2022 02 Versione Digitale

haematologica
Journal of the Ferrata Storti Foundation
Editor-in-Chief
Jacob M. Rowe (Jerusalem)
Deputy Editors
Carlo Balduini (Pavia), Jerry Radich (Seattle)
Associate Editors
Hélène Cavé (Paris), Monika Engelhardt (Freiburg), Steve Lane (Brisbane), Pier Mannuccio Mannucci (Milan), Pavan
Reddy (Ann Arbor), David C. Rees (London), Francesco Rodeghiero (Vicenza), Gilles Salles (New York), Kerry Savage
(Vancouver), Aaron Schimmer (Toronto), Richard F. Schlenk (Heidelberg), Sonali Smith (Chicago)
Statistical Consultant
Catherine Klersy (Pavia)
Editorial Board
Walter Ageno (Varese), Sarit Assouline (Montreal), Andrea Bacigalupo (Roma), Taman Bakchoul (Tübingen), Pablo
Bartolucci (Créteil), Katherine Borden (Montreal), Marco Cattaneo (Milan), Corey Cutler (Boston), Kate Cwynarski
(London), Mary Eapen (Milwaukee), Francesca Gay (Torino), Ajay Gopal (Seattle), Alex Herrera (Duarte), Shai Izraeli
(Ramat Gan), Martin Kaiser (London), Marina Konopleva (Houston), Johanna A. Kremer Hovinga (Bern), Nicolaus
Kröger (Hamburg), Austin Kulasekararaj (London), Shaji Kumar (Rochester), Ann LaCasce (Boston), Anthony R. Mato
(New York), Neha Mehta-Shah (St. Louis), Alison Moskowitz (New York), Yishai Ofran (Haifa), Farhad Ravandi
(Houston), John W. Semple (Lund), Liran Shlush (Toronto), Sara Tasian (Philadelphia), Pieter van Vlieberghe (Ghent),
Ofir Wolach (Haifa), Loic Ysebaert (Toulouse)
Managing Director
Antonio Majocchi (Pavia)
Editorial Office
Lorella Ripari (Office & Peer Review Manager), Simona Giri (Production & Marketing Manager), Paola Cariati
(Graphic Designer), Giulia Carlini (Graphic Designer), Igor Poletti (Graphic Designer), Marta Fossati (Peer Review),
Diana Serena Ravera (Peer Review) , Laura Sterza (Account Administrator)
Assistant Editors
Britta Dorst (English Editor), Rachel Stenner (English Editor), Bertie Vitry (English Editor), Massimo Senna
(Information technology), Idoya Lahortiga (Graphic artist)
haematologica
Brief information on Haematologica

Haematologica (print edition, pISSN 0390-6078, eISSN 1592-8721) publishes peer-reviewed papers on all areas of experi-
mental 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 guide-
lines (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/#con-
flicts), “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 conse-
quences of any inaccurate or misleading data, opinion or statement. Whilst all due care is taken to ensure that drug doses
and other quantities are presented accurately, readers are advised that new methods and techniques involving drug usage,
and described within this journal, should only be followed in conjunction with the drug manufacturer’s own published
literature.
Direttore responsabile: Prof. Carlo Balduini; Autorizzazione del Tribunale di Pavia n. 63 del 5 marzo 1955.
Printing: Press Up, zona Via Cassia Km 36, 300 Zona Ind.le Settevene - 01036 Nepi (VT)
Associated with USPI, Unione Stampa Periodica Italiana.

Premiato per l’alto valore culturale dal Ministero dei Beni Culturali ed Ambientali
haematologica
Table of Contents
Volume 107, Issue 2: February 2022
About the Cover
351 Images from the Haematologica Atlas of Hematologic Cytology: dyserythropoiesis
Rosangela Invernizzi
https://doi.org/10.3324/haematol.2021.280385
Landmark Papers in Hematology

352 The platelet aggregometer
Carlo L. Balduini
Editorials
353 Too much and not enough: revisiting maintenance rituximab in indolent lymphomas
Sonali M. Smith
354 An exciting RXRA mutant revives interest in retinoids for acute myeloid leukemia
Fang Qiu and Hugues De The
356 Are clinical pharmacology studies still needed in childhood acute lymphoblastic leukemia?
Valentino Conter and Francesco Ceppi
Review Article
358 Integrating genetic and epigenetic factors in chronic myeloid leukemia risk assessment: toward gene expression-based biomarkers
Vaidehi Krishnan
Articles
Acute Lymphoblastic Leukemia
371 Comprehensive analysis of dose intensity of acute lymphoblastic leukemia chemotherapy
Seth E. Karol et al.
Hematopoiesis
381 Reversible switching of leukemic cells to a drug-resistant, stem-like subset via IL-4-mediated cross-talk with mesenchymal stroma
Hae-Ri Lee et al.
393 Aging of human hematopoietic stem cells is linked to changes in Cdc42 activity
Amanda Amoah et al.
Acute Myeloid Leukemia

403 APR-246 induces early cell death by ferroptosis in acute myeloid leukemia
Rudy Birsen et al.
Haematologica 2022; vol. 107 no. 2 - February 2022

http://www.haematologica.org/
haematologica
417 RXRA DT448/9PP generates a dominant active variant capable of inducing maturation in acute myeloid leukemia cells
Orsola di Martino et al.
Cell Therapy & Immunotherapy

427 Human invariant natural killer T cells promote tolerance by preferential apoptosis induction of conventional dendritic cells
Hannes Schmid et al.
437 CD38 knockout natural killer cells expressing an affinity optimized CD38 chimeric antigen receptor successfully target acute myeloid leukemia
with reduced effector cell fratricide
Mark Gurney et al.
446 Successful gene therapy of Diamond-Blackfan anemia in a mouse model and human CD34+ cord blood hematopoietic stem cells using a
clinically applicable lentiviral vector
Yang Liu et al.
Complications in Hematology
457 Long term follow-up of pediatric-onset Evans syndrome: broad immunopathological manifestations and high treatment burden
Thomas Pincez et al.
Iron Metabolism & its Disorders

467 Risk factors for endocrine complications in transfusion-dependent thalassemia patients on chelation therapy with deferasirox: a risk
assessment study from a multi-center nation-wide cohort
Maddalena Casale et al.
478 UBA6 and NDFIP1 regulate the degradation of ferroportin

Lisa Traeger et al.
Non-Hodgkin Lymphoma
489 Early detection of T-cell lymphoma with T follicular helper phenotype by RHOA mutation analysis
Rachel Dobson et al.
500 Efficacy and safety assessment of prolonged maintenance with subcutaneous rituximab in patients with relapsed or refractory indolent
non-Hodgkin lymphoma: results of the phase III MabCute study
Simon Rule et al.
Myeloid Biology
510 Unique ethnic features of DDX41 mutations in patients with idiopathic cytopenia of undetermined significance, myelodysplastic syndrome, or
acute myeloid leukemia
Eun-Ji Choi et al.
Platelet Biology & its Disorders

519 Sequence-specific 2'-O-methoxyethyl antisense oligonucleotides activate human platelets through glycoprotein VI, triggering formation of
platelet-leukocyte aggregates
Martina H. Lundberg Slingsby et al.

haematologica
Red Cell Biology & its Disorders

532 Sulfated non-anticoagulant heparin derivative modifies intracellular hemoglobin, inhibits cell sickling in vitro, and prolongs survival of
sickle cell mice under hypoxia
Osheiza Abdulmalik et al.
Letters to the Editor

541 SARS-CoV-2 infection in aplastic anemia
Daniele Avenoso et al.
544 The insecticides permethrin and chlorpyrifos show limited genotoxicity and no leukemogenic potential in human and murine hematopoietic
stem progenitor cells
Virginia C. Rodriguez-Cortez et al.
550 Somatic STAT3 mutations in CD8+ T cells of healthy blood donors carrying human T-cell leukemia virus type 2
Daehong Kim et al.
555 Incidence and outcome of SARS-CoV-2 infection in patients with monoclonal gammopathy of undetermined significance: a case-control study
Nicola Sgherza et al.
558 Interferon α-induced SAMHD1 regulates human cultured megakaryocyte apoptosis and proplatelet formation
Seema Bhatlekar et al.
562 Detection of ABL1 kinase domain mutations in therapy-naïve BCR-ABL1-positive acute lymphoblastic leukemia
Constance Baer et al.

haematologica
The origin of a name that reflects Europe’s cultural roots.
Ancient Greek
Scientific Latin haematologicus (adjective) = related to blood
Scientific Latin haematologica (adjective, plural and neuter,

used as a noun) = hematological subjects
Modern English The oldest hematology journal,

publishing the newest research results.
2020 JCR impact factor = 9.94
ABOUT THE COVER
Images from the Haematologica Atlas of Hematologic Cytology: dyserythropoiesis
Rosangela Invernizzi
University of Pavia, Pavia, Italy
E-mail: ROSANGELA INVERNIZZI - rosangela.invernizzi@unipv.it
doi:10.3324/haematol.2021.280385
T
he most frequently altered lineage in myelodysplastic syndromes is the erythroid one. Some of the morphological
abnormalities of erythroid precursors are displayed in the Figure, showing representative bone marrow smears. In the
top image erythroid hyperplasia with megaloblastoid changes is evident; in addition, note a detached nuclear fragment
within the cytoplasm of an erythroblast (center) and a very large, late erythroblast with a pyknotic, irregularly shaped nucleus
(top left). Giant multinucleated erythroblasts should also be considered dysplastic (bottom image), but variable degrees of
dyserythropoiesis are commonly observed in various hematologic and non-hematologic disorders and have low diagnostic
power. It is of critical importance that dyserythropoiesis is not assessed in isolation and, if dysplasia is confined to erythroid
cells, other causes of erythroid dysplasia should be considered.1
Disclosures
No conflicts of interest to disclose
Reference
1. Invernizzi R. Myelodysplastic syndromes. Haematologica. 2020;105(Suppl 1):78-97.
haematologica | 2022; 107(2) 351

LANDMARK PAPERS IN HEMATOLOGY
The platelet aggregometer
Carlo L. Balduini
Ferrata-Storti Foundation, Pavia, Italy
E-mail: carlo.balduini@unipv.it
doi:10.3324/haematol.2021.280198
TITLE Aggregation of blood platelets by adenosine diphosphate and its reversal
AUTHORS Gustav Victor Rudolph Born
JOURNAL Nature. 1962;194:927-929. PMID: 13871375
M
ax Schultze identified platelets in 1865 and described interested in hemostasis. It is thanks to the aggregometer that
them as "small colorless spherules often grouped Weiss & Aledort demonstrated that the intake of aspirin
together".1 A few years later, in 1882, Giulio Bizzzero inhibits platelet aggregation, John Vane and his colleagues
realized that the ability of platelets to cluster at the site of vas- showed that aspirin is an inhibitor of the prostaglandin-form-
cular damage is the basis of the hemostatic process.2 Although ing cyclooxygenase, and Hamberg & Samuelsson identified the
these two fathers of platelets noted that platelets tend to aggre- pro-aggregating substance (now known as) thromboxane A2.
gate, the mechanisms of this phenomenon remained mysteri- Summarizing his work in the field of platelet aggregation, Born
ous for nearly a century because no technique for studying wrote "It is gratifying that the feedback hypothesis of platelet
platelet function was available. The turning point occurred in aggregation turned out to be explanatory of the remarkable
1962, when Gustav Born, at that time Professor of effectiveness of antiplatelet drugs of the aspirin type in the pre-
Pharmacology at the Royal College of Surgeons of England, vention of heart attacks and strokes". There is no doubt that
published the paper ‘Aggregation of blood platelets by adeno- the discovery of antiplatelet agents is one of the most remark-
sine diphosphate and its reversal' in Nature. In this Letter to the able achievements of pharmacological research of the twenti-
Editor, Born described a simple instrument consisting of a pho- eth century. Moreover, it should not be forgotten that the iden-
tometer that measures the passage of light through a sili- tification of numerous hemorrhagic diseases due to a function-
conized tube in which platelets are kept in suspension by a al platelet defect is largely due to Born’s aggregometer, which
rotating magnetic bar. When platelets aggregate, the optical is still in use in most clinical and research laboratories con-
density of the platelet suspension decreases and light transmis- cerned with thrombosis and hemostasis. The great fame
sion increases. In the same article Born reported that ATP and derived from the aggregometer must not make us forget the
AMP inhibit platelet aggregation and concluded that 'it is con- many other scientific merits of Born, including important stud-
ceivable that AMP or some other substance could be used to ies on atherosclerotic plaques, pioneering use of intravital
inhibit or to reverse platelet aggregation in thrombosis'. The microscopy and direct recording of white blood cell behavior
idea of antiplatelet agents was born! Being strictly against in small vessels. For his achievements Born received honorary
patenting anything of potential medical value for mankind, MD or DSci degrees from ten European and US universities,
Born did not patent his invention that, in a few years, was mar- and the Centre for Vascular Research at the University of
keted by more than ten firms and spread to all laboratories Edinburgh was named after him.
Born's platelet aggregometer. (A) The

first aggregometer produced by Born
in the workshop of the Royal College of
Surgeons in London. (B) Tracings of
platelet aggregation from the paper
published in Nature in 1962. The
addition of ADP to platelet-rich plasma
causes a reduction in the latter’s opti-
cal density which is proportional to the
extent of platelet aggregation. With
low doses of ADP the aggregation is
reversible, while with higher doses it
becomes irreversible (Reproduced
from Nature, with permission).
References
1. Schultze, M. Ein heizbarer Objecttisch und seine Verwendungbei Untersuchungen des Blutes. Archiv für mikroscopische Anatomie. 1865;1:1-42.
2. Bizzozero, J. Ueber einen neuen Formbestandtheil des Blutes und dessen Rolle bei der Thrombose und der Blutgerinnung. Archiv für Pathologische
Anatomie und Physiologie und für klinische Medicin 1882;90:261-332.
352 haematologica | 2022; 107(2)

EDITORIALS
Too much and not enough: revisiting maintenance rituximab in indolent lymphomas
Sonali M. Smith
Section of Hematology/Oncology, University of Chicago, Chicago, IL, USA
E-mail: SONALI M. SMITH - smsmith@medicine.bsd.uchicago.edu
doi:10.3324/haematol.2021.279101
D
espite being slow-growing and chemosensitive dis- receive either an additional 2 years of maintenance therapy
eases, the inevitable outcome of indolent lym- with subcutaneous rituximab or active observation. With a
phomas is relapse and prevention of relapse has been primary endpoint of progression-free survival, 274 patients
an important avenue of investigation for decades. The were randomized and the median follow-up is 28 months.
advent of rituximab allowed consideration of a less toxic Adverse events were slightly increased in the extended main-
(compared to chemotherapy) opportunity to maintain remis- tenance arm. During the observation period, the number of
sion following induction regimens without the need for pro- progression events was quite low in both arms, and the trial
longed exposure to cytotoxic agents. The common ques- is now closed without a clear signal of improved progres-
tions have focused on both the schedule and the duration of sion-free survival for the extended maintenance arm; no con-
maintenance therapy, with the goal to improve progression- clusions can be made regarding survival.
free or overall survival; in an ideal world, maintenance strate- The MabCute trial can thus be added to the list of trials
gies might even seek to cure. showing no advantage from prolonged maintenance with
Rituximab pharmacokinetics and/or impact on B-cell anti-CD20 targeting strategies if disease control, toxicity, and
depletion and subsequent recovery provide sufficient ration- overall survival are considered collectively. As one example,
ale for delivering maintenance rituximab with a variety of the SAKK 35/03 trial randomized a heterogeneous group of
approaches: one dose every 8 weeks, one dose every 12 patients with follicular lymphoma to receive either rituximab
weeks, or four weekly doses every 6 months.1,2 Although every other month for four administrations or rituximab
none of these schedules has been directly compared, small every other month for 5 years;7 patients in this trial could
retrospective reviews suggest relative equivalence in terms of have had treatment-naïve, relapsed, or refractory disease and
efficacy and small differences in terms of toxicity.3 In front- all received induction therapy with rituximab monotherapy
line follicular lymphoma, the PRIMA trial established one and not chemoimmunotherapy. While event-free survival
dose of maintenance rituximab every 8 weeks based on was improved, there was more toxicity and no impact on
achieving a trough level of 25 mg/mL in the majority of survival in the prolonged treatment arm. Other key trials
patients, and this has arguably become the most common have shown no benefit from rituximab treatment at relapse
schedule.4 It is important to note that there remains a lack of compared to a maintenance approach;8 specifically, the
a survival advantage for maintenance rituximab in frontline RESORT trial found no difference in time to next treatment
follicular lymphoma, but 10-year data show impressive and and no improvement in overall survival between mainte-
persistent disease control and validate maintenance ritux- nance and retreatment in low tumor burden indolent lym-
imab for 2 years as an appropriate option to improve pro- phomas. Finally, the induction chemotherapy backbone also
gression-free survival in patients with high-tumor burden influences the risk-benefit profile; for example, the GALLI-
follicular lymphoma. UM trial observed that patients receiving bendamustine-
While the schedule of rituximab maintenance can be justi- based induction had more toxicity and even increased mor-
fied based on pharmacokinetics, the duration of rituximab tality, particularly during the maintenance component of
maintenance is more empirically derived. Given the relative- therapy.9 It is worth noting that the majority of patients in
ly favorable toxicity profile even with prolonged administra- the current trial also received bendamustine-based regimens
tion, studies have ranged from several limited doses to 5 and no toxicity signal was seen with the extended mainte-
years of treatment to indefinite treatment. In the relapsed nance, but this may be because of drop-out during the initial
setting, a meta-analysis from a decade ago suggested maintenance component.
improved overall survival for maintenance rituximab in The COVID-19 pandemic has forced us to re-evaluate data
relapsed/refractory follicular lymphoma, although it is criti- with an additional critical lens related to treatment-associat-
cal to acknowledge that the majority of trials included in this ed B-cell suppression. Early studies during this pandemic
study had involved chemotherapy induction and not showed that patients on immune suppression or those on
chemoimmunotherapy induction.5 chemotherapy suffer more severe complications related to
Based on improved progression-free survival and a lure of SARS-CoV2, and mortality in patients with hematologic
improved overall survival, is more maintenance better? In malignancies is high.10 Since maintenance rituximab has yet
this issue of Haematologica,6 Rule and colleagues present the to offer cure or improved overall survival after chemoim-
final results of the MabCute trial. This prospective, interna- munotherapy induction, it seems more appropriate to identi-
tional, randomized phase III trial sought to determine the fy the minimum number of doses rather than trying to
added benefit of extended rituximab dosing beyond 2 years expand or extend treatment. Indeed, only a minority of
in responding patients with indolent lymphomas. All patients with blood cancers mount a sufficient response to
patients had relapsed or refractory indolent lymphomas, and vaccines;11 accordingly, recent scientific society
could receive any chemoimmunotherapy induction followed (https://www.hematology.org/covid-19/ash-astct-covid-19-and-
by 2 years of rituximab maintenance. Using 2007 response vaccines) and advocacy guidelines12 highlight that patients
criteria, responding patients at 2 years were randomized to receiving B-cell-directed therapies have attenuated or even

Editorials
absent responses to vaccination against SARS-CoV2 and 4. Salles G, Seymour JF, Offner F, et al. Rituximab maintenance for 2
years in patients with high tumour burden follicular lymphoma
cannot abandon social distancing and masking precau- responding to rituximab plus chemotherapy (PRIMA): a phase 3, ran-
tions. domised controlled trial. Lancet. 2011;377(9759):42-51.
In the end, there is no advantage from prolonged main- 5. Vidal L, Gafter-Gvili A, Salles G, et al. Rituximab maintenance for the
tenance in indolent lymphomas, and 2 years should treatment of patients with follicular lymphoma: an updated systemat-
ic review and meta-analysis of randomized trials. J Natl Cancer Inst.
remain the standard duration if maintenance rituximab is 2011;103(23):1799-1806.
offered. The MabCute trial thus supports that extended 6. Rule S, Barreto W, Briones J, et al. Efficacy and safety assessment of
dosing of rituximab is both too much, and also not prolonged maintenance with subcutaneous rituximab in patients with
relapsed or refractory indolent non-Hodgkin lymphoma: results of the
enough to offer cure or improve overall survival. Future phase III MabCute study. Haematologica. 2022;107(2):500-509.
studies should focus on identifying the minimum number 7. Taverna C, Martinelli G, Hitz F, et al. Rituximab maintenance for a
of maintenance doses needed to improve outcomes, par- maximum of 5 years after single-agent rituximab induction in follicu-
ticularly in light of a pandemic threat. lar lymphoma: results of the randomized controlled phase III trial
SAKK 35/03. J Clin Oncol. 2016;34(5):495-500.
8. Kahl BS, Hong F, Williams ME, et al. Rituximab extended schedule or
Disclosures re-treatment trial for low-tumor burden follicular lymphoma: Eastern
No conflicts of interest to disclose. Cooperative Oncology Group protocol E4402. J Clin Oncol.
2014;32(28):3096-3102.
9. Hiddemann W, Barbui AM, Canales MA, et al. Immunochemotherapy
References with obinutuzumab or rituximab for previously untreated follicular
lymphoma in the GALLIUM study: influence of chemotherapy on
1. Berinstein NL, Grillo-Lopez AJ, White CA, et al. Association of serum efficacy and safety. J Clin Oncol. 2018;36(23):2395-2404.
rituximab (IDEC-C2B8) concentration and anti-tumor response in the 10. Mato AR, Roeker LE, Lamanna N, et al. Outcomes of COVID-19 in
treatment of recurrent low-grade or follicular non-Hodgkin's lym- patients with CLL: a multicenter international experience. Blood.
phoma. Ann Oncol. 1998;9(9):995-1001. 2020;136(10):1134-1143.
2. Gordan LN, Grow WB, Pusateri A, Douglas V, Mendenhall NP, Lynch 11. Herishanu Y, Avivi I, Aharon A, et al. Efficacy of the BNT162b2
JW. Phase II trial of individualized rituximab dosing for patients with mRNA COVID-19 vaccine in patients with chronic lymphocytic
CD20-positive lymphoproliferative disorders. J Clin Oncol. leukemia. Blood. 2021;137(23):3165-3173.
2005;23(6):1096-1102. 12. Proceedings of the COVID-19 Vaccine Panel: a Lymphoma Research
3. Nabhan C, Ollberding NJ, Villines D, et al. A systematic review of Foundation White Paper. https://lymphoma.org/wp-
comparative schedule-related toxicities with maintenance rituximab content/uploads/2021/04/Proceedings-of-the-COVID-19-Vaccine-
in follicular and mantle cell lymphomas. Leuk Lymphoma. Panel_LRF-White-Paper_April-2021.pdf; 2021. (Last accessed May 12,
2014;55(6):1288-1294. 2021)
An exciting RXRA mutant revives interest in retinoids for acute myeloid leukemia
Fang Qiu1 and Hugues De The1-3
1
INSERM UMR 944, CNRS UMR 7212, Université de Paris, IRSL, Hôpital Saint Louis; 2Collège de France, Oncologie Cellulaire et
Moléculaire, PSL University, INSERM UMR 1050, CNRS UMR 7241, and 3Department of Hematology, Hôpital Saint Louis (Assistance
publique Hôpitaux de Paris) and Paris University, Paris, France
E-mail: HUGUES DE THÉ - hugues.dethe@inserm.fr
doi:10.3324/haematol.2021.279152
T
he idea that "one size fits all" is obviously outdated silent receptors, allowing sequence-specific DNA recog-
for acute myeloid leukemia (AML) therapy: tomor- nition.4 Hence, therapeutic targeting of RXR could be a
row’s treatments will depend on phenotypically or strategy to activate targets under the control of the
genetically defined subtypes. The most striking example is RXR/RAR transcriptional complex. However, in princi-
acute promyelocytic leukemia (APL), driven by the PML- ple, RXR/RAR signaling cannot be activated by RXR lig-
RARA fusion protein. In APL, a subtype that accounts for ands alone, at least in part because co-repressors remain
5% of cases of AML, a combination of two targeted firmly bound to RAR. This may be modulated by other
agents, all trans-retinoic acid (ATRA) and arsenic trioxide, signaling cascades/second messengers, such as cAMP.5
cures over 90% of patients through PML-RARA driver In AML, this simple view has been challenged. RXR lig-
degradation, differentiation and restoration of PML-depen- ands (rexinoids, such as bexarotene) may exert some dif-
dent senescence.1 These clinical successes have spurred ferentiating effects ex vivo and in vivo.5,6 Hematopoietic
attempts to harness the power of retinoids in other can- cells and some AML express endogenous RXRA ligands.7,8
cers. Unfortunately, ATRA treatment alone remains poorly Two recent studies have revived interest in RXRA signal-
effective in most non-APL AML.2 ing in AML. The first demonstrated that, in AML driven
Retinoid signaling is complex and still incompletely by KMT2A-MLLT3, rexinoids partially suppressed AML
understood.3 ATRA acts primarily through heterodimeric growth and triggered differentiation.8 Moreover, genetic
complexes of retinoic acid receptors (RAR) assembled ablation of RXR accelerated AML growth, while concomi-
with retinoid X receptors (RXR). These RXR are key het- tant activation of both RXRA and RARA precipitated dif-
erodimerization partners of many class-II nuclear recep- ferentiation or apoptosis. It is hoped that dual activation
tors and may be ligand-dependent transcription factors or of these key regulators may harness retinoids more effi-

Editorials
Figure 1. Schematic summary of the effects of constitutively active RXRA DT448/9PP. (A) Under normal circumstances, the transcriptional activity of RXRA het-
erodimerized with other nuclear receptors (NR), including RARA, remains silent, because of co-repressor binding. Selective agonists activate RXRA/NR-driven tran-
scription, resulting in cellular differentiation and growth arrest. (B) Mutations of residues 488/9 in RXRA allow potent ligand-independent transcriptional activation
and drive differentiation.
ciently in AML.2 In the second study, published last year in theme of retinoid sensitivity in non-APL AML2 could be
Haematologica, di Martino et al. report a serendipitously particularly important in the context of combinations of
identified activating mutation in RXRA (RXRA treatment, particularly with decitabine, as encouraging
DT448/9PP), which potently activates rexinoid/retinoid clinical trials have been published recently,10 with more
downstream signaling and suffices to induce terminal dif- likely to come.
ferentiation of KMT2A-MLLT3-transformed cells.9 The C-
terminal helix 12 or AF-2 helix of RXRA, is a critical deter- Disclosures
minant of ligand-dependent transcriptional activity No conflicts of interest to disclose.
through control of co-activator/co-repressor binding.
Surprisingly, di Martino et al. demonstrated that RXRA References
DT448/9PP overexpression resulted in enhanced tran-
1. de Thé H, Pandolfi PP, Chen Z. Acute promyelocytic leukemia: a par-
scriptional activity leading to multiple features of differen- adigm for oncoprotein-targeted cure. Cancer Cell. 2017;32(5):552-560.
tiation, notably loss of colony-forming ability, in KMT2A- 2. Geoffroy MC, Esnault C, de Thé H. Retinoids in hematology: a timely
MLLT3-transformed AML cells. Amazingly, this constitu- revival? Blood. 2021;137(18):2429-2437.
3. de Thé H. Differentiation therapy revisited. Nat Rev Cancer.
tively active RXRA variant binds co-activators completely 2018;18(2):117-127.
independently of ligands. Accordingly, transactivation 4. De Bosscher K, Desmet SJ, Clarisse D, Estébanez-Perpiña E, Brunsveld
could not be abrogated or further boosted by selective L. Nuclear receptor crosstalk-defining the mechanisms for therapeutic
antagonists of RXR or other nuclear receptors, or their innovation. Nat Rev Endocrinol. 2020;16(7):363-377.
5. Altucci L, Rossin A, Hirsch O. Rexinoid-triggered differentiation and
agonists, respectively (Figure 1). tumor-selective apoptosis of acute myeloid leukemia by protein
These intriguing observations imply that even though kinase A-mediated desubordination of retinoid X receptor. Cancer
rexinoids and retinoids synergize for myeloid differentia- Res. 2005;65(19):8754-8765.
6. Sanchez PV, Glantz ST, Scotland S, Kasner MT, Carroll M. Induced dif-
tion of those AML,8 more profound "unconventional" ferentiation of acute myeloid leukemia cells by activation of retinoid
activation by RXRA can initiate terminal differentiation. X and liver X receptors. Leukemia. 2014;28(4):749-760.
This master transcriptional regulatory complex deserves 7. Niu H, Fujiwara H, di Martino O. Endogenous retinoid X receptor lig-
further studies to mechanistically decipher how it can ands in mouse hematopoietic cells. Sci Signal. 2017;10(503):eaan1011.
8. Di Martino O, Niu H, Hadwiger G. Endogenous and combination
become so potent in the absence of ligands. Issues of part- retinoids are active in myelomonocytic leukemias. Haematologica.
ner proteins, post-translational modifications or non-cod- 2021;106(4):1008-1021.
ing RNA, all come to mind. Whatever the molecular 9. Di Martino O, Ferris MA, Hadwiger G, et al. RXRA DT448/9PP gen-
erates a dominant active variant capable of inducing maturation in
mechanism, these observations suggest that the acute myeloid leukemia cells. Haematologica. 2022;107(2):417-426.
RXRA/RARA axis, when super-activated, has the poten- 10. Lübbert M, Grishina O, Schmoor C. Valproate and retinoic acid in
tial to initiate terminal differentiation of some AML cells. combination with decitabine in elderly nonfit patients with acute
myeloid leukemia: results of a multicenter, randomized, 2 x 2, phase
Further studies should determine which AML exhibit this II trial. J Clin Oncol. 2020;38(3):257-270.
exquisite sensitivity to RXRA signaling. This re-emerging

Editorials
Are clinical pharmacology studies still needed in childhood acute lymphoblastic leukemia?
Valentino Conter1 and Francesco Ceppi2
1
Pediatric Hemato-Oncology, Fondazione MBBM, University Milano Bicocca, Ospedale San Gerardo, Monza, Italy and 2Pediatric
Hematology-Oncology Unit, Division of Pediatrics, Department Woman-Mother-Child, University Hospital of Lausanne & University
of Lausanne, Lausanne, Switzerland
E-mail: VALENTINO CONTER - valentino.conter@gmail.com
doi:10.3324/haematol.2021.279059
I
n this issue of Haematologica, Karol et al. report a study it in patients who cannot be treated with the drug. DFCI
on dose intensities for all drugs in two consecutive studies showed that these patients have a poorer out-
acute lymphoblastic leukemia (ALL) clinical trials at St. come. In this context it quite interesting the finding that
Jude Children’s Research Hospital, which differed in their patients with low asparaginase dose intensity, a higher
asparaginase formulation and intensity.1 The amount of systemic methotrexate dose intensity compensated for
data is impressive, with more than 500,000 dosing the low asparaginase dose intensity. The often neglected
records. The main message of the manuscript is that the and yet very relevant aspect of oral medications adminis-
lack of benefit from increased asparaginase intensity may tered at home is also of note. In the study reported in this
be due to the decrease of dose intensity of other drugs, issue, there is the apparent paradox of higher relapse rate
induced by the additional treatment with asparaginase. associated with higher dose intensity for mercaptopurine,
It is widely recognized that intensity of chemotherapy which the authors suggest might reflect low treatment
delivered has an impact on outcome and that drug inter- adherence for oral medications at home (not measured in
actions, which are difficult to assess, can influence anti- this study), in keeping with the findings of the COG
cancer activity and acute and/or late toxicity too. The fast AALL03N1 study, in which it was shown that an adher-
improvement of outcome in childhood ALL in the last ence rate below 90% to maintenance therapy was associ-
three decades of the last century were strictly associated ated with an increased relapse risk.10
with progressive treatment intensity. Dr. Riehm was the Although the expectation for further improvements in
pioneer in this historical process, which was thereafter the treatment of childhood ALL is mostly based on inno-
pursued by all major pediatric oncology groups. In the vative immunological or targeted therapies, pharmaco-
early 1990s, Sallan summarized the Dana-Faber Cancer logical studies remain crucial to improve the therapeutic
Institute (DFCI) experience, largely based on treatment index of combinations of antineoplastic agents. To this
intensification with asparaginase, with the words “More purpose, it must be considered that simple measurement
is better!”,2 and Niemeyer (with Riehm and Sallan) sug- of duration of treatment phases, incidence of severe
gested that merging the intensive elements of Berlin- adverse effects, and dose intensities of single agents may
Frankfurt-Münster (BFM) and DFCI protocols would be a be inadequate or even misleading. What is needed in
logical program to improve outcomes.3 Various attempts order to optimize precision personalized treatment in
were made in this frame, sometimes successfully, such as childhood ALL are comprehensive investigations of com-
in the Children’s Cancer Study Group (CCSG) study with pliance/adherence for all drugs, drug interactions and
Augmented BFM.4 Most studies did not, however, show bioavailability, and germline and tumor sensitivity.
any benefit in intensive BFM-oriented protocols, either
from additional asparaginase treatment as done in Disclosures
Associazione Italiana Ematologia Oncologia Pediatrica No conflicts of interest to disclose.
(AIEOP) ALL 9102,5 European Organization for Research
and Treatment of Cancer - Children Leukemia Group Contributions
(EORTC-CLG) 58951,6 Nordic Society of Pediatric The two authors contributed equally.
Hematology and Oncology (NOPHO) ALL-2008,7 and
BFM ALL 90 trials,8 or from the marked intensification in
the Children’s Oncology Group (COG) AALL1131 trial References
with clofarabine, which was interrupted early due to an
1. Karol SE, Pei D, Smith CA, et al. Comprehensive analysis of dose
excess of toxicity.9 intensity of acute lymphoblastic leukemia chemotherapy.
This general experience has led to a consensus that Haematologica. 2022;107(2):371-380.
treatment intensity in childhood ALL may have reached 2. Sallan SE, Gelber RD, Kimball V, Donnelly M, Cohen HJ. More is
better! Update of Dana-Farber Cancer Institute/Children's Hospital
the maximum tolerated doses, so that further improve- childhood acute lymphoblastic leukemia trials. Haematol Blood
ment can only be obtained by precision medicine based Transfus. 1990;33:459-466.
on targeted therapies. However, most children with ALL 3. Niemeyer CM, Reiter A, Riehm H, Donnelly M, Gelber RD, Sallan
are cured with conventional chemotherapy, which can be SE. Comparative results of two intensive treatment programs for
childhood acute lymphoblastic leukemia: the Berlin-Frankfurt-
further optimized and tailored thanks to the progressive Münster and Dana-Farber Cancer Institute protocols. Ann Oncol.
improvement of biology-based stratification. 1991;2(10):745-749.
The study by Karol et al. shows that room remains for 4. Nachman J, Sather HN, Gaynon PS, Lukens JN, Wolff L, Trigg ME.
Augmented Berlin-Frankfurt-Munster therapy abrogates the adverse
improvement of chemotherapy, although this cannot be prognostic significance of slow early response to induction
achieved by a simple protocol therapy intensification.1 chemotherapy for children and adolescents with acute lymphoblas-
Asparaginase is a drug with a unique mechanism of tic leukemia and unfavorable presenting features: a report from the
action, and there are no suggested alternatives to replace Children's Cancer Group. J Clin Oncol. 1997;15(6):2222-2230.
5. Rizzari C, Valsecchi MG, Aricò M, et al. Effect of protracted high-

Editorials
dose L-asparaginase given as a second exposure in a Berlin-Frankfurt- 8. Schrappe M, Reiter A, Ludwig WD, et al. Improved outcome in
Münster-based treatment: results of the randomized 9102 intermedi- childhood acute lymphoblastic leukemia despite reduced use of
ate-risk childhood acute lymphoblastic leukemia study--a report anthracyclines and cranial radiotherapy: results of trial ALL-BFM 90.
from the Associazione Italiana Ematologia Oncologia Pediatrica. J German-Austrian-Swiss ALL-BFM Study Group. Blood.
Clin Oncol. 2001;19(5):1297-1303. 2000;95(11):3310-3322.
6. Mondelaers V, Suciu S, De Moerloose B, et al. Prolonged versus stan- 9. Salzer WL, Burke MJ, Devidas M, et al. Toxicity associated with
dard native E. coli asparaginase therapy in childhood acute lym- intensive postinduction therapy incorporating clofarabine in the
phoblastic leukemia and non-Hodgkin lymphoma: final results of very high-risk stratum of patients with newly diagnosed high-risk B-
the EORTC-CLG randomized phase III trial 58951. Haematologica. lymphoblastic leukemia: a report from the Children's Oncology
2017;102(10):1727-1738. Group study AALL1131. Cancer. 2018;124(6):1150-1159.
7. Toft N, Birgens H, Abrahamsson J, et al. Results of NOPHO ALL2008 10. Bhatia S, Landier W, Hageman L, et al. 6MP adherence in a multira-
treatment for patients aged 1-45 years with acute lymphoblastic cial cohort of children with acute lymphoblastic leukemia: a
leukemia. Leukemia. 2018;32(3):606-615. Children's Oncology Group study. Blood. 2014;124(15):2345-2353.

REVIEW ARTICLE
Ferrata Storti Foundation Integrating genetic and epigenetic factors in

chronic myeloid leukemia risk assessment:
toward gene expression-based biomarkers
Vaidehi Krishnan,1,2 Dennis Dong Hwan Kim,2,3 Timothy P. Hughes2,4,5,6
Susan Branford2,4,7,8 and S. Tiong Ong1,2,9,10,11
1
Cancer and Stem Cell Biology Signature Research Program, Duke-NUS Medical School,
Singapore, Singapore; 2International Chronic Myeloid Leukaemia Foundation, Bexhill on
Haematologica 2022 Sea, UK; 3Department of Medical Oncology and Hematology, Princess Margaret Cancer
Center, University Health Network, University of Toronto, Toronto, Ontario, Canada; 4School
Volume 107(2):358-370 of Medicine, University of Adelaide, Adelaide, Australia; 5South Australian Health & Medical
Research Institute, Adelaide, Australia; 6Department of Hematology, Royal Adelaide
Hospital, Adelaide, Australia; 7Department of Genetics and Molecular Pathology, Center for
Cancer Biology, SA Pathology, Adelaide, Australia; 8School of Pharmacy and Medical
Science, University of South Australia, Adelaide, Australia; 9Department of Haematology,
Singapore General Hospital, Singapore; 10Department of Medical Oncology, National
Cancer Centre Singapore, Singapore and 11Department of Medicine, Duke University
Medical Center, Durham, NC, USA
ABSTRACT
C
ancer treatment is constantly evolving from a one-size-fits-all
towards bespoke approaches for each patient. In certain solid can-
cers, including breast and lung, tumor genome profiling has been
incorporated into therapeutic decision-making. For chronic phase chronic
myeloid leukemia (CML), while tyrosine kinase inhibitor therapy is the
standard treatment, current clinical scoring systems cannot accurately
predict the heterogeneous treatment outcomes observed in patients.
Biomarkers capable of segregating patients according to outcome at diag-
nosis are needed to improve management, and facilitate enrollment in
clinical trials seeking to prevent blast crisis transformation and improve
the depth of molecular responses. To this end, gene expression (GE) pro-
filing studies have evaluated whether GE signatures at diagnosis are clin-
ically informative. Patient material from a variety of sources has been pro-
filed using microarrays, RNA sequencing and, more recently, single-cell
Correspondence: RNA sequencing. However, differences in the cell types profiled, the tech-
S. TIONG ONG nologies used, and the inherent complexities associated with the interpre-
sintiong.ong@duke-nus.edu.sg tation of genomic data pose challenges in distilling GE datasets into bio-
markers with clinical utility. The goal of this paper is to review previous
Received: July 16, 2021. studies evaluating GE profiling in CML, and explore their potential as risk
assessment tools for individualized CML treatment. We also review the
Accepted: September 28, 2021.
contribution that acquired mutations, including those seen in clonal
Pre-published: October 7 2021. hematopoiesis, make to GE profiles, and how a model integrating contri-
butions of genetic and epigenetic factors in resistance to tyrosine kinase
https://doi.org/10.3324/haematol.2021.279317 inhibitors and blast crisis transformation can define a route to GE-based
biomarkers. Finally, we outline a four-stage approach for the develop-
ment of GE-based biomarkers in CML.
©2022 Ferrata Storti Foundation
Material published in Haematologica is covered by copyright.
All rights are reserved to the Ferrata Storti Foundation. Use of Introduction
published material is allowed under the following terms and
conditions:
https://creativecommons.org/licenses/by-nc/4.0/legalcode. Chronic myeloid leukemia (CML) is a clonal disorder of the hematopoietic stem
Copies of published material are allowed for personal or inter- cell compartment defined and driven by the BCR-ABL1 gene rearrangement and
nal use. Sharing published material for non-commercial pur- the tyrosine kinase it encodes.1 Clinically, it is accompanied by an expansion of
poses is subject to the following conditions:
mostly myelo-erythroid progenitors that maintain the ability to differentiate ter-
https://creativecommons.org/licenses/by-nc/4.0/legalcode,
sect. 3. Reproducing and sharing published material for com- minally into neutrophils. Prior to the introduction of ABL1 tyrosine kinase
mercial purposes is not allowed without permission in writing inhibitors (TKI), most patients would progress to a terminal blast crisis (BC) stage
from the publisher. marked by the acquisition of additional genetic abnormalities within an average
of 5-7 years.2 In this stage, the clinico-pathological features were the inexorable
accumulation of either myeloid or lymphoid progenitors that had acquired aber-

GE-based biomarkers in CML
rant self-renewal properties, broad resistance to cytotoxic achieve a deep molecular response, those who will be
therapies, and eventual patient demise from bone mar- able to stop TKI successfully (Figure 1). The biomarker
row failure.2 The arrival of TKI at the turn of the century would be informative from the time of diagnosis and
resulted in remarkable responses, such that most individ- prior to TKI initiation, since this would enable early strat-
uals treated in chronic phase (CP) CML can expect to ification of patients for therapy with a first-generation
achieve near-normal life expectancies.3 Nevertheless, versus a second/third-generation TKI, allosteric BCR-
CML-related deaths are still reported, mainly due to ABL1 inhibitor, a clinical trial, or preparation for allogene-
resistance and progression to BC, especially in the first ic transplantation. Additionally, among patients who
few years of treatment.4 meet the criteria for stopping TKI therapy, the ideal bio-
marker would identify additional therapies that would
enhance treatment-free remissions. Finally, gene expres-
Current treatment aims and features of an sion (GE)-based biomarkers should be clinically robust,
ideal biomarker and widely available among centers and regions in both
low and high Human Development Index countries.9
Current therapeutic aims are directed at achieving suf-
ficiently deep molecular responses that the risks of BC
transformation are effectively negligible and, in the Why gene expression-based biomarkers?
longer-term, increasing the rates of treatment-free remis-
sion.5-7 Clinical guidelines toward achieving deep molecu- Contributions from genetic and epigenetic mediators to
lar responses have been reviewed elsewhere,8 and at their TKI resistance and BC transformation are well document-
core, prescribe the measurement of BCR-ABL1 transcript ed,10-14 and it is axiomatic that genetic or epigenetic factors
levels using the International Scale (IS) every 3 months as mediating these outcomes will contribute to a cell’s GE
a readout of the depth of the response to TKI. In turn, the signature. Accordingly, GE signatures offer a molecular
depth of TKI response serves as a critical biomarker guid- profile that integrates risk factors encoded by both muta-
ing patient management and prognostication (Figure 1). tions and epigenetic states. However, faithfully extracting
Given current treatment goals, an ideal biomarker and interpreting GE-based information in clinical settings
would accurately predict patients who will achieve a is challenging. Barriers to adoption include technical lim-
deep molecular response with first-line TKI, or require a itations, logistical factors, as well as differences in study
switch to alternative therapy, and, among those who design and data analysis, and are described below.
Figure 1. Features of an ideal chronic myeloid leukemia biomarker. Curves indicate changes in BCR-ABL1 transcript levels, meausred using the International Scale
(BCR-ABL1IS), following initation of tyrosine kinase inhibitor (TKI) therapy in patients with chronic phase (CP) chronic myeloid leukemia (CML). The corresponding
molecular response (MR) value is provided next to the BCR-ABLIS value. Green, orange, and red curves are representative of patients in European LeukemiaNet 2020
‘optimal’, ‘warning’, and ‘failure’ cateogories respectively. Major (MMR) and deep (DMR) molecular remissions are defined as 0.1% (MR 3) and 0.01% (MR4) BCR-
ABLIS, respectively. Green, orange, and yellow bullet points indicate guidelines for each category of response. Gray boxes describe predictive capabilities of an ideal
biomarker. HCT: hematopoietic stem cell transplant.

V. Krishnan et al.
Despite the barriers, recent advances in technological failure based on 17 genes and the signature was validated
and computational platforms are enabling the interroga- in an independent cohort. Of these, eight genes IGFBP2,
tion of patient samples on an unprecedented scale, and SRSF11, BAX, CDKN1B, BNIP3L, FZD7, PRSS57, and
are being translated into robust technical assays on RPS28 intersected with findings of previous CML TKI-
patient material that are reproducible in clinical laborato- resistance and progression studies. This study demon-
ries.15 Such advances may eventually result in the identifi- strated that GE information from diagnostic samples
cation of pretreatment biomarkers that not only predict could predict events long in the future, including major
TKI resistance but suggest alternative non-BCR-ABL1-tar- molecular response (MMR) at 24 months, MR4.5 at 5
geting therapies to pre-empt the emergence of clinical years, and BC transformation.
resistance. Accordingly, it is timely to review the results
of GE studies using primary patient material annotated Gene expression using bone marrow
for clinical outcomes, and assess how genetic and epige- Independently, a series of studies used unselected bone
netic factors associated with treatment outcome con- marrow for comparisons of GE between groups of
tribute to GE signatures. In doing so, it is also important patients with different treatment responses. Frank et al.
to develop models incorporating the interplay between identified a 128 GE signature associated with imatinib
genetic and epigenetic factors, and determine how best to resistance, specifically in an interferon-a pre-treated
use the resulting GE outputs to understand and predict cohort. Differentially expressed genes were involved in
CML drug resistance and transformation. Finally, it is apoptosis (CASP9, TRAP1), DNA repair (MSH3, DDB2),
incumbent on the CML community to outline the practi- oxidative stress protection (GSS, PON2, VNN1) and cen-
cal steps needed for the clinical development of GE-based trosomes (ID1).23 Villuendas et al.24 identified 46 differen-
biomarkers in CML. tially expressed genes of which a six-gene prediction score
(BIRC4, FZD7, IKBKB, IL-7R, TNC, VWF) that correlated
with imatinib resistance after interferon-a failure devel-
Gene expression signatures associated with oped. Differentially expressed genes were involved in cell
resistance to tyrosine kinase inhibitors adhesion (TNC and SCAM-1), drug metabolism (COX1 or
PTGS1), protein tyrosine kinases (MKNK1), and phos-
Since the beginning of the TKI era, a variety of diagnos- phatases (BTK and PTPN22). Notably, the MKNK1/2
tic material from CP patients has been used to discover kinases have been shown by two independent groups to
TKI-resistance GE signatures (Table 1). Here, we review be involved in BC transformation.25,26 In contrast to the
the key conclusions from these studies. prior studies, Crossman et al. found no differentially
expressed genes between the imatinib responder cate-
Gene expression using peripheral blood gories. The use of mixed peripheral blood and bone mar-
In the earliest research by Kaneta et al.16 and McLean et row samples, unselected white blood cells and a heteroge-
al.17 microarray studies were conducted on blood from neous cohort of patients in late CP and heavily pre-treat-
imatinib responders and non-responders. Apart from ed, were suggested as potential reasons for the negative
CBLB, which was downregulated in responders, there was results.27 The important conclusion was that GE compar-
no overlap between the two datasets. De Lavallade et al. isons should be made on purified CD34+ cells. Indeed, in a
conducted microarray studies on peripheral blood meta-analysis comparing six published GE studies in
mononuclear cells to identify a 105-gene set that was CML, DDX11, MSH5, and RAB11FIP3 were the only
enriched in imatinib non-responders, comprising mainly genes coincident between any two of the studies.28 The
genes in cell cycle and DNA repair pathways.18 However, small differences in differential GE between responder
the GE signature could be validated only in an imatinib- groups, different GE platforms, different statistical meth-
treated cohort but not in a cohort treated with interferon- ods and different sources of cells profiled were suggested
a. As a targeted approach, the expression of 21 genes asso- reasons for the poor intersection. The disappointing
ciated with TKI responses and disease progression was results from unselected peripheral blood and bone mar-
studied by Zhang et al.19 Increased PTGS1 expression was row provided the impetus to isolate and study CD34+ frac-
the only gene that differentiated primary imatinib-resis- tions.
tant patients from responders, while 15 genes distin-
guished CP from BC. Twelve genes distinguished ima- Gene expression using CD34+ cells
tinib-responsive from secondary imatinib-resistant CML McWeeney et al. were the first group to use CD34+ cells
without BCR-ABL1 mutations, of which LYN, JAK2, from diagnostic bone marrow.11 Cell adhesion genes were
PTPN22 and CEBPA downregulation was shared with BC upregulated in imatinib-resistant patients suggesting that
samples. The study concluded that at least some features CD34+ cells may establish more adhesive interactions
of secondary imatinib resistance overlap with BC transfor- with the bone marrow milieu. The enrichment for b-
mation. catenin binding targets suggested activated Wnt/b-catenin
More recently, Kok et al. conducted microarray-based signaling in imatinib-resistant patients, a feature shared
analysis on diagnostic blood from 96 CP patients from with CD34+ progenitors from BC.26,29 The authors conclud-
the TIDEL-II trial to predict failure of early molecular ed that primary resistance to imatinib might reflect more
response,20 which correlates with inferior long-term out- advanced disease progression. A 75-probe minimal gene
comes.21,22 Three hundred sixty-five differentially classifier predicted 88% of responders and 83% of non-
expressed genes were identified which were enriched for responders in a validation cohort. Importantly, the authors
‘cell cycle’ and ‘stemness’ (MYC, HOXA9, b-catenin) but of this paper compared their GE signatures to those pre-
depleted for ‘immune-response’ categories in the group dicting early BC transformation, as discussed below, and
with early molecular response failure. A binary classifica- provided an important resource for validation and com-
tion model was built to predict early molecular response parison of other CD34+-based GE datasets.

Table 1. Gene expression profiling studies comparing responders and non-responders to tyrosine kinase inhibitors.
Kaneta et al., McLean et al., Crossman et al., Villuendas et al., Frank et al., McWeeney et al., Zhang et al., de Lavallade et al., Kok et al.,
2002 2004 2004 2006 2006 2009 2009 2010 2019
Stage & CP 18; CP 66 CP 29, included CP 32, CP 23 R; CP 12 R; CP 63; CP 15 CP 96 (discovery);
numbers AP 2; patients previously 12 validation 11 NR 24 NR (discovery); AP 5; CP 88 (validation);
BC 2 on IFN-a CP17 R secondary TKI-R 29; CP 132 (nilotinib Rx).
6 NR (validation) BC 27
Time sample Diagnostic Diagnostic
Prior to TKI, but could Diagnostic BM Diagnostic Diagnostic PB & BM Diagnostic blood Diagnostic PB Diagnostic PB
taken have been on IFN-a PB & BM
Unselected/ Unselected; Unselected; Unselected; Unselected; Unselected; CD34+ selected BM Unselected; Unselected; Unselected;
CD34+ cells; PB & whole blood total WBC BM total WBC MNC & CD34+ PB total WBC MNC MNC
haematologica | 2022; 107(2)

PB/BM MNC from PB & BM from BM & PB MNC in validation group
Platform cDNA Microarray Microarray Microarray Microarray Microarray TaqMan LDA Microarray Microarray
Microarray (Affymetrix (Affymetrix (CNIO OncoChip) (Affymetrix (Affymetrix HG-U133 (Affymetrix HG-U133 (Illumina HT-12v4)
HG_U95Av2) HG_U95Av2) HG-U133A) Plus 2.0) Plus 2.0) TaqMan LDA
N. of DEG 79 55 - 46 128 885 21 105 365
Time of R=MCyR R=CCyR R=CCyR within MCyR at 12 months R=MCyR R=CCyR at CCyR at 12 months CCyR at 12 months; EMR at 3 months
predicted event (<35% Ph+); (0% Ph+); 9 months; (≤35% Ph+); 12 months; NR (failed to achieve
NR= >65% NR= >65% Ph+ NR= >35% Ph+ NR= ≥35% Ph+ NR= >66% Ph+ any cytogenetic response)
Ph+ at 5 months at 12 months after 1 year at 12 months at 12 months
Biological insights First evidence Predictive genes GE comparisons Predictive genes Predictive genes Predictive genes Predictive genes Predictive genes GSEA indicated genes
that GE profiles can enriched for should be made associated with enriched for enriched for cell involved in TKI enriched for DNA associated with
predict sensitivity cell adhesion, on purified CD34+ Wnt signaling, transcriptional adhesion and influx/efflux, BC repair by poorer outcome
to imatinib mitogenic cells cell adhesion, regulation of targets of the progression, recombination enriched for cell
signaling, NK-kB, apoptosis, apoptosis, oxidative Wnt/b-catenin BCR-ABL1 signaling; cycle, stem cell
apoptosis DNA repair stress, DNA repair, pathway Secondary TKI-R function, &
centrosomal genes genes similar to BC depleted for
genes but not primary immune function.
TKI-R
Comments 79 DEG were 31 genes were No DEG were A 6-gene prediction A 128-gene A 75-probe set 15 genes Identified a set A binary
identified. used to develop identified between model was predictor of classifier that distinguished of genes whose classification
15 or 30 genes were a classifier to TKI responders and constructed which primary separated the CP from BC. expression was model based
used to develop a separate TKI- non-responders could predict cytogenetic responder groups. 12 genes differentially on 17 genes.
prediction score to responders from MCyR at 12 months resistance to PPV 87.7%; NPV 73.7% distinguished regulated HR-GES: 77% failure,
separate TKI- non-responders. imatinib was PPV 94.4%; NPV 75%. between in patients but missed 2/9.
responders from identified CD34+ cell selection secondary TKI-R resistant to LR-GES: 95% did well,
non-responders & microarray analysis vs. optimal imatinib but missed 4/79.
possible, successful responders. 64% sensitivity;
in 71% of patients. PTGS1 97% specificity.
Predictive genes predicted primary HR-GES had lower
overlapped with three TKI-R rate of EMR failure
independent datasets with nilotinib
for BC genes (Zheng
et al., 2006), genes
prediciting early BC
transformation (Yong et al.,
2006 ), PRC target genes
in BC (Ko et al., 2020).
CP: chronic phase; AP: accelerated phase; BC: blast crisis; R: responder; NR: non-responder; TKI-R: resistance to tyrosine kinase inhibitors; Rx: treatment; TKI: tyrosine kinase inhibitor; IFN-: interferon-alpha; BM: bone marrow; PB: peripheral blood; PBMC:
peripheral blood mononuclear cells; MNC: mononuclear cells; WBC: white blood cells; N.; number; DEG: differentially expressed genes; MCyR: major cytogenetic response; CCyR: complete cytogenetic response; Ph+: Philadelphia chromosome-positive; EMR:
early molecular response; GE: gene expression; GSEA: gene set enrichment analysis; DEG: differentially expressed genes; PPV: positive predictive value; NPV: negative predictive value; HR-GES: high-risk gene expression signature; LR-GES: low-risk gene expres-
361
sion signature.
V. Krishnan et al.
Single-cell-based gene expression analysis Gene expression signatures associated with

Recent advances in single-cell analysis have enabled blast crisis progression
novel GE-based insights on the roles of tumor cell hetero-
geneity and clonal evolution under the selective pressure Transcriptomic comparisons between the CP and BC
of therapeutics, with obvious applications in biomarker stages have uncovered progression-related signatures that
development.30 Leukemia stem cell (LSC) heterogeneity can herald BC transformation (Table 2). In the pre-TKI era,
was characterized by Warfvinge et al. by combining high- the time to BC transformation from CP varied between
throughput immunophenotyping with single-cell GE pro- patients, and to understand this difference, Yong et al. com-
filing with a defined panel of genes.31 LSC sub-fractions pared CD34+ cells from leukapheresis samples provided by
with more primitive and quiescent signatures had a high- patients who progressed to BC within 3 years (aggressive
er persistence after TKI therapy with the most TKI-insen- leukemia) or after more than 7 years (indolent leukemia) fol-
sitive population identified as Lin–CD34+CD38low/– lowing diagnosis.13 The study identified that lower CD7
CD45RA–cKIT–CD26+ stem cells. Giustacchini et al. used with higher PR3 and ELA2 expression at diagnosis was asso-
the Smart-seq2 platform to combine single-cell RNA- ciated with longer survival. Intriguingly, when the GE signa-
sequencing analysis with BCR-ABL1 transcript detection tures identified by Yong et al. and McWeeney et al. were
using purified stem cells. A sub-population of BCR-ABL1+ compared, a significant overlap was found.11 This important
quiescent stem cells enriched for hematopoietic stem cell study demonstrated that biological processes associated
(HSC) signatures was found to persist during TKI with TKI resistance and early BC transformation over-
therapy.32 Intriguingly, the BCR-ABL1- cells in CML lapped, and that CD34+ cells from different sources (bone
patients were enriched for inflammatory, tumor growth marrow vs. peripheral blood) contained this information.
factor-b and tumor necrosis factor-a hallmarks and dis- In the post-TKI era, a landmark study by Radich et al.
criminated between the TKI-responder groups. identified distinct transcriptional programs during BC pro-
In addition to LSC-derived signatures, the GE signature gression.25 About 3,000 genes were associated with the BC
of immune cells can be equally instructive. For example, stage with a dysregulated WNT/b-catenin pathway,
plasmacytoid dendritic cells, the major producers of inter- decreased Jun B and FOS, and higher PRAME expression.
feron-a in vivo, promoted resistance to nilotinib in CML The Radich dataset was subsequently used to compute a
patients.33 These studies imply that the cytokines released six-gene signature comprising NOB1, DDX47, IGSF2,
by immune cells in the bone marrow microenvironment, LTB4R, SCARB1, and SLC25A3 to predict progression.35
and the transcriptomic changes that they bring about on Independently, Zheng et al. isolated CD34+ cells and iden-
the LSC, may activate cytokine-dependent TKI resistance tified 34 differentially expressed genes as cells transited
programs.34 Together, these single-cell studies demon- from CP to BC. Among the misregulated genes, SOCS2
strate that GE signatures within malignant and non- and CD52 were downregulated while HLA-related genes
malignant compartments in CML are prognostically were overexpressed in BC.36
informative. We anticipate the discovery of additional To understand the biological mechanisms underlying GE
biomarkers among discrete cell types which have the changes in TKI resistance and BC, a recent study tested the
potential to be assayed by platforms available in standard hypothesis that prognostically important genes were
pathology laboratories, e.g., by flow cytometry or enriched for targets of the polycomb repressive complex
immunohistochemistry. (PRC; see below).10 Importantly, target genes of PRC-asso-
Table 2. Gene expression profiling studies comparing chronic phase and acute phase.
Yong et al., Radich et al., Zheng et al., Oehler et al., Ko et al.,
2006 2006 2006 2009 2020
Stage & numbers CP 68 CP 42; CP 11; CP 42; CP 16;
AP 17; BC 9 AP 17; MBC 13;
BC 32 BC 34 LBC 5
Unselected/CD34+; CD34+; Unselected; CD34+; Unselected; CD34+;
PBMC/BM PBMC BM PBMC BM PBMC
Platform Microarray Microarray Microarray Microarray Microarray
N. of genes 20 3000+ 114 6 431 Upregulated
LBC 522 downregulated
Comments Identifies early (≤3 years) Identifies TKI-R Genes that Discriminates between Identifies a core BC
vs. late (≥7 years) BC in CP (had BC-like distinguish early & late CP gene expression
transformation. signature) CP and BC signature common to
Low CD7 & high PR-3 MBC and LBC.
predicts higher OS. PRC-driven
transcriptional
reprogramming is
enriched for poor
prognostic genes in CP
in the CD34+ datasets
of Yong et al. (2006) and
McWeeney et al. (2009).
CP: chronic phase; AP: accelerated phase; BC: blast crisis; MBC: myeloid blast crisis; LBC: lymphoid blast crisis; PBMC: peripheral blood mononuclear cells; BM: bone marrow; PR-
3: proteinase-3: OS: overall survival; TKI-R: resistance to tyrosine kinase inhibitors; PRC: polycomb repressive complex.

ciated silencing in BC progression were enriched for down- carry a somatic mutation that is at least 2% of the variant
regulated genes identified in the datasets of both allele fraction.55 The common genes comprise DNMT3A,
McWeeney et al. and Yong et al.10 The cross-validation of TET2, and ASXL1, and others also found in CML individ-
these three independent datasets suggests important les- uals, including RUNX1, BCORL1, and TP53.56 Individuals
sons for the development of GE-based risk assessment: (i) with clonal hematopoiesis are at increased risk of devel-
the discovery of reproducible GE-based biomarkers is pos- oping hematologic malignancies, and it is therefore per-
sible when homogeneous CD34+ populations are used; (ii) tinent to ask whether clonal hematopoiesis-related
the processes of TKI resistance and BC transformation are mutations also confer increased risk of TKI resistance or
biologically convergent despite genetic heterogeneity:10 progression. A study by Kim et al. has highlighted impor-
and (iii). PRC-regulated processes contribute to silencing of tant features of clonal hematopoiesis-related mutations
prognostically informative genes. in CML.57 Firstly, they may occur in a non-Philadelphia
chromosome-positive clone and predate the develop-
ment of CML, and are unrelated to the CML clone.
Contribution of somatic mutations to gene Secondly, even when a specific mutation occurs in the
expression signatures Philadelphia chromosome-positive clone, it only confers
a relative risk of TKI resistance or progression. Indeed,
Recent reviews have described the range and frequency patients with RUNX1 mutations have been documented
of specific genetic mutations in patients who developed to achieve MMR (personal observation, Dr. Dennis Kim).
TKI resistance and/or BC.37 For many of these genes, there Nevertheless, Kim et al. concluded that mutations in
is strong preclinical information indicating that their asso- genes regulating epigenetic function (TET2, ASXL1
ciated mutations contribute to or are even sufficient to pro- among them) were associated with a higher risk of infe-
duce TKI resistance or transformation phenotypes (sum- rior TKI responses.
marized in Table 3).38-48 These studies imply that genetic There are also strong preclinical data indicating that
mutations alter GE profiles, and here we review their con- clonal hematopoiesis-related mutations result in subtle
tributions to GE changes in CML since these changes may but important changes in GE in HSC. For example,
represent useful GE-based biomarkers. Dnmt3a-deficient HSC show a loss of DNA methylation
For RUNX1 mutations, the Mustjoki group identified an in regions enriched for self-renewal genes such as Meis1,
accompanying GE signature in BC samples.49 They found Evi1 and HoxA9.58 In Tet2-deficient mice, the loss of DNA
that RUNX1 mutations were associated with the upregula- demethylation is accompanied by an expansion of the
tion of stemness, B-cell markers, interferon and immune stem and progenitor cell compartments, and eventual
signaling and transcription factors regulating plasmacytoid myeloproliferation.45 In ASXL1-deficient mice, an
dendritic cell development. increase in self-renewal capacity of stem cells is
In analogous work, the overexpression of an IKZF1 observed, through the loss of PRC1-mediated gene
dominant-negative mutant in CD34+ cells from CP patients repression.59 Another interesting aspect of hematopoietic
increased STAT5 expression, a pathway associated with stem and progenitor cells harboring inactivating muta-
imatinib resistance,50 and enhanced transformation.51 RAG tions of DNMT3A and TET2 is that they both led to
expression status was recently assessed in diagnostic sam- increased cytokine production in peripheral myeloid
ples, given the role of RAG recombination as a mediator of cells, including interleukin-6 and interleukin-1b.60,61
IKZF1 deletions.52 Notably, RAG1/2 and DNTT upregula- Furthermore, mutations associated with clonal
tion at diagnosis suggested imminent lymphoid BC trans- hematopoiesis are frequently found in monocytes, gran-
formation within 12 months (8/8 patients), demonstrating ulocytes, and natural killer cells compared to B or T cells,
that GE signatures can reliably predict transformation. suggesting that their effects may also be manifest in mul-
Despite limited functional interrogation of ASXL1 using tiple differentiated cell types within the hematopoietic
CML patient material, insertion sites within ASXL1 pro- compartment.62 Together, these observations are relevant
moted BC progression in a CP mouse model subjected to to the search for prognostic GE signatures in CML for the
transposition-based mutagenesis.53 Transgenic expression following reasons: (i) increased inflammation and
of truncated protein ASXL1aa1-587 in mice increased HSC cytokine production is associated with LSC persistence,32
self-renewal, and Brd4 occupancy and chromatin accessi- and disease progression;10,63 (ii) prognostic GE changes
bility around genes required for stemness, and predisposed may be found in both CD34+ and CD34– fractions of
mice to myeloid malignancies.54 However, the clinical rele- peripheral blood or bone marrow mononuclear cells; and
vance of diagnostic ASXL1 mutations is still unclear (iii) changes in natural killer cell function and number
because some patients with ASXL1 variants at diagnosis may predict treatment-free remissions, and presumably
can achieve a MMR after TKI therapy.14 Furthermore, contain informative natural killer cell GE signatures.64-66
ASXL1 mutations frequently disappeared when monitored
in the long-term during TKI therapy (personal observation by
Dr. Dennis Kim). Meanwhile, direct evidence for contribu- Epigenetic contributions to gene expression
tions of other mutations to CML GE signatures is currently signatures
lacking, and we have to infer them from studies in other
malignancies (Table 3). Polycomb repressive complex-associated gene
expression changes
Among the most well studied epigenetic complexes in
Lessons from clonal hematopoiesis CML are the polycomb group (PcG) proteins.67 The poly-
comb group proteins assemble into two complexes, PRC2
Clonal hematopoiesis is the clinical phenomenon by and PRC1, which modify histones through repressive
which populations of hematopoietic cells expand and H3K27 trimethylation (H3K27me3) and H2AK119

V. Krishnan et al.
monoubiquitination (H2AK119ub1),respectively, and in ASXL1, IKZF1, and BCORL1 drive aberrant PRC recruit-
general repress gene expression.68 ment and GE in CML remains to be determined.
From a mechanistic standpoint, the most commonly In this respect, a recent study determined that lym-
occurring mutations in CML appear to converge in their phoid and myeloid BC transcriptomes are highly congru-
ability to interact with and function in conjunction with ent, and that both undergo PRC-driven epigenetic repro-
the polycomb group proteins. ASXL1 functions in tran- gramming towards a convergent transcriptomic state.10
scriptional repression through its interaction with PRC2 PRC-dependent epigenetic reprogramming was attrib-
and BAP1.69 BCORL1 is a transcriptional co-repressor that uted to gain- and loss-of-function mutations in members
interacts with PCGF1, a core complex of the PRC1.1 com- of the PRC1 and PRC2 complexes, respectively. Of these,
plex.70 The RUNX1-CBF-b heterodimer mediates tran- ongoing BMI1/PRC1 activity contributes to maintaining
scription by binding to RUNX sites, but also represses the BC transcriptome, while EZH2/PRC2-binding was
transcription by interacting and recruiting BMI1 of the instructional for DNA hypermethylation-dependent gene
PRC1 complex to target sites.71 IKZF1 regulates transcrip- repression. Importantly, the integrative model proposed
tion by interacting with repressive epigenetic complexes by Ko et al. suggests that enrichment for PRC-dependent
such as HDAC1, HDAC2, CHD3, CHD4, and SWI/SNF GE signatures at diagnosis can predict disease transforma-
complex, and also recruits PRC2 to target gene loci in T tion and TKI resistance, as highlighted above.10
cells.72 Thus, while the commonly mutated genes in CML We also note that dysregulated regulation of PRC has
have their own exclusive roles in transcriptional regula- been identified as a key feature of TKI-resistant LSC in
tion, they also share a striking commonality as modula- CP. EZH2 expression was higher in CML LSC than in nor-
tors of the PRC. Whether mutated variants of RUNX1, mal HSC, and CML LSC have a stronger dependence on
Figure 2. Diagrammatic representation of the ‘seed and soil’ model of chronic myeloid leukemia. The model proposes that both acquired mutations and the cell
state of the mutation-acquiring cell contribute to the process of full transformation to blast crisis (BC). A ‘strong’ mutation is defined as being able to confer self-
renewal function on a progenitor cell that does not possess inherent self-renewal capacity. A ‘weak’ mutation is unable to confer self-renewal function and can only
transform a cell with native self-renewal ability, i.e., a stem cell. For both strong and weak mutations, it is likely that additional genetic and epigenetic events are nec-
essary to confer the full suite of features required for BC transformation. The model is also based on the recent finding that BC progenitors which harbor different
somatic mutations share a common or core transcriptome enriched for stemness, quiescence, and inflammatory gene expression signatures.20 HSC: hematopoietic
stem cell; LSC: leukemia stem cell; MPP: multipotential progenitor; LMPP: lymphoid-primed multipotent progenitor; CMP: common myeloid progenitor; GMP: granu-
locyte-macrophage progenitor; TKI-S/R: tyrosine kinase inhibitor-sensitive/resistant cells. Rx: treatment; CP: chronic phase; BC: blast crisis; CML: chronic myeloid
leukemia; GE: gene expression.

