel (Cilta-cell). 95% CI: 95% confidence interval; OR: odds ratio; IPW: inverse probability weighting; ATT: average treatment effect in the treated population; ORR: overall response rate; VGPR: very good partial response; CR: complete response; NE: not estimable.

Table 2. Summary of observed and adjusted rates of clinical response in the infused/aligned population.

in most of the baseline characteristics. Imbalances can bias an unadjusted comparison both in favor of and against the intervention. Here, we found imbalances in both directions, however, the most important differences were observed for refractory status, time to progress on prior line of treatment, and average duration of prior lines of treatment, which suggest that the CARTITUDE-1 population had more aggressive myeloma. After IPW-ATT weighting, imbalances between the cilta-cel and RWCP cohorts were greatly reduced with all standardized mean differences <0.2, except for extramedullary disease (standardized mean difference =0.23) (Table 1; Online Supplementary Appendix S5). Online Supplementary Appendix S5 further illustrates that propensity score distributions were different before reweighting and became very similar after reweighting.

Response endpoints

Table 2 summarizes the observed rates and adjusted treatment comparisons for ORR, ≥VGPR and ≥CR. The ORR for cilta-cel was 97.9% versus 42.9% for RWCP. Rates of ≥VGPR and ≥CR were 94.8% and 82.5%, respectively, with cilta-cel, compared to 17.6% and 0.6%, respectively, with RWCP. IPWATT adjusted comparisons favored cilta-cel for ORR (RR=3.12, 95% CI: 2.24-4.00; P<0.0001) and ≥VGPR (RR=5.67, 95% CI: 3.25-8.08; P<0.0001). As only one patient (0.6%) in the RWCP group achieved ≥CR compared to 80 (82.5%) with cilta-cel, an IPW-ATT adjusted comparison could not be estimated; however, the extreme difference in observed ≥CR rates between the cilta-cel and RWCP groups reflects the significantly higher efficacy of cilta-cel.

Progression-free survival

The median PFS in the cilta-cel group was not reached, while the median PFS for patients in RWCP was 4.34

months (95% CI: 3.65-5.55). Figure 2A presents the observed and adjusted Kaplan-Meier curves for PFS for both groups; the IPW-ATT adjusted HR for cilta-cel versus RWCP was 0.15 (95% CI: 0.08-0.29; P<0.0001). The proportional hazards assumption was met for all analyses.

Overall survival

In both the cilta-cel and RWCP populations, median OS was not reached when unadjusted. The median OS for the adjusted RWCP population was 11.33 months. Figure 2B presents Kaplan-Meier curves for OS in both groups. Following IPW-ATT-based adjustment, the HR comparing groups was 0.20 (95% CI: 0.09-0.41; P<0.0001), favoring cilta-cel. The proportional hazards assumption was met for all analyses.

Enrolled populations

Prior to reweighting of the enrolled population, examination of standardized mean differences found imbalances in nine baseline characteristics; however, after IPW-ATT weighting had been performed, imbalances between the cilta-cel and RWCP cohorts were greatly reduced, with all standardized mean differences below 0.20 (Online Supplementary Appendix S4 and S5).

Response endpoints

Table 3 summarizes the observed rates and adjusted treatment comparisons for ORR, ≥VGPR and ≥CR. The ORR for cilta-cel was 84.1% versus 29.8% for RWCP. Rates of ≥VGPR and ≥CR were 81.4% and 70.8%, respectively, with cilta-cel, compared to 12.5% and 0.4%, respectively, with RWCP. After IPW-ATT adjustment, comparisons favored cilta-cel for each of ORR (RR=4.34, 95% CI: 2.69-6.00; P<0.0001) and ≥VGPR (RR=8.08, 95% CI: 3.63-12.53; P<0.0001). As only one patient in LocoMMotion achieved

Observed and adjusted data comparing rates of clinical response in the enrolled population between patients treated with ciltacabtagene autoleucel or real-world clinical practice (RWCP). Adjusted comparisons account for the effects of refractory status, International Staging System stage, time to progression on prior line of treatment, presence of extramedullary disease, number of prior lines of treatment, years since MM diagnosis, average duration of prior lines of treatment, patients’ age and sex, hemoglobin at index date, lactate dehydrogenase at index date, creatinine clearance at index date, Eastern Cooperative Oncology Group Performance Status, and type of multiple myeloma. All comparisons favored ciltacabtagene autoleucel (Cilta-cel). RWCP: real-world clinical practice; 95% CI: 95% confidence interval; OR: odds ratio; IPW: inverse probability weighting; ATT: average treatment effect in the treated population; ORR: overall response rate; VGPR: very good partial response; CR: complete response; NE: not estimable.

Table 3. Summary of observed and adjusted rates of clinical response in the enrolled population.

≥CR, IPW-ATT adjusted comparison could not be derived; however, the extreme difference in observed CR rates between the cilta-cel and RWCP groups reflects a significant difference between therapies.

Progression-free survival and overall survival

Results of the unadjusted comparison produced an estimate of effect for PFS that favored cilta-cel (HR=0.23, 95% CI: 0.16-0.33; P<0.0001). After IPW-ATT reweighting, the PFS HR was 0.19 (95% CI: 0.11-0.32; P<0.0001). The unadjusted comparison for OS between cilta-cel and RWCP favored cilta-cel (HR=0.32, 95% CI: 0.20-0.50; P<0.0001). Following IPW-ATT based adjustment, the OS HR was 0.32 (95% CI: 0.17-0.58; P<0.0001), again supporting the unadjusted results (Figure 3B). The proportional hazards assumption was met for all analyses.

Sensitivity analyses

To assess the robustness of the findings, sensitivity analyses were performed by using overlap weighting and multivariable regression (Online Supplementary Appendix S1) and by including additional baseline characteristics

(race, history of stem cell transplant and cytogenetic risk; see Online Supplementary Appendix S1 for details regarding these variables) for the adjusted analyses. Figure 3 shows consistent results for IPW-ATO and multivariable regression with the main analyses (IPW-ATT), for both response (Figure 3A) and survival endpoints (Figure 3B). The impact of additionally including race, history of stem cell transplantation and cytogenetic risk on results was minimal (Online Supplementary Appendix S7), which can be explained by their low prognostic value. However, including these additional covariates caused imbalances in the baseline characteristics in the ATT-based results, which were balanced in the main analyses. Online Supplementary Appendix S7 shows results of relative treatment comparisons related to use of the extended model including all available baseline characteristics.

Results from the sensitivity analyses excluding patients who were not treated with a regimen that included a novel therapy and excluding patients treated with a single or combination of two therapies were less stable, but generally consistent with the overall results (Online Supplementary Appendix S7).

Figure 3. Summary of unadjusted and adjusted comparisons for response and survival outcomes. (A) Forest plots of response outcomes showing response ratios with corresponding 95% confidence intervals (95% CI) comparing ciltacabtagene autoleucel (cilta-cel) versus real-world clinical practice (RWCP) with different analytical methods and for the infused/aligned and enrolled patient populations. Values >1 favor cilta-cel, values <1 favor RWCP. (B) Forest plots of survival outcomes showing hazard ratios (HR) with corresponding 95% confidence intervals comparing cilta-cel versus RWCP with different analytical methods and for the infused/aligned and enrolled patient populations. Values <1 favor cilta-cel, values >1 favor RWCP. RR: response rate ratio; ORR: overall response rate; IPW: inverse probability weighting; ATT: average treatment effect in the treated population; ATO: average treatment effect in the overlap population; VGPR: very good partial response; RWCP: real-world clinical practice; OS: overall survival; PFS: progression-free survival.

Comparative analyses of patient-reported outcomes

RWCP groups demonstrated improved outcomes over time. Patients treated with cilta-cel experienced continuously improving quality of life over time, as measured by absolute differences versus baseline on a 0-100 standardized scale for EQ VAS and GHS, which increased from 4.0

Haematologica | 108 August 2023 2199 ARTICLE - CARTITUDE-1 versus LocoMMotion M.V. Mateos et al.

Figure

patient-reported

EuroQoL

EQ visual analog scale (EQ VAS) of patients who were alive and did not initiate subsequent therapy. (B) Comparison of EORTC QLQ-C30 global health status (GHS) of patients who were alive and did not initiate subsequent therapy. (C) Comparison of EQ VAS of patients who were alive and did not initiate subsequent therapy or died, i.e., adjusted for informative dropout analysis. (D) Comparison of EORTC QLQ-C30 GHS of patients who were alive and did not initiate subsequent therapy or died, i.e., adjusted for informative dropout analysis; ciltacel: ciltacabtagene autoleucel; RWCP: real-world clinical practice.

4. Comparisons of

outcomes. (A) Comparison of

Group’s

and 3.0 at week 4 to 12.6 and 15.6 at week 52, respectively. Improvements with RWCP were considerably smaller (from -1.4 and -0.7 at week 4 to 2.3 and 2.2 at week 52). The differences in improvement versus baseline between cilta-cel and RWCP increased up to 10.3 (P=0.0076) for EQ VAS and 13.4 (P=0.0081) for GHS at week 52. The “adjusted-for-informative-dropout” analysis, which included death as additional information regarding patients’ health status,24 showed thatimprovements for cilta-cel versus RWCP were 23.7 (P<0.0001) and 30.8 (P<0.0001) for the EQ VAS and GHS, respectively (Figure 4C, D), illustrating the conservative nature of the main analysis.

Comparison of safety outcomes

Detailed safety findings for CARTITUDE-1 18 and LocoMMotion20,21 have been previously reported elsewhere. Unadjusted comparison of all adverse events showed higher rates of adverse events for cilta-cel versus RWCP across organ classes. All patients treated with cilta-cel experienced at least one adverse event, while 83.5% of patients treated with RWCP had at least one adverse event. This was also the case for grade 3/4 events (93.8% vs . 49.2%)

(Table 4). Mateos et al 21 indicated that adverse events had been underreported for RWCP in LocoMMotion because of the observational nature of this study. Six (6.2%) patients treated with cilta-cel and 19 (7.7%) patients with RWCP experienced an adverse event with an outcome of death. Cytokine release syndrome and CART therapy-related neurotoxicites occurred in 95% and 21% of patients in CARTITUDE-1, respectively, and were manageable.

Discussion

Despite improvements in treatments for patients with MM in recent years, there is still a pressing need for novel therapies to address unmet treatment needs for patients with triple-class exposed RRMM. Patients treated with cilta-cel have demonstrated early, deep and durable clinical responses and the therapy had a manageable safety profile within the recent CARTITUDE-1 trial. Due to a lack of an established standard of care and clinical equipoise, CARTITUDE-1 was designed as a single-arm trial. Hence, the

Hematologic AE occurring in ≥25%

aAdverse events underreported for real-world clinical practice (RWCP). bEvents not reported as immune effector cell-associated neurotoxicity syndrome (ICANS) in CARTITUDE-1 (i.e., onset after a period of recovery from cytokine release syndrome [CRS] and/or ICANS); cNo chimeric antigen receptor (CAR) T-cell treatments used in LocoMMotion. Adverse events (AE) ≥25% and of special interest (CRS, CAR-T cell neurotoxicities) are reported for ciltacabtagene autoleucel (cilta-cel) and RWCP for any grade and for grade 3/4 adverse events. NA: not applicable; ICANS: immune effector cell-associated neurotoxicity syndrome; AST: aspartate aminotransferase; ALT: alanine aminotransferase.

Table 4. Summary of adverse events observed with an incidence >25% and of special interest.

benefits of cilta-cel need to be compared to those of the group of therapies used in the context of current clinical practice. In such situations, adjusted comparisons versus data from the real world are needed, and this approach has become highly relevant for health technology assessment.25,26 Such adjusted comparisons use statistical methods to overcome the lack of randomization and the related potential for confounding bias. The LocoMMotion study was designed as the first prospective, multinational study of RWCP interventions in patients with advanced MM. The LocoMMotion cohort can be considered highly representative of the clinical population of interest as it includes patients from nine European countries and the USA, reflecting RWCP across different settings. The prospective design and alignment with CARTITUDE-1 guaranteed that eligibility criteria and definitions of all clinically important baseline characteristics and endpoints were identical in both studies, which allowed the most robust comparisons between ciltacel and current RWCP on all relevant endpoints, including response, PFS, OS, patient-related outcomes and safety. Given these strengths along with the rigorous analytical approaches, the findings of this study represent the highest quality comparative evidence on the benefits of cilta-cel for patients with triple-class exposed RRMM.

Based upon the analyses performed, clinically and statistically significant advantages with cilta-cel versus outcomes of RWCP were found for ORR, ≥VGPR, ≥CR, PFS, OS and patient-reported outcomes. Patients treated with cilta-cel were 3.1 times more likely to achieve a response (ORR) compared to RWCP patients and 5.7 times more likely to achieve ≥VGPR. Hazard ratios for OS demonstrated a reduced risk of death by 80% and an improvement in PFS of 85%. Gains in quality of life over 52 weeks were significantly better for cilta-cel than for RWCP, ranging between 10.3% and 30.8%, depending on the endpoint and the analytical approach. Unadjusted safety comparisons indicate higher rates of hematologic adverse events and CAR-T therapy-specific adverse events for cilta-cel.

Additional strength is conferred to the findings in this study, both by the internal consistency of results, as well as by the consistency with comparison to similar analyses of cilta-cel versus other external cohorts.27-29

While randomized controlled trials remain the gold-standard design when evaluating the benefits and safety of new medical interventions, such trials may not be feasible and/or ethical when clinical equipoise is lacking, and no established standard-of-care therapy exists. For these reasons, CARTITUDE-1 was performed as a single-arm study, and adjusted comparisons as presented here represent high quality evidence on the comparative effectiveness of cilta-cel relative to RWCP.

As with any non-randomized study, the potential for residual confounding for unobserved patients’ characteristics cannot be ruled out. However, in the current study the prospective

collection of patients’ characteristics at baseline in LocoMMotion was broad, which allowed data analyses to adjust for clinically important factors. Accounting for these characteristics was a key step in addressing differences between the two cohorts to avoid confounding bias in the comparative analyses, and represents an important strength of this study as opposed to naïve treatment comparisons or comparisons with existing data sources, which do not include all clinically relevant prognostic baseline variables. While three baseline characteristics (race, history of stem cell transplantation, cytogenetic risk) were not adjusted for in the main analysis, they were included in sensitivity analyses that showed consistent results. Although cytogenetic risk at baseline was previously shown to be a relevant predictive factor,30 missingness in LocoMMotion was high (37.9%), which reflects that cytogenetic testing is not routinely performed in clinical practice. As cytogenetic testing cannot be mandated in a non-interventional study, missingness could not be reduced. However, in the case of LocoMMotion, no association of outcomes with cytogenetic risk was observed in patients for whom data were available. Similar missingness in cytogenetic risk was observed in other real-world evidence sources and even in clinical trials.31,32 Similar challenges for LocoMMotion were also observed for CR rates and adverse events. The LocoMMotion study was performed with no restrictions on the types of treatments that could be received by patients, thereby allowing treating physicians to prescribe patients with the therapy they deemed most appropriate. The wide variety of treatment regimens used in the LocoMMotion cohort illustrates the absence of an established standardof-care therapy for patients with triple-class exposed RRMM, and is representative of current clinical practice. Although new therapies for the population of triple-class exposed RRMM patients have recently emerged, the RWCP group in the current study included only limited numbers of patients receiving selinexor or belantamab mafodotin, as these treatments only became available following their approvals in the USA and European Union toward the end of the recruitment period in LocoMMotion. Given the rapidly changing treatment landscape of MM and the heterogeneity of patients, further clinical and real-world studies are needed to better compare cilta-cel against these and other emerging therapies.

Comparisons of cilta-cel to individual therapies were not possible because of the highly varied treatments selected by physicians for their patients. However, two sensitivity analyses, excluding patients from the LocoMMotion cohort who received a regimen without a novel component and who received one or two treatments in combination, were performed. Due to the smaller sample sizes, results from these analyses were less stable, but generally consistent with the overall results illustrating that the comparative efficacy estimates for cilta-cel versus RWCP were consistent across treatment combinations, and were not being driven

by the heterogeneity in the LocoMMotion study or by patients receiving a particular therapy/combination.

The comparison of CARTITUDE-1 to the prospective LocoMMotion study on RWCP designed to match CARTITUDE-1 provides the highest quality possible comparative evidence for a single-arm trial. Findings from the adjusted treatment comparisons showed clinically and statistically significant improvements with cilta-cel compared to RWCP in MM patients exposed to three classes of treatment (proteasome inhibitors, immunomodulatory agents, anti-CD38 antibodies) and highlight cilta-cel’s potential as a novel and highly effective therapy to address unmet treatment needs in patients with triple-class exposed RRMM.

Disclosures

MVM has received honoraria from and is a member on boards of directors/advisory committees for Janssen, Celgene, Takeda, and Amgen; has received honoraria from Adaptive; and is a member on boards of directors/advisory committees for GSK, AbbVie, EDO, and Pharmamar. KW is a member on boards of directors/advisory committees for Amgen, Celgene, Janssen, and Sanofi; is a consultant for and receives honoraria from BMS and Takeda); has provided consultancy for Adaptive Biotech and Juno. TM has received research funding from Janssen, Amgen, and Sanofi and acts as a consultant for Oncopeptides and GSK. JGB has received research funding from AbbVie, Amgen, Acetylon, Bluebird, Bristol Myers Squibb, Celgene, Cellularity, Constellation, CRISPR, Therapeutics, CURIS, EMD Serono, Genentech, Glenmark, Janssen, Kesios, Lilly, Novartis, Poseida, Teva, Takeda Pharmaceuticals, and Vivolux; and has acted as a consultant for Amgen, Bioclinica, Bristol Myers Squibb, Celgene, CRISPR Therapeutics, Janssen, Karyopharm, Kite Pharma, Legend, Prothena, Servier, Takeda Pharmaceuticals, and SecuraBio. AJ has received consulting fees and honoraria from AbbVie, Adaptive, Amgen, Bristol Myers Squibb, Celgene, GlaxoSmithKline, Janssen, Juno, Karyopharm, and Sanofi; and is a member of a board of directors or advisory committee for AbbVie, Amgen, Bristol Myers Squibb, Celgene, GlaxoSmithKline, Janssen, Karyopharm, and Sanofi. AKS has received honoraria from Aventis, Janssen, Amgen, Oncopeptides, Bristol Myers Squibb, GlaxoSmithKline, and Sanofi; and is a member of a board of directors or advisory committee for Genomics England and Tempus. SJ is a consultant for Bristol Myers Squibb, Janssen, Karyopharm Therapeutics, Merck, Sanofi, Legend Biotech, and Takeda Pharmaceuticals. YL has provided consultancy services for Bluebird Bio, Celgene, Gamida Cells, Janssen, Huno, Kite, Novartis, Sorrento, Legend BioTech, and Vineti; and has received

research funding from Bluebird Bio, Celgene, Janssen, Kite, Merck, and Takeda Pharmaceuticals. JD, FG, PT, NJP, JC, NEY, CH, JMS, and VS are employees of Janssen. BH was a previous employee of Janssen. CCJ is employed with Janssen; and is a consultant physician at the Memorial Sloan Kettering Cancer Center (New York, NY, USA). TN and LP are employees of Legend Biotech. HE has no conflicts of interest to disclose. PM has received consultancy fees and/or honoraria from Celgene, Janssen, Amgen, Takeda, and AbbVie.

Contributions

MVM, KW, JD, BH, NEY, CH, and VS conceived the study. JD, FG, PT, BH, and NEY were responsible for the methodology. JD, FG, PT, NJP, and JC were responsible for software and formal analysis. MVM, KW, TM, JGB, AJ, AKS, SJ, YL, CCJ, JMS, VS, HE, and PM conducted the investigation. MVM, KW, TM, JGB, AJ, AKS, SJ, YL, JD, CH, CCJ, JMS, VS, HE, and PM organized resources. MVM, KW, JD, FG, PT, NJP, JC, CCJ, JMS, VS, and PM curated the data. JD, FG, PT, and BH wrote and prepared the original draft of the manuscript. MVM, KW, TM, JGB, AJ, AKS, SJ, YL, JD, FG, PT, NJP, JC, BH, NEY, CH, CCJ, JMS, VS, TN, LP, HE, and PM wrote, reviewed and/or edited the manuscript. JD, FG, PT, and BH contributed to the visualization. MVM, KW, JD, CH, JMS, and PM supervised the study. The project administrators were BH and NEY. All authors read and agreed to the published version of the manuscript.

Acknowledgments

We thank all patients who participated in the CARTITUDE-1 and LocoMMotion studies, their families and caregivers, the physicians and nurses who cared for patients and supported the studies, as well as staff members of CARTITUDE-1 and LocoMMotion sites, and staff members involved in data collection, data analysis and interpretation. We also thank staff from Eversana Inc. for their support in the development of this work.

Funding

This study was funded by Janssen Pharmaceutica NV and Legend Biotech Inc.

Data-sharing statement

Data used for this study were derived from the CARTITUDE-1 and LocoMMotion studies. CARTITUDE-1 data-sharing is governed by the Janssen Pharmaceutical Companies of Johnson & Johnson’s data-sharing policy that is available online. As noted in the policy, requests for access to the study data can be submitted through Yale Open Data.

References

1. Kyle RA, Rajkumar SV. Criteria for diagnosis, staging, risk stratification and response assessment of multiple myeloma. Leukemia. 2009;23(1):3-9.

2. Raab MS, Cavo M, Delforge M, et al. Multiple myeloma: practice patterns across Europe. Br J Haematol. 2016;175(1):66-76.

3. Benyamini N, Adir S, Gatt ME, et al. Real-life data on the Haematologica | 108

outcome of daratumomab-refractory myeloma patients: multicenter experience. Blood. 2018;132(Suppl_1):3259.

4. Haefliger B, Diels J, Ghilotti F, et al. Baseline characteristics and survival outcomes of patients with tri-exposed multiple myeloma in a German registry. Proceedings of the EHA2021-European Hematology Association Conference. https://oncologyinformationservice.com/wpcontent/uploads/pdf/ Publikationen%20PDFs/Haefliger_et_al_EBMT_EHA_Poster_OIs.pd f Accessed 15 April 2023

5. Ferlay J, Steliarova-Foucher E, Lortet-Tieulent J, et al. Cancer incidence and mortality patterns in Europe: estimates for 40 countries in 2012. Eur J Cancer. 2013;49(6):1374-1403.

6. Usmani SZ, Nahi H, Plesner T, et al. Daratumumab monotherapy in patients with heavily pretreated relapsed or refractory multiple myeloma: final results from the phase 2 GEN501 and SIRIUS trials. Lancet Haematol. 2020;7(6):e447-e455.

7. Miguel JS, Weisel K, Moreau P, et al. Pomalidomide plus lowdose dexamethasone versus high-dose dexamethasone alone for patients with relapsed and refractory multiple myeloma (MM-003): a randomised, open-label, phase 3 trial. Lancet Oncol. 2013;14(11):1055-1066.

8. San-Miguel JF, Hungria VT, Yoon SS, et al. Panobinostat plus bortezomib and dexamethasone versus placebo plus bortezomib and dexamethasone in patients with relapsed or relapsed and refractory multiple myeloma: a multicentre, randomised, double-blind phase 3 trial. Lancet Oncol. 2014;15(11):1195-1206.

9. Orlowski RZ, Nagler A, Sonneveld P, et al. Randomized phase III study of pegylated liposomal doxorubicin plus bortezomib compared with bortezomib alone in relapsed or refractory multiple myeloma: combination therapy improves time to progression. J Clin Oncol. 2007;25(25):3892-3901.

10. Dimopoulos MA, Moreau P, Palumbo A, et al. Carfilzomib and dexamethasone versus bortezomib and dexamethasone for patients with relapsed or refractory multiple myeloma (ENDEAVOR): a randomised, phase 3, open-label, multicentre study. Lancet Oncol. 2016;17(1):27-38.

11. Dimopoulos MA, Chen C, Spencer A, et al. Long-term follow-up on overall survival from the MM-009 and MM-010 phase III trials of lenalidomide plus dexamethasone in patients with relapsed or refractory multiple myeloma. Leukemia. 2009;23(11):2147-2152.

12. Dimopoulos MA, Moreau P, Terpos E, et al. Multiple myeloma: EHA-ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2021;32(3):309-322.

13. Gandhi UH, Cornell RF, Lakshman A, et al. Outcomes of patients with multiple myeloma refractory to CD38-targeted monoclonal antibody therapy. Leukemia. 2019;33(9):2266-2275.

14. Robak P, Drozdz I, Szemraj J, Robak T. Drug resistance in multiple myeloma. Cancer Treat Rev. 2018;70:199-208.

15. Kumar V, Ailawadhi M, Dutta N, et al. Trends in early mortality from multiple myeloma: a population-based analysis. Clin Lymphoma Myeloma Leuk. 2021;21(5):e449-e455.

16. Majzner RG, Mackall CL. Clinical lessons learned from the first leg of the CAR T cell journey. Nat Med. 2019;25(9):1341-1355.

17. Madduri D, Berdeja JG, Usmani SZ, et al. CARTITUDE-1: phase 1b/2 study of ciltacabtagene autoleucel, a B-cell maturation antigen-directed chimeric antigen receptor T cell therapy, in relapsed/refractory multiple myeloma. Blood. 2020;136(Suppl 1):22-25.

18. Berdeja JG, Madduri D, Usmani SZ, et al. Ciltacabtagene

autoleucel, a B-cell maturation antigen-directed chimeric antigen receptor T-cell therapy in patients with relapsed or refractory multiple myeloma (CARTITUDE-1): a phase 1b/2 openlabel study. Lancet. 2021;398(10297):314-324.

19. A study of JNJ-68284528, a chimeric antigen receptor T cell (CAR-T) therapy directed against B-cell maturation antigen (BCMA) in participants with relapsed or refractory multiple myeloma (CARTITUDE-1). 2020.

https://www.clinicaltrials.gov/ct2/show/NCT03548207 Accessed 14 December 2021

20. Mateos M-V, Weisel K, Stefano VD, et al. LocoMMotion: a prospective, non-interventional, multinational study of real-life current standards of care in patients with relapsed/refractory multiple myeloma (RRMM) receiving ≥3 prior lines of therapy. J Clin Oncol. 2021;39(Suppl 15):S8041.

21. Mateos MV, Weisel K, De Stefano V, et al. LocoMMotion: a prospective, non-interventional, multinational study of real-life current standards of care in patients with relapsed and/or refractory multiple myeloma. Leukemia. 2022;36(5):1371-1376.

22. Li F, Thomas LE, Li F. Addressing extreme propensity scores via the overlap weights. Am J Epidemiol. 2019;188(1):250-257.

23. Martin T, Lin Y, Agha M, et al. Health-related quality of life in the CARTITUDE-1 study of ciltacabtagene autoleucel for relapsed/refractory multiple myeloma. Blood. 2020;136(Suppl_1):41-42.

24. Ratcliffe J, Young T, Longworth L, Buxton M. An assessment of the impact of informative dropout and nonresponse in measuring health-related quality of life using the EuroQol (EQ5D) descriptive system. Value Health. 2005;8(1):53-58.

25. Patel D, Grimson F, Mihaylova E, et al. Use of external comparators for Health Technology Assessment submissions based on single-arm trials. Value Health. 2021;24(8):1118-1125.

26. Goring S, Taylor A, Müller K, et al. Characteristics of nonrandomised studies using comparisons with external controls submitted for regulatory approval in the USA and Europe: a systematic review. BMJ Open. 2019;9(2):e024895.

27. Costa LJ, Lin Y, Martin TG, et al. Cilta-cel versus conventional treatment in patients with relapse/refractory multiple myeloma. J Clin Oncol. 2021;39(15_suppl):8030.

28. Weisel K, Martin T, Krishnan A, et al. Comparison of ciltacabtagene autoleucel (cilta-cel) in CARTITUDE-1 versus standard of care in triple-class exposed multiple myeloma patients in clinical trials of daratumumab. Proceedings of the EHA2021-European Hematology Association Conference. Abstract EP977.

29. Martin TG, Krishnan AY, Yong K, et al. Comparison of outcomes with ciltacabtagene autoleucel (cilta-cel) in CARTITUDE-1 versus real-world standard of care (RW SOC) for patients (pts) with triple-class exposed relapsed/refractory multiple myeloma (RRMM). J Clin Oncol. 2021;39(Suppl 15):S8045.

30. Perrot A, Lauwers-Cances V, Tournay E, et al. Development and validation of a cytogenetic prognostic index predicting survival in multiple myeloma. J Clin Oncol. 2019;37(19):1657-1665.

31. Dimopoulos MA, Terpos E, Boccadoro M, et al. Daratumumab plus pomalidomide and dexamethasone versus pomalidomide and dexamethasone alone in previously treated multiple myeloma (APOLLO): an open-label, randomised, phase 3 trial. Lancet Oncol. 2021;22(6):801-812.

32. Merz M, Goldschmidt H, Hari P, et al. Adjusted comparison of outcomes between patients from CARTITUDE-1 versus multiple myeloma patients with prior exposure to PI, IMiD and anti-CD38 from a German Registry. Cancers (Basel). 2021;13(23):5996.