Table 3. Functional effects of frequently mutated genes in blast crisis.
RUNX1 IKZF1 ASXL1 BCORL1 GATA2 TET2 DNMT3A
Frequency: diagnosis/ 2.6/18.3 6.1/16.0 9.7/15.1 0.9/8.6 0/8.4 0.9/6.7 2.3/4.5
progression (%)
Function Transcription Transcription Transcriptional Transcriptional Transcription Methylcytosine DNA methylation
factor factor repression co-repressor factor dioxygenase
Mode of action and DNA binding via DNA binding via zinc Regulator of H3K27me3 Represses transcription DNA binding via zinc Conversion of Transfer of methyl group
interactions RUNT domain. finger domain. & H2AK119ub1 marks by binding to class II finger domain 5-methylcytosine to to cytosines on DNA
HDAC & CTBP1 5-hydroxymethylcytosine

Interactions P300, CBP, PRC1, NuRD, HDAC1, HDAC2, CHD3, PRC2 , BAP1 complexes PCGF1, the core PRC1.1 FOG1 through N-terminal 2-HG, vitamin C, OGT, PRC2, EVI1, ISGF3, AP2a,
with other complexes SWI/SNF, MLL/TrxG CHD4, PRC2, CtBP1, component Zinc finger domain WT1, VPRBP, IDAX ZEB1, HDAC1
SWI/SNF
Aberration in CML RUNT domain mutations Deletions- Majority are frameshift Frameshift, nonsense Zinc finger domain Missense, nonsense DNMT3A mutations are
deletions and fusions exon D4–7 (IK6), and nonsense mutations in mutations variants and frameshift mutations mostly CHIP mutations
exon D2–7 exon 12 TET2 mutations may be since they are also
CHIP-related or a part present in the Ph- clone
of the Ph+ clone
Effects on gene Up: Interferon signaling, Up: JAK-STAT signaling, Increased Brd4 occupancy Transcriptional repressor CML L359V mutant Impaired DNMT3A-deficient HSC
expression immune molecules, self-renewal genes. and chromatin accessibility of E-cadherin. inhibits 5-methylcytosine show loss of DNA
pDC- TF. Down: B-cell lineage around genes Other targets unknown transactivation by PU.1. hydroxylation and methylation at the edge
Down: DNA repair and DNA repair genes GATA2 MDS and AML decreased methylationof hypomethylated
mutants have altered at CpG sites in myeloid
canyon regions enriched
transactivation activity cancers with mutant for self-renewal genes
TET2 such as MEIS1, EVI1,
HOXA9
Effect of mutant protein/ RUNX1 H78Q orV91 IK6 expression in CD34+ Truncated ASXL1 increased BCORL1 depletion MDS and AML GATA2 TET2 silencing in human Nearly a third of CHIP-
gene knock out in vitro fs-ter94 in 32D-BCR-ABL1 cells isolated from CP-CML proliferation, and decreased increased the re-plating mutants inhibit CD34+ cells increased the related DNMT3A
model blocked patients enhances their differentiation along capacity of Runx1- differentiation and monocytic lineage at the mutations reduce protein
differentiation in vitro expansion megakaryocyte and erythroid depleted Lin− cells apoptosis expense of erythroid and stability
lineages lymphoid lineages
Effect of mutant protein/ RUNX1 H78Q or Ikaros DNA binding AML, MPN, MDS-like Effect of CML BCORL1 GATA2 deficiency has Conditional TET2 loss Dnmt3a ablation in HSC
gene knock out in vivo V91fs-ter94 mutants domain inactivation diseases variants unknown been recognized as a in the hematopoietic predisposes mice to
induced a BC or in early pre-B cells major MDS compartment leads develop a spectrum of
accelerated leads to ALL predisposition syndrome to increased stem cell myeloid and lymphoid
phase-like phenotype in humans self-renewal malignancies
in mice
Effect on CML Yes Yes Yes No No No No
variant studied
References Zhao et al., 2012 Joshi et al., 2014 Yang et al., 2018 Pagan et al., 2007 Zhang et al., 2009 Kim et al., 2017 Mayle et al., 2015
Awad et al., 2020 Beer et al., 2015 Katoh, 2013 Wong et al., 2016 Branford et al., 2018 Pronier et al., 2011 Hervouet et al., 2018
Branford et al., 2018 Balasubramani et al., 2015 Crusio et al., 2011 Huang et al., 2018
Kim et al., 2017
Branford et al., 2018
HDAC: histone deacetylases; CML: chronic myeloid leukemia; CHIP: clonal hematopoiesis of indeterminate potential; MDS: myelodysplastic syndrome; AML: acute myeloid leukemia; HSC: hematopoietic stem cell; CP: chronic phase; ALL: acute lymphoblastic
leukemia; MPN: myeloproliferative neoplasm.
365
V. Krishnan et al.
the PRC2-EZH2 axis for survival and TKI resistance.73,74 stem cell fate, differentiation and leukemia-related func-
Likewise, higher BMI1 levels at diagnosis correlated with tions.10 Mechanistically, differential DNA methylation pat-
disease progression from CP to BC.12 while BMI1 overex- terns in CML have been attributed to underlying
pression in CP CD34+ cells increased proliferation and DNMT3A/TET2 mutations, PRC2-dependent epigenetic
self-renewal,75 and transformed B-lymphoid progenitors re-programming, and cytosolic sequestration of Tet2 by
in vivo.76 BCR-ABL1.79 Notably, the physiological targeting of DNA
hypermethylation using 5-aza-2’-deoxycytidine amelio-
DNA methylation-associated gene expression changes rated disease phenotypes in a mouse model of CP dis-
Many studies have examined the role of DNA methyla- ease,80 while low-dose decitabine displayed clinical activi-
tion as a regulator of aberrant GE in CML pathogenesis. In ty in patients refractory to imatinib,81 suggesting DNA
candidate-based approaches, genes involved in cell cycle methylation does indeed contribute to TKI resistance.
regulation (P16, P53, PLCD1, PER3, HIC1), differentiation Based on the biological insights gleaned so far, it is pos-
(HOXA4, DLX4, DDIT3, SPI1) proliferation (CDH13, sible that progression-related DNA methylation signa-
DAPK1), apoptosis (BIM), Wnt regulation (sFRP1, CBY1), tures may already be evident at diagnosis, particularly in
LSC maintenance (MTSS1), and cell signaling (Jun B, patients presenting with advanced CP.10 The DNA methy-
SOCS2) were identified as targets of DNA methylation.67,77 lation status of specific target genes might therefore be
Recent unbiased genome-wide methylome analyses useful in the timely identification of such patients for
have solidified the concept of aberrant DNA methylation more aggressive therapies. Furthermore, given that DNA
as a driver of resistance and transformation. The number methylation is a relatively stable epigenetic and biochem-
of differentially methylated regions in CP increased from ical mark, there are practical advantages to developing
~600 to ~6,500 CpG sites in BC.78 BC was associated with DNA methylation-based biomarkers rather than tran-
heightened DNA hypermethylation, and to a lesser extent script-based readouts, especially for the development of
hypomethylation, around promoters of genes involved in robust clinical-grade tests (Figure 3).
Figure 3. Stages of development of gene expression-based biomarkers. In chronic myeloid leukemia (CML), the development of gene expression-based biomarkers
can be divided into three stages following an initial discovery phase. These stages will each determine the analytical validity, clinical validity, and clinical utility of the
tests in question. Examples of CML-specific issues or questions that are pertinent to each stage are outlned in boxes under each stage. GE: gene expression; IHC:
immunohistochemistry; FC: flow cytometry; RT-PCR: reverse transcriptase polymerase chain reaction; ISH: in-situ hybridization; LCM: laser capture microdissection;
scRNA-seq: single-cell RNA sequencing; ATAC-seq: assay for transposase-accessible chromatin sequencing; BM: bone marrow; PBMC: peripheral blood mononuclear
cell; MNC: mononuclear cells; FFPE: formalin-fixed paraffin-embedded tissues; PB: peripheral blood; NK: natural killer cells; MDSC: myeloid-derived suppressor cells;
TKI: tyrosine kinase inhibitor; NCCN: National Comprehensive Cancer Network; ELN: European LeukemiaNet; BC: blast crisis; EFS: event-free survival; DFS: disease-
free survival; PFS: progression-free survival; OS: overall survival, TFR: treatment-free remission; 95% CI: 95% confidence interval; DMR: deep molecular response..

Gene expression profiles and mutations: ‘seed another is seen when the TET2 variant allele frequency
and soil’ revisited reduces after TKI treatment in patients with disease pro-
gression. In other cases, TET2 mutations were also detect-
As described above, it will be important to develop ed within Philadelphia chromosome-negative cells, and
CML models that integrate the interaction between here, patients showed complex outcomes following TKI
genetic and epigenetic factors in driving drug resistance therapy, with some achieving MMR and others showing
and disease transformation. In this respect, the effects of TKI resistance. These observations suggest that the effect
specific mutations may be cell-context dependent, with of TET2 mutations are highly contextual.
differential effects on GE and function depending on the
cell type being examined. This is particularly the case for
mutations affecting transcription factors, for which cell Challenges ahead but room for optimism
states, and their attendant chromatin accessibility profile,
determine whether the mutated transcription factor has As described above, the discovery of a limited and
access to its target genes. tractable set of genes that is prognostic across a majority
To integrate contributions from both the above fea- of CML patients has been challenging for clinical, biolog-
tures, we propose a model in which the cell of origin, ical, and technical reasons. Nevertheless, there is room for
with its attendant epigenetic and transcriptional program, optimism. In the setting of breast cancer, GE panels com-
determines the ability of specific mutations to contribute prising 21 genes that encompass various aspects of breast
to biological and clinical outcomes (Figure 2). This model cancer biology have been found to be predictive of thera-
is a derivative of the ‘seed and soil’ hypothesis of cancer peutic response, and minimized the use of additional
initiation.82 The model will be useful for hypothesis test- therapy without compromising survival.85 Among liquid
ing, and likely explains an important feature of BCR- tumors, a recent study in acute myeloid leukemia demon-
ABL1 itself. It has been shown in murine models that strated that a parsimonious 17-gene GE score, derived
only when expressed in HSC, but not more committed from a larger set of stemness-conferring genes, predicts
progenitors, can BCR-ABL1 induce a myeloproliferative resistance to initial therapy.86 Interestingly, this score was
disorder. This is likely because BCR-ABL1 is incapable of independent of cytogenetic and mutational risk factors,
conferring self-renewal capacity upon committed progen- and suggests that biological factors (e.g., stemness) tran-
itors, indicating that CML cells rely on BCR-ABL1-inde- scend traditional genetics-based groupings.87
pendent mechanisms for stemness programs. These find- Encouragingly in CML, two recent reports suggest that
ings are in contrast to those for other leukemia fusion it is possible, using peripheral blood samples taken at
genes (e.g., MLL-ENL, MLL-AF9, MOZ-TIF2) which are diagnosis or 3 months after diagnosis, to predict deep
capable of conferring self-renewal and transform progen- molecular responses and also sustained treatment-free
itor cells.83 Relatedly, the model may also explain a natu- remissions. In the first study, the Adelaide group showed
rally occurring phenomenon whereby normal individuals that the rate of decline of BCR-ABL1 transcripts during
found to carry the BCR-ABL1 fusion in their peripheral first-line TKI therapy (calculated from baseline and 3-
blood mononuclear cells apparently never develop month BCR-ABL1 transcript levels) predicts success of
CML.84 Here, the model would posit that the BCR-ABL1 treatment-free remission.88 The time taken for BCR-ABL1
fusion is occurring in a long-lived progenitor without self- transcripts to halve was the strongest independent predic-
renewal function. tor of sustained treatment-free remission: 80% in patients
Analogous to the situation regarding cancer initiation with a halving time of <9.35 days versus 4% if the halving
by leukemia fusion genes, mutations devoid of self- time was >21.85 days (P<0.001). In a separate study,
renewal function may only confer an increased risk of BC Radich et al. reported that GE signatures from peripheral
transformation when they occur in a target cell that blood taken prior to TKI initiation can distinguish indi-
already possesses physiological self-renewal function. viduals who will achieve a deep molecular response
According to this model, mutations in RUNX1 that are (MR4.5) at 5 years from those who will have suboptimal
sufficient to induce BC-like disease in mice (Table 3) may responses.89 Thus, biological information encoded in GE
be deemed a ‘strong’ biological seed that can transform data can predict very long-term clinical outcomes in
many cell types within the hematopoietic hierarchy. Such CML, and it is therefore conceivable that GE-based data
mutations would be expected to induce disease progres- will be able to identify patients in whom TKI therapy can
sion in the majority of patients who harbor such muta- be safely discontinued. More importantly, these early
tions, which is indeed the case.37 However, a minority of reports suggest that despite the likely existence of diverse
CP patients with RUNX1 mutations continue to enjoy resistance mechanisms within the study populations,
sustained deep maolecular responses,57 suggesting the final common paths, readout either as dynamic measures
existence of other important factors that modulate of BCR-ABL1 transcript levels, or peripheral blood GE sig-
RUNX1 function. Along the same lines, ASXL1 was natures are indeed discoverable.
recently identified as the most frequently mutated gene at
diagnosis in nine patients, the majority (n=6) of whom
eventually developed BC, while a minority (n=2) Stages in developing gene expression-based
achieved a MMR.14 risk assessment
In contrast to the above examples, the prognostic
impact of ‘weak’ seeds is much less clear. In a study by The stages of developing GE-based tests has been out-
Kim et al., at least four different patterns were observed lined in recent reviews and consensus statements, and
for TET2 mutations.57 One pattern is seen in patients with comprise at least three phases that assess: analytical
TKI resistance when both TET2 and ABL1 variant allele validity (reliably measuring the genotype of interest),
frequencies increased following TKI therapy, while clinical validity (ability to segregate patients into biologi-

V. Krishnan et al.
cally and clinically important subsets), and clinical utility pliance) to overall outcomes. Given the relative rarity of
(ability to improve clinical decision making).90,91 In this CML, it is envisaged that this will be a multicenter inter-
section, we summarize the pertinent stages and highlight national study.
issues of particular relevance to GE-based biomarker
development in CML (Figure 3).
Stage 0 is the discovery stage, which is where the field Conclusion
is currently. Here, we highlight three important compo-
nents, which include the use of technical approaches for Genetic and epigenetic events contribute to the emer-
unbiased discovery, the simultaneous interrogation of gence of BCR-ABL1-independent clones that result in
leukemic and non-leukemic clones from the same sample clinical TKI resistance and, if unopposed, BC transforma-
(since both have been shown to be prognostic), and the tion. Long-term TKI responses, including successful TKI
use of robust statistical and computational pipelines to stoppage, can be predicted by slower declines in BCR-
discover minimal prognostic genes sets. The advent of ABL1 transcript levels during first-line TKI therapy,88 sug-
single-cell-based technologies and their application to gesting that genetic and epigenetic factors contributing
well-annotated cohorts will facilitate this step. to TKI resistance are present at diagnosis. Recent studies
In stage I, the minimal gene set has to be converted into describe a convergent GE signature common to the
a clinical test that accurately and reproducibly measures majority of BC progenitors.10 Elements of this common
the GE phenotype. The test platform needs to be robust, or core transcriptome can be detected in CD34+ cells
as well as sensitive, specific and reliable. The assay from CP patients at risk of TKI resistance or early trans-
should be developed for tissues that are collected as part formation,11,13 and specific mutations have been shown to
of routine clinical care. Ideally, any additional processing contribute additional nuances to the core transcrip-
of material beyond what is routine should be minimized, tome.49 These observations are consistent with a ‘seed
e.g., CD34+ selection, and should utilize standard proce- and soil’ model that may be helpful for hypothesis gen-
dures available in clinical laboratories, such as flow eration (Figure 2). Emerging technologies, particularly
cytometry and bone marrow immunohistochemistry. An multimodal single-cell-based approaches, will facilitate
example would be detecting GE signatures of interest by the discovery of genetic and epigenetic biomarkers at
a panel of antibodies for use in flow cytometry or presentation. This initial discovery phase has to be fol-
immunohistochemistry applications. It is preferable that lowed by the translation of GE-based information into
the samples used for analytical validation are from well- validated analytical tests, and subsequently, the determi-
characterized patients representative of ‘real-world’ set- nation of clinical validity and utility. This process will be
tings and, ideally, validated in at least one independent a multi-year, multi-institution international effort akin to
cohort. Sample size and power calculations should be that for the development of a genetic-based risk assess-
determined prior to starting the study, and analytic sensi- ment.85,90,91 The integration of both gene mutation- and
tivity and specificity for the test should be available at the gene expression-based biomarkers into the care of CML
end of the study. At the end of stage I, a locked-down test patients will be an important step to achieving the ulti-
should be evaluated in stage II, that of clinical validation. mate goal of CML research: the cure of the majority of
In stage II, the locked-down test will be evaluated for our patients.
its ability to differentiate between clinically meaningful
outcomes in modern CML practice. The samples to be Disclosures
tested should be obtained from well-annotated cohorts DK is a member of the advisory boards of Novartis, Pfizer,
representative of the broader population, and the test Paladin, and has received honoraria from Novartis, Pfizer and
conducted on tissues in a blinded manner with respect to Paladin, as well as research funding from Novartis, Bristol-
testing and result reporting. Ideal populations include Myers Squibb, Pfizer, and Paladin. TH is a member of a
patients who have been treated uniformly in clinical tri- Novartis advisory board and receives research support from
als. At the end of this stage, the ability of the test to pre- Novartis and Bristol-Myers Squibb. SB is a member of the advi-
dict clinical outcome should be available as a test score, sory boards of Qiagen, Novartis, and Cepheid and has received
with clearly defined positive and negative predictive val- honoraria from Qiagen, Novartis, Bristol-Myers Squibb, and
ues. Cepheid, as well as research support from Novartis and
The final stage, stage III, will be the determination of Cepheid.
clinical utility. This stage would entail the use of the GE-
based test to improve clinical decision-making, and Contributions
would require the study to demonstrate that meaningful VK and STO conceived the topic for review, and wrote the
outcomes are improved when the test is used compared first draft of the manuscript. DDHK, TPH, and SB contributed
to when the test is not used. Besides clinical outcomes by the addition of new sections and critical discussions through-
such as improved progression-free survival and overall out the writing of the review.
survival, additional measures such as cost-effectiveness,
avoidance of toxicities, quality of life and psychological Funding
parameters should also be assessed. Such studies may VK and STO are supported by the National Medical
also incorporate the contribution of pharmacological fac- Research Council Singapore (MOH-CSASI18may-0002,
tors (e.g., drug metabolism and side effects, patient com- MOH-CIRG20nov-0003, NMRC/CIRG/1468/2017).

References genome-wide cDNA microarray analysis. 2017;23(6):692-702.

Jpn J Cancer Res. 2002;93(8):849-856. 33. Inselmann S, Wang Y, Saussele S, et al.
17. McLean LA, Gathmann I, Capdeville R, Development, function, and clinical signifi-
1. Rowley JD. Letter: A new consistent chro- Polymeropoulos MH, Dressman M. cance of plasmacytoid dendritic cells in
mosomal abnormality in chronic myeloge- Pharmacogenomic analysis of cytogenetic chronic myeloid leukemia. Cancer Res.
nous leukaemia identified by quinacrine response in chronic myeloid leukemia 2018;78(21):6223-6234.
fluorescence and Giemsa staining. Nature. patients treated with imatinib. Clin Cancer 34. Sinnakannu JR, Lee KL, Cheng S, et al.
1973;243(5405):290-293. Res. 2004;10(1 Pt 1):155-165. SRSF1 mediates cytokine-induced impaired
2. Perrotti D, Jamieson C, Goldman J, Skorski 18. de Lavallade H, Finetti P, Carbuccia N, et al. imatinib sensitivity in chronic myeloid
T. Chronic myeloid leukemia: mechanisms A gene expression signature of primary leukemia. Leukemia. 2020;34(7):1787-1798.
of blastic transformation. J Clin Invest. resistance to imatinib in chronic myeloid 35. Oehler VG, Yeung KY, Choi YE, et al. The
2010;120(7):2254-2264. leukemia. Leuk Res. 2010;34(2):254-257. derivation of diagnostic markers of chronic
3. Druker BJ. Translation of the Philadelphia 19. Zhang WW, Cortes JE, Yao H, et al. myeloid leukemia progression from
chromosome into therapy for CML. Blood. Predictors of primary imatinib resistance in microarray data. Blood. 2009;114(15):3292-
2008;112(13):4808-4817. chronic myelogenous leukemia are distinct 3298.
4. Holyoake TL, Vetrie D. The chronic from those in secondary imatinib resist- 36. Zheng C, Li L, Haak M, et al. Gene expres-
myeloid leukemia stem cell: stemming the ance. J Clin Oncol. 2009;27(22):3642-3649. sion profiling of CD34+ cells identifies a
tide of persistence. Blood. 2017;129(12): 20. Kok CH, Yeung DT, Lu L, et al. Gene molecular signature of chronic myeloid
1595-1606. expression signature that predicts early leukemia blast crisis. Leukemia. 2006;20(6):
5. Hochhaus A, Baccarani M, Silver RT, et al. molecular response failure in chronic-phase 1028-1034.
European LeukemiaNet 2020 recommenda- CML patients on frontline imatinib. Blood 37. Branford S, Kim DDH, Apperley JF, et al.
tions for treating chronic myeloid Adv. 2019;3(10):1610-1621. Laying the foundation for genomically-
leukemia. Leukemia. 2020;34(4):966-984. 21. Marin D, Ibrahim AR, Lucas C, et al. based risk assessment in chronic myeloid
6. Mahon FX, Rea D, Guilhot J, et al. Assessment of BCR-ABL1 transcript levels leukemia. Leukemia. 2019;33(8):1835-1850.
Discontinuation of imatinib in patients at 3 months is the only requirement for pre- 38. Zhao LJ, Wang YY, Li G, et al. Functional
with chronic myeloid leukaemia who have dicting outcome for patients with chronic features of RUNX1 mutants in acute trans-
maintained complete molecular remission myeloid leukemia treated with tyrosine formation of chronic myeloid leukemia and
for at least 2 years: the prospective, multi- kinase inhibitors. J Clin Oncol. 2012;30(3): their contribution to inducing murine full-
centre Stop Imatinib (STIM) trial. Lancet 232-238. blown leukemia. Blood. 2012;119(12):2873-
Oncol. 2010;11(11):1029-1035. 22. Hughes TP, Saglio G, Kantarjian HM, et al. 2882.
7. Ross DM, Hughes TP. Treatment-free Early molecular response predicts out- 39. Joshi I, Yoshida T, Jena N, et al. Loss of
remission in patients with chronic myeloid comes in patients with chronic myeloid Ikaros DNA-binding function confers inte-
leukaemia. Nat Rev Clin Oncol. 2020;17 leukemia in chronic phase treated with grin-dependent survival on pre-B cells and
(8):493-503. frontline nilotinib or imatinib. Blood. progression to acute lymphoblastic
8. Branford S, Fletcher L, Cross NC, et al. 2014;123(9):1353-1360. leukemia. Nat Immunol. 2014;15(3):294-
Desirable performance characteristics for 23. Frank O, Brors B, Fabarius A, et al. Gene 304.
BCR-ABL measurement on an international expression signature of primary imatinib- 40. Balasubramani A, Larjo A, Bassein JA, et al.
reporting scale to allow consistent interpre- resistant chronic myeloid leukemia Cancer-associated ASXL1 mutations may
tation of individual patient response and patients. Leukemia. 2006;20(8):1400-1407. act as gain-of-function mutations of the
comparison of response rates between clin- 24. Villuendas R, Steegmann JL, Pollan M, et al. ASXL1-BAP1 complex. Nat Commun.
ical trials. Blood. 2008;112(8):3330-3338. Identification of genes involved in imatinib 2015;6:7307.
9. Malhotra H, Radich J, Garcia-Gonzalez P. resistance in CML: a gene-expression pro- 41. Pagan JK, Arnold J, Hanchard KJ, et al. A
Meeting the needs of CML patients in filing approach. Leukemia. 2006;20(6): novel corepressor, BCoR-L1, represses tran-
resource-poor countries. Hematology Am 1047-1054. scription through an interaction with CtBP.
Soc Hematol Educ Program. 2019;2019 25. Radich JP, Dai H, Mao M, et al. Gene J Biol Chem. 2007;282(20):15248-15257.
(1):433-442. expression changes associated with pro- 42. Kazenwadel J, Secker GA, Liu YJ, et al.
10. Ko TK, Javed A, Lee KL, et al. An integra- gression and response in chronic myeloid Loss-of-function germline GATA2 muta-
tive model of pathway convergence in leukemia. Proc Natl Acad Sci U S A. tions in patients with MDS/AML or
genetically heterogeneous blast crisis 2006;103(8):2794-2799. MonoMAC syndrome and primary lym-
chronic myeloid leukemia. Blood. 2020;135 26. Lim S, Saw TY, Zhang M, et al. Targeting of phedema reveal a key role for GATA2 in
(26):2337-2353. the MNK-eIF4E axis in blast crisis chronic the lymphatic vasculature. Blood.
11. McWeeney SK, Pemberton LC, Loriaux myeloid leukemia inhibits leukemia stem 2012;119(5):1283-1291.
MM, et al. A gene expression signature of cell function. Proc Natl Acad Sci U S A. 43. Zhang SJ, Shi JY, Li JY. GATA-2 L359V
CD34+ cells to predict major cytogenetic 2013;110(25):E2298-2307. mutation is exclusively associated with
response in chronic-phase chronic myeloid 27. Crossman LC, Mori M, Hsieh YC, et al. In CML progression but not other hematolog-
leukemia patients treated with imatinib. chronic myeloid leukemia white cells from ical malignancies and GATA-2 P250A is a
Blood. 2010;115(2):315-325. cytogenetic responders and non-responders novel single nucleotide polymorphism.
12. Mohty M, Yong AS, Szydlo RM, Apperley to imatinib have very similar gene expres- Leuk Res. 2009;33(8):1141-1143.
JF, Melo JV. The polycomb group BMI1 sion signatures. Haematologica. 2005;90(4): 44. Pronier E, Almire C, Mokrani H, et al.
gene is a molecular marker for predicting 459-464. Inhibition of TET2-mediated conversion of
prognosis of chronic myeloid leukemia. 28. Burguillo FJ, Martin J, Barrera I, Bardsley 5-methylcytosine to 5-hydroxymethylcy-
Blood. 2007;110(1):380-383. WG. Meta-analysis of microarray data: the tosine disturbs erythroid and granu-
13. Yong AS, Szydlo RM, Goldman JM, case of imatinib resistance in chronic myel- lomonocytic differentiation of human
Apperley JF, Melo JV. Molecular profiling of ogenous leukemia. Comput Biol Chem. hematopoietic progenitors. Blood. 2011;
CD34+ cells identifies low expression of 2010;34(3):184-192. 118(9):2551-2555.
CD7, along with high expression of pro- 29. Jamieson CH, Ailles LE, Dylla SJ, et al. 45. Moran-Crusio K, Reavie L, Shih A, et al.
teinase 3 or elastase, as predictors of longer Granulocyte-macrophage progenitors as Tet2 loss leads to increased hematopoietic
survival in patients with CML. Blood. candidate leukemic stem cells in blast-crisis stem cell self-renewal and myeloid trans-
2006;107(1):205-212. CML. N Engl J Med. 2004;351(7):657-667. formation. Cancer Cell. 2011;20(1):11-24.
14. Branford S, Wang P, Yeung DT, et al. 30. Lim B, Lin Y, Navin N. Advancing cancer 46. Mayle A, Yang L, Rodriguez B, et al.
Integrative genomic analysis reveals cancer- research and medicine with single-cell Dnmt3a loss predisposes murine
associated mutations at diagnosis of CML genomics. Cancer Cell. 2020;37(4):456-470. hematopoietic stem cells to malignant
in patients with high-risk disease. Blood. 31. Warfvinge R, Geironson L, Sommarin transformation. Blood. 2015;125(4):629-
2018;132(9):948-961. MNE, et al. Single-cell molecular analysis 638.
15. Malone ER, Oliva M, Sabatini PJB, Stockley defines therapy response and immunophe- 47. Hervouet E, Peixoto P, Delage-Mourroux R,
TL, Siu LL. Molecular profiling for precision notype of stem cell subpopulations in Boyer-Guittaut M, Cartron PF. Specific or
cancer therapies. Genome Med. 2020;12 CML. Blood. 2017;129(17):2384-2394. not specific recruitment of DNMTs for
(1):8. 32. Giustacchini A, Thongjuea S, Barkas N, et DNA methylation, an epigenetic dilemma.
16. Kaneta Y, Kagami Y, Katagiri T, et al. al. Single-cell transcriptomics uncovers dis- Clin Epigenetics. 2018;10:17.
Prediction of sensitivity to STI571 among tinct molecular signatures of stem cells in 48. Huang Y-H, Tovy A, Sundaramurthy V, et
chronic myeloid leukemia patients by chronic myeloid leukemia. Nat Med. al. Nearly a third of clonal hematopoiesis-

V. Krishnan et al.
associated DNMT3A mutations reduce evolutionary dynamics of clonal Genet. 2018;56(3):149-175.

protein stability and may be associated hematopoiesis. Leukemia. 2018;32(9):1908- 78. Heller G, Topakian T, Altenberger C, et al.
with poorer prognosis. Blood. 2018;132 1919. Next-generation sequencing identifies
(Suppl 1):1315. 63. Welner RS, Amabile G, Bararia D, et al. major DNA methylation changes during
49. Adnan Awad S, Dufva O, Ianevski A, et al. Treatment of chronic myelogenous progression of Ph+ chronic myeloid
RUNX1 mutations in blast-phase chronic leukemia by blocking cytokine alterations leukemia. Leukemia. 2016;30(9):1861-1868.
myeloid leukemia associate with distinct found in normal stem and progenitor cells. 79. Mancini M, Veljkovic N, Leo E, et al.
phenotypes, transcriptional profiles, and Cancer Cell. 2015;27(5):671-681. Cytoplasmatic compartmentalization by
drug responses. Leukemia. 2021;35(4): 64. Hughes A, Yong ASM. Immune effector Bcr-Abl promotes TET2 loss-of-function in
1087-1099. recovery in chronic myeloid leukemia and chronic myeloid leukemia. J Cell Biochem.
50. Warsch W, Kollmann K, Eckelhart E, et al. treatment-free remission. Front Immunol. 2012;113(8):2765-2774.
High STAT5 levels mediate imatinib resist- 2017;8:469. 80. Amabile G, Di Ruscio A, Muller F, et al.
ance and indicate disease progression in 65. Hughes A, Clarson J, Tang C, et al. CML Dissecting the role of aberrant DNA
chronic myeloid leukemia. Blood. 2011;117 patients with deep molecular responses to methylation in human leukaemia. Nat
(12):3409-3420. TKI have restored immune effectors and Commun. 2015;6:7091.
51. Beer PA, Knapp DJ, Miller PH, et al. decreased PD-1 and immune suppressors. 81. Issa JP, Gharibyan V, Cortes J, et al. Phase II
Disruption of IKAROS activity in primitive Blood. 2017;129(9):1166-1176. study of low-dose decitabine in patients
chronic-phase CML cells mimics myeloid 66. Yong AS, Keyvanfar K, Hensel N, et al. with chronic myelogenous leukemia resist-
disease progression. Blood. 2015;125(3): Primitive quiescent CD34+ cells in chronic ant to imatinib mesylate. J Clin Oncol.
504-515. myeloid leukemia are targeted by in vitro 2005;23(17):3948-3956.
52. Thomson DW, Shahrin NH, Wang PPS, et expanded natural killer cells, which are 82. Visvader JE. Cells of origin in cancer.
al. Aberrant RAG-mediated recombination functionally enhanced by bortezomib. Nature. 2011;469(7330):314-322.
contributes to multiple structural rearrange- Blood. 2009;113(4):875-882. 83. Huntly BJ, Shigematsu H, Deguchi K, et al.
ments in lymphoid blast crisis of chronic 67. Koschmieder S, Vetrie D. Epigenetic dys- MOZ-TIF2, but not BCR-ABL, confers
myeloid leukemia. Leukemia. 2020;34(8): regulation in chronic myeloid leukaemia: a properties of leukemic stem cells to com-
2051-2063. myriad of mechanisms and therapeutic mitted murine hematopoietic progenitors.
53. Giotopoulos G, van der Weyden L, Osaki options. Semin Cancer Biol. 2018;51:180- Cancer Cell. 2004;6(6):587-596.
H, et al. A novel mouse model identifies 197. 84. Bose S, Deininger M, Gora-Tybor J,
cooperating mutations and therapeutic tar- 68. Sparmann A, van Lohuizen M. Polycomb Goldman JM, Melo JV. The presence of typ-
gets critical for chronic myeloid leukemia silencers control cell fate, development and ical and atypical BCR-ABL fusion genes in
progression. J Exp Med. 2015;212(10):1551- cancer. Nat Rev Cancer. 2006;6(11):846-856. leukocytes of normal individuals: biologic
1569. 69. Katoh M. Functional and cancer genomics significance and implications for the assess-
54. Yang H, Kurtenbach S, Guo Y, et al. Gain of of ASXL family members. Br J Cancer. ment of minimal residual disease. Blood.
function of ASXL1 truncating protein in the 2013;109(2):299-306. 1998;92(9):3362-3367.
pathogenesis of myeloid malignancies. 70. Wong SJ, Gearhart MD, Taylor AB, et al. 85. Sparano JA, Gray RJ, Makower DF, et al.
Blood. 2018;131(3):328-341. KDM2B recruitment of the polycomb Adjuvant chemotherapy guided by a 21-
55. Jaiswal S, Ebert BL. Clonal hematopoiesis group complex, PRC1.1, requires coopera- gene expression assay in breast cancer. N
in human aging and disease. Science. 2019; tion between PCGF1 and BCORL1. Engl J Med. 2018;379(2):111-121.
366(6465):eaan4673. Structure. 2016;24(10):1795-1801. 86. Ng SW, Mitchell A, Kennedy JA, et al. A 17-
56. Schmidt M, Rinke J, Schafer V, et al. 71. Yu M, Mazor T, Huang H, et al. Direct gene stemness score for rapid determina-
Molecular-defined clonal evolution in recruitment of polycomb repressive com- tion of risk in acute leukaemia. Nature.
patients with chronic myeloid leukemia plex 1 to chromatin by core binding tran- 2016;540(7633):433-437.
independent of the BCR-ABL status. scription factors. Mol Cell. 2012;45(3):330- 87. Bill M, Nicolet D, Kohlschmidt J, et al.
Leukemia. 2014;28(12):2292-2299. 343. Mutations associated with a 17-gene
57. Kim T, Tyndel MS, Kim HJ, et al. Spectrum 72. Oravecz A, Apostolov A, Polak K, et al. leukemia stem cell score and the score's
of somatic mutation dynamics in chronic Ikaros mediates gene silencing in T cells prognostic relevance in the context of the
myeloid leukemia following tyrosine through Polycomb repressive complex 2. European LeukemiaNet classification of
kinase inhibitor therapy. Blood. 2017;129 Nat Commun. 2015;6:8823. acute myeloid leukemia. Haematologica.
(1):38-47. 73. Xie H, Peng C, Huang J, et al. Chronic 2020;105(3):721-729.
58. Jeong M, Sun D, Luo M, et al. Large con- myelogenous leukemia-initiating cells 88. Shanmuganathan N, Pagani IS, Ross DM, et
served domains of low DNA methylation require polycomb group protein EZH2. al. Early BCR-ABL1 kinetics are predictive
maintained by Dnmt3a. Nat Genet. Cancer Discov. 2016;6(11):1237-1247. of subsequent achievement of treatment-
2014;46(1):17-23. 74. Scott MT, Korfi K, Saffrey P, et al. free remission in chronic myeloid
59. Abdel-Wahab O, Gao J, Adli M, et al. Epigenetic reprogramming sensitizes CML leukemia. Blood. 2021;137(9):1196-1207.
Deletion of Asxl1 results in myelodysplasia stem cells to combined EZH2 and tyrosine 89. Radich JP, Larson R, Kantarjian H, et al.
and severe developmental defects in vivo. J kinase inhibition. Cancer Discov. Gene expression signature predicts deep
Exp Med. 2013;210(12):2641-2659. 2016;6(11):1248-1257. molecular response (DMR) in chronic
60. Sano S, Oshima K, Wang Y, et al. CRISPR- 75. Rizo A, Horton SJ, Olthof S, et al. BMI1 col- myeloid leukemia (CML): an exploratory
mediated gene editing to assess the roles of laborates with BCR-ABL in leukemic trans- biomarker analysis from ENESTnd
Tet2 and Dnmt3a in clonal hematopoiesis formation of human CD34+ cells. Blood. [Abstract]. Blood. 2019;34;(Suppl_1):665.
and cardiovascular disease. Circ Res. 2010;116(22):4621-4630. 90. Kwa M, Makris A, Esteva FJ. Clinical utility
2018;123(3):335-341. 76. Sengupta A, Ficker AM, Dunn SK, Madhu of gene-expression signatures in early stage
61. Sano S, Oshima K, Wang Y, et al. Tet2- M, Cancelas JA. Bmi1 reprograms CML B- breast cancer. Nat Rev Clin Oncol.
mediated clonal hematopoiesis accelerates lymphoid progenitors to become B-ALL- 2017;14(10):595-610.
heart failure through a mechanism involv- initiating cells. Blood. 2012;119(2):494-502. 91. Teutsch SM, Bradley LA, Palomaki GE, et al.
ing the IL-1beta/NLRP3 inflammasome. J 77. Behzad MM, Shahrabi S, Jaseb K, et al. The Evaluation of Genomic Applications in
Am Coll Cardiol. 2018;71(8):875-886. Aberrant DNA methylation in chronic Practice and Prevention (EGAPP) Initiative:
62. Arends CM, Galan-Sousa J, Hoyer K, et al. myeloid leukemia: cell fate control, progno- methods of the EGAPP Working Group.
Hematopoietic lineage distribution and sis, and therapeutic response. Biochem Genet Med. 2009;11(1):3-14.

Acute Lymphoblastic Leukemia ARTICLE
Comprehensive analysis of dose intensity of Ferrata Storti Foundation

acute lymphoblastic leukemia chemotherapy
Seth E. Karol,1,2 Deqing Pei,3 Colton A. Smith,1 Yiwei Liu,1 Wenjian Yang,1 Nancy
M. Kornegay,1 John C. Panetta,1 Kristine R. Crews,1 Cheng Cheng,3 Emily R.
Finch,1 Hiroto Inaba,2 Monika L. Metzger,2,4 Jeffrey E. Rubnitz,2 Raul C. Ribeiro,2
Tanja A. Gruber,5,6 Jun J. Yang,1 William E. Evans,1 Sima Jeha,2,4 Ching-Hon Pui2
and Mary V. Relling1
Department of Pharmaceutical Sciences, St. Jude Children’s Research Hospital,
1
Memphis, TN; 2Department of Oncology, St. Jude Children’s Research Hospital, Haematologica 2022
Memphis, TN; 3Department of Biostatistics, St. Jude Children’s Research Hospital,
Memphis, TN; 4Department of Global Pediatric Medicine, St. Jude Children’s Research
Volume 107(2):371-380
Hospital, Memphis, TN; 5Department of Pediatrics, Stanford University School of
Medicine, Stanford, CA, and 6Stanford Cancer Institute, Stanford University School of
Medicine, Stanford, CA, USA
ABSTRACT
C
hemotherapy dosages are often compromised, but most reports
lack data on dosages that are actually delivered. In two consecutive
acute lymphoblastic leukemia trials that differed in their asparagi-
nase formulation, native E. coli L-asparaginase in St. Jude Total 15 (T15,
n=365) and pegaspargase in Total 16 (T16, n=524), we tallied the dose
intensities for all drugs on the low-risk or standard-risk arms, analyzing
504,039 dosing records. The median dose intensity for each drug ranged
from 61-100%. Dose intensities for several drugs were more than 10%
higher on T15 than on T16: cyclophosphamide (P<0.0001 for the stan-
dard-risk arm), cytarabine (P<0.0001 for the standard-risk arm), and mer-
captopurine (P<0.0001 for the low-risk arm and P<0.0001 for the standard-
risk arm). We attributed the lower dosages on T16 to the higher asparagi-
nase dosages on T16 than on T15 (P<0.0001 for both the low-risk and
standard-risk arms), with higher dose-intensity for mercaptopurine in
those with anti-asparaginase antibodies than in those without (P=5.62x10-
3
for T15 standard risk and P=1.43x10-4 for T16 standard risk). Neutrophil
count did not differ between protocols for low-risk patients (P=0.18) and Correspondence:
was actually lower for standard-risk patients on T16 than on T15
MARY V. RELLING
(P<0.0001) despite lower dosages of most drugs on T16. Patients with low mary.relling@stjude.org
asparaginase dose intensity had higher methotrexate dose intensity with
no impact on prognosis. The only dose intensity measure predicting a
higher risk of relapse on both studies was higher mercaptopurine dose Received: January 21, 2021.
intensity, but this did not reach statistical significance (P=0.03 T15; P=0.07 Accepted: April 28, 2021.
T16). In these intensive multiagent trials, higher dosages of asparaginase Pre-published: July 1, 2021.
compromised the dosing of other drugs for acute lymphoblastic leukemia,
particularly mercaptopurine, but lower chemotherapy dose intensity was
not associated with relapse. https://doi.org/10.3324/haematol.2021.278411

Introduction Material published in Haematologica is covered by copyright.
All rights are reserved to the Ferrata Storti Foundation. Use of
Dosages of chemotherapy drugs used in treatment regimens for acute lymphoblas- conditions:
tic leukemia (ALL) vary widely.1-8 There is, however, a lack of data comparing admin- https://creativecommons.org/licenses/by-nc/4.0/legalcode.
istered dosages to planned dosages, and the last comprehensive analysis (for only Copies of published material are allowed for personal or inter-
209 patients) was published in 1991.9 Hence, it is difficult to compare feasibility of nal use. Sharing published material for non-commercial pur-
protocol delivery among cooperative treatment groups, between adults and children, https://creativecommons.org/licenses/by-nc/4.0/legalcode,
and across different countries. Most ALL drugs can cause myelosuppression, and sect. 3. Reproducing and sharing published material for com-
thus comparisons of administered dose intensity across protocols that are limited to mercial purposes is not allowed without permission in writing
a single agent, e.g., mercaptopurine or asparaginase, may be misleading if the from the publisher.
dosages of other possibly “compensating” agents are not also accounted for across
protocols. As newer immune-based and less myelosuppressive agents are added to

S.E. Karol et al.
existing ALL regimens,10 predicting the tolerability of new report focuses on the ~90% of patients who were treated on
combination protocols has proven challenging without either the LR or SR arms. Ninety patients on T15 who received
accurate information on how current conventional ALL most of their therapy at a collaborating institution11 were also
therapy is actually administered. excluded, because their drug administration data were difficult
Our aim was to tabulate the actual dosages and dose to verify (Online Supplementary Figure S2). The studies were
intensities of conventional chemotherapy in two consecu- approved by the Institutional Review Board. Informed consent
tive multiagent front-line pediatric St. Jude ALL trials that was obtained from either the parents or the patients, consistent
used highly similar backbones but different formulations of with the Declaration of Helsinki.
asparaginase: E. coli L-asparaginase in Total 15 (T15) and
pegaspargase in Total 16 (T16). Treatment
Drug administration data were recorded prospectively on pro-
tocol-specific forms, generally on a daily basis for induction and
Methods reinduction and on a weekly basis for other phases, by clinical
staff and research nurses, and entered into centralized St. Jude
Patients databases by protocol-specific research data managers. Reasons
Patients were enrolled on St. Jude Children’s Research for dose modifications were protocol-specified (Online
Hospital protocols T15 (ClinicalTrials.gov ID: NCT00137111)11 Supplement).
and T16 (ClinicalTrials.gov ID: NCT00549848)12 for newly diag- Treatment regimens have been described previously11,12 and
nosed ALL (Online Supplementary Figure S1). T15 and T16 thera- are summarized in Online Supplementary Figure S1 and Online
py included remission induction therapy followed by consolida- Supplementary Table S1. The two protocols differed primarily by
tion therapy and 120 weeks of continuation therapy (146 weeks first-line asparaginase formulation, although there were a few
for boys on T15) which included two phases of reinduction other differences (Online Supplementary Figure S1, Online
(Online Supplementary Table S1). After remission induction, Supplementary Tables S1 and S2).11 Patients received E. coli
patients were classified for risk-adapted therapy as low-risk (LR, asparaginase (Elspar) in T15, and pegaspargase (Oncaspar) in
about 40% of patients), standard-risk (SR, about 50% of T16. In the case of allergic reaction to either E. coli asparaginase
patients) or high-risk (HR, about 10% of patients).11,12 Patients or pegaspargase, an asparaginase formulation change was per-
with Down syndrome received altered methotrexate and leu- mitted by the protocol (Online Supplement). To compare planned
covorin regimens, and HR patients received intensification phas- dosages roughly between protocols, Elspar doses were convert-
es, and thus both groups were excluded from this analysis. This ed to comparable doses of pegaspargase based on protocol-spec-
Figure 1. Planned cumulative doses of drugs used in the low- and standard-risk arms of T15 and T16. The only drugs with at least a 10% difference in planned
cumulative dosages between T16 and T15 (within risk arms) were asparaginase (34% higher in the standard risk arm), dexamethasone (17.4% lower in the stan-
dard-risk arm), and doxorubicin (50% less in the low-risk arm). Thus, only asparaginase was planned for higher dosages on T16 than on T15. There were no differ-
ences between T15 and T16 in planned cumulative dosages of prednisone, vincristine, daunorubicin, cytarabine, cyclophosphamide, mercaptopurine; there was a
planned decrease of only 6.0% in low-dose methotrexate dosages in the standard-risk arm on T16 versus T15. See Online Supplementary Table S2 for details on
planned dosages.

Dose intensity in ALL treatment
ified conversions (i.e., pegaspargase 2,500 units/m2 were consid- Results

ered comparable to 50,000 units/m2 of Elspar and 100,000
units/m2 of Erwinia asparaginase). Total planned dosages were Differences in dosage intensity and delivered dosages
identical for T15 LR and T16 LR as well as for T15 SR and T16 between the Total 15 and Total 16 trials
SR for consolidation high-dose methotrexate, cyclophos- The median cumulative dose intensity for each drug
phamide, cytarabine, daunorubicin, continuation mercapto- ranged from 61% to 100%, with the largest interpatient
purine, and vincristine; they were higher on T16 than on T15 for variability observed for mercaptopurine and methotrexate.
asparaginase, and were lower on T16 than on T15 for dexam- Post-induction, dose intensity median absolute deviations
ethasone (both LR and SR arms) and for doxorubicin (LR arms; ranged from 16% to 23% for mercaptopurine (Online
Figure 1, Online Supplementary Table S2). In Total therapy proto- Supplementary Table S3). The cumulative dose intensities for
cols, particularly for the 120 weeks of continuation, dosages of several drugs were higher on T15 than on T16 (Figure 2,
therapy were adjusted or omitted if they caused toxicity (partic- Online Supplementary Table S3), and median cumulative
ularly myelosuppression), as detailed (Online Supplementary Table dose intensities were over 10% higher on T15 than on T16
S2), but therapy was not typically delayed. for cyclophosphamide (P<0.0001 for SR patients), cytara-
Dose intensities were calculated as the delivered dosage divid- bine (P<0.0001 for SR patients), mercaptopurine (P<0.0001
ed by the protocol-specified dosage for each risk group and pro- for LR and SR patients), and consolidation methotrexate
tocol. Dose intensity was tabulated per phase, and cumulative (P<0.0001 for LR patients). Examining phases instead of
dose intensity was estimated as the total prescribed dosage per cumulative dose intensity, mercaptopurine exhibited the
patient divided by the total cumulative protocol-specified lowest dose intensities of any drug in all protocols and
dosage (Online Supplement). Absolute neutrophil count (ANC) arms, and its dose intensity was higher on T15 than on T16
was generally checked at least weekly throughout treatment for all phases (P<0.0001 for all) except induction (Online
(Online Supplement). Time to complete therapy was the time Supplementary Table S4). The dose intensity of cyclophos-
from start of first chemotherapy in induction to last dose of phamide was higher on T15 than on T16 during the contin-
chemotherapy at 120 weeks. uation phase (P<0.0001), while that of cytarabine was high-
er at reinduction II and during continuation (P<0.0001 for
Additional methods both phases; Online Supplementary Table S5 and S6).
Details of genetic ancestry assessment, TPMT phenotype Tolerated dosages of mercaptopurine were lower on T16
assignment, and statistical methods are available in the Online than they had been on T15, with as much as 27% higher
Supplement. dose intensity on T15 than on T16 (Figure 3, Online
Figure 2. Administered cumulative dose intensities for drugs on T15 versus T16, by risk arm. The only significant differences (***P<0.0001) in dose intensities
with more than a 10% difference between protocols were for cyclophosphamide and cytarabine (standard-risk arms) and for mercaptopurine (both risk arms), all of
which were higher on T15 than on T16 (see Online Supplementary Table S3 for details on all drugs). Bars and whiskers indicate medians and median absolute devi-
ations among each patient population. There was a total of 16 statistical comparisons, thus the Bonferroni significance threshold=0.003.

S.E. Karol et al.
Figure 3. Thiopurine dose intensity

by phase in T15 versus T16. The box
and whisker plots show the quartiles
and nonoutlier ranges. Dose intensi-
ty was significantly lower on T16
than on T15 (P<0.0001) for all phas-
es except induction (see Online
Supplementary Table S4 for exact P
values). The largest difference was
27% during continuation weeks 10-
16 in standard-risk patients. There
was a total of 10 statistical compar-
isons, thus the Bonferroni signifi-
cance threshold=0.005.
Supplementary Table S4). The median cumulative dosage of tive dosage was higher on T15 than on T16, reflecting the
mercaptopurine delivered was 7,420 mg/m2 higher on the fact that the planned dosage (i.e., the denominator) was
LR arm and 9,130 mg/m2 higher on the SR arm of T15 com- lower on T16 than on T15.
pared to T16 (Table 1). Despite these markedly higher For most drugs on each of the LR and SR arms, the
dosages on T15 (Table 1), ANC did not differ between the planned dosages were identical on T15 and T16 (Figure 1,
protocols for the LR arm (P=0.18) (Figure 4), and were actu- Online Supplementary Table S2); the drug with the largest
ally lower on T16 than on T15 for the SR arm (P<0.0001), planned differences in dosages between the two protocols
indicating that our clinicians were not titrating to a higher was asparaginase, with an 8% planned increase in the LR
ANC on T16 than they had on T15. Interestingly, the only arms and a 34% increase in the SR arms on T16 compared
drug for which dose intensity was significantly higher on to T15. In actuality, asparaginase was the only drug for
T16 than on T15 was dexamethasone (P<0.0001) (Online which the planned or administered dosages were higher on
Supplementary Table S3), although the administered cumula- T16 than on T15, with higher cumulative dosages given on

Table 1. Actual cumulative dosages administered for drugs on T15 and T16 by risk group.
T15 T16 T15
Drug Risk N Median 5th-%tile 95th-%tile CV% N Median 5th-%tile 95th-%tile CV% minus
(units) arm P Value T16
Median n
Asparaginase LR 192 12000 8500 15500 6% 254 13100 10600 17200 3% <0.0001 -1100
(U/m2) SR 173 26600 15700 40900 7% 270 43300 12700 58700 27% <0.0001 -16700
Cyclophosphamide LR 190 1000 917 1040 <1% 252 1000 969 1040 <1% 1 0
(mg/m2) SR 172 4440 1190 4650 6% 269 3830 2080 4670 25% <0.0001 613
Cytarabine LR 190 599 227 636 2% 253 600 295 638 4% 1 -1.03
(mg/m2) SR 172 11300 3550 12300 13% 269 9460 5510 10400 10% <0.0001 1830
Daunorubicin LR 192 50 25 51.9 2% 254 49.4 24.7 50.9 2% 1 0.58
(mg/m2) SR 173 50.3 38.4 52.3 2% 270 49.5 25 50.8 2% 2.07 x10-4 0.85
Doxorubicin LR 189 60 57.7 60.9 1% 250 30 29.2 30.9 1% <0.0001 30
(mg/m2) SR 171 179 120 183 2% 265 162 118 182 16% 0.22 17.7
Vincristine LR 192 60.8 38.9 66.1 8% 253 62.1 38.6 66.8 7% 0.33 -1.3
(mg/m2) SR 173 52.2 19.8 67.2 34% 269 55.3 26 67.9 29% 0.33 -3.1
Thiopurine.Induction LR 190 817 260 936 9% 253 813 354 915 11% 1 4.1
(mg/m2) SR 172 813 354 918 10% 269 796 376 886 12% 0.52 16.5
ConsolHDMTX LR 190 11000 7410 14800 19% 250 9970 8990 10300 2% <0.0001 1020
(mg/m2) SR 171 18500 11100 23400 12% 266 18400 13000 20900 10% 0.92 94.7
Dexamethasone LR 190 1060 586 1160 9% 251 1020 481 1110 8% 6.18 x10-4 37
(mg/m2) SR 171 1330 287 1580 22% 265 1260 365 1400 13% 1.29 x10-4 74.2
MP Consol_to_Wk120 LR 191 51400 29600 69000 21% 251 43900 22700 61400 26% <0.0001 7420
(mg/m2) SR 171 37100 9340 48600 23% 266 28000 8000 42200 40% <0.0001 9130
MTX Cont_to_Wk120 LR 190 3250 2210 4600 15% 251 2980 1840 3960 17% <0.0001 273
(mg/m2) SR 165 2440 1280 3440 17% 261 2090 854 3070 24% <0.0001 347
Drugs excluded were prednisone from the induction phase (because of variability in formulations and documentation) and mercaptopurine and methotrexate for boys in T15 from week 120-
146 (which were not present in T16). The CV% is calculated as median absolute deviation divided by median dosage. Total of 11(drugs)x2(risk arms)=22 comparisons. Bonferroni significance
threshold=0.002. T15: Total therapy 15 trial, T16: Total therapy 16 trial; Consol: consolidation; cont: continuation; HDMTX: high-dose methotrexate; CV%: percentage coefficient of variation; MP: mer-
captopurine; MTX: methotrexate; wk: week.
T16 than on T15 for both the LR and SR arms (P<0.0001 for Supplementary Table S7). Importantly, using this prospective
both) (Table 1). The higher exposure was due to the precision medicine approach of genetically-driven prescrib-
planned higher dosages on T16 than on T15, rather than to ing prevented excessive thiopurine-induced cytopenias. As
a change in prescribing practices, in that the dose intensities a result, neither the dose intensities of other drugs (anthra-
for asparaginase did not differ on T16 versus T15 for either cyclines, asparaginase, methotrexate, cyclophosphamide,
the LR (P=1) or the SR (P=0.77) arms (Online Supplementary cytarabine, and dexamethasone) (Online Supplementary
Table S3). Table S7) nor the ANC (Figure 6) differed (or differed only
The dose intensity for mercaptopurine was higher in marginally) by TPMT status, demonstrating that pinpoint-
those who had antibodies against asparaginase (against ing the correct drug for dosage adjustments prevented com-
Elspar for T15 and against pegaspargase for T16) than in promising the dosages of other chemotherapeutic agents.
those who did not (P=0.12 for T15 LR, P=0.0056 for T15 When mercaptopurine dose intensity was re-estimated
SR, P=0.00027 for T16 LR, and P=0.00014 for T16 SR) using a TPMT-specific denominator for expected mercap-
(Figure 5). The higher dose intensity was likely in response topurine dosages (see the Online Supplement for details), the
to a higher ANC in those with antibodies compared to dose intensity for mercaptopurine showed a much smaller
those without antibodies, with the difference only reaching difference between those with and those without a defect
nominal statistical significance for T15 SR (P=0.032) (Figure in TPMT (Online Supplementary Figure S4) than when the
5). denominator was not adjusted downward for the expected
decrease in dosage due to TPMT status (Online
Differences in dose intensity by inherited genetics and Supplementary Figure S5).
other characteristics of the patients (age, sex, race) We compared dose intensity for all drugs by sex, ances-
The only dosage modification based on a pharmacoge- tral group (white, Black, Hispanic, and other), and age,
netic characteristic for both protocols was that for thio- adjusting for protocol and risk arm. Differences in dose
purines based on TPMT status (Online Supplementary Table intensity by race and by sex were relatively modest (data not
S2), and this modification was made prospectively. Thus, as shown), particularly compared to the differences by protocol
expected, the median dose intensity was lower for those or risk group. There were a few agents for which dose
with a TPMT abnormality (poor or intermediate metaboliz- intensity differed by age (Online Supplementary Figure S6).
ers) than for those without a defective TPMT allele (normal For T15 SR, T16 SR, and T16 LR, the dose intensities for
metabolizers): 0.73 versus 0.83 for T15 SR, 0.68 versus 0.84 dexamethasone and vincristine were significantly inversely
for T15 LR, 0.48 versus 0.63 for T16 SR, and 0.60 versus 0.75 correlated to age and the dose intensity for methotrexate
for T16 LR (P=0.001, P=4.3x10-4, P<0.0001, and P=1.92x10- was positively correlated to age; for T15 LR, only the dose
4
, respectively) (Online Supplementary Figure S3, Online intensity of vincristine was inversely correlated with age,

S.E. Karol et al.
Figure 4. Absolute neutrophil count by risk group on T15 versus T16. The graphs show the average of fitted absolute neutrophil count (ANC; cells/mm3) data per
phase per patient with thick solid lines representing the median per risk group. Based on 46,310 and 64,549 ANC records for T15 and T16, respectively.
but the effect did not reach statistical significance after cor- worse outcomes was also true when including patients
rection for multiple testing. who did not complete 120 weeks of therapy (Online
Supplementary Table S12). Associations between ANC and
Differences in dose intensity between risk arms outcomes were not reproducible across phases or protocols.
The most striking difference in dose intensity between Notably, there was no association between dose intensities
risk arms was for mercaptopurine (Figure 2). The median of any other medications, including asparaginase (Online
dose intensity was similar between risk groups for most Supplementary Figure S8) and outcomes; however, it should
phases on T15 (Online Supplementary Table S8) including the be noted that there was also little interpatient variability in
phase immediately after Elspar asparaginase ended (weeks the dose intensity for asparaginase (median absolute devia-
20-47, P=0.1). In contrast, on T16, in the weeks immediate- tions, 1% to 7%; Online Supplementary Table S3).
ly following reinduction and the completion of pegaspar-
gase (which ended week 29), the median dose intensity Absolute neutrophil count versus dose intensity
was 41% versus 69% for the SR versus the LR arms For each drug, protocol, and risk arm, we analyzed
(P<0.0001) (Online Supplementary Table S9). The cumulative whether dose intensity was related to ANC for each phase.
dose intensities for all the drugs by protocol and risk arm The strongest associations were for mercaptopurine; in all
are shown in Online Supplementary Tables S10 and S11. instances with nominal associations, ANC and dose inten-
sity were positively correlated (Online Supplementary Table
Relapse S13). This reflects that clinicians followed protocol recom-
We explored relationships between dose intensities for all mendations (Online Supplementary Table S1) to increase the
drugs during all phases and cumulatively, and ANC for all mercaptopurine dosage in those with high ANC, and to
phases and cumulatively, and treatment outcomes. No sig- decrease the dosage for those with low ANC. For all asso-
nificant associations between dose intensity or ANC and ciations between dose intensity of other drugs and ANC,
outcomes were observed after adjusting for multiple test- the correlations were also positive, with a few exceptions;
ing. Only one dose intensity or ANC variable was nominal- dexamethasone dose intensity was inversely correlated
ly associated with outcome in the same direction for both with ANC in some phases (although not statistically signif-
protocols in both continuous and tertile analyses: for those icantly after corrections for multiple testing; data not shown).
who finished 120 weeks of therapy: a higher dose intensity
for mercaptopurine was nominally associated with a higher Dose intensity versus time on therapy
risk of any relapse (unadjusted P=0.03 for T15, P=0.07 for Although the practice was to avoid delays in therapy,
T16, tertile analysis) (Online Supplementary Table S12, Online there was observed variability in the time required to com-
Supplementary Figure S7). For T15, this tendency for an asso- plete all therapy up to week 120 of continuation (median
ciation between higher mercaptopurine dose intensity and time to complete therapy including induction, consolida-

tion, and continuation was 135.5 weeks for both T15 and
T16; interquartile range, 134.4-137.3 weeks). To assess the
possible impact of delivered dose intensity on time to com-
plete therapy, we analyzed the association between the
dose intensity of asparaginase and mercaptopurine and
time to complete therapy in patients completing treatment.
We found an inverse association between dose intensity
and time to complete therapy (P<0.0001 for both drugs),
however, the correlation was relatively weak (r2 = 0.03 for
both asparaginase and mercaptopurine). More importantly,
there was no association between time to complete therapy
and relapse risk (P=0.7 on T15, P=0.4 on T16). Of the 889
patients included on the study, 17 (1.9%) discontinued
treatment early due to toxicity. With a median follow-up of
7.5 years from diagnosis for these patients, only one of 17
patients (5.9%) experienced a relapse (at 3.9 years after
diagnosis).
Discussion
This study is the first to comprehensively evaluate
dosages of all conventional drugs constituting modern ALL
therapy. Our most striking observation was that mercap-
topurine dose intensity was lower on T16 than on T15,
despite no planned changes to mercaptopurine dosages on
T16 versus T15. The most likely explanation for this
decreased ability to administer full dosages of mercaptop-
urine is higher planned and administered doses of asparag-
inase on T16 than on T15. We suggest that asparaginase
interfered with mercaptopurine delivery based on several
findings: an inverse association between delivered asparag-
inase and mercaptopurine dosages between protocols
(Figure 2, Table 1, Online Supplementary Table S4), a higher
delivered mercaptopurine dosage in those with antibodies
to asparaginase than in those without antibodies (Figure 5),
and a temporal decrease in dose intensity of mercaptop-
urine that corresponds with the timing of asparaginase use
(Online Supplementary Tables S8 and S9, Online
Supplementary Figures S4 and S5). Moreover, we and others
Figure 5. Mercaptopurine cumulative dose intensity and absolute neutrophil
have shown that asparaginase can decrease the clearance of count in patients negative or positive for anti-asparaginase antibodies.
other drugs, such as dexamethasone, putatively through its Mercaptopurine cumulative dose intensity (DI; left y axes) and absolute neu-
hypoproteinemic effects on hepatic drug metabolizing trophil count (ANC) in cells/mm3 (right y axes) for continuation weeks 10-16 in
patients who were negative or positive for anti-asparaginase antibodies against
enzymes and transporters. We found that this interaction is Elspar (T15) or Oncaspar (T16) measured at continuation week 7. Boxes and
associated with an increased risk of at least one adverse whiskers represent quartiles and non-outlier ranges. Nominal P values were
effect of dexamethasone (osteonecrosis) and higher drug *P<0.05; **P<0.01, ***P<0.001. There was a total of four comparisons, thus
the Bonferroni significance threshold=0.01
exposure both in the clinic13 and in preclinical models.14
Others have also hypothesized that asparaginase can influ-
ence the dose intensity or toxicity of thiopurines and/or current analysis, we did not find that asparaginase dose
methotrexate.15,16 intensity was related to disease free-survival in either T15
Asparaginase is an important part of ALL therapy, which (using primarily native E. coli asparaginase) or T16 (using
is the reason that we chose to increase exposure to asparag- primarily pegaspargase) (Online Supplementary Table S12,
inase on T16 compared to that on T15. Early discontinua- Online Supplementary Figure S8). There are several possible
tion of asparaginase was associated with lower event-free explanations for this finding. First, we used more asparagi-
survival in the Dana-Farber 91-01 study1 and in nase on our studies than others did, thus perhaps exceeding
ETV6/RUNX1 ALL,17 and patients treated with E. coli some threshold value for exposure, consistent with the lack
asparaginase had fewer relapses than those treated with of influence of pegaspargase dosage on relapse in T16.12
Erwinase.18 Patients with allergy to pegaspargase treated on Second, in those with allergy to their front-line asparagi-
recent Children’s Oncology Group protocols who did not nase preparation, substitution with another formulation
receive all asparaginase therapy had a lower disease-free was aggressive, such that asparaginase dose intensity was
survival, and drug shortages of Erwinase contributed to not not lower for those with or without allergy on T15 or on
being able to compensate fully for missed pegaspargase T16 (Online Supplementary Figure S9) and interpatient vari-
doses.19 However, prolonged asparaginase did not improve ability in dose intensity for asparaginase was low, i.e., <7%
outcomes in two trials,2,20 and higher doses of pegaspargase (Online Supplementary Table S3). Another possibility is that
did not improve outcome relative to standard doses.12 In the even in those patients with relatively low asparaginase dose

378
S.E. Karol et al.
Figure 6. Absolute neutrophil count according to TPMT status on the T15 and T16 trials. Absolute neutrophil count (ANC, cells/mm3) was similar regardless of TPMT status on the T15 (top) and T16 (bottom) trials. ANC is depicted for the
low-risk and standard-risk arms. The nominal P values are from the Wilcoxon rank sum test. Boxplots show the 25th and 75th percentiles; whiskers extend from 1.5 times the interquartile range, and data points outside the whiskers are
depicted as dots. There was a total of 28 statistical comparisons, thus the Bonferroni significance threshold=0.002.

intensity, the dosages of other medications were increased, there was a modest association between higher ANC and
and this could have compensated for the lower exposure to increased relapse, but this association was not statistically
asparaginase. In fact, we found that patients with low significant after correction for multiple testing, and was not
asparaginase dose intensity had significantly higher dose reproducible, as no such association was observed on T16,
intensity for methotrexate on T16 (Online Supplementary and in subanalyses for different therapy phases, higher
Figure S10), consistent with our protocol recommendations ANC tended to associate with both increased and
for substituting methotrexate in those who could not decreased relapse.
receive asparaginase due to allergy or pancreatitis. These In both our studies, thiopurine starting dose was adjusted
findings represent what is unique about the current study, based on TPMT status,24 which appropriately resulted in
because we have data to indicate how dosages of each lower dosages of mercaptopurine in those with a genetic
medication may have influenced each other. This is the sit- defect in TPMT; herein, we show for the first time that this
uation clinicians face: when pressing ahead with one med- allowed for uncompromised dosing of the other
ication is thwarted for some reason (drug shortages, chemotherapeutic agents (Online Supplementary Table S7)
adverse effects), other medications are often substituted. In and no difference in ANC based on TPMT status (Figure 6).
these trials, it appears that the substitutions made were It is also worth noting that the dose intensity for mercap-
effective in maintaining efficacy, in that there were no asso- topurine, especially in the first 6 months of therapy, was
ciations between low exposure to any one drug and out- lower than reported by others;23 after reinduction, the medi-
come. an dose intensity was as low as 49% for one phase on T16,
In our studies, as in other ALL studies, the primary drugs suggesting that our planned dosages may have been too
for which dosage adjustments are routinely made are mer- high (Online Supplementary Table S4). Interestingly, the
captopurine and methotrexate, which are titrated to a median dose intensity for the comparable time period on
desired ANC, and indeed the greatest variability in dosages T15, when native asparaginase rather than pegaspargase
was observed for these two drugs (Online Supplementary was used, was higher at 76%. Had we realized a priori the
Table S3). There is controversy as to whether increasing the impact of asparaginase on patients’ tolerance of thiopurine
dose intensity of mercaptopurine and/or methotrexate therapy, we could have designed a more realistic dosage
increases,21 decreases22 or has no effect16 on the risk of regimen, and this finding has implications for future proto-
relapse. Complicating the interpretation of the data is that col design.
most studies do not systematically assess adherence, and There are conflicting data on the importance of “inten-
most protocols use both drugs orally; thus, patients with sive” non-antimetabolite therapy in ALL.3,5-7,25-29 Although
the highest prescribed dose intensity may be the patients the lack of association of outcomes with dose intensity for
who are actually taking the lowest percentage of their ther- most drugs in our study is fairly consistent with studies
apy, thereby complicating any interpretation of how dose touting deintensification strategies, St. Jude differs from
intensity affects relapse. For example, those with lower many other centers in that patients receive all weekly
measured mercaptopurine adherence had higher relapse methotrexate parenterally (and thus return to the clinic
risk, but there was no association with mercaptopurine every week, in contrast to many centers that see patients
dose intensity.23 It is possible that low adherence to mer- only every 4-6 weeks during continuation), LR and SR
captopurine also translates into lower adherence with patients received vincristine/dexamethasone pulses
other drugs not measured (e.g., glucocorticoids, methotrex- throughout continuation, and a high percentage of patients
ate, supportive care drugs), thus compounding the risk of (~50%) received therapy on the SR and HR arms, which
relapse due to poor adherence. In the current analysis, include more asparaginase and other non-antimetabolite
dosages of all antileukemic drugs were captured, and the agents (cytarabine, cyclophosphamide) than many other
only drug whose prescribed dose intensity was associated treatment protocols. Thus, the lack of association between
with relapse in unadjusted analyses on both T15 and T16 dose intensity and relapse we observed may not be exten-
was mercaptopurine (Online Supplementary Table S12, sible to centers with less intensive monitoring and/or less
Online Supplementary Figure S7), albeit not in the same chemotherapy-dense and diverse regimens. It also suggests
phases of therapy for both studies. It should be noted that that, for patients intolerant to specific components of ther-
the direction of association was that higher dose intensity apy (e.g., asparaginase due to pancreatic or hepatic toxici-
was associated with higher relapse, indicating that it is ty), substitution with alternative chemotherapy may miti-
likely that high dose intensity identified patients who were gate the adverse prognosis associated with early discontin-
either noncompliant (and thus clinicians needed to push uation of the offending agent. Given that therapy-limiting
the dosage to achieve desired neutropenia) or had very fast toxicities may preclude delivery of prescribed chemothera-
drug clearance. However, it should be noted than no met- py in 12-25% of patients receiving treatment on modern tri-
rics of dose intensity or of ANC were significantly associ- als,19 prospective evaluation of chemotherapy substitution
ated with relapse or outcome using P value thresholds to address therapy-limiting toxicity should be considered.
adjusted for the large number of exploratory comparisons. We conclude that intentional changes to the dose intensi-
Although our data comprehensively capture prescribed ty of one agent, e.g., asparaginase, can have dramatic con-
mercaptopurine dose, one limitation of these data is that sequences on the ability to administer other conventional
we did not assess adherence to administration of pre- agents. Comprehensive data on chemotherapy actually
scribed drug in patients who were not under direct medical delivered in cancer clinical trials are needed to fully interpret
supervision, which applied to most of the oral doses of glu- results and further optimize therapy.
cocorticoids and mercaptopurine. Any association
between dose intensity and outcomes is likely to be affect- Disclosures
ed by the extent to which adherence is emphasized, the MVR and HI and St. Jude Children’s Research Hospital
extent of thiopurine monitoring, and the rigor with which receive investigator-initiated research funding from Servier
ANC targets are pursued. It should be noted that on T15, Pharmaceutical.

S.E. Karol et al.
Contributions Funding
SEK, WEE, CHP, SJ and MVR conceived and designed the This work was supported by National Institutes of Health
study; KRC, ARM, CC, DP, SEK, HI, JER, MLM, RCR, GM115279, CA35401, CA142665, CA21765, and
TAG, JJY, WEE, SJ, CHP, and MVR provided study material K08CA250418, and the American Lebanese Syrian Associated
or patients and collected and/or assembled data; NMK, YL, Charities. The content is solely the responsibility of the authors and
JCP, ERF, CAS, WY, CC, YL and DP, analyzed and interpreted does not necessarily represent the official views of the National
data; MVR wrote the first draft of the manuscript; all authors Institutes of Health.
contributed to the revision of the manuscript; and all authors
approved the final submitted version of the manuscript. Data-sharing statement
De-identified data used in the preparation of this manuscript are
Acknowledgments available upon request.
The authors thank the patients and parents who participated in
the clinical protocols included in this study, and the participating cli-
nicians and research staff.
References single-arm, phase 2 study. Lancet Oncol. Haematologica. 2017;102(10):1727-1738.

2018;19(2):240-248. 21. Schmiegelow K, Nielsen SN, Frandsen TL,
11. Pui CH, Campana D, Pei D, et al. Treating Nersting J. Mercaptopurine/methotrexate
1. Silverman LB, Gelber RD, Dalton VK, et al. childhood acute lymphoblastic leukemia maintenance therapy of childhood acute
Improved outcome for children with acute without cranial irradiation. N Engl J Med. lymphoblastic leukemia: clinical facts and
lymphoblastic leukemia: results of Dana- 2009;360(26):2730-2741. fiction. J Pediatr Hematol Oncol. 2014;36
Farber Consortium Protocol 91-01. Blood. 12. Jeha S, Pei D, Choi J, et al. Improved CNS (7):503-517.
2001;97(5):1211-1218. control of childhood acute lymphoblastic 22. Relling MV, Hancock ML, Boyett JM, Pui
2. Albertsen BK, Grell K, Abrahamsson J, et al. leukemia without cranial irradiation: St CH, Evans WE. Prognostic importance of 6-
Intermittent versus continuous PEG- Jude Total Therapy Study 16. J Clin Oncol. mercaptopurine dose intensity in acute
asparaginase to reduce asparaginase-associ- 2019;37(35):3377-3391. lymphoblastic leukemia. Blood. 1999;93(9):
ated toxicities: a NOPHO ALL2008 ran- 13. Kawedia JD, Liu C, Pei D, et al. 2817-2823.
domized study. J Clin Oncol. 2019;37(19): Dexamethasone exposure and asparaginase 23. Bhatia S, Landier W, Hageman L, et al.
1638-1646. antibodies affect relapse risk in acute lym- Systemic exposure to thiopurines and risk
3. Moricke A, Zimmermann M, Reiter A, et phoblastic leukemia. Blood. 2012;119(7): of relapse in children with acute lym-
al. Long-term results of five consecutive tri- 1658-1664. phoblastic leukemia: a Children's
als in childhood acute lymphoblastic 14. Liu C, Janke LJ, Kawedia JD, et al. Oncology Group study. JAMA Oncol.
leukemia performed by the ALL-BFM Asparaginase potentiates glucocorticoid- 2015;1(3):287-295.
study group from 1981 to 2000. Leukemia. induced osteonecrosis in a mouse model. 24. Relling MV, Schwab M, Whirl-Carrillo M,
2010;24(2):265-284. PLoS One. 2016;11(3):e0151433. et al. Clinical Pharmacogenetics
4. Smid EJ, Driessen AJ, Konings WN. 15. Merryman R, Stevenson KE, Gostic WJ Implementation Consortium guideline for
Mechanism and energetics of dipeptide 2nd, et al. Asparaginase-associated myelo- thiopurine dosing based on TPMT and
transport in membrane vesicles of suppression and effects on dosing of other NUDT15 genotypes: 2018 update. Clin
Lactococcus lactis. J Bacteriol. 1989;171(1): chemotherapeutic agents in childhood Pharmacol Ther. 2019;105(5):1095-1105.
292-298. acute lymphoblastic leukemia. Pediatr 25. Vora A, Goulden N, Mitchell C, et al.
5. Veerman AJ, Kamps WA, van den Berg H, Blood Cancer. 2012;59(5):925-927. Augmented post-remission therapy for a
et al. Dexamethasone-based therapy for 16. Chessells JM, Harrison G, Lilleyman JS, minimal residual disease-defined high-risk
childhood acute lymphoblastic leukaemia: Bailey CC, Richards SM. Continuing subgroup of children and young people
results of the prospective Dutch Childhood (maintenance) therapy in lymphoblastic with clinical standard-risk and intermedi-
Oncology Group (DCOG) protocol ALL-9 leukaemia: lessons from MRC UKALL X. ate-risk acute lymphoblastic leukaemia
(1997-2004). Lancet Oncol. 2009;10(10): Medical Research Council Working Party in (UKALL 2003): a randomised controlled
957-966. Childhood Leukaemia. Br J Haematol. trial. Lancet Oncol. 2014;15(8):809-818.
6. Matloub Y, Bostrom BC, Hunger SP, et al. 1997;98(4):945-951. 26. Lange BJ, Bostrom BC, Cherlow JM, et al.
Escalating intravenous methotrexate 17. Usami I, Imamura T, Takahashi Y, et al. Double-delayed intensification improves
improves event-free survival in children with Discontinuation of L-asparaginase and event-free survival for children with inter-
standard-risk acute lymphoblastic leukemia: poor response to prednisolone are associat- mediate-risk acute lymphoblastic
a report from the Children's Oncology ed with poor outcome of ETV6-RUNX1- leukemia: a report from the Children's
Group. Blood. 2011;118(2):243-251. positive pediatric B-cell precursor acute Cancer Group. Blood. 2002;99(3):825-833.
7. Seibel NL, Steinherz PG, Sather HN, et al. lymphoblastic leukemia. Int J Hematol. 27. Conter V, Valsecchi MG, Silvestri D, et al.
Early postinduction intensification therapy 2019;109(4):477-482. Pulses of vincristine and dexamethasone in
improves survival for children and adoles- 18. Duval M, Suciu S, Ferster A, et al. addition to intensive chemotherapy for
cents with high-risk acute lymphoblastic Comparison of Escherichia coli-asparagi- children with intermediate-risk acute lym-
leukemia: a report from the Children's nase with Erwinia-asparaginase in the phoblastic leukaemia: a multicentre ran-
Oncology Group. Blood. 2008;111(5):2548- treatment of childhood lymphoid malig- domised trial. Lancet. 2007;369(9556):123-
2555. nancies: results of a randomized European 131.
8. Boissel N, Baruchel A. Acute lymphoblastic Organisation for Research and Treatment 28. De Moerloose B, Suciu S, Bertrand Y, et al.
leukemia in adolescent and young adults: of Cancer-Children's Leukemia Group Improved outcome with pulses of vin-
treat as adults or as children? Blood. phase 3 trial. Blood. 2002;99(8):2734-2739. cristine and corticosteroids in continuation
2018;132(4):351-361. 19. Gupta S, Wang C, Raetz EA, et al. Impact therapy of children with average risk acute
9. Gaynon PS, Steinherz PG, Bleyer WA, et al. of asparaginase discontinuation on out- lymphoblastic leukemia (ALL) and lym-
Association of delivered drug dose and outcome in childhood acute lymphoblastic phoblastic non-Hodgkin lymphoma (NHL):
come for children with acute lymphoblastic leukemia: a report from the Children's report of the EORTC randomized phase 3
leukemia and unfavorable presenting fea- Oncology Group. J Clin Oncol. trial 58951. Blood. 2010;116(1):36-44.
tures. Med Pediatr Oncol. 1991;19(4):221- 2020;38(17):1897-1905. 29. Stock W, La M, Sanford B, et al. What
227. 20. Mondelaers V, Suciu S, De Moerloose B, et determines the outcomes for adolescents
10. Kantarjian H, Ravandi F, Short NJ, et al. al. Prolonged versus standard native E. coli and young adults with acute lymphoblastic
Inotuzumab ozogamicin in combination asparaginase therapy in childhood acute leukemia treated on cooperative group pro-
with low-intensity chemotherapy for older lymphoblastic leukemia and non-Hodgkin tocols? A comparison of Children's Cancer
patients with Philadelphia chromosome- lymphoma: final results of the EORTC- Group and Cancer and Leukemia Group B
negative acute lymphoblastic leukaemia: a CLG randomized phase III trial 58951. studies. Blood. 2008;112(5):1646-1654.

Hematopoiesis ARTICLE
Reversible switching of leukemic cells to a drug- Ferrata Storti Foundation

resistant, stem-like subset via IL-4-mediated
cross-talk with mesenchymal stroma
Hae-Ri Lee,1 Ga-Young Lee,1 Eung-Won Kim,1 Hee-Je Kim,2 Min-Ho Lee,3 R. Keith
Humphries4,5 and Il-Hoan Oh1
1
Catholic High-Performance Cell Therapy Center & Department of Medical Life Science,
College of Medicine, The Catholic University, Seoul, Republic of Korea; 2Division of
Hematology, Department of Internal Medicine, St Mary’s Hematology Hospital, College
of Medicine, The Catholic University of Korea, Seoul, Republic of Korea; 3Department of Haematologica 2022
Life Science, Dongguk University-Seoul, Goyang-si, Gyeonggi-do, Republic of Korea; Volume 107(2):381-392
4
Terry Fox Laboratory, British Columbia Cancer Agency, Vancouver, British Columbia,
Canada and 5Department of Medicine, University of British Columbia, Vancouver, British
Columbia, Canada
ABSTRACT
C
hemoresistance of leukemic cells has largely been attributed to
clonal evolution secondary to accumulating mutations. Here, we
show that a subset of leukemic blasts in contact with the mes-
enchymal stroma undergo cellular conversion into a distinct cell type
that exhibits a stem cell-like phenotype and chemoresistance. These
stroma-induced changes occur in a reversible and stochastic manner
driven by cross-talk, whereby stromal contact induces interleukin-4 in
leukemic cells that in turn targets the mesenchymal stroma to facilitate
the development of new subset. This mechanism was dependent on
interleukin-4-mediated upregulation of vascular cell adhesion molecule-
1 in mesenchymal stroma, causing tight adherence of leukemic cells to
mesenchymal progenitors for generation of new subsets. Together, our
study reveals another class of chemoresistance in leukemic blasts via
functional evolution through stromal cross-talk, and demonstrates
dynamic switching of leukemic cell fates that could cause a non-homol-
ogous response to chemotherapy in concert with the patient-specific
microenvironment. Correspondence:
IL-HOAN OH
iho@catholic.ac.kr
Introduction
Received: August 16, 2020.
Acute myeloid leukemia (AML) is a heterogeneous, clonal hematopoietic disorder
characterized by excessive proliferation of stem cell-like progenitor cells in the bone Accepted: December 22, 2020.
marrow (BM). AML has a highly variable prognosis1 and a very high risk of relapse Pre-published: January 14, 2021.
particularly in elderly patients.2
Leukemia progression and relapse are widely viewed to occur via clonal evolution
from preleukemic cells to overt leukemia driven by genetic mutations,3 followed by https://doi.org/10.3324/haematol.2020.269944
additional mutations leading to treatment-resistant, relapsed clone(s).4 However, in-
depth clonal analyses have revealed the persistence of founding clones,4 and func-
tional heterogeneity among the developed leukemic clones,5 suggesting that other
mechanisms may be involved. Material published in Haematologica is covered by copyright.
Several studies have highlighted leukemic stem cell (LSC) properties contributing All rights are reserved to the Ferrata Storti Foundation. Use of
to drug resistance:6 AML patients whose leukemic blast exhibit higher levels of stem conditions:
cell signatures are at greater risk of relapse and have a poorer prognosis.7,8 However, https://creativecommons.org/licenses/by-nc/4.0/legalcode.
the specific relationship between stemness and functional heterogeneity of LSC Copies of published material are allowed for personal or inter-
related to drug resistance, remains poorly understood.7,9 nal use. Sharing published material for non-commercial pur-
There is increasing awareness that the microenvironment, including growth fac- https://creativecommons.org/licenses/by-nc/4.0/legalcode,
tors, cytokines and niche stromal cells, can provide protection to leukemic cells and sect. 3. Reproducing and sharing published material for com-
thereby contribute to the acquisition of chemoresistance.10,11 For example, leukemic mercial purposes is not allowed without permission in writing
cell subsets surviving chemotherapy were localized to the surface of osteoblasts in from the publisher.
the BM.12-15 Subsequently, multiple protective signals from the stroma have been
shown to enhance leukemic cell survival through activation of receptor tyrosine

H.R. Lee et al.
kinases16 or interaction with the extracellular matrix. immunoglobulin G (IgG) from rat serum (Bio X cell).
However, despite these protective signals, a role for the Cytotoxicity of in vivo leukemic cells was examined by treat-
stroma in the clonal development of leukemic blasts for the ment with 100 mg/kg of Ara-C (Sigma-Aldrich, MO, USA) and
acquisition of chemoresistance has not been demonstrated. 3 mg/kg of doxorubicin hydrochloride (Sigma).
Here, we show that subsets of leukemic cells in stromal
contact undergo reversible changes associated with a stem Gene expression analysis
cell-like phenotype and drug-resistant state. These changes Sequencing libraries of two subjects were prepared according
are stochastic, and distinct from changes induced by other to the TruSeq Stranded Total RNA Sample Preparation guide.
mechanisms of chemoresistance, thus representing a new Aligned reads were quantified using HTSeq-count.23
class of drug-resistant cells developed in the leukemic Differentially expressed genes, fold ratio, P-value, and false dis-
microenvironment. covery rate were identified by edgeR algorithm24 for each sub-
ject. Enriched KEGG pathways were identified by GSEA-P.25
Methods Statistical analysis

In order to compare the generation of CD90+ subsets from
Human sample collection individual primary human leukemia patients’ samples, or the
Primary leukemic blasts were collected from newly diagnosed responses of individual patients’ leukemia cells to chemothera-
AML patients without prior treatment history. Part of the BM py, we used Mann-Whitney test. In order to compare the differ-
samples are from AML patients who had complete medical ences of means in specific experimental settings, we used a stan-
records during 5 years of follow-up in clinical courses. Human dard unpaired, two-tailed student t-test. The frequencies of
mesenchymal stromal cells (MSC) were separated from the BM leukemia-initiating cells in limiting dilution analysis were calcu-
of normal donors under informed consent. This study was lated by applying Poisson statistics with 95% Confidence
approved by the Institutional Review Boards of St. Mary’s Interval (CI) representing ±2 standard error of the mean (SEM).
Hospital and Catholic University of Korea.
Animals Results
C57/BL6 mice were obtained from the Jackson Laboratories
(Bar Harbor, ME). Bis+/+, Bis+/-, Bis-/- mice17 were provided by Dr. A subset of leukemic cells acquires a stem cell-like
Jeong-Hwa Lee (Catholic University of Korea). Mice with dis- phenotype by contact with mesenchymal stroma
ruption of interleukin-4 (IL-4) receptor18 were provided by Dr. In order to investigate the influence of stromal cells on
Chang Yul Kang (Seoul National University). the function of leukemic cells, we employed an in vitro co-
culture model of murine leukemic cells in contact with
Mouse and human acute myeloid leukemia cells and BM-derived MSC. Murine AML cells were generated by
mesenchymal stromal cells transducing BM mononuclear cells (MNC) with menin-
Fresh murine or human MSC were analyzed in the BM using gioma-1 (MN1)21 or HoxA9-Meis1 (H9M1)22 (Figure 1A).
flowcytometry. Cultured MSC (passage five to eight) were When co-cultured with MSC a subset of MN1 leukemic
obtained by serial plating of BM cells in the DMEM containing cells acquired a Sca-1(+) phenotype (Lin-c-kit+sca-1+; LSK)
10% fetal bovine serum as described.19,20 For generation of murine mimicking normal hematopoietic progenitors, while the
AML cells, fluorouracil (5-FU) treated BM cells were transduced majority remained Sca-1(-) (Lin-c-kit+sca-1-) (Figure 1B). The
with MN-1 or Meis1/HoxA9 through retroviral infection as acquisition of Sca-1(+) phenotypes was similarly observed
described.21,22 For co-culture, MSC were irradiated (15 Gy) 18-24 in other types of leukemic cells (H9M1) or leukemia cell
hours prior to use and leukemic cells were seeded on the MSC for line (C1498) independent of irradiation (Online
co-culture. For co-culture with transwell, MSC were seeded into Supplementary Figure S1A and B). The emergence of the Sca-
the upper chamber (6-well type, polyethylene terephthalate [PET] 1(+) subset was dependent on direct contact with the mes-
membrane with 0.4 mm pores; BD Bioscience, San Diego, USA) enchymal stroma (Online Supplementary Figure S1C), as
and leukemic cells were seeded into the lower well. these cells were not observed in stroma-free conditions or
in stromal co-culture with a transwell filter (Figure 1B).
Flow cytometry of leukemic cells and mesenchymal In order to determine if acquisition of the Sca-1(+) phe-
stromal cells notype occurs in vivo, MN1 leukemogenic cells (Lin-c-
Murine leukemic cells were analyzed by flow cytometry using kit+sca-1-) were transplanted into mice. Consistent with
the following antibodies: CD45.1-APC (BD PharMingen, USA), the in vitro results, a subset of leukemic cells (GFP+) in
Lineage cocktail (StemCell Technologies Inc, Canada), Sca-1-PE- recipient mice acquired a Sca-1(+) (Lin-c-kit+sca-1+) pheno-
Cy7 and c-kit-PE (BD PharMingen). For human leukemic cells, type (Figure 1C).
CD45-APC, CD34-BV421, CD90-FITC (BD PharMingen) anti- In order to determine whether acquisition of the Sca-
bodies were used. For MSC, anti-CD106 (VCAM-1)-biotin, 1(+) phenotype in leukemic cells originated from their
CD51-PE (eBioscience, CA, USA.), CD140a (PDGFRa)-APC, Sca- fusion with stromal cells, as implicated previously,26 we
1-PE-Cy7 (BD PharMingen) were used. co-cultured MN1 leukemic cells (GFP+) with MSC trans-
duced with YFP. None of the GFP+ leukemic cells co-
Treatment of antibody and cytotoxic drug expressed YFP (Figure 1D). Moreover, there was no differ-
Leukemic cells seeded on irradiated MSC were treated with ence in cell size between Sca-1(+) and Sca-1(-) cells, as
anti-IL-4 antibody (R&D Systems Inc., USA), anti-CD106 determined by identical forward scatter in flow cytome-
(VCAM-1) (R&D Systems Inc.) for 3 days. For in vivo antibody try, and no increase in tetraploidy in the Sca-1(+) cells
injections, mice received intraperitoneal injection of anti-IL-4 Ab (Figure 1E). Similarly, there was no evidence of cell fusion
(1 mg/kg) (R&D Systems Inc.) or intravenous injection of anti- in this in vivo generated Sca-1(+) subset (Online
VCAM-1 antbody (10 mg/kg) (Bio X cell, USA) along with Supplementary Figure S1D).

Functional switching of leukemic cells by stromal contact
A recent study implicated mitochondrial transfer from ma, and replated for a second round of co-culture with or
MSC to leukemic cells during acquisition of chemoresis- without stroma. The purified Sca-1(-) cell fraction again
tance.27,28 In order to examine this, MSC were labeled generated Sca-1(+) cells during the second round selec-
with a mitochondrial tracker and co-cultured with tively in the presence of stroma, whereas purified Sca-
leukemic cells. There was no difference in mitochondrial 1(+) cells co-cultured with stroma rapidly decreased in
tracker intensity between the Sca-1(-) and Sca-1(+) sub- frequency (Figure 2A and B) with the emergence of a
sets (Figure 1F). major Sca-1(-) cell population. Thus, final stable ratios of
Altogether, this emergence of a new leukemic subset Sca-1(+) and Sca-1(-) cells were similarly maintained
with a stem cell-like phenotype (Sca-1(+)) represents an under secondary stromal co-culture conditions regardless
intrinsic cellular evolution of leukemic cells that occurs of the phenotype of the initial cell population (Figure 2C).
independently of cell fusion or mitochondrial transfer The changes in cell populations occurred rapidly within 3
during in vivo leukemogenesis and in vitro culture with days of co-culture suggesting that the conversion
stromal cells. between Sca-1(-) and Sca-1(+) cells occurs by phenotypic
switching rather than selective proliferation in the cul-
Switching to the Sca-1(+) phenotype is reversible ture.
In order to determine if the Sca-1(+) subset is a stable Thus, the emergence of Sca-1(+) leukemic cells during
phenotype, we sort-purified Sca-1(+) (LSK) and Sca-1(-) stromal contact occurs in a reversible manner in any sub-
(LK) leukemic cells generated during co-culture with stro- sets of leukemic cells without clonal predisposition (sto-
A B
C D
E F
Figure 1. Generation of a stem cell-like phenotype in a subset of leukemic cells. (A) Schematic illustration of the experiment. Murine acute myeloid leukemia (AML)
cells were generated by transduction of fluorouracil (5-FU)-treated bone marrow (BM) cells with retrovirus encoding oncogene (MN1, or HoxA9/Meis1). Shown are
retroviral vectors, experimental procedure for transplantation into mice, and the light microscopy morphology of transformed leukemic cells visualized by Giemsa
staining. (B) Generation of Sca-1(+) (Lin-c-kit+sca-1+: LSK) leukemic cells during co-culture with murine mesenchymal stromal cells (mMSC). Co-cultures with mMSC
for 3 days were performed in the presence (transwell) or absence (direct contact) of a transwell membrane between the cells in comparison to stroma-free (SF) cul-
ture. Phenotypes of leukemic cells (CD45+GFP+) from co-cultured MSC (CD45-GFP-) were analyzed by flow cytometry. Shown are the representative profile (left) and
quantification (right) (mean ± standard error of the mean [SEM] , n=7, *P<0.05). (C) In vivo generation of Sca-1(+) leukemic subsets. MN1 leukemic cells (Lin-c-kit+)
were transplanted into mice and generation of Sca-1(+) subsets among BM engrafted leukemic cells were examined at 2 weeks post-transplantation (95% green
fluorescent protein postive [GFP+] leukemic cells at the point). Representative flowcytometry plot (left) and quantification (right) are shown (mean ± SEM, n=10,
*P<0.05). (D) Experimental scheme for analyzing cell fusion between MSC and leukemic cells. MSC transduced with a retroviral vector encoding yellow fluorescent
protein (YFP), and leukemic cells transduced with a vector encoding GFP were co-cultured for 3 days. Shown are the experimental scheme (left) and representative
flow cytometry profiles showing the absence of double positive (YFP/GFP) populations before and after co-culture (right). (E) Flow cytometry profiles for comparison
between LK (Sca-1(-)), and LSK (Sca-1(+)) cell populations of cell size by forward scattering (left), and of DNA content (right) (n=5). (F) Experimental scheme to com-
pare mitochondrial transfer between Sca-1(+) and Sca-1(-) cell populations. Murine MSC were pre-labeled with MitoTracker and co-cultured with MN1 leukemic cells
for 3 days. Shown are representative flow cytometry plots from the experiments, each indicated leukemic cell subset (LK or LSK) of leukemic cells (CD45+GFP+) was
gated and analyzed for MitoTracker and quantified for difference in mitochondrial transfer (n=4).

H.R. Lee et al.
B C
Figure 2. Reversible and equipotent nature for generation of LSK leukemic subsets. LSK (Sca-1(+)) or LK (Sca-1(-)) subsets of MN1 leukemic cells generated by co-
culture were sort-purified and then replated for 3 days in the absence (SF) or presence of mesenchymal stromal cells (+MSC). (A) Flow cytometry profiles. (B and C)
Quantitative analysis for expansion of cell numbers for LSK or LK subsets from input numbers of sorted LSK or LK cell populations during the second round of co-
culture with murine MSC. Shown are the fold increases of cell numbers compared with input in the second co-culture (B) and final frequencies for LSK from each
set of the second co-culture (C) (mean ± standard error of the mean, n=6, *P<0.05).
chastic), but with similar probability of each cells leukemia stem cells30 in the Sca-1(+) leukemic subset in
(equipotent) for conversion among the total leukemic cell comparison to the other subsets. Thus, subsets of Lin(+)
populations. cells, Lin-c-kit-, LK (Lin-c-kit+sca-1-) cells and LSK (Lin-c-kit+
sca-1+) leukemic cells generated in the BM of MN1 trans-
Functional heterogeneity acquired in stem cell-like planted mice were sort purified and transplanted into sec-
leukemic subsets ondary recipient mice in a limiting dilution assay. However,
We next determined whether the Sca-1(+) stem cell-like the LK and LSK populations exhibited a similar frequency
leukemic subset arising by stromal contact was functionally of LIC, while exhibited significantly higher frequencies
distinct. When MN1 leukemic cells were treated with the than the other cell populations (Figure 3D; Online
chemotherapeutic Ara-C during co-culture with mesenchy- Supplementary Figure S2A). These two populations (LK and
mal cells, Sca-1(-) subsets exhibited significant decrease of LSK) also exhibited comparable levels of in vivo leukemic
cell numbers, but the Sca-1(+) subset exhibited higher engraftment or in vitro leukemia colony formation (Online
resistance compared to the Sca-1(-) subsets, with no signif- Supplementary Figure S2B and C), indicating that the Sca-1(+)
icant changes in cell numbers (Figure 3A). Drug resistance in subset developed during in vivo leukemogenesis comprise a
the Sca-1(+) subset was similarly reproduced in other subset of LIC that does not display significantly different
leukemia cell types tested (HoxA9/Meis1-induced leukemic leukemogenic activity compared to their Sca-1(-) counter-
cells or C1498 leukemia cell line) (Figure 3A). parts.
The chemoresistance of the Sca-1(+) (LSK) leukemic pop- Thus, the Sca-1(+) leukemia subset generated from
ulation compared to the rest of the Sca-1(-) (LK) cells was leukemic cells represents a distinct leukemic cell population
similarly observed in vivo with mice engrafted with MN1 that has acquired drug-resistance without altering their
leukemic cells and treated with chemotherapeutic drug leukemogenic activity.
(Ara-C and doxorubicin)29 (Figure 3B). Thus, enhanced drug
resistance is a common feature of leukemic subsets acquir- Interleukin-4 plays a role in the emergence of drug-
ing a Sca-1(+) phenotype upon stromal contact in a range of resistant Sca-1(+) cells
leukemic cell models. We next sought to identify possible signals from the stro-
In order to further investigate the drug resistance of the ma that induce emergence of Sca-1(+) cells. Given that
Sca-1(+) cells, we analyzed their cell cycling in BM and altered production of cytokines and/or growth factors are
found that % of quiescent cell population (G0) was higher frequently observed in leukemic cells,10 we examined the
in LSK cells (Figure 3C). We also compared the frequency of cytokine/growth factor gene expression induced by stromal
leukemic initiating cells (LIC), a functional assay for contact of leukemic cells (Online Supplementary Figure S3A).

A B
C D
Figure 3. Property of stem cell-like leukemic subsets exhibiting drug resistance without changes in leukemic initiating cell frequencies. (A) Murine leukemic cells
were cultured in the presence of murine mesenchymal stromal cells (MSC), and numbers of surviving cells (annexinV- propidium iodide [PI]) in Sca-1(+) or Sca-1(-)
subsets were measured after 2 days of treatment with Ara-C (100 nM for MN1 leukemia, 30 nM for H9M1, 500 nM for C1498 leukemic cells). (B) Mice engrafted
with MN1 leukemic cells were treated with chemotherapeutic agents (Ara-C + doxorubicin) for 3 days by intraperitoneal injection at the indicated times and then
examined for changes in the numbers of Sca-1(+) (LSK) or Sca-1(-) (LK) subsets in the bone marrow (BM). Experimental scheme (upper) and relative fold changes
(lower) in the cell numbers after chemotherapy (chemoTx) compared to the control groups (mean ± standard error of the mean, n=5, *P<0.05). (C) Cell cycling of
leukemic subsets in BM. MN1 leukemic cells were transplanted into mice and cell cycling of Sca-1(+) and Sca-1(-) subsets in BM were analyzed by
Hoechst33342/pyronin staining. Shown are the representative flow cytometry plots with % of cell population (upper) and quantification of cells in G0 (quiescent cell
population) and Non-G0 (G1/S/G2M) phase (lower) (n=3, *P<0.05). (D) Comparisons of leukemia-initiating cell (LIC) frequencies for each leukemic subset. MN1
leukemic cells were transplanted into mice and each subset of leukemic cells in recipient BM were sort-purified for transplantation into secondary recipients in a
limiting dilution dose. Shown is the plot of limiting dilution analysis for frequencies of LIC in each leukemic subset analyzed by Poisson statistics. The resulting LIC
frequencies are shown in the Online Supplementary Figure S2A with 95% Confidence Intervals in parenthesis.
Upon contact with stroma, the murine leukemic cells of recipients without changes in overall engraftment levels
exhibited a notable induction of cytokines and growth fac- (Figure 4B). Moreover, Il-4-neutralizing antibody abrogated
tors implicated in leukemogenic activity, including IL-4, resistance of the LSK population to chemotherapeutic drugs
PDGF-A, PDGF-D, CCL-2, CCL-5, CXCL-1 and stem cell (Ara-C and doxorubicin), markedly decreasing the LSK pop-
factor,31-39 but not in the presence of transwell filters (Online ulation in recipient BM (Figure 4C), which caused a
Supplementary Figure S3B). Among those cytokines, IL-4 was decrease in the residual burden of surviving LIC that can ini-
selectively induced in LSK subsets, but not in the majority tiate leukemogenesis (Online Supplementary Figure S4).
of remaining cells (LK) as determined by its transcript and Together, these results support a key role for IL-4 in the
protein level (Online Supplementary Figure S3C and D). Thus, generation of drug resistant Sca-1(+) subset upon stromal
we examined whether IL-4 acts as an autocrine signal for contact.
generating Sca-1(+) subsets. Addition of recombinant IL-4
increased the frequency of Sca-1(+) subsets (LSK) in a dose- Interleukin-4-dependent generation of Sca-1(+)
dependent manner (Figure 4A). Conversely, addition of an leukemic cells is generated by stromal cross-talk
IL-4-neutralizing antibody significantly decreased the fre- In order to investigate the mechanisms underlying IL-4-
quency of LSK during co-culture (Figure 4A). Injection of an mediated generation of Sca-1(+) subsets, we examined the
antibody against IL-4 into recipient mice along with MN1 cellular target of IL-4 during the co-culture of leukemic cells
leukemic cells also decreased the LSK population in the BM and stroma (Figure 5A). First, to see if IL-4 acts directly on

H.R. Lee et al.
Figure 4. Role of stroma-induced interleukin-4 in the generation of the stem cell-like leukemic subset. (A) Effects of interleukin-4 (IL-4) during co-culture of leukemic
cells. Left: experimental scheme; middle: % LSK generated during 3-day co-culture of leukemic cells with stroma supplemented with recombinant IL-4 (mean ± stan-
dard error of the mean [SEM], n= 6); Right: effects of antibody against IL-4 on generation of LSK during stromal co-culture of leukemic cells. Shown are the
mean±SEM for % LSK in leukemic cells (green fluorescent protein positive [GFP+] CD45+) (n=7, *P<0.05). (B) Effects of IL-4 antibody on in vivo generation of LSK.
Left: experimental design. Antibody against IL-4 was intraperitoneally administered into recipient mice at each indicated time point before and after transplantation
of MN1 leukemic cells. Middle: numbers of LSK leukemic (GFP+) cells in the BM (two femurs and two tibia) of recipient mice. Middle and right: % of MN1 leukemic
cell engraftment determined by total leukemic cells (GFP+ cells) (middle) and cells with leukemia-initiating cell properties (LK and LSK) (right), respectively (mean
± SEM, n=6, *P<0.05). (C) Effects of IL-4 antibody on the chemosensitivity of the leukemic subsets. Left: experimental design. After engraftment of MN1 leukemic
cells (10 days after transplantation), recipients were injected with IL-4 antibody and chemotherapeutic drug (AraC+doxorubicin) at the indicated times. Right:
changes in the chemosensitivity of leukemic subsets by in vivo injected IL-4 antibody. Relative fold decrease in the cell numbers of each leukemic subset compared
to the control (phosphate buffered saline) group 3 days after exposure to drug and antibody (mean ± SEM, n=5).
leukemic cells, we established MN1 leukemic cells from molecules 1 (VCAM-1) in MSC, which mediate stromal
hematopoietic progenitors of mice lacking IL-4 receptor a adherence of leukemic cells,40 were up-regulated by IL-4 in
(IL-4Ra knockout [KO]). Co-culture of leukemic cells from WT MSC, but not in IL-4Ra KO MSC (Figure 5D).
IL-4Ra KO or wild-type (WT) with stromal cells led to com- Conversely, in vivo injection of IL-4-neutralizing antibody
parable frequencies of LSK or LK subsets in each group caused a significant decrease of VCAM-1 expressions in BM
(Figure 5B). In contrast, when mesenchymal stromal cells mesenchymal cells including subsets enriched for mes-
from IL-4Ra KO mice were co-cultured with MN1 leukemic enchymal progenitors (CD44(-)PDGFRa(+))41-43 (Figure 5E).
cells, significantly lower frequencies of LSK, but not LK sub- In order to further examine the influences of stromal
sets, were observed compared to the WT stroma group VCAM-1 expression level on the generation of LSK subsets,
(Figure 5C). Thus, IL-4 signals target mesenchymal stromal leukemic cells were co-cultured with sort-purified MSC
cells, rather than leukemic cells, to facilitate stroma-mediat- fractions for different levels of VCAM-1. MSC with higher
ed generation of the LSK subset, indicating that IL-4-medi- VCAM-1 levels increased LSK generation during co-culture,
ated cross-talk promotes the functional evolution of whereas MSC expressing lower levels of VCAM-1
leukemic cells. decreased it, in comparison to LSK cells from unsorted MSC
Next, to investigate the effects of IL-4 on mesenchymal co-cultures (Figure 5F). Similarly, VCAM-1-blocking anti-
stroma, we examined whether the mode of cellular interac- body significantly decreased stromal adherence of leukemic
tion between MSC and leukemic cells is influenced by IL-4. cells (Online Supplementary Figure S5B), which led to a con-
LSK subsets were predominantly generated among the comitant decrease in the generation of the LSK subset
leukemic cells tightly adherent to the mesenchymal cells, (Figure 5G). Moreover, in vivo administration of VCAM-1
for both MN1 or H9M1 leukemic cells, but seldom among antibody caused a significant decrease of LSK numbers in
the loosely adherent/suspension leukemic cells (Online BM (Figure 5H). These data, together with positive expres-
Supplementary Figure S5A). Supporting the influence of IL-4 sion of VCAM-1 ligands in leukemic cells (Online
on stromal adherence, the level of vascular cell adhesion Supplementary Figure S6) indicates that tight adherence of

A B
C D
E F
G H
Figure 5. Interleukin-4-mediated cross-talk in mesenchymal stromal cells controlling VCAM-1-mediated generation of leukemic subsets. (A to C) Identification of
interleukin-4 (IL-4) target in cross-talk of leukemic cells and mesenchymal stromal cells (MSC). (A) Experimental design. IL-4 receptor (IL-4Ra) knockout (Leukemia-
KO) or wild-type (WT) leukemic cells were co-cultured with WT mice-derived murine MSC (MSC WT) (left), or MSC from IL-4Ra KO mice (MSC KO) or WT (right) were
co-cultured with leukemic cells from WT mice. (B) Effects of IL-4Ra KO out in leukemic cells on the generation of LSK subsets. Shown are the quantification of %
LSK and LK cells during co-culture (mean ± standard error of the mean[SEM], n=9, *P<0.05). (C) Effects of IL-4R KO in MSC on the generation of LSK subsets.
Shown are the quantification of % LSK and % LK from the co-culture (mean± SEM, n=15, *P<0.05). (D to H) IL-4 targeting of MSC facilitates generation of leukemic
subsets by controlling VCAM-1 expression in MSC. (D) Effects of IL-4 signals on VCAM-1 expression levels of MSC. Murine MSC from WT or IL-4R KO mice were treated
with recombinant IL-4 and the fold increase of % VCAM-1(+) were analyzed in comparison to the control group (mean± SEM, n=6, *P<0.05). (E) In vivo changes of
VCAM-1(+) cells in the bone marrow (BM) of mice injected with IL-4-neutralizing antibody. Mice were intraperitoneally injected with IL-4 antibody for 4 days and ana-
lyzed for % VCAM-1(+) cells in BM stromal cells. Shown are the relative fold differences of % VCAM-1(+) cells in the indicated subsets of BM mesenchymal stromal
cells relative to the IgG treated mice group (n=5, *P<0.05). (F) Influence of VCAM-1 expression in MSC for generation of LSK subsets. Murine MSC were sort-purified
for differences in VCAM-1 expression levels, and the generation of LSK subsets from MN1 leukemic cells co-cultured with each fraction was analyzed. Shown are
the flow cytometry plots for sorting of MSC (left) and relative folds for LSK numbers generated in each co-culture group compared to co-culture with unsorted MSC
(right) (n=6). (G) Effects of blocking antibody against VCAM-1 on the in vitro generation of LSK subsets. During co-culture of leukemic cells with stroma, the indicated
amount of antibody against VCAM-1 was added and changes in the numbers of LSK generated in the co-culture were analyzed (n= 6, *P<0.05). (H) Effects of block-
ing antibody against VCAM-1 on the in vivo generation of LSK subsets. Mice transplanted with MN1 leukemic cells were injected with rat immunoglobulin (Ig) or
VCAM-1-blocking antibody (intravenous 10 mg/kg) (7 and 10 days after leukemic cell transplantation). Three days after antibody injection, generation of LSK in recip-
ient BM was analyzed (n=3 for IgG, n=4 for anti-VCAM1, *P<0.05).

H.R. Lee et al.
A C D
B E F
Figure 6. Generation of stem-like, drug-resistant leukemic cells in human leukemia models. (A) Generation of stem-like (CD90(+)) leukemic subsets from each indi-
cated human leukemia cell line during stromal co-culture. Leukemic cell populations (CD45(+))were gate separated from mesenchymal stromal cells (MSC) (CD45(-
) for analysis. Shown are the quantification of the frequency of CD90(+) cells in the co-culture (n=7, † <0.2%, *P<0.05). (B) Comparison of drug sensitivity between
each leukemic subset. Each indicated human leukemic cell line was exposed to Ara-C (200 nM) for 2 days. Shown are the numbers of surviving (annexinV- propidium
iodide-) cells in the culture (mean ± standard error of the mean [SEM], n=6). (C to G) Generation of stem-like, drug-resistant leukemic subsets in human leukemic
cells from acute myeloid leukemia (AML) patients. (C) Human leukemia cells from AML patients were co-cultured in the presence or absence of human MSC. Shown
are the % CD90 (+) cells in total leukemic cells after co-culture for 3 days and differences were analyzed by Mann-Whitney U test (n=14 from seven individual
patients’ samples, P<0.05). (D) Effects of interleukin-4 (IL-4) on the generation of CD90(+) subsets. Leukemic cells were co-cultured with human MSC in the pres-
ence or absence of IL-4 (100 ng/mL). Shown are the relative fold increases in % of CD90(+) subsets in leukemic cells (mean± SEM, n= 10 from five individual
patient’s samples, *P<0.05). (E) Reversible switching of subsets of human primary leukemic cells to maintain constant equilibrium. CD90(+) and CD90(-) subsets
generated during co-culture were sort-purified and re-plated in the co-culture with human MSC. Shown are the fold changes of each subset after plating each purified
subset after 3 days of co-culture (n=6, *P<0.05). (F) Comparisons of drug sensitivity between the CD90(+) subset and the rest of the CD90(-) population in primary
AML cells after exposure to Ara-C (200 nM). Shown are the relative fold changes in numbers of surviving cells of each population determined by numbers of
annexinV-propidium iodide- (PI) cells and differences were analyzed by Mann-Whitney U test (n=14 from seven individual patients’ samples, P<0.05). (G) Enrichment
of stem cell signatures in CD90(+) human leukemic cells. Primary human leukemic cells from two AML patients (#4 and #2) were co-cultured with human MSC for
3 days, and the generated CD90(+) and CD90(-) cells were subjected to RNA sequencing analysis. Differentially expressed genes were analyzed by gene set enrich-
ment analysis (GSEA-P) for enrichment of 259 genes specific for leukemia stem cells.47 Shown are the plots of enrichment scores (upper) and ranked list of each
gene in the order of log folds ratio (CD90+/90-) with position at zero indicated (lower).
2
leukemic cells to VCAM-1 in MSC facilitates emergence of can be seen in human leukemic cells, we examined human
LSK subsets. AML cells for acquisition of CD90(+) as a phenotype for
Consistent with these findings, gene expression changes stem-like subsets based on findings that a subset of
in MSC induced by IL-4 treatment during culture revealed CD90(+) cells amongst CD34(+) cells represent long-term
41 differentially expressed genes (DEG), the most profound repopulating hematopoietic stem cells (HSC)44 and that
changes of which were in the gene ontology group related CD90 expression in human leukemic cells represents high-
to the ‘binding’ molecular function, supporting their role in risk leukemia with stem cell properties.45,46 We first exam-
the cellular interaction with leukemic cells (Online ined human leukemic cell lines, MOLM-14 and MV4-11
Supplementary Figure S7A, B). Thus, IL-4 enhances the cellu- (M5 type FAB), and HL-60 (M3 type FAB). For each
lar interaction of stroma and leukemic cells to facilitate stro- leukemic cell line tested, co-culture with human BM-
ma-dependent evolution of the Sca-1(+) leukemic subset derived MSC resulted in the emergence of leukemic subsets
exhibiting drug resistance. with the CD90(+) phenotype, albeit to variable levels
(Figure 6A). Moreover, when chemoresistance was com-
Stroma-induced changes in human leukemic cell pared between leukemic subsets, significant resistance to
models Ara-C treatment was observed selectively for CD90(+) cells
In order to investigate whether a similar phenomenon in all tested leukemic cells (Figure 6B) similarly exhibiting

A B
C D
Figure 7. Dependency on the mesenchymal progenitor cells for generation of stem-like leukemic subsets. (A) Identification of VCAM-1-expressing mesenchymal
stromal cells in the bone marrow (BM) of mice. % of VCAM-1(+) cells among murine mesenchymal cells (CD45-31-Ter119-) in fresh BM were compared for CD44(+)
and CD44(-) cells. Shown are the representative flow cytometry plots (left) and quantification (right) for frequency of VCAM-1(+) cells among the indicated murine
mesenchymal stromal subsets in fresh BM. (B) Comparisons of frequency of VCAM-1(+) cells in mice BM between the mesenchymal progenitor and non-progenitor
subsets of murine mesenchymal stromal cells. Mesenchymal progenitor subsets in fresh mice BM were defined by PDGFRa(+)/Sca-1(+) or PDGFRa(+)CD51(+) based
on published reports.41-43,50 (C) Leukemogenesis in the Bis knockout (KO) mouse model. MN1 leukemic cells were transplanted into Bis KO mice, where mesenchymal
progenitor populations are selectively decreased. Two weeks after transplantation into neonates of each mice model, engraftment of leukemic cells in BM and %
LSK among engrafted leukemic cells were analyzed. Shown are the experimental design (upper) and % LSK leukemic subsets among engrafted leukemic cells for
each indicated mice recipient (lower, left) and % engraftment of leukemic cells (GFP (+)) in BM (lower, right) (mean ± standard error of the mean [SEM], n=6 for wild-
type [WT], n=27 for hetero, n=10 for KO). (D) Comparisons of mesenchymal progenitor cell numbers in BM of acute myeloid leukemia (AML) patients with respect
to the clinical course. (upper) Experimental design. Fresh uncultured BM of AML patients without prior treatment were analyzed for cytogenetic abnormalities of
leukemic blasts and content of mesenchymal progenitor cells (MPC; CD45-31-235a-146+166-) in fresh BM. Five years after the initial analysis, MPC numbers in
patients’ fresh BM were compared with subsequent clinical courses (maintenance of complete remission or relapse) with respect to the karyotype of leukemic blasts.
(lower) Mean numbers of MPC (CD146+166-) in fresh BM of AML patients for each indicated clinical course and karyotype (mean ± SEM, n=14 for normal karyotype,
n=5 for mixed-lineage leukemia [MLL], n=10 for others).
quiescence in cell cycle (Online Supplementary Figure S8) than the remaining CD90(-) cells (Figure 6G), and enriched
In order to examine these findings in primary leukemic with gene sets involved in the interaction with the extracel-
cells, we examined the response of primary AML blasts lular matrix (ECM) or focal adhesion (Online Supplementary
from five to seven individual patients to mesenchymal stro- Figure S9).
ma. Primary AML blasts exhibited a significant induction of Thus, subsets of human leukemic cells in contact with
CD90(+) cells upon stromal contact, which was further stroma exhibit a stem-like properties to acquire drug-resis-
increased by IL-4 treatment during co-culture (Figure 6C tance through interaction with stroma.
and D). Sort-purified subsets of CD90(+) and CD90(-)
leukemic cells exhibited similar switching of phenotypes to Stromal heterogeneity for generation of stem cell-like
maintain constant ratios in CD90(+) subsets in total leukemic subsets
leukemic cells, as observed for murine leukemic cells (Figure Extensive heterogeneity has been documented among
6E). Moreover, the CD90 (+) subset generated during stro- mesenchymal populations in BM stroma.48,49 Therefore, we
mal contact exhibited higher resistance to Ara-C (Figure 6F) investigated the mesenchymal subpopulations responsible
than the remaining CD90(-) cell population in the same co- for generation of Sca-1(+) cells. Given that VCAM-1
culture, demonstrating a similar drug-resistance of newly expressing MSC played a role in the generation of drug-
emerging leukemic subsets in primary human leukemic resistant subsets, we examined VCAM-1 expression among
cells. stromal cell populations in the BM. VCAM-1(+) mesenchy-
Gene expression study on two independent patients mal cells were predominantly enriched by a CD44(-) popu-
showed that CD90(+) subsets of primary human AML cells lation, where colony-forming mesenchymal progenitor
are significantly enriched with gene sets specific for LSC47 cells (MPC) are exclusively localized41-43,50 (Figure 7A).