CD169-CD43 interaction is involved in erythroblastic island formation and erythroid differentiation

Jian Bai,1*

1State Key Laboratory of Cancer Biology, Department of Medical Genetics and Developmental Biology, Fourth Military Medical University, Xi’an; 2Department of Aerospace Physiology, Fourth Military Medical University, Xi’an; 3Department of Hematology, The First Affiliated Hospital of Zhengzhou University, Zhengzhou and 4Department of Transfusion Medicine, Xijing Hospital, Fourth Military Medical University, Xi'an, Shaanxi, China

*JB, FF, CG and SL contributed equally as first authors.

Abstract

Correspondence: H. Qin hyqin@fmmu.edu.cn

Received: September 28, 2022.

Accepted: February 23, 2023.

Early view: March 2, 2023.

https://doi.org/10.3324/haematol.2022.282192

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

CD169, a specific marker for macrophages, is a member of the sialic acid-binding immunoglobulin-like lectin (Siglec) family which acts as an adhesion molecule implicated in cell–cell interaction via sialylated glycoconjugates. Although CD169+ macrophages have been found to participate in erythroblastic island (EBI) formation and support erythropoiesis under homeostasis and stress, the exact role of CD169 and its counter receptor in EBI remains unknown. Herein, we generated CD169-CreERT knock-in mice and investigated the function of CD169 in EBI formation and erythropoiesis using CD169-null mice. EBI formation was impaired in vitro by both blockade of CD169 using anti-CD169 antibody and deletion of CD169 on macrophages. Furthermore, CD43 expressed by early erythroblasts (EB) was identified as the counter receptor for CD169 in mediating the EBI formation via surface plasmon resonance and imaging flow cytometry. Interestingly, CD43 was proven to be a novel indicator of erythroid differentiation due to the progressive decrease of CD43 expression as EB mature. Although CD169-null mice did not display defects in bone marrow (BM) EBI formation in vivo, CD169 deficiency impeded BM erythroid differentiation probably via CD43 under stress erythropoiesis, in concert with the role of CD169 recombinant protein in hemin-induced K562 erythroid differentiation. These findings have shed light on the role of CD169 in EBI under steady and stress erythropoiesis through binding with its counter receptor CD43, suggesting that CD169-CD43 interaction might be a promising therapeutic target for erythroid disorders.

Introduction

Erythropoiesis is a complicated progress wherein hematopoietic stem cells proliferate and differentiate into mature red blood cells (RBC) via multiple developmental stages regulated by various factors.1 The last several decades have witnessed the emerging role of macrophages in supporting erythroid development. As the first described hematopoietic niche, the erythroblastic island (EBI) is composed of a central macrophage and surrounding developing erythroblasts (EB), seen in the fetal liver, bone marrow (BM), and spleen.2,3 Central macrophages function as “nursing” cells in this niche, which anchor EB within EBI and provide cytokines and growth factors to promote the proliferation and differentiation of EB.3–7 Moreover, these macrophages can transfer iron to attached EB for heme synthesis and phagocytose the nuclei extruded by EB at terminal differentiation.3–7 Such functions are based on the adhesion of central macrophages

and EB,3–5 but the mechanisms mediating EBI formation have not been fully elucidated. Several adhesion molecule pairs participate in EBI formation, such as erythroblast macrophage protein (EMP)-EMP,8–12 vascular cell adhesion molecule-1 (VCAM-1)-integrin α4β113,14 and integrin αv-intercellular adhesion molecule-4 (ICAM-4).15,16 However, recent studies found that not all of these molecules are essential for EBI formation in vivo. 12 EMP expressed by macrophages, not EB, is necessary for maintaining EBI.12 Another adhesion molecule, VCAM-1, is dispensable for EBI formation in vivo 12 In addition, CD163 expressed by macrophages might be involved in EBI formation in vitro 17 Nevertheless, little is known regarding the interactions between other adhesion molecules mediating macrophages and EB.

CD169, also known as sialoadhesion (Sn) or sialic acidbinding immunoglobulin-like lectin 1 (Siglec1), is a specific marker for tissue resident macrophages.18,19 As the largest member of the Siglec family, CD169 contains an extracel-

Fan Fan,1* Chunchen Gao,1* Shaohua Li,2* Wei Li,3 Tiaoxia Wei,1 Shilin Cheng,1 Jinmin Yu,1 Chao Zheng,1 Junlong Zhao,1 Linru Zou,1 Lei Feng,1 Jing Yi4 and Hongyan Qin1

lular region made up of 16 C2-set immunoglobulin domains and 1 V-set immunoglobulin domain but lacks intracellular signaling motifs.18,19 CD169 was originally named sheep erythrocyte receptor because of its ability to interact with sialylated structures on sheep erythrocytes, but later it was found to bind murine EB probably via sialylated glycoconjugates which is located at macrophagesEB contact zones.20–22 Further in vivo investigations have shown that depletion of CD169+ macrophages strikingly impairs EBI and markedly decreases the number of EB in BM, suggesting that CD169+ macrophages are crucial for EBI formation in vivo. 23 In addition, the contribution of CD169+ macrophages to stress and pathological erythropoiesis suggests that targeting CD169+ macrophages has therapeutic implications in anemia, polycythemia vera, and β-thalassemia.23,24 These studies emphasize the pivotal role of CD169+ macrophages in supporting erythropoiesis under homeostasis and stress. A recent study has reported that EPOR+ EBI central macrophages express a higher level of CD169 than EPOR macrophages.25 However, the mechanism of CD169 as an adhesion molecule for mediating macrophages and EB interaction remains unclear. Furthermore, the counter receptor on EB for CD169 has not been studied.

In this study, we found that CD43 is the counter receptor on EB for CD169 expressed by macrophages in EBI. In addition, similar to CD44, CD43 expression progressively decreased as EB mature, and thus it could be a novel marker to distinguish erythroid differentiation. Although blocking CD169 disrupted the adhesion between macrophages and EB in vitro, CD169 deletion did not impair EBI formation in vivo, suggesting that CD169 might play a dispensable role in the BM EBI niche. Moreover, in the model of high-altitude polycythemia (HAPC), CD169 might slightly promote BM erythroid differentiation through binding with CD43. Consistently, CD169 recombinant protein promoted hemin-induced K562 erythroid differentiation mildly with reduced CD43 expression. Collectively, our findings unravel the counter receptor on EB for CD169 and elucidate the role of CD169 expressed by macrophages in EBI formation and erythroid differentiation under homeostasis and stress conditions, replenishing the molecular mechanisms of adhesion molecules participating in EBI formation.

Methods

Mice

CD169-CreERT mice were generated on a C57BL/6 background by using CRISPR/Cas9 to knock CreERT into the exon 1 of CD169 (Biocytogen Biotechnology Company, China). Genotyping was performed by polymerase chain reaction (PCR) with two pairs of primers: WT-F: CATGCCACCAAGTGAGAG-

CATTTCC; WT-R: TGCACATTCTTGGGACTGGAGACACC; MUT-F: AGGAGACCAATTTCCGGTGCTTACG; MUT-R: GGCTTGCAGGTACAGGAGGTAG. The first pair of primers amplified a 539 bp wild-type (WT) band while the second pair amplified a 641 bp mutant (MUT) band from the CD169-CreERT allele. Rosa26tdTomato (Ai9) mice were kindly provided by Dr. Yuqiang Ding. In order to induce gene recombination in CD169-CreERT crossed with Ai9 mice, tamoxifen (100 mg/kg, Sigma) was administered intraperitoneally daily for 5 consecutive days, and analysis was performed 3 weeks after the last injection. In order to establish a high-altitude polycythemia (HAPC) model,26 mice were exposed to hypoxia in a hypobaric chamber corresponding to an altitude of 5,000 m for 7 days. All mice with a C57BL/6 background were maintained in a specific-pathogen-free facility. All animal experiments were approved by the Animal Experiment Administration Committee of the Fourth Military Medical University. Adult mice 2–3 months of age were used for the studies.

Cell isolation and culture

Murine BM-derived macrophages (BMDM) were induced and cultured as described previously.27 The BM EB (CD45CD11b-Ter119+) were sorted by fluorescence activated cell sorting (FACS) using BD AriaIII or magnetic activated cell sorting (MACS) with the BD IMagTM cell separation system.

Erythroblastic island formation in vitro

EBI formation was performed as previously described. 25 Briefly, BMDM were pretreated with isotype or antiCD169 antibody (5 μ g/mL) for 1 hour (h), and were then washed with phosphate-buffered saline (PBS). Pretreated BMDM were mixed with sorted EB at a 1:20 ratio for 12 h in Iscove’s modified Dulbecco’s medium (IMDM) containing 10% fetal bovine serum (FBS), 5 mM Mg2+, and 5 mM Ca2+ . Afterwards, cells were collected for cytospin and May–Grunwald–Giemsa (MGG) staining to assess the ability of EBI formation in vitro . In some cases, EB were pretreated with isotype or anti-CD43 antibody for 1 h before co-culture or BM macrophages were sorted from CD169 WT and knockout (KO) mice using BD AriaIII or SONY MA900.

Preparation of native erythroblastic islands from bone marrow

Native murine BM EBI were prepared according to the described methods,25,28 then stained with different antibodies listed in the Online Supplementary Table S1. Cells were washed and resuspended for flow cytometry or imaging flow cytometry analysis.

Flow cytometry and imaging flow cytometry

Isolated murine BM and spleen cells were stained with different antibodies listed in the Online Supplementary

Table S1. Flow cytometry analyses were performed with routine protocols using a BD CantoPlus flow cytometer. Dead cells were excluded by 7-AAD staining. Data was analyzed using the FlowJo 7.6.1 software. Imaging flow cytometry was done using a Luminex Amnis Imagestream Mk II instrument at 40× magnification with IDEAS 6.2 software for data analyses.

Surface plasmon resonance

BIAcore T200 (Cytiva) was used to determine the binding ability of CD169 to CD43. Based on the binding active unit of rmSiglec1-mFc (RD, 5610-SL) in N-terminal Ig-like V-set domain, mouse antibody capture kit (Cytiva, BR-1008-38) was utilized for rmSiglec1-mFc immobilization on a CM3 sensor chip via the C-terminal rmFc domain. Immobilization conditions and running conditions were set according to the mouse antibody capture kit instructions, and rmCD43-hFc (SinoBiological, 50735-M02H) was injected at a series of concentrations diluted with HBS-EP buffer. The association time was 60 seconds (sec), and the dissociation time was also 60 sec. The equilibrium dissociation constant (KD) value was obtained using a steady state affinity model on the BIAcore evalution software program.

Statistics

The data were quantitatively analyzed using the GraphPad Prism 7 software. All experiments were replicated at least three times and all data were reported as mean ± standard error of the mean (SEM). Statistical significance was assessed using the Student’s t-test and one-way ANOVA with Tukey’s multiple comparison tests, as appropriate. P<0.05 was considered statistically significant.

Results

Blockade of CD169 expression on bone marrow-derived marcrophages disrupts erythroblastic island formation in vitro

CD169+ macrophages are reportedly involved in BM EBI formation.23 In order to further assess the exact role of CD169 on EBI macrophages, we first examined CD169 expression during murine BM monocyte differentiation into macrophages at different time points. CD169 expression was found to increase at both the RNA and protein levels (Figure 1A–C) in the process of monocyte-to-macrophage differentiation, which is in line with the differentiation of healthy human peripheral blood monocytes to macrophages.29 Meanwhile, the expression of CD169 on mature BMDM was confirmed by immunofluorescence staining (Figure 1D). Afterwards, we developed an in vitro EBI formation assay (Online Supplementary Figure S1A) to evaluate adhesion between CD169+ BMDM and BM EB. BM EB were sorted (Online Supplementary Figure S1B) and

then co-cultured with BMDM pretreated with isotype or anti-CD169 antibody. Representative images and quantitative analyses showed that the ability of EBI formation in vitro was impaired when the expression of CD169 on BMDM was blocked by anti-CD169 antibody (Figure 1E, F). These results demonstrated the crucial role of CD169 in EBI formation in vitro

Regarding the lack of VCAM-1 on CD169+ BMDM in vitro (Online Supplementary Figure S1C), it is reasonable to speculate whether overexpression of VCAM-1 on CD169+ BMDM may rescue EBI formation upon blocking CD169. In order to prove this hypothesis, VCAM-1 was induced in cultured CD169+ BMDM by lentivirus overexpression (Online Supplementary Figure S1D). Unexpectedly, the result showed that overexpression of VCAM-1 on CD169+ BMDM might not rescue the impairment of EBI formation upon blocking CD169 in vitro, indicating that VCAM-1 may not compensate for the effect of CD169 deletion in EBI formation at least in vitro (Online Supplementary Figure S1E, F).

Generation of CD169-CreERT mice to investigate the role of CD169 in erythroblastic island formation

In order to enunciate the role of CD169 in EBI, we constructed CD169-CreERT knock-in mice on a C57BL/6 background using CRISPR/Cas9 technology. As shown in the schematic model, CreERT was knocked into exon 1 to allow the independent expression of CreERT under the control of the CD169 promoter (Figure 2A), and the genotype was confirmed by PCR (Online Supplementary Figure S2A). Meanwhile, the different genotype of CD169-CreERT mice was determined by FACS using BM cell suspension. As shown in Figure 2B and C, CD169 expression was reduced by about half in heterozygotes, and CD169 was knocked out successfully in CD169-CreERT double knockin mutant mice (CD169 KO mice). Further analysis revealed that CD169 deletion did not affect BM macrophage percentage and numbers (Online Supplementary Figure S2B). Because CD169+ macrophages maintain the hematopoietic stem cells niche,30,31 we then examined the hematopoietic stem cell development by FACS. The percentage and cell numbers of hematopoietic stem and progenitor cells showed negligible changes in CD169 KO mice (Online Supplementary Figure S2C, D).

Moreover, we observed CD169 distribution in immune cells by crossing CD169-CreERT mice and Rosa26tdTomato mice with tdTomato followed by the loxp-flanked STOP sequence inserted into the Rosa26 locus. After tamoxifen administration, tdTomato expression was observed in the EBI multiplets (Figure 2D), but not in T cells, B cells, or NK cells (Online Supplementary Figure S2E), suggesting that CD169 is a macrophage-restricted marker and that CD169-CreERT could be utilized for tracing EBI central macrophages. However, the percentage of tdTomato in EBI central macrophages was not as high as we

Figure 1. Blockade of CD169 expression on bone marrow-derived macrophages disrupts erythroblastic island formation in vitro. (A)

The relative expression of Cd169 was determined by quantitative real-time polymerase chain reaction (qRT-PCR) at different time points of bone marrow-derived macrophages (BMDM) differentiation in vitro (N=3). (B) The percentage of CD11b+F4/80+ macrophages (upper panel) and the CD169 expression of CD11b+F4/80+ macrophages (lower panel) were analyzed by fluorescence activated cell sorting (FACS) during BMDM differentiation. The representative FACS plot is shown (N=3). (C) The percentage of CD169+ macrophages was quantitatively analyzed as shown in (B) (N=3). (D) BMDM on day 7 (D7) were stained with CD169 (green), F4/80 (red), and Hoechst (blue) using immunofluorescence staining (N=3). (E) Sorted BM erythroblasts (EB) were cultured with BMDM pretreated with immunoglobulin G (IgG) isotype or anti-CD169 antibody, then in vitro EBI formation was examined by May–Grunwald–Giemsa (MGG) staining (N=3). (F) Quantitative analyses of in vitro erythroblastic island (EBI) formation by counting the number of surrounding EB associated with BMDM in (E) (N= 3). Data are shown as mean ± standard error of the mean; *P<0.05; **P<0.01; ***P<0.001.

expected, probably due to the efficiency of tamoxifen induction.

We then sorted BM macrophages from WT and CD169 KO mice as well as BM EB from WT mice, and performed EBI formation in vitro. EBI formation was undermined in vitro in the CD169 KO group (Figure 2E, F), which resembled the impairment of EBI by blocking CD169 using anti-CD169 antibody (Figure 1E, F). Based on the abovementioned results, CD169-CreERT double knock-in mice could be utilized as CD169 KO mice for studying CD169 function in EBI formation in vivo.

CD43 is the counter receptor on erythroblasts for CD169

As the counter receptor on EB for CD169 in EBI has not been determined, we attempted to determine its potential counter receptor from some reported sialylated mol-

ecules interacting with CD169, such as MUC1 and CD43.32,33 Since the counter receptor candidate is likely to share a similar proportion with that of recombinant CD169-Fc protein binding to BM EB, we first measured the binding of recombinant mouse CD169-Fc protein to sorted BM EB. In fact, recombinant CD169-Fc bound only to about 10% of BM EB (Figure 3A; Online Supplementary Figure S1G). Next, we examined the expression of MUC1 and CD43 on EB by western blotting. We found that CD43, but not MUC1, was expressed on sorted BM EB (Online Supplementary Figure S1H). Accordingly, we also observed approximately 16% CD43+ BM EB by FACS (Figure 3B), indicating that CD43 might be the counter receptor for CD169 on EB.

For further insight regarding the interaction between CD169 and CD43, surface plasmon resonance was per-

Figure 2. Generation of CD169-CreERT mice for investigating the function of CD169 in erythroblastic island formation. (A) Schematic model for the generation of CD169-CreERT knock-in mice. (B) The expression of CD169 in bone marrow (BM) macrophages from CD169 wild-type (WT), heterozygous (Het), and mutant (Mut) mice was analyzed by fluorescence activated cell sorting (FACS) (N=3). (C) The deletion efficiency of CD169 on BM macrophages was determined by mean fluorescence intensity (MFI) as shown in (B) (N=3). (D) Representative flow cytometry plots of BM erythroblasts (EB) gated as Ter119+F4/80+ live multiplets and representative histograms showing tdTomato expression of erythroblastic island (EBI) central macrophages in CD169-CreERT / Rosa26tdTomato+/ and CD169-CreERT+/ Rosa26tdTomato+/ mice (N=3). (E) Sorted BM EB were cultured with BM macrophages sorted from CD169 WT and KO mice, followed by May–Grunwald–Giemsa (MGG) staining for evaluating in vitro EBI formation (N=3). (F) Quantitative analyses of in vitro EBI formation by counting the number of surrounding EB associated with BM macrophages from CD169 WT and KO mice as shown in (E) (N=3). Data are shown as mean ± standard error of the mean; *P<0.05; **P<0.01; ***P<0.001.

formed to examine the binding affinity of CD169 to CD43. As shown in Figure 3C, their binding was dose-dependent and exhibited a fast association-dissociation process with a KD value of 107.1 nM. Meanwhile, imaging flow cytometry analysis showed that in a typical EBI structure one CD169+F4/80+ central macrophage was surrounded by sev eral Ter119+ EB, of which one or two were CD43+ (Figure 3D). Furthermore, blockade of CD43 on BM EB with anti-CD43 antibody disrupted EBI formation in vitro (Figure 3E, F), which was consistent with the effect of CD169

blockade and CD169 deletion on macrophages (Figure 1E, F; Figure 2E, F). Taken together, our findings strongly suggest that CD43 is the counter receptor on EB for CD169 in EBI.

CD43 might be a novel marker for distinguishing erythroid differentiation

Given that not all EB express CD43, we next investigated whether the kinetics of CD43 expression are relevant to the specific developmental stages of erythroid cells. We

Figure 3. The counter receptor on erythroblasts for CD169 is CD43. (A) The binding ability of recombinant mouse CD169-Fc protein to sorted bone marrow (BM) erythroblasts (EB) was detected by fluorescence activated cell sorting (FACS) with recombinant mouse Fc protein as control stained by anti-mouse CD169 antibody (N=7). (B) The CD43 expression of CD45 CD11b Ter119+ BM EB was analyzed by FACS (N= 12). (C) The binding affinity of CD169 to CD43 was determined by surface plasmon resonance (SPR). (D) The typical erythroblastic island (EBI) structure was observed by using imaging flow cytometry (IFC): CD169+F4/80+ macrophage was surrounded by several Ter119+ EB, among which 1 to 2 EB were CD43+. (E) Representative May–Grunwald–Giemsa (MGG) staining images of in vitro EBI formation between sorted BM macrophages and sorted BM EB pretreated with isotype or anti-CD43 antibody (N=3). (F) Quantitative analyses of in vitro EBI formation by counting the number of surrounding EB pretreated with isotype or anti-CD43 antibody associated with macrophages (N=3). Data are shown as mean ± standard error of the mean; *P<0.05; **P<0.01; ***P<0.001.

adopted a putative method to identify distinct EB populations34,35 and examined the expression of CD43 in different developmental stages. During the successive differentiation progress from immature EB to mature RBC, CD43 displayed a progressively decreasing pattern from region I (highest level) to region V (lowest level) (Figure 4A). In addition, the difference among the mean fluorescence in-

tensity (MFI) of CD43 in distinct EB population was statistically significant (Online Supplementary Figure S3A), and the MFI of CD43 was positively correlated with the MFI of CD44 (Online Supplementary Figure S3B). Consistently, imaging flow cytometry also revealed the progressive decrease of CD43 fluorescence as well as CD44 from Pro-EB to RBC (Figure 4B). Then we used CD43 and FSC-A to sep-

Figure 4. CD43 may be a novel marker for distinguishing erythroid differentiation. (A) Representative histograms showing CD43 expression of bone marrow (BM) erythroblasts (EB) at various stages gated by CD44 and FSC-A (I, pro-EB; II, basophilic EB; III, polychromatic EB; IV, orthochromatic EB and reticulocytes; V, mature red blood cells [RBC]). (B) Representative images of imaging flow cytometry (IFC) analysis showing the expression trend of CD43 as well as changes in cell size, CD44, and Ter119 from pro-EB to RBC (C) Representative gating strategy using CD43 and FSC-A of CD45 CD11b Ter119+ BM EB and May–Grunwald–Giemsa (MGG) staining images of EB cytospins sorted from distinct clusters. (D) The expression level of CD43 during human erythroid differentiation was analyzed from published transcriptomic data (poly vs. ortho false discovery rate [FDR] =0.001). (E) CD43 expression during human erythroid differentiation from published proteomic analysis (ANOVA significant–FDR <0.05). RPKM: reads per kilobase million.

arate EB at distinct stages and observed similar morphology of EB staged by CD44 (Figure 4C). Moreover, both published transcriptomic36 and proteomic37 data proved that CD43 expression decreased during human erythroid differentiation (Figure 4D, E). Similarly, the protein level of CD43 progressively decreased in K562 cells induced by hemin towards erythroid differentiation, whereas the positive cell rate of benzidine staining and erythroid surface marker CD235a expression increased gradually (Online Supplementary Figure S3C–F). Collectively, these

results indicate that CD43 might be a novel surface marker to distinguish erythroid differentiation.

CD169 plays a dispensable role in the bone marrow erythroblastic island niche in vivo

In order to further dissect the role of CD169 in BM EBI in vivo, we evaluated EBI formation and erythroid differentiation in CD169 WT and KO mice. Unexpectedly, we neither observed paler bones (Figure 5A) nor striking defects of BM EBI percentages in CD169 KO mice versus WT mice

Figure 5. CD169 plays a dispensable role in the bone marrow erythroblastic island niche in vivo. (A) Representative photograph of femurs dissected from CD169 wild-type (WT) and knockout (KO) mice. (B) CD169 expression bone marrow (BM) erythroblastic islands (EBI) gated as Ter119+F4/80+ live multiplets was determined by fluorescence activated cell sorting (FACS) from CD169 WT and KO mice. (C) The percentage of BM EBI as shown in (B) in CD169 WT and KO mice was quantitatively compared. (D) Representative flow cytometry gating strategy of BM EB from CD169 WT and KO mice. (E) The percentage and cell numbers of BM EB as shown in (D) were quantitatively compared between CD169 WT and KO mice. (F) The percentages of BM erythroblasts (EB) at distinct stages of EB differentiation in (D) were quantitatively compared (N=3). Data are shown as mean ± standard error of the mean.

(Figure 5B, C). There was no significant difference in BM EB quantification between CD169 WT and KO mice, except for marginally reduced BM EB numbers in CD169 KO mice (Figure 5D, E). Moreover, CD169 WT versus KO mice showed almost no difference in BM EB maturation measured by CD44 (Figure 5F). Furthermore, CD43 expression of total EB in BM was not different, either (Online Supplementary Figure S4A). Notably, CD169 deletion did neither influence the expression of other adhesion molecules, such as VCAM-1 and integrin αv on BM EBI central macrophages (Online Supplementary Figure S4B), nor integrin α4 and integrin β1 on BM EB (Online Supplementary Figure S4C, D), implying that other adhesion molecules might compensate for the effect of CD169 deletion in EBI formation. In addition, spleen EB and macrophages were not significantly altered and the peripheral RBC compartments at steady state were also normal (Online Supplementary Figure S4E–M).

Next, in order to verify whether CD169 deficiency would affect the nursing cell phenotype of BM EBI macrophages, we compared the expression level of genes supporting erythropoiesis between CD169 WT and KO BM macrophages, such as adhesion molecules, nucleus engulfment, iron handling, transcription factors, growth factors, and membrane receptors according to the literature25 and our RNA sequencing data (data not shown). Overall, almost no significant difference was observed on the nursing cell phenotype of EBI macrophages mentioned above except different expression levels of Slc40a1 and Spic between CD169 WT and KO group, indicating that CD169 deletion might not affect the EBI nursing macrophage phenotype (Online Supplementary Figure S5).

Taken together, CD169 plays a dispensable role in the BM EBI niche in vivo, resembling the function of VCAM-1 in BM EBI formation in vivo as reported by Wei Q et al 12

CD169 slightly promotes bone marrow erythroid differentiation probably via CD43 in high-altitude polycythemia

Next, we investigated the functional role of CD169 in stress erythropoiesis. Similar to the stress erythropoiesis model induced by erythropoietin (EPO) treatment,38 the HAPC model was chosen as our experimental model owing to the hypoxia-induced increase in EPO levels. After 7 days of exposure to high altitude corresponding to 5,000 m, mice exhibited an increase in RBC production with elevated BM EBI and EB, as well as enhanced spleen erythropoiesis (Online Supplementary Figure S6). We then induced HAPC in CD169 WT and KO mice and observed slightly paler bones in CD169 KO HAPC mice (Online Supplementary Figure S7A). CD169 KO HAPC mice showed a slight decrease in BM EBI (Figure 6A, B) and BM EB quantification with no significance (Figure 6C, D). However, further analysis to distinguish the EB subpopulation revealed that the per-

centage of regions I and II increased, while the percentage of region V decreased in CD169 KO HAPC mice (Figure 6E). These results indicate a partial blockade of differentiation at a relatively early EB stage in CD169 KO HAPC mice. Additionally, a small but significant increase in the CD43 level of BM total EB was observed in CD169 KO HAPC mice (Figure 6F). The BM macrophages and monocytes of CD169 KO HAPC mice did not expand significantly versus controls (Online Supplementary Figure S7B, C). Although stress erythropoiesis primarily occurs in the spleen,39,40 in CD169 KO HAPC mice, mild splenomegaly (Online Supplementary Figure S7D, E) was accompanied by a faint increase in spleen EB (Online Supplementary Figure S7F-I) and slightly augmented spleen macrophages (Online Supplementary Figure S7J, K), but no alterations in the peripheral RBC parameters (Online Supplementary Figure S7L).

We then validated whether CD169 could promote erythroid differentiation in hemin-induced K562 cells. We observed a higher positive cell rate in benzidine staining in the hemin plus CD169 group during the fi rst 2 days (Online Supplementary Figure S8A) and lower expression of CD43 in the hemin plus CD169 group versus the hemin group ( Online Supplementary Figure S8C, D). However, CD235a percentage showed no remarkable changes (Online Supplementary Figure S8B), in accordance with no prominent differences in the BM EB percentage between CD169 WT and KO HAPC mice, suggesting that CD169 only weakly promotes erythroid differentiation. In contrast, CD43 expression exhibited significant alterations in both the HAPC mice model and hemin-induced K562 cells model.