H.R. Lee et al.
Similarly, VCAM-1 is exclusively enriched in subsets for homogenous response to chemotherapy among the
MPC as defined by PDGFRa(+)/Sca-1(+) or leukemic cell populations.
PDGFRa(+)/CD51(+) subsets,41-43,50 indicating that the The mechanisms for dynamic equilibrium among differ-
VCAM-1(+) cells that can drive emergence of drug-resistant ent subsets of leukemic cells remains still unclear. One pos-
subsets predominantly overlap with mesenchymal progen- sibility is a feedback control mechanism that maintains a
itor cells (Figure 7B). constant ratio of stem-like versus non-stem-like leukemic
Supporting this finding, when MN1 leukemic cells were cells, probably through cellular interaction between distinct
transplanted into the homozygous Bis KO mice, where self- leukemic subsets, as inferred from clonal interactions
renewing MPC are decreased in the BM,17,51 a significant between heterogenous subsets.55 Similarly, studies on can-
decrease in the Sca-1(+) subset (LSK) among BM engrafted cer stem cells have suggested that non-tumorigenic cells
leukemic cells was observed compared to WT or heterozy- regulate the maintenance of cancer stem cells influencing
gote Bis KO mice (Figure 7C). This decrease was not associ- their relative frequencies in the population.56 Since clonal
ated with altered overall engraftment levels (Figure 7C), con- heterogeneity of leukemia or cancer cells underlies the dif-
sistent with the differences between LSK and LK subsets. ferential response to chemotherapy and emergence of
Similarly, supporting the role of MPC for the develop- relapsing clones,55,57,58 the kinetics of generating these stem-
ment of drug resistant leukemic cells, AML patients wo like subsets could be a factor for differential response to
relapsed after treatment exhibited higher numbers of mes- chemotherapy.
enchymal progenitor subsets (MPC: CD146+/166-)19, 43, 52 in Interestingly, we show that the development of these
BM, i.e., retrospective studies on AML patients who had drug-resistant leukemic subset is facilitated by bi-direction-
undergone relapse within 1 year after complete remission al cross-talk between stroma and leukemic cells mediated
exhibited higher numbers of mesenchymal progenitor sub- by IL-4, exhibiting resistance to apoptosis (Online
sets in the BM than those who maintained complete remis- Supplementary Figure S11). While IL-4 was implicated in
sion for 5 years (Figure 7D). Notably, this difference was inhibition of leukemic cells and apoptosis,59 we did not find
observed regardless of the underlying cytogenetic abnor- increased apoptosis of the non-stem-like population pre-
mality of the leukemic blasts, indicating that the hetero- cluding the selective enrichment of stem-like subsets by IL-
geneity of stromal cells could be an additional factor drug- 4 (Online Supplementary Figure S12). Moreover, rather than
resistance of leukemic cells. acting directly on the leukemic cells, IL-4 targets stromal
Altogether, this heterogeneity in the BM MPC can influ- cells, which facilitate the generation of LSK subsets. How
ence the stroma-dependent generation of stem cell-like IL-4 acts on stromal cells to facilitate the generation of a
leukemic subsets. drug-resistant leukemic subset remains unclear. However,
we demonstrated a key role for VCAM-1 downstream of
IL-4 in MSC leading to tight adherence of leukemic cells
Discussion and MSC, which was necessary for generation of the LSK
subset. Similarly, we found IL-4-dependent induction of
Leukemic cell evolution has prevented the effective man- gene clusters in MSC whose functions are related to ‘bind-
agement of a diverse spectrum of leukemic disease. ing function’. This suggests that the mode of interaction
Here, using a variety of murine and human leukemia cells between MSC and leukemic cells is altered by IL-4 acting
both in vitro and in vivo, we show that subsets of leukemic on MSC, which facilitates the development of the drug-
cells can undergo a phenotypic conversion into a stem-like resistant leukemic subset.
phenotype that exhibit a higher resistance to chemotherapy Interestingly, we also found functional differences
in the context of stromal contact. This development of between stromal cells in terms of their capacity to drive
chemoresistant subsets by stromal contact was not depend- development of drug-resistant leukemic cells. We found
ent on cell fusion or changes in leukemogenic activities that expression of VCAM-1 in stromal cell was important
observed in mitochondrial transfer.27,28 The acquisition of for adherence-dependent generation of leukemic subsets,
the stem cell-like phenotype was reversible, being rapidly while other adhesion molecules we tested did not influence
reverted to the non-stem cell phenotype under stroma-free the process (Online Supplementary Figure S13). Importantly,
conditions independent of difference in cell cycles or apop- the VCAM-1-expressing stromal cells were selectively
tosis (Online Supplementary Figure S10), unlike the stable enriched in mesenchymal progenitors. Since mesenchymal
maintenance of chemoresistance in leukemic clones gener- stromal cells undergo various degenerative changes during
ated by clonal evolution. leukemia,19,60,61 it is possible that patient-to-patient hetero-
Moreover, the frequencies of LSK cells among leukemic geneity in BM mesenchymal progenitor cell content could
cells in contact with stroma were maintained constant differentially contribute to the development of drug-resis-
regardless of the phenotype of the initial cell populations. tant clones. Consistent with this, AML patients whose BM
This suggests that the stroma-mediated development of has higher levels of a primitive (CD146+) subset of mes-
the stem cell-like, drug-resistant subpopulation occurs in a enchymal cells19,43,52 that express higher levels of VCAM-1
stochastic and reversible manner in leukemic cells with tend to have a higher risk of leukemic relapse compared to
similar probabilities among leukemic cells (equipotent) those who maintained complete remission. Thus, inde-
without clonal predisposition. Reminiscent of these find- pendent of oncogenic mutations or cytogenetic abnormali-
ings, recent studies showed that non-stem cancer cells can ties in the blasts,62 heterogeneity per se in mesenchymal pro-
be spontaneously converted to stem-like state, and these genitors in BM could be another factor for development of
plasticity of cancer cells allows cellular switching between drug-resistant leukemic subsets.
distinct functional states.53,54 Together, these studies raise In summary, our study reveals an additional mechanism
the possibility that the stochastic development of of functional evolution of leukemic cells induced by contact
chemoresistant clones by stromal contact is an intrinsic with the mesenchymal stroma that can cause a reversible
process of leukemogenesis that could cause a non- switch to a stem cell-like, drug-resistant subset independent

of mutation-driven clonal evolution in leukemic blasts materials and wrote the manuscript; IHO conceptualized idea and
(Online Supplementary Figure S14). These findings thus pro- research, supervised research, wrote the manuscript and provided
vide further insight into the multiple mechanisms for devel- financial support
opment of drug-resistance that could generate leukemic
cells with distinct characteristics and chemoresistance, Acknowledgements
highlighting the importance of the microenvironment in We thank Dr. Lee, Jeong-Hwa (College of Medicine, Catholic
this process. This supports the need for better defining the University of Korea) for the kind supply of bis knock-out mice and
mechanisms of drug resistance in leukemia patients, and Dr. Kang, Chang-Yul (College of Pharmacy, Seoul National
could lead to the development of more comprehensive University) for the kind supply of mice lacking the IL-4 receptor
management of leukemic diseases. and Life Science Editors for manuscript editing. We thank Dr. Jin-
A Kim for help in clinical data processing. We also thank the
Disclosures Department of Biostatistics of the Catholic Research Coordinating
No conflicts of interest to disclose. Center for statistical support.
Contribution Funding
HRL, GYL, EWK, HJK performed experiments and collected This study was supported by the NRF of Korea and funded
data, MHL performed experiments on genomics and statistical by Ministry of Science, ICT, & Future Planning
analysis of data; RHK conceptualized research, provided study (2017M3A9B3061947)
References eficient mice. Leukemia. 2007;21(1):136- edgeR: a Bioconductor package for differ-
142. ential expression analysis of digital gene
1. Estey E, Dohner H. Acute myeloid 14. Saito Y, Uchida N, Tanaka S, et al. expression data. Bioinformatics. 2010;26
leukaemia. Lancet. 2006;368(9550):1894- Induction of cell cycle entry eliminates (1):139-140.
1907. human leukemia stem cells in a mouse 25. Subramanian A, Kuehn H, Gould J,
2. Craddock C, Tauro S, Moss P, Grimwade model of AML. Nat Biotechnol. Tamayo P, Mesirov JP. GSEA-P: a desktop
D. Biology and management of relapsed 2010;28(3):275-280. application for Gene Set Enrichment
acute myeloid leukaemia. Br J Haematol. 15. Ishikawa F, Yoshida S, Saito Y, et al. Analysis. Bioinformatics. 2007;23(23):
2005;129(1):18-34. Chemotherapy-resistant human AML stem 3251-3253.
3. Welch JS, Ley TJ, Link DC, et al. The origin cells home to and engraft within the bone- 26. Cogle CR, Goldman DC, Madlambayan
and evolution of mutations in acute marrow endosteal region. Nat Biotechnol. GJ, et al. Functional integration of acute
myeloid leukemia. Cell. 2012;150(2):264- 2007;25(11):1315-1321. myeloid leukemia into the vascular niche.
278. 16. Doepfner KT, Boller D, Arcaro A. Targeting Leukemia. 2014;28(10):1978-1987.
4. Ding L, Ley TJ, Larson DE, et al. Clonal receptor tyrosine kinase signaling in acute 27. Moschoi R, Imbert V, Nebout M, et al.
evolution in relapsed acute myeloid myeloid leukemia. Crit Rev Oncol Protective mitochondrial transfer from
leukaemia revealed by whole-genome Hematol. 2007;63(3):215-230. bone marrow stromal cells to acute
sequencing. Nature. 2012;481(7382):506- 17. Youn DY, Lee DH, Lim MH, et al. Bis defi- myeloid leukemic cells during chemothera-
510. ciency results in early lethality with meta- py. Blood. 2016;128(2):253-264.
5. Klco JM, Spencer DH, Miller CA, et al. bolic deterioration and involution of spleen 28. Wang J, Liu X, Qiu Y, et al. Cell adhesion-
Functional heterogeneity of genetically and thymus. Am J Physiol Endocrinol mediated mitochondria transfer con-
defined subclones in acute myeloid Metab. 2008;295(6):E1349-1357. tributes to mesenchymal stem cell-induced
leukemia. Cancer Cell. 2014;25(3):379-392. 18. Kim IK, Kim BS, Koh CH, et al. chemoresistance on T cell acute lym-
6. Dick JE. Acute myeloid leukemia stem Glucocorticoid-induced tumor necrosis fac- phoblastic leukemia cells. J Hematol Oncol.
cells. Ann N Y Acad Sci. 2005;1044:1-5. tor receptor-related protein co-stimulation 2018;11(1):11.
7. Hope KJ, Jin L, Dick JE. Acute myeloid facilitates tumor regression by inducing IL- 29. Zuber J, Radtke I, Pardee TS, et al. Mouse
leukemia originates from a hierarchy of 9-producing helper T cells. Nat Med. models of human AML accurately predict
leukemic stem cell classes that differ in self- 2015;21(9):1010-1017. chemotherapy response. Genes Dev.
renewal capacity. Nat Immunol. 2004;5 19. Kim JA, Shim JS, Lee GY, et al. 2009;23(7):877-889.
(7):738-743. Microenvironmental remodeling as a 30. Bonnet D, Dick JE. Human acute myeloid
8. Eppert K, Takenaka K, Lechman ER, et al. parameter and prognostic factor of hetero- leukemia is organized as a hierarchy that
Stem cell gene expression programs influ- geneous leukemogenesis in acute myeloge- originates from a primitive hematopoietic
ence clinical outcome in human leukemia. nous leukemia. Cancer Res. cell. Nat Med. 1997;3(7):730-737.
Nat Med. 2011;17(9):1086-1093. 2015;75(11):2222-2231. 31. Kittang AO, Hatfield K, Sand K, Reikvam
9. Guenechea G, Gan OI, Dorrell C, Dick JE. 20. Kim JH, Lee HS, Choi HK, et al. H, Bruserud O. The chemokine network in
Distinct classes of human stem cells that Heterogeneous niche activity of ex-vivo acute myelogenous leukemia: molecular
differ in proliferative and self-renewal expanded MSCs as factor for variable out- mechanisms involved in leukemogenesis
potential. Nat Immunol. 2001;2(1):75-82. comes in hematopoietic recovery. PloS and therapeutic implications. Curr Top
10. Ayala F, Dewar R, Kieran M, Kalluri R. One. 2016;11(12):e0168036. Microbiol Immunol. 2010;341:149-172.
Contribution of bone microenvironment to 21. Heuser M, Argiropoulos B, Kuchenbauer F, 32. Hassan HT, Zander A. Stem cell factor as a
leukemogenesis and leukemia progression. et al. MN1 overexpression induces acute survival and growth factor in human nor-
Leukemia. 2009;23(12):2233-2241. myeloid leukemia in mice and predicts mal and malignant hematopoiesis. Acta
11. Konopleva M, Konoplev S, Hu W, ATRA resistance in patients with AML. Haematol. 1996;95(3-4):257-262.
Zaritskey AY, Afanasiev BV, Andreeff M. Blood. 2007;110(5):1639-1647. 33. Mazur G, Wrobel T, Butrym A, Kapelko-
Stromal cells prevent apoptosis of AML 22. Kroon E, Krosl J, Thorsteinsdottir U, Baban Slowik K, Poreba R, Kuliczkowski K.
cells by up-regulation of anti-apoptotic S, Buchberg AM, Sauvageau G. Hoxa9 Increased monocyte chemoattractant pro-
proteins. Leukemia. 2002;16(9):1713-1724. transforms primary bone marrow cells tein 1 (MCP-1/CCL-2) serum level in acute
12. Katsumi A, Kiyoi H, Abe A, et al. FLT3/ through specific collaboration with Meis1a myeloid leukemia. Neoplasma. 2007;54(4):
ITD regulates leukaemia cell adhesion but not Pbx1b. EMBO J. 1998;17(13):3714- 285-289.
through alpha4beta1 integrin and Pyk2 sig- 3725. 34. Yang J, Liu X, Nyland SB, et al. Platelet-
nalling. Eur J Haematol. 2011;86(3):191- 23. Anders S, Pyl PT, Huber W. HTSeq--a derived growth factor mediates survival of
198. Python framework to work with high- leukemic large granular lymphocytes via an
13. Ninomiya M, Abe A, Katsumi A, et al. throughput sequencing data. autocrine regulatory pathway. Blood.
Homing, proliferation and survival sites of Bioinformatics. 2015;31(2):166-169. 2010;115(1):51-60.
human leukemia cells in vivo in immunod- 24. Robinson MD, McCarthy DJ, Smyth GK. 35. Akashi K, Harada M, Shibuya T, et al.

H.R. Lee et al.
Effects of interleukin-4 and interleukin-6 on cells from bone marrow lack expression of Proc Natl Acad Sci U S A. 2011;108
the proliferation of CD34+ and CD34- CD44 protein. J Biol Chem. 2012;287(31): (19):7950-7955.
blasts from acute myelogenous leukemia. 25795-25807. 54. da Silva-Diz V, Lorenzo-Sanz L, Bernat-
Blood. 1991;78(1):197-204. 44. Peault B, Weissman IL, Buckle AM, Peguera A, Lopez-Cerda M, Munoz P.
36. Piccaluga PP, Rossi M, Agostinelli C, et al. Tsukamoto A, Baum C. Thy-1-expressing Cancer cell plasticity: impact on tumor pro-
Platelet-derived growth factor alpha medi- CD34+ human cells express multiple gression and therapy response. Semin
ates the proliferation of peripheral T-cell hematopoietic potentialities in vitro and in Cancer Biol. 2018;53:48-58.
lymphoma cells via an autocrine regulatory SCID-hu mice. Nouv Rev Fr Hematol. 55. McGranahan N, Swanton C. Biological and
pathway. Leukemia. 2014;28(8):1687-1697. 1993;35(1):91-93. therapeutic impact of intratumor hetero-
37. Tsuyada A, Chow A, Wu J, et al. CCL2 45. Buccisano F, Rossi FM, Venditti A, et al. geneity in cancer evolution. Cancer Cell.
mediates cross-talk between cancer cells CD90/Thy-1 is preferentially expressed on 2015;27(1):15-26.
and stromal fibroblasts that regulates breast blast cells of high risk acute myeloid 56. Sprouffske K, Athena Aktipis C, Radich JP,
cancer stem cells. Cancer Res. 2012;72(11): leukaemias. Br J Haematol. 2004;125(2): Carroll M, Nedelcu AM, Maley CC. An
2768-2779. 203-212. evolutionary explanation for the presence
38. Ding W, Knox TR, Tschumper RC, et al. 46. Yamazaki H, Nishida H, Iwata S, Dang NH, of cancer nonstem cells in neoplasms. Evol
Platelet-derived growth factor (PDGF)- Morimoto C. CD90 and CD110 correlate Appl. 2013;6(1):92-101.
PDGF receptor interaction activates bone with cancer stem cell potentials in human 57. Jamal-Hanjani M, Quezada SA, Larkin J,
marrow-derived mesenchymal stromal T-acute lymphoblastic leukemia cells. Swanton C. Translational implications of
cells derived from chronic lymphocytic Biochem Biophys Res Commun. tumor heterogeneity. Clin Cancer Res.
leukemia: implications for an angiogenic 2009;383(2):172-177. 2015;21(6):1258-1266.
switch. Blood. 2010;116(16):2984-2993. 47. Saito Y, Kitamura H, Hijikata A, et al. 58. Quek L, David MD, Kennedy A, et al.
39. Schulz A, Toedt G, Zenz T, Stilgenbauer S, Identification of therapeutic targets for qui- Clonal heterogeneity of acute myeloid
Lichter P, Seiffert M. Inflammatory escent, chemotherapy-resistant human leukemia treated with the IDH2 inhibitor
cytokines and signaling pathways are asso- leukemia stem cells. Sci Transl Med. enasidenib. Nat Med. 2018;24(8):1167-
ciated with survival of primary chronic 2010;2(17):17ra19. 1177.
lymphocytic leukemia cells in vitro: a dom- 48. Oh IH, Kwon KR. Concise review: multiple 59. Peña-Martínez P, Eriksson M,
inant role of CCL2. Haematologica. niches for hematopoietic stem cell regula- Ramakrishnan R, et al. Interleukin 4
2011;96(3):408-416. tions. Stem Cells. 2010;28(7):1243-1249. induces apoptosis of acute myeloid
40. Juneja HS, Schmalsteig FC, Lee S, Chen J. 49. Wei Q, Frenette PS. Niches for hematopoi- leukemia cells in a Stat6-dependent man-
Vascular cell adhesion molecule-1 and etic stem cells and their progeny. Immunity. ner. Leukemia. 2018;32(3):588-596.
VLA-4 are obligatory adhesion proteins in 2018;48(4):632-648. 60. Schepers K, Pietras EM, Reynaud D, et al.
the heterotypic adherence between human 50. Mendez-Ferrer S, Michurina TV, Ferraro F, Myeloproliferative neoplasia remodels the
leukemia/lymphoma cells and marrow et al. Mesenchymal and haematopoietic endosteal bone marrow niche into a self-
stromal cells. Exp Hematol. 1993;21(3):444- stem cells form a unique bone marrow reinforcing leukemic niche. Cell Stem Cell.
450. niche. Nature. 2010;466(7308):829-834. 2013;13(3):285-299.
41. Morikawa S, Mabuchi Y, Kubota Y, et al. 51. Kwon KR, Ahn JY, Kim MS, Jung JY, Lee JH, 61. Zhang B, Ho YW, Huang Q, et al. Altered
Prospective identification, isolation, and sys- Oh IH. Disruption of bis leads to the dete- microenvironmental regulation of leukemic
temic transplantation of multipotent mes- rioration of the vascular niche for and normal stem cells in chronic myeloge-
enchymal stem cells in murine bone mar- hematopoietic stem cells. Stem Cells. nous leukemia. Cancer Cell. 2012;21(4):
row. J Exp Med. 2009;206(11):2483-2496. 2010;28(2):268-278. 577-592.
42. Pinho S, Lacombe J, Hanoun M, et al. 52. Sacchetti B, Funari A, Michienzi S, et al. 62. Byrd JC, Mrozek K, Dodge RK, et al.
PDGFRalpha and CD51 mark human Self-renewing osteoprogenitors in bone Pretreatment cytogenetic abnormalities are
nestin+ sphere-forming mesenchymal stem marrow sinusoids can organize a predictive of induction success, cumulative
cells capable of hematopoietic progenitor hematopoietic microenvironment. Cell. incidence of relapse, and overall survival in
cell expansion. J Exp Med. 2013;210(7): 2007;131(2):324-336. adult patients with de novo acute myeloid
1351-1367. 53. Chaffer CL, Brueckmann I, Scheel C, et al. leukemia: results from Cancer and
43. Qian H, Le Blanc K, Sigvardsson M. Normal and neoplastic nonstem cells can Leukemia Group B (CALGB 8461). Blood.
Primary mesenchymal stem and progenitor spontaneously convert to a stem-like state. 2002;100(13):4325-4336.

Hematopoiesis ARTICLE
Aging of human hematopoietic stem cells is Ferrata Storti Foundation

linked to changes in Cdc42 activity
Amanda Amoah,1 Anja Keller,1° Ramiz Emini,2° Markus Hoenicka,2 Andreas
Liebold,2 Angelika Vollmer,1 Karina Eiwen,1 Karin Soller,1 Vadim Sakk,1 Yi
Zheng,3 Maria Carolina Florian1° and Hartmut Geiger1
1
Institute of Molecular Medicine, Ulm University, Ulm, Germany; 2Department of
Cardiothoracic and Vascular Surgery, Ulm University Hospital, Ulm, Germany and
3
Cincinnati Children’s Hospital Medical Center and University of Cincinnati, Cincinnati,
OH, USA
Haematologica 2022
°AK current address: Zentrum für Integrative Psychiatrie, Lübeck Campus, Lübeck, Germany Volume 107(2):393-402
°RE current address: Herzzentrum, Uniklinik Köln, Köln, Germany
°MCF current address: Bellvitge Institute for Biomedical Research, IDIBELL, Barcelona, Spain
ABSTRACT
I
n this study, we characterize age-related phenotypes of human
hematopoietic stem cells (HSC). We report increased frequencies of
HSC, hematopoietic progenitor cells and lineage negative cells in the
elderly but a decreased frequency of multi-lymphoid progenitors. Aged
human HSC further exhibited a delay in initiating division ex vivo though
without changes in their division kinetics. The activity of the small
RhoGTPase Cdc42 was elevated in aged human hematopoietic cells and
we identified a positive correlation between Cdc42 activity and the fre-
quency of HSC upon aging. The frequency of human HSC polar for
polarity proteins was, similar to the mouse, decreased upon aging, while
inhibition of Cdc42 activity via the specific pharmacological inhibitor of
Cdc42 activity, CASIN, resulted in re-polarization of aged human HSC
with respect to Cdc42. Elevated activity of Cdc42 in aged HSC thus con-
tributed to age-related changes in HSC. Xenotransplant, using NBSGW
mice as recipients, showed elevated chimerism in recipients of aged
compared to young HSC. Aged HSC treated with CASIN ex vivo dis-
played an engraftment profile similar to recipients of young HSC. Taken
Correspondence:
together, our work reveals strong evidence for a role of elevated Cdc42
activity in driving aging of human HSC, and similar to mice, this presents HARTMUT GEIGER,
a likely possibility for attenuation of aging in human HSC. hartmut.geiger@uni-ulm.de

Introduction
Accepted: December 22, 2020.
Aging is associated with tissue degeneration, aging-related diseases and an Pre-published: January 14, 2021.
increased susceptibility to infections.1,2 These hallmarks of aging have been linked
to aging-related changes within somatic stem cell compartments, and primarily
investigated in animal models like mice.3,4 One of the most extensively studied https://doi.org/10.3324/haematol.2020.269670
somatic stem cell-based system is the hematopoietic system. Hematopoietic stem
cells (HSC) maintain blood homeostasis and show an age-related decline in overall
function in mice,5 which includes an increase in myelopoiesis,6 accumulation of
DNA damage,7 changes in epigenomic and transcriptional programs,8 decreased Material published in Haematologica is covered by copyright.
cell polarity and aberrant activity of the small RhoGTPase Cdc42.9 Although sig- All rights are reserved to the Ferrata Storti Foundation. Use of
nificant progress has been achieved in elucidating mechanisms of aging of murine conditions:
HSC, it remains unclear whether these mechanisms can be simply extrapolated to https://creativecommons.org/licenses/by-nc/4.0/legalcode.
other species, including humans. Early studies on HSC number and function in Copies of published material are allowed for personal or inter-
larger mammals showed, based on stochastic modeling, clear differences in HSC nal use. Sharing published material for non-commercial pur-
biology and aging to the murine model.10,11 These data suggested that HSC from https://creativecommons.org/licenses/by-nc/4.0/legalcode,
non-human primates cycle more slowly and that fewer numbers of HSC clones sect. 3. Reproducing and sharing published material for com-
actively contribute to hematopoiesis in humans at steady-state than in mice. This mercial purposes is not allowed without permission in writing
connotes that non-human primates and murine HSC may undergo different stress from the publisher.
intensities such as the rate of accruing DNA damage and, as a result, may exhibit
dissimilarities in aging.

A. Amoah et al.
In addition, mice show an increase in HSC frequency Colony forming unit assay
with age, while the rhesus monkey, shows a decrease In order to assess the myeloid and erythroid generative poten-
with age.12 Moreover, even among distinct murine inbred tial of samples, 200 HSC were seeded in methylcellulose medi-
strains, HSC number and function is distinct upon aging, um (further details are provided in the Online Supplementary
like C57BL/6 mice present with an elevated number of Methods).
HSC upon aging, but not so in DBA/2 animals.13,14 For
these reasons, novel studies into understanding mecha- Xenotransplantation
nisms of aging of human HSC are warranted and are a pre- All animal experiments were carried out in accordance to
requisite to bolster the transition of this knowledge into institutional guidelines and approved by the
the clinic. Regierungspräsidium Tübingen (TVA 1412). Five hundred HSC
Age-related changes in the frequency and function of were injected via the tail vein into non-conditioned or low dose
HSPC have been in part previously described by a small irradiated (1.6 Gy) NBSGW mice. At 8 and 12 weeks, aspirates
number of groups. One study for example reported no were drawn from the bone marrow (BM) of mice after adminis-
changes in the re-population potential of aged HSC and a tering anesthesia. Human cells were identified using human-spe-
decreased propensity for myeloid differentiation while cific antibodies (see the Online Supplementary Methods for further
another recorded a decline in the reconstitution capacity details) and analyzed on LSR Fortessa SORP flow cytometer (BD
of aged HSC with an increased myeloid differentiation Biosciences). Human chimerism was determined as a percentage
potential.15,16 Both groups used NOD.Cg-Prkdcscid of total CD45+ cells and mature cells, as a percentage of human
Il2rgtm1Wjl/SzJ (NSG) animals as recipients for in vivo xeno- CD45+ cells.
transplantation to study the function of human HSC.
This model requires irradiation of the recipient animal for CASIN treatment
the successful establishment of xenotransplants17 which HSC were collected in serum-free expansion media and incu-
might contribute to variable secondary effects not linked bated at 37°C, 3% oxygen for 1 hour. Cells were then trans-
to the transplanted HSC. New and improved mouse ferred into media ± CASIN, incubated for 4 hours and washed.
models have been created that do not require pre-condi- Cells were then used in subsequent experiments.
tioning of the recipients for achieving human xeno-
chimerism.18,19 NOD.Cg-KitW-41J Tyr +
Prkdcscid Immunofluorescent staining
Il2rgtm1Wjl/ThomJ (NBSGW) animals bear in addition to the Cells were seeded in serum-free expansion media, fixed and
NSG genotype a mutation in the Kit gene.19 The Kit muta- polarity for Cdc42 or tubulin assessed as previously described by
tion enables donor cells to efficiently engraft without irra- Florian et al.9 Cdc42-GTP in HSC was determined using the anti-
diation.20 We characterize here aged human HSC with a body described by Althoff et al.22 (see the Online Supplementary
special focus on likely shared hallmarks of age-related Methods for further details).
changes among human and murine HSC and describe a
novel approach to attenuate aging of human HSC. Our Western blot
data support the possibility of rejuvenating the function A Rac/Cdc42 assay reagent (# 14-325, Millipore) was used in
of aged human HSC due to similarities between aging of pull down assays according to the manufacturer’s protocol (see
murine and human HSC. the Online Supplementary Methods for further details).
Statistical analysis
Methods Statistical analyses were performed with GraphPad Prism 8
(version 8.1.2) and are presented as mean ± standard deviation
Primary cells (SD) or mean ± standard error of the mean (SEM) and box plots
Bone marrow cells were isolated from young (range, 23-39 as minimum and maximum points. Kendall’s correlation analy-
years; median age 27 years) donors acquired from Cincinnati sis was performed with R version 4.0.3, RStudio Team (2020)
Children’s Hospital Medical Center and aged (range, 58-82 version 1.3.1093. *P<0.05, **P<0.005, ***P<0.0005.
years; median age 66 years) individuals undergoing heart sur-
gery at the Ulm University Clinic, Department of Heart,
Thoracic and Vascular Surgery (additional details of age strata Results
are provided in the Online Supplementary Methods). All donors
were hematologically healthy. Sample collection and investiga- Changes in the immunophenotypic frequencies of
tion was approved by the Internal Review Board hematopoietic populations occur with age
(Ethikkomission) of Ulm University (392/16). Data on whether there are changes in the frequency of
HSC in the BM of humans upon aging remains controver-
Flow cytometric analysis and cell sorting sial. This might be, at least in part, due to different gating
Mononuclear cells (MNC) were thawed and stained in phos- strategies employed to identify human HSC.15,16 Using a
phate-buffered saline (PBS) supplemented with 3% fetal bovine more recently established and improved marker profile
serum (FBS) with human specific antibodies (see the Online for the identification of human HSC21 (Online
Supplementary Methods for details). Different cell populations Supplementary Figure S1A), we first determined the fre-
were identified and sorted on a BD FACS ARIA II 4L SORP (BD quency of HSC (Lin-CD34+SSc low CD38-CD90+CD45ra-),
Biosciences) according to the markers used by van Galen et al.21 HSPC (Lin-CD34+SSc low CD38-) and Lin-CD34+SSc low cells
in BM cells from the sternum of the elderly. While the fre-
Single cell division assay quency of the Lin-CD34+SSc low population did not change
Single HSC were sorted into Terasaki plates and checked with age within our cohort, the HSPC frequency within
every 12 hours under a light microscope (further details are pro- Lin-CD34+SSc low population and HSC frequency within
vided in the Online Supplementary Methods). HSPC population increased with age (Figure 1A to C).

Cdc42 and aging of human HSC
A B C D
E F
Figure 1. Flow cytometric analysis of different bone marrow populations isolated from young (white) and aged (grey) donors. Populations of interest within the low-
density mononuclear cell fraction (MNC) were identified and frequencies of (A) Lin-CD34+ssc low, (B) hematopoietic stem progenitor cells (HSPC), (C) hematopoietic
stem cells (HSC) and (D) Lin- cells from young and aged donors were determined. *P<0.04, **P<0.005; Mann-Whitney and t-test with Welch’s correction. Bars rep-
resent the mean ± standard error of the mean (SEM). (E) Representative image of common myeloid progenitor/ megakaryocyte–erythroid progenitor (CMP/MEP),
multipotent progenitor (MPP) and multipotent lymphoid progenitor (MLP) gates. (F) The frequency of CMP/MEP, MPP and MLP in the Lin- fraction of donors.
**P=0.003; Mann-Whitney test. Bars represent the mean ± SEM. 17< n >20; 23< n >29. Donor age: young =23-39 years (yr), median =27 yr; aged =58-82 yr,
young aged
median =65 yr. n : number of young donors; n : number of aged donors.

young aged
These results support an increase in the HSPC population Aged hematopoietic stem cells are delayed in
with age previously described for the iliac crest.16,23 Our the initiation of division
observations further indicate that the age-related increase We next tested whether the age-related increase in the
in frequency is not restricted to a single anatomical site or, HSC frequency might be linked to an elevated division
in our studies, influenced by sex (Online Supplementary rate of aged HSC. To this end, we determined the dynam-
Figure S1B and C). In addition, the frequency of HSC ics of first or second divisions of individual HSC ex vivo
within the Lin-CD34+SSc low population in the sternum was (Figure 2A). In general, BM-derived HSC showed a
also significantly higher than in the young16 (Online delayed initiation of division when compared to HSC
Supplementary Figure S1D). We also found a not yet from cord blood (CB) or HSC mobilized to blood (Online
described increase in the frequency of lineage negative Supplementary Figure S2A and B). Surprisingly, aged HSC
cells within the MNC population in aged donors (Figure actually showed a delay until the first 50% of HSC under-
1D). The frequency of common myeloid progenitor/ went their first division compared to young HSC (Online
megakaryocyte–erythroid progenitor (CMP/MEP, Lin- Supplementary Figure S2B and C). This delay in initiation of
CD34+SSc low CD38+CD90-CD45ra-) and multipotent pro- division of aged HSC was still imminent in the presence of
genitor (MPP, Lin-CD34+SSc low CD38-CD90-CD45ra-) did a different combination of cytokines (Online Supplementary
not change upon aging, while the frequency of Figure S2C to E) as well as under normoxic conditions
immunophenotypic multipotent lymphoid progenitor (Online Supplementary Figure S2 F and G). The overall rate
(MLP, Lin-CD34+SSc low CD38-CD90-/low CD45ra+) with- of division after initiation though was similar for both
in the Lin- population decreased significantly (Figure 1E young and aged HSC for both the first division (Figure 2E)
and F). Our results demonstrate and confirm that aging is as well as the second division (Figure 2D to E; Online
associated with an increase in the frequency of Supplementary Figure S2E). We next examined the propor-
hematopoietic progenitor and HSC, but with a decrease tion of cells in the distinct phases of the cell cycle by stain-
in the frequency of MLP. ing for DNA (Hoechst 33342) and Ki-67 (Ki-67 antibody)

A. Amoah et al.
A B C
D E
Figure 2. Kinetics of single cell division of young (blue) and aged (green) hematopoietic stem cells cultured ex vivo. (A) Experimental design of cells singly sorted
into plates containing M4 media and incubated at 37°C, 3% oxygen. (B) Cumulative first division of live young and aged hematopoietic stem cells (HSC) and (C) time
by which 50% of the cells have undergone the first division. *P=0.03; t-test with Welch’s correction. Bars represent the mean ± standard error the mean (SEM). (D)
Cumulative second division of live young and aged HSC. (E) Slope was derived as linear regression fits and probability values calculated from the correlation coeffi-
cients (M4 calculations in the box). Each curve was derived from cumulative gaussian fits with robust regression. n =6; n =5 different donors. Donor age: young
young aged
=27-39 years (yr), median =28 yr; aged =64-75 yr, median =69 yr.
(Online Supplementary Figure S3A and B). The proportions standard pull down assay (Figure 3A; Online Supplementary
of young and aged HSC in the G , G and S-G -M phases of
0 1 2 Figure S4A). Pull-down replicates on the same samples
the cell cycle were similar, hence the delayed initiation of from randomized donors demonstrated overall high
division was not simply driven by a higher frequency of reproducibility of the pull-down assay also on primary
quiescent cells upon aging. In aggregation, aged HSC human samples (Online Supplementary Figure S4B). We
show a delayed initiation of division that is not linked to observed an approximate 4-fold increase in the relative
a higher frequency of cells in G or G , while once division
0 1 Cdc42 activity in LDBM cells from elderly donors com-
is initiated, there is no difference in the overall kinetics of pared to the young (Figure 3B) which was re-affirmed in
the division. correlation analysis of Cdc42 activity and age (Online
Supplementary Figure S4C). Using immunofluorescence
Cdc42 activity is increased in human hematopoietic imaging, we found aged HSC had significantly higher
stem cells from the elderly and correlates with an Cdc42-GTP expression than young HSC (Figure 3C to E;
increase in hematopoietic stem cell frequency Online Supplementary Figure S4D), implying primitive stem
The small RhoGPTase Cdc42 cycles between an active cells also undergo Cdc42 activity changes upon aging. In
(GTP bound) and inactive (GDP bound) state24 and has addition, we identified a positive association between
been shown to have essential roles in HSC regulation.25 Cdc42 activity and HSC frequency but not between
We previously demonstrated that the active form (GTP- Cdc42 activity and HSPC frequency (Figure 3F and G).
bound form) of the small RhoGTPase Cdc42 was Furthermore, the spread of values and, thus, the standard
increased in murine low-density bone marrow (LDBM) deviation (Figure 3B) was higher in the aged group than in
cells as well as in HSC upon aging, and that this increase the young, which points to a strong increase in hetero-
in HSC resulted in the age-related increase in HSC fre- geneity of the hematopoietic system in individuals upon
quency in mice.9 We therefore determined the level of aging. Heterogeneity upon aging in mice is, for example,
activity of Cdc42 in human LDBM from the elderly by a less pronounced due to their inbred nature.26–28 In aggrega-

A B
C D E
F G
Figure 3. Relative expression of Cdc42 activity in young and aged hematopoietic stem cells. (A) Representative image of western blot. (B) Quantitative expression
of relative Cdc42 activity normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH). *P= 0.010, if the two points with activity higher than 20 are excluded,
then P=0.017; Mann-Whitney test. Bars represent the mean ± standard error of the mean (SEM). n =13; n =41. Donor age: young =23-39 years (yr), median = 27
young aged
yr; aged = 60-82 yr, median = 66 yr. (C) Representative confocal image of Cdc42-GTP expression in hematopoietic stem cells (HSC). Cdc42-GTP quantification of
young and aged HSC normalized to (D) DAPI intensity and (E) cell size. ***P=0.0001, **P=0.0012; Mann-Whitney test. Bars = mean ± SEM and n =66; n =67,
young aged
from three different donors per cohort. Donor age: young =27-31 years (yr), median =27 yr; aged =63-76 yr, median =76 yr. Scale bar represents 2 mm. (F) Correlation
analysis (Spearman) of relative Cdc42 activity and HSC frequency (r=0.4, P= 0.05, n=37) and (G) hematopoietic stem progenitor cell (HSPC) frequency (r= 0.3,
P=0.175, n=22), broken grey lines represent 95% Confidence Interval.
tion, the data show that similar to mice, primitive Cdc42 and tubulin by immunofluorescence analyses
hematopoietic cells from elderly humans show elevated (Figure 4B to D). Approximately 70% of young HSC
Cdc42 activity. The level of Cdc42 activity in individuals showed a polar distribution of Cdc42 while approximately
correlates positively with the frequency of HSC, support- 70% of the aged HSC showed an apolar distribution
ing a possible role for Cdc42 activity in causing the elevat- (Figure 4C). The frequency of aged HSC polar for tubulin
ed HSC frequency, similar to what has been described in was also reduced (Figure 4D). Our findings establish that
the mouse.8,29 aged human HSC present with a reduced frequency of cells
polar for polarity proteins. Oddly, we did not find a signif-
The frequency of hematopoietic stem cells polar for icant association between Cdc42 polarity and tubulin
Cdc42 and tubulin declines upon aging polarity in human HSC (Online Supplementary Figure S5A)
Another established age-related hallmark for murine suggesting these two parameters might not be directly cor-
HSC is the reduction in frequency of cells polar for cytoso- related as has been shown for murine HSC.9 Nonetheless,
lic polarity proteins like tubulin and Cdc42 (Figure 4A). we observed a strong negative association between Cdc42
This “apolarity” of aged murine HSC is a direct conse- polarity and age (Online Supplementary Figure S5B), imply-
quence of the elevated activity of Cdc42 itself in aged HSC9 ing the frequency of cells that remain polar decreases with
and likely results in a change in the mode of the division of increasing age. In order to determine whether the level of
aged murine HSC.8 We therefore determined the frequency Cdc42 activity in human cells might be linked to the fre-
of aged human HSC that showed a polar distribution of quency of cells polar for Cdc42, linear regression analyses

A. Amoah et al.
of Cdc42 activity and Cdc42 polarity were performed on for Cdc42, though not fully to the frequency seen in
human HSC. The data revealed that indeed Cdc42 activity young HSC (Figure 4C). This data confirms that elevated
and the frequency of polar HSC were inversely correlated activity of Cdc42 is, at least in part, causative for the
(Figure 4E), suggesting a causative link between Cdc42 reduced polarity of aged human HSC similar to aged
activity and polarity also for human HSC. murine HSC.8,9,32
Aged hematopoietic stem cells are repolarized CASIN treatment induces early division onset in aged
by pharmacological inhibition of Cdc42 activity hematopoietic stem cells
CASIN (Cdc42 activity specific inhibitor) is a pharma- We previously demonstrated that HSC from aged
cological compound that specifically inhibits the activity donors were delayed in division onset in comparison to
of Cdc42.30,31 We thus tested whether inhibition of Cdc42 HSC isolated from the young (Figure 2B and C), hence we
activity in aged human HSC via CASIN might result in assessed the effect CASIN treatment would have on the
an increase in the frequency of polar HSC. CASIN indeed division of aged HSC. Our results show that upon CASIN
increased the frequency of HSC polar for Cdc42 in a treatment, aged HSC commence division hours before
dose-dependent fashion (Figure 4F) but not for tubulin untreated HSC (Figure 4G) with no additional effect on
(Online Supplementary Figure S5C). Inhibition of Cdc42 the second division (Online Supplementary Figure S6D to E).
activity, thus, increases the frequency of aged HSC polar We observed that the time taken for 50% of untreated
A B
C D E
F G H
Figure 4. Hematopoietic stem cells polarity assessment and division kinetics in the presence and absence of CASIN. (A) Illustration of a polar and apolar cell. (B)
Representative immunofluorescent images taken with a confocal microscope. Quantification of the proportion of young (white) and aged (grey) donor cells polar for
(C) Cdc42 and (D) tubulin. ***P=0.00018, **P=0.002; t-test, false discovery rate approach. Bars represent the mean ± standard error of the mean (SEM). Scale
bar represents 2 mm; n >5; n >8. Donor age: young =26-39 years (yr), median =28 yr; aged =58-82 yr, median =65 yr. (E) Correlation analysis (Pearson) of Cdc42
young aged
polarity and relative Cdc42 activity (r=0.6, P=0.03, n =13), broken grey lines represent 95% Confidence Interval. (F) Quantification of Cdc42 polarity of hematopoietic
stem cells (HSC) after treatment with CASIN. *P=0.03; Mann-Whitney test. Bars represen the mean ± SD. n >2. Donor age: aged =61-81 years (yr), median = 62
aged
yr. (G) Cumulative first division of aged HSC with and without CASIN and (H) time by which 50% of the cells have undergone the first division. Curve was derived from
cumulative gaussian fits with robust regression. n and n
aged =4. Donor age: aged =63-71 yr, median =66 yr.
aged+CASIN

HSC to undergo first division was considerably longer derived HSC result in a much lower level of chimerism in
than cells treated with 5 mM or 10 mM CASIN albeit not comparison to transplants with CB-derived HSC (data not
statistically significant (P=0.06, Figure 4H) suggesting the shown). Animals transplanted with aged human HSC
inhibition of Cdc42 activity may facilitate or support early showed a significantly higher level of engraftment (mean
division onset. 3.7%) compared to recipients that received young HSC
(mean 0.6%) and the aged + CASIN groups (mean 1.6%,
Aged hematopoietic stem cells show an elevated level 2.1%) (Figure 5A). Chimerism driven by aged HSC further
of chimerism but balanced differentiation in NBSGW increased at 12 weeks post-transplantation (mean 5%)
xenotransplants compared to chimerism driven by young HSC (Figure 5B).
Murine xenotransplant models that are used to test the This result was surprising, as the small number of previous
potential of human HSC usually require irradiation of the studies in which aged human cells were transplanted into
recipient animal for establishing human engraftment.17 irradiated recipients demonstrated a lower16 or at least sim-
Irradiation of the recipients though might contribute to ilar chimerism15 stemming from aged compared to young
variable secondary effects not linked to the potential of human HSC. In order to test whether irradiation of recipi-
the transplanted HSC. In order to circumvent the necessi- ents might influence the reconstitution potential of aged
ty for irradiation of the recipients in xenotransplants, we HSC, we also transplanted young and aged human HSC
utilized NBSGW animals as recipients33 to assess the func- into irradiated (1.6 Gy) NBSGW mice. Recipients of aged
tion of young, aged and CASIN-treated aged human HSC HSC displayed similar levels of chimerism compared to
in vivo. recipients of young HSC, and only marginal differences in
To this end, 500 HSC were injected via the tail vein into their differentiation profile (Online Supplementary Figure
mice and BM chimerism was analyzed at week 8 and 12 S6A to C). Irradiation of recipients might thus influence the
post xenotransplantation. Mice were considered to be suc- reconstitution potential of aged human HSC in xenotrans-
cessfully engrafted when the level of human hematopoiet- plantation experiments. In summary, our results demon-
ic cells (huCD45+) detected was higher than 0.1% of total strate that aged HSC show an overall elevated contribution
CD45+ cells.34 In general, transplants with human BM- to BM upon transplantation into NBSGW animals.
A B
C D
Figure 5. Analysis of xenotransplantation experiment under non-irradiated settings. Frequency of human cells in the bone marrow (BM) of NBSGW mice at (A) 8
weeks and (B) 12 weeks post-transplant. (C) Emergence of human myeloid cells and (D) B cells post-transplant from young (blue), aged (green) and CASIN-treated
aged (orange, brown) -recipients. Red line represents averages of non-transplanted controls. Bars represent the mean "±" standard deviation (SD). n = 6 different
donors per cohort, 46 mice at 8 weeks and 42 mice at 12 weeks. Donor age: young =24-35 years (yr), median =27 yr; aged =60-81 yr, median =62 yr.

A. Amoah et al.
Aged HSC treated with CASIN at 5 mM or 10 mM were competition is quickly resolved, with only one axis matur-
more similar in their contribution to chimerism to young ing and recruiting other proteins such as Bem1, to facilitate
HSC at 8 weeks post transplantation. Aged HSC treated division.37 It is therefore a possibility that also in aged
with the higher concentration of CASIN, 10 mM, were human HSC, the correct allocation of proteins to distinct
able to sustain a significantly lower, young-like contribu- positions within cells that are critical for proper initiation
tion to chimerism at 12 weeks post transplantation. This of division, might be altered upon elevated activity of
mirrors our finding that aged HSC, treated with 10 mM of Cdc42, resulting in a delay.
CASIN, showed the highest increase in the repolarization Knowledge on the molecular mechanisms underlying
of aged HSC (Figure 4F). Assessment of the frequency of aging of HSC have supported the development of thera-
human myeloid (CD33+) and B cells (CD19+) in the BM peutic approaches to mitigate aging of HSC. For example,
of recipient mice revealed no significant differences CASIN shows great promise in rejuvenating aged HSC9
among any of the experimental groups (Figure 5C and D) and also in reverting aging-associated immune remodel-
implying that aged human HSC, at least in NBSGW ani- ing (AAIR) in murine models.32 A multivariate analysis
mals, show no differentiation skewing as reported for model of gene expression profiles and biological age iden-
aged murine HSC. tified Cdc42 as a strong predictor of survival and that a
Colony forming unit assays also revealed the myelo- higher Cdc42 level is associated with higher mortality.38
erythroid potential of aged and CASIN-treated aged HSC We showed previously that the level of Cdc42 activity in
were preserved (Online Supplementary Figure S6F to H) human blood cells correlates positively with age39 and in
which is in line with the report by Pang et al.16 this study, we identified a similar association in LDBM
In summary, our xenotransplantation experiments cells. Aged human HSC also show elevated activity of
demonstrate that, to our surprise, aged human HSC are Cdc42 and a low frequency of HSC polar for polarity pro-
able to confer an elevated level of chimerism upon trans- teins. We also report a negative correlation between the
plantation into NBSGW animals, which is likely a result of activity of Cdc42 in individual donors and the frequency
the elevated activity of Cdc42 in aged HSC, as aged HSC of HSC polar for polarity proteins like Cdc42 in that
treated with 10 mM of CASIN are more similar to young donor. We could demonstrate that the inhibition of the
HSC than to aged HSC. activity of Cdc42 is sufficient to increase the frequency of
chronologically aged HSC polar for Cdc42 but not for
tubulin which suggests in human cells, Cdc42 and tubulin
Discussion axes may not be closely linked as shown in murine cells.9
Furthermore, the number of HSC in a given donor and
In this study, we characterized aging-related changes in the activity of Cdc42 in hematopoietic cells of that donor
human HSC and tested an intervention to target these are positively correlated, which suggests that Cdc42
aging-related changes. Using a well-established marker activity may affect the frequency of human HSC upon
panel for the identification of primitive hematopoietic cells aging and could therefore, directly or indirectly, con-
in BM, we show that there is an increase in the HSC fre- tribute to the increased HSC frequency observed. Given
quency and a decrease in lymphoid progenitors in BM of that in mice, apolar distribution of Cdc42 (a consequence
the sternum of the elderly. We specifically selected the low of elevated activity of Cdc42) drives HSC to divide sym-
side scatter CD34+ cells as our reference population in metrically,8 and that elevated activity of Cdc42 in human
accordance to the guidelines of the International Society of leukemia stem cells is linked to more symmetric divi-
Hematotherapy and Graft Engineering (ISHAGE).35 sions,40 we postulate here an increased frequency of sym-
Additionally, the gating of low side scatter CD34+ cells metric divisions also for HSC of the elderly, which
minimizes variation across labs thereby increasing repro- though will require further investigations.
ducibility and reliability.35 It is a possibility that the dispar- In order to assess changes in function upon aging, HSC
ities in previous reports on changes in the frequency of were transplanted into NBSGW recipients. By employing
human HSC15,16 may result from the use of different refer- this model, secondary elements that may be introduced
ence populations. Our data also demonstrates a delayed due to irradiation were avoided.41 We noted higher
entry of aged HSC into division, an observation similarly chimerism levels in non-conditioned recipients of aged
made by Flach but not us for aged murine HSC,8,36 but no HSC whereas conditioned recipients showed no increase.
overall delay in division kinetics once the cells start to These observations also corroborate findings that irradia-
divide. This finding is in contrast to observations by Pang tion, even at low doses, induces functional changes in mes-
et al.16 which concluded that aged HSC are more prone to enchymal stromal cells which influences their hematopoi-
divide. Here, Pang determined the proportion of G and 0 etic-supporting ability.42 Our data suggests that the elevat-
non-G cells using Hoechst 33342 and Pyronin Y while we
0 ed Cdc42 activity in aged HSC drives the age-related high
used Hoechst 33342 and Ki-67. While Pyronin Y correlates level of chimerism observed in non-irradiated NBSGW
with RNA content, Ki-67 is seen as a marker for prolifera- recipients since recipients of young and 10 mM CASIN-
tion in itself. It is thus a possibility that upon aging the cor- treated aged HSC exhibited close similarities in the level of
relation between elevated RNA content and proliferation human chimerism. The higher level of chimerism driven
might be somewhat diminished. The molecular mecha- by aged human HSC further supports a model in which
nisms of the delayed entry of aged human HSC into cell human aged HSC might predominantly undergo symmet-
division upon stimulation might be linked to elevated rical divisions. Another possibility remains that an elevated
Cdc42 activity as cells treated with a pharmacological repopulation potential of aged human HSC could be driven
inhibitor of Cdc42 activity (CASIN), proceeded into divi- by the presence of individual clones43 linked to aging-relat-
sion ahead of their untreated counterparts. In mutant ed clonal hematopoiesis reported for a certain percentage
yeast, hyperactive levels of Cdc42 result in the formation of the elderly.44 Most human studies though demonstrated
of multiple budding sites, however in wild- type cells, foci clonal hematopoiesis only in peripheral blood and not

among HSC, thus it remains unclear whether changes in RE, MK and AL provided aged samples; AK and KS per-
clonality exist already among aged HSC.44–46 We did not formed experiments; KS, AV and KE assisted in cell sorting
determine in this study clonality among the small number procedures; VS supported in transplantation and bone marrow
of HSC transplanted into animals, as this remains techni- aspiration; YZ and MCF assisted in study design; AL, RE,
cally very challenging. MK and MCF reviewed and edited the manuscript.
In summary, we identified novel age-related phenotypes
of human HSC and provide evidence of inter-species paral- Acknowledgements
lels as well as differences to support translational studies in We would like to thank the Flow Cytometry Core and
the aging field. Our data supports that age-related pheno- Imaging Core Facilities and the Tierforschungszentrum at the
types that are indicators of the function of aged HSC University of Ulm for their support. We also thank Jeffrey
(Cdc42 activity, polarity, reconstitution potential in Bailey for excellent training in the bone marrow aspiration tech-
xenografts) are malleable in human HSC by inhibition of nique, Aishlin Hassan and Kalpana Nattamai for their techni-
the age-related elevated activity of Cdc42. This might cal support at Cincinnati Children’s Hospital Medical Center
therefore present a new possibility to improve autologous (CCHMC).
stem cell transplants of aged donors.
Funding
Disclosures This study was supported with funding from the DFG,
No conflicts of interest to disclose. GRK1789 (CEMMA) to HG and AA.
Contributions Data sharing statemen

AA, HG was involved in study design, interpretation and The data that support the findings of this study are available
manuscript writing; AA performed and analyzed experiments; from the corresponding author upon reasonable request.
References 13. Geiger H, Rennebeck G, Van Zant G. tor in hematopoietic stem cell maintenance.
Regulation of hematopoietic stem cell aging Cell Cycle. 2007;6(12):1444-1449.
1. Kovtonyuk L V., Fritsch K, Feng X, Manz in vivo by a distinct genetic element. Proc 26. Gekas C, Graf T. CD41 expression marks
MG, Takizawa H. Inflamm-aging of Natl Acad Sci. 2005;102(14):510-5107. myeloid-biased adult hematopoietic stem
hematopoiesis, hematopoietic stem cells, 14. Geiger H, True JM, De Haan G, Van Zant G. cells and increases with age. Blood.
and the bone marrow microenvironment. Age- and stage-specific regulation patterns in 2013;121(22):4463-4472.
Front Immunol. 2016;7:502. the hematopoietic stem cell hierarchy. 27. Glauche I, Thielecke L, Roeder I. Cellular
2. Ponnappan S, Ponnappan U. Aging and Blood. 2001;98(10):2966-2972. aging leads to functional heterogeneity of
immune function: molecular mechanisms to 15. Kuranda K, Vargaftig J, de la Rochere P, et al. hematopoietic stem cells: a modeling per-
interventions. Antioxid Redox Signal. Age-related changes in human hematopoiet- spective. Aging Cell. 2011;10(3):457-465.
2011;14(8):1551-1585. ic stem/progenitor cells. Aging Cell. 28. Akunuru S, Geiger H. Aging, clonality, and
3. De Haan G, Lazare SS. Aging of hematopoi- 2011;10(3):542-546. rejuvenation of hematopoietic stem cells.
etic stem cells. Blood. 2018;131(5):479-487. 16. Pang WW, Price EA, Sahoo D, et al. Human Trends Mol Med. 2016;22(8):701-712.
4. Geiger H, de Haan G, Florian MC. The age- bone marrow hematopoietic stem cells are 29. Klose M, Florian MC, Gerbaulet A, Geiger H,
ing haematopoietic stem cell compartment. increased in frequency and myeloid-biased Glauche I. Hematopoietic stem cell dynam-
Nat Rev Immunol. 2013;13(5):376-389. with age. Proc Natl Acad Sci U S A. ics are regulated by progenitor demand: les-
5. Chambers SM, Shaw CA, Gatza C, Fisk CJ, 2011;108(50):20012-20017. sons from a quantitative modeling approach.
Donehower LA, Goodell MA. Aging 17. Doulatov S, Notta F, Laurenti E, Dick JE. Stem Cells. 2019;37(7):1-22.
hematopoietic stem cells decline in function Hematopoiesis: a human perspective. Cell 30. Liu W, Du W, Shang X, et al. Rational identi-
and exhibit epigenetic dysregulation. PLoS Stem Cell. 2012;10(2):120-136. fication of a Cdc42 inhibitor presents a new
Biol. 2007;5(8):1750-1762. 18. Cosgun KN, Rahmig S, Mende N, et al. Kit regimen for long-term hematopoietic stem
6. Liang Y, Van Zant G, Szilvassy SJ. Effects of regulates HSC engraftment across the cell mobilization. Leukemia. 2019;33(3):749-
aging on the homing and engraftment of human-mouse species barrier. Cell Stem 761.
murine hematopoietic stem and progenitor Cell. 2014;15(2):227-238. 31. Du W, Liu W, Mizukawa B, et al. A non-
cells. Blood. 2005;106(4):1479-1487. 19. McIntosh BE, Brown ME, Duffin BM, et al. myeloablative conditioning approach for
7. Beerman I, Seita J, Inlay MA, Weissman IL, Nonirradiated NOD,B6.SCID Il2rγ-/- long-term engraftment of human and mouse
Rossi DJ. Quiescent hematopoietic stem kitW41/W41(NBSGW) mice support multi- hematopoietic stem cells. Leukemia. 2018;32
cells accumulate DNA damage during aging lineage engraftment of human hematopoiet- (9):2041-2046.
that is repaired upon entry into cell cycle. ic cells. Stem Cell Rep. 2015;4(2):171-180. 32. Leins H, Mulaw M, Eiwen K, et al. Aged
Cell Stem Cell. 2014;15(1):37-50. 20. Waskow C, Madan V, Bartels S, Costa C, murine hematopoietic stem cells drive aging-
8. Florian MC, Klose M, Sacma M, et al. Aging Blasig R, Rodewald HR. Hematopoietic stem associated immune remodeling. Blood.
alters the epigenetic asymmetry of HSC divi- cell transplantation without irradiation. Nat 2018;132(6):565-576.
sion. PLoS Biol. 2018;16(9):1-35. Methods. 2009;6(4):267-269. 33. McIntosh BE, Brown ME, Duffin BM, et al.
9. Florian MC, Dörr K, Niebel A, et al. Cdc42 21. van Galen P, Kreso A, Wienholds E, et al. Nonirradiated NOD,B6.SCID Il2rγ-/-
activity regulates hematopoietic stem cell Reduced lymphoid lineage priming pro- kitW41/W41 (NBSGW) mice support multi-
aging and rejuvenation. Cell Stem Cell. motes human hematopoietic stem cell lineage engraftment of human hematopoiet-
2012;10(5):520-530. expansion. Cell Stem Cell. 2014;14(1):94- ic cells. Stem Cell Rep. 2015;4(2):171-180.
10. Abkowitz JL, Golinelli D, Harrison DE, 106. 34. Martin MG, Welch JS, Uy GL, et al. Limited
Guttorp P. In vivo kinetics of murine hemo- 22. Althoff MJ, Nayak RC, Hegde S, et al. Yap1- engraftment of low-risk myelodysplastic
poietic stem cells. Blood. 2000;96(10):3399- Scribble polarization is required for syndrome cells in NOD/SCID gamma-C
3405. hematopoietic stem cell division and fate. chain knockout mice. Leukemia. 2010;24(9):
11. Catlin SN, Busque L, Gale RE, Guttorp P, Blood. 2020;136(16):1824-1836. 1662-1664.
Abkowitz JL. The replication rate of human 23. Beerman I, Maloney WJ, Weissmann IL, 35. Sutherland DR, Anderson L, Keeney M,
hematopoietic stem cells in vivo. Blood. Rossi DJ. Stem cells and the aging Nayar R, Chin-Yee I. The ISHAGE guide-
2011;117(17):4460-4466. hematopoietic system. Curr Opin Immunol. lines for CD34+ cell determination by flow
12. Lee CCI, Fletcher MD, Tarantal AF. Effect of 2010;22(4):500-506. cytometry. J Hematother. 1996;5(3):213-226.
age on the frequency, cell cycle, and lineage 24. Etienne-Manneville S, Hall A. Rho GTPases 36. Flach J, Bakker ST, Mohrin M, et al.
maturation of rhesus monkey (Macaca in cell biology. Nature. 2002;420(6916):629- Replication stress is a potent driver of func-
mulatta) CD34+and hematopoietic progeni- 635. tional decline in ageing haematopoietic stem
tor cells. Pediatr Res. 2005;58(2):315-322. 25. Yang L, Zheng Y. Cdc42: A signal coordina- cells. Nature. 2014;512(7513):198-202.

A. Amoah et al.
37. Witte K, Strickland D, Glotzer M. Cell cycle The cell polarity determinant CDC42 con- Donor-to-donor heterogeneity in the clonal
entry triggers a switch between two modes trols division symmetry to block leukemia dynamics of transplanted human cord blood
of Cdc42 activation during yeast polariza- cell differentiation. Blood. 2017;130(11): stem cells in murine xenografts. Biol Blood
tion. Elife. 2017;6:e26722. 1336-1346. Marrow Transplant. 2020;26(1):16-25.
38. Kerber RA, O’Brien E, Cawthon RM. Gene 41. Costa S, Reagan MR. Therapeutic irradia- 44. Jaiswal S, Natarajan P, Silver AJ, et al. Clonal
expression profiles associated with aging and tion: consequences for bone and bone mar- hematopoiesis and risk of atherosclerotic car-
mortality in humans. Aging Cell. row adipose tissue. Front Endocrinol diovascular disease. N Engl J Med. 2017;377
2009;8(3):239-250. (Lausanne). 2019;10:587. (2):111-121.
39. Florian MC, Klenk J, Marka G, et al. 42. Preciado S, Muntión S, Rico A, et al. 45. Jaiswal S, Fontanillas P, Flannick J, et al. Age-
Expression and activity of the small Mesenchymal stromal cell irradiation inter- related clonal hematopoiesis associated with
RhoGTPase Cdc42 in blood cells of older feres with the adipogenic/osteogenic differ- adverse outcomes. N Engl J Med. 2014;371
adults are associated with age and cardiovas- entiation balance and improves their (26):2488-2498.
cular disease. J Gerontol Ser. A hematopoietic-supporting ability. Biol Blood 46. Sano S, Wang Y, Walsh K. Clonal
2017;72(9):1196-1200. Marrow Transplant. 2018;24(3):443-451. hematopoiesis and its impact on cardiovas-
40. Mizukawa B, O’Brien E, Moreira DC, et al. 43. Belderbos ME, Jacobs S, Koster TK, et al. cular disease. Circ J. 2019;83(1):2-11.