Discussion

EBI are specialized niches for erythropoiesis, consisting of a central macrophage surrounded by developing EB.2,3 EBI central macrophages, characterized by specific markers and transcriptional factors,25,41–44 nurse and support erythroid development through adhesion molecule interaction pairs.3–7 Previously, adhesion molecules were validated in vitro by neutralizing antibodies to elucidate their function in mediating macrophages-EB interaction.8,9,13 However, the latest study on this declared that not all adhesion molecules are indispensable for EBI formation in vivo using conditional gene KO mice,12 implicating the need to comprehensively revisit and reassess the functional role of adhesion molecules in EBI formation. CD169 is suggested to be such an adhesion molecule participating in this progress,4,5,25,45 but its exact role and counter receptor in the EBI have not been deeply explored. In the current study, we find that CD169 exerts functions on EBI central macrophages by interacting with CD43 on EB, which is a novel marker to distinguish ery-

Figure 6. CD169 slightly promotes bone marrow erythroid differentiation probably via CD43 in high-altitude polycythemia. CD169 wild-type (WT) and knockout (KO) mice were exposed to hypoxia in a hypobaric chamber (altitude: 5,000 m) for 7 days to establish high-altitude polycythemia (HAPC) models. (A) CD169 expression in bone marrow (BM) erythroblastic island (EBI) was determined by fluorescence activated cell sorting (FACS) from CD169 WT and KO HAPC mice, respectively. (B) The percentage of BM EBI as shown in (A) in CD169 WT and KO HAPC mice was quantitatively compared (N=6). (C) Representative flow cytometry gating strategy of BM erythroblasts (EB) from CD169 WT and KO HAPC mice. (D) The percentage and cell numbers of BM EB as shown in (C) were quantitatively compared between CD169 WT and KO HAPC mice (N=6). (E) The percentage of BM EB at different stages of EB differentiation in (C) was quantitatively compared (N=6). (F) CD43 expression of CD45 CD11b Ter119+ BM EB as shown in (C) was detected by FACS from CD169 WT and KO HAPC mice, and the mean fluorescense intensisty (MFI) of CD43 was quantitatively compared (N=4). Data are shown as mean ± standard error of the mean; *P<0.05; **P<0.01.

throid differentiation. This CD169-CD43 interaction might promote BM EB differentiation in stress erythropoiesis (Figure 7). Although CD169 deletion may not affect the EBI nursing macrophage phenotype with the exception of Slc40a1 and Spic, which are reportedly involved in ironrecycling,46 the role of CD169 in scavenging extruded EB nuclei and iron-recycling still awaits further investigation. This study evaluated the function of CD169 in EBI formation through both in vitro and in vivo studies. Blocking CD169 and CD169 deletion could impair EBI formation in vitro, but CD169 KO mice did not display defects in BM EBI function in vivo, indicating that other adhesion molecules may compensate for the effect of CD169 deletion in BM EBI formation in vivo. Furthermore, in vitro cultured murine BMDM lack VCAM-1 expression, which is consistent with the report of Li W et al. in which VCAM-1 expression is not observed in umbilical cord blood CD34+ cell-derived macrophages in vitro. 25 Unexpectedly, overexpression of VCAM1 on CD169+ BMDM might not rescue the impairment of EBI formation upon blocking CD169 in vitro. However, whether VCAM-1 can synergize with CD169 on nursing

macrophages for EBI formation in vivo still needs more experiments to verify.

The counter receptor on EB for CD169 in EBI formation has been unknown in this field. MUC1 is anticipated to bind CD169, as evidenced from the study of human erythroleukemia K562 cells.32 However, murine BM EB do not express MUC1. Therefore, we turned to another possible candidate, CD43 (also called sialophorin [Spn]), which was originally identified as the counter receptor for CD169 on T cells.33 It has been reported that CD43 is expressed in hematopoietic cells, including T lymphocytes, B-cell subpopulations, and natural killer cells,47 but it had so far remained unclear whether CD43 is expressed by EB. We confirmed the expression of CD43 on BM EB and found that the proportion of CD43+ BM EB bore a striking similarity to that of recombinant CD169 protein binding to BM EB. Using multiple approaches, we have identified CD43 as the counter receptor for CD169 in the EBI. Our current results strongly authenticate the new finding that CD169 on macrophages and CD43 expressed by both classical and immune-prone erythroid cells are involved in the EBI based

Figure 7. Schematic model of CD169–CD43 interaction in bone marrow erythroblastic island formation and erythroid differentiation. CD43 is identified as the counter receptor on erythroblasts (EB) for CD169 expressed by macrophages in erythroblastic island (EBI). As bone marrow (BM) EB mature, they gradually reduce and even lose CD43 expression, and their binding abilities to CD169 correspondingly attenuate in steady erythropoiesis (left). However, in high-altitude polycythemia (HAPC) stress erythropoiesis, CD169 deficiency impedes BM erythroid differentiation probably via CD43, with the increased percentage of pro-EB and baso-EB and decreased percentage of erythrocytes (right).

on single-cell RNA sequencing combined with CellPhoneDB analysis.48 Additionally, we observed a progressive decrease of CD43 expression during murine EB maturation similar to the decrease of CD44. Moreover, both published transcriptomic36 and proteomic37 data also showed reduced CD43 expression during human erythroid differentiation, which needs further proof to determine whether CD43 could indicate human BM erythroid differentiation, especially by using human BM cell samples coupled with integrin α4 and band 3 antibodies.49 Collectively, these findings suggest the potential of CD43 as a new marker for distinguishing erythroid maturation. However, it remains to be addressed whether CD43 deletion affects erythroid differentiation by generating CD43 KO mice.

Although CD169 plays a dispensable role in the EBI formation in vivo, CD169 deletion impedes erythroid differentiation under stress erythropoiesis, as seen in our mice HAPC model. In line with this, in a burn injury-associated anemia model, Hasan et al. highlighted that stagnant early EB of BM may be attributed to CD169 downregulation of EBI central macrophages, leading to the retardation of late EB and reticulocyte differentiation.50 Our findings provide further evidence to uncover the underlying mechanism behind the phenomenon that CD169 deletion or downregulation of macrophages hinders BM erythroid differentiation probably via upregulated CD43 on EB in stress erythropoiesis. However, in our HAPC model, compensatory splenic erythropoiesis of CD169 KO mice might mask

the BM deficit. Similarly, recombinant CD169 protein could enhance hemin-induced K562 erythroid differentiation, along with reduced CD43 expression. However, the mechanism of CD169 promoting erythroid differentiation remains to be determined. Combined with previous studies using CD169-DTR mice in a polycythemia vera model,23 our current findings may indicate that the effect of CD169 molecule deletion in stress erythropoiesis is weaker than that of CD169+ marcrophage depletion.

In summary, we have demonstrated that CD169 participates in the interaction between macrophages and EB. CD43 is also identified to be the counter receptor on EB for CD169 expressed by EBI central macrophages. More importantly, CD43 might be a novel indicator of erythroid differentiation and maturation. Despite the dispensable role of CD169 in EBI formation in vivo, CD169 could promote erythroid differentiation probably via CD43 under stress erythropoiesis, which might be a promising target for the treatment of erythroid disorders.

Disclosures

No conflicts of interest to disclose.

Contributions

JB, FF, CG and SL performed most of the experiments. TW, SC, JY and LZ helped perform the FACS experiments. CZ and JZ performed data analyses WL, JY, and LF provided technical support. HQ supervised the study. JB and HQ

drafted the manuscript. All authors discussed the results and contributed to the manuscript.

Acknowledgments

We would like to thank the Military Medical Innovation Center, Fourth Military Medical University for providing SPR and IFC instruments. We thank Jinmei Xu and Ning An from the Department of Transfusion Medicine, Xijing Hospital, Fourth Military Medical University for technical assistance. We also thank Tingting Liu, Ying He, Xingyue An, and Ting Bian from the Department of Neurology, Xijing Hospital, Fourth Military Medical University for cytospin and MGG staining.

References

1. Palis J. Primitive and definitive erythropoiesis in mammals. Front Physiol. 2014;5:3.

2. Bessis M. Erythroblastic island, functional unity of bone marrow. Rev Hematol. 1958;13(1):8-11.

3. Chasis JA, Mohandas N. Erythroblastic islands: niches for erythropoiesis. Blood. 2008;112(3):470-478.

4. Li W, Guo R, Song Y, Jiang Z. Erythroblastic island macrophages shape normal erythropoiesis and drive associated disorders in erythroid hematopoietic diseases. Front Cell Dev Biol. 2020;8:613885.

5. Klei TRL, Meinderts SM, van den Berg TK, van Bruggen R. From the cradle to the grave: the role of macrophages in erythropoiesis and erythrophagocytosis. Front Immunol. 2017;8:73.

6. Lévesque J-P, Summers KM, Bisht K, Millard SM, Winkler IG, Pettit AR. Macrophages form erythropoietic niches and regulate iron homeostasis to adapt erythropoiesis in response to infections and inflammation. Exp Hematol. 2021;103:1-14.

7. May A, Forrester LM. The erythroblastic island niche: modeling in health, stress, and disease. Exp Hematol. 2020;91:10-21.

8. Hanspal M, Hanspal JS. The association of erythroblasts with macrophages promotes erythroid proliferation and maturation: a 30-kD heparin-binding protein is involved in this contact. Blood. 1994;84(10):3494-3504.

9. Hanspal M, Smockova Y, Uong Q. Molecular identification and functional characterization of a novel protein that mediates the attachment of erythroblasts to macrophages. Blood. 1998;92(8):2940-2950.

10. Soni S, Bala S, Gwynn B, Sahr KE, Peters LL, Hanspal M. Absence of erythroblast macrophage protein (Emp) leads to failure of erythroblast nuclear extrusion. J Biol Chem. 2006;281(29):20181-20189.

11. Javan GT, Salhotra A, Finley SJ, Soni S. Erythroblast macrophage protein (Emp): past, present, and future. Eur J Haematol. 2018;100(1):3-9.

12. Wei Q, Boulais PE, Zhang D, Pinho S, Tanaka M, Frenette PS. Maea expressed by macrophages, but not erythroblasts, maintains postnatal murine bone marrow erythroblastic islands. Blood. 2019;133(11):1222-1232.

13. Sadahira Y, Yoshino T, Monobe Y. Very late activation antigen 4vascular cell adhesion molecule 1 interaction is involved in the formation of erythroblastic islands. J Exp Med. 1995;181(1):411-415.

14. Hamamura K, Matsuda H, Takeuchi Y, Habu S, Yagita H, Okumura

Funding

This work was supported by grants from the National Natural Science Foundation of China (81530018, 31970829, 82102224, and 82173082); Shaanxi Science and Technology Program (2020ZDLSF03-05); Shaanxi Science and Technology Innovation Team Program (2021TD-36).

Data-sharing statement

All data generated in this study are included in the present article and its Online Supplementary Appendix, which are available from the corresponding author on reasonable request.

K. A critical role of VLA-4 in erythropoiesis in vivo. Blood. 1996;87(6):2513-2517.

15. Lee G, Lo A, Short SA, et al. Targeted gene deletion demonstrates that the cell adhesion molecule ICAM-4 is critical for erythroblastic island formation. Blood. 2006;108(6):2064-2071.

16. Lee G, Spring FA, Parsons SF, et al. Novel secreted isoform of adhesion molecule ICAM-4: potential regulator of membraneassociated ICAM-4 interactions. Blood. 2003;101(5):1790-1797.

17. Fabriek BO, Polfliet MMJ, Vloet RPM, et al. The macrophage CD163 surface glycoprotein is an erythroblast adhesion receptor. Blood. 2007;109(12):5223-5229.

18. O’Neill ASG, van den Berg TK, Mullen GED. Sialoadhesin - a macrophage-restricted marker of immunoregulation and inflammation. Immunology. 2013;138(3):198-207.

19. Macauley MS, Crocker PR, Paulson JC. Siglec-mediated regulation of immune cell function in disease. Nat Rev Immunol. 2014;14(10):653-666.

20. Crocker PR, Gordon S. Properties and distribution of a lectinlike hemagglutinin differentially expressed by murine stromal tissue macrophages. J Exp Med. 1986;164(6):1862-1875.

21. Crocker PR, Gordon S. Mouse macrophage hemagglutinin (sheep erythrocyte receptor) with specificity for sialylated glycoconjugates characterized by a monoclonal antibody. J Exp Med. 1989;169(4):1333-1346.

22. Crocker PR, Werb Z, Gordon S, Bainton DF. Ultrastructural localization of a macrophage-restricted sialic acid binding hemagglutinin, SER, in macrophage-hematopoietic cell clusters. Blood. 1990;76(6):1131-1138.

23. Chow A, Huggins M, Ahmed J, et al. CD169+ macrophages provide a niche promoting erythropoiesis under homeostasis and stress. Nat Med. 2013;19(4):429-436.

24. Ramos P, Casu C, Gardenghi S, et al. Macrophages support pathological erythropoiesis in polycythemia vera and βthalassemia. Nat Med. 2013;19(4):437-445.

25. Li W, Wang Y, Zhao H, et al. Identification and transcriptome analysis of erythroblastic island macrophages. Blood. 2019;134(5):480-491.

26. Li P, Huang J, Tian H, Huang Q, Jiang C, Gao Y. Regulation of bone marrow hematopoietic stem cell is involved in highaltitude erythrocytosis. Exp Hematol. 2011;39(1):37-46.

27. Ma P-F, Gao C-C, Yi J, et al. Cytotherapy with M1-polarized macrophages ameliorates liver fibrosis by modulating immune microenvironment in mice. J Hepatol. 2017;67(4):770-779.

28. Cao W, Fan W, Wang F, et al. GM-CSF impairs erythropoiesis by

disrupting erythroblastic island formation via macrophages. J Transl Med. 2022;20(1):11.

29. Zhang Y, Li J-Q, Jiang Z-Z, Li L, Wu Y, Zheng L. CD169 identifies an anti-tumour macrophage subpopulation in human hepatocellular carcinoma. J Pathol. 2016;239(2):231-241.

30. Chow A, Lucas D, Hidalgo A, et al. Bone marrow CD169+ macrophages promote the retention of hematopoietic stem and progenitor cells in the mesenchymal stem cell niche. J Exp Med. 2011;208(2):261-271.

31. Zhang D, Gao X, Li H, et al. The microbiota regulates hematopoietic stem cell fate decisions by controlling iron availability in bone marrow. Cell Stem Cell. 2022;29(2):232-247.

32. Rughetti A, Biffoni M, Pierelli L, et al. Regulated expression of MUC1 epithelial antigen in erythropoiesis. Br J Haematol. 2003;120(2):344-352.

33. van den Berg TK, Nath D, Ziltener HJ, et al. Cutting edge: CD43 functions as a T cell counterreceptor for the macrophage adhesion receptor sialoadhesin (Siglec-1). J Immunol. 2001;166(6):3637-3640.

34. Chen K, Liu J, Heck S, Chasis JA, An X, Mohandas N. Resolving the distinct stages in erythroid differentiation based on dynamic changes in membrane protein expression during erythropoiesis. Proc Natl Acad Sci U S A. 2009;106(41):17413-17418.

35. 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-e49.

36. An X, Schulz VP, Li J, et al. Global transcriptome analyses of human and murine terminal erythroid differentiation. Blood. 2014;123(22):3466-3477.

37. Gautier E-F, Ducamp S, Leduc M, et al. Comprehensive proteomic analysis of human erythropoiesis. Cell Rep. 2016;16(5):1470-1484.

38. Wang J, Hayashi Y, Yokota A, et al. Expansion of EPOR-negative macrophages besides erythroblasts by elevated EPOR signaling in erythrocytosis mouse models. Haematologica. 2018;103(1):40-50.

39. Chen Y, Xiang J, Qian F, et al. Epo receptor signaling in macrophages alters the splenic niche to promote erythroid

differentiation. Blood. 2020;136(2):235-246.

40. Liao C, Prabhu KS, Paulson RF. Monocyte-derived macrophages expand the murine stress erythropoietic niche during the recovery from anemia. Blood. 2018;132(24):2580-2593.

41. Mukherjee K, Xue L, Planutis A, Gnanapragasam MN, Chess A, Bieker JJ. EKLF/KLF1 expression defines a unique macrophage subset during mouse erythropoiesis. Elife. 2021;10:e61070.

42. Mukherjee K, Bieker JJ. Transcriptional control of gene expression and the heterogeneous cellular identity of erythroblastic island macrophages. Front Genet. 2021;12:756028.

43. Okreglicka K, Iten I, Pohlmeier L, et al. PPARγ is essential for the development of bone marrow erythroblastic island macrophages and splenic red pulp macrophages. J Exp Med. 2021;218(5):e20191314.

44. Seu KG, Papoin J, Fessler R, et al. Unraveling macrophage heterogeneity in erythroblastic islands. Front Immunol. 2017;8:1140.

45. Falchi M, Varricchio L, Martelli F, et al. Dexamethasone targeted directly to macrophages induces macrophage niches that promote erythroid expansion. Haematologica. 2015;100(2):178-187.

46. Haldar M, Kohyama M, So AY-L, et al. Heme-mediated SPI-C induction promotes monocyte differentiation into iron-recycling macrophages. Cell. 2014;156(6):1223-1234.

47. Tuccillo FM, de Laurentiis A, Palmieri C, et al. Aberrant glycosylation as biomarker for cancer: focus on CD43. Biomed Res Int. 2014;2014:742831.

48. Xu C, He J, Wang H, et al. Single-cell transcriptomic analysis identifies an immune-prone population in erythroid precursors during human ontogenesis. Nat Immunol. 2022;23(7):1109-1120.

49. Hu J, Liu J, Xue F, et al. Isolation and functional characterization of human erythroblasts at distinct stages: implications for understanding of normal and disordered erythropoiesis in vivo. Blood. 2013;121(16):3246-3253.

50. Hasan S, Johnson MC, Kini AR, Baldea AJ, Muthumalaiappan K. A shift in myeloid cell phenotype via down regulation of Siglec-1 in island macrophages of bone marrow is associated with decreased late erythroblasts seen in anemia of critical illness. Front Med (Lausanne). 2019;6:260.

Treatment patterns and real-world effectiveness of rituximab maintenance in older patients with mantle cell lymphoma: a population-based analysis

Despite advances in management over recent years,1-4 mantle cell lymphoma (MCL) remains an incurable disease. 5 The primary objective remains to achieve a longlasting remission with first-line therapy. In order to achieve this goal, younger patients typically receive induction chemoimmunotherapy followed by consolidative autologous stem cell transplant (ASCT) with or without rituximab maintenance (RM). While many older patients may not be candidates for ASCT due to comorbidities and frailty, RM following induction chemoimmunotherapy is often considered. The clinical benefit of RM for MCL in older, ASCT-ineligible patients was demonstrated in a randomized controlled trial.6 When compared to interferon, RM prolongs both progression-free (PFS) and overall survival (OS) after R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, prednisone) as induction therapy (rituximab vs . interferon: median PFS: 5.4 vs . 1.9 years, median OS: 9.8 vs . 7.1 years).6, 7

However, since the publication of the efficacy data on RM in 2012, increasing evidence has raised the concern that R-CHOP for induction might be associated with inferior outcomes in MCL compared to bendamustine-rituximab (BR).8 Consequently, BR is recommended as the preferred first-line regimen for older, ASCT-ineligible patients with MCL in contemporary clinical practice guidelines.9 With the availability of more effective first-line treatment options (e.g., BR), the clinical benefit of RM has become less certain. In order to address this knowledge gap, we conducted a population-based study using the linked Surveillance, Epidemiology and End Results (SEER)-Medicare 2020 database and hypothesized that i) BR has replaced R-CHOP as the most used first-line regimen for older patients with MCL and ii) despite an evolution of the preferred first-line regimen, RM remains beneficial.

In order to assess the real-world treatment patterns for first-line MCL therapy, we selected adults ≥66 years old who were diagnosed with MCL in 2007-2017, had continuous Medicare A/B/D coverage, and had received ≥ one MCL therapy. We included 1,579 patients (Figure 1A). The median age was 76 years (interquartile range: 71-81 years), 65% were men, 95% were white, and 25% were described as frail with 21% having a comorbidity score ≥3 (Table 1). The median follow-up was 68 months.

We examined details of the fi rst-line regimens among those receiving treatment in the outpatient setting, as chemoimmunotherapy information was only available in

the outpatient claims (drug codes included in the Online Supplementary Table S1). We evaluated the practice patterns of RM and second-line therapy in a sub-population of patients who received R-CHOP or BR as first-line therapy. We defined RM as rituximab given as a single agent within 200 days after completion of rituximab containing first-line regimen, for ≥ two consecutive doses and lasting for ≥28 days. We used similar criteria to define lines of therapy as those applied in the Flatiron Health dataset.10 For example, any non-rituximab chemoimmunotherapy agent given within a 30-day window or rituximab within a 90-day window was considered as the same line. Among patients receiving treatment in the outpatient setting (n=1,367; 87%), the most common first-line therapies were bendamustine-based regimens (n=630; 46%), followed by anthracycline-based regimens (n=304; 22%). Novel therapies (including BTK and BCL-2 inhibitors) were rarely administered in the first line (n=72; 5.3%), and only 16 patients (1.2%) received cytarabine-based regimens as first-line therapy. Fifty patients (3.2% among the 1,579 patients) underwent ASCT following the first-line therapies. In order to examine the shift of first-line practice (R-CHOP vs. BR), we applied logistic regression models by incorporating year of diagnosis. Use of R-CHOP decreased substantially over time (2007: 43%, 2017: 7.3%, P for trend <0.001), with a significant increase in use of BR (2007:1.5%, 2017: 60%, P<0.001; Figure 1B). Among patients receiving R-CHOP or BR, 28% (n=244) received RM following the completion of the first-line therapies, with a median number of doses and duration of 9.5 doses and 18 months, respectively. Novel agents have become the most common therapies for the second-line setting (Figure 1C). In order to examine the real-world effectiveness of RM, we limited the overall study population to those who received R-CHOP or BR as first-line treatment and did not receive consolidative ASCT (transplant codes are included in the Online Supplementary Table S1). In order to minimize the potential immortal bias, for the non-RM group, we included patients who had a treatment gap (recipients of second-line therapy) or survival (no second-line therapy given) of ≥200 days after the completion of the firstline therapy, to ensure a sufficient amount of time to have been considered for RM.

We then created a matched study sample using propensity score matching (PSM) (ratio=1:1, greedy nearest neighbor with caliper=0.10) based on age, sex, race, mari-

tal status, Medicaid dual coverage, residence, poverty, frailty,11 comorbidities (modified Elixhauser index),12 year of diagnosis, extranodal disease, stage, first-line regimen (R-CHOP vs. BR), and duration of first-line therapy. In the PSM cohort, we included 262 patients, with a median age of 75 years, 67% men, >91% White, and 76% receiving first-line BR. The median number of doses and duration of RM in the “intervention arm” was 9.0 doses and 17 months, respectively. All baseline variables were balanced between the RM and non-RM groups (Table 1; P values for

all variables >0.1). The distributions of probabilities of receiving RM after PSM became very similar between the two comparison groups (Online Supplementary Figure S1). The standardized mean differences of all covariates between the two groups were smaller than (or very close to) 10%, which also suggested optimal matching.

In the PSM cohort, we applied the Cox regression model to compare overall survival (OS) and “approximated progression-free survival” (“approximated-PFS”) (survival and free of second-line therapy), respectively, based on the re-

Figure 1. Selection of study population and patterns of real-world practice for management of older patients with mantle cell lymphoma. (A) Flow diagram of patient selection. (B) Trend in use of BR and R-CHOP over time. (C) Real-world practice patterns of rituximab maintenance and second-line therapies following BR and R-CHOP in older, transplant ineligible patients with mantle cell lymphoma (MCL). BR: bendamustine+ rituximab; R-CHOP: rituximab+ cyclophosphamide+ doxorubicin+ vincristine+ prednisone; RM: rituximab maintenance; CIT: chemoimmunotherapy; R: rituximab. CIT includes anthracycline-, bendamustine-, and cytarabine-based chemoimmunotherapy. Novel therapies include ibrutinib, acalabrutinib, venetoclax, lenalidomide, and bortezomib. Other includes chemoimmunotherapy not included in the “CIT” category, rituximab single agent, and inpatient treatment (unknown regimens). Among patients receiving rituximab maintenance, 178 and 66 received BR and R-CHOP, respectively. Among patients who did not receive second-line therapy, 251 remained alive at the end of the study follow-up.

ceipt of RM and reported hazard ratio (HR) with 95% confidence interval (CI). We conducted competing-risk analysis for initiation of second-line therapy, reporting subdistribution HR (sHR) and 95% CI (all-cause mortality as the competing events). We followed patients from the initiation of the first-line therapy until the events of interest (all-cause mortality and initiation of second-line therapy), or the end of follow-up on December 31, 2019, whichever occurred first. Compared to the non-RM group, patients receiving RM had significantly longer OS and “approximated-PFS” and lower likelihood of receiving secondline therapy. In the subgroup analysis in patients receiving first-line BR, there was also significant clinical benefit in all three outcomes with the use of RM (Figure 2). In the sensitivity analyses using i) different definitions of the non-RM group (treatment gap following the completion of the first-line therapy of 150, 180, and 210 days, respectively); ii) earlier definition of RM (RM received ≤120 days following completion of first-line therapy); and iii) follow-

up from completion of first-line therapy rather than initiation – our findings were essentially the same (Online Supplementary Table S2). Given the missing stage information, we performed simple imputation and the results following imputation were also largely unchanged (data not shown). The effectiveness of RM following BR in MCL has been examined in several previous studies.10,13,14 However, due to the inconsistency of findings and some limitations of these studies, there remains uncertainty in the clinical benefit. Two prior observational studies10,13 showed similar benefits in the multivariable analysis. In contrast, the MAINTAIN trial showed no significant difference in PFS or OS between the RM and non-RM groups, despite the longer median PFS observed with RM (not reached vs. 55 months for non-RM).14 However, multivariable analysis might not be sufficient to control most potential confounding effects within observational data.10,13 The non-significant results in the MAINTAIN trial14 might be attributed to its relatively small sample size (total n=120).

RM: rituximab maintenance, PSM: propensity score matching; IQR: interquartile range; BR: bendamustine-rituximab; BTKi era: diagnosis in 2014-2017, pre-BTKi era: diagnosis in 2007-2011. *Prioritizing data from SEER record > census tract linkage > zip code linkage. **In 3 categories: fit, unfit, and frail with no missing data in the study population. $In 3 categories: BTKi era, pre-BTKi era, and washout period (2012-2013) with no missing data in the study population. $$No missing data in extranodal/nodal disease in the study population. §Among patients who received chemoimmunotherapy in the outpatient setting; these were the patients who had treatment regimen information available in the Medicare database. Some of the actual numbers are not reported in the table (e.g., >120 for White) in compliance with the reporting policy of the National Cancer Institure. P values for all baseline variables are >0.1

Table 1. Baseline characteristics of the overall cohort for evaluation of first-line therapy and of the propensity score matched cohort for evaluation of real-world effectiveness of rituximab maintenance.

Figure 2. Comparison of patient outcomes between the rituximab maintenance and non-maintenance groups. (A) Overall survival, (B) survival and free of second-line therapy, (C) receipt of second-line therapy in patients receiving either BR or R-CHOP (n at risk=131 for each comparison group), (D) overall survival, (E) survival and free of 2nd line therapy, (F) receipt of second-line therapy in patients receiving BR only (n at risk= 98 for maintenance group and 100 for non-maintenance group). HR: hazard ratio; Sub-HR: subdistribution hazard ratio; CI: confidence interval; BR: bendamustine+ rituximab; R-CHOP: rituximab+ cyclophosphamide+ doxorubicin+ vincristine+ prednisone; R= rituximab.

Our analysis rigorously addressed most potential biases. These include i) the incorporation of most prognostic factors, including comorbidities and frailty; ii) the application of causal inference approach in the comparative effectiveness analysis and use of multiple methods to examine/ensure comparability between the matching groups; iii) the consideration and adjustment for immortal bias in receipt of RM; and iv) the conduction of multiple sensitivity analyses showing robust results. In addition, our study used the population-based US database and focused on the older population, contributing complementarily with the previous observational studies.

Despite the rigorous study design and bias control, our population-based analysis has several limitations. Firstly, we were unable to adjust some potential confounders which were not available in the SEER-Medicare database (e.g., TP53 aberrance status). Although there was a relatively high level of missing data in the lymphoma

stage, results following imputation were essentially unchanged. Secondly, in the absence of data on responses to first-line induction from SEER-Medicare, we were unable to examine the potential differential effectiveness of RM based on prior responses. Thirdly, we were unable to compare the duration of remission, as date of relapse was not available in SEER-Medicare. Lastly, our population was not set to evaluate the effectiveness of RM following intensive regimens (e.g., cytarabine-based15) in fit, older patients, which should be examined in future studies.

In conclusion, our population-based real-world analysis showed significant benefits of RM in survival and disease control among older patients with MCL who did not receive ASCT, despite the shift from R-CHOP to BR as firstline induction. While prospective randomized trials would help validate the benefit of RM following BR, our study adds to the growing observational data supporting the benefit of RM in this setting.

Authors

Mengyang Di,1,2 Jessica B. Long,2 Shalin K. Kothari,1 Tarsheen Sethi,1 Amer M. Zeidan,1,2 Nikolai A. Podoltsev,1,2 Rory M. Shallis,1,2 Rong Wang,2,3 Xiaomei Ma2,3 and Scott F. Huntington1,2

1Section of Hematology, Department of Internal Medicine, Yale School of Medicine; 2Yale Cancer Outcomes, Public Policy and Effectiveness Research (COPPER) Center and 3Department of Chronic Disease Epidemiology, Yale School of Public Health, New Haven, CT, USA

Correspondence:

M. DI - mengyang.di@yale.edu

https://doi.org/10.3324/haematol.2022.282252

Received: October 20, 2022.

Accepted: January 5, 2023.

Early view: January 19, 2023.