Acute Myeloid Leukemia ARTICLE
APR-246 induces early cell death by Ferrata Storti Foundation

ferroptosis in acute myeloid leukemia
Rudy Birsen,1,2 Clement Larrue,3 Justine Decroocq,1,2 Natacha Johnson,1
Nathan Guiraud,4,5 Mathilde Gotanegre,4,5 Lilia Cantero-Aguilar,1
Eric Grignano,1 Tony Huynh,1 Michaela Fontenay,1,6 Olivier Kosmider,1,6
Patrick Mayeux,1 Nicolas Chapuis,1,6 Jean Emmanuel Sarry,4
Jerome Tamburini,1,2,3 and Didier Bouscary1,2
1
Université de Paris, Institut Cochin, CNRS UMR8104, INSERM U1016, Paris, France;
2
Assistance Publique-Hôpitaux de Paris, Centre-Université de Paris, Service Haematologica 2022
d’Hématologie Clinique, Hôpital Cochin, Paris, France; 3Translational Research Center
in Onco-Hematology, Faculty of Medicine, University of Geneva, Geneva, Switzerland; Volume 107(2):403-416
4
Centre de Recherches en Cancérologie de Toulouse, UMR1037, INSERM, Equipe
Labellisée LIGUE 2018, Toulouse, France; 5Université de Toulouse, Institut National des
Sciences Appliquées de Toulouse, INSERM, Toulouse, France and 6Assistance Publique-
Hôpitaux de Paris, Centre-Université de Paris, Service d’Hématologie Biologique,
Hôpital Cochin, Paris, France
ABSTRACT
A
PR-246 is a promising new therapeutic agent that targets p53
mutated proteins in myelodysplastic syndromes and in acute
myeloid leukemia (AML). APR-246 reactivates the transcriptional
activity of p53 mutants by facilitating their binding to DNA target sites.
Recent studies in solid cancers have found that APR-246 can also induce
p53-independent cell death. In this study, we demonstrate that AML cell
death occurring early after APR-246 exposure is suppressed by iron
chelators, lipophilic antioxidants and inhibitors of lipid peroxidation,
and correlates with the accumulation of markers of lipid peroxidation,
thus fulfilling the definition of ferroptosis, a recently described cell death
process. The capacity of AML cells to detoxify lipid peroxides by
increasing their cystine uptake to maintain major antioxidant molecule
glutathione biosynthesis after exposure to APR-246 may be a key deter-
minant of sensitivity to this compound. The association of APR-246 Correspondence:
with induction of ferroptosis (either by pharmacological compounds, or RUDY BIRSEN
genetic inactivation of SLC7A11 or GPX4) had a synergistic effect on the rudy.birsen@inserm.fr
promotion of cell death, both in vivo and ex vivo.
DIDIER BOUSCARY
didier.bouscary@aphp.fr
Received: May 18, 2020.
Introduction
Acute myeloid leukemias (AML) are highly heterogeneous diseases with a con- Pre-published: January 7, 2021.
stant activation of oncogenic signaling.1 Recent years have witnessed major break-
throughs in their treatment with the approval of midostaurin, venetoclax and IDH
mutant inhibitors.2-5 However, AML has a poor prognosis and there is still an urgent https://doi.org/10.3324/haematol.2020.259531
need for new treatments. APR-246, also known as PRIMA-1MET, is a promising
new therapeutic agent that targets TP53 mutated cancers.6-8 This compound is
being evaluated in AML and myelodysplastic syndromes (MDS) with TP53 muta-
tions and appears to be highly effective against this poor prognosis disease.8-11 Material published in Haematologica is covered by copyright.
Mechanistically, APR-246 is converted to a reactive product (methylene quinuclidi- All rights are reserved to the Ferrata Storti Foundation. Use of
none, MQ) that reacts with nucleophiles and thus alkylates thiol groups in conditions:
proteins.12 APR-246 reactivates the transcriptional activity of p53 mutants by facil- https://creativecommons.org/licenses/by-nc/4.0/legalcode.
itating their binding to DNA target sites. Specific cysteines in the core domain of Copies of published material are allowed for personal or inter-
mutant p53 proteins are critical targets for their reactivation by APR-246/MQ.13 nal use. Sharing published material for non-commercial pur-
APR-246 also triggers p53-independent cell death mechanisms.14,15 Accordingly, https://creativecommons.org/licenses/by-nc/4.0/legalcode,
using esophageal cancer as a model, it has been shown that APR-246 causes a sect. 3. Reproducing and sharing published material for com-
decrease in glutathione (GSH) content resulting in an increased amount of reactive mercial purposes is not allowed without permission in writing
oxygen species (ROS) and of lipid peroxides in particular.16 In this report, we inves- from the publisher.
tigated the mechanisms of cell death induced by APR-246 in AML and we demon-
strated that early cell death in AML is due to ferroptosis induction.

R. Birsen et al.
Methods Biotracker cystine in 1 mL of warm complete medium for 120

minutes in a tissue culture incubator (37°C, 5% CO2) in the dark.
Cell lines and reagents FCM data were collected using a C6 Accuri flow cytometer
HL60, MOLM14, SET2, MV4-11, OCI-AML2, OCI-AML3, (Becton Dickinson, Le Pont de Claix, France) with CFlow Plus
K562, THP1, UT7-EPO, SKM1, NB4 and KASUMI AML cell software. 10,000 events were captured for subsequent analysis
lines were used. Patients provided written informed consent in with CFlow Plus software (Becton Dickinson, Le Pont de Claix,
accordance with the Declaration of Helsinki. Bone marrow (BM) France).
samples were obtained from five patients with newly diagnosed
AML (characteristics provided in the Online Supplementary Table Western blotting
S1). Cells were cultured in RPMI with glutamine (Gibco61870, Whole-cell extraction and western blotting were performed as
Life Technologies® Saint Aubin, France) supplemented with previously described.17 Anti-GPX4 antibody was purchased
10% fetal bovine serum (FBS) and 4 mM glutamine. All AML cell from Proteintech (Manchester, UK). Anti-PARP, caspase 8, cas-
lines were certified using their microsatellite identity (character- pase 3, cleaved caspase 3, MLKL, pMLKL, p53 and SLC7A11
istics provided in Online Supplementary Table S2). Ferrostatin-1, antibodies were sourced from Cell Signaling Technology
necrostatin-1, necrostatin-1S, necrosulfonamide, QVD-OPH, (Danvers, MA, USA).
APR-246 for the in vitro study, erastin and RSL3 were sourced
from Selleckchem (Houston, TX). Chloroquine and doxycycline Viability assay
were obtained from Sigma–Aldrich (Saint-Louis, MO). FINO2 AML cells were plated at 20x104/mL in 100 µl of 10% FBS-
was purchased from Cayman Chemicals (Ann Arbor, MI). The supplemented RPMI prior to the addition of compounds. Cells
APR-246 reagent used in the in vivo study was provided by were cultured in the presence of the test compounds for 24 to 48
APREA therapeutics (Solna, Sweden). h at 37°C. Viability was quantified using the fluorescence based
Uptiblue assay (Interchim, Montluçon, France). Uptiblue was
Constructs added to each well in 10 mL aliquots. Fluorescence was then
Inducible short hairpin RNA (shRNA) targeting SLC7A11 or measured with a Typhoon FLA9500 scanner (GE Healthcare; IL).
GPX4 were constructed as previously described17 using the fol- Fluorescence values were normalized to dimethyl sulfoxide
lowing sequences: SLC7A11#2, GCTGAATTGGGAACAAC- (DMSO)-treated controls for each AML cell line. Half maximal
TATA; SLC7A11#3, GCAGTTGCTGGGCTGATTTAT; inhibitory concentration (IC50) values were calculated using a
GPX4#1, GTGAGGCAAGACCGAAGTAAA; GPX4#2, CTA- four parameter non-linear regression curve with Graph Pad
CAACGTCAAATTCGATAT. Prism v8 (GraphPad, La Jolla, CA, USA). For primary AML cells,
viability was assayed by FCM analysis using forward scatter
Lentivirus production and acute myeloid leukemia cell (FSC) versus side scatter (SSC).
line infection
293-T packaging cells were used to produce lentiviral constructs Measurement of synergistic effects
through co-transfection with plasmids encoding lentiviral pro- Cell viability was calculated for every dose combination of
teins. Supernatants were collected and ultracentrifuged for 48 APR-246 and ferroptosis inducer using the Synergy Finder
hours (h) after transfection over two consecutive days and subse- webtool (https://synergyfinder.fimm.fi/) and compared to each
quently stored at -80°C. AML cell lines were plated at 2x106/mL agent alone. Calculations were based on the ZIP model.18
and 2-10 mL aliquots of lentiviral supernatants were added for 3 h.
Cells were then grown in 10% fetal calf serum medium and fur- Tumor xenografts in NOD/SCID IL-2 receptor
ther selected with puromycin. For shRNA induction, 200 mg/mL g-chain-null mice
doxycycline was added to the culture medium. Xenograft tumors were generated by randomly injecting
1×106 MOLM14 shCTRL or shSLC7A11 cells into the tail veins
Flow cytometry-based assay of NOD/SCID IL-2 receptor g-chain-null mice (NSG) aged 6–9
Data acquisition and data analysis were conducted at the weeks. Fourteen days after injection, doxycycline (200 mg/mL)
Cochin Cytometry and Immunobiology Facility. For glutathione and sucrose (1% weight:volume) were added to the drinking
measurements using monochlorobimane (MCB; Thermofischer; water of these animals. After 3 days, the mice were randomly
Waltham, MA), 2x105 cells were labeled with 40 mM MCB in 1 treated with a daily intraperitoneal injection of APR-246 (100
mL of warm complete medium for 20 minutes (min) in a tissue mg/kg) or vehicle (phosphate-buffered saline [PBS]) for 4 days.
culture incubator (37°C, 5% CO2) in the dark. The reaction was All experiments were conducted in accordance with the guide-
terminated using 1 mL of cold complete medium, followed by lines of the Association for Assessment and Accreditation of
centrifugation (200xg, 1 min). The pelleted cells were then re- Laboratory Animal Care International. Animals were used in
suspended in 0.5 mL of cold complete medium and placed on ice accordance with a protocol reviewed and approved by the
in the dark until analysis by flow cytometry (FCM). The MCB- Institutional Animal Care and Use Committee of Région Midi-
GSH signal was detected using a 355 nm laser through a 450/50 Pyrénées (France). BM (mixed from tibias and femurs) were dis-
nm filter. FCM data were collected using a BD Fortessa flow sected and flushed in Hanks balanced salt solution with 1% FBS.
cytometer with DIVA software. 10,000 events were recorded for MNC from BM were labeled with PE-conjugated anti-hCD33,
analysis. Data analysis was then carried out with KALUZA soft- PerCP-Cy5.5-conjugated anti-mCD45.1 and APC-conjugated
ware. For lipid peroxide production measurements using C11- anti-hCD45 (all antibodies from Becton Dickinson, BD) to deter-
BODIPY (581/591) (2 mM) (Thermofischer, Waltham, MA), mine the fraction of human blasts (hCD45+mCD45.1−hCD33+
2x105 cells were labeled with C11-BODIPY in 1 mL of warm cells) using FCM. Acquisition of data was performed on a
complete medium for 10 min in a tissue culture incubator (37°C, CytoFLEX (Beckman Coulter) flow cytometer with CytExpert
5% CO2) in the dark. Cells were then washed twice and resus- software. The number of AML cells in the BM was determined
pended in 200 mL of fresh PBS. For cystine uptake measurements using CountBright beads (Invitrogen, CA, USA) in accordance
using BioTracker Cystine-FITC Live Cell Dye (5 mM) with the manufacturer’s protocol. Data analysis was performed
(Thermofischer, Waltham, MA), 2x105 cells were labeled with with flowJo software.

APR-246 induced ferroptosis in AM
Figure 1. APR-246 induces cell death in acute myeloid leukemia cells irrespective of their TP53 mutational status. (A) Viability curves for the indicated cells at 24
hours (h) post APR-246 treatment. Error bars, ± standard deviation. (B) Viability curves for the indicated primary acute myeloid leukemia (AML) cells at 24 h post
APR-246 treatment. Patient characteristics are provided in the Online Supplementary Table S1. (C) Half maximal inhibitory concentration percentage (IC50%) of APR-
246 treatment for 24 h across a panel of AML cell lines based on cell viability (n=3). In our subsequent experiments, we selected five AML cell lines sensitive to APR-
246 in these concentration ranges, with or without TP53 mutations (highlighted in bold font). wt: wild-type; fs: frameshift; del: deletion.

R. Birsen et al.
B C
Figure 2. Legend on following page.

Figure 2. APR-246 induces ferroptosis in acute myeloid leukemia cells. (A) Cell viability (%) for the indicated cells at 16 hours (h) post-APR-246 treatment (60 mM)
with or without ferrostatin-1 (10 mM), deferoxamine (DFO) (100 mM), necrostatin-1 (20 mM), chloroquine (20 mM) or QVD-OPH (25 mM) (n=3). Error bars ± standard
error of the mean [SEM]. All compounds were added 2 h (h) prior to APR-246 in the medium. Statistics, 2-way ANOVA; *P<0.05, **P<0.01, ***P<0.0001. (B)
Immunoblotting analysis of PARP, caspase 8 and caspase 3 in MOLM-14 cells treated for 16 h with dimethyl sulfoxide (DMSO), APR-246 (60 µM) or puromycin (1
mg/mL). b-actin was used as a loading control (n=2). (C) Viability curves for the indicated cells at 16 h post APR-246 treatment with or without ferrostatin-1 (10 mM),
DFO (100 mM), necrostatin-1 (20 mM), chloroquine (20 mM) or QVD-OPH (25 mM) (n=3). Error bars ± SEM. (d) Cell death (%) of the indicated cells at 16 h and 24 h
post-APR-246 treatment (50 mM) with or without ferrostatin-1 (10 mM) (n=3). Error bars ± standard deviation.
Figure 3. APR-246 induces ferroptosis in acute myeloid leukemia cells. (A) Electron
microscopy analysis of MOLM-14 cells treated with or without APR-246 (60 mM, H16). The
white arrowhead indicates a mitochondrion showing membrane rupture and reduced cristae.
(B and C) Detection of lipid peroxidation using C11-BODIPY and flow cytometry (FCM) at 14
hours post APR-246 treatment in acute myeloid leukemia (AML) cell lines (B) and in primary
AML cells (C). APR-246 was used at a 100 mM concentration for MOLM-14 and 50 µM for other
AML cell lines. Left panels show representative FCM quantification (n=3). Error bars ± stan-
dard deviation.

R. Birsen et al.
C D
Figure 4. APR-246 induces glutathione depletion in acute myeloid leukemia cells. (A) Summary of the cellular pathways involved in ferroptosis. Ferroptosis execution
is triggered by an iron-catalyzed excessive peroxidation of polyunsaturated fatty acids (PUFA)-containing phospholipids (PL-PUFA). Glutathione (GSH) and glutathione
peroxidase 4 (GPX4) are the two key elements controlling the elimination of lipid peroxides. Solute carrier family 7 member 11 (SLC7A11) encodes the transporter
subunit of the heterodimeric cystine-glutamate antiporter named system xc-. System xc- mediates cystine entry into the cell in exchange for glutamate.26 Once inside
the cell, cystine is rapidly reduced to cysteine which is the limiting amino acid for GSH synthesis. SLC7A11 inhibition results in cellular cysteine depletion, which leads
to the exhaustion of intracellular pool of GSH. GPX4 is a pleiotropic selenoprotein that uses GSH to selectively reduce lipid hydroperoxides to lipid alcohols, in order
to protect the cells against membrane lipid peroxidation.29 GPX4 inhibition is either due to its direct inhibition or downregulation, or to GSH depletion via direct or
indirect processes. The inhibition of GPX4 results in uncontrolled polyunsaturated fatty acid phospholipid (PL-PUFA) oxidation and fatty acid radical generation, lead-
ing to ferroptotic cell death. ACSL4: acyl-CoA synthetase long chain family member 4; LPCAT 3: lysophosphatidylcholine acyltransferase 3; ALOX: arachidonate lipoxy-
genase; PUFA: polyunsaturated fatty acid; PL: phospholipids; PE: phosphatidylethanolamine; GPX4: glutathione peroxidase 4. (B) GSH (mBCI) measurement in acute
myeloid leukemia (AML) cell lines by flow cytometry (FCM) at 14 hours (h) post APR-246 and fer1 (10 mM) treatment. APR-246 was used at 100 mM for MOLM-14
and 50 mM for other AML cell lines. Fer1 was associated to prevent cell death and allowed analysis of GSH depletion. Left panels show representative FCM quantifi-
cation. n=3. Error bars + standard deviation. Statistics by t-test. *P<0.05, **P<0.01, ***P<0.0001. (C and D) Cell death (%) (C) and GSH (mBCI) measurement (D)
for MOLM-14 at 24 h post-APR-246 treatment (60 mM) with or without B-ME (50 mM), cysteine (50 mM) or Fer1 (10 mM). Error bars + standard deviation. Statistics
by t-test; *P<0.05, **P<0.01, ***P<0.0001.

Results vivo. This increase in lipid ROS was fully reversed by Fer1,
indicating that lipid peroxides had been newly generated
APR-246 induces ferroptosis in acute myeloid (Figure 3B and C). All these results allowed to conclude
leukemia cells that APR-246 induces early cell death by ferroptosis, a
In order to determine the activity of APR-246 in AML, recently described non-apoptotic form of regulated cell
we assayed a set of 12 AML cell lines carrying diverse and death that links together membrane lipid peroxidation,
representative molecular abnormalities, and five primary cysteine and iron metabolism, glutathione peroxidase
AML samples (Online Supplementary Table S1). Most of the activity and oxidative stress (as summarized in Figure
AML cell lines and primary AML cells were sensitive to 4A23–25). As previously reported,14,16 we observed that
cell death induction by APR-246 (Figure 1A and B). The APR-246 treatment induced a dramatic decrease in GSH
IC50 of APR-246 at 24 h for each cell line ranged from 11 levels in AML cell lines (Figure 4B). Cysteine is the main
mM to more than 200 mM, independently of their TP53 biosynthetic precursor of GSH. Cysteine can be trans-
mutational status (Figure 1C). Previous in vivo human data ported into cells via membrane transporters for neutral
have shown that the plasma concentrations of APR-246 in amino acids. However, in the extracellular space, cysteine
the 12 h following its intravenous administration range is rapidly reduced to cystine. Thus the main source of
from 50 to 500 mM, suggesting that concentrations above intracellular cysteine comes from the entry of cystine into
50 mM are suitable for in vitro studies of the early effects of the cell via system xc-.26 b-mercaptoethanol (b-ME) can
APR-246.19 In our subsequent experiments, we selected promote cystine uptake through an alternative pathway.27
five AML cell lines sensitive to APR-246 in these concen- b-ME was able to completely rescue the cell death and
tration ranges, with or without TP53 mutations. GSH depletion induced by APR-246 (Figure 4C and D).
In order to investigate the mechanisms underlying Cysteine treatment showed similar results as Fer1 treat-
APR-246 activity against AML cells, we exposed the cells ment (Figure 4C and D). Altogether, these results demon-
treated with APR-246 to inhibitors of various cell death strate that APR-246 induces GSH depletion which
pathways (Figure 2A). The decrease in cell viability was induces ferroptosis in AML cells irrespective of their TP53
almost completely rescued by either iron chelation via mutational status, and that ferroptosis induction is the
deferoxamine (DFO) or by the lipophilic antioxidant fer- main mechanism of cell death after early exposure to
rostatin-1 (Fer1), demonstrating that cell death is both APR-246.
iron and reactive- oxygen species (ROS) dependent.
Necrostatin-1 (Nec1) also consistently prevented cell Cystine uptake determined the sensitivity of acute
death induced by APR-246 after short-term incubation in myeloid leukemia cells to APR-246
all AML cells lines. Nec1 has been used to define necrop- Since cysteine is a biosynthetic precursor of GSH, we
totic cell death, but can also protect against ferroptosis asked whether the ability of cells to provide cysteine for
through a target which is as yet unknown.20 Induction of GSH synthesis underlies the sensitivity to APR-246.
necroptosis after treatment with APR-246 in our cells was Using FITC-labeled cystine, we showed that after expo-
excluded due to the absence of protection by more specif- sure to APR-246, AML cells increased their uptake of cys-
ic necroptosis inhibitors (Necrostatin 1s and tine from the extracellular space (Figure 5A). Western blot
Necrosulfonamide),21,22 and due to the absence of MLKL analysis of the protein levels of SLC7A11 showed an
phosphorylation, a key marker in necroptosis21 (Online increased amount of SLC7A11 (Figure 5B). This suggests
Supplementary Figure S1A and B). Inhibitors of autophagy that the cells enhanced cystine uptake by increasing
(chloroquine) or apoptosis (QVD-OPH) failed to block SLC7A11 protein levels to maintain intracellular GSH lev-
APR-246-induced cell death. We confirmed that the els after exposure to APR-246. We modulated the cystine
mechanism of APR-246 induced cell death is distinct from uptake through SLC7A11 overexpression or inhibition.
apoptosis, as evidenced by the absence of caspases 3/8 or SLC7A11 overexpression in MOLM-14 and OCI-AML2
PARP1 cleavage, including in TP53 mutated AML cell (Online Supplementary Figure S4) decreased the cell death
lines (Figure 2B; Online Supplementary Figure S2A to C). and prevented the depletion of GSH induced by APR-246
Notably, the protection against cell death observed with (Figure 5C and D). Then we showed that targeting the
Fer1, DFO or Nec1 was partially lost at higher doses of SLC7A11 cystine transporter by RNA interference
APR-246 (Figure 2C) and at later time points (Figure 2D). reduced the basal uptake of cystine (Online Supplementary
However, apoptosis did not appear to be the mechanism Figure S4B and C) and had a very little effect on cell death
of this late death, since caspases 3/8 and PARP1 were not (Online Supplementary Figure S4D), but strongly reduced
cleaved and QVD-OPH was still unable to prevent the cell proliferation in AML cells in vitro (Online
cell death (Online Supplementary Figure S2A to E). Supplementary Figure S4E). Inhibition of SLC7A11 with
Examination of the ultrastructural changes induced by RNA interference increased cell death and viability
APR-246 treatment did not reveal any characteristic fea- impairment, GSH depletion, and the accumulation of
tures of apoptosis (i.e., no plasma membrane blebbing, lipid peroxides induced by APR-246 (Figure 6A to C;
chromatin condensation or nuclear fragmentation) or Online Supplementary Figure S4F). Interestingly, basal GSH
autophagy (absence of autophagolysosomes). Necrotic levels were not consistently affected by b−ME addition,
cells were rare, and some mitochondria showed mem- cysteine addition, or the overexpression or inhibition of
brane rupture and reduced cristae (Figure 3A; Online SLC7A11, which suggests that the amount of GSH in
Supplementary Figure S3). We analyzed the levels of basal conditions is not a reliable marker of cell cystine
endogenous lipid peroxidation – a hallmark of ferroptosis uptake ability (Figures 4D, 5D and 6C). Finally, targeting
induction – following APR-246 treatment by flow cytom- SLC7A11 with erastin, a potent inhibitor of system xc-28
etry analysis with C11-BODIPY staining. We observed a that showed variable sensitivity in our AML cell lines
high induction of lipid ROS in AML cell lines and primary (Online Supplementary Figure S4G), had synergistic activity
AML cells from two patients treated with APR-246 ex with APR-246 both on cell death and on cell viability

R. Birsen et al.
impairment (Figure 6D and E; Online Supplementary Figure exposure to APR-246 is a key determinant of the sensitiv-
S5). The association of erastin and APR-246 also had a ity to this compound.
synergistic effect on cell viability in five primary AML
samples (Figure 6F). There was no correlation between The association of APR-246 and ferroptosis inducers
the basal levels of GPX4 and SLC7A11 proteins and the has a synergistic anti-leukemic activity in vitro
sensitivity to APR-246 (Online Supplementary Figure S6A We next determined whether targeting ferroptosis path-
and B). Altogether, these data suggest that the ability of ways in combination with APR-246 can increase the anti-
AML cells to prevent lipid peroxides accumulation by leukemic activity of this compound, mimicking SLC7A11
increasing their cystine uptake to support GSH after inhibition. Downregulation of GPX4 by RNA interference
Figure 5. SLC7A11 overexpression prevents glutathione depletion and cell death following APR-246 exposure. (A) Cystine uptake in MOLM-14 and OCI-AML2 cells
lines at 16 hours (h) post-APR-246 (100 mM) and Fer1 (10 mM) treatment. Fer1 was associated to prevent cell death and allowed analysis of cystine uptake. (B)
Immunoblotting analysis of SLC7A11 in MOLM-14 cells treated for 16 h with dimethyl sulfoxide (DMSO) or APR-246 (n=2). b-actin was used as a loading control. (C
and D) Cell death (%) (C) and glutathione (GSH) (mBCI) measurement (D) of the indicated cells at 20 h post-APR-246 treatment (n=3). For GSH measurement, Fer1
was associated to prevent cell death and allowed analysis of GSH depletion. Error bars ± standard deviation. Statistics by t-test. *P<0.05, **P<0.01, ***P<0.0001.

E F
Figure 6. Legend on following page.

R. Birsen et al.
Figure 6. SLC7A11 inhibition sensitizes cells to APR-246. (A) Viability curves for MOLM-14 and OCI-AML2 with small hairpin RNA control (shSCR) or doxycycline-
inducible shRNA (shSLC7A11) cells at 16 hours (h) post APR-246 treatment. Prior to adding APR-246, the cells were treated for 3 days with doxycycline (n=3). Error
bars ± standard deviation (SD). (B and C) Cell death (%) (B) and glutathione (GSH) (mBCI) measurement (C) of the indicated cells at 20 h post-APR-246 (MOLM-14
30 mM, OCI-AML2 10 mM) treatment (n=3). For GSH measurement, Fer1 was associated to prevent cell death and allowed analysis of GSH depletion. Prior to adding
APR-246, the cells were treated for 3 days with doxycycline (n=3). Error bars ± SD. Statistics by t-test. *P<0.05, **P<0.01, ***P<0.0001. (D) Cell death (%) for the
indicated cell types at 24 h post APR-246 (MOLM-14 30 mM, OCI-AML2 10 µM) and or erastin (MOLM-14 100 nM, OCI-AML2 1 mM). Error bars ± SD. Statistics by
t-test; *P<0.05, **P<0.01. (E) Illustrative synergy map (left panel) of 24 h co-treatment of MOLM-14 cells with APR-246 and erastin. The mean cell viability of three
independent experiments was used. Mean synergy scores of the most synergistic area of 24 h co-treatment of AML cell lines with APR-246 and erastin (n=3). (F)
Mean synergy score of the 48 h co-treatment of primary AML cells with APR-246 and erastin (n=1).
resulted in cell death (Online Supplementary Figure S7A and cancer, APR-246 induces cell death by impairing
B). Our AML cell lines panel showed variable sensitivity to GSH/ROS balance and acts synergistically with L-buthio-
two ferroptosis-inducing drugs: RSL3, a direct GPX4 nine sulphoximine to inhibit myeloma growth in vivo.14 In
inhibitor;29 and FINO2, an indirect GPX4 inhibitor and ARID1A-deficient cancers, GSH was the major target of
direct iron oxidant30 (Online Supplementary Figure S8). We APR-246 and was the basis of the high sensitivity of these
observed that knockdown of GPX4 increased the impair- cancer cells to this coumpound.38 In esophageal cancer, Liu
ment of cell viability induced by APR-246 (Figure 7A). et al. showed that mutants p53 bind to the antioxidant
RSL3 and FINO2 in association with APR-246 synergisti- transcription factor NRF2, leading to a decreased expres-
cally decreased cell viability in AML cell lines (Figure 7B sion of SLC7A11 which sensitizes cells to GSH depletion
and C; Online Supplementary Figures S9 and 10). by APR-246.16
Collectively, these results show that pharmacological or Paradoxically, while APR-246 clinical development is
genetic activation of ferroptosis enhances the anti- the most advanced in AML with TP53 mutation, the
leukemic activity of APR-246 in AML. effects of APR-246 have been little studied in this disease.
Two studies showed that APR-246 induced in vitro cell
Genetic invalidation of SLC7A11 has synergistic death in a large number of leukemic cells from patients,
anti-leukemic activity with APR-246 in vivo alone or in association with chemotherapies.39,40 In both
We then examined whether inhibition of GSH synthesis studies, the cytotoxicity of APR-246 was independent of
through SLC7A11 inhibition could interfere with AML the TP53 mutational status. The mechanisms of action of
persistence and could enhance APR-246 activity in vivo. APR-246 was investigated in AML cell lines with TP53
We engrafted MOLM14 cells transduced with either con- mutations, and more specifically studied its association
trol (shSCR) or anti-SLC7A11 (shSLC7A11) doxycycline- with 5-azacytidine which is currently used in clinical tri-
inducible shRNA (Figure 8A). After induction of shRNA als.8 It was shown that in TP53-mutated myelodysplastic
expression in vivo, we treated the mice with a 4-day APR- syndromes (MDS) and AML, APR-246 can reactivate the
246 regimen in order to mimic the therapeutic schedule p53 pathway and induce an apoptotic transcriptional pro-
used in clinical trials of APR-246.16 This treatment scheme gram, with synergistic effects of APR-246 and azacytidine.
varied from those previously published in mice in terms of In this context, our study strongly showed that APR-246
treatment duration. Indeed, in these studies, APR-246 was induced cell death in AML cells irrespective of their TP53
administrated for 7 to 28 days, and reduction of tumor mutational status. APR-246 depleted intracellular GSH
volume after 4 days of treatment was minimal at and induced lipid peroxide production, which led to fer-
best.7,8,14,16 However, our aim was not to assess the efficacy roptosis induction. The ability of AML cells to detoxify
of APR-246 alone but to demonstrate that its association lipid peroxides primed their sensitivity to APR-246 treat-
with SLC7A11 inhibition enhanced its anti-leukemic ment. Additionally, we uncovered that inhibition of anti-
activity. As expected, APR-246 alone did not reduce BM ferroptosis mechanisms enhanced the anti-leukemic activ-
tumor cell burden. SLC7A11 knockdown significantly ity of APR-246 both in vitro and in vivo. We confirm the
reduced tumor cell burden in the BM (Figure 8B and C). TP53 independence and GSH depletion and we demon-
Moreover, the decrease in BM tumor cell burden was strated that APR-246 induces ferroptosis.14,16 Ferroptosis is
enhanced when APR-246 treatment was combined to rapidly induced after GPX4 inactivation and cell death
SLC7A11 knockdown (Figure 8B and C). Overall, these occurs in the first 24 hours post-treatment or administra-
results showed that inhibition of anti-ferroptosis mecha- tion.23,29 The observation that protection against cell death
nisms enhanced the anti-leukemic activity of APR-246 in by ferroptosis inhibitors decreases after 24 hours of expo-
vivo. sure suggests that other cell death mechanisms might be
involved after this early phase and that they may have
masked the earlier induction of ferroptosis. The effect of
Discussion APR-246 might be also different in AML cells in compari-
son to solid cancers.
APR-246 can restore the wild-type conformation of Our study might have several important implications
mutant p53 protein, therefore inducing apoptosis and for the management of MDS and AML patients. First,
inhibition of tumor growth in mice.6 Thus, APR-246 is one since APR-246 acts independently of TP53 mutational sta-
of the most promising compounds in clinical development tus, this treatment could be used in a broader panel of
for TP53 mutated cancers. Controversies exist over the AML patients. Future study will need to identify predic-
TP53 mutation status dependencies of APR-246.31 Some tive elements of the sensitivity of AML to APR-246 and
studies reported that APR-246 acts independently of its the induction of ferroptosis. The mechanism of action of
ability to reactivate mutant p53 protein.12,32–37 Tessoulin APR-246, which is based on GSH depletion and induction
et al. demonstrated that myeloma cells are highly sensitive of ferroptosis, makes it the first ferroptosis-inducing agent
to APR-246, independently of their TP53 status.14 In this currently used therapeutically in humans. Using ferropto-

Figure 7. The combination of APR-246 with ferroptosis inducers has synergistic anti-leukemic effects in acute myeloid leukemia in vitro. (A) Viability curves for
MOLM-14 and OCI-AML2 cells with or without GPX4 inducible small hairpin RNA (shRNA) at 16 hours (h) post APR-246 treatment. Prior to adding APR-246, the cells
were treated for 2 days with doxycycline (n=3). Error bars ± standard deviation. (B) Illustrative synergy map (left panel) of 24 h co-treatment of MOLM-14 cells with
APR-246 and RSL3. The mean cell viability of three independent experiments was used. Mean synergy scores of the most synergistic area of 24 h co-treatment of
acute myeloid leukemia (AML) cell lines with APR-246 and RSL3 (n=3).(C) Illustrative synergy map (left panel) of 24 h co-treatment of MOLM-14 cells with APR-246
and FINO2. The mean cell viability of three independent experiments was used. Mean synergy scores of the most synergistic area of 24 h co-treatment of AML cell
lines with APR-246 and FINO2 (n=3).

R. Birsen et al.
Figure 8. The combination of APR-246 with SLC7A11 inhibition has synergistic anti-leukemic effects in acute myeloid leukemia in vivo. (A) Design of the in vivo
experiment. (B) Representative flow cytometry analysis of bone marrow cells marked with mCD45.1 and hCD45 from each treatment subgroup. (C) Viable mCD45.1-
hCD45+ hCD33+ cell counts in the bone marrow of the different treatment subgroups. Bars represent mean of all experiments and errors denote ± standard devia-
tions. Statistics by t-test; *P<0.05, **P<0.01, ***P<0.0001.

sis induction to treat cancer is an emerging field in oncol- induction may represent a new target in AML, opening
ogy research. Renal cancer cells have been reported as new therapeutic strategies based on disease-specific vul-
highly dependent on the GSH pathway for ROS detoxifi- nerabilities. The effect of ferroptosis induction-based
cation, including lipid peroxides, and targeting compo- treatments on normal hematopoietic cells and their value
nents of this pathway is an effective strategy for the treat- compared to standard-of-care AML therapies will be
ment of this disease.41 Other studies have elegantly high- important to evaluate in the future.
lighted the higher sensitivity to ferroptosis of cancer cells
that are resistant to conventional therapy.42,43 In AML, data Disclosures
about ferroptosis are scarce. An in vitro study showed that No conflicts of interest to disclose.
the ferroptosis inducer erastin enhances sensitivity of
AML cells to chemotherapeutic agents.44 Jones et al. Contributions
recently reported that cysteine depletion leads to GSH JD, LCA, EG, TH performed experiments; CL, JES, JT, NG
exhaustion and ROS-low leukemic stem cell eradication in and MG performed in vivo experiments; NC, OK and MF pro-
AML.45 Thus, APR-246 could act on these cell pools that vided AML samples; NC, PM, JES, NJ and JT analyzed the
are poorly sensitive to conventional therapy, and which results and corrected the manuscript; RB performed experiments,
are at the origin of frequent therapeutic failures in AML. analyzed data, and wrote the manuscript; DB designed and
From a clinical perspective, this mechanism of action supervised the research program, analyzed data, and wrote the
might be relevant. Indeed, iron chelators are frequently manuscript. All authors approved the final version of the manu-
used for the treatment of iron overload due to red blood script.
cell transfusions and dyserythropoïesis in MDS/AML
patients. Several studies have reported beneficial effects of Acknowledgments
iron chelation therapy on overall survival in MDS patients We thank Alain Schmitt and the cell imagery department at
with iron overload.46 However, iron chelators are recog- the Cochin Institute for performing the transmission electron
nized as canonical ferroptosis inhibitors. Therefore, cau- microscopy. We thank Tata Jojo for manuscript proofreading. We
tion should be exercised regarding co-administration of also thank the CYBIO cytometry-department at the Cochin
iron chelators which may antagonize the anti-leukemic Institute. We further thank Aprea Therapeutics for providing the
activity of APR-246, as we observed in vitro. Moreover, the APR-246 used in the in vivo study.
phase II studies of APR-246 in MDS/AML reported the
occurrence of neurological adverse events in over a third Funding
of patients treated with APR-246.10,11 Recently, ferroptosis This work was supported by grants from the Association de
has been implicated in the pathogenesis of several neuro- Recherche Contre le Cancer (ARC; aides doctorales RB, grant
logical disorders, especially neurodegenerative disor- n°DOC20170505807 and DOC20190508975; aides jeune
ders.47,48 One hypothesis could be that the neurological chercheur TH, grant n°M2R20180507379), from the Institut
side effects observed after administration of APR-246 are National du Cancer (JD, grant n° ASC16046KSA), from the
linked to its ability to deplete GSH in neuronal cells. Ligue Nationale Contre le Cancer (LNCC; DB, Equipe
Consequently, anti-ferroptosis agents, such as iron chela- Labellisée EL2017. N° Projet: ELFUZ17337; NG, grant n°
tors or vitamin E, could be valuable drugs to treat these IP/SCG/JD-16129) and from association Laurette Fugain (grant
side effects. Finally, our study highlights that ferroptosis n°ALF2018/02).
References H. PRIMA-1Met/APR-246 displays high covalent binding to the core domain.

antitumor activity in multiple myeloma by Cancer Cell. 2009;15(5):376-388.
induction of p73 and Noxa. Mol Cancer 13. Zhang Q, Bykov VJN, Wiman KG,
1. Döhner H, Estey E, Grimwade D, et al. Ther. 2013;12(11):2331-2341. Zawacka-Pankau J. APR-246 reactivates
Diagnosis and management of AML in 8. Maslah N, Salomao N, Drevon L, et al. mutant p53 by targeting cysteines 124 and
adults: 2017 ELN recommendations from Synergistic effects of PRIMA-1Met (APR- 277. Cell Death Dis. 2018;9(5):1-12.
an international expert panel. Blood. 246) and Azacitidine in TP53-mutated 14. Tessoulin B, Descamps G, Moreau P, et al.
2017;129(4):424-447. myelodysplastic syndromes and acute PRIMA-1Met induces myeloma cell death
2. Stone RM, Mandrekar SJ, Sanford BL, et al. myeloid leukemia. Haematologica. 2020; independent of p53 by impairing the
Midostaurin plus chemotherapy for acute 105(6):1539-1551. GSH/ROS balance. Blood. 2014;
myeloid leukemia with a FLT3 Mutation. N 9. Nahi H, Merup M, Lehmann S, et al. 124(10):1626-1636.
Engl J Med. 2017;377(5):454-464. PRIMA-1 induces apoptosis in acute 15. Bykov VJN, Zhang Q, Zhang M, Ceder S,
3. Stein EM, DiNardo CD, Pollyea DA, et al. myeloid leukaemia cells with p53 gene Abrahmsen L, Wiman KG. Targeting of
Enasidenib in mutant-IDH2 relapsed or deletion. Br J Haematol. 2006;132(2):230- mutant p53 and the cellular redox balance
refractory acute myeloid leukemia. Blood. 236. by APR-246 as a strategy for efficient can-
2017;130(6):722-731. 10. Sallman DA, DeZern AE, Garcia-Manero cer therapy. Front Oncol. 2016;6:21.
4. DiNardo CD, Stein EM, de Botton S, et al. G, et al. Phase 2 results of APR-246 and 16. Liu DS, Duong CP, Haupt S, et al. Inhibiting
Durable remissions with Ivosidenib in Azacitidine (AZA) in patients with TP53 the system xC-/glutathione axis selectively
IDH1-mutated relapsed or refractory AML. mutant myelodysplastic syndromes (MDS) targets cancers with mutant-p53 accumula-
N Engl J Med. 2018;378(25):2386-2398. and oligoblastic acute myeloid leukemia tion. Nat Commun. 2017;8:14844.
5. DiNardo CD, Pratz K, Pullarkat V, et al. (AML). Blood. 2019;134(Suppl 1):S676. 17. Jacque N, Ronchetti AM, Larrue C, et al.
Venetoclax combined with decitabine or 11. Cluzeau T, Sebert M, Rahmé R, et al. APR- Targeting glutaminolysis has antileukemic
azacitidine in treatment-naive, elderly 246 combined with Azacitidine (AZA) in activity in acute myeloid leukemia and syn-
patients with acute myeloid leukemia. TP53 mutated myelodysplastic syndrome ergizes with BCL-2 inhibition. Blood.
Blood. 2019;133(1):7-17. (MDS) and acute myeloid leukemia (AML). 2015;126(11):1346-1356.
6. Bykov VJN, Issaeva N, Shilov A, et al. A phase 2 study by the Groupe 18. Ianevski A, He L, Aittokallio T, Tang J.
Restoration of the tumor suppressor function Francophone Des Myélodysplasies (GFM). SynergyFinder: a web application for ana-
to mutant p53 by a low-molecular-weight Blood. 2019;134(Suppl 1):S677. lyzing drug combination dose-response
compound. Nat Med. 2002;8(3):282-288. 12. Lambert JMR, Gorzov P, Veprintsev DB, et matrix data. Bioinformatics. 2017;33(15):
7. Saha MN, Jiang H, Yang Y, Reece D, Chang al. PRIMA-1 reactivates mutant p53 by 2413-2415.

R. Birsen et al.
19. Lehmann S, Bykov VJN, Ali D, et al. death by GPX4. Cell. 2014;156(1):317-331. and synergism with conventional
Targeting p53 in vivo: a first-in-human 30. Gaschler MM, Andia AA, Liu H, et al. chemotherapeutic drugs in acute myeloid
study with p53-targeting compound APR- FINO2 initiates ferroptosis through GPX4 leukemia cells. Eur J Haematol. 2011;86
246 in refractory hematologic malignancies inactivation and iron oxidation. Nat Chem (3):206-215.
and prostate cancer. J Clin Oncol. Biol. 2018;14(5):507-515. 40. Nahi H, Lehmann S, Mollgard L, et al.
2012;30(29):3633-3639. 31. Perdrix A, Najem A, Saussez S, et al. Effects of PRIMA-1 on chronic lymphocytic
20. Friedmann Angeli JP, Schneider M, Proneth PRIMA-1 and PRIMA-1Met (APR-246): leukaemia cells with and without hemizy-
B, et al. Inactivation of the ferroptosis regu- from mutant/wild type p53 reactivation to gous p53 deletion. Br J Haematol. 2004;
lator Gpx4 triggers acute renal failure in unexpected mechanisms underlying their 127(3):285-291.
mice. Nat Cell Biol. 2014;16(12):1180-1191. potent anti-tumor effect in combinatorial 41. Miess H, Dankworth B, Gouw AM, et al.
21. Sun L, Wang H, Wang Z, et al. Mixed line- therapies. Cancers (Basel). 2017;9(12):172. The glutathione redox system is essential
age kinase domain-like protein mediates 32. Bykov VJN, Zache N, Stridh H, et al. to prevent ferroptosis caused by impaired
necrosis signaling downstream of RIP3 PRIMA-1(MET) synergizes with cisplatin lipid metabolism in clear cell renal cell car-
kinase. Cell. 2012;148(1-2):213-227. to induce tumor cell apoptosis. Oncogene. cinoma. Oncogene. 2018;37(40):5435-5450.
22. Takahashi N, Duprez L, Grootjans S, et al. 2005;24(21):3484-3491. 42. Tsoi J, Robert L, Paraiso K, et al. Multi-stage
Necrostatin-1 analogues: critical issues on 33. Lambert JMR, Moshfegh A, Hainaut P, differentiation defines melanoma subtypes
the specificity, activity and in vivo use in Wiman KG, Bykov VJN. Mutant p53 reacti- with differential vulnerability to drug-
experimental disease models. Cell Death vation by PRIMA-1 MET induces multiple induced iron-dependent oxidative stress.
Dis. 2012;3(11):e437. signaling pathways converging on apopto- Cancer Cell. 2018;33(5):890-904.e5.
23. Dixon SJ, Lemberg KM, Lamprecht MR, et sis. Oncogene. 2010;29(9):1329-1338. 43. Zou Y, Palte MJ, Deik AA, et al. A GPX4-
al. Ferroptosis: an iron-dependent form of 34. Duan W, Gao L, Wu X, et al. MicroRNA- dependent cancer cell state underlies the
nonapoptotic cell death. Cell. 2012; 34a is an important component of PRIMA- clear-cell morphology and confers sensitiv-
149(5):1060-1072. 1-induced apoptotic network in human ity to ferroptosis. Nat Commun.
24. Stockwell BR, Friedmann Angeli JP, Bayir lung cancer cells. Int J Cancer. 2010;127 2019;10(1):1617.
H, et al. Ferroptosis: a regulated cell death (2):313-320. 44. Yu Y, Xie Y, Cao L, et al. The ferroptosis
Nexus linking metabolism, redox biology, 35. Roh J-L, Kang SK, Minn I, Califano JA, inducer erastin enhances sensitivity of
and disease. Cell. 2017;171(2):273-285. Sidransky D, Koch WM. p53-Reactivating acute myeloid leukemia cells to chemother-
25. Feng H, Stockwell BR. Unsolved mysteries: small molecules induce apoptosis and apeutic agents. Mol Cell Oncol. 2015;
How does lipid peroxidation cause ferrop- enhance chemotherapeutic cytotoxicity in 2(4):e1054549.
tosis? PLoS Biol. 2018;16(5):e2006203. head and neck squamous cell carcinoma. 45. Jones CL, Stevens BM, D’Alessandro A, et
26. Koppula P, Zhang Y, Zhuang L, Gan B. Oral Oncol. 2011;47(1):8-15. al. Cysteine depletion targets leukemia
Amino acid transporter SLC7A11/xCT at 36. Izetti P, Hautefeuille A, Abujamra AL, et al. stem cells through inhibition of electron
the crossroads of regulating redox home- PRIMA-1, a mutant p53 reactivator, transport complex II. Blood. 2019;134(4):
ostasis and nutrient dependency of cancer. induces apoptosis and enhances 389-394.
Cancer Commun (Lond). 2018;38(1):12. chemotherapeutic cytotoxicity in pancreat- 46. Hoeks M, Yu G, Langemeijer S, et al.
27. Ishii T, Bannai S, Sugita Y. Mechanism of ic cancer cell lines. Invest New Drugs. Impact of treatment with iron chelation
growth stimulation of L1210 cells by 2- 2014;32(5):783-794. therapy in patients with lower-risk
mercaptoethanol in vitro. Role of the 37. Liu DSH, Read M, Cullinane C, et al. APR- myelodysplastic syndromes participating in
mixed disulfide of 2-mercaptoethanol and 246 potently inhibits tumour growth and the European MDS registry.
cysteine. J Biol Chem. 1981;256(23):12387- overcomes chemoresistance in preclinical Haematologica. 2020;105(3):640-651.
12392. models of oesophageal adenocarcinoma. 47. Wu J-R, Tuo Q-Z, Lei P. Ferroptosis, a
28. Dixon SJ, Patel DN, Welsch M, et al. Gut. 2015;64(10):1506-1516. recent defined form of critical cell death in
Pharmacological inhibition of cystine-glu- 38. Ogiwara H, Takahashi K, Sasaki M, et al. neurological disorders. J Mol Neurosci.
tamate exchange induces endoplasmic Targeting the vulnerability of glutathione 2018;66(2):197-206.
reticulum stress and ferroptosis. Elife. metabolism in ARID1A-deficient cancers. 48. Cardoso BR, Hare DJ, Bush AI, Roberts BR.
2014;3:e02523. Cancer Cell. 2019;35(2):177-190.e8. Glutathione peroxidase 4: a new player in
29. Yang WS, SriRamaratnam R, Welsch ME, et 39. Ali D, Jönsson-Videsäter K, Deneberg S, et neurodegeneration? Mol Psychiatry. 2017;
al. Regulation of ferroptotic cancer cell al. APR-246 exhibits anti-leukemic activity 22(3):328-335.

Acute Myeloid Leukemia ARTICLE
RXRA DT448/9PP generates a dominant Ferrata Storti Foundation

active variant capable of inducing maturation
in acute myeloid leukemia cells
Orsola di Martino,1* Margaret A. Ferris,2* Gayla Hadwiger,1 Soyi Sarkar,1
Anh Vu,1 María P. Menéndez-Gutiérrez,3 Mercedes Ricote3 and John S. Welch
1
Department of Internal Medicine, Washington University, St Louis, MO, USA;
2
Department of Pediatrics, Washington University, St Louis, MO, USA and 3Myocardial
Pathophysiology Area, Centro Nacional de Investigaciones Cardiovasculares (CNIC),
Madrid, Spain Haematologica 2022
*OdM and MAF contributed equally as co-first authors. Volume 107(2):417-426
ABSTRACT
R
ARA and RXRA contribute to myeloid maturation in both mice
and humans, and deletion of Rxra and Rxrb augments leukemic
growth in mice. While defining the domains of RXRA that are
required for anti-leukemic effects in murine KMT2A-MLLT3 leukemia
cells, we unexpectedly identified RXRA DT448/9PP as a constitutively
active variant capable of inducing maturation and loss of their prolifera-
tive phenotype. RXRA DT448/9PP was associated with ligand-indepen-
dent activity in reporter assays, with enhanced co-activator interactions,
reduced engraftment in vivo, and activation of myeloid maturation tran-
scriptional signatures that overlapped with those of cells treated with the
potent RXRA agonist bexarotene, suggestive of constitutive activity that
leads to leukemic maturation. Phenotypes of RXRA DT448/9PP appear
to differ from those of two other RXRA mutations with forms of consti-
tutive activity (F318A and S427F), in that DT448/9PP activity was resist-
ant to mutations at critical ligand-interacting amino acids
(R316A/L326A) and was resistant to pharmacological antagonists, sug-
gesting it may be ligand-independent. These data provide further evi-
dence that activated retinoid X receptors can regulate myeloid matura-
tion and provide a novel constitutively active variant that may be ger-
mane for broader studies of retinoid X receptors in other settings. Correspondence:
JOHN S. WELCH
Introduction jwelch@wustl.edu
Retinoid receptors are highly conserved transcription factors that direct hematopoi- Received: February 17, 2021.
etic self-renewal and differentiation.1,2 Retinoids (vitamin A metabolites) bind directly
to retinoid receptors, converting the receptors from transcriptional repressors to tran- Accepted: May 7, 2021.
scriptional activators. There are six types of retinoid receptor (RARA, RARB, RARG, Pre-published: June 17, 2021.
RXRA, RXRB, RXRG) with distinct tissue expression and subtle differences in their
binding and response to different ligands. The retinoid receptors RARA and RXRA
undergo remarkable upregulation during myeloid maturation in both mice and https://doi.org/10.3324/haematol.2021.278603
humans, whereas RXRG is not detected.3-5
Retinoid treatments, both in vitro and in vivo, facilitate hematopoietic stem cell mat-
uration and lineage commitment.1,2 Until now, the clinical application of retinoids in
hematology has been restricted to acute promyelocytic leukemia (treated with all-trans Material published in Haematologica is covered by copyright.
retinoic acid, or tretinoin, a pan-RAR ligand) and cutaneous T-cell leukemia (treated All rights are reserved to the Ferrata Storti Foundation. Use of
with bexarotene, a pan-RXR ligand). However, recent studies have observed activity conditions:
of all-trans retinoic acid when combined with chemotherapy in acute myeloid https://creativecommons.org/licenses/by-nc/4.0/legalcode.
leukemias other than the promyelocytic form, and the RARA super-enhancer has Copies of published material are allowed for personal or inter-
emerged as a potential biomarker of retinoid sensitivity.6-10 nal use. Sharing published material for non-commercial pur-
To better understand the molecular determinants of anti-leukemic retinoid activity, https://creativecommons.org/licenses/by-nc/4.0/legalcode,
we evaluated a series of RXRA truncations and mutations to determine which might sect. 3. Reproducing and sharing published material for com-
rescue phenotypes observed in Rxra/Rxrb-deficient leukemia cells.5 We unexpectedly mercial purposes is not allowed without permission in writing
identified RXRA DT448/9PP as a constitutively active variant capable of inducing mat- from the publisher.
uration and loss of proliferative capacity in leukemia cells. In this study, we character-
ized the activity of this variant using cell culture assays, transcription reporter assays,

O. di Martino et al.
in vivo engraftment, and RNA sequencing, and present evi- package edgeR was used to determine genes that were differen-
dence of ligand-independent loss of proliferative capacity, tially expressed between groups: RXRA WT cells, RXRA WT cells
transcriptional signatures of myeloid maturation that over- treated with bexarotene , and RXRA DT448/9PP-transduced cells.
lap with signatures induced by the potent RXR ligand Differentially expressed genes were defined as having an absolute
bexarotene, and inhibition of engraftment and leukemic log fold change >2 and a P-value <0.0001. Functional pathway
2
expansion in vivo. These data again suggest that programs of analysis was performed using Panther Gene Ontology Analysis
leukemic cell growth and maturation may be susceptible to software14 with a Fisher exact test using the Bonferroni correction
retinoids and provide a novel constitutively active tool for for multiple testing. Pathways were called if P<0.05 and fold
further delineating the function and activity of retinoid enrichment was >2. Gene set enrichment analysis was performed
receptors. using gene sets curated in the MSigDB database and GSEA 4.0.3
software; RRID:SCR_003199.15
Methods Study approval

All animal procedures were approved by the Institutional
Mice Animal Care and Use Committee of Washington University.
Mice were maintained in a specific pathogen-free barrier facility
with a 12 h light-dark cycle. Upon weaning, all mice were housed
in groups of up to five mice of the same sex per cage. Food and a Results
water bottle were provided in a recess of the metal wire lid at the
top of the cage. Cages were changed once every week. Six- to ten- Effects of RXRA DT448/9PP on maturation
week-old mice (C57Bl/6 background) were typically used for the We recently evaluated a series of RXRA truncations and
experiments. Equal numbers of male and female mice were used; mutations to map the domains of RXRA required for anti-
no gender biases were noted. leukemic effects in murine KMT2A-MLLT3 (formerly MLL-
AF9) leukemia cells derived from Mx-Cre x Rxraflox/flox x
Hematopoietic cell culture Rxrbflox/flox bone marrow cells (RXR-KO KMT2A-MLLT3
Murine bone marrow Kit+ cells were isolated using an leukemia).5 Ghosh et al. had previously observed that RXRA
Automacs Pro (Miltenyl Biotec, San Diego, CA, USA) according to DT448/9PP was associated with increased binding to an
the manufacturer’s protocol. Kit+ cells were plated in progenitor NCOR2 co-repressor (nuclear receptor corepressor 2, for-
expansion medium (RPMI 1640 medium, 15% fetal bovine serum, merly SMRT) peptide in a GST pull-down assay.16 We
stem factor [50 ng/mL], interleukin 3 [10 ng/mL], Flt3L [25 ng/mL], included this variant as part of the screen to evaluate co-
thrombopoietin [10 ng/mL], L-glutamine [2 mM], sodium pyru- repressor requirements. DT448/9 is positioned at the end of
vate [1 mM], HEPES buffer [10 mM], penicillin/streptomycin [100 the turn between helix 11 and 12. Helix 11 directly binds to
units/mL], b-mercaptoethanol [50 mM]) overnight and transduced co-repressors and co-activators.17 Helix 12 (the AF2 domain)
with MSCV-KMT2A-MLLT3 retrovirus by spinfection with 10 adopts an extended and unstructured conformation in the
µg/mL polybrene and 10 mM HEPES at 2400 rpm, 30°C for 90 min absence of ligand and assumes a repositioned and helical
in an Eppendorf 5810R centrifuge. Cells were transplanted into conformation following ligand binding (Figure 1A).18
sublethally irradiated mice and subsequent leukemia harvested 4 Unexpectedly, retroviral expression of RXRA DT448/9PP
to 6 months later. KMT2A-MLLT3 leukemia cells were cultured in (MSCV-RXRA DT448/9PP-IRES-mCherry) in RXR-KO
vitro using similar media, but without Flt3L or thrombopoietin. KMT2A-MLLT3 leukemia cells induced the normally round
KMT2A-MLLT3-RXR-knockout (KO) leukemia cells were derived cells to form adherent clumps that were mCherry+ on the
from Mx1-Cre x Rxraflox/flox x Rxrbflox/flox bone marrow cells11,12 trans- tissue culture plate (Figure 1B). mCherry+ cells were also
duced with MSCV-KMT2A-MLLT3 retrovirus and generated as associated with loss of leukemic colony-forming capacity,
described above. RXR deletion was induced by injecting Mx1-Cre lack of proliferation, and macrophage-like cytomorphology
x Rxraflox/flox x Rxrbflox/flox mice intraperitoneally with pIpC 300 with ruffled borders and vacuoles (Figure 1C-E). These
mg/mouse; four doses were given, every other day. RXR deletion effects were absent in RXR-KO KMT2A-MLLT3 cells retro-
was confirmed by polymerase chain reaction analysis 4 weeks virally transduced with RXRA WT (MSCV-RXRA-IRES-
after mice had been treated with pIpC. THP1, and K562 cells were mCherry), with other tested RXRA domain deletions or
obtained from the American Type Culture Collection. Monomac- point mutations, or with treatment combinations of RXRA
6 and OCI-AML3 cells were obtained from the Deutsche and RARA activating ligands5 (Figure 1B). In addition, we
Sammlung von Mikroorganismen und Zellkulturen. found that retroviral expression of RXRA DT448/9PP
(MSCV-RXRA DT448/9PP-IRES-mCherry) in KMT2A-
RNA sequencing MLLT3 WT leukemia cells induced the same phenotype as
KMT2A-MLLT3 leukemia cells were transduced with MSCV- that observed in RXR-KO KMT2A-MLLT3 leukemia cells
RXRA-IRES-mCherry or MSCV-RXRA DT448/9PP-IRES-mCherry (lack of proliferation and macrophage-like cytomorpholo-
and treated with or without 250 nM bexarotene. After 24 h, gy) (data not shown). We examined the resulting proteins on
mCherry+ cells were sorted and total RNA was extracted with western blot after transduction into KMT2A-MLLT3 WT
TRIzol LS (Ambion Life Technologies) and isolated on PureLink cells (Figures 1F). RXRA DT448/99PP was associated with a
RNA Kit columns (Thermo Fisher Scientific). The quality of RNA similar sized band as wild-type RXRA, but it appeared
was measured on a 2100 Bioanalyzer (Agilent). Sequencing resistant to the proteolytic cleavage that results in the for-
libraries, each with individual Illumina indexes, were constructed mation of a truncated 36 kDa fragment.19 Endogenous Rxra
using the TruSeq Stranded mRNA procedure (Sample Prep Kit v2; levels were significantly lower than levels of retrovirally
Illumina). Libraries were sequenced as paired-end 151 bp reads on expressed RXRA (Figure 1G). Thus, retroviral transduction
an Illumina NovaSeq instrument. Reads were aligned to the mm10 led to similar levels of expression of RXRA WT and RXRA
mouse reference genome, and transcription was quantified using DT448/9PP, at supraphysiological protein levels compared
kallisto version 0.43.1 13 and Ensembl transcripts version 95. The R to endogenous RXRA.

RXRA active mutation
A B
C D
E F G
Figure 1. Monocytic maturation induced by RXRA DT448/9PP. (A) DT448/9 amino acids are highlighted in red within the structure of the RXRA ligand binding domain
(PDB 4K4J). An Ncoa2 peptide is highlighted in orange. Note that DT448/9 occurs at the beginning of helix 12, the AF2 domain, and the rigid proline substitution
could enhance helix formation. (B, C) RXR-KO KMT2A-MLLT3 leukemia cells were transduced with MSCV-RXRA DT448/9PP-IRES-mCherry or with MSCV-RXRA WT-
IRES-mCherry. Cells were evaluated under fluorescent and light microscopy at 72 h. (B) Transduced cells were sorted for mCherry+ cells and plated in methylcellulose,
and colonies assessed in technical triplicates on day 7. Statistical significance evaluated using the t-test, ***P<0.001 (C). (D) RXR-KO KMT2A-MLLT3 leukemia cells
were transduced with MSCV-RXRA DT448/9PP-IRES-mCherry, stained with FxCycle violet, and retention of the dye was assessed by flow cytometry at the indicated
time points comparing mCherry+ (RXRA DT 448/9 PP) versus mCherry– (RXR-KO) cells. (E) Cytomorphology of mCherry+ (RXRA DT 448/9 PP) versus mCherry– (RXRA
WT endogenous) sorted KMT2A-MLLT3 leukemia cells after transduction with MSCV-RXRA DT448/9PP-IRES-mCherry. (F) RXR-KO KMT2A-MLLT3 leukemia cells were
transduced with MSCV-RXRA DT448/9PP-IRES-mCherry or with MSCV-RXRA-IRES-mCherry and protein expression was evaluated through western blot analysis using
anti-RXRA antibody (H-10, Santa Cruz). GAPDH was used as a loading control. (G) KMT2A-MLLT3 WT leukemia cells were transduced with MSCV-RXRA WT-IRES-
mCherry as indicated, and protein expression was evaluated through western blot analysis using anti-RXRA antibody (5388, Cell Signaling). GAPDH was used as a
loading control. (H) KMT2A-MLLT3 WT leukemia cells were transduced as indicated and analyzed by flow cytometry. RXRA WT cells were analyzed with and without
treatment with 250 nM bexarotene for 24 h. RAW 264.7 cells were used as a positive staining control, but not included in the statistical comparisons. Statistical sig-
nificance was evaluated using analysis of variance with the Tukey correction for multiple comparisons, *P<0.05. **P<0.01. ***P<0.001.

Because the RXRA DT448/9PP-transduced cells acquired for ligand-dependent interactions of PGC1a with nuclear
cell adhesion properties and ruffled borders suggestive of receptors. Thus, RXRA DT448/9PP appears to constitutive-
macrophage-like maturation, we assessed immunopheno- ly engage PGC1a using domains outside the canonical N-
typic changes. Bexarotene treatment of RXRA WT trans- terminal LXXLL motifs.
duced cells resulted in an increased percentage of cells
expressing F4/80, CD11b, CD14, and CD64, as determined Comparison with other active RXRA variants
by flow cytometry, and increased the median fluorescence Two other RXRA variants with constitutively active prop-
intensity (MFI) of F4/80, CD11b, CD14, CD64, Ly6c, and erties have been described. First, mouse Rxra F318A exhibits
Gr1, but had little effect on CD115 expression. RXRA constitutively active phenotypes.22 However, when a crystal
DT448/9PP transduction resulted in increased percentages structure was generated, the ligand-binding pocket con-
of cells expressing CD11b, CD14, and CD64, with tained oleic acid, and Rxra F318A activity could be inhibited
increased MFI of CD11b, CD14, CD64, Ly6c and Gr1, con- by the pan-RXR antagonist HX531, suggesting that the
sistent with myeloid and monocytic maturation, but lack- mutation leads to augmented responsiveness to a natural lig-
ing the classical macrophage marker CD115 (Figure 1H, and present in the tissue culture, and is not completely lig-
Online Supplementary Figure S1). and-independent.23
Transduction of RXRA DT448/9PP, but not RXRA WT We assessed the sensitivity of RXRA DT448/9PP to two
into two KMT2A-MLLT3-associated human cells (THP1 pan-RXR antagonists (HX531 and UVI3003) across a series
and MonoMac-6) also resulted in adherence to plastic with of assays. In RXR-WT KMT2A-MLLT3 leukemia cells trans-
the development of podocytes and ruffled borders (Online duced with RXRA DT448/9PP, cell adhesion, clumping, and
Supplementary Figure S2). These effects required 4 to 6 days loss of proliferation phenotypes were not abrogated by
to emerge, whereas adhesion and cell clumping in murine either compound (Online Supplementary Figure S4).
KMT2A-MLLT3 cells typically required only 3 or 4 days. In To further assess whether RXRA DT448/9PP pheno-
contrast, Kit+ murine bone marrow cells cultured in stem types may result from hyper-responsiveness to intracellu-
cell media (stem cell factor, interleukin 3, thrombopoietin, lar natural ligands, we mutated two amino acids (R316
Flt3L) and K562 cells were resistant to the effects of RXRA and L326) that form critical ionic bonds with the car-
DT448/9PP, and the cells remained round and non-adherent boxylic acid group in ligands (e.g., bexarotene, 9-cis
(Online Supplementary Figures S2D and S3A, B). retinoic acid, and long-chain fatty acids).5 RXRA
R316A/L326A has been previously shown to abrogate lig-
RXRA DT448/9PP is constitutively active and-dependent activation and is unable to rescue response
We assessed the functional activity of RXRA DT448/9PP to bexarotene in RXR-KO KMT2A-MLLT3 leukemia
using multiple reporter systems. First, we used a UAS/Gal4 cells.5,24 Retroviral expression of the compound variant
reporter.11 KMT2A-MLLT3 WT leukemia cells derived from RXRA R316A/L326A/DT448/9PP again led to overexpres-
UAS-GFP bone marrow cells were retrovirally transduced sion on western blot and resistance to proteolytic cleav-
with a fusion of the Gal4 DNA binding domain and the age, but did not abrogate cell clumping, loss of prolifera-
RXRA wild-type ligand-binding domain (Gal4-RXRA) or a tion, or loss of colony formation (Figure 3A-D).
fusion of the Gal4 DNA-binding domain and the RXRA Second, recurrent RXRA hot-spot mutations (S427F/Y)
DT448/9PP ligand-binding domain (Gal4-RXRA occur in patients with bladder cancer, and these augment the
DT448/9PP). Gal4-RXRA DT448/9PP resulted in constitu- activity of the PPARG:RXRA heterodimer, but are not capa-
tive activity uninfluenced by increasing concentrations of ble of independently activating RXRA reporters.25 Retroviral
bexarotene (Figure 2A). Constitutive activity of RXRA expression of RXRA S427F also led to strong overexpression
DT449/8PP was also observed in Kit+ UAS-GFP bone mar- of the variant, which retained sensitivity to proteolytic
row cells transduced with Gal4-RXRA retroviruses (Online cleavage, although cell adhesion, cell clumping, cell prolifer-
Supplementary Figure S3C). ation, and colony-forming properties were not altered
Second, we evaluated three different luciferase reporters (Figure 3E-H).
in 293T cells. 293T cells are known to express endogenous To determine whether similar heterodimerization with
RXRA as well as various of its partners; RARA, PPARG, and PPARG or with other nuclear receptors may play essential
VDR.20 Reporters assays included a synthetic direct repeat 1 roles in the activity of RXRA DT448/9PP, we evaluated phe-
(DR1) peroxisome proliferator response element (PPRE); a notypes using a series of receptor antagonists. Concurrent
DR1 response element from the ApoA1 promoter; and a treatment with potent antagonists of retinoic acid receptors,
DR5 response element from the RARB promoter. We noted LXR, PPARA, and PPARG did not affect clumping, adhesion
that RXRA DT448/9PP transfection consistently led to lig- to plastic, or proliferation phenotypes (Online Supplementary
and-independent activation of all three constructs (Figure Figure S5), demonstrating that the DT448/9PP creates a non-
2B). A control RXRA construct contained a deletion of the permissive receptor that is functionally active, independent-
AF2 domain (RXRA DAF2), and this construct is unrespon- ly of other activated nuclear receptors.
sive to ligand. Using this approach we assessed whether
RXRA DT448/9PP might be sensitive to inhibition by two RXRA DT448/9PP inhibits leukemic engraftment
pan-RXR antagonists, HX531 and UVI3003. Neither com- and expansion
pound inhibited activation of the UAS/Gal4 reporter or the KMT2A-MLLT3 WT cells were transduced with RXRA
RARE-Luc reporter (Figure 2C). WT or RXRA DT448/9PP retroviruses labeled with IRES-
We used a mammalian two-hybrid assay to assess the mCherry cassettes (Figure 4A). To limit maturation effects in
interaction of RXRA DT448/9PP with the co-activator vitro, the populations were immediately transplanted into
PGC1a, noting constitutive, ligand-independent binding recipient mice and residual cells were subsequently ana-
(Figure 2D, E). This activity remained present with the lyzed by flow cytometry. After 4 weeks of engraftment and
L2/3A variant (L147A, L148A, L210A, L211A),21 which con- expansion, leukemia cells in the peripheral blood were
tains point mutations in the critical LXXLL motifs required assessed and we observed that mice transplanted with

RXRA WT displayed increased leukocytosis, an increase in 0.095%-0.36%) (Figure 4E). When we assessed moribund
neutrophils, and a decrease in lymphocytes compared to mice transplanted with cells transduced with RXRA WT,
mice transplanted with RXRA DT448/9PP-transduced cells we observed further reduction in the absolute proportion of
(Figure 4B-D). Compared with the pre-engraftment popula- mCherry+ cells in the bone marrow rather than an expan-
tion (RXRA WT, 55% mCherry+; RXRA DT448/9PP, 62% sion: RXRA WT, average 1% (range, 1.21%-1.46%); RXRA
mCherry+), at 4 weeks there was reduced engraftment in DT448/9PP, average 0.741 (range, 0.022%-1.46%) (Figure
RXRA DT448/9PP cells relative to RXRA WT cells: RXRA 4G). Survival was shorter in mice transplanted with RXRA
WT average of 7.6% mCherry+ cells (range, 0.94%-16.4%) WT-transduced cells than in those transplanted with RXRA
versus RXRA DT448/9PP 0.25% mCherry+ cells (range, DT448/9PP-transduced cells (Figure 4H). At the time of sac-
A B C
Figure 2. Constitutive activity of RXRA DT448/9PP. (A) KMT2A-MLLT3 WT leukemia cells derived from UAS-GFP bone marrow cells were transduced with MSCV-Gal4-
RXRA DT448/9PP-IRES-mCherry or MSCV-Gal4-RXRA-IRES-mCherry, treated with increasing concentrations of bexarotene, and the ratio of GFP+ cells (responding)
to total mCherry+ cells (capable of responding) was determined. Data from biological duplicates are shown. (B, C) 293T cells were transfected with the indicated plas-
mids, treated with bexarotene, HX531, or UVI3003 (all 1 mM), and luciferase was measured after 40 h. Data from biological triplicates are shown. PPRE: peroxisome
proliferator response element. ApoA1: DR1 element from the ApoA1 promoter. RARE: retinoic acid receptor response element from the RARB promoter. The ∆AF2
deletion acts as a negative control. (D, E) Mammalian two-hybrid. 293T cells were transfected with the indicated plasmids, treated with bexarotene and GFP was
measured by flow cytometry after 40 h. Data from biological triplicates are shown. GFP-N1: a positive control CMV-GFP expression vector. GFP-N0: a negative control
derived from GFP-N1 after the deletion of the CMV promoter. Statistical significance was evaluated using analysis of variance with the Bonferroni comparison to con-
trol, *P<0.05, **P<0.01, ***P<0.001.

rifice, all but one mouse had developed splenomegaly (aver- consequences with RXRA WT activation by bexarotene,
age spleen weight 0.5 g) and bone marrow lacked erythroid KMT2A-MLLT3 WT leukemia cells were transduced with
elements, features that are typical of overwhelming second- RXRA WT or RXRA DT448/9PP and the transcriptional pro-
ary leukemic engraftment (Online Supplementary Figure S6). files of the transduced cells were assessed by RNA sequenc-
ing (Figure 5A). We noted significant overlap of regulated
Transcriptional consequences of RXRA DT448/9PP genes between RXRA WT cells treated with bexarotene and
To assess the maturation effects of RXRA DT448/9PP cells transduced with RXRA DT448/9PP (Figure 5B-D). Of
more comprehensively and to investigate the overlapping 598 genes that were upregulated in RXRA WT cells treated
A B
C D
E F
G H
Figure 3. Comparison with other RXRA variants with activity. KMT2A-MLLT3 WT leukemia cells were transduced with MSCV-3xFlag-RXRA R316A/L326A/DT448/9PP-
IRES-mCherry or with MSCV-RXRA S427F-IRES-mCherry and evaluated under fluorescent and light microscopy at 72 h (A and E). RXR-KO KMT2A-MLLT3 leukemia
cells were transduced with MSCV-3xFlag-RXRA R316A/L326A/DT448/9PP-IRES-mCherry or MSCV-RXRA S427F-IRES-mCherry, stained with FxCycle Violet, and
retention of the dye was assessed by flow cytometry at the indicated time points (B and F). RXR-KO KMT2A-MLLT3 leukemia cells were transduced with MSCV-RXRA
WT-IRES-mCherry or MSCV-3xFlag-RXRA R316A/L326A/DT448/9PP-IRES-mCherry or MSCV-RXRA S427F-IRES-mCherry, mCherry+ cells were then sorted and plated
in methylcellulose, and colonies assessed in technical triplicates on day 7 (C and G). RXR-KO KMT2A-MLLT3 leukemia cells were transduced with MSCV-RXRA-IRES-
mCherry or MSCV-RXRA DT448/9PP-IRES-mCherry or MSCV-RXRA S427F-IRES-mCherry and protein expression was evaluated through western blot analysis using
anti-RXRA antibody (H-10, Santa Cruz). GAPDH was used as a loading control (D and H). Statistical significance was evaluated using the t-test, ***P<0.001.

with bexarotene, 525 overlapped with those of RXRA downregulation of “Down” genes) compared to untreated
DT448/9PP-transduced cells. Likewise, of 74 downregulated WT cells (Figure 5E, F).
genes in RXRA WT cells treated with bexarotene, 28 over- Within this set of differentially expressed genes, we eval-
lapped with those in RXRA DT448/9PP-transduced cells. uated transcripts associated with cell surface proteins.
Gene ontology pathway analysis showed enrichment of Sixty-seven cell surface transcripts were differentially
lipid metabolism and immune system signaling and down- expressed, according to RNA sequencing, between RXRA
regulation of RNA processing (Online Supplementary Tables WT cells versus RXRA WT cells treated with bexarotene or
S1 and S2). Gene set enrichment analysis of a curated list of RXRA DT448/9PP-transduced cells (Online Supplementary
genes regulated during myeloid development26 indicated Figure S7A). As in the overall RNA sequencing analysis,
that treatment of RXRA WT cells with bexarotene and these genes were associated with similar or greater upregu-
expression of RXRA DT448/9PP both lead to a myeloid dif- lation in DT448/9PP-transduced cells than in the wild-type
ferentiation phenotype (upregulation of “Up” genes and cells treated with bexarotene (Online Supplementary Figure
B C D
E F G
Figure 4. In vivo engraftment with RXRA WT cells

versus RXRA DT448/9PP-transduced cells. (A)
Experimental schema. KMT2A-MLLT3 WT
leukemia cells were transduced with MSCV-RXRA
H WT-IRES-mCherry or MSCV-RXRA DT448/9PP-
IRES-mCherry and transplanted immediately into
sublethally irradiated mice. (B-D) Peripheral blood
assessment at 4 weeks. WBC: white blood cells.
(E) Assessment of proportion of mCherry+ cells
before transplantation. (F) Assessment of mCherry
in peripheral blood at 4 weeks. (G) Assessment of
mCherry in bone marrow of moribund mice (of
note, 2 RXRA DT448/9PP mice were found dead
and were unable to be assessed and 1 was sacri-
ficed but did not have splenomegaly). (H) Survival
of mice after transplantation. *P<0.05,
**P<0.01. The Mann-Whitney test was used in (B,
F and G) with unequal variance. Otherwise, the t-
test was used.

S7B-F). The list of upregulated genes included CD14 and bexarotene and RXRA DT448/9PP-transduced cells (Figure
CD64 (Fcgr1), which at the protein level were observed to 5B-D), there were 11 genes that were differentially
increase by immunofluorescence (Figure 1H, Online expressed between these two groups (Online Supplementary
Supplementary Figure S1C-D). The most highly differentiat- Figure S8A). In five, RXRA DT448/9PP augmented the
ed transcript out of the cell surface markers was the com- bexarotene-induced response: (upregulated Vsig8, Fcgr1
plement factor C3 (Online Supplementary Figure S7D), and (Cd64), Camp, Gp6; downregulated: Pik3ip1) (Online
additional complement-receptors (C5ar1 and CD59a) also Supplementary Figure S8B). The other six displayed three dif-
underwent upregulation with RXRA activation (Online ferent patterns: not expressed in treated or untreated RXRA
Supplementary Figure S7B, F), which correlate with mono- WT cells and increased in DT448/9PP-transduced cells
cytic differentiation. We also noted a variety of upregulated (Gpr84, S100a8); downregulated from untreated to treated
integrins (Bcam, Ceacam1, Itgam, Itgb2, Itgb2l, and Itgb7), RXRA WT cells and expressed in DT448/9PP-transduced
and these may play a role in the aggregation phenotype of cells (Mpo, Gm28438, Gm10359); or not expressed in
the DT448/9PP variant (Online Supplementary Figure S7C-F). untreated RXRA WT with an increase upon bexarotene
While the majority of transcripts that were up- or down- treatment but not with DT448/9PP (Cited1) (Online
regulated overlapped between RXRA WT cells treated with Supplementary Figure S8C). This list is too small for pathway
A B
C D
E F
Figure 5. RXRA DT448/9PP transcriptome changes. (A) Schema of experimental design: KMT2A-MLLT3 WT leukemia cells were transduced with the indicated virus-
es and cultured for 24 h with or without 250 nM bexarotene; mCherry+ cells were sorted, and total RNA was isolated for RNA sequencing analysis. (B) Heatmap of
differentially expressed genes (DEG) between all three groups. Each condition was evaluated by biological triplicates. WT: RXRA WT cells; WT + bex: RXRA WT cells
treated with bexarotene; DT448/9PP: RXRA DT448/9PP-transduced cells. (C) Venn diagram with the number of upregulated unique DEG from RXRA WT cells versus
RXRA WT cells treated with bexarotene (red) and RXRA WT cells versus RXRA DT448/9PP-transduced cells (blue) and DEG common to both groups (purple). (D) Venn
diagram with the number of downregulated unique DEG from RXRA WT cells versus. RXRA WT cells treated with bexarotene (red) and RXRA WT cells versus RXRA
DT448/9PP-transduced cells (blue) and DEG in common between both groups (purple). (E, F) Gene set enrichment analysis comparing the untreated RXRA WT cells
to the bexarotene-treated and DT448/9PP-transduced cells upregulated (E) and downregulated (F) during myelopoiesis (published gene sets from Brown, et al.26).
Heatmaps are shown as row z-scores with the color key below the panel (B). FDR: false discovery rate; NES: normalized enrichment score.

analysis, but manual curation found that the most of these lines with KMT2A-MLLT3 (THP1 and MonoMac-6) were
11 transcripts are related to immune function and activity. susceptible, whereas the heterogeneous stem/progenitor
populations of Kit+ murine bone marrow cells and the blast
phase chronic myeloid leukemia cell line K562 were not
Discussion (Online Supplementary Figures S2 and S3). Maintained in
stem cell cytokines, Kit+ bone marrow cells have multipo-
Retinoic X receptors and retinoic acid receptors have tent potential and their medium lacks cytokines that might
been shown to influence myeloid maturation.1,2,4,27,28 RXRA provide monocytic maturation support. K562 cells are sus-
protein levels and activity have been linked to cell fate deci- ceptible to erythroid maturation stimuli rather than
sions at the neutrophil versus monocyte decision,29 and dele- myeloid maturation. Other groups have observed differ-
tion of Rxra and Rxrb prevents osteoclast maturation.12 and ences in sensitivity and resistance to retinoids across cell
augments KMT2A-MLLT3 leukemogenesis and cell expan- lines.9,30,31 Thus, different external signals and/or internal
sion in vivo,5 suggesting a role for retinoic X receptors in reg- priming states may affect the susceptibility to retinoids and
ulating myeloid maturation. to RXRA DT448/9PP.
RXRA DT448/9PP is a serendipitously discovered muta- We have previously noted that Rxra and Rxrb expression
tion that results in constitutive activity, leukemic cell matu- negatively regulates KMT2A-MLLT3 leukemia and that
ration, and loss of proliferative capacity. This variant these cells are exposed to low levels of natural RXRA lig-
demonstrates ligand-independent activation, augmented ands in vivo.5,32 Similarly, here we found that cells that over-
co-activator binding, induced maturation transcripts by express RXRA WT consistently engraft in recipient mice,
RNA sequencing, and reduced engraftment in vivo. To our but are associated with a competitive disadvantage relative
knowledge, RXRA DT448/9PP has not been spontaneously to untransduced, mCherry-negative cells (Figure 4). This
observed in cancer or other pathological states. phenotype was augmented by transduction with RXRA
Two other RXRA mutations with activating phenotypes DT448/9PP, which further limited engraftment and
have been described. Mouse Rxra F318A has increased tran- leukemic outgrowth in vivo, again suggesting the potential
scriptional activity, potentially via increased responsiveness of activated retinoic X receptors to inhibit growth of
to cell-available oleic acid.22 However, Rxra F318A retains leukemia cells in vivo.
sensitivity to RXRA antagonists, whereas RXRA RXRA DT448/9PP resulted in several maturation-related
DT448/9PP did not (Figure 2C, Online Supplementary Figure phenotypes, including lack of proliferation, loss of colony
S2). Furthermore, R316 and L326 are critical amino acids formation, acquisition of ruffled borders and podocytes,
that interact directly with bound ligand and are required for and increased expression of cell surface markers of myeloid
ligand activation of RXRA.5,24 Combining these ligand- and monocytic maturation. Transcriptional analysis of
blocking mutations with DT448/9PP, we noted retained bexarotene-treated cells versus RXRA DT448/9PP-trans-
KMT2A-MLLT3 maturation phenotypes (Figure 3A-C), fur- duced cells suggested strong overlapping myeloid and
ther suggesting ligand-independent activity of RXRA monocytic maturation signatures, consistent with ligand-
DT448/9PP. Recurrent RXRA hot-spot mutations (S427F/Y) independent, constitutively active effects of DT448/9PP.
have been noted in patients with bladder cancer. These Many maturation-related transcripts were more effectively
mutations augment the activity of the PPARG:RXRA het- induced by RXRA DT448/9PP than by bexarotene, and a
erodimer and are not capable of independently activating few myeloid-related transcripts were uniquely induced by
RXRA in reporter assays.25 We previously found that com- RXRA DT448/9PP. Thus, DT448/9PP may more effectively
binations of RARA and RXRA ligands lead to leukemic activate the same loci as ligand-activated RXRA, and activ-
maturation and apoptosis, whereas PPARG and RXRA lig- ity at select novel loci may enable additional phenotypes.
ands did not.5 Here we found that RXRA S427F did not Like other nuclear receptors, the retinoic X receptors are
recapitulate leukemia maturation phenotypes in KMT2A- ligand-dependent transcription factors and their function
MLLT3 leukemia cells (Figure 3E-G), and the PPARG antag- and activity change from transcriptional repressors to tran-
onist T0070907 did not abrogate RXRA DT448/9PP pheno- scriptional activators in the presence of active ligands.
types (Online Supplementary Figure S5). Thus, in contrast to Multiple natural ligands have been proposed for the retinoic
Rxra F318A and RXRXA S427F, RXRA DT448/9PP may be X receptors,6 and it is often difficult to know which cells
constitutively active, independently of endogenous, avail- and settings contain active, natural ligands, and when an
able, natural ligands, and also may not depend on activation observed effect of retinoic X receptors may be related to the
through PPARG:RXR heterodimers or other major nuclear receptor function in the absence versus presence of ligand. A
receptor heterodimers. constitutively active variant provides a helpful genetic com-
RXRA DT448/9PP constitutive activity may be related to parator and could be used with diverse forms of previously
at least two phenotypes. RXRA DT448/9PP is resistant to characterized non-functional (e.g., DDBD, E153G/G154S)
enzymatic cleavage, and this may enable augmented func- and ligand-non-responsive variants (e.g., DAF2,
tional protein levels (Figures 1F and 3D, H), but cannot R316A/326A).
explain resistance to pan-RXR antagonists or co-mutation Retinoid receptors have been an attractive therapeutic
with R316A/L326A (Figures 2C and 3A-C, Online target in acute myeloid leukemia; however their effects in
Supplementary Figure S4). RXRA DT448/9PP was also asso- clinical trials have been modest. The maturation effects
ciated with ligand-independent co-activator binding (Figure (loss of proliferation and colony-forming potential, and
2D), which may enable augmented, or even constitutive morphological changes) of DT448/9PP on KMT2A-MLLT3
activity. A limitation of these studies is that they require a leukemia cells are more robust than the maximal effects of
retroviral overexpression system and therefore the effect of retinoid ligands such as all-trans retinoic acid and
DT448/9PP RXRA at physiological levels is unknown. bexarotene. Although further delineation of mechanistic
Not all cells tested were susceptible to RXRA DT448/9PP. activity and cell-type susceptibility rules will be required,
The two human myelomonocytic acute myeloid leukemia these data suggest that current small-molecule retinoids

may incompletely activate the maturation pathways regu- Acknowledgments

lated by the retinoid receptors and that additional, unreal- We thank the Alvin J. Siteman Cancer Center at Washington
ized potential may exist for retinoids in the treatment of University School of Medicine and Barnes-Jewish Hospital in St.
forms of acute myeloid leukemia other than non-acute Louis, (MO, USA). for the use of the Flow Cytometry Core. The
promyelocytic leukemia. In particular, myeloid maturation Siteman Cancer Center is supported in part by a National Cancer
programs may be augmented by correctly activated retinoic Institute Cancer Center Support grant (P30 CA91842). We thank
X receptors and cells with monocytic potential may be sus- Christopher Miller, Sai Ramakrishnan, Deborah Laflamme,
ceptible to retinoid-induced maturation. An efficient, con- Conner York, and Sangeetha Vadivelu for technical assistance.
stitutively active RXRA variant may enable further elucida-
tion of this potential. Funding
This work was supported by National Institutes of Health grant
Disclosures R01 HL128447 (JSW), by the Siteman Investment Program
No conflicts of interest to disclose. (JSW), the Washington University SPORE DRP (JSW and MAF),
the Children’s Discovery Institute (JSW), the Alex’s Lemonade
Contributions Stand Foundation Young Investigator Award (MAF), the National
JSW, OdM and MAF designed and performed experiments, Institutes of Health 5K12HD07622408 (MAF), and grants from
and wrote the manuscript. GH, SS, AV, MPMG, and MR the Spanish Ministerio de Ciencia e Innovación (MCI) (SAF2017-
designed and performed experiments. 90604-REDT-NurCaMeIn, RTI2018-095928-BI00) (MR).
References mouse bone marrow cells and implications 22. Kersten S, Dong D, Lee W, Reczek PR, Noy
for PML-RARA transformation. Blood N. Auto-silencing by the retinoid X receptor.
Cancer J. 2015;5(2):e284. J Mol Biol. 1998;284(1):21-32.
1. Oren T, Sher JA, Evans T. Hematopoiesis 12. Menendez-Gutierrez MP, Roszer T, Fuentes 23. Bourguet W, Vivat V, Wurtz JM, Chambon P,
and retinoids: development and disease. L, et al. Retinoid X receptors orchestrate Gronemeyer H, Moras D. Crystal structure
Leuk Lymphoma. 2003;44(11):1881-1891. osteoclast differentiation and postnatal bone of a heterodimeric complex of RAR and
2. Evans T. Regulation of hematopoiesis by remodeling. J Clin Invest. 2015;125(2):809- RXR ligand-binding domains. Mol Cell.
retinoid signaling. Exp Hematol. 2005;33(9): 823. 2000;5(2):289-298.
1055-1061. 13. Bray NL, Pimentel H, Melsted P, Pachter L. 24. Hiromori Y, Aoki A, Nishikawa J, Nagase H,
3. Ricote M, Snyder CS, Leung HY, Chen J, Near-optimal probabilistic RNA-seq quan- Nakanishi T. Transactivation of the human
Chien KR, Glass CK. Normal hematopoiesis tification. Nat Biotechnol. 2016;34(5):525- retinoid X receptor by organotins: use of
after conditional targeting of RXRalpha in 527. site-directed mutagenesis to identify critical
murine hematopoietic stem/progenitor cells. 14. Ashburner M, Ball CA, Blake JA, et al. Gene amino acid residues for organotin-induced
J Leukoc Biol. 2006;80(4):850-861. ontology: tool for the unification of biology. transactivation. Metallomics. 2015;7(7):
4. Welch JS, Klco JM, Gao F, et al. Combination The Gene Ontology Consortium. Nat 1180-1188.
decitabine, arsenic trioxide, and ascorbic Genet. 2000;25(1):25-29. 25. Halstead AM, Kapadia CD, Bolzenius J, et
acid for the treatment of myelodysplastic 15. Subramanian A, Tamayo P, Mootha VK, et al. Bladder-cancer-associated mutations in
syndrome and acute myeloid leukemia: a al. Gene set enrichment analysis: a knowl- RXRA activate peroxisome proliferator-acti-
phase I study. Am J Hematol. 2011;86(9): edge-based approach for interpreting vated receptors to drive urothelial prolifera-
796-800. genome-wide expression profiles. Proc Natl tion. Elife. 2017;6:e30862.
5. Di Martino O, Niu H, Hadwiger G, et al. Acad Sci U S A. 2005;102(43):15545-15550. 26. Brown AL, Wilkinson CR, Waterman SR, et
Endogenous and combination retinoids are 16. Ghosh JC, Yang X, Zhang A, et al. al. Genetic regulators of myelopoiesis and
active in myelomonocytic leukemias. Interactions that determine the assembly of leukemic signaling identified by gene profil-
Haematologica. 2021;106(4):1008-1021. a retinoid X receptor/corepressor complex. ing and linear modeling. J Leukoc Biol.
6. Martino OD, Welch JS. Retinoic acid recep- Proc Natl Acad Sci U S A. 2002;99(9):5842- 2006;80(2):433-447.
tors in acute myeloid leukemia therapy. 5847. 27. Purton LE. Roles of retinoids and retinoic
Cancers. 2019;11(12):1915 17. Lee WY, Noy N. Interactions of RXR with acid receptors in the regulation of
7. Lubbert M, Grishina O, Schmoor C, et al. coactivators are differentially mediated by hematopoietic stem cell self-renewal and
Valproate and retinoic acid in combination helix 11 of the receptor's ligand binding differentiation. PPAR Res. 2007;2007:87934.
with decitabine in elderly nonfit patients domain. Biochemistry. 2002;41(8):2500- 28. Mullen EM, Gu P, Cooney AJ. Nuclear
with acute myeloid leukemia: results of a 2508. receptors in regulation of mouse ES cell
multicenter, randomized, 2 x 2, phase II trial. 18. Cordeiro TN, Sibille N, Germain P, et al. pluripotency and differentiation. PPAR Res.
J Clin Oncol. 2020:38(3):257-270. Interplay of protein disorder in retinoic acid 2007;2007:61563.
8. vSchlenk RF, Frohling S, Hartmann F, et al. receptor heterodimer and its corepressor 29. Taschner S, Koesters C, Platzer B, et al.
Phase III study of all-trans retinoic acid in regulates gene expression. Structure. Down-regulation of RXRalpha expression is
previously untreated patients 61 years or 2019;27(8):1270-1285. essential for neutrophil development from
older with acute myeloid leukemia. 19. Gao W, Liu J, Hu M, et al. Regulation of pro- granulocyte/monocyte progenitors. Blood.
Leukemia. 2004;18(11):1798-1803. teolytic cleavage of retinoid X receptor- 2007;109(3):971-979.
9. McKeown MR, Corces MR, Eaton ML, et al. alpha by GSK-3beta. Carcinogenesis. 30. Yadav B, Pemovska T, Szwajda A, et al.
Super-enhancer analysis defines novel epige- 2013;34(6):1208-1215. Quantitative scoring of differential drug sen-
nomic subtypes of non-APL AML Including 20. Fadel L, Reho B, Volko J, et al. Agonist bind- sitivity for individually optimized anticancer
an RARalpha dependency targetable by SY- ing directs dynamic competition among therapies. Sci Rep. 2014;4:5193.
1425, a potent and selective RARalpha ago- nuclear receptors for heterodimerization 31. Bianchi N, Ongaro F, Chiarabelli C, et al.
nist. Cancer Discov. 2017;7(10):1136-1153. with retinoid X receptor. J Biol Chem. Induction of erythroid differentiation of
10. Sakamoto K, Imamura T, Yano M, et al. 2020;295(29):10045-10061. human K562 cells by cisplatin analogs.
Sensitivity of MLL-rearranged AML cells to 21. Devarakonda S, Gupta K, Chalmers MJ, et Biochem Pharmacol. 2000;60(1):31-40.
all-trans retinoic acid is associated with the al. Disorder-to-order transition underlies the 32. Niu H, Fujiwara H, di Martino O, et al.
level of H3K4me2 in the RARalpha promot- structural basis for the assembly of a tran- Endogenous retinoid X receptor ligands in
er region. Blood Cancer J. 2014;4(4):e205. scriptionally active PGC-1alpha/ERRgamma mouse hematopoietic cells. Sci Signal.
11. Niu H, Chacko J, Hadwiger G, Welch JS. complex. Proc Natl Acad Sci U S A. 2017;10(503):eaan1011.
Absence of natural intracellular retinoids in 2011;108(46):18678-18683.

Cell Therapy & Immunotherapy ARTICLE
Human invariant natural killer T cells promote Ferrata Storti Foundation

tolerance by preferential apoptosis induction
of conventional dendritic cells
Hannes Schmid,1* Emmanuelle M. Ribeiro,1* Kathy-Ann Secker,1
Silke Duerr-Stoerzer,1 Hildegard Keppeler,1 Ruoyun Dong,1 Timo Munz,1
Klaus Schulze-Osthoff,2 Stephan Hailfinger,2 Corina Schneidawind1
and Dominik Schneidawind1
1
Department of Medicine II, University Hospital Tübingen, Eberhard Karls University, and
2
Interfaculty Institute of Biochemistry, Eberhard Karls University, Tübingen, Germany Haematologica 2022
*HS and EMR contributed equally as co-first authors.
Volume 107(2):427-436
ABSTRACT
G
raft-versus-host disease (GvHD) is a major cause of morbidity and
mortality after allogeneic hematopoietic cell transplantation. We
recently showed in murine studies and in vitro human models that
adoptively transferred invariant natural killer T (iNKT) cells protect from
GvHD and promote graft-versus-leukemia effects. The cellular mechanisms
underlying GvHD prevention by iNKT cells in humans, however, remain
unknown. In order to study relevant cellular interactions, dendritic cells (DC)
were either generated from monocytes or isolated directly from blood of
healthy donors or GvHD patients and co-cultured in a mixed lymphocyte
reaction (MLR) with T cells obtained from healthy donors or transplantation
bags. Addition of culture-expanded iNKT cells to the MLR-induced DC apop-
tosis in a cell contact-dependent manner, thereby preventing T-cell activation
and proliferation. Annexin V/propidium iodide staining and image stream
assays showed that CD4+CD8–, CD4–CD8+ and double negative iNKT cells
are similarly able to induce DC apoptosis. Further MLR assays revealed that
conventional DC (cDC) but not plasmacytoid DC (pDC) could induce allore-
active T-cell activation and proliferation. Interestingly, cDC were also more
susceptible to apoptosis induced by iNKT cells, which correlates with their
higher CD1d expression, leading to a bias in favor of pDC. Remarkably, these
Correspondence:
results could also be observed in GvHD patients. We propose a new mecha-
nism how ex vivo expanded human iNKT cells prevent alloreactivity of T cells. DOMINIK SCHNEIDAWIND
iNKT cells modulate T-cell responses by selective apoptosis of DC subsets, dominik.schneidawind@med.uni-tuebingen.de
resulting in suppression of T-cell activation and proliferation while enabling
beneficial immune responses through pDC. Received: July 22, 2020.
Pre-published: January 14, 2021.
Introduction
Despite significant advances in the field of allogeneic hematopoietic cell trans- https://doi.org/10.3324/haematol.2020.267583
plantation (HCT), graft-versus-host disease (GvHD) still represents a major compli-
cation after allogeneic HCT, leading to substantial morbidity and mortality.1,2
GvHD is mediated by donor T cells activated through antigen-presenting cells
(APC).3 Dendritic cells (DC) are professional APC that precisely orchestrate adap- Material published in Haematologica is covered by copyright.
tive immune responses and their significant role in GvHD pathophysiology has All rights are reserved to the Ferrata Storti Foundation. Use of
been established previously.4-6 Both donor and host DC present host antigens and conditions:
promote activation and proliferation of alloreactive donor T cells, which conse- https://creativecommons.org/licenses/by-nc/4.0/legalcode.
quently home to GvHD target sites, resulting in tissue destruction and clinical man- Copies of published material are allowed for personal or inter-
ifestations of GvHD.7,8 The ability of DC to elicit or prevent T-cell responses is nal use. Sharing published material for non-commercial pur-
tuned by the concomitant expression of stimulatory or inhibitory molecules as well https://creativecommons.org/licenses/by-nc/4.0/legalcode,
as immunomodulatory cytokines.9 sect. 3. Reproducing and sharing published material for com-
DC also express antigen-presenting molecules such as the major histocompatibil- mercial purposes is not allowed without permission in writing
ity complex-I (MHC-I)-like molecule CD1d that allows for interactions with invari- from the publisher.
ant natural killer T (iNKT) cells. iNKT cells are a small subset of T lymphocytes
characterized by the expression of an invariant T-cell receptor in both humans and

H. Schmid et al.
mice.10 Upon activation through glycolipids, iNKT cells Biolegend, San Diego, CA, USA) according to the manufacturer’s
regulate immune responses by the instant release of instructions and tested in a mixed lymphocyte reaction (MLR).
immunoregulatory cytokines or by direct cell killing.11-13
Several studies have shown the ability of iNKT cells to Mixed lymphocyte reaction
reduce the incidence of GvHD. In murine models, iNKT Major mismatched mo-DC or blood DC were plated together
cells prevent acute and chronic GvHD, while promoting with allogeneic CD3+ T cells at a 1:1 ratio. Culture-expanded
beneficial graft-versus-leukemia (GvL) effects.14-16 In iNKT cells were added to the MLR at different doses, either direct-
humans, clinical studies have demonstrated that high ly or separated from the MLR by 0.4 µm TC-Inserts (Sarstedt,
iNKT-cell numbers are associated with a diminished Nuembrecht, Germany). Cells were analyzed by flow cytometry
occurrence of GvHD.17-19 Moreover, we recently showed for activation markers (CD69 and CD25) and proliferation (CFSE).
that culture-expanded human iNKT cells are able to pre- Alternatively, T cells were incubated with anti-CD3/CD28-coated
vent T-cell activation and proliferation while exerting beads (ThermoFisher Scientific, Waltham, MA, USA) in the pres-
potent anti-leukemic activity.13,20 ence or absence of iNKT cells. For blocking assays, iNKT cells or
Nevertheless, the complex cellular and molecular mech- DC were pre-treated with the respective antibodies or IgG control
anisms of immune tolerance induction through iNKT cells (Online Supplementary Appendix).
remain poorly understood. In this study, we focused on
how culture-expanded human iNKT cells modulate allore- Apoptosis assays
active T-cell responses through DC in healthy volunteers Apoptosis was assessed with an annexin V-FITC/propidium
and GvHD patients. iodide (PI) Staining Kit (BD Bioscience, Franklin Lakes, NJ, USA),
by cell cycle analysis modified according to Nicoletti21 or by image
stream analysis (Online Supplementary Appendix). The percentage
Methods of apoptotic cells was determined by flow cytometry.
Research subjects Cytokine analysis

Human buffy coats from healthy volunteers were obtained Cell culture supernatants from MLR were collected after 4 and
from the Center of Clinical Transfusion Medicine Tuebingen. 24 hours (h), respectively. In order to analyze cytokine production
Samples from hematopoietic cell grafts and peripheral blood bead-based immunoassays were performed according to the man-
mononuclear cells (PBMC) from patients with GvHD were isolat- ufacturer’s instructions. Cytokine release was measured by a
ed after written informed consent had been obtained. Human LEGENDplex human CD8/NK-cell panel (BioLegend). Data were
leukocyte antigen (HLA) typing was performed by the Center of acquired using the Lyric flow cytometer with autosampler (BD
Clinical Transfusion Medicine Tuebingen or the HLA laboratory Biosciences).
of the Department of Medicine II of the University Hospital
Tuebingen. The study was approved by our Institutional Review Statistical analysis
Board to be in accordance with ethical standards and with the Student’s t-test and analysis of variance (ANOVA) were used for
Helsinki Declaration of 1975, as revised in 2013 (IRB approvals statistical analysis. P<0.05 was considered statistically significant.
483/2015BO2 and 137/2017BO2). Data were analyzed with Prism 8 (GraphPad Software, La Jolla,
CA, USA). All experiments were performed at least in technical
Flow cytometry duplicates and repeated independently at least three times using
Antibodies and reagents used for flow cytometric analyses are different iNKT-cell donors.
described in the Online Supplementary Appendix.
Invariant natural killer T-cell expansion and enrichment Results

iNKT cells were expanded from third-party PBMC with some
minor modifications as previously described (Online Supplementary Invariant natural killer T cells inhibit T-cell activation
Appendix).13 Culture-expanded iNKT cells were purified with anti- and proliferation in a cell contact-dependent manner
iNKT MicroBeads (Miltenyi Biotech, Bergisch Gladbach, Human culture-expanded iNKT cells suppress alloreac-
Germany). Alternatively, iNKT cells were stained with DAPI (4',6- tive T-cell responses when T lymphocytes are stimulated
diamidino-2-phenylindole, Merck, Darmstadt, Germany), anti- by MHC-mismatched DC.13 As iNKT cells exhibit potent
CD3, anti-CD4, anti-CD8 antibodies and PBS57-loaded CD1d immunoregulatory properties through a rapid release of
tetramer allowing for enrichment of iNKT cells and their different humoral mediators, we wondered whether this effect
subsets by fluorescence-activated cell sorting (FACS). might be related to the inhibition of T-cell function.22
Therefore, we measured early (CD69 expression) and late
Generation of monocyte-derived dentritic cells activation (CD25 expression) as well as proliferation
and isolation of blood dendritic cells (CFSE dilution) of T cells co-incubated with DC in pres-
Monocyte-derived dentritic cells (Mo-DC) were generated as ence or absence of iNKT cells. iNKT cells were either
described previously.13 Blood DC from healthy donors and added directly to the MLR or separated through a TC-
patients were isolated using Blood Dendritic Cell Isolation Kit II insert (transwell [TW]). Direct addition of iNKT cells at
(Miltenyi Biotech). Where indicated, HLA-DR+ blood DC were different ratios to the MLR reduced T-cell activation and
further sorted either as CD1c+ conventional DC (cDC) or CD303+ proliferation in a dose-dependent manner. However,
plasmacytoid DC (pDC). iNKT cells which were separated by a TW did not prevent
T-cell activation and proliferation (Figure 1A to C; Online
CD3+ T-cell isolation Supplementary Figure S1). We conclude from these experi-
CD3+ T cells were isolated from human PBMC with anti-CD3 ments that iNKT cells mostly rely on direct cell contact to
MicroBeads (Miltenyi Biotech). For proliferation analysis, T cells efficiently suppress alloreactive T-cell responses. In addi-
were marked with CFSE (carboxyfluorescein succinimidyl ester, tion, T-cell activation and proliferation initiated by artifi-

iNKT cells promote tolerance by cDC apoptosis
Figure 1. Culture-expanded invariant natural

A killer T cells inhibit T-cell activation and prolifer-
ation. Representative dot plots and histograms
showing (A) early activated T cells (CD69+, day
1), (B) late activated T cells (CD25+, day 3) and
(C) proliferating T cells (carboxyfluorescein suc-
cinimidyl ester [CFSE], day 7). T-cell activation
and proliferation was measured after incubation
with monocyte-derived dendritic cell (mo-DC) in
the presence or absence of invariant natural
killer T (iNKT) cells. iNKT cells were added to the
culture either directly or separately through a
transwell insert (TW). (D) Representative dot
plots and histograms showing late activated T
cells (CD25+, day 3) and (E) proliferating T cells
(CFSE, day 7) after stimulation with anti-
CD3/CD28-coated beads in the presence or
B absence of iNKT cells. All events were gated on
single cells and living lymphocytes. iNKT cells
were excluded from the analysis by gating on
CD3+ PBS57-loaded CD1d tetramer+ popula-
tions. Histograms show the mean of three inde-
pendent experiments (n=3). Error bars indicate
standard error of the mean. ns: not significant,
*P<0.05, **P<0.01, ***P<0.001,
****P<0.0001. DC: dendritic cells; T: T cells.

H. Schmid et al.
C D
F
Figure 2. Culture-expanded invariant natural killer T cells
induce dendritic cell apoptosis in a dose-dependent
manner. (A) Representative dot plots showing absence of
dendritic cells (DC) (CD11c+HLA-DR+) after co-culture with
invariant natural killer T (iNKT) cells. (B) Representative
dot plots and pooled data of living DC (annexin V-/propid-
ium iodide [PI]-) after 4 hours (h) of co-culture with
increasing numbers of T cells or iNKT cells. Indicated are
the ratios of DC to T or iNKT cells. (C) Representative dot
plots showing DC apoptosis in co-cultures with iNKT cells
after 1, 2, 4, 6 and 8 h of incubation. (D) Histograms
showing increased DNA fragmentation in DC after 4 h and
18 h of co-culture with iNKT cells. The gates on the left of
each plot show the percentage of apoptotic nuclei. (E)
Representative image stream assay and dot plots showing
DC apoptosis induced by iNKT cells after 4 h and 18 h
(annexin V+, green; 7-AAD+, red). (F) Representative dot
plots, relative and absolute numbers of living DC (annexin
V-/PI-) after 4 h of co-culture with different iNKT-cell sub-
sets. Histograms show the mean of three independent
experiments (n=3). Error bars indicate standard error of
the mean. ns: not significant. HLA-DR: human leukocyte
antigen DR-isotype.

C D
Figure 3. Culture-expanded invariant natural killer T cells require cell contact to induce apoptosis through degranulated effector molecules. (A) Representative
image stream assay illustrating direct cellular contact between invariant natural killer T (iNKT) cells (PBS57-loaded CD1d tetramer+, yellow) and dendrtic cells (DC)
(HLA-DR+, pink) and subsequent DC apoptosis induction (annexin V+, green) after 4 hours (h) of co-incubation. (B) Representative dot plots showing DC apoptosis
and pooled data of living DC (annexin V-propidium iodide [PI]-) after co-incubation with iNKT cells either directly or separated by a transwell insert (TW). (C) Percentage
of DC apoptosis inhibition after blocking of the receptors CD1d, FasL, TRAIL, NKG2D and applying the inhibitors zVAD-fmk (N-benzyloxycarbonyl-Val-Ala-Asp(O-Me) flu-
oromethylketone), CMA (concanamycin A) and monensin/brefeldin A. (D) Representative dot plots showing DC apoptosis after co-culture with non-degranulated and
degranulated iNKT-cell supernatant. (E) IFN-γ, granzyme B, perforin and granulysin release by iNKT cells after encountering DC analyzed by bead-based immunoas-
say. Histograms show the mean of three independent experiments (n=3). Error bars show standard error of the mean. ns: not significant; *P<0.05, **P<0.01,
***P<0.001; HLA-DR: human leukocyte antigen DR-isotype.

H. Schmid et al.
cial antigen-presenting cells (aAPC, e.g., Dynabeads) was reduced DC apoptosis significantly indicating that T-cell
only affected when higher numbers of iNKT cells were receptor engagement contributes to efficient lysis. Further,
added to the culture (Figure 1D and E; Online blocking apoptosis via caspase inhibitor zVAD-fmk (N-
Supplementary Figure S2). In particular, proliferation speed benzyloxycarbonyl-Val-Ala-Asp(O-Me) fluoromethylke-
was decreased with a predominance of early daughter tone) or inhibition of the perforin pathway via CMA (con-
generations (Online Supplementary Figure S2). Our findings canamycin A) also diminished iNKT-cell-mediated cell
suggest that the interaction of iNKT cells with DC largely death of DC. Moreover, the inhibition of iNKT-cell
contributes to the control of alloreactive T cells although a degranulation by monensin and brefeldin A was shown to
minor direct impact of iNKT cells on T cells could be impede apoptosis induction most efficiently (Figure 3C;
observed. Online Supplementary Figure S3). By adding supernatant of
DC-triggered degranulated iNKT cells to viable DC, we
Invariant natural killer T cells induce apoptosis of could show that iNKT cells released cytotoxic factors dur-
allogeneic dendritic cells in a dose-dependent manner ing degranulation which further induced apoptosis in DC
We performed flow cytometry to determine the pheno- (Figure 3D). In order to identify these factors, we per-
type of DC challenged with iNKT cells. Notably, DC formed bead-based multiplex assays and thereby revealed
numbers were highly reduced (Figure 2A) and we suspect- the release of interferon-g (IFN-g), granzyme B, perforin
ed induction of apoptosis through iNKT cells. Annexin V and granulysin (Figure 3E).
assays showed that culture-expanded iNKT cells rapidly
induced apoptosis of allogeneic DC, while co-culture of Invariant natural killer T cells induce preferential
DC with conventional allogeneic CD3+ T cells did not, apoptosis of blood conventional dendritic cells in
suggesting that apoptosis induction is not only dose- healthy donors and graft-versus-host disease patients
dependent but also specific to iNKT cells (Figure 2B and Our previous observations are based on ex vivo cultured
C). The Nicoletti assay revealed that DC start to defrag- mo-DC. In order to support our findings, we additionally
ment their DNA after co-culture with iNKT cells, which performed MLR and apoptosis assays using blood DC iso-
represents a further hallmark of apoptosis (Figure 2D). lated from PBMC of healthy donors and GvHD patients
Interestingly, an increase of spontaneous DNA defrag- following allogeneic HCT. Blood DC are mainly com-
mentation could be observed in DC without iNKT cells posed of cDC and pDC with the latter expressing lower
after 18 h which could be explained by the lack of specific levels of CD1d (Online Supplementary Figure S4A). Also,
stimuli. Image stream analysis also confirmed morpholog- human blood DC of healthy volunteers induce activation
ic changes in DC after co-culture with iNKT cells. and proliferation of MHC-mismatched T cells that can be
Whereas DC cultured alone presented a healthy and diminished through the addition of iNKT cells (Figure 4A).
round morphology, DC co-cultured with iNKT cells were Given that iNKT cells interact with DC through CD1d
small, squashed and with a blobbing membrane. Further, and CD1d engagement contributes to efficient lysis of tar-
upregulation of Annexin V and loss of nuclear integrity in get cells, we aimed to determine how human blood pDC
DC co-incubated with iNKT cells could be observed in and cDC are susceptible to iNKT-cell apoptosis induction.
image stream assays, confirming our previous assumption For this purpose, we isolated HLA-DR+ pDC (CD303+)
(Figure 2E). Next, we investigated whether the induction and cDC (CD1c+) by FACS, co-cultured them separately
of apoptosis is specific to certain iNKT-cell subpopula- with iNKT cells for 4 h and stained with annexin V and PI.
tions. Therefore, culture-expanded iNKT cells were sorted We observed preferential apoptosis induction of cDC,
into double negative, CD4+CD8– and CD4–CD8+ subsets while pDC were less affected by the addition of iNKT
and co-cultured separately with allogeneic DC: all iNKT- cells (Figure 4B). Further, we wondered whether preferen-
cell subsets were able to induce apoptosis of DC with tial apoptosis of cDC by iNKT cells would also affect T-
comparable efficiency (Figure 2F). cell alloreactivity using fresh human blood DC as stimula-
tors. We observed that only allogeneic cDC in contrast to
Induction of dendritic cell apoptosis is cell pDC could induce significant T-cell activation and prolif-
contact-dependent and mediated by cytotoxic effector eration (Online Supplementary Figure S4B). Consequently,
molecules co-culture of these distinct blood DC subsets with allo-
In order to further elucidate the cellular and molecular geneic T cells and iNKT cells revealed that iNKT cells
mechanisms responsible for iNKT-cell-induced DC apop- were also able to suppress activation and proliferation of
tosis, we first analyzed image stream data visualizing dou- alloreactive T cells induced by cDC (Figure 4C).
blets consisting of DC and iNKT cells. Image stream Finally, we tested whether our findings also apply to
analysis revealed a direct binding of iNKT cells (PBS57- patients with acute GvHD having received grafts from
loaded CD1d tetramer+) to the surface of allogeneic DC HLA-matched donors. Therefore, we isolated blood DC
(HLA-DR+), which subsequently revealed positive surface from PBMC obtained from patients with clinical manifes-
staining for the apoptosis marker annexin V (Figure 3A). In tations of acute GvHD grade ≥2 prior to induction of sys-
order to test whether this direct cellular interaction is temic treatment with steroids. At the time point of blood
required, iNKT cells and DC were separated by a tran- collection patients had complete donor chimerism in their
swell insert demonstrating that iNKT cells were unable to peripheral blood. cDC from GvHD patients also showed
induce DC apoptosis anymore (Figure 3B). We further per- higher expression levels of CD1d (Online Supplementary
formed blocking experiments of common key molecules Figure S4C) and were more susceptible to iNKT-cell-
to identify critical pathways responsible for apoptosis induced apoptosis than pDC (Figure 4D), similarly as
induction through iNKT cells. It was observed that block- demonstrated in our previous experiments with cells from
ing FasL, TRAIL or NKG2D did not significantly reduce healthy donors. Next, T cells derived from donors prior to
apoptosis of DC exposed to iNKT cells. However, block- transplantation were co-cultured with blood DC from
ing the CD1d and invariant T-cell receptor interaction GvHD patients. Importantly, adding iNKT cells from

A B
Figure 4. Preferential apoptosis induction of blood conven-

D tional dendritic cells by invariant natural killer T cells. (A)
Representative dot plots and pooled data of activated (CD69,
day 1; CD25, day 3) and proliferating (CFSE, day 7) major
histocompatibility complex (MHC)-matched T cells after co-
culture with blood dendritic cells (DC) from healthy volunteers
in presence or absence of invariant natural killer T (iNKT)
cells. (B) Representative dot plots showing increased blood
DC apoptosis (upper row) and the frequency of plasmacytoid
(pDC, CD303+) and conventional DC (cDC, CD1c+) among liv-
ing blood DC (annexin V-/propidium iodide [PI]-, lower row)
after co-culture with iNKT cells. (C) Representative dot plots
showing early and late T-cell activation (CD69, day 1; CD25,
day 3) and proliferating T cells (carboxyfluorescein succin-
E imidyl ester [CFSE], day 7) after co-culture with sorted pDC
and cDC from healthy volunteers. (D) Representative dot plots
showing apoptosis of blood DC from patients with graft-ver-
sus-host disease (GvHD) after co-culture (4 hours) with cul-
ture-expanded third-party donor iNKT cells and the frequency
of cDC and pDC among living DC (annexin V-/PI-). (E)
Representative dot plots showing early and late activation
(CD25, day 3) and proliferation (CFSE, day 7) of
MHC-matched donor T cells after co-culture with blood DC
from GvHD patients in presence or absence of third-party
donor iNKT cells. Histograms show the mean of three inde-
pendent experiments. Error bars indicate standard error of
the mean. *P<0.05.