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Disclosures

MD, JBL, TS, and RW do not have any conflict of interest to report. SKK reports consultancy and honoraria from Incyte and Karyopharm Pharmaceuticals. AMZ reports consultancy and research funding from AbbView, Acceleron, Amgen, Aprea, Boehringer Ingelheim, Cardiff Oncology, BMS, Incyte, Novartis; research funding from ADC Therapeutics, Astex, Pfizer; consultancy from Agios, Astellas, BeyondSpring, Daiichi Sankyo, Epizyme, Genentech, Gilead, Kura, Ionis, Loxo Oncology, Janssen, AstraZeneca, Jasper, Jazz; serving on clinical trial committees at AbbVie, BioCryst, BMS, Geron, Gilead, Kura, Loxo Oncology, Novartis; travel support from Cardiff Oncology, Novartis, Pfizer.

References

1. Tam CS, Opat S, Simpson D, et al. Zanubrutinib for the treatment of relapsed or refractory mantle cell lymphoma. Blood Adv. 2021;5(12):2577-2585.

2. Wang M, Munoz J, Goy A, et al. Three-year follow-Up of KTE-X19 in patients with relapsed/refractory mantle cell lymphoma, including high-risk subgroups, in the ZUMA-2 Study. J Clin Oncol. 2023;41(3):555-567.

3. Wang M, Rule S, Zinzani PL, et al. Acalabrutinib in relapsed or refractory mantle cell lymphoma (ACE-LY-004): a single-arm, multicentre, phase 2 trial. Lancet. 2018;391(10121):659-667.

4. Wang ML, Rule S, Martin P, et al. Targeting BTK with ibrutinib in relapsed or refractory mantle-cell lymphoma. N Engl J Med. 2013;369(6):507-516.

5. Di M, Cui C, Kothari SK, et al. Survival of mantle cell lymphoma in the era of Bruton tyrosine kinase inhibitors: a population-

NAP reports honoraria from Pfizer, Blueprint Medicines, Incyte, Novartis, Celgene, Bristol-Myers Squib, CTI BioPharma, PharmaEssentia, AbbVie, Constellation pharmaceuticals; serving on independent data monitoring committee at Cogent biosciences. RMS reports divested equity in a private or publicly-traded company at Curis and advisory boards for BMS and Gilead Sciences, Inc. XM reports consultancy with Bristol Myers Squibb. SFH reports consultancy from Abbvie, ADC Therapeutics, Arvinas, AstraZeneca, BeiGene, Epizyme, Flatiron Health, Genentech, Janssen, Merck, Novartis, Pharmacyclics, Seagen, Servier, TG therapeutics, and Thyme Inc.; research funding from Celgene, DTRM Biopharm, TG Therapeutics; honoraria from Pharmacyclics, AstraZeneca, Bayer.

Contributions

Conceptualization, formal analysis, data interpretation, writingoriginal draft, writing-review and editing by MD. Conceptualization, formal analysis, data interpretation, writing-review and editing by SFH. Conceptualization, data accrual, formal analysis, data interpretation, writing-review and editing by JBL. Conceptualization, data interpretation, writing-review and editing by SKK, TS, AMZ, NAP, RMS, RW, and XM.

Acknowledgments

This study used the linked SEER-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. The authors acknowledge the efforts of the National Cancer Institute, the Office of Research, Development and Information, CMS, Information Management Services (IMS), Inc. and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database.

Data-sharing statement

This study used Surveillance, Epidemiology, and End ResultsMedicare database, which can be available upon application for the database through the National Cancer Institute.

based analysis. Blood Adv. 2022;6(11):3339-3342.

6. Kluin-Nelemans HC, Hoster E, Hermine O, et al. Treatment of older patients with mantle-cell lymphoma. N Engl J Med. 2012;367(6):520-531.

7. Kluin-Nelemans HC, Hoster E, Hermine O, et al. Treatment of older patients with mantle cell lymphoma (MCL): long-term follow-up of the randomized European MCL Elderly Trial. J Clin Oncol. 2020;38(3):248-256.

8. Flinn IW, van der Jagt R, Kahl B, et al. First-line treatment of patients with indolent non-Hodgkin lymphoma or mantle-cell lymphoma with bendamustine plus rituximab versus R-CHOP or R-CVP: results of the BRIGHT 5-year follow-up Study. J Clin Oncol. 2019;37(12):984-991.

9. National Comprehensive Cancer Network. NCCN Clinical Practice Guidelines in Oncology: B-cell lymphoma.

https://www.nccn.org/professionals/physician_gls/pdf/b-cell.pdf

Accessed 26 June 2022.

10. Martin P, Cohen JB, Wang M, et al. Treatment outcomes and roles of transplantation and maintenance rituximab in patients with previously untreated mantle cell lymphoma: results from large realworld cohorts. J Clin Oncol. 2023;41(3):541-554.

11. Kim DH, Schneeweiss S, Glynn RJ, Lipsitz LA, Rockwood K, Avorn J. Measuring frailty in medicare data: development and validation of a claims-based frailty index. J Gerontol A Biol Sci Med Sci. 2018;73(7):980-987.

12. van Walraven C, Austin PC, Jennings A, Quan H, Forster AJ. A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data. Med Care. 2009;47(6):626-633.

13. Hill BT, Switchenko JM, Martin P, et al. Maintenance rituximab improves outcomes in mantle cell lymphoma patients who respond to induction therapy with bendamustine+ rituximab without autologous transplant. Blood. 2019;134(Suppl 1):S1525.

14. Rummel MJ, Knauf W, Goerner M, et al. Two years rituximab maintenance vs. observation after first-line treatment with bendamustine plus rituximab (B-R) in patients with mantle cell lymphoma: first results of a prospective, randomized, multicenter phase II study (a subgroup study of the StiL NHL72008 MAINTAIN trial). J Clin Oncol. 2016;34(Suppl 15):S7503.

15. Bega G, Olivieri J, Riva M, et al. Rituximab and bendamustine (BR) compared with rituximab, bendamustine, and cytarabine (R-BAC) in previously untreated elderly patients with mantle cell lymphoma. Cancers (Basel). 2021;13(23):6089.

Insights into dasatinib use and outcomes in real-world patients with chronic myeloid leukemia

Patients with chronic myeloid leukemia (CML) in routine clinical practice likely differ substantially from clinical trial populations, which can have implications for the extrapolation of trial outcomes to the general population. In the DASISION trial, mean patient age was 46 years and the study excluded patients with Eastern Cooperative Oncology Group (ECOG) performance status ≥3, uncontrolled or serious medical disorders (including cardiovascular disease) or active infections, hepatic or kidney dysfunction, and previous or concurrent cancer.1 Over 30% of patients failed to achieve deep molecular response (DMR) and 20% experienced adverse drug reactions (ADR) necessitating treatment discontinuation within 5 years.1 The now well-recognised cardiovascular and pulmonary toxicities associated with dasatinib were reported in subsequent studies.2 The aim of this was study to investigate prescribing patterns, tolerability, and effectiveness of dasatinib in patients with CML in real-world clinical practice, with a focus on patients considered ineligible for a pivotal CML clinical trial. In this retrospective observational study, patients with CML who had at least 3 months of dasatinib treatment (2006-2018) were identi fi ed through hematology registries at two University hospitals in Sydney, Australia. Demographics, CML disease characteristics, and dasatinib prescribing patterns were collected and described according to indication (treatment-naïve vs. second-line or later). Concomitant medicines, including complementary and alternative medicines, were documented. Molecular response endpoints were defi ned using quantitative BCR-ABL1 transcript levels according to the international scale.3 All documented ADR during dasatinib treatment were defined using the Common Terminology Criteria for Adverse Events version 5.0,4 and were evaluated for causality to dasatinib using the Naranjo algorithm.5 Time to first dasatinib dose modification, time to treatment discontinuation, overall survival (OS) and progression-free survival (PFS) were evaluated using the Kaplan-Meier method, with a log-rank test comparing between-group differences. The cumulative incidences of major molecular response (MMR), deep molecular response (DMR) and dasatinib-related ADR were modeled using were modeled using the cumulative incidence competing risk method with Gray’s weighted log-rank test comparing between-group differences. Fine-Gray subdistribution hazard model was used to assess the independent factors associated with achievement of DMR and grade ≥3 ADR occurrence, with the subdistribution hazard ratios (SHR) reported. The Fine-Gray subdistribution ha-

zard model, compared with traditional Cox models, considers competing risks that may hinder the observation of the event of interest such as treatment discontinuation or death prior to DMR or observation of the ADR. The hazard of recurrent ADR was modeled using the Prentice, Williams and Peterson Total Time model, and reported as hazard ratios (HR). Subgroup analyses were performed based on whether patients did or did not meet the original eligibility criteria for the DASISION1 trial. The study was conducted in accordance with ethical requirements of local institutions. The research included in this study was approved by the Sydney Local Health District Human Research Ethics Committee (reference number: LNR/17/CRGH/248) on October 13, 2017. Site specific approval was also obtained for Concord Repatriation General Hospital (reference number: LNRSSA/17/CRGH/249) and Royal North Shore Hospital (reference number: RESP/18/146). A waiver of consent according to the National Statement on Ethical Conduct in Human Research was granted by the Human Research Ethics Committee. The statistical analyses were performed using R (version 3.3.3).6 A detailed description of the study design and statistical methods are described in Adattini et al.7 Fifty-two patients who started dasatinib from 2006 to 2018 met eligibility criteria and were included in the analysis, with data collected on a total of 54 dasatinib treatment courses (22 treatment-naïve, 32 second-line or later). Most treatments (n=48) were initiated at the standard dasatinib dose of 100 mg/day (Table 1). Applying the eligibility criteria from DASISION, 30 (56%) patients would likely have been excluded due to serious or poorly controlled medical conditions (n=26, 50% of patients), hepatic or kidney dysfunction (n=4, 8%), concurrent cancer (n=3, 6%), or Eastern Cooperative Oncology Group ≥3 (n=1, 3%). This group also had a higher Charlson Comorbidity Index (CCI) compared with patients meeting eligibility criteria (median CCI score 5 vs. 2, P<0.001), and were significantly older at treatment initiation (median age 67 vs. 41 years, P<0.001; Table 1). A larger proportion of patients in the ineligible group were receiving one or more potentially interacting medicines during dasatinib treatment (87% vs. 41% if eligible, P<0.001; Table 1).

Within the first 12 months of treatment, 45% of patients (95% confidence interval [CI]: 29–57) required a dasatinib dose modification (any type), 41% (95% CI: 25–54) a dose reduction or temporary treatment interruption, and 15% (95% CI: 4–25) a dose escalation. Occurrence of an ADR was the stated reason for 90% of dose reductions/inter-

ruptions, whilst dose escalation was due to poor clinical response in 73%. After a median follow-up of 26 months (interquartile range [IQR]: 7–48 months), 59% of patients were still receiving dasatinib. Estimated rates of dasatinib discontinuation were 32% (95% CI: 17–44) at 2 years and 47% (95% CI: 28–60) at 5 years. Reasons for dasatinib discontinuation were occurrence of ADR (64% of discontinuations), poor clinical response (23%), relapse or disease progression (14%), achievement of sustained DMR for ≥2

years (9%), and death (9%). There were no significant differences in the risk of dasatinib dose modifications or discontinuation between patients receiving dasatinib first-line compared to second-line or later.

After 2 years of dasatinib treatment, 71% (95% CI: 56–82) of patients had experienced at least one dasatinib-related ADR requiring modification of existing long-term medicines or commencement of new medicines, 51% (95% CI: 32–59) a grade ≥3 ADR, and 38% (95% CI: 24–52) an ADR

BM: bone marrow; CAM: complementary or alternative medicine; CCI: Charlson Comorbidity Index; CYP: cytochrome P450; ECOG PS: Eastern Cooperative Oncology Group Performance Status; IQR: interquartile range; PPI: proton pump inhibitor. †Hepatic function missing 4 observations (2 eligible, 2 ineligible), ECOG PS missing 1 observation (eligible), additional BM karyotype abnormalities missing 11 observations (4 eligible, 7 ineligible). §Comparison is between ECOG PS of 0 vs. 1 or more.

Table 1. Baseline demographic and chronic myeloid leukemia disease characteristics, including comparison by likely eligibility for the DASISION trial.

resulting in hospitalisation (Figure 1; Table 2). The most frequent dasatinib-related grade ≥3 ADR at 24 months included cardiovascular disorders, neutropenia, pleural effusion or pulmonary edema, and infection (Online Supplementary Table S1).

Independent risk factors for a grade ≥3 ADR on dasatinib included poorly controlled hypertension (SHRadjusted: 6.24; 95% CI: 2.70–14.20), higher ECOG (SHRadjusted: 1.65; 95% CI: 1.16–2.36), and higher dasatinib starting dose of 70 mg twice daily compared with 50 mg/day and 100 mg/day (SHRadjusted: 9.09; 95% CI: 2.63–25.00 and 7.69; 95% CI: 3.13–20.00), respectively). Independent risk factors for recur-

rent grade ≥3 ADR included age >60 years (HRadjusted: 4.28; 95% CI: 1.73–10.57), higher ECOG (HRadjusted: 2.15; 95% CI: 1.37–3.37), higher dasatinib starting dose of 70 mg twice daily (HRadjusted, 3.45; 95% CI: 1.16–10.00 vs. 100 mg/day) and geographic ancestry (HRadjusted: 2.73; 95% CI: 1.06–7.02 for East Asian compared with European ancestry).

The 2-year cumulative incidences of MMR and DMR were 81% (95% CI: 61–92) and 73% (95% CI: 55–85), respectively (Online Supplementary Figure S1). Poorer ECOG was the only independent predictor of inferior DMR rates with dasatinib (SHRadjusted: 0.34; 95% CI: 0.12–0.96). Estimated 3year OS and PFS rates of dasatinib-treated patients were

Figure 1. Cumulative incidence of dasatinib-related adverse drug reactions. Cumulative incidence of dasatinib-related adverse drug reactions (ADR) by 3 years (95% confidence intervals) calculated using the cumulative incidence competing risk method.

Table 2. Cumulative incidence of dasatinib-related adverse drug reactions, including comparison by likely eligibility for the DASISION trial.

94% (95% CI: 87–100) and 94% (CI:87-100), respectively (Online Supplementary Figure S2).

Patients likely ineligible for the DASISION trial had a significantly higher risk of requiring a dasatinib dose reduction or interruption compared with the likely eligible patients (HR: 2.27; 95% CI: 1.00–5.24; P<0.05). Furthermore, the likely ineligible cohort had a higher risk of ADR resulting in dasatinib dose changes or treatment discontinuation (SHR: 3.02; 95% CI: 1.44–6.33) and ADR resulting in hospitalization (SHR: 2.65; 95% CI: 1.11–6.33; Table 2). Recurrent grade ≥3 infection was more likely to occur in patients considered ineligible for DASISION compared to those considered eligible (HR: 4.09; 95% CI: 1.00–17.03; P<0.05). There were no significant differences in molecular response rates or survival in the likely ineligible versus eligible groups.

Reassuringly, rates of MMR and DMR with dasatinib treatment in this real-world study were at least comparable to, if not higher than previously reported in controlled clinical trials.1,8,9 We observed a higher incidence of many non-hematologic ADR than is reported in clinical trials. In 3-year follow-up of the DASISION trial,10 grade ≥3 fluid retention (pleural effusion and superficial edema) was reported in only 3% (vs. a 3-year cumulative incidence of 13% in this study; Online Supplementary Table S1), with all other non-hematological grade ≥3 ADR occurring in 3% or less of patients. Additionally, we observed a 3-year cumulative incidence of 14% for grade ≥3 cardiovascular events compared with less than 5% in published studies.9-11

The higher rates of observed ADR in this study compared with controlled clinical trials may be explained by the differences in patient characteristics in this real-world population who were of older age, with poorer performance status, more likely to have poorly controlled hypertension and other concomitant medical conditions and more likely to be co-prescribed medications with potential interactions. Our population also included some patients of East Asian ancestry, a population known to be more susceptible to dasatinib-related ADR likely due to increased drug exposure.12

An important insight was the signi ficant healthcare resources utilized for the management of dasatinib-related ADR, including a substantial proportion requiring hospitalization. The incidence of ADR-related short-term therapeutic interventions observed in this study (e.g., glucocorticosteroids, diuretics, antimicrobials and thoracocentesis) is higher than those reported by Fox et al., 13 whereby 21% of patients treated in routine clinical practice experienced dasatinib-related pleural effusion requiring therapeutic intervention. Most patients in our study experienced a dasatinib-related ADR requiring commencement of (or changes to) long-term medications; a clinically important implication of ADR which has not been highlighted in previous studies. Increased medication burden has the potential to increase adverse events, in-

crease medical costs, reduce medication adherence, and negatively affect health outcomes such as frailty and mortality.14

The sample size of this study was limited by the number of patients diagnosed and treated with dasatinib at the two clinical centers over the period of data collection. As such, certain baseline predictor variables could not be evaluated in multivariable regression. Furthermore, as anticipated in a CML study, there were a small number of events such as death, disease progression, CML transformation or relapse on dasatinib treatment. Larger patient samples and longer observation times would be ideal in a future study to identify predictors of survival. Despite the small sample size of this study, it is notable that statistically significant findings were made, however caution should be applied in the interpretation. Importantly, the patients in our study were representative of patients receiving dasatinib treatment for CML in Australia with respect to age and sex distribution.15 This real-world data on 54 dasatinib treatments represents a total of 154 patient years of experience with dasatinib treatment in CML. The detail and depth of collection of comorbidity data, patient outcomes and disposition, are some of the novel contributions of our study. Here, we present new insights on the relationship between patient and disease characteristics and real-world dasatinib treatment outcomes. We also identify the impact of clinical trial eligibility criteria on obtaining a patient sample representative of the real-world clinical population and on real-world treatment outcomes. These results can be used to inform the joint clinical decision-making process by patient and doctor as to the individualized benefits and risks of dasatinib. Biological and clinical factors should be considered prior to tyrosine kinase inhibitor selection to aid in dose and drug selection, and to identify those who require careful monitoring or intervention to optimize risk factors for the development of ADR.

Authors

1Sydney Pharmacy School, The University of Sydney; 2Clinical Pharmacology Modeling & Simulation, GlaxoSmithKline R&D and 3Concord Cancer Center, Concord Repatriation General Hospital, Sydney, Australia

Correspondence:

A. MCLACHLAN - andrew.mclachlan@sydney.edu.au

https://doi.org/10.3324/haematol.2022.282237

Josephine A. Adattini,1 Annette S. Gross,1,2 Nicole Wong Doo3 and Andrew J. McLachlan1

Received: November 13, 2022.

Accepted: January 5, 2023.

Early view: January 19, 2023.

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Disclosures

No conflicts of interest to disclose.

Contributions

All authors conceived the study. JA collected, analyzed and interpreted the data, and wrote the manuscript. ASG, AJM and NWD also contributed to the interpretation of the data and revised the manuscript. All authors approved the final manuscript.

References

1. Cortes JE, Saglio G, Kantarjian HM, et al. Final 5-year study results of DASISION: the dasatinib versus imatinib study in treatment-naive chronic myeloid leukemia patients trial. J Clin Oncol. 2016;34(20):2333-2340.

2. Medeiros BC, Possick J, Fradley M. Cardiovascular, pulmonary, and metabolic toxicities complicating tyrosine kinase inhibitor therapy in chronic myeloid leukemia: strategies for monitoring, detecting, and managing. Blood Rev. 2018;32(4):289-299.

3. Deininger MW, Shah NP, Altman JK, et al. Chronic myeloid leukemia, Version 2.2021, NCCN clinical practice guidelines in oncology. J Natl Compr Canc Netw. 2020;18(10):1385-1415.

4. National Cancer Institute (U.S.). Common Terminology Criteria for Adverse Events (CTCAE) v5.0: U.S. Department of Health and Human Services; 2017. Available from: https://ctep.cancer.gov/protocoldevelopment/electronic_applica tions/ctc.htm.

5. Naranjo CA, Busto U, Sellers EM, et al. A method for estimating the probability of adverse drug reactions. Clin Pharmacol Ther. 1981;30(2):239-245.

6. R Core Team. R: a language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria; 2020.

7. Adattini JA, Gross AS, Wong Doo N, McLachlan AJ. Real-world efficacy and safety outcomes of imatinib treatment in patients with chronic myeloid leukaemia: an Australian experience. Pharmacol Res Perspect. 2022;10(5):e01005. doi: 10.1002/prp2.1005.

8. Marin D, Hedgley C, Clark RE, et al. Predictive value of early molecular response in patients with chronic myeloid leukemia

Acknowledgments

The authors would like to thank pharmacy students from the University of Sydney (Maddison Mansfield, Maisah Joarder, Samantha Brown and Zahraa Falfaly) for their assistance with data collection, the Hematology Departments at Concord Repatriation General Hospital and Royal North Shore Hospital for facilitating access to data, and Dejana Munjiza for her assistance with R programming.

Funding

This work was supported by the Peter Coates Postgraduate Scholarship in Ethnopharmacology provided by GlaxoSmithKline.

Data-sharing statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available because of privacy or ethical restrictions.

treated with first-line dasatinib. Blood. 2012;120(2):291-294.

9. Cortes JE, Jiang Q, Wang J, et al. Dasatinib vs. imatinib in patients with chronic myeloid leukemia in chronic phase (CMLCP) who have not achieved an optimal response to 3 months of imatinib therapy: the DASCERN randomized study. Leukemia. 2020;34(8):2064-2073.

10. Jabbour E, Kantarjian HM, Saglio G, et al. Early response with dasatinib or imatinib in chronic myeloid leukemia: 3-year follow-up from a randomized phase 3 trial (DASISION). Blood. 2014;123(4):494-500.

11. Shah NP, Kim DW, Kantarjian H, et al. Potent, transient inhibition of BCR-ABL with dasatinib 100 mg daily achieves rapid and durable cytogenetic responses and high transformation-free survival rates in chronic phase chronic myeloid leukemia patients with resistance, suboptimal response or intolerance to imatinib. Haematologica. 2010;95(2):232-240.

12. Chuah CT, Nakamae H, Shen ZX, Bradley-Garelik MB, Kim DW. Efficacy and safety of dasatinib versus imatinib in the East Asian subpopulation of the DASISION trial of newly diagnosed chronic myeloid leukemia in chronic phase. Leuk Lymphoma. 2014;55(9):2093-2100.

13. Fox LC, Cummins KD, Costello B, et al. The incidence and natural history of dasatinib complications in the treatment of chronic myeloid leukemia. Blood Adv. 2017;1(13):802-811.

14. Nightingale G, Skonecki E, Boparai MK. The impact of polypharmacy on patient outcomes in older adults with cancer. Cancer J. 2017;23(4):211-218.

15. Australian Institute of Health and Welfare. Cancer data in Australia. Canberra: Australian Institute of Health and Welfare; 2020.

TIM3, a human acute myeloid leukemia stem cell marker, does not enrich for leukemia-initiating stem cells in B-cell acute lymphoblastic leukemia

Here, we prompted to determine in molecularly distinct B-cell acute lymphoblastic leukemia (B-cell ALL) patients whether TIM3 represents a leukemia-initiating stem cell (LIC) marker enabling the prospective isolation of LIC-enriched B-cell ALL cells and found that in contrast to what has been shown in acute myeloid leukemia (AML), TIM3 does not enrich for LIC in B-ALL. Relapse remains a major challenge in the clinical management of both AML and B-cell ALL and is driven by rare therapy-resistant LIC that reside in specific bone marrow (BM) niches.1 The clinical implications of LIC are beyond any doubt, as evidenced by the large number of preclinical and clinical studies elucidating the phenotype and molecular determinants of LIC.2,3 In human AML, where the hierarchical leukemic stem cell model is well established, multiple surface proteins that have been proposed to enrich for AML-LIC.4-8 Among these, stands out T-cell immunoglobulin mucin-3 (TIM3), a human AML stem cell marker which has been shown to enable the prospective isolation of LIC-enriched AML cells.9 In fact, TIM3 has been sug-

gested to be a promising target to selectively eliminate AML-LIC and several TIM3 inhibitors are being clinically tested in patients with advanced AML.10,11 The stem cell model picture is less clear in B-cell ALL. Conflicting studies could not resolve the phenotype of the B-cell ALL-LIC yet. Consequently, whether B-cell ALL follows a hierarchical leukemogenic model, driven by a rare population of LIC, or a stochastic leukemogenic model where most of the blasts (even at different maturational stages) can reconstitute in serial xenotransplantation assays and re-establish the complete leukemic phenotype remains unresolved. Here, we initially profiled by fluorescence-activated cell sorting (FACS) the expression of TIM3 in 85 BM samples from both pediatric and adult B-cell patients (Online Supplementary Table S1) and found that TIM3 protein is heterogeneously expressed in B-cell ALL blasts from both diagnostic (Dx, n=47) and relapsed (n=38) patients (Figure 1A-C). The proportion of TIM3-expressing CD34+CD19+ Bcell blasts at Dx was ∼2-fold higher than that observed in normal B-cell/B-cell progenitor counterparts from healthy

Figure 1. Expression of TIM3 in blasts from B-cell acute lymphoblastic leukemia patients at disease presentation and relapse. (A) Representative fluorescence-activated cell sorting analysis showing the expression of TIM3 in B-cell acute lymphoblastic leukemia (ALL) blasts (CD45low/+ CD34+CD19+). (B, C) Percentage (B) and mean fluorescence intensity (MFI) (C) of TIM3+ B-cell ALL cells in bone marrow (BM) at diagnosis (Dx) and relapse. CD19+ B-cell progenitors and mature B cells were analyzed in healthy BM. Healthy BM, n=21; Dx B-cell ALL, n=47; relapsed B-cell ALL, n=38. (D) Bulk RNA sequencing-based TIM3 expression obtained at Dx from n=10 pediatric MLL4-AF4+ proB-cell ALL patients and in the corresponding matched B-cell ALL (no lineage switch) relapse (n=5) or myeloid lineage-switched relapse (n=5).

Figure 2. Leukemia reconstitution in vivo of TIM3+ and TIM3- B-cell acute lymphoblastic leukemia blasts in cell dose limiting dilution primograft xenotransplantation assays. (A) Detailed experimental design for assessing in vivo the leukemia-initiating stem cells (LIC) frequency in cell dose-limiting dilution primograft xenotransplantation assays using 5 primary B-cell acute lymphoblastic leukemia (B-ALL) samples (2 ETV6-RUNX1+ and 3 MLL-AF4+). Decreasing doses (200,000 down to 5,000) of TIM3+ and TIM3- B-cell ALL blasts were intra-bone marrow (intra-BM) transplanted into NSG mice. Mice health and leukemia development was monitored over 20 weeks. Mice were sacrificed when i) signs of disease were evident, ii) B-cell ALL graft was >10% in peripheral blood in the absence of signs of disease or iii) at day 140 (end point) in the absence of symptoms or leukemia engraftment. (B) Kaplan-Meier event-free and overall survival curves for each cell fraction (TIM3+ vs. TIM3-) and cell dose (200,000 down to 5,000). N=150 mice studied in total: 30 mice/leukemia; 75 mice/cell fraction, 10 mice/cell dose. (C) Estimated frequency (and 95% confidence interval) of LIC in primografts transplanted with TIM3+ and TIM3- blasts. (D) Penetrance of leukemic mice at end point (number engrafted mice/total number transplanted mice). IBMT: intra-bone marrow transplant.

BM (20%, range, 0.3-56.4 vs. 12%, range, 2.8-28.4; P=0.0028), and the proportion of TIM3+CD34+CD19+ cells further increased significantly in B-cell blasts at relapse (32%, range, 2-80; P=0.0008) (Figure 1A, B). The levels of TIM3 expression in CD34+CD19+ B-cell blasts, measured by mean intensity fluorescence, showed a very similar trend to the proportion of TIM3+CD34+CD19+ cells, further confirming an upregulation of TIM3 expression in B-cell ALL patients during disease progression (Dx>relapse) (Figure 1C). We next analyzed the TIM3 RNA expression in ten

KMT2A-AFF1+ proB-cell ALL patients at Dx and in matched relapses (Figure 1D).12 Half of these MLL-AF4+ proB-cell ALL patients relapsed as CD19+ B-cell ALL while the other half relapsed as a CD19- myeloid lineage-switched. Very interestingly, the expression levels of TIM3 were dramatically higher in all KMT2A-AFF1+ myeloid lineage-switched relapses than in KMT2A-AFF1+ CD19+ B-cell relapses or the Dx samples, further linking TIM3 with AML-LIC/AML progression.

We next interrogated the ability of highly purified (FACS

Figure 3. TIM3 expression does not enrich for B-cell acute lymphoblastic leukemia-initiating stem cells capacity in secondary recipients. (A) Detailed experimental design for assessing in vivo the leukemia-initiating stem cells (LIC) frequency in serial xenotransplantation assays. Bone marrow (BM) cells were retrieved from 20 primografts engrafted (>80%) with TIM3+ (n=10) or TIM3- (n=10) B-cell acute lymphoblastic leukemia (B-ALL) blasts and transplanted at 2 different doses (50,000 and 25,000) in secondary irradiated recipients (1 primograft into 4 secondary; n=80 secondary mice, 16 mice/leukemia, 40 mice/cell fraction). Mice health and leukemia development was monitored over 20 weeks. Mice were sacrificed when i) signs of disease were evident, ii) B-cell ALL graft was >10% in peripheral blood (PB) in the absence of signs of disease or iii) at day 140 (end point) in the absence of symptoms or leukemia engraftment. (B) Kaplan-Meier event-free and overall survival curves for each cell fraction (TIM3+ vs. TIM3-) and cell dose (50,000 and 25,000). N=80 mice studied in total: 16 mice/leukemia; 40 mice/cell fraction, 40 mice/cell dose. (C) Estimated frequency (and 95% confidence interval) of LIC in secondary mice. (D) Penetrance of leukemic mice at end point (number engrafted mice/total number transplanted mice). IBMT: intra-bone marrow transplant.

purity >98%, data not shown) TIM3+ and TIM3- blasts to initiate B-cell ALL in vivo upon intra-BM transplantation into NSG mice (n=150) in limiting dilution assays (LDA) (Figure 2A). All primary leukemias (2 ETV6-RUNX1+ and 3 KMT2A-AFF1+) engrafted onto primografts reproducing the Dx phenotype (data not shown ). As expected, both the event-free survival (EFS) and the overall survival (OS) decreased with increasing doses of transplanted blasts (Figure 2B); however, no significant differences were found in either EFS or OS between TIM3+ and TIM3- populations across cell doses (Figure 2B). Importantly, the estimated LIC frequency, calculated in LDA using the ELDA software13 was similar between TIM3+ and TIM3- B-cell blast populations (Figure 2C). Similarly, despite a trend towards a slightly higher frequency of engrafted mice (leukemia penetrance) in primografts transplanted with lower doses of TIM3- blasts, no significant differences were observed overall in the frequency of engrafted mice between TIM3+ and TIM3- B-cell blast populations (Figure 2D).

For serial transplantation experiments, 50,000 and 25,000 B-cell ALL cells from primografts were intra-BM transplanted into secondary mice, rendering a significantly lower EFS and OS (higher aggressiveness) (Figure 3A, B) than that observed in primary recipients; however, no differences in either EFS or OS were observed between secondary recipients transplanted with TIM3+ or TIM3- B-cell blast populations (Figure 3B). Similarly, the estimated LIC frequency (Figure 3C) and leukemic penetrance (frequency of engrafted mice) were very similar between secondary recipients transplanted with TIM3+ or TIM3- B-cell blast populations (Figure 3D). Taken together, our data demonstrate that despite an increased expression of TIM3 in Bcell ALL blasts during disease progression, TIM3 does not enrich for LIC in B-cell ALL.

We report here that, in contrast to what has been shown in AML, TIM3 does not represent a stem cell marker capable of prospectively isolating LIC in either high-risk KMT2A-AFF1+ and low/standard-risk ETV6-RUNX1+ B-cell ALL. All individual high-risk and standard-risk leukemic primary BM samples engrafted in NSG mice even at very low doses, validating this immunodeficient mouse model to functionally assess for candidate human leukemia stem cell populations. Furthermore, our intra-BM transplantation assay provides a highly sensitive and specific assay for interrogating LIC in acute leukemia because it overcomes potential survival and BM homing intrinsic deficiencies of transplanted cells.14 These findings are in line with previous reports and reinforce that distinct immunophenotypically defined B-cell ALL blast populations, even at different maturation stages, have stem cell properties,14,15 reinforcing that some hematopoietic malignancies (as for AML and other myeloid neoplasms) are maintained by a rare population of LIC (hierarchical model) whereas in B-cell ALL most of the blasts possess “stemness” features (stochastic

model) being capable of initiating and recapitulating the disease in vivo. Further work is needed to understand the role of, and alteration in, the expression of the immune checkpoint receptor TIM3 in blasts patients with acute leukemia, especially for rationalizing and interpreting current clinical trials testing TIM3 inhibitors in relapsing/refractory AML and myelodysplastic syndrome patients.11

Authors

Clara Bueno,1,2,3 Alba Martínez,1,2 Paola Alejandra Romecin,1,2 Samanta Romina Zanetti,1 Ricky Tirtakusuma,4 Eulalia Genesca,1 Mireia Camós,5,6 Manuel Ramírez-Orellana,2,7 Eduardo Anguita,8 Paola Ballerini,9 Franco Locatelli,10 José Luis Fuster11 and Pablo Menéndez1,2,3,12,13

1Josep Carreras Leukemia Research Institute, Barcelona, Spain; 2Red Española de Terapias Avanzadas (TERAV)-Instituto de Salud Carlos III (ISCIII) (RICORS, RD21/0017/0029), Madrid, Spain; 3CIBER-ONC, ISCIII, Barcelona, Spain; 4Newcastle University, Newcastle upon Tyne, UK; 5Hematology Laboratory, Hospital Sant Joan de Déu, Barcelona, Spain; 6Leukemia and other Pediatric Hemopathies, Developmental Tumor Biology Group, Instiutut de Recerca Hospital Sant Joan de Déu, Barcelona, Spain; 7Oncohematología, Hospital Niño Jesús, Madrid, Spain; 8Servicio de Hematología, Hospital Clínico San Carlos, IdISSC, Medicina UCM, Madrid, Spain; 9Pediatric Hematology, Armand Trosseau Hospital, Paris, France; 10Department of Hematology and Oncology, Ospedale Bambino Gesú, Rome, Italy; 11Sección de Oncohematología Pediátrica, Hospital Clínico Universitario Virgen de la Arrixaca and Instituto Murciano de Investigación Biosanitaria (IMIB), El Palmar, Murcia, Spain; 12Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain and 13Department of Biomedicine, School of Medicine, University of Barcelona, Barcelona, Spain

Correspondence:

P. MENENDEZ - pmenendez@carrerasresearch.org

C. BUENO - cbueno@carrerasresearch.org

https://doi.org/10.3324/haematol.2022.282394

Received: November 22, 2022.

Accepted: January 5, 2023.

Early view: January 19, 2023.

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Disclosures

PM is founder of the spin-off OneChain Immunotherapeutics which has no connection with the present research. All other authors have no conflicts of interest to disclose.

Contributions

CB conceived the study, designed, and performed experiments, analyzed/interpreted data, prepared figures, wrote the manuscript and financially supported the study. AM, PAR and RT performed experiments. SRZ conceived the study, performed experiments and analyzed/interpreted data. EG, MC, MRO, EA, PB, FL and JLF contributed patient’s samples and clinical data. PM conceived the study, designed experiments, interpreted data, wrote the manuscript, and financially supported the study. All authors have read and agreed to publish the manuscript.

Acknowledgments

We thank Francisco Gutierrez-Agüera for technical help. We thank CERCA/Generalitat de Catalunya and Fundació Josep Carreras-Obra Social la Caixa for core support.

References

1. Ishikawa F, Yoshida S, Saito Y, et al. Chemotherapy-resistant human AML stem cells home to and engraft within the bonemarrow endosteal region. Nat Biotechnol. 2007; 25(11):1315-1321.

2. Hoang VT, Zepeda-Moreno A, Ho AD. Identification of leukemia stem cells in acute myeloid leukemia and their clinical relevance. Biotechnol J. 2012; 7(6):779-788.

3. Mohamed MMI, Aref S, Agdar MA, Mabed M, El-Sokkary AMA. Leukemic stem cell (CD34(+)/CD38(-)/TIM3(+)) frequency in patients with acute myeloid leukemia: clinical implications. Clin Lymphoma Myeloma Leuk. 2021;21(8):508-513.

4. Bonnet D, Dick JE. Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat Med. 1997;3(7):730-737.

5. Ding Y, Gao H, Zhang Q. The biomarkers of leukemia stem cells in acute myeloid leukemia. Stem Cell Investig. 2017;4:19.

6. Jaiswal S, Jamieson CH, Pang WW, et al. CD47 is upregulated on circulating hematopoietic stem cells and leukemia cells to avoid phagocytosis. Cell. 2009;138(2):271-285.

7. Jin L, Lee EM, Ramshaw HS, et al. Monoclonal antibody-mediated targeting of CD123, IL-3 receptor alpha chain, eliminates human acute myeloid leukemic stem cells. Cell Stem Cell. 2009;5(1):31-42.

8. van Rhenen A, van Dongen GA, Kelder A, et al. The novel AML stem cell associated antigen CLL-1 aids in discrimination between normal and leukemic stem cells. Blood. 2007;110(7):2659-2666.

Funding

Competitive financial support for this work was obtained from Spanish Ministry of Economy and Competitiveness (PID2019108160RB-I00) to PM and (PLE2021-007518) to CB, the Carlos III Health Institute (ISCIII/FEDER PI20/00822) to CB, ISCIII-RICORS within the Next Generation EU program (plan de recuperación, transformación y resiliencia) to PM and Asociación Española Contra el Cancer (AECC) (PRYGN211192BUEN) to CB. SRZ was supported by a Marie Sklodowska Curie Fellosip (GA795833).

Data-sharing statement

Original data is available upon reasonable request to the corresponding author.

9. Jan M, Chao MP, Cha AC, et al. Prospective separation of normal and leukemic stem cells based on differential expression of TIM3, a human acute myeloid leukemia stem cell marker. Proc Natl Acad Sci U S A. 2011;108(12):5009-5014.

10. Bosch M, Sanchez-Alvarez M, Fajardo A, et al. Mammalian lipid droplets are innate immune hubs integrating cell metabolism and host defense. Science. 2020;370(6514):eaay8085.

11. Rezaei M, Tan J, Zeng C, Li Y, Ganjalikhani-Hakemi M. TIM-3 in leukemia; immune response and beyond. Front Oncol. 2021;11:753677.

12. Tirtakusuma R, Szoltysek K, Milne P, et al. Epigenetic regulator genes direct lineage switching in MLL/AF4 leukaemia. Blood. 2022;140(17):1875-1890.

13. Hu Y, Smyth GK. ELDA: extreme limiting dilution analysis for comparing depleted and enriched populations in stem cell and other assays. J Immunol Methods. 2009;347(1-2):70-78.

14. Prieto C, Lopez-Millan B, Roca-Ho H, et al. NG2 antigen is involved in leukemia invasiveness and central nervous system infiltration in MLL-rearranged infant B-ALL. Leukemia. 2018;32(3):633-644.

15. le Viseur C, Hotfilder M, Bomken S, et al. In childhood acute lymphoblastic leukemia, blasts at different stages of immunophenotypic maturation have stem cell properties. Cancer Cell. 2008;14(1):47-58.

Clinical and molecular response of acute myeloid leukemia harboring non-canonical FLT3 N676K driver mutations to contemporary FLT3 inhibitors

Treatment of acute myeloid leukemia (AML) has been enhanced by the development and regulatory approval of several novel agents, including midostaurin and gilteritinib (FLT3 inhibitors), venetoclax (BCL2 inhibitor), ivosidenib (IDH1 inhibitor), and enasidenib (IDH2 inhibitor).1 A challenge during the era of molecular therapies, however, is determining the efficacy of these agents for patients with AML harboring atypical driver mutations. These atypical drivers were underrepresented in seminal clinical trials that led to the approval of targeted AML therapies, thereby limiting availability of data for clinical decision making.2 The non-canonical FLT3 N676K variant was initially described as an acquired resistance mechanism in patients with FLT3 internal tandem duplication (ITD) mutations treated with midostaurin.3 In vivo studies demonstrated FLT3 N676K-mutated AML is sensitive to midostaurin and quizartinib, but suggested that co-operating ITD mutations confer resistance to both agents.4 Clinical reports of FLT3 N676K-mutated AML are limited to those of two individuals, both of whom developed FLT3 N676K mutations at relapse.4,5 Treatment outcomes for de novo disease with FLT3 N676K mutations are lacking, and limited data exist regarding the utility of FLT3 inhibitors for FLT3 N676K-mutated AML patients. The aim of this study was to use clinical and genomic data to investigate the efficacy of FLT3 inhibitors, both as monotherapy and in combination with other agents, for FLT3 N676K-mutated AML. We performed a retrospective analysis of patients with AML receiving care at the University of Chicago. The study was approved by the Institutional Review Board and conducted according to the Declaration of Helsinki. Our practice utilizes a validated 1,213 gene next-generation sequencing (NGS) panel that has been previously described.6-8 NGS is employed at presentation and at subsequent time points to assess response or disease status. In cases of morphologic remission, NGS is not performed due to the anticipated lack of detectable tumor DNA. We identified nine patients with AML and FLT3 N676K mutations. N676K was the only FLT3 mutation detected in seven patients, whereas two patients had co-incident ITD or tyrosine kinase domain (TKD) mutations at some point during their clinical course. Two patients in the study were referred from outside of our institution and had an unknown FLT3 mutation status at first presentation. The remaining seven patients were assessed by upfront NGS, and all patients had molecular reassessment longitudi-

nally, including at relapse, by our in-house assay. The median age at AML diagnosis was 41 years (range, 17-79), with a mean presenting white blood cell (WBC) count of 53,300/µL. Four of seven (57%) patients with available cytogenetic data had normal cytogenetics. Laboratory and clinical data can be found in Table 1.

In order to infer antileukemic activity of FLT3 inhibitors for FLT3 N676K-mutated AML and to characterize relapse dynamics, we analyzed FLT3 N676K variant allele frequency (VAF) kinetics in patients for whom longitudinal NGS data were available. Regardless of disease status at the time of FLT3 inhibitor use (newly diagnosed vs. relapse), patients receiving FLT3 inhibitors had declines in FLT3 N676K VAF. For individuals in whom FLT3 N676K was the only FLT3 mutation (patients 1, 2, 9), a mean treatment time of 95 days led to undetectable FLT3 N676K. Suppression of FLT3 N676K VAF generally paralleled clinical response and likelihood of survival at the time of our analysis (Figure 1A). One exception was patient 9, who achieved morphologic and molecular remission with gilteritinib salvage therapy but who unfortunately died of post-transplant veno-occlusive disease after exposure to gemtuzumab ozogamicin and subsequent allogeneic transplant.

One older patient pursued comfort care immediately after diagnosis. Six patients were treated with “7+3” induction therapy, with three of six patients having FLT3 inhibitors added to induction chemotherapy (midostaurin, n=2; sorafenib, n=1). Patient 1 had a complete morphologic and molecular remission. They proceeded to allogeneic transplantation after induction 7+3 therapy with sorafenib (discontinued for gastrointestinal toxicity) and subsequent midostaurin during pretransplant consolidation. Day 30 post-transplant bone marrow studies demonstrated a complete morphologic and molecular remission on midostaurin maintenance therapy.

Two additional patients (patients 5 and 6) harbored de novo disease with co-occurring FLT3 mutations, one with a FLT3 ITD mutation (patient 5) and the other a FLT3 TKD mutation (patient 6). Each had midostaurin added to 7+3 induction. By days +71 and +32 after midostaurin treatment, respectively, both demonstrated remission with no detectable FLT3 VAF (Figure 1B).

Patient 5 had de novo disease with co-occurring FLT3 mutations (FLT3 ITD F612_G613ins25 and N676K). She received 7+3+midostaurin induction and had a morphologic

and molecular complete remission 71 days after initiating midostaurin. Induction was complicated by fungal pneumonia and repeated episodes of acute kidney injury. She was not a candidate for cytotoxic consolidation therapy

and started gilteritinib. She continued to experience multiple episodes of acute kidney injury unrelated to gilteritinib. Gilteritinib was held during these episodes, and she eventually presented with 25% circulating blasts after

Pt: patient; F: female; M: male; WBC: white blood count; BM: bone marrow; PB: peripheral blood; FLT3i: FLT3 inhibitor; HCT: hematopoietic cell transplant; NA: not available/not applicable; CR: complete remission; PD: progressive disease; NR: not reached; MFC: multiparameter flow cytometry; MRD: measurable residual disease. *For living patients, survival calculated as of manuscript submission and included in parentheses.

Table 1. Clinical and laboratory characteristics of patients with FLT3 N676-mutated acute myeloid leukemia.

approximately 100 days of intermittent gilteritinib administration. NGS at relapse showed ascendancy of the same FLT3 ITD clone (VAF 43%) that was present at diagnosis, but an absence of the FLT3 N676K mutation. The patient chose comfort measures.

Patient 6 had de novo disease with co-occurring FLT3 TKD (D835V) and N676K mutations. She underwent induction with 7+3+midostaurin, which led to morphologic and molecular remission 32 days after initiating midostaurin. She proceeded to hematopoietic cell transplantation in first remission. Despite MRD negativity at transplant, she relapsed after 6 months. She did not receive post-transplant FLT3 inhibitor therapy. NGS performed at relapse demonstrated expansion of the FLT3 N676K population (VAF 33%) and absence of the original FLT3 TKD clone. Salvage measures with donor lymphocyte infusion and high dose cytarabine/mitoxantrone were unsuccessful. She died of complications from central nervous system leukemic infiltration.

We also analyzed the spectrum of other pathogenic mutations co-existing with FLT3 N676K in our cohort (Online Supplementary Figure S1). Co-mutational clusters were most notable for FLT3 N676K and either FLT3 ITD or FLT3 TKD mutations (Figure 1C). In order to understand the

structural properties of therapeutic inhibition of FLT3 in the presence of the N676K mutation, we utilized PyMOL (Schrödinger), an open source molecular graphics tool that is commonly used for visualization of macromolecules, to study the FLT3 TKD harboring N676K in the presence and absence gilteritinib.9 Upon activation, three residues, Asp-Phe-Gly (DFG), shift inward (DFG-in) from the inactive state (DFG-out). Mutations at D835 within the TKD favor the active DFG-in state and promote resistance to type II FLT3 inhibitors.10 Gilteritinib and other type I FLT3 inhibitors bind directly to the ATP-binding site, maintaining their activity regardless of DFG conformation.11 As shown in Figure 2, the N676K mutation does not prohibit the transition from DFG-out to DFG-in or the interaction of gilteritinib with the ATP-binding site.

A recent analysis of the mutational landscape of patients with FLT3-mutated AML treated on CALGB 10603 (RATIFY) showed 26 of 275 (5.5%) patients harbored non-canonical FLT3 mutations.2 Ten of these 26 patients (38%) had FLT3 N676K-mutated AML.2 Growing clinical application of NGS will increase the identification of atypical driver mutations. Robust clinical series focused on FLT3 N676K-mutated AML patients are lacking, and the benefit of FLT3 inhibitor therapies in this population was previously unknown.

Figure 1. Clinical and molecular characterization of acute myleoid leukemia harboring atypical FLT3 N676K mutations. (A) Longitudinal detection of FLT3 N676K by next-generation sequencing from the time of FLT3 inhibitor initiation. (B) Clinical course and relapse dynamics of 2 patients with FLT3 N676K-mutated acute myeloid leukemia (AML) and coincident internal tandem duplication or tyrosine kinase domain mutations. (C) Co-mutational networks of FLT3 N676K-mutated AML seen in 2 or more patients within the cohort. Chord thickness reflects the number of co-occurrences between 2 genes.

visualization models of the uninhibited FLT3 tyrosine

Figure

domain with N676K (residue shown in orange) and activation loop (yellow). Residues F612 and D835 (corresponding to internal tandem duplication and tyrosine kinase domain mutations, respectively) are shown in magenta. (B) Upon activation, the 18-fluoro-deoxyglucose (FDG)-containing α helix (salmon) translocates inward. Inhibition by gilteritinib at the ATP-binding site is depicted in the inactive, FDG-out (C) and activated, FDG-in (D) state with the FDG-containing α helix shown in red.

Here, we used clinical and genomic data to assess the utility of FLT3 inhibitors in the largest series of FLT3 N676K AML patients described to date. Although previously described to be enriched in populations of core binding factor AML,4 none of the seven patients in our cohort with available metaphase cytogenetic data available had corebinding factor AML. Seven of nine (78%) individuals received intensive induction chemotherapy. FLT3 inhibitors were utilized in three patients during frontline induction in combination with cytotoxic chemotherapy, and in five patients during subsequent lines of therapy. We observed reduction, and in some instances, complete molecular suppression of detectable FLT3 N676K VAF on NGS, underscoring the activity of FLT3 inhibitors in this population, regardless of the line of therapy. All patients with FLT3 N676K mutations who were treated with FLT3 inhibitors had a best response of MRD negativity via flow cytometry or NGS at some point during their care.

Consistent with previous evidence,3,4 concurrent canonical ITD or TKD FLT3 mutations were associated with loss of treatment response. In silico modeling of FLT3 in the presence of gilteritinib suggests that the mechanism of

N676K-mediated resistance is not due to disruption of FLT3 inhibitor binding at the ATP-binding site but is likely influenced by other allosteric forces on the protein structure. Individuals with FLT3 N676K-mutated AML in our cohort whose treatment included FLT3 inhibitors had longer median survival (940 days) than those who did not (408 days, excluding patient 4 who immediately pursued comfort measures). This difference was not significant, likely because of the small size of this study (P=0.2). The three patients who remained in an ongoing remission at the time of manuscript submission, however, were all treated with FLT3 inhibitors. With emerging evidence supporting the role of post-transplant FLT3 inhibitor maintenance therapy for suppression of FLT3 ITD-mutated AML,12 further studies evaluating the durability of FLT3 inhibitor maintenance for patients with non-canonical driver mutations in both transplant and non-transplant settings is warranted.

In conclusion, this is the largest study to date demonstrating that the atypical FLT3 N676K driver mutation is sensitive to contemporary FLT3 inhibitors, such as midostaurin and gilteritinib. This mutation has been infre-

2. Macromolecular modeling of the FLT3 tyrosine kinase domain. (A) PyMol

kinase

quently detected in seminal studies of FLT3 inhibitors. However, our data demonstrate FLT3 inhibitors should be included both in upfront induction setting and relapsed/refractory settings for patients harboring the atypical FLT3 N676K mutation.

Authors

Gregory W. Roloff,1 Frank Wen,1 Aubrianna Ramsland,1 Andrew S. Artz,2 Satyajit Kosuri,1 Wendy Stock,1 Olatoyosi Odenike,1 Richard A. Larson,1 Hongtao Liu,1 Lucy A. Godley,1 Michael J. Thirman,1 Anand A. Patel,1 Christopher K. Daugherty,1 Adam S. DuVall,1 Mariam T. Nawas,1 Emily Dworkin,1 Geoffrey D. Wool,3 Sandeep Gurbuxani,3 Carrie Fitzpatrick,3 Jeremy P. Segal,3 Peng Wang3 and Michael W. Drazer1

1Section of Hematology/Oncology, The University of Chicago, Chicago, IL; 2Division of Hematology & Hematopoietic Cell Transplantation, City of Hope, Duarte, CA and 3Department of Pathology, The University of Chicago, Chicago, IL, USA

Correspondence:

M.W. Drazer - mdrazer@medicine.bsd.uchicago.edu

https://doi.org/10.3324/haematol.2022.282148

Received: September 22, 2022.

Accepted: January 9, 2023. Early view: January 19, 2023.

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Disclosures

ASA has acted as a consultant for AbbVie and Magenta Therapeutics. WS has acted as a consultant or advisor to Adaptive Biotechnologies, Jazz Pharmaceuticals, Agios, Kite, a Gilead company, Kura Oncology, GlaxoSmithKline, MorphoSys, Pfizer, Servier, has received honoraria from AbbVie, has received royalties for a chapter in UpToDate, and has received travel accommodation from Pfizer. OO has acted as a consultant for Abbvie, Impact

References

1. Estey EH. Acute myeloid leukemia: 2021 update on riskstratification and management. Am J Hematol. 2020;95(11):1368-1398.

2. Jahn N, Jahn E, Saadati M, et al. Genomic landscape of patients with FLT3-mutated acute myeloid leukemia (AML) treated within the CALGB 10603/RATIFY trial. Leukemia. 2022;36(9):2218-2227.

3. Heidel F, Solem FK, Breitenbuecher F, et al. Clinical resistance to the kinase inhibitor PKC412 in acute myeloid leukemia by

Biomedicines, Celgene, Novartis, BMS, Taiho Pharmaceutical, CTI, Threadwell therapeutics, Bristol-Myers Squibb/Celgene, and has received research support to her institution from Celgene, Daichii Sankyo, Uncyte, Astex Pharmaceuticals, NS Pharma, AbbVie, Janssen Oncology, OncoTherapy Science, Agios, AstraZeneca, CTI BioPharma Corp, Kartos Therapeutics and Aprea AB. RAL has acted as a consultant or advisor to Ariad/Takeda, Celgene/BMS, CVS/Caremark, Epizyme, Immunogen, Novartis, and Servier, and has received clinical research support to his institution from Astellas, Cellectis, Daiichi Sankyo, Forty Seven/Gilead, Novartis, and Rafael Pharmaceuticals, and royalties from UpToDate. HL has acted as a consultant or advisor to Agios, Pfizer, Nkarta, CTI Biopharm, Servier, NGM Biopharma, has acted as a speaker/lecturer for SITC, CAHON, Academy for Continued Healthcare Learning, and has received research support from Miltenyi Biotec. LAG has received royalties from UptoDate, Inc. for a co-authored article on germline predisposition to hematopoietic malignancies. MJT reports grant support from AbbVie, Merck, Syndax, and TG Therapeutics and has received personal fees from AbbVie, Adaptive Biotechnologies, AstraZeneca, Celgene, Pharmacyclics, and Genentech. ASD has acted as a consultant or advisor to Jazz Pharmaceuticals and has served on a speakers’ bureau for Jazz Pharmaceuticals. AAP has received honoraria from AbbVie and research funding from Celgene/BMS, Pfizer and Kronos Bio. CKD has received consulting/advisory fees from Daiichi Sankyo and Sun Pharma. ASD has received fees for consulting and is serving as a member of a speakers’ bureau for Jazz Pharmaceuticals. ED has received honoraria from AbbVie. GDW has received honoraria and has served on an advisory board for Diagnostica Stago. The remaining authors have no conflicts of interest to declare.

Contributions

MWD conceived the study. MWD and GWR developed the concept and design of the study. MWD, GWR, FW and AS collected and analyzed the data. MWD, GWR, ASA, SK, WS, OO, RAL, HL, LAG, MJT, AAP, CKD, ASD, MTN and ED cared for the patients described. GDW, SG, CF, JPS, PW provided pathological support. GWR and MWD drafted the manuscript. All authors contributed to editing the manuscript.

Data-sharing statement

Data from the current work are available on request.

mutation of Asn-676 in the FLT3 tyrosine kinase domain. Blood. 2006;107(1):293-300.

4. Opatz S, Polzer H, Herold T, et al. Exome sequencing identifies recurring FLT3 N676K mutations in core-binding factor leukemia. Blood. 2013;122(10):1761-1769.

5. Daver N, Price A, Benton CB, et al. First report of sorafenib in patients with acute myeloid leukemia harboring non-canonical FLT3 mutations. Front Oncol. 2020;10:1538.

6. Kadri S, Long BC, Mujacic I, et al. Clinical validation of a next-

generation sequencing genomic oncology panel via crossplatform benchmarking against established amplicon sequencing assays. J Mol Diagn. 2017;19(1):43-56.

7. Patel AA, Rojek AE, Drazer MW, et al. Therapy-related myeloid neoplasms in 109 patients after radiation monotherapy. Blood Adv. 2021;5(20):4140-4148.

8. Cahill KE, Karimi YH, Karrison TG, et al. A phase 1 study of azacitidine with high-dose cytarabine and mitoxantrone in high-risk acute myeloid leukemia. Blood Adv. 2020;4(4):599-606.

9. Kawase T, Nakazawa T, Eguchi T, et al. Effect of Fms-like tyrosine kinase 3 (FLT3) ligand (FL) on antitumor activity of gilteritinib, a FLT3 inhibitor, in mice xenografted with FL-

overexpressing cells. Oncotarget. 2019;10(58):6111-6123.

10. Smith CC, Lin K, Stecula A, Sali A, Shah NP. FLT3 D835 mutations confer differential resistance to type II FLT3 inhibitors. Leukemia. 2015;29(12):2390-2392.

11. Larrosa-Garcia M, Baer MR. FLT3 inhibitors in acute myeloid leukemia: current status and future directions. Mol Cancer Ther. 2017;16(6):991-1001.

12. Xuan L, Wang Y, Huang F, et al. Sorafenib maintenance in patients with FLT3-ITD acute myeloid leukaemia undergoing allogeneic haematopoietic stem-cell transplantation: an openlabel, multicentre, randomised phase 3 trial. Lancet Oncol. 2020;21(9):1201-1212.