H. Schmid et al.
third-party donors also inhibited alloreactive T-cell activa- hypothesizing a relevant mechanism for the modulation
tion and proliferation (Figure 4E). of immune responses and GvHD suppression.32,33 We add
significant knowledge by showing that preferential apop-
tosis induction of cDC leads to a relative expansion of
Discussion beneficial pDC. In contrast, we did not find significant
functional differences regarding distinct iNKT-cell subsets.
Allogeneic HCT is an established therapeutic option for iNKT cells are activated upon recognition of glycolipids
the treatment of advanced and high-risk hematologic presented by the MHC-I-like molecule CD1d, which is
malignancies. Efforts to optimize donor selection, tailored highly expressed on DC.11,34 Hence, T-cell receptor-CD1d
preparative conditioning regimes and advanced support- engagement induces cytokine release by iNKT cells,
ive care have significantly contributed to improved out- which confers immunoregulatory properties and the abil-
comes and enabled long-term survival even in aged and ity to orchestrate immune responses of several cell types.
comorbid patient populations. Nevertheless, GvHD and For instance, the release of cytokines such as IFN-g, tumor
relapse still represent the most important reasons for sig- necrosis factor-a (TNF-a), interleukin-2 (IL-2), IL-4, IL-17
nificant morbidity and mortality after allogeneic HCT.1,2 and IL-21 has been noted.13,22 Beyond immunoregulatory
Various strategies have been applied to prevent or treat properties, iNKT cells exert potent direct cytotoxic effects
GvHD such as immunosuppressive medications and using different pathways.13,20 In this context, stimulation
in vivo or ex vivo donor T-cell depletion. However, these via CD95 (Fas)35,36 and TRAIL pathways36,37 has been
approaches are suboptimal since they also inhibit immune demonstrated, resulting in a classical lymphocytotoxic
reconstitution, pathogen control and beneficial GvL response against tumor cells. Moreover, several studies
effects, leading to higher relapse rates.23,34 Therefore, have shown that perforin/granzyme B is involved in
strategies that prevent GvHD while preserving the capac- iNKT-cell tumor cytotoxicity.38-40 Using different blocking
ity of the graft to promote GvL effects are urgently need- reagents and specific antibodies, we could demonstrate
ed. that DC apoptosis induced by iNKT cells relies on degran-
A convincing body of evidence has demonstrated the ulation of perforin, granzyme B and granulysin and par-
potential of iNKT cells as a promising alternative for the tially on the interaction of the invariant T-cell receptor
prevention of GvHD in both mice and humans. Early with CD1d .
murine studies demonstrated that the reinfusion of DC originate from either myeloid or lymphoid
NK1.1+ T cells after transplantation resulted in GvHD sup- hematopoietic stem cell progenitors in the bone mar-
pression.25 In particular, low doses of CD4+ iNKT cells pre- row.41,42 They constitute a heterogeneous cell group of dif-
vented GvHD lethality in mice by promoting the expan- ferent subsets playing distinct roles in regulating immune
sion of Tregs while maintaining GvL effects.14 We also responses.43 DC have been categorized in cDC, pDC and
showed previously that third-party iNKT cells are equipo- mo-DC, considering their lineage and expression of tran-
tent due to the highly conserved invariant TCR of iNKT scription factors such as IFN regulatory factors 8 and 4.44
cells.15,26 Based on these findings, we used iNKT cells from In humans, cDC are potent producers of IL-12 and harbor
third-party donors in our present study. In humans, sever- excellent cross-priming properties. In the context of
al groups have shown that high numbers of iNKT cells GvHD, cDC turned out to be important stimulators of
were associated with a decreased incidence of alloreactive T-cell responses.45 Also, Markley et al. demon-
GvHD.17,18,27 Malard et al. also showed in a study of 80 strated that donor cDC are critical for allo-antigen presen-
patients that high iNKT-cell numbers in the graft correlat- tation and consequently potentiate GvHD.46 Besides, cDC
ed with an increased GvHD-free, relapse-free survival; are most likely responsible for the replenishment of tissue-
however, the frequency of Tregs did not seem to correlate specific DC such as migratory Langerhans cells of the skin
with iNKT-cell numbers.19 Moreover, Cheng et al. ana- after inflammation and therefore might contribute to the
lyzed Treg expansion after infusion of a-galactosylce- occurrence and perpetuation of skin GvHD.47,48 In con-
ramide (a-GalCer), a potent iNKT-cell stimulator, but trast, the functional hallmark of pDC is the release of high
expansion of Tregs could only be observed in a subset of quantities of type I and type III interferon (IFN) in
patients.28 Thus, the role of Treg expansion as mediator of response to viral antigen recognition.49,50 Interestingly, pre-
therapeutically used iNKT cells is not well established in cursor and fully differentiated pDC are associated with an
humans and suggests further mechanisms that contribute improved outcome after allogeneic HCT due to a
to the immunoregulatory properties of iNKT cells. decreased incidence of GvHD and optimized GvL
GvHD can be characterized as a response of donor T effects.51,52 Thus, the modulation of cDC and pDC by
cells to host antigens presented by MHC molecules iNKT cells could represent a useful approach to reduce the
through APC1: first, host APC become activated and pres- incidence of GvHD.
ent allo-antigens to donor T cells, which are stimulated In this study, we focused on human blood DC, which
and expand. Consequently, cellular effectors promote cell are mainly composed of cDC and pDC and can be easily
damage and apoptosis.1,8,29 In this context, several studies obtained from healthy volunteers and GvHD patients.
have emphasized the role of DC as potent APC in the Our results suggest an additional mechanism of how
pathogenesis of GvHD and therefore, they represent an human culture-expanded iNKT cells prevent GvHD: pref-
interesting target for prophylactic and therapeutic strate- erential apoptosis of cDC leads to a relative expansion of
gies against GvHD.30,31 In the present study, we therefore beneficial pDC. This bias results in decreased activation
focused on the cellular and humoral interplay of human and proliferation of alloreactive T cells from healthy vol-
culture-expanded iNKT cells with DC. Thereby, we could unteers and GvHD patients. However, we could also
show that iNKT cells induce DC apoptosis and conse- observe a minor direct impact of iNKT cells on T- cell acti-
quently, impair alloreactive T-cell activation and prolifera- vation and proliferation when higher numbers of iNKT
tion. Liu and Coman reported similar findings previously cells were used. Given the high plasticity and functional

diversity of iNKT cells we assume that several mecha- CS performed the research and analyzed data; EMR and DS
nisms, that are not mutually exclusive, are generally wrote the manuscript. All authors edited the manuscript for con-
involved in tolerance induction: modulation of DC func- tent.
tion, expansion of FoxP3 regulatory T cells, induction of a
Th2 bias of T-helper cells and decreased expansion of Acknowledgments
alloreactive donor T cells. Indeed, it has been observed We would like to thank the Flow Cytometry Core Facility of the
that distinct iNKT-cell subsets are associated with certain University Hospital Tuebingen for their excellent technical sup-
functional properties which might explain different find- port. Furthermore, we thank Stella Autenrieth for sharing her
ings from other groups in humans and mice. Also, culture expertise about dendritic cell biology and Kirsten Lauber for
conditions might affect the function of iNKT cells after many fruitful discussions about apoptosis.
expansion.
In conclusion, we postulate an additional mechanism by Funding
which iNKT cells prevent GvHD in humans, focusing on This study was supported by a Max Eder Research fellowship
their interaction with different DC subsets. iNKT cells of the German Cancer Aid (Deutsche Krebshilfe, 70112548), a
promote selective cDC apoptosis through the release of Junior Research Group Grant of the Interdisciplinary Center for
effector molecules such as perforin and granzyme B in a Clinical Research (IZKF, 2316-0-0) and the Clinician Scientist
cell-contact-dependent manner, which could consequent- Program of the Faculty of Medicine Tuebingen. HS received a
ly prevent GvHD. However, pDC are spared and may still grant from the Ludwig Hiermaier Foundation. CS was funded by
convey beneficial immune responses leading to efficient a Junior Research Group Grant of the Interdisciplinary Center for
GvL effects and pathogen control resulting in improved Clinical Research (IZKF, 2383-0-0), the Clinician Scientist
survival after allogenic HCT. Program of the Faculty of Medicine Tuebingen and the
Wuerttemberg Cancer Award (Wuerttembergischer Krebspreis).
Disclosures The National Institutes of Health Tetramer Core Facility kindly
No conflicts of interest to disclose provided CD1d tetramer reagents.
Contributions Data sharing statement

HS, EMR and DS designed and performed the research and Raw data and detailed protocols of the used methods used can
analyzed data; K-AS SD-S, HK, RD, TM, KS-O, SH and be obtained upon direct request to the corresponding author.
References 10. Lantz O, Bendelac A. An invariant T cell 18. Chaidos A, Patterson S, Szydlo R, et al.
receptor alpha chain is used by a unique Graft invariant natural killer T-cell dose
subset of major histocompatibility complex predicts risk of acute graft-versus-host dis-
1. Ferrara JLM, Levine JE, Reddy P, Holler E. class I-specific CD4+ and CD4-8- T cells in ease in allogeneic hematopoietic stem cell
Graft-versus-host disease. Lancet. mice and humans. J Exp Med. 1994; transplantation. Blood. 2012;119(21):5030-
2009;373(9674):1550-1561. 180(3):1097-1106. 5036.
2. Anasetti C, Logan BR, Lee SJ, et al. 11. Bendelac A, Lantz O, Quimby ME, Yewdell 19. Malard F, Labopin M, Chevallier P, et al.
Peripheral-blood stem cells versus bone JW, Bennink JR, Brutkiewicz RR. CD1 Larger number of invariant natural killer T
marrow from unrelated donors. N Engl J recognition by mouse NK1+ T lympho- cells in PBSC allografts correlates with
Med. 2012;367(16):1487-1496. cytes. Science. 1995;268(5212):863-865. improved GVHD-free and progression-free
3. Shlomchik WD, Couzens MS, Tang CB, et 12. Metelitsa LS, Naidenko OV, Kant A, et al. survival. Blood. 2016;127(14):1828-1835.
al. Prevention of graft versus host disease Human NKT cells mediate antitumor cyto- 20. Jahnke S, Schmid H, Secker KA, et al.
by inactivation of host antigen-presenting toxicity directly by recognizing target cell Invariant NKT cells from donor lympho-
cells. Science. 1999;285(5426):412-415. CD1d with bound ligand or indirectly by cyte infusions (DLI-iNKTs) promote ex
4. Wang X, Li H, Matte-Martone C, et al. producing IL-2 to activate NK cells. J vivo lysis of leukemic blasts in a CD1d-
Mechanisms of antigen presentation to T Immunol. 2001;167(6):3114-3122. dependent manner. Front Immunol. 2019;
cells in murine graft-versus-host disease: 13. Schmid H, Schneidawind C, Jahnke S, et al. 10:1542.
cross-presentation and the appearance of Culture-expanded human invariant natural 21. Riccardi C, Nicoletti I. Analysis of apopto-
cross-presentation. Blood. 2011; killer T cells suppress T-cell alloreactivity sis by propidium iodide staining and flow
118(24):6426-6437. and eradicate leukemia. Front Immunol. cytometry. Nat Protoc. 2006;1(3):1458-
5. Stenger EO, Turnquist HR, Mapara MY, 2018;9:1817. 1461.
Thomson AW. Dendritic cells and regula- 14. Schneidawind D, Pierini A, Alvarez M, et 22. Coquet JM, Chakravarti S, Kyparissoudis
tion of graft-versus-host disease and graft- al. CD4+ invariant natural killer T cells pro- K, et al. Diverse cytokine production by
versus-leukemia activity. Blood. tect from murine GVHD lethality through NKT cell subsets and identification of an
2012;119(22):5088-5103. expansion of donor CD4+CD25+FoxP3+ IL-17-producing CD4-NK1.1- NKT cell
6. Lau J, Sartor M, Bradstock KF, Vuckovic S, regulatory T cells. Blood. 2014; population. Proc Natl Acad Sci U S A.
Munster DJ, Hart DN. Activated circulating 124(22):3320-3328. 2008;105(32):11287-11292.
dendritic cells after hematopoietic stem cell 15. Schneidawind D, Baker J, Pierini A, et al. 23. Martinez C, Urbano-Ispizua A. Graft-ver-
transplantation predict acute graft-versus- Third-party CD4+ invariant natural killer T sus-host disease therapy: something else
host disease. Transplantation. 2007; cells protect from murine GVHD lethality. beyond glucocorticoids? Haematologica.
83(7):839-846. Blood. 2015;125(22):3491-3500. 2011;96(9):1249-1251.
7. Matte CC, Liu J, Cormier J, et al. Donor 16. Du J, Paz K, Thangavelu G, et al. Invariant 24. Horowitz MM, Gale RP, Sondel PM, et al.
APCs are required for maximal GVHD but natural killer T cells ameliorate murine Graft-versus-leukemia reactions after bone
not for GVL. Nat Med. 2004;10(9):987-992. chronic GVHD by expanding donor regula- marrow transplantation. Blood. 1990;
8. Ferrara JLM, Levy R, Chao NJ. tory T cells. Blood. 2017;129(23):3121- 75(3):555-562.
Pathophysiologic mechanisms of acute 3125. 25. Zeng D, Lewis D, Dejbakhsh-Jones S, et al.
graft-vs.-host disease. Biol Blood Marrow 17. Rubio MT, Moreira-Teixeira L, Bachy E, et Bone marrow NK1.1(-) and NK1.1(+) T
Transplant. 1999;5(6):347-356. al. Early posttransplantation donor-derived cells reciprocally regulate acute graft versus
9. Bourque J, Hawiger D. invariant natural killer T-cell recovery pre- host disease. J Exp Med. 1999;189(7):1073-
Immunomodulatory bonds of the partner- dicts the occurrence of acute graft-versus- 1081.
ship between dendritic cells and T cells. host disease and overall survival. Blood. 26. Brossay L, Kronenberg M. Highly con-
Crit Rev Immunol. 2018;38(5):379-401. 2012;120(10):2144-2154. served antigen-presenting function of

H. Schmid et al.
CD1d molecules. Immunogenetics. 1999; CD95/CD178-dependent and is correlated 45. Koyama M, Hashimoto D, Aoyama K, et
50(3-4):146-151. with antigenic potency. J Immunol. 2010; al. Plasmacytoid dendritic cells prime
27. Haraguchi K, Takahashi T, Hiruma K, et al. 185(5):2721-2729. alloreactive T cells to mediate graft-versus-
Recovery of Valpha24+ NKT cells after 36. Mattarollo SR, Kenna T, Nieda M, Nicol AJ. host disease as antigen-presenting cells.
hematopoietic stem cell transplantation. Chemotherapy pretreatment sensitizes Blood. 2009;113(9):2088-2095.
Bone Marrow Transplant. 2004;34(7):595- solid tumor-derived cell lines to V alpha 46. Markey KA, Banovic T, Kuns RD, et al.
602. 24+ NKT cell-mediated cytotoxicity. Int J Conventional dendritic cells are the critical
28. Chen YB, Efebera YA, Johnston L, et al. Cancer. 2006;119(7):1630-1637. donor APC presenting alloantigen after
Increased Foxp3(+)Helios(+) regulatory T 37. Nieda M, Nicol A, Koezuka Y, et al. TRAIL experimental bone marrow transplanta-
cells and decreased acute graft-versus-host expression by activated human CD4(+)V tion. Blood. 2009;113(22):5644-5649.
disease after allogeneic bone marrow trans- alpha 24NKT cells induces in vitro and in 47. Martinez-Cingolani C, Grandclaudon M,
plantation in patients receiving Sirolimus vivo apoptosis of human acute myeloid Jeanmougin M, Jouve M, Zollinger R,
and RGI-2001, an activator of invariant nat- leukemia cells. Blood. 2001;97(7):2067-2074. Soumelis V. Human blood BDCA-1 den-
ural killer T cells. Biol Blood Marrow 38. Kawano T, Cui J, Koezuka Y, et al. Natural dritic cells differentiate into Langerhans-
Transplant. 2017;23(4):625-634. killer-like nonspecific tumor cell lysis medi- like cells with thymic stromal lymphopoi-
29. Reddy P. Pathophysiology of acute graft- ated by specific ligand-activated Valpha14 etin and TGF-beta. Blood. 2014; 124(15):
versus-host disease. Hematol Oncol. 2003; NKT cells. Proc Natl Acad Sci U S A. 2411-2420.
21(4):149-161. 1998;95(10):5690-5693. 48. Ito T, Inaba M, Inaba K, et al. A
30. Duffner UA, Maeda Y, Cooke KR, et al. 39. Nicol A, Nieda M, Koezuka Y, et al. Human CD1a(+)/CD11c(+) subset of human blood
Host dendritic cells alone are sufficient to invariant valpha24+ natural killer T cells dendritic cells is a direct precursor of
initiate acute graft-versus-host disease. J activated by alpha-galactosylceramide Langerhans cells. J Immunol. 1999;163(3):
Immunol. 2004;172(12):7393-7398. (KRN7000) have cytotoxic anti-tumour 1409-1419.
31. Zhang Y, Louboutin JP, Zhu J, Rivera AJ, activity through mechanisms distinct from 49. Siegal FP, Kadowaki N, Shodell M, et al.
Emerson SG. Preterminal host dendritic T cells and natural killer cells. Immunology. The nature of the principal type 1 interfer-
cells in irradiated mice prime CD8+ T cell- 2000;99(2):229-234. on-producing cells in human blood.
mediated acute graft-versus-host disease. J 40. Beilke JN, Kuhl NR, Van Kaer L, Gill RG. Science. 1999;284(5421):1835-1837.
Clin Invest. 2002;109(10):1335-1344. NK cells promote islet allograft tolerance 50. Cella M, Jarrossay D, Facchetti F, et al.
32. Coman T, Rossignol J, D'Aveni M, et al. via a perforin-dependent mechanism. Nat Plasmacytoid monocytes migrate to
Human CD4- invariant NKT lymphocytes Med. 2005;11(10):1059-1065. inflamed lymph nodes and produce large
regulate graft versus host disease. 41. Watowich SS, Liu YJ. Mechanisms regulat- amounts of type I interferon. Nat Med.
Oncoimmunology. 2018;7(11):e1470735. ing dendritic cell specification and develop- 1999;5(8):919-923.
33. Liu TY, Uemura Y, Suzuki M, et al. Distinct ment. Immunol Rev. 2010;238(1):76-92. 51. Lu Y, Giver CR, Sharma A, et al. IFN-
subsets of human invariant NKT cells dif- 42. Manz MG, Traver D, Miyamoto T, gamma and indoleamine 2,3-dioxygenase
ferentially regulate T helper responses via Weissman IL, Akashi K. Dendritic cell signaling between donor dendritic cells and
dendritic cells. Eur J Immunol. 2008; potentials of early lymphoid and myeloid T cells regulates graft versus host and graft
38(4):1012-1023. progenitors. Blood. 2001;97(11):3333-3341. versus leukemia activity. Blood. 2012;
34. Exley M, Garcia J, Balk SP, Porcelli S. 43. Yu H, Tian Y, Wang Y, Mineishi S, Zhang Y. 119(4):1075-1085.
Requirements for CD1d recognition by Dendritic cell regulation of graft-vs.-host 52. Hassan M, Ulezko Antonova A, Li JM, et al.
human invariant Valpha24+ CD4-CD8- T disease: immunostimulation and tolerance. Flt3L Treatment of bone marrow donors
cells. J Exp Med. 1997;186(1):109-120. Front Immunol. 2019;10:93. increases graft plasmacytoid dendritic cell
35. Wingender G, Krebs P, Beutler B, 44. Collin M, Bigley V. Human dendritic cell content and improves allogeneic transplan-
Kronenberg M. Antigen-specific cytotoxici- subsets: an update. Immunology. tation outcomes. Biol Blood Marrow
ty by invariant NKT cells in vivo is 2018;154(1):3-20. Transplant. 2019;25(6):1075-1084.

Cell Therapy & Immunotherapy ARTICLE
CD38 knockout natural killer cells expressing Ferrata Storti Foundation

an affinity optimized CD38 chimeric antigen
receptor successfully target acute myeloid
leukemia with reduced effector cell fratricide
Mark Gurney,1 Arwen Stikvoort,2 Emma Nolan,1 Lucy Kirkham-McCarthy,1
Stanislav Khoruzhenko,3 Rama Shivakumar,3 Sonja Zweegman,2 Niels W.C.J.
van de Donk,2 Tuna Mutis,2 Eva Szegezdi,1 Subhashis Sarkar1# and Michael
O’Dwyer1#
Haematologica 2022
National University of Ireland Galway, Galway, Ireland; 2Cancer Center Amsterdam, VU
1
Volume 107(2):437-445
University Medical Center, Amsterdam, the Netherlands and 3MaxCyte, Inc.,
Gaithersburg, MD, USA
#
SS and MOD contributed equally as co-senior authors.
ABSTRACT
T
here is a strong biological rationale for the augmentation of allo-
geneic natural killer (NK) cell therapies with a chimeric antigen
receptor (CAR) to enhance acute myeloid leukemia (AML) target-
ing. CD38 is an established immunotherapeutic target in multiple
myeloma and under investigation as a target antigen in AML. CD38
expression on NK cells and its further induction during ex vivo NK cell
expansion represent barriers to the development of a CD38 CAR-NK cell
therapy. We set out to develop a CD38 CAR-NK cell therapy for AML,
first by using an NK cell line which has low baseline CD38 expression
and subsequently NK cells expanded from healthy donors. To overcome
anticipated fratricide due to NK cell CD38 expression when using pri-
mary expanded NK cells, we applied CRISPR/Cas9 genome editing to
disrupt the CD38 gene during expansion, achieving a mean knockdown
efficiency of 84%. The resulting CD38 knockdown expanded NK cells,
after expression of an affinity optimized CD38 CAR, showed reduced
NK-cell fratricide and an enhanced ability to target primary AML blasts.
Furthermore, the cytotoxic potential of CD38 CAR-NK cells was aug-
mented by pretreatment of the AML cells with all-trans retinoic acid
Correspondence:
which drove enhanced CD38 expression, offering a rational combination MICHAEL O’DWYER
therapy. These findings support the further investigation of CD38 michael.odwyer@nuigalway.ie
knockdown - CD38 CAR-NK cells as a viable immunotherapeutic
approach to the treatment of AML. Received: September 11, 2020.
Pre-published: December 30, 2020.
Introduction
Acute myeloid leukemia (AML) is the most common acute leukemia in adults, https://doi.org/10.3324/haematol.2020.271908
accounting for approximately 2% of all cancer deaths.1 Curative treatment
approaches remain chemotherapy-based, with allogeneic stem cell transplant con-
solidation for selected patients. The introduction of molecularly targeted thera-
pies has provided important incremental improvements for specific AML sub- Material published in Haematologica is covered by copyright.
types.2-4 Relapsed disease, mediated by the persistence of chemotherapy-resistant All rights are reserved to the Ferrata Storti Foundation. Use of
leukemic stem cells (LSC) is particularly difficult to treat, and accounts for much conditions:
of the mortality burden associated with AML. For many older patients, treatment https://creativecommons.org/licenses/by-nc/4.0/legalcode.
options that are both tolerable and efficacious do not yet exist. Anti-CD19 Copies of published material are allowed for personal or inter-
chimeric antigen receptor (CAR) T-cell therapies have provided a ground-breaking nal use. Sharing published material for non-commercial pur-
approach to cancer immunotherapy in B-cell acute lymphoblastic leukemia and B- https://creativecommons.org/licenses/by-nc/4.0/legalcode,
cell non-Hodgkin lymphomas.5,6 While there is considerable interest in applying sect. 3. Reproducing and sharing published material for com-
the principle of CAR technology in other diseases, progress in AML has been lim- mercial purposes is not allowed without permission in writing
ited to date by the absence of an ideal antigenic target, concerns about ‘on-target from the publisher.
off-tumor’ toxicity including that to normal hematopoietic stem cells, and blast
cell heterogeneity which exists both within and between patients.7-9.

M. Gurney et al.
The multifunctional cell surface glycoprotein CD38, a obtained from the American Type Culture Collection and their
breakthrough immunotherapeutic target in multiple identities confirmed by short tandem repeat profiling (Eurofins
myeloma, is also considered a potential target antigen in Genomics™). CD38 CAR and mock KHYG-1 cells were gener-
AML. In contrast to the uniformly high CD38 expression ated by retroviral transduction with genomic integration con-
on malignant plasma cells, blast cell CD38 expression is firmed by the inclusion of DsRed fluorescent protein. The
heterogeneous although frequently exceeds that of nor- development of the second-generation CD28-CD3ζ, opti-
mal cell populations.10 The CD38 monoclonal antibody mized-affinity CD38 CAR was reported previously.13 Primary
daratumumab has been investigated in AML and has NK cells were isolated from healthy donor peripheral blood
shown promising pre-clinical activity.10 CD38 CAR-T mononuclear cells after Ficoll-Paque density gradient centrifu-
cells have been evaluated mainly for their activity in mul- gation and negative immunomagnetic selection (NK Isolation
tiple myeloma and cytotoxicity against primary AML Kit, Miltenyi Biotec™). NK cells were expanded in NK MACS
samples has also been confirmed.11 However, there medium (Miltenyi Biotec™) containing NK MACS supplement,
remains concern about a potent myelosuppressive effect 5% heat-inactivated human AB serum and 100 U/mL. inter-
with a constituently expressed high-affinity anti-CD38 leukin-2 (PeproTech™). Cultures were pre-treated for 48 h with
CAR due to CD38 expression on both mature myeloid ATRA (Sigma-Aldrich™) or dimethyl sulfoxide, in the relevant
cells and their precursors.11,12 To circumvent this problem, experiments.
an affinity-optimized CD38 CAR has been developed to
minimize the targeting of positive, but low-expressing CRISPR/Cas9 gene editing
normal cell populations.13 Five days after isolation, 5x105 NK cells were electroporated
There is a strong biological rationale for natural killer with sgRNA-Cas9 complexes targeting multiple sites within the
(NK) cell-based approaches to adoptive cell transfer CD38 gene (Gene Knockout Kit V2, Synthego™) or control
immunotherapy for AML. NK cells confer a component electroporated (MaxCyte™ GT flow transfection system).
of the graft-versus-leukemia effect of allogeneic stem cell CD38-edited and control electroporated cells were expanded at
transplant and infusions of purified alloreactive NK cells a target density of 1x106 cells/mL. On day 13-15 of expansion,
have proven therapeutic potential.14-16 CAR-NK cell thera- CD38 expression was assessed by flow cytometry. Knockdown
pies are emerging as a complementary approach to CAR- efficiency was calculated as (% CD38-positive cells [mock elec-
T cells, with potential advantages including allogeneic cell troporated] - % CD38-positive cells [CRISPR/Cas9 edited]).
sources and innate antigen independent anti-leukemic
activity. An early clinical report of a cord-blood derived CD38 chimeric antigen receptor mRNA
CD19 CAR-NK cell therapy has shown promising safety electroporation
and efficacy in B-cell malignancies.17 We set out to devel- CD38 CAR mRNA was synthesized (Trilink
op and evaluate an affinity optimized CD38 CAR-NK cell Biotechnologies™) and CD38 CAR expression in primary
therapy for AML. We first used the NK cell line KHYG-1, CD38 KD and control eNK cells was achieved by electropora-
which has naturally low levels of CD38 expression. tion (100 mg/mL mRNA, Maxcyte™ GT Flow Transfection
While allogeneic expanded NK (eNK) cell approaches are System). CAR expression was confirmed by flow cytometry
more suited to clinical translation, ex vivo NK cell expan- using anti-IgG H+L specific goat anti-human antibody (Jackson-
sion has been shown to lead to upregulation of CD38, Immuno research™) and biotinylated protein L stain (ACRO
which we also encountered using a feeder-free, inter- Biosystems™).
leukin-2-based expansion protocol.18 To reduce the antic-
ipated NK cell fratricide that would occur using eNK cells, Cytotoxicity assays
we applied CRISPR/Cas9 to disrupt the CD38 gene dur- Co-culture experiments involved 10,000 target cells (cell
ing NK cell expansion, creating fratricide-resistant NK lines), or 20,000-50,000 bone marrow mononuclear cells from
cells prior to CD38 CAR expression. Both KHYG-1 and AML patients’ samples. NK cell numbers were determined by
CD38 knockdown (KD) eNK approaches lead to efficient the desired effector to target (E:T) cell ratio. After co-culture for
targeting of AML blasts upon CD38 CAR expression, 18-24 h, target cell lines or bone marrow mononuclear cells
with the degree of cytotoxicity correlating with CD38 were identified by flow cytometry, using a cell-tracking dye:
expression. Finally, we confirm a rational combination Tag-IT BV™ proliferation and cell tracking dye (Biolegend™) or
approach utilizing all-trans retinoic acid (ATRA) to VioletTraceTM (Thermo Fisher). Primary blast cell populations
enhance CD38 expression on the AML cells. Collectively, were identified as CD45int/SSClow (CD45 APC), supported by
our data support the potential of CD38 as a therapeutic additional markers chosen based on clinical immunophenotyp-
target in AML and help to define a CD38 CAR-NK cell ing data. Cell death was determined using propidium iodide (PI)
approach suited to clinical development. or LIVE/DEAD Fixable Near-IR (Life Technologies L10119)
staining and reported as ‘% specific (blast) cytotoxicity’ ([sam-
ple cytotoxicity – background cytotoxicity]/[100 – background
Methods cytotoxicity] x 100%) or ‘% blast cell cytotoxicity’ as indicated.
Ethical statement Statistical analysis

Healthy donor blood and AML patients’ bone marrow sam- GraphPad Prism 8 software (San Diego, CA, USA) was used
ples were collected with written informed consent and approval for statistical analysis. Comparisons were conducted using mul-
from the institutional review boards at each institution (ref: tiple two-sided t-tests for cytotoxicity assays at each E:T ratio
CA2219). Cryopreserved samples were obtained from the or one-way analysis of variance for cell expression data with
biobank of Blood Cancer Network Ireland. statistical significance indicated by asterisks (*P<0.05, **P<0.01,
***P<0.001 and ****P<0.0001). Flow cytometry data were
Cells and reagents acquired on a BD FACS Canto II and analyzed using Flow Jo
The cell lines THP-1, KG1a, U937 and KHYG-1 were V10 software and Infinicyt (Cytognos™).

CD38 CAR-NK cells targeting AML
Results cytotoxic potential, expressing a high concentration of

active perforin and signaling kinases.19 The low CD38
CD38 chimeric antigen receptor expression enhances expression of KHYG-1 is in contrast to that of NK-92
KHYG-1 acute myeloid leukemia targeting cells, another NK cell line which has been investigated
To assess the feasibility of targeting CD38 with a CAR- clinically as an adoptive cell therapy, but is strongly
NK cell approach in AML, we first defined the CD38 CD38-positive.18
expression profile of AML cell lines. We classified THP-1 CD38 CAR-KHYG-1 cells were generated by retroviral
and U937 as CD38-positive, and KG1a as CD38-negative transduction using an ‘affinity-optimized’, second-gener-
for further experiments (Figure 1A). We also confirmed a ation anti-CD38 CAR (CD3ζ-CD28).13 CD38 CAR-
low level of CD38 expression on the KHYG-1 cell line, KHYG-1 cells displayed similar characteristics and kinet-
previously shown to be a NK cell line with significant ics in cell culture as those of mock-transduced KHYG-1
C D
Figure 1. Affinity-optimized CD38 CAR-KHYG-1 cell cytotoxicity against acute myeloid leukemia cell lines. (A) Histograms depict CD38 expression of the KHYG-1 NK cell
line and acute myeloid leukemia (AML) cell lines THP-1 (CD38-positive), U937 (CD38-positive) and KG1a (CD38-negative) with mean fluorescence intensity (MFI) indicated.
Bars represent relative CD38 expression of KHYG-1 and AML cell lines (n=3 individual repetitions). Mean values for AML cell lines compared by an unpaired t-test. (B) Bar
chart depicting the specific cytotoxicity of mock-transduced and CD38 CAR-KHYG-1 cells at varying effector to target (E:T) ratios against the CD38-positive cell lines THP-
1 and U937. Comparisons of four independent experiments made by an unpaired t-test at each effector to target (E:T) ratio. (C) Representative histograms depicting CD38
upregulation and MFI values in KG1a cells after 48 h of treatment with all-trans retinoic acid (ATRA) at 10 nM and 20 nM concentrations, compared to dimethylsulfoxide
(DMSO) control treated KG1a cells. The bar chart summarize data from four independent experiments with comparisons by one-way analysis of variance. (D) Bar chart
depicting specific cytotoxicity of mock-transduced and CD38 CAR-KHYG-1 cells in co-culture with 48 h, 10 nM ATRA-pretreated KG1a cells at varying E:T ratios (summary
of 4 experiments). Error bars indicate standard error of mean (SEM). Statistical significance is defined as *P≤0.05, **P≤0.01, ***P≤0.001, ****P≤0.0001.

M. Gurney et al.
cells. To assess whether CD38 CAR targeting has an addi- To confirm that CD38 KD eNK cells showed greater
tional cytotoxic effect, we tested CD38 CAR-KHYG-1 resistance to fratricide than wild-type eNK cells, we intro-
cells and mock-transduced KHYG-1 cells against CD38- duced mRNA coding for an affinity-optimized CD38
positive AML cell lines (Figure 1B). CD38 CAR-KHYG-1 CAR. CAR expression was confirmed by complementary
cells demonstrated greater cytotoxic effects against staining techniques – an anti-human IgG with light chain
CD38-positive cell lines relative to mock-transduced specificity, and biotinylated protein L, with control (back-
KHYG-1 cells at all E:T ratios tested, with relatively ground) and CAR staining depicted in Figure 3C. CD38
greater increases seen against the strongly CD38-positive KD eNK cells displayed significantly less cell death than
THP-1 cells (specific cytotoxicity 58% vs. 28% for THP- wild-type eNK cells, measured 18 h after CD38 CAR
1, 10:1 E:T ratio; P<0.0001). mRNA electroporation in the absence of target cells (18%
ATRA has been shown to upregulate CD38 expression vs. 37%, P=0.002) (Figure 3D), confirming a greater resist-
across all AML subtypes, mediated by a retinoic acid ance to fratricide. Furthermore, the biphasic CD38
response element in the first intron of the CD38 gene.20 expression pattern (representing the small residual CD38-
Pretreatment with ATRA at 10 nM for 48 h led to marked positive NK cell population after CRISPR/Cas9 gene edit-
induction of CD38 expression on KG1a cells, which do ing) was lost in the CD38 KD population after CD38
not express detectable levels of CD38 in resting condi- CAR mRNA transfection, but not after non-specific
tions (Figure 1C). CD38 CAR-KHYG-1 cells were cyto- (CD16) mRNA electroporation (Figure 3E). This empha-
toxic to ATRA-pretreated KG1a cells, while mock-trans- sized the tendency of the CD38 CAR-NK cells to target
duced KHYG-1 cells showed little cytotoxicity despite CD38-positive eNK cells despite affinity-optimization of
ATRA pretreatment (Figure 1D). the CD38 CAR binding domain.
To better mimic the CD38 expression profile encoun-
tered in AML, we tested the efficacy of CD38 CAR- CD38 knockdown - CD38 chimeric antigen receptor-NK
KHYG-1 cells against primary bone marrow mononuclear cells efficiently target primary acute myeloid leukemia
cells from AML patients (Figure 2A, B). CD38 CAR- blasts
KHYG-1 cells displayed greater specific cytotoxicity To confirm that CD38 CAR expression in CD38 KD
against AML blasts relative to mock-transduced KHYG-1 eNK cells enhances the activity of alloreactive NK cells
cells across a range of blast cell CD38 expression, with against AML, CD38 KD eNK cells were electroporated
the degree of specific cytotoxicity correlating with blast with CD38 CAR mRNA or mock-electroporated prior to
cell CD38 expression (Figure 2C). co-culture with bone marrow mononuclear cells from
AML patients with a variety of molecular AML subtypes
CRISPR/Cas9 gene editing of CD38 in primary NK (Online Supplementary Table S1). CD38 KD - CD38 CAR-
cells reduces NK cell fratricide upon CD38 chimeric NK cells showed enhanced cytotoxicity relative to mock-
antigen receptor expression electroporated CD38 KD cells, with the effect being most
While alloreactive NK cell approaches have shown prominent at the highest E:T ratios tested (Figure 4A, B).
some success in treating AML, we hypothesized that Enhanced cytotoxicity was observed for all AML patients
increased expression of CD38 during ex vivo NK cell and cytotoxicity at the 5:1 E:T ratio correlated with blast
expansion could be sufficient to trigger effector cell fratri- cell CD38 expression (R2=0.81) (Figure 4C).
cide after expression of a CD38 CAR, despite affinity We investigated the potential of ATRA pretreatment as
optimization. Indeed, we observed a consistent, mean 4- a means of modulating CD38 expression and potentiating
fold increase in CD38 expression during feeder-free the effects of CD38 CAR targeting using CD38 KD -
expansion of NK cells in interlueukin-2-containing media. CD38 CAR-NK cells. ATRA pretreatment induced a mean
Increases in CD38 from baseline (mean fluorescence 5-fold upregulation of surface CD38 expression in blast
intensity [MFI] 11,903) were detectable by day 5 (MFI cells (Figure 4D). The increased CD38 expression was
40,948) and persisted to at least day 13 (MFI 38,600) associated with greater sensitivity to CD38 KD - CD38
(Figure 3A). Extrapolating from our previous work on CAR-NK cells compared to dimethylsulfoxide-treated
THP-1 cells (MFI 31,866), we concluded that this degree bone marrow mononuclear cell samples tested at the 2:1
of CD38 expression would lead to a fratricidal effect and 5:1 E:T ratios (Figure 4E).
upon CD38 CAR expression thus limiting the cytotoxic
capacity of ex vivo-expanded CD38 CAR-NK cells.
We, therefore, set out to use CRISPR/Cas9 gene edit- Discussion
ing technology to disrupt the CD38 gene in primary NK
cells. We used a multi-sgRNA format, introducing We set out to augment the potential of NK cell adoptive
sgRNA-Cas9 complexes using a high-efficiency, electro- transfer strategies in AML through expression of an affin-
poration-based approach on a platform scalable to Good ity-optimized CD38 CAR. We demonstrated two poten-
Manufacturing Practice (GMP) grade development. tial approaches to CD38 CAR-NK cell therapy in this set-
CD38 KD and mock-electroporated cells were further ting. We confirmed that CD38 KD eNK cells show
expanded for use in functional assays. A consistent KD reduced fratricide after CD38 CAR expression, allowing
effect was achieved across all NK cell donors (mean 84%; effective targeting of primary AML blasts. As an alterna-
range, 75-92%) (Figure 3B). CD38 KD was detectable 48 tive approach we modified the NK cell line KHYG-1 to
h after CRISPR/Cas9 gene editing, peaked by day 3-7 express a CD38 CAR, successfully targeting AML cell
after electroporation and was stable across the duration lines and primary samples. Both approaches could be
of expansion suggesting minimal differences in the enhanced by induction of CD38 expression using ATRA.
growth potential of CD38 KD and mock-electroporated We chose a NK cell line with naturally low CD38 expres-
NK cells in this expansion system (Online Supplementary sion to ensure viability after introducing a CD38 CAR.
Figure S1). KHYG-1 cells have previously been shown to maintain

cytotoxicity after irradiation and could be applied clinical- The expression pattern of CD38 in AML, in which there
ly in a similar manner to the NK-92 cell line.21 However, is often overlap with normal cell populations including
irradiation limits the potential for in vivo expansion and myeloid and monocytic populations, raises concerns
persistence - important variables in determining the clini- about considerable ‘on-target, off-tumor’ toxicity when a
cal efficacy of cellular therapies. This requirement for potent effector cell is directed toward CD38. High-affini-
irradiation may be avoided by using donor-derived, eNK ty CD38 CAR strategies may maximize the proportion of
cells, although this approach is further complicated by patients for whom a CD38-directed therapy is likely to
robust CD38 upregulation encountered during ex vivo have activity, at the expense of considerable myelosup-
expansion. Our CRISPR/Cas9 CD38 KD eNK cells reduce pressive effects. It is important to consider that not all off-
effector cell fratricide, representing an approach that tumor effects are undesirable: in the case of CD38, elimi-
could be explored clinically. nation of CD38-positive immunoregulatory cell subsets
CD38 was a breakthrough immunotherapeutic target may lead to a beneficial therapeutic effect.23,24 The effica-
in multiple myeloma. While there is greater variability in cy of lower-affinity CD38 CAR strategies is likely to be
CD38 expression in AML, CD38 is a potential target anti- limited to cases with strong expression or pharmacologi-
gen in this disease. Daratumumab was shown to be cal upregulation of CD38. Herein we investigated an
active in an in vivo model of AML, while isatuximab has approach to CD38 CAR targeting in AML which aims to
recently been examined in a large-scale, in vitro study.10,22 strike the balance of efficacy, applicability, and off-tumor
Figure 2. CD38 CAR-KHYG-1 activity against primary acute myeloid leukemia samples. (A) Histograms depict unstained controls (blue) and anti-CD38 stained blast
cells (red), from a range of acute myeloid leukemia (AML) patients chosen to represent a spectrum of CD38 expression. Relative mean fluorescence intensity (MFI)
figures for stained samples are reported. (B) Graphs represent specific blast cytotoxicity after co-culture assays with CD38 CAR transduced KHYG-1 (blue) and mock-
transduced KHYG-1 (black) at specified effector to target (E:T) ratios for each corresponding patient’s sample in Figure 2A. (C) The correlation plot and linear regres-
sion line depicts specific blast cell cytotoxicity at the E:T ratio of 3:1, versus CD38 expression (relative MFI) of primary AML samples from all co-culture experiments
carried out in 2A, (n=8 experiments).

M. Gurney et al.
effects. The established anti-leukemic activity of alloreac- optimized-affinity CD38 CAR variant.11
tive NK cells in AML provides a rationale for the develop- While many target antigens are being considered in
ment of CAR-NK cell approaches.14,15 Alloreactive NK AML, CD38 is also unique in the availability of a licensed
cells can be expected to retain their innate anti-leukemic and well-tolerated oral agent capable of modulating target
activity, with enhancement against CD38-positive cells antigen expression, ATRA.20 Furthermore, it has been
conferred by an anti-CD38 CAR. Toxicity against normal shown that malignant blast cells are particularly sensitive
cell populations can be minimized through the use of an to the CD38-inducing effect of ATRA, acting directly
A B
D E
Figure 3. CRISPR/Cas9 gene editing of CD38 in primary expanded natural killer cells to reduce natural killer cell fratricide upon CD38 CAR expression. (A) CD38
expression in freshly isolated (day 0), and expanded natural killer (eNK) cells at day 5 and day 13, as measured by flow cytometry and presented as a representative
histogram and summary bar chart of three unique expansions. (B) Residual CD38 expression measured on day 8-10 after CRISPR/Cas9 gene editing of CD38. Dot
plots from representative donor for mock-electroporated and CD38 knockdown (KD) conditions. Bar chart represents summary data for four donors. (C) Confirmatory
CAR staining performed 18 h after CD38 CAR mRNA electroporation. Pseudo-colored plots depict results from one representative experiment. (D) NK cell death after
18 h of culture after CD38 CAR mRNA electroporation comparing CD38 KD and control eNK cells across three individual experiments and NK cell donors. Comparison
of mean cell death by an unpaired t-test. (E) Histogram depicting a representative residual CD38 expression profile of viable CD38 KD eNK cells demonstrating loss
of residual CD38-positive eNK cell population after CD38 CAR electroporation but not after ‘dummy’ mRNA (CD16) electroporation. EP: electroporated; H + L: heavy
and light chain specific. Error bars indicate standard error of mean (SEM). Statistical significance is defined as *P≤0.05, **P≤0.01.

through a retinoic acid response element within the careful planning. Evidence suggests that NK cell exposure
CD38 gene.25 The vitamin D receptor agonist inecalcitol to ATRA may have a net inhibitory effect on NK cell
represents another investigational approach to CD38 function, suggesting that the preferred approach may be
modulation which could enhance the efficacy of CD38- ATRA prior to adoptive cell transfer.27
directed therapies in AML through a similar principle.26 CD38 targeted therapies are complicated by NK cell
The timing of a clinically applied combination therapy CD38 expression, observed clinically with the NK cell-
using ATRA and a CD38 CAR-NK cell would require depleting effects of daratumumab seen during the treat-
B C
D E
Figure 4. CD38 KD - CD38 CAR-NK cells show enhanced cytotoxicity against primary acute myeloid leukemia samples. (A) CD38 knockdown (KD) expanded natural
killer (NK) cells were electroporated with CD38 CAR mRNA or ‘dummy’ mRNA (CD16) prior to co-culture with bone marrow mononuclear cells from acute myeloid
leukemia (AML) patients. Blast cell cytotoxicity was measured by percentage of propidium iodide (PI)-positive cells (representative dot-plots are shown). (B) Summary
data of co-culture assays as described in (A), for seven AML patients compared using unpaired t-tests for each effector to target (E:T) cell ratio. (C) The CD38 expres-
sion level of the blast population was correlated with the cytotoxic effect observed at an E:T ratio of 5:1 for experiments conducted in (B), and a linear regression
model fitted using GraphPad Prism. (D) Bone marrow mononuclear cells from n=4 donors were treated with 10 nM all-trans retinoic acid (ATRA) or dimethyl sulfoxide
(DMSO) for 48 h prior to anti-CD38 staining. A representative histogram is displayed and summary data of four pooled donors were compared using an unpaired t-
test. (E) ATRA or DMSO pretreated cells were co-cultured with CD38 KD-CD38 CAR-NK cells. Summary data from three experiments. Analysis by unpaired t-test for
each E:T ratio. Statistical significance is defined as *P≤0.05, **P≤0.01, ***P≤0.001.

M. Gurney et al.
ment of multiple myeloma.28 Indeed, overcoming the frat- phosphorylation within NK cells. FT538, a NK cell prod-
ricidal effect of daratumumab through combination with uct derived from induced pluripotent stem cells being
ex vivo eNK cells is actively under investigation in multi- developed by FATE Therapeutics, incorporates a CD38
ple myeloma. In keeping with prior reports, we observed deletion to overcome fratricide when combined with
CD38 upregulation during NK cell expansion, which was daratumumab. The group also demonstrated greater
sufficient to lead to a fratricidal effect despite the use of resistance to oxidative stress conferred by deletion of
an optimized-affinity CD38 CAR design.18 While it has CD38, a characteristic likely to be favorable within the
been considered difficult to apply genetic engineering tumor microenvironment.34 These enhancements to NK
approaches to primary NK cells, we achieved a consis- cell biology suggest a broad range of applications for
tent, and high-efficiency disruption of the CD38 gene CD38 KD eNK cells beyond CD38 targeting and fratricide
using a multi-sgRNA approach coupled with a flow trans- concerns. Simple and consistent approaches to their gen-
fection system. Our findings are comparable to recent eration will likely be of clinical utility.
descriptions of CRISPR/Cas9 editing in primary NK cells In conclusion, we present two viable approaches to
but using a different sgRNA design and expansion CD38 CAR-NK cell therapies applied to AML. Both our
approach.27,29,30 The resulting CD38 KD eNK cells contin- CD38 CAR-KHYG-1 cells and CD38 KD eNK cell plat-
ued to expand and displayed reduced fratricide after forms overcome effector cell fratricide relating to NK cell
CD38 CAR expression. With the availability of CD38 expression. Furthermore, we report an efficient
CRISPR/Cas9 and the relative ease of application to pri- approach to CRISPR/Cas9 genome editing adapted to pri-
mary NK cells using clinically adaptable platforms now mary eNK cells and suitable for GMP expansion.
demonstrated by multiple groups, there are vast possibil-
ities for this technology across NK cell therapeutics. Disclosures
One potential limitation to CD38 targeting in AML is MG has received educational funding from Janssen
the limited capacity to target LSC populations, question- Pharmaceuticals and Takeda. AS and SS have received
ing the ‘curative’ potential of the therapies. LSC in AML research funding from ONK Therapeutics Limited. LKM is an
are well-established, and while our understanding has employee of ONK Therapeutics Limited. SK and RS are
evolved to include the existence of some CD38-positive employees of Maxcyte Inc. SZ has received research funding
LSC populations, it is likely that many LSC do reside from Takeda, Celgene, and Janssen and is a member of the
within the traditional CD34-positive, CD38-negative board of directors or an advisory committee for Takeda, Celgene,
compartment.31 Considering this feature of AML LSC, a and Janssen. NWCJvdD has received research funding from
CAR-NK cell targeting CD38 could be expected to have Janssen Pharmaceuticals, Amgen, Celgene, Novartis, and BMS
greater LSC targeting potential than a CAR-T cell, and has participated in advisory boards for Janssen
because of the presence of the innate activating pathways Pharmaceuticals, Amgen, Celgene, BMS, Takeda, Roche,
of NK cells above and beyond the CD38-specific CAR. Novartis, Bayer, and Servier. TM has received research funding
Indeed the potential for long-term disease control, and from Gilead, Celgene, Novartis, ONK Therapeutics Limited,
thus LSC targeting capabilities can be inferred from data Genmab, Janssen and has been a member of an advisory board
establishing the importance of NK cell KIR-ligand mis- for Janssen. ES has collaborated in research projects with
match in the efficacy of allogeneic stem cell transplanta- Janssen, Roche, Celgene, and Takeda. MOD has received
tion.16 Furthermore, a tandem CAR approach including a research funding from ONK Therapeutics Limited, BMS,
LSC-specific antigen and/or a variant of TRAIL (tumor Celgene, and Glycomimetics; is a member of the board of direc-
necrosis factor related apoptosis inducing ligand) could be tors or an advisory committee for Janssen, Abbvie, and ONK
incorporated to augment LSC targeting.32 Tailored Therapeutics Limited; and owns equity in ONK Therapeutics
approaches using CAR modified NK cells targeting com- Limited. EN has no potential conflicts of interest to disclose.
binations based on the specific identified LSC
immunophenotype in each case may ultimately be Contributions
required given the absence of an identified universal LSC SS and MOD conceived the research. EN, MG, AS and
marker. This approach is becoming feasible with current LKM performed functional assays. SK and RS contributed to
and emerging technologies. electroporation optimization. ES contributed to acquisition of
Antibody- and protein-based approaches have been patients’ samples and the primary AML assay design. TM, SZ
considered previously in attempts to overcome fratricide and NVD developed CD38 CAR-KHYG1 cells and associated
in a CD38-directed CAR-T cell platform.33 CRISPR/Cas9- functional assays. MG, SS and AS wrote the manuscript and
generated, CD38 KD eNK cells have recently and success- prepared the figures. All authors contributed to editing and
fully been applied to reducing the NK cell fratricidal effects reviewing the final manuscript prior to submission.
of daratumumab with a focus on multiple myeloma.27
Interestingly, while a magnetic separation step was uti- Acknowledgments
lized to enhance the purity of the KD population in this Reagents for electroporation were contributed by Maxcyte inc.
innovative study, our data suggest that expression of a Funding
CD38 CAR combined with a highly efficient CRISPR/ This research was supported by Irish Clinical Academic Training
Cas9 KD will likely lead to a self-selecting KD population (ICAT) Programme fellowship funding to MG. ICAT is supported
without additional processing. While not the focus of our by the Wellcome Trust and the Health Research Board (grant num-
experiments, Kararoudi et al.27 also explored the cellular ber 203930/B/16/Z), the Health Service Executive National
bioenergetic benefit of deletion of CD38 in eNK cells. Doctors Training and Planning and the Health and Social Care,
CD38 converts nicotinamide adenine dinucleotide (NAD+) Research and Development Division, Northern Ireland. Work per-
to cyclic adenosine diphosphate-ribose through an enzy- formed in collaboration with Blood Cancer Network Ireland was
matic function. Additional NAD+ availability due to loss of funded by Science Foundation Ireland and the Irish Cancer Society
CD38 supplies an important co-factor favoring oxidative (Blood Cancer Network Ireland, 14/ICS/B3042).

References 2009;32(7):737-743. and skews T-cell repertoire in multiple

13. Drent E, Themeli M, Poels R, et al. A rational myeloma. Blood.2016;128(3):384-394.
1. Surveillance, Epidemiology, and End Results strategy for reducing on-target off-tumor 24. Zhao C, Jia B, Wang M, et al. Multi-dimen-
(SEER) Program (1969-2018). U.S. effects of CD38-chimeric antigen receptors sional analysis identifies an immune signa-
Population Data with Other Software: by affinity optimization. Mol Ther. ture predicting response to decitabine treat-
(SEER Web site: www.seer.cancer.gov/pop- 2017;25(8):1946-1958. ment in elderly patients with AML. Br J
data), National Cancer Institute, DCCPS, 14. Björklund AT, Carlsten M, Sohlberg E, et al. Haematol. 2020;188(5):674-684.
Surveillance Research Program, released Complete remission with reduction of high- 25. Farber M, Arnold L, Chen Y, Möllmann M,
December 2019. risk clones following haploidentical NK-cell Duehrsen U, Hanoun M. Inhibition of CD38
2. DiNardo CD, Stein EM, de Botton S, et al. therapy against MDS and AML. Clin Cancer shows anti-leukemic activity in acute
Durable remissions with ivosidenib in Res. 2018;24(8):1834-1844. myeloid leukemia. Blood. 2018;132(Suppl
IDH1-mutated relapsed or refractory AML. 15. Nguyen R, Wu H, Pounds S, et al. A phase II 1):1456-1456.
N Engl J Med. 2018;378(25):2386-2398. clinical trial of adoptive transfer of hap- 26. Mouly E, Planquette C, Rousseau E,
3. Stein EM, DiNardo CD, Pollyea DA, et al. loidentical natural killer cells for consolida- Delansorne R. Inecalcitol respectively
Enasidenib in mutant IDH2 relapsed or tion therapy of pediatric acute myeloid induces or increases CD38 expression at the
refractory acute myeloid leukemia. Blood. leukemia. J Immunother Cancer. 2019;7 surface of CD38- or CD38+ AML cell lines
2017;130(6):722-731. (1):81. representative of all 9 FAB subtypes except
4. Stone RM, Mandrekar SJ, Sanford BL, et al. 16. Ruggeri L, Capanni M, Urbani E, et al. M6. Cancer Res. 2018;78(13 Suppl):1890.
Midostaurin plus chemotherapy for acute Effectiveness of donor natural killer cell 27. Naeimi Kararoudi M, Nagai Y, Elmas E, et al.
myeloid leukemia with a FLT3 mutation. N alloreactivity in mismatched hematopoietic CD38 deletion of human primary NK cells
Engl J Med. 2017;377(5):454-464. transplants. Science. 2002;295(5562):2097- eliminates daratumumab-induced fratricide
5. Maude SL, Laetsch TW, Buechner J, et al. 2100. and boosts their effector activity. Blood.
Tisagenlecleucel in children and young 17. Liu E, Marin D, Banerjee P, et al. Use of 2020;136(21):2416-2427.
adults with B-cell lymphoblastic leukemia. CAR-transduced natural killer cells in CD19- 28. Casneuf T, Xu XS, Adams HC, et al. Effects
N Engl J Med. 2018;378(5):439-448. positive lymphoid tumors. N Engl J Med. of daratumumab on natural killer cells and
6. Neelapu SS, Locke FL, Bartlett NL, et al. 2020;382(6):545-553. impact on clinical outcomes in relapsed or
Axicabtagene ciloleucel CAR T-cell therapy 18. Wang Y, Zhang Y, Hughes T, et al. Fratricide refractory multiple myeloma. Blood Adv.
in refractory large B-cell lymphoma. N Engl of NK cells in daratumumab therapy for 2017;1(23):2105-2114.
J Med. 2017;377(26):2531-2544. multiple myeloma overcome by ex vivo- 29. Pomeroy EJ, Hunzeker JT, Kluesner MG, et
7. Cummins KD, Gill S. Chimeric antigen expanded autologous NK cells. Clin Cancer al. A genetically engineered primary human
receptor T-cell therapy for acute myeloid Res. 2018;24(16):4006-4023. natural killer cell platform for cancer
leukemia: how close to reality? 19. Suck G, Branch DR, Smyth MJ, et al. KHYG- immunotherapy. Mol Ther. 2019;28(1):52-
Haematologica. 2019;104(7):1302-1308. 1, a model for the study of enhanced natural 63.
8. Kenderian SS, Ruella M, Shestova O, et al. killer cell cytotoxicity. Exp Hematol. 30. Kararoudi MN, Dolatshad H, Trikha P, et al.
Targeting CLEC12A with chimeric antigen 2005;33(10):1160-1171. Generation of knock-out primary and
receptor T cells can overcome the 20. Uruno A, Noguchi N, Matsuda K, et al. All- expanded human NK cells using Cas9
chemotherapy refractoriness of leukemia trans retinoic acid and a novel synthetic ribonucleoproteins. J Vis Exp. 2018;2018
stem cells. Biol Blood Marrow Transplant. retinoid tamibarotene (Am80) differentially (136):58237.
2017;23(3):S247-S248. regulate CD38 expression in human 31. Taussig DC, Miraki-Moud F, Anjos-Afonso
9. Petrov JC, Wada M, Pinz KG, et al. leukemia HL-60 cells: possible involvement F, et al. Anti-CD38 antibody-mediated clear-
Compound CAR T-cells as a double- of protein kinase C-δ. J Leukoc Biol. ance of human repopulating cells masks the
pronged approach for treating acute myeloid 2011;90(2):235-247. heterogeneity of leukemia-initiating cells.
leukemia. Leukemia. 2018;32(6):1317-1326. 21. Tang X, Yang L, Li Z, et al. First-in-man clin- Blood. 2008;112(3):568-575.
10. Naik J, Themeli M, de Jong-Korlaar R, et al. ical trial of CAR NK-92 cells: safety test of 32. Szegezdi E, Reis CR, Sloot AM van der, et al.
CD38 as a therapeutic target for adult acute CD33-CAR NK-92 cells in patients with Targeting AML through DR4 with a novel
myeloid leukemia and T-cell acute lym- relapsed and refractory acute myeloid variant of rhTRAIL. J Cell Mol Med.
phoblastic leukemia. Haematologica. 2019; leukemia. Am J Cancer Res. 2018;8(6):1083- 2011;15(10):2216-2231.
104(3):e100-e103. 1089. 33. Gao Z, Tong C, Wang Y, Chen D, Wu Z,
11. Drent E, Groen RWJ, Noort WA, et al. Pre- 22. Zabaleta A, Tomas J, Simoes C, et al. The Han W. Blocking CD38-driven fratricide
clinical evaluation of CD38 chimeric antigen mode of action of the anti-CD38 monoclon- among T cells enables effective antitumor
receptor engineered T cells for the treatment al antibody (MAB) isatuximab in elderly activity by CD38-specific chimeric antigen
of multiple myeloma. Haematologica. acute myeloid leukaemia (AML). receptor T cells. J Genet Genomics.
2016;101(5):616-625. Hemasphere. 2020;4(S1):Abstract Book 2019;46(8):367-377.
12. Mihara K, Yanagihara K, Takigahira M, et al. EP467. 34. Cichocki F, Woan K, Wu C-Y, et al. NK cells
Activated T-cell-mediated immunotherapy 23. Krejcik J, Casneuf T, Nijhof IS, et al. lacking CD38 are resistant to oxidative
with a chimeric receptor against CD38 in B- Daratumumab depletes CD38+ immune stress-induced death. Blood. 2019;134(Supp
cell non-Hodgkin lymphoma. J Immunother. regulatory cells, promotes T-cell expansion, 1):3215-3215.

ARTICLE Cell Therapy & Immunotherapy
Ferrata Storti Foundation Successful gene therapy of Diamond-Blackfan

anemia in a mouse model and human CD34+
cord blood hematopoietic stem cells using a
clinically applicable lentiviral vector
Yang Liu,1 Maria Dahl,1 Shubhranshu Debnath,1 Michael Rothe,2 Emma M.
Smith,1 Tan Hooi Min Grahn,1 Sarah Warsi,1 Jun Chen,1 Johan Flygare,1 Axel
Schambach2,3 and Stefan Karlsson1
Haematologica 2022 1
Molecular Medicine and Gene Therapy, Lund Stem Cell Center, Lund University, Lund,
Volume 107(2):446-456 Sweden; 2Institute of Experimental Hematology, Hannover Medical School, Hannover,
Germany; and 3Division of Hematology/Oncology, Boston Children’s Hospital, Harvard
Medical School, Boston, MA, USA
ABSTRACT
D
iamond-Blackfan anemia (DBA) is an inherited bone marrow fail-
ure disorder in which pure red blood cell aplasia is associated with
physical malformations and a predisposition to cancer. Twenty-
five percent of patients with DBA have mutations in a gene encoding ribo-
somal protein S19 (RPS19). Our previous proof-of-concept studies demon-
strated that DBA phenotype could be successfully treated using lentiviral
vectors in Rps19-deficient DBA mice. In our present study, we developed
a clinically applicable single gene, self-inactivating lentiviral vector, con-
taining the human RPS19 cDNA driven by the human elongation factor 1a
short promoter, which can be used for clinical gene therapy development
for RPS19-deficient DBA. We examined the efficacy and safety of the vec-
tor in a Rps19-deficient DBA mouse model and in human primary RPS19-
deficient CD34+ cord blood cells. We observed that transduced Rps19-defi-
cient bone marrow cells could reconstitute mice long-term and rescue the
bone marrow failure and severe anemia observed in Rps19-deficient mice,
Correspondence: with a low risk of mutagenesis and a highly polyclonal insertion site pat-
STEFAN KARLSSON tern. More importantly, the vector can also rescue impaired erythroid dif-
stefan.karlsson@med.lu.se ferentiation in human primary RPS19-deficient CD34+ cord blood
hematopoietic stem cells. Collectively, our results demonstrate the efficacy
YANG LIU and safety of using a clinically applicable lentiviral vector for the successful
yang.liu@med.lu.se treatment of Rps19-deficient DBA in a mouse model and in human pri-
mary CD34+ cord blood cells. These findings show that this vector can be
Received: August 27, 2020. used to develop clinical gene therapy for RPS19-deficient DBA patients.
Pre-published: January 14, 2021.
Introduction
https://doi.org/10.3324/haematol.2020.269142 Diamond-Blackfan anemia (DBA) is a congenital bone marrow (BM) failure dis-
order with erythroid hypoplasia that presents early in infancy.1-3 The classic hema-
tologic profile of DBA consists of macrocytic anemia with reticulocytopenia, nor-
©2022 Ferrata Storti Foundation mal or decreased levels of neutrophils, and a variable platelet count.3,4
Material published in Haematologica is covered by copyright. Additionally, patients with DBA can also manifest with non-hematologic symp-
All rights are reserved to the Ferrata Storti Foundation. Use of toms such as physical abnormalities and a predisposition to cancer.2,5
conditions:
The majority of DBA cases (60-70%) are caused by heterozygous loss-of-func-
https://creativecommons.org/licenses/by-nc/4.0/legalcode. tion mutations in genes coding for ribosomal proteins (RP), resulting in functional
Copies of published material are allowed for personal or inter- RP haploinsufficiency.2 Recent studies have also identified mutations in erythroid
nal use. Sharing published material for non-commercial pur- transcriptional factors GATA1 and TSR2 (a direct binding partner of RPS26) as a
cause of the DBA phenotype.6-8 Mutations in RPS19 are the most common alter-
sect. 3. Reproducing and sharing published material for com- ations among patients with putative causal mutations, contributing to over 25%
mercial purposes is not allowed without permission in writing of cases.2 The main therapeutic option for DBA patients is corticosteroids, with
from the publisher. more than 80% of the subjects responding well during early stages of treatment.
However, half of these patients become non-responsive to corticosteroid therapy
over prolonged treatment and have to be given blood transfusions.9 Importantly,

Gene therapy of DBA
none of the currently available treatments is curative, and mented with penicillin/streptomycin (Gibco), murine stem cell
the treatments are often accompanied by serious compli- factor (100 ng/mL; PeproTech), and human thrombopoietin (50
cations.1,2,10 Hematopoietic stem cell transplantation is ng/mL; PeproTech) in six-well plates at the concentration of
currently the sole curative option for the treatment of 0.5x106 cells/mL. For transduction, retronectin-coated (20
DBA. This treatment is, however, limited by the availabil- ng/mL; Takara) 12-well plates were preloaded with the viral vec-
ity of suitable donors and the potential for serious tors (multiplicity of infection [MOI]=5-10), followed by seeding
immunological complications.11 Gene therapy, using of 0.5x106 cells into each well filled with 1 mL pre-stimulation
gene-corrected hematopoietic stem cells, would be a medium.
potential alternative therapeutic strategy, as highlighted
in our previous studies.12-16 For clinical applications using Flow cytometry
this approach, the efficiency of transgene expression and A complete description of all antibodies used is listed in the
safety aspects, including potential insertional mutagene- Online Supplementary Methods.
sis, should be addressed.17-19 Our group recently demon-
strated that correction of Rps19-deficient BM cells using Human primary cord blood cells and erythroid
lentiviral vectors containing a clinically relevant promoter differentiation
could rescue BM failure and defects in erythroid develop- Human cord blood samples were obtained from the maternity
ment, while exhibiting limited risk of insertional mutage- wards of Helsingborg General Hospital and Skåne University
nesis.12 Hospital in Lund and Malmö, Sweden, after informed, written
In our previous study, we utilized a lentiviral vector consent according to guidelines approved by the regional ethical
containing the RPS19 gene as well as a GFP marker. We committee. Mononuclear cells were separated through density-
have subsequently designed a clinically applicable single gradient centrifugation. CD34+ cells were magnetically isolated
gene lentiviral self-inactivating (SIN) vector for the clinical according to the manufacturer’s description (Milteny Biotec, cat.
development of gene therapy for Rps19-deficient DBA n. 130-046-702). Cells were cultured in serum-free expansion
patients. This vector harbors a codon-optimized human medium (Stem Cell Technologies), supplemented with human
RPS19 cDNA driven by the short human elongation fac- stem cell factor, thrombopoietin, and FLT3-ligand at 100 ng/mL
tor 1a promoter and lacks a fluorescent marker. In this from Peprotech. Full descriptions of transduction and erythroid
study, we demonstrate that this vector can rescue the differentiation are provided in the Online Supplementary Methods.
anemia and lethal BM failure observed in mouse models
of Rps19-deficient DBA, with a low-risk insertion profile Other experimental details
and no evidence of clonal expansion associated with vec- Full descriptions of the quantitative reverse transcriptase poly-
tor integration near cancer-associated genes. We also merase chain reaction (qRT-PCR), determination of transduction
observed the rescue of impaired erythroid differentiation efficiency and vector copy number measurements are provided
in human RPS19-deficient CD34+ cord blood cells treated in the Online Supplementary Methods.
with this vector. Our results demonstrate the feasibility
and preclinical efficacy for treatment of RPS19-deficient Insertion site analysis
DBA using a clinically applicable SIN lentiviral vector, Whole BM cells were isolated at 16 weeks after transplanta-
which opens the possibility for the development of clini- tion. Genomic DNA was isolated from the BM of flushed femo-
cal gene therapy for RPS19-deficient DBA patients. ra using the DNA Blood & Tissue kit (Qiagen). The vector-
genome junction was amplified using the INtegration Site
PIpeline for paIRed-End reaDs (INSPIIRED) workflow as
Methods described by Sherman and colleagues.24
Lentiviral vector constructs Statistical analysis

The SIN lentiviral vector is derived from the pRRL.PPT.PGK. t-tests and one-way analysis of variance with the Tukey mul-
vector backbone.20 A codon-optimized human RPS19 cDNA was tiple-comparison test were used to determine statistical signifi-
designed as described previously21 and inserted downstream of cance. Computations were performed using GraphPad Prism
the EFS promoter. Lentiviral vectors were produced by the (version 6; GraphPad Software).
Vector Unit at Lund University as previously described.12
Mice and transplantations Results

Mice were maintained at the Lund University animal facility
and all animal experiments were approved by the Lund High transduction efficiency of the EFS-RPS19 vector
University animal ethics committee. The homozygous doxycy- We first studied the transduction efficiency of the vec-
cline-inducible Rps19-deficient mouse model used in the study tor in BM progenitor cells isolated from our established
was established as previously described.15 A detailed description Rps19-deficient DBA mouse model.15 This model contains
of transplantations is provided in the Online Supplementary the Rps19-targeting shRNA expressed under a doxycy-
Methods. cline-responsive promoter located downstream of the
collagen A1 gene (Figure 1A). Experimental animals were
Transduction bred to be either heterozygous (D/+) or homozygous
c-kit+ or lineage negative (Lin–) cells isolated from BM of trans- (D/D) for the shRNA to generate two models with inter-
genic mice were enriched by using CD117 or Lin– microbeads mediate or severe Rps19 deficiency (Figure 1B). Rps19
and magnetic-activated cell sorting separation columns (all from mRNA expression was reduced by approximately 50%,
Miltenyi Biotec) according to the manufacturer’s protocol. After and a trend toward more efficient knockdown in D/D
enrichment, cells were pre-stimulated for 24 h in StemSpan mice compared to D/+ mice was seen, as shown in our
serum-free expansion medium (Stem Cell Technologies), supple- previous studies.15 Upon induction with doxycycline,

Y. Liu et al.
transplanted recipients receiving D/D BM cells developed Gene-corrected bone marrow cells can rescue
an acute and lethal BM failure, while recipients of D/+ the Diamond-Blackfan anemia phenotype in vivo
BM cells developed a mild chronic anemia.15 As D/D mice We next assessed the function of gene-corrected BM
develop lethal BM failure shortly after doxycycline induc- cells using the EFS-RSP19 vector in vivo. As shown in
tion, the more severely affected D/D mice were used in Figure 2A, uninduced (no doxycycline) c-kit+ BM cells
the present study to investigate whether the lethal phe- from D/D mice (CD45.2) were transduced with the EFS-
notype could be fully rescued with the vector. RPS19 vector (MOI=5-10), and then transplanted into
The clinically applicable single gene lentiviral vector lethally irradiated wild-type B6SJL recipient mice
was developed using a SIN lentiviral vector design har- (CD45.1/CD45.2, named the EFS-RPS19 group). Mice
boring the codon-optimized human RPS19 cDNA driven receiving uninduced c-kit+ BM cells without vector trans-
by an internal EFS promoter (named EFS-RPS19) (Figure duction were regarded as the mock group (negative con-
1C). Compared to the human codon-optimized RPS19 trol). Following engraftment and stable donor-derived
cDNA, there are six mismatches in the shRNA construct reconstitution of the hematopoietic system, doxycycline
for generating the mouse model. Because of this, gene was administrated to all recipients to induce the DBA
expression derived from the human codon-optimized phenotype. To determine whether the vector-treated cells
RPS19 cDNA is not affected by the shRNA. could achieve a full correction, age-matched B6SJL wild-
We first examined the transduction efficiency of the type (WT) mice receiving no irradiation and no transplan-
vector. In our previous study, we transduced cells at a tation but the same doxycycline administration were
MOI of 10-20 and demonstrated the rescue of the anemia used as the control group.
and BM failure with a reduced risk of insertional mutage- Before doxycycline administration, both the mock and
nesis.12 To investigate the therapeutic effects of a lower EFS-RPS19 groups showed high overall donor reconstitu-
average vector copy number per cell, we decided to use tion, indicating minimal to absent recipient-derived
an MOI of 5-10. As shown in Online Supplementary Figure hematopoiesis (Online Supplementary Figure S3). After
S1, the transduction efficiency was 75% on average in c- induction with doxycycline for 2 weeks, recipients in the
kit+ cells isolated from D/D mice. We next examined both mock group showed a dramatic decrease in red blood cell
endogenous Rps19 and vector-derived RPS19 mRNA counts, mean corpuscular volume (MCV), and white
expression in c--kit+ cells isolated from D/D and +/+ mice blood cell and platelets counts, indicating that the mice
at a MOI of 5. As shown in Online Supplementary Figure developed BM failure shortly after doxycycline adminis-
S2, the endogenous Rps19 mRNA expression levels were tration (Figure 2B-F). In contrast, recipients in the EFS-
significantly decreased in the cells isolated from D/D RPS19 vector-treated group showed normal blood cellu-
mice compared to the levels in cells from +/+ mice after larity compared to the WT group. To assess the long-term
doxycycline administration. Cells transduced with EFS- therapeutic effects, recipients were administered doxycy-
RPS19 exhibited a 2.5-fold higher level of expression of cline for 16 weeks (Figure 3A). As shown in Figure 3B,
human RPS19 mRNA compared to the endogenous Rps19 most of the recipients in the mock group died (9 out of
expression in the D/D group. Interestingly, transduced 16) due to severe anemia or BM failure (data not shown) at
cells isolated from +/+ mice showed a significantly lower 2-3 weeks after doxycycline administration. The few
level of human RPS19 mRNA expression than the trans- remaining recipients exhibited a macrocytic anemia phe-
duced cells isolated from D/D mice at the same MOI, notype with significantly reduced red blood cell counts
indicating the internal physiological regulation of excess and increased MCV at 16 weeks. The hemoglobin levels
RPS19 production as reported by others.29,30 The overall and platelet counts were also decreased compared to
results indicated that cells transduced with the EFS-RPS19 those in the WT group (Figure 3C-G). As expected, there
vector could successfully express the human RPS19 trans- was a significantly decreased expression of endogenous
gene Rps19 in donor-derived BM cells from both the mock and
A B
Figure 1. The inducible Rps19-deficient mouse model and structure of the EFS-RPS19 self-inactivating lentiviral vector. (A) Overview of modified loci. Black arrow-
heads indicate the transcriptional start sites. (B) Breeding strategy to adjust the level of Rps19 downregulation. Homozygous mice (D/D mice) are used in the project.
(C) The self-inactivating lentiviral vector harboring a codon-optimized human RPS19 cDNA driven by human elongation factor 1a short (EFS) promoter. LTR: long ter-
minal repeat; pA: polyadenylation signal; PPT: polypurine tract; RRE: Rev response element; SA: splice acceptor.