Clinical and molecular features of familial chronic lymphocytic leukemia: a pilot monocentric study

Chronic lymphocytic leukemia (CLL) is the most common leukemia in the western world with a median age at diagnosis of 70 years. CLL is characterized by accumulation of monoclonal immune-disrupted B cells in peripheral blood, bone marrow, spleen, and lymph nodes. A family history of CLL is a risk factor to developing the disease.1 Familial CLL is defined as the presence of at least one first degree relative with CLL. The prevalence of familiarity for blood diseases, and more specifically for CLL, reported in the literature is 13% and 7%, respectively.2 A study investigating clinical features of familial CLL showed a 10year survival probability similar to sporadic cases (67% vs 66%);3 however, no data regarding molecular characteristics or genetic predisposition were provided. Genetic anticipation is the phenomenon of earlier onset or increased severity when a disease is passed to the next generation. Some studies on familial CLL showed a 10-20 year earlier onset and a more severe phenotype in younger generations,2,4,5 while some subsequent evidence refuted these results.3 Regarding molecular characteristics of sporadic and familial CLL, a study showed higher frequency of mutated IgHV in familial CLL, with intrafamilial concordance of the mutation status.6 Another study showed that 86% of familial CLL cases carried a chromosome 13q deletion,7 compared with 50-60% of patients with sporadic CLL from large case series,7 but no further association with somatic mutations has been described.

The primary aim of our study is to determine the prevalence of familial CLL and to confirm the phenomenon of anticipation. Novel topics of investigation include the comparison of clinical features, molecular biomarkers, survival analysis, and response to treatment in familial CLL compared to sporadic CLL cases.

Our retrospective monocentric study consists of collection of clinical and biological data from all patients with familial CLL. It was approved by the local ethical committee (protocol n. 0014904/22; approved on April 29, 2022) and has been carried out according to the principles of the Declaration of Helsinki. All patients gave signed informed consent. Patients’ characteristics were described by frequency tables for qualitative variables and position indicators for quantitative variables. The associations with clinical-biological parameters were analyzed using χ2 or Fisher's exact test for the qualitative variables and Wilcoxon or Kruskal-Wallis test for quantitative variables.

For the monocentric phase of the study described here, 500 patients were recruited: 46 familial CLL and 454 sporadic CLL. All patients were initially diagnosed in our

center; no patient was referred from other hospitals on the grounds of known familiarity or dismal prognostic features. The prevalence of familial CLL in our population was 9.4%, while the prevalence of familiarity for any other hemopathy was 20%.

Median age at diagnosis was 65 years in sporadic CLL and 59 years in familial CLL (P=0.018). Considering only familial CLL, the median age of the first generation of affected patients at diagnosis was 69 years (range 50-90). For the second generation of affected patients, the median age at diagnosis was 57 years (range 31-83), a statistically significant mean difference between these two populations of 12.6 years (range 17.5-7.7, 95%CI) (t test, P=0.00003) (Figure 1).

Concerning the IgHV mutational status, sporadic CLL showed mutated IgHV in 65% of cases and unmutated IgHV in 35%, while familial CLL showed mutated IgHV in 35% of cases and unmutated IgHV in 65% (P<0.001). In contrast, no statistically significant difference was found for either FISH or for molecular biomarkers. FISH was negative in 31% of patients with sporadic CLL and 24% with familial CLL; del17p: 4.2% of sporadic CLL and 8.7% of familial CLL; del13q: 27% of sporadic CLL and 37% of familial CLL; del11q: 5.3% of sporadic CLL and 6.5% of familial CLL; trisomy 12: 8.4% of sporadic CLL and 4.3% of familial CLL; other alterations or combinations of the above were found in 24% of patients with sporadic CLL and 20% of patients with familial CLL (P=0.38). TP53 was mutated in 15% of patients with sporadic CLL (291 unknown) and 16% with familial CLL (14 unknown) (P>0.99); NOTCH1 was mutated in 23% of patients with sporadic CLL (241 unknown) and 37% with familial CLL (16 unknown) (P=0.18); SF3B1 was mutated in 16% of patients with sporadic CLL (246 unknown) and 11% of patients with familial CLL (19 unknown) (P=0.67); BIRC3 was mutated in 7.7% of patients with sporadic CLL (246 unknown) and 15% of patients with familial CLL (19 unknown) (P=0.38). The molecular characteristics of our patient population are reported in Table 1.

No difference was found in clinical or laboratory parameters. The prevalence of Richter syndrome was 1.7% in sporadic CLL and 1.4% in familial CLL (P=0.46). The prevalence of a second neoplasia was 15% in sporadic CLL and 18% in familial CLL (P>0.99).

Median follow-up time was 61 months (range 0-383). Median overall survival (OS) for familial CLL was 379 months and 296 months for the control group (P=0.45). At the univariate analysis, family history for CLL did not

significantly impact OS (HR=0.73, 95%CI: 0.17-3.27; P=0.68).

Progression-free survival (PFS) was calculated from the date of start of first-line treatment to the date of disease progression or death, whichever was reported first. Patients still alive without signs of progression were censored at the last date of follow up as free from progression or lost at follow-up. Median PFS was 24 months for sporadic CLL and 18 months for familial CLL (P=0.093). Univariate analysis showed family history for

Table

CLL did not significantly impact PFS (HR=1.46, 95%CI: 0.89-2.40; P=0.13).

There was no statistically significant difference in median time to treatment (TTT) between patients with sporadic CLL (68 months) and patients with familial CLL (51 months) (P=0.94). Univariate analysis showed family history for CLL (HR=0.98, P=0.92) did not significantly impact TTT. The median time to next treatment (TTNT) was 242 months for patients with sporadic CLL and 169 months for patients with familial CLL; nevertheless, log-rank test

1. Biological characteristics in terms of IgHV mutational status, FISH, and molecular lesions (TP53, NOTCH1, SF3B1, BIRC3) of patients with familial chronic lymphocytic leukemia (CLL) compared to sporadic CLL.

Figure 1. Mean difference in terms of age at diagnosis between first and second generation of patients with familial chronic lymphocytic leukemia. The mean difference in terms of age at diagnosis between first and second generations of patients with familial chronic lymphocytic leukemia was 12.6 years (range 17.5-7.7, 95%CI, P=0.00003). (A) Histogram of age difference. (B) Probability plot of age difference.

showed this difference was not statistically significant (P=0.54) (Figure 2).

Regarding treatment regimens, 203 out of 454 patients from the sporadic CLL and 20 out of 46 patients from the familial CLL groups underwent treatment: 72% and 70% of them received chemo-immunotherapy (CIT), and 28% and 30% received new drugs, respectively. There was no difference in the treatment received by the two groups (P>0.99). Overall response rate (ORR) was 87% for sporadic CLL and 68% for familial CLL, showing that patients with familial CLL obtained significantly lower rates of overall response (P=0.034), despite the fact that the proportion of patients treated with CIT and with new drugs was similar.

Our result shows a 9.4% prevalence of familial CLL in our

population, similar to the 7% reported in the literature. Patients with familial CLL are younger at diagnosis, but, and more importantly, our data confirmed the phenomenon of anticipation: the second generation has a markedly earlier onset of disease, as reported in the literature (mean difference in our population was 12.6 years compared with 10-20 years in the literature),5,8 suggesting an additive mutational effect mediated by one or a group of defective genes.

Our analysis did not identify any significant molecular biomarker. The only statistically significant finding was a higher rate of unmutated IgHV in familial cases (65% of patients with unmutated IgHV) conferring a worse prognostic profile, the opposite trend to that previously observed in the literature (68% of patients with mutated

Figure 2. Survival analysis did not show any statistically significant difference between patients with familial and sporadic chronic lymphocytic leukemia. (A) Curves of overall survival (OS) for familial (blue) and sporadic (red) chronic lymphocytic leukemia (CLL). (B) Curves of progression-free survival (PFS) for familial (blue) and sporadic (red) CLL. (C) Curves of time to treatment (TTT) for familial (blue) and sporadic (red) CLL. (D) Curves of time to next treatment (TTNT) for familial (blue) and sporadic (red) CLL.

IgHV).6 FISH analysis showed slightly higher rates of 17p and 13q deletions and lower rates of trisomy 12 in familial cases, without statistical significance.

Rai/Binet staging, laboratory parameters and clinical characteristics were substantially identical among the two groups, confirming that, from a clinical point of view, sporadic and familial CLL are indistinguishable.2 There was no statistical difference in the rate of Richter transformation between familial and sporadic cases (1.4% vs. 1.7%), confirming data from the literature (6% vs. 5%). The rate of secondary malignancy was homogeneous in the two populations (16% vs. 18%),8 unlike that published in 2008 (8.8% vs. 16%); this difference could be dictated by the current use of target therapies which reduces the risk of secondary malignancy from chemotherapy.

Shorter PFS, TTT and TTNT, even if not statistically significant, would suggest the trend toward an aggressive course of familial CLL. The OS was superimposable, as already reported in the literature.3 Nevertheless, the ORR was significantly lower in patients with familial CLL, even though the rate of CIT and new drugs was equal in the two groups, a result which has never been reported by any previous study and which confirms the aggressive course of the disease.

To validate our statistically significant findings on anticipation and poor ORR, to define the molecular characteristics of familial CLL, and to improve the significance of the survival analysis, we need to study a larger population of patients. Our monocentric experience will soon be extended to other centers in Italy and opens the way to a better understanding of CLL clustering in families.

Authors

Giulia Benintende,1* Idanna Innocenti,2* Alberto Fresa,2 Francesco Autore,2 Annamaria Tomasso,1 Alfonso Piciocchi,3 Florenzia Vuono,2 Luca Stirparo,1 Antonio Mosca,1 Andrea Bacigalupo,1 Valter Gattei,4 Dimitar Efremov,5 Eugenio Sangiorgi6 and Luca Laurenti1

References

1. Karakosta M, Delicha EM, Kouraklis G, Manola KN. Association of various risk factors with chronic lymphocytic leukemia and its cytogenetic characteristics. Arch Environ Occup Health. 2016;71(6):317-329.

2. Goldin LR, Slager SL, Caporaso NE. Familial chronic lymphocytic leukemia. Curr Opin Hematol. 2010;17(4):350-355.

3. Mauro FR, Giammartini E, Gentile M, et al. Clinical features and outcome of familial chronic lymphocytic leukemia. Haematologica. 2006;91(8):1117-1120.

4. Horwitz M, Goode EL, Jarvik GP. Anticipation in familial leukemia. Am J Hum Genet. 1996;59(5):990-998.

5. Wiernik PH, Ashwin M, Hu XP, Paietta E, Brown K. Anticipation in

1Sezione di Ematologia, Dipartimento di Scienze Radiologiche ed Ematologiche, Università Cattolica del Sacro Cuore, Rome; 2Dipartimento di Diagnostica per Immagini, Radioterapia Oncologica ed Ematologia, Fondazione Policlinico Universitario “A. Gemelli” IRCCS, Rome; 3Data Center and Health Outcomes Research Unit, Italian Group for Adult Hematologic Disease (GIMEMA), Rome; 4Clinical Experimental Onco-Haematology Unit, Centro di Riferimento Oncologico di Aviano (CRO) IRCCS, Aviano; 5Molecular Hematology, International Centre for Genetic Engineering & Biotechnology, Science Park Padriciano 99, Trieste and 6Sezione di Medicina Genomica, Dipartimento di Scienze della Vita e Sanità Pubblica, Università Cattolica del Sacro Cuore - Fondazione Policlinico Universitario “A. Gemelli” IRCCS, Rome, Italy

*GB and II contributed equally as first authors.

Correspondence:

G. BENINTENDE - giuliabenintende97@gmail.com

https://doi.org/10.3324/haematol.2022.282268

Received: October 14, 2022.

Accepted: December 14, 2022.

Early view: December 22, 2022.

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Disclosures

No conflicts of interest to disclose.

Contributions

GB, II and AF collected data, analyzed results and wrote the manuscript. FA, AT, FV, LS, SM collected data. AP performed the statistical analysis. EG provided the genetical background. VG performed molecular biology analysis. AB, DE and LL supervised the project.

Data-sharing statement

Data only available on request due to privacy/ethical restrictions.

familial chronic lymphocytic leukaemia. Br J Haematol. 2001;113(2):407-414.

6. Crowther-Swanepoel D, Wild R, Sellick G, et al. Insight into the pathogenesis of chronic lymphocytic leukemia (CLL) through analysis of IgVH gene usage and mutation status in familial CLL. Blood. 2008;111(12):5691-5693.

7. Ng D, Toure O, Wei MH, et al. Identification of a novel chromosome region, 13q21.33-q22.2, for susceptibility genes in familial chronic lymphocytic leukemia. Blood. 2007;109(3):916-925.

8. Ishibe N, Sgambati MT, Fontaine L, et al. Clinical characteristics of familial B-CLL in the National Cancer Institute Familial Registry. Leuk Lymphoma. 2001;42(1-2):99-108.

Activity of decitabine combined with all-trans retinoic acid in oligoblastic acute myeloid leukemia: results from a randomized 2x2 phase II trial (DECIDER)

Treatment options for patients with high-risk myelodysplastic syndrome (MDS) ineligible for allogeneic stem cell transplantation (allo-SCT) are limited. DNA-hypomethylating agents (HMA) form the backbone of the treatment of these patients, but only very rarely provide long-term survival as single agents.1–5 The addition of venetoclax to HMA proved to be a highly active treatment in acute myeloid leukemia (AML)6 and this concept was successfully adapted in studies for high-risk MDS.7 Within the DECIDER trial (clinicaltrails gov. Identifier: NCT00867672), the combination of decitabine (DEC) and all-trans retinoic acid (ATRA) also resulted in an improved response rate and survival in AML compared to DEC alone, and was likewise active in patients with prior hematologic disorder (mostly MDS).8 The prospective, randomized, observer-blind, active-control, parallel-group, multicenter, phase II DECIDER trial has a 2x2 factorial design and compared treatment with DEC alone to DEC +/- ATRA and +/- valproic acid (VPA) in newly diagnosed AML patients ineligible for alloSCT. No benefit was seen when VPA was added to the treatment. We here present an exploratory, not preplanned subgroup analysis of the DECIDER trial, where we evaluated the effect of the combination of DEC and ATRA on patients with 20-30% bone marrow blasts. This subgroup is often referred to as oligoblastic AML, formerly RAEB-T according to the French-American-British classification.9,10 The new International Consensus Classification and the new classification of myeloid neoplasms of the World Health Organization are focused on biological and genetic drivers of myeloid malignancies for the prognosis of affected patients. Nevertheless, we believe that the result of this analysis could indicate whether the combination of HMA and a retinoid could also be active in high-risk MDS patients with excess blasts, since the presence of blasts is still an independent high-risk feature, which is reflected in the new classifications.

Patients and methods of the DECIDER trial have been reported previously.8 In short, patients were randomly assigned to four different arms with DEC 20 mg/m2 day 1-5 (treatment arms A/B/C/D), ATRA orally (p.o.) day 6-28 (arms C/D), VPA p.o. continuously from day 6 (arms B/D) of each 28-day course (repeated until relapse/progression, prohibitive toxicity, withdrawal or death) (Figure 1). Study endpoints were objective response rate (ORR), defined as complete remission with or without count recovery and partial remission (CR/CRi/PR), and overall survival (OS)

time. For patient characteristics and methods of the statistical analysis, we refer to the original publication, emphasizing again that all analyses were not preplanned in the study protocol and are, thus of exploratory nature.

Between December 2011 and February 2015, 200 patients were randomly assigned and received study medication at 27 centers.8 Of these, 56 fulfilled the criteria of an oligoblastic AML with 20-30% (median 24.6%) blasts by central hematopathology. Patient and disease characteristics are shown in the Online Supplementary Table S1. Of the 56, 22 (39.3%) were treated with DEC + ATRA +/- VPA, with nine (40.9%) in arm C and 13 (59.1%) in arm D, in the following referred to as the “ATRA” group, and 34 (60.7%) were treated with DEC +/- VPA, with 13 (38.2%) in arm A and 21 (61.8%) in arm B, in the following referred to as the “no ATRA” group.

The majority (76.8%) of patients was male, but sex was evenly balanced between the ATRA versus no ATRA groups. The median age (75 years) was similar in both groups, however the proportion of patients >= 75 years of age was higher in the ATRA group (68.2% vs. 41.2%). Regarding genetic risk according to the 2010 European LeukemiaNet genetic risk classification, the ATRA group displayed a higher proportion of intermediate-risk (77.3% vs. 55.9%) and a lower proportion of adverse-risk (13.6% vs. 32.4%) patients, which constitutes a certain limitation to this analysis. All other characteristics like Eastern Cooperative Oncology Group performance status (ECOG PS), comorbidities, prior hematologic disorder or white blood count (WBC) were evenly balanced.

For the entire subgroup, a median of five DEC courses were administered (a median of 2 in arm A, 5 in arm B, 11 in arm C and 4 in arm D), resulting in a median of 7.5 courses in the ATRA group and 3.5 courses in the no ATRA group. In total, six patients attained a CR (9.1% vs. 11.8% in the ATRA vs. no ATRA group), seven patients a CRi (18.2% vs 8.8% respectively), one patient a PR (4.6% vs. 0.0% respectively), ten patients had an antileukemic effect (ALE) (18.2% vs. 17.7% respectively), ten patients had stable disease (27.3% vs. 11.8% respectively) and 22 patients progressive disease (22.7% vs. 50.0% respectively) (Online Supplementary Table S2). The ORR was 25%, with a difference between the ATRA and the no ATRA groups of 31.8% versus 20.6% with an OR of 1.85 (95% confidence interval [CI]: 0.54- 6.37) and a two-sided P value of 0.33 (Table 1). The ORR of patients additionally treated with VPA versus

Figure 1. Effects of the addition of all-trans retinoic acid and valproic acid to decitabine on overall survival. (A) Overall survival (OS) according to treatment arms: A: black, B: red, C: blue, D: green. (B) OS according to treatment with decitabine plus all-trans retinoic acid (DEC + ATRA) (+/- valproic acid [VPA]) (red curves) compared to DEC (+/- VPA) (blue curves, Kaplan-Meier method). Solid curves: unadjusted; broken curves: adjustment for Eastern Cooperative Oncology Group performance status, comorbidities (hematopoietic cell transplantation index), serum lactate dehydrogenase, hemoglobin, genetic risk (ELN 2010). (C) OS according to treatment with DEC + VPA (+/- ATRA) (red curves) compared to DEC (+/- ATRA) (blue curves, Kaplan-Meier method). Solid curves: unadjusted; broken curves: Eastern Cooperative Oncology Group performance status, comorbidities (hematopoietic cell transplantation index), serum lactate dehydrogenase, hemoglobin, genetic risk (ELN 2010).

*Adjustment for Eastern Cooperative Oncology Group performance status, hematopoietic cell transplantation index, serum lactate dehydrogenase, hemoglobin, genetic risk. ORR: objective response rate; CI: confidence interval; DEC: decitabine; ATRA: all-trans retinoic acid; VPA: valproic acid.

no VPA was 29.4% versus 18.2% with an OR of 1.93 (95% CI: 0.51-7.24) and a two-sided P value of 0.33 (Table 1). The median follow-up time for OS was 6.2 years. With 48 deaths of 56 patients, the median OS time of the whole subgroup analysis population was 9.1 months (arm A: 7.6 months, arm B: 8.9 months, arm C: 19.0 months, arm D: 11.2 months) (Table 1; Figure 1A). A comparison of the ATRA and the no ATRA group resulted in median OS time of 11.5 versus 7.6 months, respectively, with an unadjusted hazard ratio [HR] of 0.71 (95% CI: 0.40-1.29) and a two-sided P value of 0.26 (Table 1; Figure 1B). Adjustment for ECOG PS, hematopoietic cell transplantation comorbidity index (HCT-CI), serum lactate dehydrogenase (sLDH), hemoglobin, and genetic risk led to similar results (adjusted HR=0.61; 95% CI: 0.32-1.16; P=0.13) (Table 1; Figure 1B). Although no statistically significant difference could be detected, the survival curves of the ATRA and the no ATRA group were separating in the first 2 years after therapy initiation (Figure 1B). By the addition of VPA to the treatment, no difference in OS was observed (median OS: VPA: 10.0 months vs. no VPA: 8.4 months, unadjusted HR=0.89; 95% CI: 0.49-1.61; P=0.71, adjusted HR=0.91; 95% CI: 0.431.93; P=0.80) (Table 1; Figure 1C) and the survival curves did not separate at any given time point.

In this subgroup analysis of the DECIDER trial, the addition of ATRA to DEC (+/-VPA) resulted in higher ORR and OS rates in elderly patients with oligoblastic AML ineligible for induction chemotherapy, with no added toxicity, as shown in the original publication.8 Although the results were not statistically significant, the Kaplan-Meier plots are suggestive of a clinically relevant difference.

In our analysis of the whole study population of the DECIDER trial it was shown that the addition of ATRA did not only improve the ORR but also led to a prolonged response duration.8 We thus reasoned that ATRA stabilizes the response to HMA therapy and leads to delayed emergence

of resistance. The number of patients with oligoblastic leukemia was not sufficient to perform this type of analysis, but we also did not observe any hint that this population should respond differently.

A high proportion of the patients (67.9%) with oligoblastic leukemia had a preceding hematologic disorder, mainly MDS. Due to of the diagnostic continuum from MDS to secondary AML separated by the arbitrary number of 20% bone marrow or peripheral blood blasts at the time of this study, the higher proportion of patients in this subgroup, compared to 51% in the whole study population, does not come as a surprise. The fact that the addition of ATRA to DEC is also leading to clinical benefit of the subgroup of oligoblastic leukemia patients could therefore be sufficient to claim that this combination treatment could be efficacious in MDS patients with excess blasts.

The effort to establish retinoic acid as a therapeutic agent in MDS dates back to the 1980s, however without proof of single-agent activity.11,12 Over 10 years ago, two phase II trials from Paris and the MD Anderson Cancer Center demonstrated clinical activity of HMA + ATRA + VPA in AML and MDS.13,14 Our recent study was the first randomized trial to prove the feasibility of the combination of HMA + ATRA in AML.8 Recently, we also found in vitro and in vivo evidence for co-operation of decitabine and ATRA.15 It is only logical to also launch a randomized trial with this combination in high-risk MDS as conducted by the group of Dr. Tong.16 Since the combination of HMA + venetoclax is likely to also play an important role in the treatment of MDS,7 a combination of HMA + venetoclax + ATRA is a rational study concept, advanced by us for AML (DECIDER-2 trial). Since the main limitation of HMA + venetoclax in AML is the emergence of resistance6 and ATRA delayed emergence of resistance in the entire DECIDER study population without adding any toxicity, it appears as a rational partner for a reduced-toxicity triple therapy.

Table 1. Effects of decitabine + all-trans retinoic acid (+/- valproate) versus decitabine - all-trans retinoic acid (+/- valproate) on objective response and overall survival.

Authors

Christoph Rummelt,1 Olga Grishina,² Claudia Schmoor,2 Martina Crysandt,3 Michael Heuser,4 Katharina S. Götze,5,6 Richard F. Schlenk,7,8,9 Konstanze Döhner,7 Helmut R. Salih,10 Gerhard Heil,11 Carsten Müller-Tidow,8,12,13 Wolfram Brugger,14 Andrea Kündgen,15,16

Maike de Wit,17 Aristoteles Giagounidis,18 Sebastian Scholl,19 Andreas Neubauer,20 Jürgen Krauter,21 Gesine Bug,22,23 Haifa Kathrin Al-Ali,24

Ralph Wäsch,1 Heiko Becker,1,25 Annette M. May,26 Justus Duyster,1,25 Björn Hackanson,1,27 Arnold Ganser,4 Hartmut Döhner8 and Michael Lübbert1

1Department of Hematology, Oncology and Stem Cell Transplantation, University Medical Center Freiburg, Faculty of Medicine, University of Freiburg, Freiburg; 2Clinical Trials Unit, University Medical Center Freiburg, Faculty of Medicine, University of Freiburg, Freiburg; 3Department of Hematology, Oncology, Hemostaseology and Stem Cell Transplantation, University Hospital

RWTH Aachen University, Aachen; 4Department of Hematology, Hemostasis, Oncology and Stem Cell Transplantation, Hannover Medical School, Hannover; 5Department of Medicine III, Technical University of Munich, Munich; 6German Cancer Consortium (DKTK) and German Cancer Research Center (DKFZ), Partner Site Munic, München; 7Department of Internal Medicine III, University Hospital of Ulm, Ulm; 8Department of Internal Medicine V, Heidelberg University Hospital, Heidelberg; 9National Center of Tumor Diseases, NCT-Trial Center, German Cancer Research Center, Heidelberg; 10Clinical Collaboration Unit Translational Immunology, German Cancer Consortium (DKTK) and German Cancer Research Center (DKFZ), Partner Site Tübingen, Tübingen; 11Department of Hematology/Oncology, Klinikum Luedenscheid, Luedenscheid; 12German Cancer Consortium (DKTK) and German Cancer Research Center (DKFZ), Partner Site Heidelberg, Heidelberg; 13Department of Medicine A, University Hospital of Münster, Münster; 14Department of Hematology, Hospital Villingen-Schwenningen, VillingenSchwenningen; 15Department of Hematology, Oncology and Clinical Immunology, Heinrich-Heine-University, Faculty of Medicine, Düsseldorf; 16German Cancer Consortium (DKTK) and German Cancer Research Center (DKFZ), Partner Site Düsseldorf, Düsseldorf; 17Department of Hematology and Oncology, Vivantes Klinikum Neukölln, Berlin; 18Clinic for Oncology, Hematology and Palliative Medicine, Marien-Hospital Düsseldorf, Düsseldorf; 19Department of Hematology and Oncology, Universitätsklinikum Jena, Klinik für Innere Medizin II, Jena; 20Department of Hematology and Oncology, University Clinic Gießen/Marburg, Marburg; 21Department of Internal Medicine III, Städtisches Klinikum Braunschweig, Braunschweig; 22Department of Medicine II, Hematology and Oncology, University Hospital Frankfurt, Goethe University, Frankfurt; 23German Cancer Consortium (DKTK) and German Cancer Research Center (DKFZ), Partner Site Frankfurt, Frankfurt; 24Krukenberg Cancer Center, University Hospital Halle, Halle; 25German Cancer Consortium (DKTK) and German Cancer

Research Center (DKFZ), Partner Site Freiburg, Freiburg; 26Institute for Surgical Pathology, University Medical Center Freiburg, Faculty of Medicine, University of Freiburg, Freiburg and 27Department of Hematology/Oncology, Universitätsklinikum Augsburg, Augsburg, Germany

Correspondence: M. LUEBBERT - michael.luebbert@uniklinik-freiburg.de

https://doi.org/10.3324/haematol.2022.282258

Received: October 10, 2022.

Accepted: December 23, 2022.

Early view: January 5, 2023.

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Disclosures

HB received honoraria from BMS, Novartis, Pierre Faber GmbH, Roche and Servier. All other authors have no conflicts of interest to disclose.

Contributions

ML, OG, CS, BH, HD and AG developed the concept and designed the study; CR, OG, MC, MH, KG, RS, KD, HS, CM-T, WB, AK, MW, AG, SS, AN, JK, GB, HA-A, RW, HB, BH, AG, HD and ML collected and assembled data; ML, OG, CS, CR, RS, HS, CM-T, WB, AG, AN, JK, MW, RW, HB, JD, BH and HD analyzed and interpreted data; ML, RS, EJ, MH, HS, AK, KG, GH, SS, GB, AG, AN, JK, WB, MW, RW, KD, BH and HD provided study materials or recruited patients; OG and BH provided administrative support. All authors wrote the manuscript, are accountable for all aspects of the work and approved the final version of the manuscript.

Acknowledgments

We wish to thank the members of the Independent Data Monitoring Committee for providing guidance during the trial. At the Clinical Trials Unit (Freiburg, Germany), we thank Caroline Cieslik for excellent organizational support and Angelika Gerlach and Inga Steinbrenner for assistance in statistical calculations.

Funding

The study was supported by the German Federal Ministry of Education and Research (BMBF, Clinical Trials Program Grant No 01KG0913). Decitabine was provided by Janssen-Cilag, valproic acid was provided by TEVA. The study was further supported by the German Research Foundation (DFG consortia FOR2674, CRC992) for translational studies conducted by ML (A05, C04), HB (A05), HD (A02).

Data-sharing statement

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

References

1. Silverman LR. Demakos EP, Peterson BL, et al. Randomized controlled trial of azacitidine in patients with the myelodysplastic syndrome: a study of the cancer and leukemia group B. J Clin Oncol. 2002;20(10):2429-2440.

2. Kantarjian H, Issa JPJ, Rosenfeld CS, et al. Decitabine improves patient outcomes in myelodysplastic syndromes: results of a phase III randomized study. Cancer. 2006;106(8):1794-1803.

3. Lübbert M, Suciu S, Baila L, et al. Low-dose decitabine versus best supportive care in elderly patients with intermediate- or high-risk myelodysplastic syndrome (MDS) ineligible for intensive chemotherapy: final results of the randomized phase III study of the European Organisation for Research and Treatment of Cancer Leukemia Group and the German MDS Study Group. J Clin Oncol. 2011;29(15):1987-1996.

4. Zeidan AM, Davidoff AJ, Long JB, et al. Comparative clinical effectiveness of azacitidine versus decitabine in older patients with myelodysplastic syndromes. Br J Haematol. 2016;175(5):829-840.

5. Maakaron, J. E. et al. Hypomethylating agents super-responders: challenging the dogma of long-term remission for acute myeloid leukemia. Ann. Hematol. 2020;99:1411-1413.

6. DiNardo CD, Jonas BA, Pullarkart V, et al. Azacitidine and venetoclax in previously untreated acute myeloid leukemia. N Engl J Med. 2020;383(7):617-629.

7. Ball BJ, Famulare CA, Stein EM, et al. Venetoclax and hypomethylating agents (HMAs) induce high response rates in MDS, including patients after HMA therapy failure. Blood Adv. 2020;4(13):2866-2870.

8. Lübbert M, Grishina O, Schmoor C, et al. Valproate and retinoic acid in combination with decitabine in elderly nonfit patients with acute myeloid leukemia: results of a multicenter, randomized, 2 × 2, phase II trial. J Clin Oncol. 2020;38(3):257-270.

9. Bacher U, Alpermann T, Schnittger S, et al. Prognosis in patients with MDS or AML and bone marrow blasts between 10% and 30% is not associated with blast counts but depends on cytogenetic and molecular genetic characteristics. Leukemia. 2011;25(8):1361-1364.

10. Lichtman MA. Does a diagnosis of myelogenous leukemia require 20% marrow myeloblasts, and does <5% marrow myeloblasts represent a remission? The history and ambiguity of arbitrary diagnostic boundaries in the understanding of myelodysplasia. Oncologist. 2013;18(9):973-980.

11. Gold E J, Mertelsmann RH, Itri LM, et al. Phase I clinical trial of 13cis-retinoic acid in myelodysplastic syndromes. Cancer Treat Rep. 1983;67(11):981-986.

12. Koeffler HP, Heitjan D, Mertelsmann R, et al. Randomized study of 13-cis retinoic acid v placebo in the myelodysplastic disorders. Blood. 1988;71(3):703-708.

13. Raffoux E, Cras A, Recher C, et al. Phase 2 clinical trial of 5azacitidine, valproic acid, and all-trans retinoic acid in patients with high-risk acute myeloid leukemia or myelodysplastic syndrome. Oncotarget. 2010;1(1):34-42.

14. Soriano AO, Yang H, Faderl S, et al. Safety and clinical activity of the combination of 5-azacytidine, valproic acid, and all-trans retinoic acid in acute myeloid leukemia and myelodysplastic syndrome. Blood. 2007;110(7):2302-2308.

15. Meier R, Greve G, Zimmer D, et al. The antileukemic activity of decitabine upon PML/RARA-negative AML blasts is supported by all-trans retinoic acid: in vitro and in vivo evidence for cooperation. Blood Cancer J. 2022;12(8):122.

16. Zhou X, Meng F, Lin Y, et al. Combination of decitabine and ATRA in newly diagnosed myelodysplastic syndromes subtype EB-interim analysis of a multicenter, randomized, open-label trial. Blood. 2021;138(Suppl 1):S539.

Efficacy, safety, and cost of mobilization strategies in multiple myeloma: a prospective, observational study

High-dose therapy followed by autologous hematopoietic cell transplant (AHCT) has been the standard-of-care for eligible patients with newly diagnosed multiple myeloma.1,2 Mobilization of peripheral blood hematopoietic stem and progenitor cells (HSPC) can be accomplished using either granulocyte colony-stimulating factor (G-CSF) alone or in combination with chemotherapy (cyclophosphamide 3 g/m2) or plerixafor.3,4 None of these strategies has been prospectively compared. Given the higher drug costs of plerixafor, its use has been limited in several centers as a rescue in patients at high risk of mobilization failure with G-CSF alone (the so-called “just-in-time” [JIT] approach). In this prospective, multicenter, observational study, we sought to evaluate the efficacy of various mobilization regimens and their impact on post-AHCT outcomes and conduct a health economics analysis in Medicare beneficiaries.

Eligibility criteria included patients 18-70 years of age undergoing AHCT within 12 months of induction treatment for multiple myeloma and reported consecutively from participating Center for International Blood and Marrow Transplant Research (CIBMTR) centers (N=20) (NCT 03200626). For the health economic analysis, we used an independent dataset from the Centers for Medicare and Medicaid Services (CMS) of Medicare patients merged with the CIBMTR database who underwent first AHCT in 2016 in the USA.

The primary objective of the prospective study was to compare the total CD34+ cell yield of HSPC mobilization in patients with multiple myeloma undergoing mobilization with: (i) a G-CSF-based strategy (alone or using JIT plerixafor; G±JIT cohort), (ii) a planned G-CSF and plerixafor combination (G+P), and (iii) G-CSF plus chemomobilization (G+C), (Online Supplementary Table S1). The objective of the health economics analysis was to determine service utilization and costs associated with mobilization regimens and was a retrospective analysis of a separate cohort of Medicare beneficiaries who underwent transplants in 2016 comparing: (i) G-CSF alone (G), (ii) GCSF plus plerixafor, planned or JIT (G+P or JIT), and (iii) G-CSF ± plerixafor plus chemomobilization (G+C). Patient- and transplant-related characteristics were described and compared between the three mobilization strategies. A Cox proportional hazards model and a Fine and Gray subdistribution hazards model including patient-, disease- and treatment-related variables were used to assess hematopoietic recovery, non-relapse mortality, relapse/progression, progression-free survival and overall survival.

For the cost analysis, observable total Medicare allowable costs associated with service resources utilized during mobilization were estimated in 2016 US dollars. Post-hoc analyses were conducted to compare outpatient costs and total costs between pair groups by mobilization strategy for each period. The method of cost-to-charge ratio analysis was applied to estimate total inpatient and outpatient costs from the provider’s perspective.

A total of 750 patients were enrolled of whom 744 met all inclusion criteria (Online Supplementary Table S1). The three groups were well balanced overall. Twenty patients in the G+C group received plerixafor as rescue. Twelve patients underwent a second mobilization attempt.

Table 1 summarizes the efficacy parameters of the three mobilization strategies. The median CD34+ cell yield in the first apheresis session was significantly lower in the G±JIT cohort (4.7x106/kg) than in the G+P cohort (6.4x106/kg) or G+C cohort (6x106/kg) (P<0.01). The total CD34+ cell yield was also significantly lower in the G±JIT cohort (8x106/kg vs. 8.8 x106/kg in the G+P cohort and 9.3x106/kg in the G+C cohort; P<0.01. A higher proportion of patients in the G+P cohort required two or fewer total apheresis sessions (86%) compared to the proportions in the G±JIT (70%) and G+C (74%) cohorts (P<0.01). The total number of G-CSF doses administered was higher in the G+C group (8-14 doses used in 68% of the G+C group, 18% of the G±JIT group and 9% of the G+P group; P<0.01). Overall, about 16% of patients failed to mobilize adequate HSPC for two transplants (based on the center’s definition), with no difference between the groups. About 23% of patients in the G+C group developed complications attributed to chemotherapy, including 4% who required hospitalization and 11% who had nausea/vomiting (Table 2). Among the G+C group, 12% received packed red cell transfusions, compared to 2% in the G±JIT group and 2% in the G+P group (P<0.01). Similarly, 10% of subjects received platelet transfusions in the G+C group, compared to 0.5% in the G±JIT group and 3% in the G+P group (P<0.01). On multivariate analysis, the rates of neutrophil and platelet engraftment were significantly higher in the G+P and G+C groups than in the G±JIT group but no difference was observed in non-relapse mortality, relapse, progressionfree survival or overall survival (Online Supplementary Table S2). Myeloma remained the primary cause of death in all groups.

For the health economics analysis, a total 222 patients in the CIBMTR-Medicare merged dataset met the inclusion criteria. In the pre-apheresis period, patients in the G+C

cohort had more outpatient visits and a higher number of outpatient prescriptions filled (Online Supplementary Table S3). The median costs for the entire mobilization strategy were $23,033 (interquartile range [IQR] $15,512) for the G+P or JIT group, $19,522 (IQR $10,132) for the G+C group, and $11,191 (IQR $9,695) for the G alone group (P<0.0001) (Figure 1A, Online Supplementary Table S3).

The median total costs were further examined by payment type and mobilization strategy (Figure 1B). Patients in the G+P or JIT group were responsible for 17% of the total

payments, compared to 20% for the other two mobilization groups (P<0.001). The estimated median cost for drugs charged to patients (cost-to-charge ratio) was approximately $18,000 for the G+P or JIT and G+C groups, and approximately $7,000 for the G alone group. In this largest prospective, observational study comparing three different mobilization regimens, we showed that the G+C and G+P approaches provided higher CD34+ cells/kg yield on day 1 as well as total CD34+ cell/kg yield, but there was no difference in mobilization failures or post-

Stayed at a center closer to mobilization, N (%)

G±JIT: granulocyte colony-stimulating factor (G-CSF) with or without “just in time” plerixafor; G+P: G-CSF plus plerixafor; G+C: G-CSF plus cyclophosphamide; IQR: interquartile range; AHCT: autologous hematopoietic cell transplantation. Pairwise

values:

vs

Table 1. Efficacy of mobilization strategies.

AHCT outcomes. G+C was associated with higher rates of hospitalization, gastrointestinal symptoms (nausea and vomiting) and transfusion requirements. From the health economic perspective, we showed that patients in the G+C group incurred the highest median total costs during the pre-apheresis period, while patients in the G+P or JIT group incurred the highest median total costs during the apheresis period.

G-CSF alone was found to be inferior to G+P and chemomobilization in several studies.5,6 Studies comparing the JIT plerixafor approach to chemomobilization have pro-

duced divergent results.7,8 The current observational study demonstrated superior efficacy with both G+C and routine G+P based approaches, compared to the G±JIT approach in terms of total number of cells collected, but there was no difference in mobilization failures or cells collected for two AHCT. Patients in the G+C and G±JIT groups required a higher number of apheresis sessions and a higher number of G-CSF doses to reach the collection target goal, consistent with previous studies.6 It important to acknowledge that CD34+ collection targets vary across centers, but most collect enough cells for more than two

Figure 1. Costs by mobilization strategy. (A) Median total costs by time period and mobilization strategy in the health economics set. Note that the median total costs varied by mobilization strategy in the pre-apheresis period (P<0.0001). Total costs include inpatient, outpatient, home healthcare, and outpatient pharmacy costs. (B) Median percent of total costs by payment type and mobilization strategy. Note that both Medicare payment and “patient responsibility” varied by mobilization strategy (both P<0.0001). Patient responsibility included co-payment, co-insurance, and deductibles. The number of patients with a secondary payer was less than 11 (not reported). Total costs include inpatient, outpatient, home healthcare, and outpatient pharmacy costs. G-CSF: granulocyte colony-stimulating factor.

G±JIT: granulocyte colony-stimulating factor (G-CSF) with or without “just in time” plerixafor; G+P: G-CSF plus plerixafor; G+C: G-CSF plus cyclophosphamide.

Table 2. Toxicity by mobilization strategies.

AHCT. To account for variation across centers for collection targets, we defined a good mobilizer as a patient from whom it was possible to collect ≥5.0 x106 CD34+ cells/kg in a maximum of two apheresis sessions. There were significantly higher numbers of good mobilizers in both the G+P (81%) and G+C (75%) cohorts than in the G±JIT (72%) cohort.

Chemomobilization has been shown to be associated with higher rates of hematologic and non-hematologic toxicities, consistent with our observations.9-11 Interestingly the rates of hospitalization were much lower in our study (3%) than those reported in prior studies.12

A major limitation of our analysis is that while the claims data captured the use of plerixafor, they lacked the granularity to differentiate between how the drug was incorporated into mobilization (G+P or JIT). Thus, in the cost analysis we were not able to compare the costs of G±JIT against G+P; hence, the cost analysis groups were different from the grouping for efficacy analysis. A prior analysis comparing the cost of G+P versus G±JIT showed that the average estimated cost with routine plerixafor use was significantly higher than that with JIT use.13 We showed that total costs were lowest for G-CSF alone and highest for G+P or JIT, and that drug cost, the key driver for all mobilization strategies, varied by period of mobilization. In the pre-apheresis period, G+C incurred the highest cost of the three mobilization strategies, likely associated with more outpatient visits and days of G-CSF use. During the apheresis period, G+P or JIT incurred the highest cost, driven likely by the cost of plerixafor. Prior cost analyses have shown varying results in single institution studies. For example, Afifi et al. showed that the high cost of plerixafor in the G±JIT group was offset by increased resource utilization in the G+C (cyclophosphamide 3 g/m2) group.6 However, despite more frequent episodes of febrile neutropenia, intravenous antibiotic use, and hospitalization with G-CSF plus cyclophosphamide (3-4 g/m2), Awan et al. showed a significantly lower average total cost of mobilization compared to that of a planned G+P approach.7 Costa et al . compared a G+JIT approach to cyclophosphamide (2 g/m2) plus G-CSF and granulocyte-macrophage colony-stimulating factor and showed that the estimated average cost per patient successfully completing mobilization was lower in the G+JIT cohort than in the chemotherapy cohort.14 Another limitation to the efficacy analysis was a selection bias on the type of patients selected given the lack of randomized comparisons.

In conclusion, this large prospective, observational study showed that overall success of three mobilization regimens is similar with no impact on AHCT outcomes. The resource utilization in chemomobilization was highest among the groups, with outpatient costs being the major contributor of the total costs.

Authors

Binod Dhakal,1 Mei-Jei Zhang,2,3 Linda J. Burns,3 Xiaoying Tang,3 Christa Meyer,4 Lih-Wen Mau,4 Ajay K. Nooka,5 Edward Stadtmauer,6 Ivana N. Micallef,7 Joseph McGuirk,8 Luciano Costa,9 Mark B. Juckett,10 Nina Shah,11 Richard E. Champlin,12 Saad Z. Usmani,13 Sherif S. Farag,14 Taiga Nishihori,15 Vivek Roy,16 Andrew Bodiford,17 Yvonne J. Barnes,18 Edward J. Drea,19 Parameswaran Hari3 and Mehdi Hamadani1,3

1BMT and Cellular Therapy Program, Department of Medicine, Medical College of Wisconsin, Milwaukee, WI; 2Division of Biostatistics, Institute for Health and Equity, Medical College of Wisconsin, Milwaukee, WI; 3CIBMTR® (Center for International Blood and Marrow Transplant Research), Department of Medicine, Medical College of Wisconsin, Milwaukee, WI; 4CIBMTR® (Center for International Blood and Marrow Transplant Research), National Marrow Donor Program/Be The Match, Minneapolis, MN;

5Department of Hematology and Medical Oncology, Winship Cancer Institute, Emory University School of Medicine, Atlanta, GA;

6University of Pennsylvania Abramson Cancer Center, Philadelphia, PA; 7Division of Hematology, Mayo Clinic, Rochester, MN; 8Division of Hematologic Malignancies and Cellular Therapeutics, The University of Kansas Cancer Center, Kansas City, KS; 9Division of Hematology/Oncology, Department of Medicine, University of Alabama at Birmingham, Birmingham, AL; 10Department of Medicine, Division of Hematology, Oncology and Transplantation, University of Minnesota, Minneapolis, MN; 11Division of Hematology-Oncology; University of California San Francisco, San Francisco, CA;

12Department of Stem Cell Transplantation and Cellular Therapy, Division of Cancer Medicine, The University of Texas MD Anderson Cancer Center, Houston, TX; 13Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, NY; 14Division of Hematology and Oncology, Indiana University School of Medicine, Indianapolis, IN; 15Department of Blood and Marrow Transplant and Cellular Immunotherapy (BMT CI), Moffitt Cancer Center, Tampa, FL; 16Blood and Marrow Transplant Program, Mayo Clinic, Jacksonville, FL; 17US Medical Affairs – Transplantation, Sanofi, Bridgewater, NJ; 18US Medical Affairs – Hematology Oncology, Sanofi Specialty Care, Cambridge, MA and 19US Medical – Oncology Medical Value & Outcomes, Sanofi Specialty Care, Cambridge, MA, USA

Correspondence:

M. HAMADANI - mhamadani@mcw.edu

https://doi.org/10.3324/haematol.2022.282269

Received: October 24, 2022.

Accepted: December 27, 2022.

Early view: January 5, 2023.

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Disclosures

BD has served on advisory boards for Janssen, Amgen, Takeda, Sanofi, GSK, Arcellx and Natera; he has received honoraria from GSK, BMS and Karyopharm and institutional funding from Janssen, BMS, Sanofi, GSK, Arcellx, Cartesian, Carsgen and Fate. MH has provided consultancy services to Incyte Corporation, ADC Therapeutics, Pharmacyclics, Omeros, Verastem, Genmab, Morphosys, Kite, Novartis, and Kadmon and has served on speaker’s bureau for Sanofi Genzyme, AstraZeneca, BeiGene, and ADC Therapeutics. Sanofi contracted with CIBMTR for services associated with fulfillment of this study. The CIBMTR aligns all activities through the lens of its research mission and utilizes funding sources only to expand research infrastructure and to facilitate a broad research portfolio. It contractually maintains independent review and publication rights and, as such, does not consider the services provided to be a conflict of interest. AB, YB, and ED are employees of Sanofi

Contributions

MH, PH, LJB, and BD conceived the study. MJZ, XT, CM, and LWM performed the statistical analysis. AN, ES, IM, JM, LC, MJ, NS, RC, SU, SF, TN, VR, AB, YB, and ED provided data and critically reviewed the manuscript. BD wrote the first draft, and all authors approved the final version.

Funding

The CIBMTR is supported primarily by Public Health Service U24CA076518 from the National Cancer Institute (NCI), the National Heart, Lung and Blood Institute (NHLBI) and the National Institute of Allergy and Infectious Diseases (NIAID); HHSH250201700006C from the Health Resources and Services Administration (HRSA); and N00014-20-1-2832 and N00014-21-1-2954 from the Office of Naval

References

1. Attal M, Lauwers-Cances V, Hulin C, et al. Lenalidomide, bortezomib, and dexamethasone with transplantation for myeloma. N Engl J Med. 2017;376(14):1311-1320.

2. Gay F, Musto P, Rota-Scalabrini D, et al. Carfilzomib with cyclophosphamide and dexamethasone or lenalidomide and dexamethasone plus autologous transplantation or carfilzomib plus lenalidomide and dexamethasone, followed by maintenance with carfilzomib plus lenalidomide or lenalidomide alone for patients with newly diagnosed multiple myeloma (FORTE): a randomised, open-label, phase 2 trial. Lancet Oncol. 2021;22(12):1705-1720

3. Wood WA, Whitley J, Moore D, et al. Chemomobilization with etoposide is highly effective in patients with multiple myeloma and overcomes the effects of age and prior therapy. Biol Blood Marrow Transplant. 2011;17(1):141-146.

4. Dhakal B, Veltri LW, Fenske TS, et al. Hematopoietic progenitor cell mobilization with ifosfamide, carboplatin, and etoposide chemotherapy versus plerixafor-based strategies in patients with Hodgkin and non-Hodgkin lymphoma. Biol Blood Marrow Transplant. 2016;22(10):1773-1780.

5. DiPersio JF, Stadtmauer EA, Nademanee A, et al. Plerixafor and

Research. Support is also provided by Be the Match Foundation, the Medical College of Wisconsin, the National Marrow Donor Program, and from the following commercial entities: AbbVie; Actinium Pharmaceuticals, Inc.; Adaptimmune; Adaptive Biotechnologies Corporation; ADC Therapeutics; Adienne SA; Allogene; Allovir, Inc.; Amgen, Inc.; Angiocrine; Anthem; Astellas Pharma US; AstraZeneca; Atara Biotherapeutics; BeiGene; bluebird bio, inc.; Bristol Myers Squibb Co.; CareDx Inc.; CRISPR; CSL Behring; CytoSen Therapeutics, Inc.; Eurofins Viracor, DBA Eurofins Transplant Diagnostics; Gamida-Cell, Ltd.; Gilead; GlaxoSmithKline; HistoGenetics; Incyte Corporation; Iovance; Janssen Research & Development, LLC; Janssen/Johnson & Johnson; Jasper Therapeutics; Jazz Pharmaceuticals, Inc.; Kadmon, a Sanofi Company; Karius; Kiadis Pharma; Kite, a Gilead Company; Kyowa Kirin; Legend Biotech; Magenta Therapeutics; Mallinckrodt Pharmaceuticals; Medac GmbH; Medexus Pharma; Merck & Co.; Mesoblast; Millennium, the Takeda Oncology Co.; Miltenyi Biotec, Inc.; MorphoSys; Novartis Pharmaceuticals Corporation; Omeros Corporation; OptumHealth; Orca Biosystems, Inc.; Ossium Health, Inc.; Pfizer, Inc.; Pharmacyclics, LLC, An AbbVie Company; Pluristem; PPD Development, LP; Sanofi; Sobi, Inc.; Stemcyte; Takeda Pharmaceuticals; Talaris Therapeutics; Terumo Blood and Cell Technologies; TG Therapeutics; Vertex Pharmaceuticals; Vor Biopharma Inc.; and Xenikos BV.

Data-sharing statement

CIBMTR supports accessibility of research in accord with the National Institutes of Health (NIH) data-sharing policy and the National Cancer Institute (NCI) Cancer Moonshot public access and data-sharing policy. The CIBMTR only releases de-identified datasets that comply with all relevant global regulations regarding privacy and confidentiality.

G-CSF versus placebo and G-CSF to mobilize hematopoietic stem cells for autologous stem cell transplantation in patients with multiple myeloma. Blood. 2009;113(23):5720-5726.

6. Afifi S, Adel NG, Devlin S, et al. Upfront plerixafor plus G-CSF versus cyclophosphamide plus G-CSF for stem cell mobilization in multiple myeloma: efficacy and cost analysis study. Bone Marrow Transplant. 2016;51(4):546-552.

7. Awan F, Kochuparambil ST, Falconer DE, et al. Comparable efficacy and lower cost of PBSC mobilization with intermediatedose cyclophosphamide and G-CSF compared with plerixafor and G-CSF in patients with multiple myeloma treated with novel therapies. Bone Marrow Transplant. 2013;48(10):1279-1284.

8. Costa LJ, Abbas J, Hogan KR, et al. Growth factor plus preemptive ('just-in-time') plerixafor successfully mobilizes hematopoietic stem cells in multiple myeloma patients despite prior lenalidomide exposure. Bone Marrow Transplant. 2012;47(11):1403-1408.

9. Antar A, Otrock ZK, Kharfan-Dabaja MA, et al. G-CSF plus preemptive plerixafor vs hyperfractionated CY plus G-CSF for autologous stem cell mobilization in multiple myeloma: effectiveness, safety and cost analysis. Bone Marrow

Transplant. 2015;50(6):813-817.

10. Gupta S, Zhou P, Hassoun H, et al. Hematopoietic stem cell mobilization with intravenous melphalan and G-CSF in patients with chemoresponsive multiple myeloma: report of a phase II trial. Bone Marrow Transplant. 2005;35(5):441-447.

11. Koc ON, Gerson SL, Cooper BW, et al. Randomized cross-over trial of progenitor-cell mobilization: high-dose cyclophosphamide plus granulocyte colony-stimulating factor (G-CSF) versus granulocyte-macrophage colony-stimulating factor plus G-CSF. J Clin Oncol. 2000;18(9):1824-1830.

12. Costa LJ, Alexander ET, Hogan KR, et al. Development and

validation of a decision-making algorithm to guide the use of plerixafor for autologous hematopoietic stem cell mobilization. Bone Marrow Transplant. 2011;46(1):64-69.

13. Veltri L, Cumpston A, Shillingburg A, et al. Hematopoietic progenitor cell mobilization with "just-in-time" plerixafor approach is a cost-effective alternative to routine plerixafor use. Cytotherapy. 2015;17(12):1785-1792.

14. Costa LJ, Miller AN, Alexander ET, et al. Growth factor and patient-adapted use of plerixafor is superior to CY and growth factor for autologous hematopoietic stem cells mobilization. Bone Marrow Transplant. 2011;46(4):523-528.

H syndrome mimicking Erdheim Chester disease: new entity and therapeutic perspectives

Biallelic pathogenic variants in the SLC29A3 gene are responsible for autosomal recessive H syndrome, variably combining hyperpigmentation, hypertrichosis, hearing loss, hepatosplenomegaly, heart anomalies, hypogonadism, low height, diabetes mellitus and camptodactyly.1 Tocilizumab (anti-IL-6 receptor) is considered to improve patient outcomes.2

However, other heterogenous SLC29A3-associated phenotypes have been reported with overlapping features, including hematologic disorders, auto-inflammatory manifestations and histiocytic infiltrates (HI) such as Rosai Dorfman disease (RDD) which have been classified in the R-group of the revised classification of histiocytoses.3,4 Indeed, skin histiocytic infiltrate and lymphadenopathy were frequently described (68% and 24% respectively in a 79-patient cohort), whereas retroperitoneal fibrosis and mediastinal mass were unusual.1

SLC29A3 encodes for equilibrative nucleoside transporter 3 (ENT3) which allows membrane translocation of nucleosides in late endosomes, lysosomes and mitochondria.5 A SLC29A3-/- mouse model displays systemic infiltrates in several organs notably associated with upregulation of the mechanistic target of rapamycin (mTOR) signaling.6 We herein describe four adult patients with pathogenic variants in SLC29A3 and atypical presentation of HI mimicking Erdheim Chester disease (ECD) rather than R-group histiocytosis. Tocilizumab appeared effective in the regulation of their inflammatory manifestations, but ineffective regarding HI. One of these ECD-like patients was treated with cobimetinib (MEK inhibitor), allowing a fast reduction of HI, probably through the regulation of the mTOR pathway.

Immunohistochemistry was performed on an automated immunostainer (Ventana BenchMark Ultra, Roche, Meylan, France) using Ultraview Universal Kit. Sections were incubated with anti-pS6 (Ser235/236) (1:50, Cell Signaling, 4858), anti-p4E-BP1 (Thr37/46) (1:500, Cell Signaling, 2855) and anti-CD163 (10D6) (1:100, Novocastra-Leica, NCL-L-CD163). Staining was visualized with DAB solution. Patients from French centers were identified and retrospectively included. Informed signed consent was obtained from each patient.

Next-generation sequencing (NGS) of a targeted panel of auto-inflammatory diseases-associated genes was performed in blood samples and revealed homozygous SLC29A3 (NM_018344.6) disease-causing variants (Table 1).

Case 1

A 37-year-old female with a medical history of bilateral

cochlear hearing loss presented anorexia, asthenia, and weight loss, as well as abdominal pain. Clinical examination found a cutaneous hyperpigmentation and hypertrichosis overlying the lower limbs (Figure 1A), hallux valgus, and camptodactyly. Blood test results showed a 100 mg/L C reactive protein (CRP), a characterized diabetes m ellitus (hemoglobin HbA1c 7,7%) and low IGF-1 level 85 μg/L (normal range, 109-350 μg/L). Computed tomography (CT) scan and cardiac magnetic resonance imaging (MRI) found a right atrial, interatrial septum, periaortic, pleural, and perirenal tissue infiltrates (Figure 1B). Pleural and skin biopsy were performed and found CD163 + /CD68 + /S100 - /CD1a- /OCT-2 + histiocyte infi ltrates. Investigations on biopsy samples found no mutation in the BRAF-MEK-ERK pathway, or in PIK3CA, and negative p-ERK immunostaining.

The patient was treated with low-dose corticosteroids and tocilizumab (8 mg/kg/4 weeks) resulting in inflammation normalization, skin infiltration regression, but no other HI improvement. Although MAPK upregulation was not found, considering ECD-like HI, the patient was treated with cobimetinib (28-day cycle therapy: 40 mg/day for 21 days, and no treatment for 7 days), allowing a fast clinical response (within 3 months) with a significant reduction of interatrial septum infiltrate (26 mm to 13 mm), and resolution of periaortic and perirenal infiltrates (Figure 1C-F), as well as skin histiocytic infiltrate decrease in biopsy (Figure 1G, H) with no relapse at 1-year follow-up.

Case 2

A 42-year-old female presented a subacute 5-month evaluative spastic paraparesis. She had a history of bilateral cochlear hearing loss, micromelia, dysphonia and camptodactyly (Figure 2A). She also presented a history of child-onset inflammatory arthritis considered as juvenile idiopathic arthritis with corticosteroid dependence. Clinical examination revealed lower limbs pyramidal signs and sensory-motor deficit. Spinal cord MRI revealed an epidural mass with contrast enhancement between C5 and L5 complicated by dorsal medullar compression (Figure 2B), CT scan was unremarkable. Blood tests results found fluctuating inflammatory syndrome (CRP >10 mg/L). Infectious triggers were ruled out, and immunological assays were unremarkable.

A surgical biopsy was performed and revealed inflammatory infiltrates composed of plasma cells (CD138+) and CD68+/S100-/CD1a- macrophages. NGS analysis was per-

Genetic analysis

SLC29A3 (NM_018344.6)

*Data not available. **Imaging assessed (CT-scan, MRI or PET-scan). *** Skin biopsy was the only available tissue for this patient, and it revealed fibrotic tissue without histiocytes. x: Presence; 0: Absence; F: female; M: male; NGS: next-generation sequencing; Hb: hemoglobin;

mean corpuscular volume; CRP+: C-reactive protein.

Table 1. Main characteristics of the enrolled patients in the study with mutation status.

formed in biopsy and found no somatic mutation in BRAF or PIK3CA.

The patient was treated with corticosteroids (1 mg/kg/day) allowing rapid improvement of the neurological symptoms and epidural mass regression. Progression was marked by several flare-ups under corticosteroids at 15 mg/day. Tocilizumab (8 mg/kg/4 weeks) for 4 months allowed inflammatory syndrome normalization and corticosteroid sparing without arthritis flare. Epidural infiltrate persisted

after treatment, but remained stable in spite of corticosteroid discontinuation.

Case 3

A 43-year-old man was refered to the Internal Medicine Department for suspected auto-inflammatory syndrome associated with arthritis. He was born in Morocco and had a history of bilateral cochlear hearing loss, camptodactyly complicating a child onset arthritis and papillary thyroid

Figure 1. Patient #1 imaging and immunostaining before and during treatment. (A) Patient #1 lower limbs cutaneous infiltrates. (B) Heart cine 4 cavity magnetic resonance imaging (MRI) revealing right atrial histiocytic infiltrate of patient #1 before treatment (arrows). (C) Patient #1 heart cine 4 cavity MRI with measure of inter-atrial infiltrate before cobimetinib treatment and (D) under cobimentinib treatment (1-year duration). (E) Patient #1 injected abdominal computed tomography (CT) scan before cobimetinib treatment (arrows), and (F) non-injected abdominal CT scan under cobimentinib treatment (1-year duration). (G) Patient #1 skin biopsy with CD163+ immunostaining (magnification x2.4) showing histiocytic infiltrates (arrows) before cobimetinib treatment and (H) under cobimentinib treatment (1-year duration). (I) Patient #1 skin biopsy with p-S6 immunostaining (magnification x1.4), and (K) p-4EBP1 immunostaining (magnification x1.4) with mTOR pathway upregulation (arrows). (J) Patient #1 skin biopsy under cobimentinib treatment (1-year duration) with p-S6 immunostaining (magnification x1.4), and (L) p-4EBP1 immunostaining with disappearance of mTOR pathway upregulation (magnification x1.4). (M) Patient #1 pleural biopsy before treatment with CD163 + immunostaining (magnification x1.2), (N) p-S6 immunostaining (magnification x1.2), and (O) p-4E-BP1 immunostaining (magnification x1.2). (P) Patient #1 skin biopsy before treatment with OCT-2 immunostaining (magnification x400).

carcinoma. Clinical examination revealed a gynecomastia, a hallux valgus, and depigmented back skin lesions.

CT scan found peri-aortic, perirenal, peritoneal and underskin infiltrates. Blood test results found a fluctuating inflammatory syndrome (CRP >20 mg/L), an elevated luteinizing hormone and prolactinemia (14.7 UI/L; 417 UI/L respectively) with unremarkable immunological assay. A skin biopsy was performed in depigmented back lesions, but unfortunately uniquely contained fibrotic lesions without histiocytic infiltrate.

The patient was treated with tocilizumab (8 mg/kg/2 weeks) for 2 years with clinical improvement regarding auto-inflammatory manifestations (arthritis and inflammatory syndrome disappearance) but no improvement regarding the tissue infiltrate at CT scan.

Case 4

A 48-year-old women presented with multiple organ infiltrates and inflammatory syndrome. She had a history of childhood-onset bilateral hearing loss, bilateral micromelia with cubitus valgus deformity and camptodactyly. During adulthood, she developed systemic disorders such as a cutaneous infiltrate involving the back and left breast; mediastinum (Figure 2C, D), retroperitoneal (Figure 2E) and interstitial lung infiltrates on CT scan and recurrent pericarditis with fluctuating inflammatory syndrome (CRP ≥100 mg/L). She also developed endocrinopathy: diabetes mellitus and multi-nodular goiter. Immunological assay was unremarkable and infectious diseases were ruled out. A skin biopsy was performed and revealed CD163+ /CD68 + /S100 + /CD1a- /OCT-2 + histiocyte infiltrates. NGS analysis was performed in biopsy and found no somatic mutation in BRAF She was treated with tocilizumab (8 mg/kg/4 weeks) for 3 years allowing a regression of the cutaneous infiltrate and normalization of the inflammatory syndrome, but mediastinum and peri-renal infiltrate persisted.

Clinical improvement of patient #1 with cobimetinib could not be explained by a normalization of the MAPK pathway which was not upregulated.

Importantly, activation of mTOR pathway has been reported to play a pathological role in SLC29A3-deficiency and in ECD.6,7 Moreover, cobimetinib has been shown to impact the mTOR pathway either by inhibiting MAPK which is interconnected with mTOR pathway or directly by inhibiting AKT.8 Consequently, mTOR pathway downstream effectors (p-4E-BP and p-S6) were analyzed by immunohistochemistry on patient #4 before treatment on skin biopsy and on patient #1 before treatment on skin and pleural samples and during cobimetinib treatment (1-year therapy) on skin sample. We observed important p-4E-BP and p-S6 immunostainings before treatment, mainly colocalized with CD163+ cells in patient #1 pleural and skin biopsy samples (Figura 1G-O) and in patient #4 skin biopsy

sample (Figure 2F, G) Conversely, under therapy, histiocytic infiltrates, p4E-BP and p-S6 were dramatically decreased (Figure 1J, L). These findings strongly suggest that upregulation of the mTOR pathway contribute to HI pathogenesis, which can be normalized by cobimetinib treatment. We describe herein a case series of adult patients showing ECD-like HI phenotypes in the context of SLC29A3 disease-causing variants. Until now, heterogeneous H syndrome clinical presentations were described across SLC29A3-mutated cases, including HI which are R-group classified.1 Patients in this study presented classical clinical features of H syndrome, with childhood or adolescent onset driven by previously described SLC29A3 mutations.3 In addition, three patients presented peri-renal and pericardiac/-aortic infiltrates, as commonly observed in ECD, whereas the absence of bone lesions and presence of skin

abdominal CT-scan revealing retroperitoneal infiltrate. (F, G) Patient #4 skin biopsy before treatment with p-S6 and p-4EBP immunostaining (magnification x8.9). (H) Patient #4 skin biopsy before treatment with OCT-2 immunostaining (magnification x200).

Figure 2. Clinical, imaging and pathological features of patients #2 and #4. (A) Patient #2 camptodactyly. (B) Patient #2 epidural infiltrate on T2 magnetic resonance imaging sequence (arrows). (C, D) Patient #4 thoracic computed tomography (CT) scan revealing mediastinum infiltrate. (E) Patient #4 injected

infiltrates with hyperpigmentation are unusual in ECD. A recent study of 101 ECD focusing on clinical features and somatic mutations driving ECD in classical ECD (typical osseous involvement) and non-classical ECD (absence of typical osseous involvement) showed that 14% of patients had non-classical ECD.9 Among these patients, 35% did not have alterations in MAPK-ERK pathway, more than 30% presented hairy kidney, and 14% presented cardiac or cutaneous involvement.9 Considering these genotypic and phenotypic common points with our cases, we can hypothesize that a part of non-classical ECD could be driven by SLC29A3 mutations or alterations in mTOR pathway. One patient presented an epidural mass inducing spinal cord compression, which is a rare complication of ECD.10 Histopathological characteristics were ECD-like features, for patients #1 and #2 with CD68+/CD163+/S100-/CD1a- infiltrates, whereas for patient #4 S100 was positive. No emperipolesis was observed, contrasting with observations classically reported in R-group histiocytosis and some H syndrome histiocytic infiltrates. Interestingly, OCT-2 immunostaining was positive for patients #1 (Figure 1P) and #4 (Figure 2H). While OCT-2 staining should be negative on ECD samples, it is not specific of RDD, and this has been shown in some cases of Langerhans cell histiocytosis (510%) and of ALK+ histiocytosis (60%).11 Thus, OCT-2 immunostaining could be a pathological feature of interest in H syndrome histiocytosis and helpful to guide diagnosis towards H syndrome in case of histiocytic infiltrate with evocative clinical signs.

A SLC29A3-/- mouse model showed that this mutation affects macrophage and T-lymphocyte function via AMPKmTOR-ULK signaling, leading to systemic abnormalities such as hypertrichosis, skeletal deformities, endocrinopathy and histiocytic infiltrates.6 Indeed, because of lysosome trafficking disturbance, ENT3-deficient macrophages displayed impaired apoptotic cell clearance and proliferated via the MCSF/MCSFR (macrophage colony-stimulating factor) axis.12

Genetic analysis of ECD patients revealed that the BRAF V600E mutation is present in about half of the patients, while about 11% of patients displayed PIK3CA mutations.7 Thus, the upregulation of RAS downstream pathway RAF/MEK/ERK as well as of the PI3K/AKT/mTOR pathway could induce ECD. Moreover, a 10-patient cohort clinical trial associating sirolimus and prednisone found clinical improvement (3 were BRAF V600E-mutated and 7 without documented mutation), as well as a 20-patient cohort long-term study (imaging or metabolic response in 65% of patients) underlying the role of PI3K/AKT/mTOR pathway in some ECD cases.13,14 Considering these elements, we treated one patient with cobimentinib as a second-line therapy, which allowed rapid regression of HI despite the fact that the MAPK pathway was not hyperactivated. Immunostaining revealed that HI is associated with mTOR

pathway upregulation which was normalized by cobimetinib. As the MAPK pathway is known to be involved in inflammatory signaling, MEK inhibitor could play here a double role with control of anti-inflammatory features via MAPK signaling and histiocytic infiltrate via mTOR signaling.15

Our report shows that an inherited monogenic condition can be associated to at least two types of histiocytoses and that a phenotype highly reminiscent of ECD can be associated with monogenic disease. Thus, it underlines the interest in searching for SLC29A3 mutations in the context of atypical ECD-like histiocytosis. Morevoer, it expands the spectrum of overlapping histiocytoses, raising the question of reconsidering H syndrome histiocytosis as R-group members. Noteworthy, cobimetinib appears as a promising second-line therapy after tocilizumab. Finally, mTOR inhibitors could probably be tested to improve ECD-like histiocytosis in SLC29A3-mutated patients.

Authors

Hippolyte Lequain,1 Mathieu Gerfaud-Valentin,1 Jean-François Emile,2 Yann-Gaël Gangloff,3 Guilaine Boursier,4 Christophe Deligny,5 Guillaume Le Guenno,6 Juliet Tantot,7 Julie Valantin,8 Lea Savey,9 Claude Bachmeyer,9 Yvan Jamilloux,1 Laurent Schaeffer,3 Pascal Leblanc3 and Pascal Sève10

1Department of Internal Medicine, Hôpital de la Croix-Rousse, Hospices Civils de Lyon, Université Claude Bernard-Lyon1, Lyon; 2Université Paris-Saclay, Université de Versailles SQY, EA4340BECCOH, Assistance Publique–Hôpitaux de Paris (AP-HP), Ambroise-Paré Hospital, Smart Imaging, Service de Pathologie, Boulogne; 3Institut NeuroMyoGene (INMG-PGNM), CNRS UMR 5261, INSERM U1315, Université de Lyon, Lyon; 4Department of Molecular Genetics and Cytogenomics, Rare and Auto-inflammatory Diseases, CHU Montpellier, Université Montpellier, CeRéMAIA, Montpellier; 5Department of Internal Medicine, CHU de Fort de France, Fort de France; 6Department of Internal Medicine, CHU Estaing, Clermont-Ferrand; 7Department of Pathology, Institut de Pathologie Multisite, Groupement Hospitalier Sud, Hospices Civils de Lyon, Pierre-Bénite; 8Research Pathology Platform, INSERM 1052, CNRS 5286, Centre Léon Bérard, Centre De Recherche En Cancérologie De Lyon, Lyon; 9AP-HP, Hôpital Tenon, Sorbonne Université, Service de Médecine Interne, Centre de Référence des Maladies Auto-inflammatoires et des Amyloses d'Origine Inflammatoire (CEREMAIA), Paris and 10Pôle IMER, HESPER EA 7425, Lyon, France

Correspondence:

P. SEVE - pascal.seve@chu-lyon.fr

https://doi.org/10.3324/haematol.2022.282040

Received: September 4, 2022.

Accepted: January 17, 2023.

Early view: February 2, 2023.

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Disclosures

No conflicts of interest to disclose.

Contributions

HL, MG-V, CD, GLG, YJ, LS, CB and PS followed the patients, analyzed the data and draft the manuscript. J-FE analyzed biopsies and performed the molecular analysis of gene mutation in patient

References

1. Molho-Pessach V, Ramot Y, Camille F, et al. H syndrome: The first 79 patients. J Am Acad Dermatol. 2014;70(1):80-88.

2. Bloom JL, Lin C, Imundo L, et al. H syndrome: 5 new cases from the United States with novel features and responses to therapy. Pediatr Rheumatol. 2017;15(1):76.

3. Morgan NV, Morris MR, Cangul H, et al. Mutations in SLC29A3, encoding an equilibrative nucleoside transporter ENT3, cause a familial histiocytosis syndrome (Faisalabad histiocytosis) and familial Rosai-Dorfman disease. PLoS Genet 2010;6(2):e1000833.

4. Emile J-F, Abla O, Fraitag S, et al. Revised classification of histiocytoses and neoplasms of the macrophage-dendritic cell lineages. Blood. 2016;127(22):2672-2681.

5. Young JD, Yao SYM, Baldwin JM, Cass CE, Baldwin SA. The human concentrative and equilibrative nucleoside transporter families, SLC28 and SLC29. Mol Aspects Med. 2013;34(2-3):529-547.

6. Nair S, Strohecker AM, Persaud AK, et al. Adult stem cell deficits drive Slc29a3 disorders in mice. Nat Commun. 2019;10(1):2943.

7. Emile J-F, Diamond EL, Hélias-Rodzewicz Z, et al. Recurrent RAS and PIK3CA mutations in Erdheim-Chester disease. Blood. 2014;124(19):3016-3019.

8. Muranushi H, Shindo T, Chen-Yoshikawa TF, et al. Dual inhibition of the MEK/ERK and PI3K/AKT pathways prevents pulmonary

#1 biopsy. GB analyzed SLC29A3 mutations of patients and wrote the genetics paragraph. JT analyzed and collected biopsies. JV and Y-GG performed and analyzed pS6 and p-4E-BP immunostaining. All authors were involved in the critical analysis and final version of the manuscript.

Acknowledgments

The authors gratefully acknowledge patients for their agreement to participate in this study.

Data-sharing statement

The data that support the findings of this study are available from the corresponding author upon reasonable request

GVHD suppressing perivenulitis and bronchiolitis. Blood Adv. 2023;7(1):106-121.

9. Hazim AZ, Acosta Medina AA, Young JR, et al. Classical and non-classical phenotypes of ERDHEIM–CHESTER disease: correlating clinical, radiographic and genotypic findings. Br J Haematol. 2022;199(3):454-457.

10. Hwang BY, Liu A, Kern J, Goodwin CR, Wolinsky JP, Desai A. Epidural spinal involvement of Erdheim–Chester disease causing myelopathy. J Clin Neurosci. 2015;22(9):1532-1536.

11. Goyal G, Tazi A, Go RS, et al. International expert consensus recommendations for the diagnosis and treatment of Langerhans cell histiocytosis in adults. Blood. 2022;139(17):2601-2621.

12. Hsu C-L, Lin W, Seshasayee D, et al. Equilibrative nucleoside transporter 3 deficiency perturbs lysosome function and macrophage homeostasis. Science. 2012;335(6064):89-92.

13. Pegoraro F, Maniscalco V, Peyronel F, et al. Long-term follow-up of mTOR inhibition for Erdheim-Chester disease. Blood. 2020;135(22):1994-1997.

14. Gianfreda D, Nicastro M, Galetti M, et al. Sirolimus plus prednisone for Erdheim-Chester disease: an open-label trial. Blood. 2015;126(10):1163-1171.

15. Liu S, Ma H, Zhang H, Deng C, Xin P. Recent advances on signaling pathways and their inhibitors in rheumatoid arthritis. Clin Immunol. 2021;230:108793.

Venetoclax for treating refractory autoimmune hemolytic anemia in chronic lymphocytic leukemia: report of two cases in Spain

Chronic lymphocytic leukemia (CLL) is often associated with autoimmune complications. Indeed, autoimmune cytopenias (AIC) such as autoimmune hemolytic anemia (AIHA), immune thrombocytopenia (ITP), pure red cell aplasia and autoimmune granulocytopenia may be present in 5% to 9% of CLL patients.1,2 Among the different AIC, AIHA is the most prevalent in CLL patients.1 AIC refractory to steroids supose a criteria of treatment in patients with CLL according to iwCLL18.3

In the era of patient-tailored approach therapies, some targeted drugs such as ibrutinib, acalabrutinib, idelalisib, and venetoclax have been used in CLL treatment. There is evidence reporting positive outcomes in patients with preexisting AIC, especially with ibrutinib, an irreversibly inhibitor of the Bruton tyrosine kinase.2,4 Regarding B-cell lymphoma 2 (BCL-2) inhibitors, there is little evidence of about their use in CLL-associated autoimmune cytopenias.2 To our knowledge, no clinical trials have directly studied the role of these novel signal inhibitors in the management of AIC in patients with CLL.1

The current paper aims to present two patients with CLL who developed AIHA and underwent treatment with the BCL-2 inhibitor venetoclax.

Case 1

A 63-year-old woman diagnosed with CLL in May 2014, without alterations in fluorescence in situ hybridization (FISH), normal karyotype, wild-type TP53 and unknown immunoglobulin heavy chain (IGHV). In March 2016, the patient developed non-autoimmune anemia (hemoglobin: 8.5 g/dL) and splenomegaly (18.7 cm) that was treated with a total of six cycles of fludarabine, cyclophosphamide, and rituximab (FCR), achieving complete remission for 4 years. In September 2021, she presented a relapse with AIHA (hemoglobin 6 g/dL, lactate dehydrogenase [LDH] 666 UI/L, indirect bilirubin 3.04 mg/dL) and IgG C3d in direct Coombs test, associated with generalized lymphadenopathy, and hepatomegaly (17 cm) and splenomegaly (20 cm). After not responding to the first-line treatment with corticosteroids (1 mg/kg/day of prednisone), the patient began treatment with venetoclax plus rituximab according to the MURANO trial scheme.5 To date, the administration of venetoclax has been maintained at a dose of 400 mg per day. At the last follow-up visit (December 10, 2022), the patient has shown complete remission according to iwCLL 2018,3 over the 16 months of follow-up. There has been a clinically significant improve-

ment, with normalization of red blood count (13.2 g/dL of hemoglobin) (Figure 1), biochemical parameters (LDH 232 UI/L, indirect bilirubin 0.8 mg/dL), and resolution of lymphadenopathy and hepatosplenomegaly; without relevant adverse effects. Table 1 summarizes the main clinical outcomes.

Case 2

A 70-year-old man diagnosed with CLL in May 2016, with chromosome 13q deletion. He presented an episode of AIHA, serum hemoglobin of 8.5 g/dL (IgG in direct Coombs test), refractory to corticosteroids (1 mg/kg/day of prednisone) that responded to rituximab (375 mg/m2 per week for 4 weeks), reaching levels of 15 g/dL. In June 2018, there was a worsening of the disease, with a general deterioration of his performance status and relapsed AIHA, that was treated with the combination of obinituzumab-chlorambucil.7 After the first dose of obinutuzumab, the patient suffered worsening of AIHA, with hemoglobin of 5.1 g/dL, LDH 331 UI/L and indirect bilirubin of 2.89 mg/dL, and non–ST-elevation acute myocardial infarction, which led to obinutuzumab withdrawal. The patient started treatment with corticosteroids plus rituximab (4 cycles) that led to a complete remission according to iwCLL 2018.3

In April 2020, the patient presented a relapsed of AIHA, with hemoglobin 6.9 g/dL, LDH 318 UI/L, indirect bilirubin 1.5 mg/dL and IgM C3d in direct Coombs test, refractory to corticoids (prednisone 1 mg/kg/day). Unmutated IGHV and wildtype TP53 were determined, so venetoclax plus rituximab was started following the same scheme aforementioned.5 The patient has remained asymptomatic for 32 months with the combination of venetoclax plus rituximab. At the last follow-up visit (December 14, 2022), the patient showed complete remission of the disease, with normalization of red blood count (14.6 g/dL of hemoglobin) (Figure 2) and biochemistry parameters (LDH 196 UI/L and indirect bilirubin 0.55 mg/dL), without adverse effects (Table 1).

Discussion

Sometimes the diagnosis of AIHA in CLL patients can be difficult due to hemolysis findings in blood counts or biochemical tests may be distorted in CLL due to disease progression or therapy.1

The development of targeted therapies for treating AIHA in CLL patients is continuously evolving.8 To date, there is not enough evidence to recommend one or another in CLL-as-

Yrs: years; FISH: fluorescence in situ hybridization; N: number; mth: months; M: male; F: female; AIHA: autoimmune hemolytic anemia; CLL: chronic lymphocytic leukemia; IGHV: mutational status of immunoglobulin heavy-chain variable; neg: negative; NA: not applicable; WT: wildtype; un: unmutated; CR: complete remission*(*according to the iwCLL2018 criteria3).

sociated AIC.2 Although there is evidence suggesting rapid and durable responses when ibrutinib was used to treat autoimmune cytopenias associated with CLL, clinical experience with idelalisib or venetoclax is limited.2,9

The current paper describes two patients diagnosed with CLL that present a refractory AIHA crisis, who were successfully managed with venetoclax in combination with rituximab in second-line therapy. Apart from the Lacerda et al.6 paper, these are the only clinical cases demonstrating the efficacy and safety of venetoclax in refractory AIHA in patients with CLL. In addition, venetoclax has been shown to be effective in treating multiple-refractory idiopathic thrombocytopenic purpura

and Evans syndrome in two patients with CLL10 and in other two patients with not-specified pre-existing AIC.2

Some cases of treatment emergent AIC were reported after treatment with these new targeted drugs,2,8,11,12 with two cases of treatment-emergent AIC in CLL patients during treatment with venetoclax in monotherapy.11,12 However, it has been recently pointed out that treatment-emergent AIC during administration of targeted drugs is manageable in most patients without interruption of treatment.2

Venetoclax is a molecule, capable of binding and antagonizing BCL-2 family anti-apoptotic proteins by mimicking the BH3 domain of pro-apoptotic proteins.13 Venetoclax is strongly cytotoxic to CLL lymphocytes due to the high ex-

Figure 1. Evolution of hemoglobin levels and associated treatments of a 63-year-old woman diagnosed with chronic lymphocytic leukemia throughout the patient follow-up. *Autoimmune hemolytic anemia presentation. Prednisone 1 mg/kg/day + venetoclax ramp up (20 mg). jul: July; sep: September; oct: October; nov: November; jan: January; mar: March; apr: April; dec: December.

Table 1. Overview of the main clinical characteristics and outcomes of the two clinical cases and their comparison with Lacerda et al.6

Figure 2. Evolution of hemoglobin levels and associated treatments of a 70-year-old man diagnosed with chronic lymphocytic leukemia throughout the patient follow-up. *Autoimmune hemolytic anemia presentation. Prednisone 1 mg/kg/day. †End venetoclax scheme. Jan: January; apr: April; may: May; jun: June; sep: September; oct: October; feb: February; jul: July, dec: December.

pression levels of BCL-2 family proteins.14 That could be the reason why venetoclax can easily control a first-time AIHA crisis. The microenvironment plays a big part in CLL pathogenesis and is AIHA related. CLL cells process and present red blood antigen to T cells that will induce a formation of polyclonal antibodies by normal B cells against erythrocytes.15 Furthermore, CLL cells loose tolerance to cytokine inhibition by the innate system.15 For that reason, the leukemic microenvironment induces some kinds of resistance mechanisms to venetoclax, which are either CD40 ligand resistance or BCL-XL expression.14 These mechanisms can be overcome by anti-CD20 antibodies without affecting the expression of BCL-2 family proteins.14 It is only a hypothesis, the pathophysiology mechanism could be the reason why these two cases are responding to venetoclax in combination with rituximab. Moreover, this combination showed a good safety profile. As observed in the MURANO trial, only three patients in the venetoclax-rituximab arm required treatment withdrawal.2,7

Despite the good results obtained in our patients, further prospective studies are needed to identify the best profile and the best targeted drug or combination for CLL patients with AIHA.

Authors

Pablo Galindo-Navarro,+ Alicia Delgado-García, Miguel A. Rodríguez-Gil and José M Puerta-Puerta+

Unidad de Gestión Clínica de Hematología y Hemoterapia, Hospital Universitario Virgen de las Nieves, Granada, Spain

+PG-N and JMP-P contributed equally as first authors.

Correspondence:

P. GALINDO-NAVARRO - pablo.galindo.sspa@juntadeandalucia.es

https://doi.org/10.3324/haematol.2022.281850

Received: September 24, 2022.

Accepted: February 17, 2023. Early view: March 2, 2023.

©2023 Ferrata Storti Foundation

Published under a CC BY-NC license

Disclosures

The medical writer for this manuscript was supported by AbbVie with no input into the preparation, review, approval and writing of the manuscript. The authors maintained complete control over the manuscript content, and it reflects their opinions. No other conflicts of interest to disclose.

Contributions

PG-N developed the concept, visualized the research, analysed data, supervised the project and wrote the original draft. AD-G developed the concept and methodology, and visualized the research. MÁR-G analyzed data, was responsible for project administration, developed the methodology, and reviewed and

edited the manuscript. JMP-P acquired funding, was responsible for project administration, supervised the research, and reviewed and edited the manuscript.

Acknowledgments

Medical writing and editorial assistant services have been provided by Antonio Martínez of Ciencia y Deporte S.L.

References

1. Tsang M, Parikh SA. A concise review of autoimmune cytopenias in chronic lymphocytic leukemia. Curr Hematol Malig Rep. 2017;12(1):29-38.

2. Vitale C, Salvetti C, Griggio V, et al. Preexisting and treatmentemergent autoimmune cytopenias in patients with CLL treated with targeted drugs. Blood. 2021;137(25):3507-3517.

3. Hallek M, Cheson BD, Catovsky D, et al. iwCLL guidelines for diagnosis, indications for treatment, response assessment, and supportive management of CLL. Blood. 2018;131(25):2745-2760.

4. Vitale C, Montalbano MC, Salvetti C, et al. Autoimmune complications in chronic lymphocytic leukemia in the era of targeted drugs. Cancers (Basel). 2020;12(2):282.

5. Seymour JF, Kipps TJ, Eichhorst B, et al. Venetoclax-rituximab in relapsed or refractory chronic lymphocytic leukemia. N Engl J Med. 2018;378(12):1107-1120.

6. Lacerda MP, Guedes NR, Yamakawa PE, et al. Treatment of refractory autoimmune hemolytic anemia with venetoclax in relapsed chronic lymphocytic leukemia with del(17p). Ann Hematol. 2017;96(9):1577-1578.

7. Goede V, Fischer K, Busch R, et al. Obinutuzumab plus chlorambucil in patients with CLL and coexisting conditions. N Engl J Med. 2014;370(12):1101-1110.

8. Albiol N, Moreno C. Autoimmune cytopenia in CLL: prognosis

Funding

PG-N has received a grant from AbbVie during the conduct of the study.

Data-sharing statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

and management in the era of targeted therapies. Cancer J. 2021;27(4):286-296.

9. Vitale C, Ahn IE, Sivina M, et al. Autoimmune cytopenias in patients with chronic lymphocytic leukemia treated with ibrutinib. Haematologica. 2016;101(6):e254-258.

10. Gordon MJ, Maldonado E, Danilov AV. Refractory autoimmune cytopenias treated with venetoclax. Hemasphere. 2019;3(3):e202.

11. Carriles C, Ordóñez-Fernández L, Arias-Martínez A, MenárguezBlanc R, Rosado-María MC. Autoimmune hemolytic anemia, adverse event to venetoclax. Farm Hosp. 2019;43(5):166-167.

12. Abdel-Samad N, Sughayar R. Can treatment with venetoclax for chronic lymphocytic leukemia (CLL) result in autoimmune hemolytic anemia? Am J Case Rep. 2021;22:e928514.

13. Bose P, Gandhi V, Konopleva M. Pathways and mechanisms of venetoclax resistance. Leuk Lymphoma. 2017; 58(9):1-17.

14. Thijssen R, Slinger E, Weller K, et al. Resistance to ABT-199 induced by microenvironmental signals in chronic lymphocytic leukemia can be counteracted by CD20 antibodies or kinase inhibitors. Haematologica. 2015;100(8):e302-306.

15. Hodgson K, Ferrer G, Montserrat E, Moreno C. Chronic lymphocytic leukemia and autoimmunity: a systematic review. Haematologica. 2011;96(5):752-761.

haematologica

Journal of the Ferrata Storti Foundation