Gene therapy of DBA
EFS-RPS19 groups after 16 weeks of doxycycline admin- recipient cells (CD45.1/CD45.2) in the few surviving
istration (Online Supplementary Figure S4). Strikingly, all mice in the mock group than in the EFS-RPS19 group, per-
the mice in the EFS-RPS19 group survived without any haps explaining why these animals did not develop
signs of anemia and with normal BM cellularity com- severe BM failure.
pared to the WT group. These results indicate that the
lethal BM failure can be prevented by the vector. The vec- EFS-RPS19 vector-treated Rps19-deficient bone
tor copy number at 16 weeks after transplantation was marrow cells provide long-term reconstitution
5.2±1.6 and 4.7±1.0 on average in gene-corrected cells We next investigated whether the EFS-RPS19 vector-
isolated from peripheral blood and BM, respectively treated Rps19-deficient BM cells could generate long-term
(Figure 4A, B). We also analyzed the fraction of myeloid- engraftment and reconstitution in doxycycline-induced
erythroid compartments by flow cytometry (Online lethally irradiated WT recipients (Figure 5A). To this end,
Supplementary Figure S5). Donor-derived hematopoiesis D/D mice were induced with doxycycline for 1 week and
was observed in the EFS-RPS19 group, and the mean per- red blood cell counts and hemoglobin levels were meas-
centage of donor cells (CD45.2) in every progenitor pop- ured to confirm the DBA phenotype (Online Supplementary
ulation was significantly higher in the EFS-RPS19 group Figure S6). The isolated Lin– BM cells from doxycycline-
than in the mock group (Figure 4C-I). Unlike in the EFS- induced mice were transduced with the vector and trans-
RPS19 group, the transplanted cells in the mock group planted into the doxycycline-induced lethally irradiated
had limited reconstituting ability. In addition, we WT recipients (EFS-RPS19 group). Recipients receiving
observed a significantly higher reconstitution of resident untransduced Rps19-deficient Lin– BM cells from doxycy-
B C D
E F
Figure 2. Effective correction of anemia by

the EFS-RPS19 vector at 2 weeks after
induction of the Diamond-BLackfan pheno-
type. (A) The scheme of the uninduced gene-
corrected cell transplantation model and
plan for examining short-term therapeutic
effects. (B-F) Blood cellularity at 2 weeks
after doxycycline induction (n=13-16, error
bars represent the standard deviation,
*P<0.05, **P<0.01, ***P<0.005
****P<0.001 by one-way analysis of vari-
ance). BM: bone marrow; MOI: multiplicity of
infection; WT: wild-type; RBC: red blood cells;
MCV: mean corpuscular volume; WBC: white
blood cells.

Y. Liu et al.
cline-induced D/D mice were regarded as the mock group. group, which was significantly higher than in the mock
The setting of the WT group was the same as described group (Figure 6C-I). Taken together, our results demon-
above. Doxycycline was administrated to all the recipients strate that the EFS-RPS19 vector-treated group obtained
directly after transplantation. After induction with doxycy- full correction of anemia and BM failure.
cline for 2-3 weeks, the majority of mice in the mock group
died (13 out of 16 animals) due to severe anemia or BM Gene-corrected bone marrow cells showed polyclonal
failure (Figure 5B, Online Supplementary Figure S7). The few hematopoiesis and had a typical lentiviral insertion
surviving mice exhibited significantly decreased levels of profile
endogenous Rps19 expression in donor-derived BM cells at The risk of insertional mutagenesis is a major concern
16 weeks after doxycycline administration (Online for future applications of gene therapy in the clinic. To
Supplementary Figure S8) with the concomitant develop- assess the safety of the EFS-RPS19 vector integration pro-
ment of a severe anemia phenotype in the mock group file, as well as the clonal dynamics of the transduced cells,
(Figure 5C-G). In contrast, all recipients in the EFS-RPS19 insertion site analysis was performed using the INSPI-
group survived with normal blood cellularity compared to IRED workflow.24 BM cells from uninduced donors
the WT group. The vector copy number was on average (Figure 3A) of cohort 1 (animals 5-8) and cohort 2 (ani-
8.3±4.0 and 10.9±3.9 in gene-corrected Rps19-deficient mals 13-16), or those from doxycycline-induced donors
cells isolated from peripheral blood and BM, respectively (Figure 5A) of cohorts 3 (animals 7-11) and cohort 4 (ani-
(Figure 6A, B). By analyzing the fraction of myeloid-ery- mals 17-20) were isolated 16 weeks after disease induc-
throid compartments at 16 weeks, we observed almost tion in the recipients. Detailed information on pool size
complete donor-derived hematopoiesis in the EFS-RPS19 estimation and sequence diversity in each sample is
B C D
E F G
Figure 3. Effective long-term correction of the anemia and bone marrow failure in mice treated with the EFS-RPS19 vector. (A) The scheme of the induced gene-
corrected cell transplantation model and the plan for examining long-term therapeutic effects. (B) Surviral rate analysis. (C-G) Blood cellularity at 16 weeks after doxy-
cycline induction (n=13-16, error bars represent the standard deviation, *P<0.05, **P<0.01, ***P<0.005 by one-way analysis of variance). BM: bone marrow; MOI:
multiplicity of infection; WT: wild-type; RBC: red blood cells; MCV: mean corpuscular volume; WBC: white blood cells.

Gene therapy of DBA
shown in Online Supplementary Tables S2 and S3. proto-oncogenes (Figure 7B-D, Online Supplementary
Particularly, the top ten integrations with the highest Figure S9B-D). More integrations were detected in a dis-
sequence contribution in each sample are depicted in tance of 100 kb relative to CpG islands, marking actively
Figure 7A and Online Supplementary Figure S9A. Overall, transcribed regions but not in the direct vicinity of CpG
EFS-RPS19-transduced cells showed a highly polyclonal islands (1-10 kb), hence not close to the promoter region.
insertion site pattern, reflecting the overall integration Our data showed that GC rich regions (marking promoter
preferences of lentiviral vectors. No integration site con- regions of genes) and long intergenic regions were gener-
tributed with more than 5.37% in uninduced gene-cor- ally disfavored by the vector. The integrations inside tran-
rected cells (Figure 7A) or 4.4% in gene-corrected Rps19- scriptional units and in or close to proto-oncogenes (with-
deficient cells (Online Supplementary Figure S9A) to the in a 100 kb window = onco.100k) are displayed relative
overall sequence pool, and there were no integrations in to the matched random controls (Figure 7B, Online
or close to known high-risk proto-oncogenes (Lmo2, Supplementary Figure S9B).
Ccnd2 or Hmga2). However, one integration at 16.5 kb
upstream of the high-risk locus Ikzf1 was detected The EFS-RPS19 vector rescued impaired erythroid
(accounting for 1.65%), and another integration at 20.7 differentiation of human RPS19-deficient CD34+
kb upstream of the high-risk gene Mecom was found cord blood cells
(accounting for 0.04%) in recipients of gene-corrected We next examined the therapeutic effects of the EFS-
Rps19-deficient BM cells. For the analysis of overlaps RPS19 vector using human primary CD34+ cord blood
between integrations in or near the same genes among cells. Since primary CD34+ cells from DBA patients are dif-
BM samples, we observed common lentiviral integration ficult to obtain, we utilized previously validated lentiviral
sites identified in previous integration site analysis shRNA vectors that silence RPS19 expression in human
(Online Supplementary Tables S4 and S5).31,32 We identified CD34+ cells to induce a DBA phenotype.23
11 integrations in or near the same refSeq genes between Two lentiviral vectors expressing shRNA (shRNA1 and
cohorts 1+2 and cohorts 3+4 (Online Supplementary Figure shRNA2) targeting different regions of the human RPS19
S10). Four of these shared common insertion sites in or mRNA sequence were used to induce the DBA pheno-
near Hgf, Kdm6a, Lnpep and Mef2c also listed as proto- type, and a vector expressing a scrambled shRNA
oncogenes in the All Onco database.25 The sole occur- sequence (Scr) was used as a healthy control. To ensure
rence of integration sites in or near high-risk loci was not that the human construct in the vector would not be
an indication of a higher risk of insertional mutagenesis if degraded by the targeting shRNA, alignments were per-
no dominant clones were detected. The detected integra- formed and there were two mismatches (out of 19
tions might simply reflect that lentiviral vectors were nucleotides) for shRNA1 and four mismatches (out of 19
capable of integrating at these genomic sites. We per- nucleotides) for shRNA2. Thus, it is very unlikely that the
formed analysis of insertion site profile including param- gene expression derived from the human codon-opti-
eters of the integration site preferences close to CpG mized RPS19 cDNA would be affected by either shRNA.
islands, GC-rich regions, in or near transcription units, Since the shRNA vectors also contain a GFP marker gene,
the transcriptional start site of genes, gene boundaries or transduced GFP+ cells were sorted for further examination.
A B C D
E F G H I
Figure 4. Gene-corrected bone marrow cells show a competitive advantage in contributing to long-term hematopoiesis in vivo. (A, B) Vector copy number in periph-
eral blood (A) and bone marrow (B). (C, D) Donor reconstitution in peripheral blood (C) and bone marrow (D). (E–I) The percentage of transduced cells in hematopoi-
etic stem cells (E), megakaryocyte progenitors (F), pre-granulocyte-macrophage and granulocyte-macrophage progenitors (G), pre-megakaryocyte-erythroid (H), and
pre-colony-forming unit erythroid and colony-forming unit erythroid (I) (n=13-16, error bars represent the standard deviation, black asterisks indicate the statistical
significance of the comparison of recipient-derived cells between the mock and EFS-RPS19 groups, orange asterisks indicate the statistical significance of the com-
parison of donor-derived cells between the mock and EFS-RPS19 groups. *P<0.05, **P<0.01, ***P<0.005, ****P<0.001 by one-way analysis of variance). VCN:
vector copy number; PB: peripheral blood; BM: bone marrow; HSC: hematopoietic stem cells; MkP: megakaryocyte progenitors; pre-GM/GMP: pre-granulocyte
macrophage and granulocyte macrophage progenitors; preMegE: pre-megakaryocyte-erythroid; preCFU-E/CFU-E: pre-colony-forming unit–erythroid (CFU-E)/CFU-E.

Y. Liu et al.
As shown in Figure 8A, both shRNA1 and shRNA2 signif- and GFP– cells (Online Supplementary Figure S12A-B). By
icantly decreased RPS19 mRNA expression, with slightly further analysis of cell distribution, significantly decreased
more efficient knockdown being obtained with shRNA1 outputs of CD71+CD235A– cells on day 6 and CD71–
than shRNA2, which resembles our previous observations CD235A+ cells on day 10 were observed in GFPhigh popula-
in D/+ and D/D mouse models.15 In order to examine the tions of RPS19-deficient groups (Online Supplementary
function of EFS-RPS19, we transduced sorted GFP+ cells Figure S12C, D). The impaired differentiation was rescued
and cultured them in erythroid differentiation medium for by EFS-RPS19, with significantly increased GFPhigh popula-
48 h after transduction. The PCR results indicated success- tions (1.6-fold for shRNA1 and 1.8-fold for shRNA2)
ful integration of the vector into human cells, as shown in (Figure 8B) and progenitor populations (CD71+CD235A–)
Online Supplementary Figure S11. It is a well-known phe- (Online Supplementary Figure S12C) compared to the
nomenon that lentiviral vector-mediated transgene RPS19-deficient groups on day 6. Particularly, during ter-
expression is silenced in a fraction of CD34+ cells, and that minal erythropoiesis on day 16, cells in the RPS19-defi-
this fraction increases during differentiation of cells, likely cient groups showed reduced red blood cell production
due to changes in chromatin accessibility. As shown in (especially in the shRNA2 group) and few GFPhigh cells
Figure 8B, transgene silencing was evident in a fraction of (<1%) could be detected (Figure 8C-E). However, EFS-
the Scr-transduced CD34+ cells in which the fraction of RSP19 vector-treated groups produced more red blood
GFPhigh shRNA expressing cells decreased from 100% in cells and maintained significantly higher GFPhigh popula-
the sorted cells to around 90% on day 6 and day 10, with tions (~4%) than the RPS19-deficient groups. These
a further reduction to around 25% at day 16. However, in results collectively demonstrate that the EFS-RPS19 vector
the RPS19-deficient groups, the loss of GFPhigh cells was can rescue the impaired erythroid differentiation in
evident from day 6, with an increased fraction of GFPlow human primary RPS19-deficient CD34+ cord blood cells.
B C D
E F G
Figure 5. Amelioration of disease phenotype in Rps19-deficient animals transplanted with gene-corrected cells. (A) Scheme of the gene-corrected Rps19-deficient
cell transplantation model and plan for examining short-term and long-term therapeutic effects. (B) Survival rate analysis. (C-G) Blood cellularity at 4 and 16 weeks
after doxycycline induction (n=14-16, error bars represent the standard deviation, *P<0.05, **P<0.01, ***P<0.005 ****P<0.001 by one-way analysis of variance).
WT: wild-type; RBC: red blood cells; MCV: mean corpuscular volume; WBC: white blood cells.

Gene therapy of DBA
Discussion growth. Since single-copy insertion of the therapeutic

gene in the target cells is suggested to avoid the risk of
Currently, hematopoietic stem cell transplantation is genotoxicity in clinical gene therapy manipulation,19 we
the sole curative option for DBA patients, but suitable plan to examine the therapeutic effects using transduced
donors are often unavailable and there can be serious cells with a lower MOI (e.g. MOI=1) in future studies. Our
immunological complications.3,33 Gene therapy using results also demonstrated that the vector could rescue the
gene-corrected hematopoietic stem cells has been shown impaired erythroid differentiation of RPS19-deficient cord
to be a promising therapeutic strategy for genetic blood blood cells by increasing red blood cell production.
disorders in recent years.19,34,35 Our previous proof-of-con- Overall, we showed that the EFS-RPS19 vector could res-
cept studies also demonstrated the feasibility of applying cue the anemia and BM failure of RPS19-deficient DBA.
gene therapy to cure DBA.12-16 In the present study, a clin- Apart from efficacy, vector safety is the other essential
ically applicable lentiviral vector was used to investigate factor to assess when applying gene therapy. The risk of
the efficacy and safety for treating anemia and lethal BM insertional mutagenesis is a concern for future applications
failure in Rps19-deficient mice and for ameliorating the of gene therapy in the clinic. To prevent this risk, we uti-
impaired erythroid differentiation in human primary lized a third-generation SIN lentiviral vector that lacks
RPS19-deficient CD34+ cord blood cells. The next step potent enhancers in the long terminal repeat regions, since
towards clinical gene therapy will be to perform toxicol- such vectors were shown to exhibit a safer integration
ogy and biodistribution analyses, and thereafter proceed profile in previous clinical trials19,34,36 and also in our previ-
with submitting an application to regulatory authorities ous animal studies.12,15 By using the state-of-the-art INSPI-
in order to initiate a phase I/II clinical trial with 6-12 IRED workflow, which can provide better quantification
patients focusing on safety. of clonal abundance compared to a linear amplification-
For successful development of clinical gene therapy, vec- mediated PCR approach, we found that gene-corrected
tor efficacy, generating long-term therapeutic effects, is BM cells in both models exhibited a low risk of mutagen-
crucial. To examine the therapeutic effects of the vector esis with no evidence of clonal expansion associated with
with lower copies per cell, we decreased the MOI to 5-10 vector integration near cancer-associated genes. In our
in the present study and demonstrated that the EFS-RPS19 study, no hematologic abnormalities were observed due
vector has robust therapeutic effects with no evidence of to enforced expression of RPS19. The results collectively
clonal expansion associated with vector integration near demonstrate the safety of the EFS-RPS19 vector for clinical
cancer-associated genes, as we showed in our previous gene therapy development. The bioinformatic pipeline of
study.12 It has already been shown that ribosomal proteins the INSPIIRED workflow is a more automated approach,
are produced in excess of the needs of the ribosome making it well suited for monitoring patients in gene ther-
assembly, and that the excess protein is subjected to pro- apy trials in the future.
teasomal degradation.22,29,30 Similarly, in the present study, Increased MCV, due to macrocytic anemia, is a classic
our transduced Rps19-deficient cells had physiological lev- clinical observation in patients with DBA. It is, at least in
els of expression of RPS19. Hence, it is unlikely that the part, caused by the stabilization of p53 and activation of
ectopic expression of RPS19 would promote uncontrolled p53 targets (e.g., p21, Bax), which are responsible for cell
A B C D
E F G H I
Figure 6. EFS-RPS19 vector-treated Rps19-deficient cells show a competitive advantage in contributing to long-term hematopoiesis in vivo. (A, B) Vector copy num-
ber in peripheral blood (A) and bone marrow (B). (C, D) Donor reconstitution in peripheral blood (C) and bone marrow (D). (E–I) The percentage of transduced cells
in hematopoietic stem cells (E), megakaryocyte progenitors (F), pre-granulocyte-macrophage and granulocyte-macrophage progenitors (G), pre-megakaryocyte-ery-
throid (H), and pre-colony-forming unit erythroid and colony-forming unit erythroid (I) (n=14-16, error bars represent the standard deviation, black asterisks indicate
the statistical significance of the comparison of recipient-derived cells between the mock and EFS-RPS19 groups, orange asterisks indicate the statistical signifi-
cance of the comparison of donor-derived cells between the mock and EFS-RPS19 groups. *P<0.05, **P<0.01, ***P<0.005, ****P<0.001 by one-way analysis of
variance). VCN: vector copy number; PB: peripheral blood; BM: bone marrow; HSC: hematopoietic stem cells; MkP: megakaryocyte progenitors; pre-GM/GMP: pre-
granulocyte macrophage and granulocyte macrophage progenitors; preMegE: pre-megakaryocyte-erythroid; preCFU-E/CFU-E: pre-colony-forming unit –erythroid (CFU-
E)/CFU-E.

Y. Liu et al.
cycle arrest in G0/G1 phases leading to larger cell size.9,37 absolute numbers of hematopoietic stem and progenitor
We also observed a significantly increased MCV in the cells in BM due to Rps19 deficiency,15 which led to the
mock group after induction with doxycycline for more lethal BM failure we observed in untreated animals short-
than 4 weeks (Figures 3E and 5E). In particular, during the ly after doxycycline administration. This is also supported
initial phase after doxycycline induction, the majority of by the limited reconstitution ability of progenitor com-
red blood cells in the circulation were produced before the partments in the mock group (Figure 4C-I) and impaired
DBA phenotype was induced. Since old red blood cells are erythroid differentiation of human primary RPS19-defi-
smaller than the newly produced cells, the MCV is cient cord blood cells. RPS19-deficient patients who
decreased compared to normal for a short period during develop thrombocytopenia and neutropenia also experi-
the first 2 weeks after doxycycline induction (Figure 2E). ence similar progressive phenotypes of hypocellularity in
In addition to the severe anemia phenotype after induc- the BM.38,39 Moreover, DBA is a very heterogeneous dis-
tion of Rps19 deficiency, we observed decreased white ease. It is unknown why family members with the same
blood cell and platelet counts in the mock group. Mild genetic mutation in RPS19 may have very different phe-
thrombocytopenia and neutropenia (low levels of neu- notypes, ranging from no anemia to severe anemia with
trophilic granulocytes) have also been observed in about progression to multilineage BM failure.40 The current
25% of DBA patients during the course of the disease.4,37,38 understanding of phenotype-genotype correlations is far
In addition, several patients with RPS19-deficient DBA from comprehensive and needs to be studied further.
have developed myelodysplastic syndrome with multilin- Although the majority of mice in the mock group died,
eage cytopenia, which suggests a multilineage defect.37 a few mice did survive until the planned endpoint at 16
This observation correlates with a reduction in the weeks. One of the possible reasons for this may be the
A B
Figure 7. Gene-corrected bone marrow cells show a vector integration pattern that indicates low risk of mutagenesis and a highly polyclonal insertion site pattern.
(A) The top ten integration sites in each sample (*indicates that the integration was within a transcription unit, ~ indicates that the insertion was within 50 kb of a
cancer-related gene). (B, C) Percent of all integrations inside transcriptional units (B) and percent of integrations within 100 kb of proto-oncogenes compared to
matched random control sites (C). (D) Genomic heatmap analysis of the insertion site profile. mrc: matched randon control. ***P<0.001 by an unpaired t-test.

Gene therapy of DBA
A B
C D
Figure 8. Impaired erythroid differentiation of RPS19-deficient CD34+ cord blood cells can be rescued by the EFS-RPS19 vector. (A) RPS19 mRNA expression in
CD34+ cord blood cells transduced with shRNA. (B) Percentage of GFPhigh population in RPS19-deficient CD34+ cord blood cells treated or not with EFS-RPS19 during
erythroid differentiation from stage I to stage III. (C) Fluorescnce activated cell sorting analysis of erythroid differentiation of RPS19-deficient cells treated or not with
EFS-RPS19 on day 16. (D) Percentage of indicated cell outputs of GFPhigh populations on day 16. (E) Red blood cell pellets at the end of stage III initiated with equal
numbers of CD34+ cord blood cells (data shown as mean ± standard deviation, ^P<0.05 compared to the shRNA1 group, #P<0.05 compared to the shRNA2 group,
*P<0.05, **P<0.01, ***P<0.005 by a t-test, 3 independent experiments).
emergence of resident hematopoietic stem cells derived that the clinically applicable SIN lentiviral vector, EFS-
from the recipients, which contribute to the reconstitution RPS19, has the potential to be employed in a clinical gene
of hematopoiesis as a protective mechanism against therapy strategy for RPS19-deficient DBA patients.
stress-induced exhaustion in the BM.41-43 In support of this
concept, recipient-derived hematopoietic stem cells were Disclosures
observed in BM of the mock group at 16 weeks after trans- No conflicts of interest to disclose.
plantation, even though full lethal irradiation was per-
formed prior to transplantation. Other unknown reasons Contributions
may also contribute to this observation and the underlying SK and YL conceived the project and directed the research;
mechanism is unknown. YL, MD, MR, ES, THMG, SW, and JC performed the exper-
As mentioned before, the majority of DBA patients iments; YL, MR, AS, and SK analyzed the data; and YL and
have mutations in genes coding for ribosomal proteins, SK wrote the manuscript. Other co-authors provided feedback
and 25% of them are RPS19-deficient (mostly because of on the manuscript.
point mutations or small deletions). Our findings indicate
the possibility of developing SIN lentiviral vectors also tai- Acknowledgments
lored for other DBA mutations (e.g., RPL5) in the future. The authors thank Beata Lindqvist and Xiaojie Xian for
In conclusion, our data show the safety and efficacy of lentivirus production, Zhi Ma for technical assistance, and
a clinically applicable SIN lentiviral vector for the success- Alexander Doyle for English language editing.
ful treatment of Rps19-deficient DBA in our mouse model
and in human primary CD34+ cord blood cells. We did not Funding
observe any hematologic abnormalities in vivo due to This work was supported by a Hemato-Linne grant from the
enforced expression of RPS19. Our present study suggests Swedish Research Council Linnaeus, project grants from the

Y. Liu et al.
Swedish Research Council, the Swedish Cancer Society and the University Hospital (to SK), European Union project grants
Swedish Children’s Cancer Society (to SK), the Tobias Prize STEMEXPAND and PERSIST (to SK), a grant from The
awarded by the Royal Swedish Academy of Sciences financed Royal Physiographic Society of Lund, Sweden (to YL), and a
by the Tobias Foundation, a clinical research grant from Lund grant from Stiftelsen Lars Hiertas Minne (to YL).
References erythroid development in ribosomal protein populations. Biometrics. 1947;3(4):145-167.

S19-deficient Diamond-Blackfan anemia. 29. Lam YW, Lamond AI, Mann M, Andersen JS.
1. Lipton JM, Ellis SR. Diamond-Blackfan ane- Blood. 2002;100(8):2724-2731. Analysis of nucleolar protein dynamics
mia: diagnosis, treatment, and molecular 15. Jaako P, Flygare J, Olsson K, et al. Mice with reveals the nuclear degradation of ribosomal
pathogenesis. Hematol Oncol Clin North ribosomal protein S19 deficiency develop proteins. Curr Biol. 2007;17(9):749-760.
Am. 2009;23(2):261-282. bone marrow failure and symptoms like 30. Devlin EE, Dacosta L, Mohandas N, Elliott
2. Ulirsch JC, Verboon JM, Kazerounian S, et patients with Diamond-Blackfan anemia. G, Bodine DM. A transgenic mouse model
al. The genetic landscape of Diamond- Blood. 2011;118(23):6087-6096. demonstrates a dominant negative effect of
Blackfan anemia. Am J Hum Genet. 2018; 16. Naldini L. Ex vivo gene transfer and correc- a point mutation in the RPS19 gene associat-
103(6):930-947. tion for cell-based therapies. Nat Rev Genet. ed with Diamond-Blackfan anemia. Blood.
3. Vlachos A, Muir E. How I treat Diamond- 2011;12(5):301-315. 2010;116(15):2826-2835.
Blackfan anemia. Blood. 2010;116(19):3715- 17. Anguela XM, High KA. Entering the modern 31. Cesana D, Ranzani M, Volpin M, et al.
3723. era of gene therapy. Annu Rev Med. Uncovering and dissecting the genotoxicity
4. Willig TN, Niemeyer CM, Leblanc T, et al. 2019;70:273-288. of self-inactivating lentiviral vectors in vivo.
Identification of new prognosis factors from 18. Biffi A, Bartolomae CC, Cesana D, et al. Mol Ther. 2014;22(4):774-785.
the clinical and epidemiologic analysis of a Lentiviral vector common integration sites in 32. Deichmann A, Brugman MH, Bartholomae
registry of 229 Diamond-Blackfan anemia preclinical models and a clinical trial reflect a CC, et al. Insertion sites in engrafted cells
patients. DBA group of Societe benign integration bias and not oncogenic cluster within a limited repertoire of genom-
d'Hematologie et d'Immunologie selection. Blood. 2011;117(20): 5332-5339. ic areas after gammaretroviral vector gene
Pediatrique (SHIP), Gesellshaft fur 19. Cavazzana M, Bushman FD, Miccio A, therapy. Mol Ther. 2011;19(11):2031-2039.
Padiatrische Onkologie und Hamatologie Andre-Schmutz I, Six E. Gene therapy tar- 33. Da Costa L, O'Donohue MF, van
(GPOH), and the European Society for geting haematopoietic stem cells for inherit- Dooijeweert B, et al. Molecular approaches
Pediatric Hematology and Immunology ed diseases: progress and challenges. Nat to diagnose Diamond-Blackfan anemia: The
(ESPHI). Pediatr Res. 1999;46(5):553-561. Rev Drug Discov. 2019;18(6):447-462. EuroDBA experience. Eur J Med Genet.
5. Narla A, Vlachos A, Nathan DG. Diamond 20. Dull T, Zufferey R, Kelly M, et al. A third- 2018;61(11):664-673.
Blackfan anemia treatment: past, present, generation lentivirus vector with a condi- 34. High KA, Roncarolo MG. Gene therapy. N
and future. Semin Hematol. 2011;48(2):117- tional packaging system. J Virol. Engl J Med. 2019;381(5):455-464.
123. 1998;72(11):8463-8471. 35. Dunbar CE, High KA, Joung JK, Kohn DB,
6. Gripp KW, Curry C, Olney AH, et al. 21. Moreno-Carranza B, Gentsch M, Stein S, et Ozawa K, Sadelain M. Gene therapy comes
Diamond-Blackfan anemia with mandibulo- al. Transgene optimization significantly of age. Science. 2018;359(6372):eaan4672.
facial dystostosis is heterogeneous, including improves SIN vector titers, gp91phox 36. Marktel S, Scaramuzza S, Cicalese MP, et al.
the novel DBA genes TSR2 and RPS28. Am J expression and reconstitution of superoxide Intrabone hematopoietic stem cell gene ther-
Med Genet A. 2014;164A(9):2240-2249. production in X-CGD cells. Gene Ther. apy for adult and pediatric patients affected
7. Ludwig LS, Gazda HT, Eng JC, et al. Altered 2009;16(1):111-118. by transfusion-dependent ss-thalassemia.
translation of GATA1 in Diamond-Blackfan 22. Jaako P, Debnath S, Olsson K, et al. Gene Nat Med. 2019;25(2):234-241.
anemia. Nat Med. 2014;20(7):748-753. therapy cures the anemia and lethal bone 37. Da Costa L, Leblanc T, Mohandas N.
8. Sankaran VG, Ghazvinian R, Do R, et al. marrow failure in a mouse model of RPS19- Diamond-Blackfan anemia. Blood. 2020;136
Exome sequencing identifies GATA1 muta- deficient Diamond-Blackfan anemia. (11):1262-1273.
tions resulting in Diamond-Blackfan anemia. Haematologica. 2014;99(12):1792-1798. 38. Giri N, Kang E, Tisdale JF, et al. Clinical and
J Clin Invest. 2012;122(7):2439-2443. 23. Flygare J, Kiefer T, Miyake K, et al. laboratory evidence for a trilineage
9. Vlachos A, Ball S, Dahl N, et al. Diagnosing Deficiency of ribosomal protein S19 in haematopoietic defect in patients with
and treating Diamond Blackfan anaemia: CD34+ cells generated by siRNA blocks ery- refractory Diamond-Blackfan anaemia. Br J
results of an international clinical consensus throid development and mimics defects seen Haematol. 2000;108(1):167-175.
conference. Br J Haematol. 2008;142(6):859- in Diamond-Blackfan anemia. Blood. 39. Casadevall N, Croisille L, Auffray I, Tchernia
876. 2005;105(12):4627-4634. G, Coulombel L. Age-related alterations in
10. Boria I, Garelli E, Gazda HT, et al. The ribo- 24. Sherman E, Nobles C, Berry CC, et al. INSPI- erythroid and granulopoietic progenitors in
somal basis of Diamond-Blackfan anemia: IRED: a pipeline for quantitative analysis of Diamond-Blackfan anaemia. Br J Haematol.
mutation and database update. Hum Mutat. sites of new DNA integration in cellular 1994;87(2):369-375.
2010;31(12):1269-1279. genomes. Mol Ther Methods Clin Dev. 40. Engidaye G, Melku M, Enawgaw B.
11. Myers KC, Davies SM. Hematopoietic stem 2017;4:39-49. Diamond Blackfan anemia: genetics, patho-
cell transplantation for bone marrow failure 25. Berry CC, Nobles C, Six E, et al. INSPIIRED: genesis, diagnosis and treatment. EJIFCC.
syndromes in children. Biol Blood Marrow quantification and visualization tools for 2019;30(1):67-81.
Transplant. 2009;15(3):279-292. analyzing integration site distributions. Mol 41. Shi W, Vu T, Boucher D, et al. Ssb1 and Ssb2
12. Debnath S, Jaako P, Siva K, et al. Lentiviral Ther Methods Clin Dev. 2017;4:17-26. cooperate to regulate mouse hematopoietic
vectors with cellular promoters correct ane- 26. Chao A, Chazdon RL, Colwell RK, Shen TJ. stem and progenitor cells by resolving
mia and lethal bone marrow failure in a Abundance-based similarity indices and replicative stress. Blood. 2017;129 (18):2479-
mouse model for Diamond-Blackfan ane- their estimation when there are unseen 2492.
mia. Mol Ther. 2017;25(8):1805-1814. species in samples. Biometrics. 2006;62(2): 42. Baumgartner C, Toifl S, Farlik M, et al. An
13. Flygare J, Olsson K, Richter J, Karlsson S. 361-371. ERK-dependent feedback mechanism pre-
Gene therapy of Diamond Blackfan anemia 27. Haemmerle R, Phaltane R, Rothe M, et al. vents hematopoietic stem cell exhaustion.
CD34(+) cells leads to improved erythroid Clonal dominance with retroviral vector Cell Stem Cell. 2018;22(6):879-892.e6.
development and engraftment following insertions near the ANGPT1 and ANGPT2 43. Singh SK, Singh S, Gadomski S, et al. Id1
transplantation. Exp Hematol. 2008;36(11): genes in a human xenotransplant mouse ablation protects hematopoietic stem cells
1428-1435. model. Mol Ther Nucleic Acids. 2014;3: from stress-induced exhaustion and aging.
14. Hamaguchi I, Ooka A, Brun A, Richter J, e200. Cell Stem Cell. 2018;23(2):252-265.e8.
Dahl N, Karlsson S. Gene transfer improves 28. DeLury BD. On the estimation of biological

Complications in Hematology ARTICLE
Long term follow-up of pediatric-onset Evans Ferrata Storti Foundation

syndrome: broad immunopathological
manifestations and high treatment burden
Thomas Pincez,1,2 Helder Fernandes,1,3 Thierry Leblanc,4 Gérard Michel,5
Vincent Barlogis,5 Yves Bertrand,6 Bénédicte Neven,7,8,9 Wadih Abou Chahla,10
Marlène Pasquet,11 Corinne Guitton,12 Aude Marie-Cardine,13
Isabelle Pellier,14 Corinne Armari-Alla,15 Joy Benadiba,16 Pascale Blouin,17
Eric Jeziorski,18 Frédéric Millot,19 Catherine Paillard,20 Caroline Thomas,21
Nathalie Cheikh,22 Sophie Bayart,23 Fanny Fouyssac,24 Christophe Piguet,25 Haematologica 2022
Marianna Deparis,26 Claire Briandet,27 Eric Doré,28 Capucine Picard,9,29 Volume 107(2):457-466
Frédéric Rieux-Laucat,8,9 Judith Landman-Parker,30 Guy Leverger30 and
Nathalie Aladjidi1,3 on the behalf of members of the French Reference Center
for Pediatric Autoimmune Cytopenia (CEREVANCE) and of collaborators from
the French Reference Center for Adult Autoimmune Cytopenia (CERECAI).
1
Centre de Référence National des Cytopénies Auto-immunes de l’Enfant (CEREVANCE),
Bordeaux, France; 2Division of Pediatric Hematology-Oncology, Charles-Bruneau Cancer
Center, Department of Pediatrics, Sainte-Justine University Hospital, Université de
Montréal, Montréal, Québec, Canada; 3Pediatric Oncology Hematology Unit, University
Hospital, Plurithématique CIC (CICP), Centre d’Investigation Clinique (CIC) 1401,
INSERM, Bordeaux, France; 4Pediatric Hematology Unit, Robert Debré University
Hospital, AP-HP, Paris, France; 5Department of Pediatric Hematology, La Timone
Hospital, Marseille University Hospital, Marseille, France; 6Institute of Pediatric
Hematology and Oncology, Lyon University Hospital, Lyon, France; 7Pediatric Immuno-
Hematology and Rheumatology Department, Necker-Enfants Malades University
Hospital, AP-HP, Paris, France; 8Laboratory of Immunogenetics of Pediatric Autoimmune
Diseases, Paris, France; 9Imagine Institute, UMR 1163 INSERM, University of Paris,
Paris, France; 10Department of Pediatric Hematology, Jeanne de Flandre Hospital, Lille
University Hospital, Lille, France; 11Pediatric Oncology Immunology Hematology Unit,
Children’s University Hospital, Toulouse, France; 12Department of Pediatrics, Bicêtre
University Hospital, AP-HP, Le Kremlin-Bicêtre, France; 13Department of Pediatric
Hematology and Oncology, Rouen University Hospital, Rouen, France; 14Pediatric Unit,
Angers University Hospital, Angers, France; 15Pediatric Oncology Hematology Unit,
Grenoble University Hospital, Grenoble, France; 16Department of Hemato-Oncology
Pediatric, Nice University Hospital, Nice, France; 17Department of Pediatric Hematology-
Oncology, Clocheville Hospital, Tours University Hospital, Tours, France; 18Pediatric
Oncology Hematology Unit, Arnaud de Villeneuve University Hospital, Montpellier,
France; 19Department of Pediatric Hematology, Poitiers University Hospital, Poitiers, Correspondence:
France; 20Department of Pediatric Hematology and Oncology, Hautepierre University NATHALIE ALADJIDI
Hospital, Strasbourg, France; 21Pediatric Hematology Unit, Nantes University Hospital, nathalie.aladjidi@chu-bordeaux.fr
Nantes, France; 22Department of Pediatric Hematology-Oncology, Besançon University
Hospital, Besançon, France; 23Pediatric Hematology Unit, Rennes University Hospital,
Rennes, France; 24Pediatric Hematology Unit, Nancy University Hospital, Nancy, France; Received: August 31, 2020.
25
Pediatric Oncology Hematology Unit, Limoges University Hospital, Limoges, France;
26
Pediatric Oncology-Hematology Unit, Caen University Hospital, Caen, France; Accepted: December 22, 2020.
27
Department of Pediatrics, Dijon University Hospital, Dijon, France; 28Pediatric Unit, Pre-published: January 14, 2021.
Clermont-Ferrand University Hospital, Clermont-Ferrand, France; 29Study Center for
Primary Immunodeficiencies, Necker-Enfants Malades University Hospital, AP-HP, Paris,
France and 30Pediatric Oncology Immunology Hematology Unit, Armand-Trousseau https://doi.org/10.3324/haematol.2020.271106
University Hospital, AP-HP, Paris, France

ABSTRACT Material published in Haematologica is covered by copyright.
P
ediatric-onset Evans syndrome (pES) is defined by both immune throm- published material is allowed under the following terms and
conditions:
bocytopenic purpura (ITP) and autoimmune hemolytic anemia (AIHA) https://creativecommons.org/licenses/by-nc/4.0/legalcode.
before the age of 18 years. There have been no comprehensive long-term Copies of published material are allowed for personal or inter-
studies of this rare disease, which can be associated to various immunopatho- nal use. Sharing published material for non-commercial pur-
logical manifestations (IM). We report outcomes of the 151 patients with pES https://creativecommons.org/licenses/by-nc/4.0/legalcode,
and more than 5 years of follow-up from the nationwide French prospective sect. 3. Reproducing and sharing published material for com-
OBS’CEREVANCE cohort. Median age at final follow-up was 18.5 years mercial purposes is not allowed without permission in writing
from the publisher.
(range, 6.8–50.0 years) and the median follow-up period was 11.3 years
(range, 5.1–38.0 years). At 10 years, ITP and AIHA were in sustained complete

T. Pincez et al.
remission in 54.5% and 78.4% of patients, respectively. The frequency and number of clinical and biological IM
increased with age: at the age of 20 years, 74% had at least one clinical IM (cIM). A wide range of cIM occurred,
mainly lymphoproliferation, dermatological, gastrointestinal/hepatic and pneumological IM. The number of cIM
was associated with a subsequent increase in the number of second-line treatments received (other than steroids
and immunoglobulins; hazard ratio 1.4, 95% Confidence Interval: 1.15–1.60, P=0.0002, Cox proportional haz-
ards method). Survival at 15 years after diagnosis was 84%. Death occurred at a median age of 18 years (range,
1.7–31.5 years), and the most frequent cause was infection. The number of second-line treatments and
severe/recurrent infections were independently associated with mortality. In conclusion, long-term outcomes of
pES showed remission of cytopenias but frequent IM linked to high second-line treatment burden. Mortality
was associated to drugs and/or underlying immunodeficiencies, and adolescents-young adults are a high-risk
subgroup.
Introduction Methods
The presence of both immune thrombocytopenic pur- OBS’CEREVANCE prospective national cohort
pura (ITP) and autoimmune hemolytic anemia (AIHA) Inclusion and exclusion criteria are shown in the Online
defines Evans syndrome (ES). Pediatric-onset ES (pES) is a Supplementary Table S1.1,8,15 Data collected have been previously
rare disease, and approximately ten new cases are diag- detailed.1,8 Patients were included if <18 years old at first cytope-
nosed every year in France, which has a population of 66 nia diagnosis. The coordinating center gathered and analyzed all
million.1 Since its first description in 1951 by Robert data from the medical team in charge in real time, enabling
Evans,2 our understanding of pES has been based on small prospective follow-up even after the pediatric-to-adult transition.
retrospective cohorts with limited follow-up.3-7 In 2004, The CEREVANCE group recommends scheduling clinical and bio-
the French Rare Disease Center CEREVANCE set up the logical follow-up at least every 6-12 months. Some patients under-
prospective national cohort OBS’CEREVANCE, which went genetic analyses, as previously described.13 Written informed
includes children with AIHA, chronic ITP persisting for consent was obtained from parents and eligible patients. The
more than 12 months (cITP), and pES.8 cohort study was approved by the Institutional Ethics Committee
Preliminary reports from this cohort and previously (CPPRB-A; Bordeaux, France) and the database was registered
published studies showed that pES is a chronic disease with the national data protection authority (CNIL, 1396823V0).
with a high rate of relapse for both types of cytope-
nias.1,3,4,7,9 Mortality rates across studies have ranged from Patient selection
7–36%.1,3-7 In addition to cytopenias, immunopathologi- Patients with pES, defined as the simultaneous (within 1 month)
cal manifestations (IM) such as autoimmune/autoinflam- or sequential association of ITP and AIHA, were included if at
matory organ diseases, lymphoproliferation, and least 5 years of follow-up data were available after the first cytope-
hypogammaglobulinemia have been reported in 70–80% nia diagnosis. In order to provide a complete mortality report, all
of patients with pES.4,5,8,10 In an undetermined number of patients, including those with less than 5 years of follow-up data,
cases, pES is thought to be “secondary” and caused by an were included in the survival analyses. The data were extracted on
underlying disease, classically systemic lupus erythe- 21 June 2019.
matosus (SLE) or autoimmune lymphoproliferative syn-
drome (ALPS).1,11,12 Recently, genetic analyses found a Definitions
heterogeneous genetic background in up to 65% of a sub- Initial cytopenia refers to the onset of ITP or AIHA (whichever
set of 80 patients from the OBS’CEREVANCE cohort. occurred first) and does not take autoimmune neutropenia (AIN)
These patients carried variants in genes that are linked to into account. The IM categories were separated in clinical (cIM)
primary immunodeficiencies (PID) or involved in and biological (bIM). pES was defined as secondary if a diagnosis
immune responses.13 of SLE or PID was made during the follow-up period. SLE diagno-
Overall, outcomes and the long-term course of pES are sis was made according to the Systemic Lupus International
poorly understood. There have been no comprehensive Collaborating Clinics Classification criteria (SLICC).16 ALPS diag-
longitudinal studies including both cytopenia and IM. In nosis was based on international criteria.17 Second-line treatments
addition, the transition to adulthood is often particularly were all immunomodulatory or immunosuppressive treatments,
challenging for patients with chronic pediatric diseases.14 including splenectomy but excluding steroids and therapeutic
Adolescents–young adults (AYA) outcomes have not intravenous immunoglobulins (IVIG). Sustained complete remis-
been investigated in patients with pES, and whether the sion (CR) was defined as remission persisting until final follow-up,
disease improves with age is unknown. In a clinical set- regardless of ongoing treatments.
ting, the possibility to identify high-risk patients would For analyses by age, patients were assessed annually from birth
be extremely helpful in the management of this complex (for IM and treatments) or from cytopenia onset (for AIHA and
disease. Here, we describe the long-term course of hema- ITP), until final follow-up. Occasional treatments (e.g., splenecto-
tological IM and treatments received throughout child- my and rituximab) were considered as ongoing if these occurred
hood into adulthood in patients with pES from the during the previous year. Further details are stated in the Online
OBS’CEREVANCE cohort. We aimed to identify clinical- Supplementary Table S1.
ly relevant factors associated to the occurrence of IM, the
number of second-line treatments received and mortali- Statistical analyses
ty. Particularly, we investigated the impact of the number Continuous and categorical variables were compared using
second-line treatment received and splenectomy on mor- Wilcoxon–Mann–Whitney non-parametric test and Fisher’s exact
tality. test, respectively. Correlations were tested using the Pearson cor-

Pediatric-onset Evans syndrome outcomes
relation coefficient. Survival and cumulative incidence estimates Table 1. Patient characteristics.
were based on the Kaplan–Meier method and compared using Number of patients 151
log-rank test. Patients of OBS’CEREVANCE cohort with isolated Sex ratio (male/female) 1.40 (88/63)
cITP or AIHA were used for comparison in survival analyses
Consanguinity, n (%) 12 (7.9)
(unpublished data). The Cox proportional hazards method was
used to analyze factors associated with time-dependent variables cIM/cancer in first-degree relative, n (%) 43 (28)
(i.e., time to CR, AIN, cIM, and second-line treatment, as well as Median age (years) at
survival). The potential cumulative and/or time-dependent nature First cytopenia (min-max) 5.4 (0.2-16.0)
of variables was taken in account. Proportionality of hazard was ITP diagnosis (min-max) 6.7 (0.2-17.1)*
assessed for each variable. Logistic regression was used to analyze
factors associated with severe or recurrent infections. Variables AIHA diagnosis (min-max) 7.8 (0.2-21.5)*
that were statistically significant in the univariate analyses were ES diagnosis (min-max) 8.9 (0.2-21.5)
included in the multivariate analyses. We investigated associations Sequence of ES
with the following characteristics and events: sex, consanguinity, Simultaneous, n (%) 52 (34.4)
cIM/cancer in a first-degree relative, age at first and second
ITP then AIHA, n (%) 62 (41.1)
cytopenia, sequence of cytopenia, AIN, hypogammaglobulinemia,
time to ITP/AIHA CR, severe/recurrent infections, number of cIM, AIHA then ITP, n (%) 37 (24.5)
number of second-line treatments. The 95% Confidence Intervals Time between first and second cytopenia (years)
(CI) for hazard ratios (HR) and odds ratios (OR) were not adjusted Median (min-max)** 2.5 (0.1-15.8)
for multiple testing and should not be used to infer definitive Direct antiglobulin test at AIHA diagnosis
effects. All tests were two-sided and a P-value <0.05 was consid-
IgG, n (%) 73 (48.3)
ered statistically significant. Statistical analyses were performed
using R (ver. 4.0; R Development Core Team) and GraphPad Prism IgG + C3, n (%) 61 (40.4)
(ver. 8; GraphPad Software, Inc., San Diego, CA, USA) software. Unspecified, n (%) 11 (7.3)
C3, n (%) 4 (2.6)
IgA + C3, n (%) 1 (0.7)
Results
IgM then IgG, n (%) 1 (0.7)
Population Duration of follow-up after first cytopenia (years)
Of the 216 patients with pES, 151 were included in this Median (min-max) 11.3 (5.1-38.0)
study (Online Supplementary Figure S1). They were fol- Mean ± SD 12.5 ± 6.0
lowed from 25 different centers. Patient characteristics are
Age at last follow-up (years)
shown in Table 1. The median (min–max) follow-up time
from the first cytopenia diagnosis was 11.3 years (range, Median (min-max) 18.5 (6.8-50.0)
5.1–38.0 years). Median age at final follow-up was 18.5 Mean ± SD 19.1 ± 6.8
years (range, 6.8–50.0 years). In 20 cases (15%), follow-up *P=0.0076. **Considering the 99 patients with sequential cytopenias. CIM: clinical
was discontinued because the patient was considered immunopathological manifestation; Ig: immunoglobulin; SD: standard deviation; ITP:
immune thrombocytopenic purpura; AIHA: autoimmune hemolytic anemia; ES: Evans
cured (n=11) or lost to follow-up (n=9). Median age at loss syndrome.
to follow-up was 18.4 years (range, 6.8-25.1 years).
patients (34%) had no treatment ongoing at last follow-
Hematological outcomes up. No particular characteristic was associated with AIHA
AIN developed in 43 patients (28.5%). It was diagnosed or ITP CR, including cIM and bIM.
within 1 month before or after first cytopenia onset in 23 Over the first three decades, the proportions of patients
of 43 cases (53.5%), more than 1 month before in two achieving sustained CR increased with age (Figure 1B). ITP
cases (4.7%), and more than 1 month after in 18 cases and AIHA were in CR in 26% and 30% of cases at 10
(41.9%; maximal delay, 12.4 years). In all cases, the diag- years compared to 50% and 72% at 20 years, respectively
nosis was made before the age of 18 years (median age, (P<0.001 for both comparisons).
6.8 years; range, 0.6–16.2 years).
ITP and AIHA flare rates at 5 years of follow-up were Immunopathological manifestations
calculated for the 61 alive patients who did not receive a A total of 122 of 151 patients (81%) had at least one IM.
second-line treatment during this period. Forty-eight The data for each category and specific diagnosis are
patients (79%) had experienced an ITP flare and seven shown in the Online Supplementary Table S2.
(11%) an AIHA flare. cIM developed in 100 of 151 patients (66%). A total of
The proportion of patients achieving sustained CR for 47 patients (31%) had two or more IM and 22 (15%)
ITP and AIHA steadily increased after cytopenia onset patients had three or more IM (Online Supplementary Figure
(Figure 1A). At 5 and 10 years, ITP was in sustained CR in S2A). Patients with no cIM had shorter median follow-up
40.5% and 62.3% of patients (P=0.02) and AIHA was in times (9.7 years vs. 13 years; P=0.0002) and were younger
sustained CR in 54.5% and 74.1% of patients (P=0.001), when data were collected (15 years vs. 20 years;
respectively. Sustained CR was achieved earlier for AIHA P<0.0001). A cIM was diagnosed before the first cytopenia
than ITP (median time to CR, 4.0 years vs. 7.0 years; in 21 of 100, simultaneously in 13, and after in 66 cases
P=0.01). At the final follow-up of the 135 surviving (median delay, 3.7 years [range, 0.2–20.5 years]; Figure
patients, the numbers of patients in CR, partial remission, 2A). Among the 185 cIM, 29 (16%) were diagnosed before
and no remission were 126 (83%), five (3%), and one (1%) any second-line treatment. No cIM category had a statisti-
for AIHA and 119 (79%), eight (5%), and five (3%) for ITP, cally significant difference in frequency before and after
respectively (missing data in three cases). Forty-six first second-line treatment. The number of cIM increased

T. Pincez et al.
with age. At 10 compared to 20 years old, 37% and 74% the most frequently diagnosed bIM (n=54), including 44
of patients had at least one cIM and 9% and 34% of cases diagnosed prior to any anti-CD20 treatment. Among
patients had at least two cIM, respectively (P<0.001 for those 54 patients, 25 (46%) received immunoglobulin
both comparisons; Figure 2B). replacement therapy. SLE and ALPS biomarkers were
The most common cIM categories were lymphoprolifer- present (regardless of whether patients met the diagnostic
ation (n=71), dermatological (n= 26), gastrointestinal/hepat- criteria) in 42 and 24 patients, respectively. At 10 and 20
ic (n=23) and pneumological manifestations (n=16, Figure 3; years of age, 39% and 75% of patients had at least one
Online Supplementary Figure S2B). The most frequent cIM bIM, respectively (P<0.001).
diagnosis are shown in Table 2. Thirteen patients devel- Patients with bIM were more likely to have cIM (79%
oped a neurological manifestation as previously described.10 vs. 40%; P<0.001), and patients with cIM were more likely
Four patients had a hematological malignancy (age at diag- to have bIM (80% vs. 41%; P<0.001) but the correlation
nosis): Hodgkin lymphoma (16 years), juvenile between the number of bIM and cIM was low (r=0.27;
myelomonocytic leukemia (20 years), large granular lym- P<0.001).
phocytic leukemia (21 years) and angioimmunoblastic T-
cell lymphoma (29 years). Older age at ES diagnosis (HR Secondary pediatric-onset Evans syndrome
1.09; 95% CI: 1.01–1.17; P=0.02), cIM/cancer in a first- In 37 patients (24.5%), pES eventually revealed a SLE or
degree relative (HR 1.64; 95% CI: 1.1–2.4; P=0.006), and the a PID unknown at cytopenia onset.
presence of AIN were independently associated with the Eleven patients (7.3%) eventually met the SLE SLICC
number of cIM (HR 2.41; 95% CI: 1.5–3.8; P=0.0002). diagnostic criteria.16 These patients were older at first
Biological IM (bIM) were diagnosed in 101 of 151 cytopenia (median age 13 years vs. 5 years; P=0.007) and
patients (67%), and the frequency of bIM also increased almost exclusively female (one of 88 males [1%] and ten
with the age (Figure 2C). Hypogammaglobulinemia was of 63 females [16%]); P<0.001).
A
Figure 1. Hematological outcomes.
(A) Cumulative incidence of patients
achieving a sustained complete
remission (CR). Among the 23
patients without sustained CR for
autoimmune hemolytic anemia
(AIHA), four (17%) had achieved sus-
tained CR for immune thrombocy-
topenic purpura (ITP). Conversely,
among the 32 patients without sus-
tained CR for ITP, 13 (40%) had
achieved sustained CR for AIHA. (B)
Percentage of patients with a sus-
tained complete remission according
to age.

Seven patients (4.6%) met the diagnostic criteria for pathogenic variant in CTLA4 (n=7), TNFRSF6 (germline
ALPS after pES onset which prompted targeted genetic n=6, somatic n=1), STAT3 (n=5), PIK3CD (n=1), CBL
analysis.17 Overall, 66 of 151 patients (44%) underwent (n=1), and KRAS (somatic n=1) or a homozygous/com-
genetic analyses as previously described.13 Among them, pound heterozygous pathogenic variants in LRBA (n=3)
26 (39%) patients were considered to have a PID (includ- and RAG1 (n=1). Compared to the 40 other patients, the
ing the seven with ALPS). They carried a heterozygous 26 with a PID had more cIM (2 [range, 1-5] vs. 1 [range,
C
Figure 2. Immunopathological manifestations.
(A) Age at first clinical immunopathological man-
ifestation (cIM) diagnosis and at first cytopenia.
Pearson correlation coefficient r=0.42,
P<0.0001. There was no difference in the medi-
an age at first cIM and at first cytopenia in terms
of the number of cIM (data not shown). (B)
Cumulative incidence of cIM according to age.
Half of the patients had developed a cIM by the
age of 13.5 years and a second IM by the age of
27 years. (C) Cumulative incidence of any bio-
logical IM (bIM), as well as each category. Half
of the patients had at least one bIM diagnosed
by 13.2 years of age. The biological workup was
not standardized and was made at the clini-
cian’s discretion. SLE: systemic lupus erythe-
matosus; ALPS: autoimmune lymphoprolifera-
tive syndrome.

T. Pincez et al.
Figure 3. Immunopathological manifestations and other associated manifestations. Individual occurrence of autoimmune neutropenia, clinical immunopathological
manifestations (cIM), biological IM (bIM), atopy, severe or recurrent infections, and malignancies. Each column represents a patient. The patients are ordered accord-
ing to their cIM, from the most (lymphoproliferation) to the least (hematological, other) frequent. Hypoγ: hypogammaglobulinemia; SLE: systemic lupus erythemato-
sus; ALPS: autoimmune lymphoproliferative syndrome.
Table 2. Most frequent clinical immunopathological manifestations diagnosis.

cIM n (%) cIM n (%)
Superficial (palpable) adenopathies 61 (40) Lymphoid enteropathy 5 (3)
Splenomegaly 49 (33) Chronic gastritis 5 (3)
Deep (abdominal or thoracic) adenopathies 16 (11) Polyarthritis 5 (3)
Granulomatous–lymphocytic interstitial lung disease 16 (11) Vitiligo 4 (3)
Cutaneous lupus erythematosus involvement 8 (5) Eczema 4 (3)
Autoimmune hepatitis 7 (5) Keratitis 4 (3)
Subtentorial inflammatory lesions 7 (5) Uveitis 4 (3)
Diagnosis present in at least four patients are shown and ordered by frequency. Complete diagnosis list is provided in the Online Supplementary Table S2. cIM: clinical
immunopathological manifestation.
0-4], P=0.008) and a trend toward more bIM as shown by new first and/or second-line treatments and 31% had
Wilcoxon–Mann–Whitney sum ranks comparison but received a new second-line treatment (Online
same medians (1 [range, 0-3] vs. 1 [range, 0-2], P=0.029). Supplementary Figure S3C).
There was no statistically significant difference in the The number of second-line treatments received
median time to ITP CR (4.7 years vs. 8.0 years, P=0.26) increased with age, particularly after the first decade
and to AIHA CR (5.5 years vs. 5.5 years, P>0.9), the num- (Figure 4A). At 10 and 20 years, 47% and 88% of patients
ber of second-line treatment received (3 [range, 0-9] vs. 2 had received a second-line treatment, respectively
[range, 0-6]; P=0.057) and mortality (two of 26 [7.7%] vs. (P<0.001). The number of patients receiving ongoing treat-
three of 40 [7.5%]; P>0.9). ments also increased with age (Figure 4B). At 10 and 20
years, 27% and 69% of patients had received an active
Treatments second-line treatment, respectively (P<0.001). At the final
All except two patients (98.6%) had received at least follow-up, patients with a cIM had received more second-
one first-line treatment course. Second-line treatments line treatments (median, 3 vs. 1; P<0.0001).
(regardless of the hematological and/or extra-hematologi- The most frequently used second-line treatments were
cal indication) were required in 117 of 151 (77%) patients rituximab (n=79; 52%), azathioprine (n=55; 36%),
(Online Supplementary Figure 3A). Patients who did not splenectomy (n=36; 24%), and mycophenolate (n=29;
receive any second-line treatment had shorter median fol- 19%; Online Supplementary Table S3).
low-up times (10.5 years vs. 12.3 years; P=0.017). The The number of cIM was associated with a subsequent
median number of second-line treatments received was increase in the number of second-line treatments received
two (range, 0–9). (HR 1.4; 95% CI: 1.15–1.60; P=0.0002). On the contrary,
The number of second-line treatments received the number of second-line treatment was not associated
increased with the time elapsed since first cytopenia with- to a subsequent increase in the number of cIM in univari-
out reaching a plateau (Online Supplementary Figure 3B). ate analysis (HR 1.09; 95% CI: 0.98–1.22; P=0.11).
After a sustained CR for both ITP and AIHA achieved, the
number of treatments received had continued to increase: Infections
at 5 years after CR of both cytopenias, 67% of patients In total, 53 (35%) patients had severe or recurrent infec-
who achieved CR for both ITP and AIHA had received a tions (Online Supplementary Table S4). The most frequent

Figure 4. Second-line treatments. (A) Total number of second-line treatments received according to the age. (B) Number of second-line treatments ongoing according
to age.
were herpes zoster (n=17), sinusitis/otitis media (n=15), higher than those in patients with cITP or AIHA alone
pneumopathy (n=12), and bronchiectasis (n=11). Patients (P<0.0001 for both comparisons).
with infections had more cIM (median, 2 vs. 1; P<0.0001), Deaths occurred regularly throughout the follow-up
a higher incidence of hypogammaglobulinemia (53% vs. period (median delay after first cytopenia diagnosis, 8.9
28%; P=0.003), and received more second-line treatments years [range, 0.1–24.3 years]) and at a median age of 18.0
(median 3 vs. 1; P<0.0001). Among the 16 patients with years (range, 1.7–31.5 years) (Figure 5B). In the majority of
severe infection, nine (63%) were receiving an active treat- these patients, cytopenia was under control at the time of
ment at infection time. death: 15 (65%) and 19 (83%) patients had CR or partial
Severe/recurrent infections were independently associ- remission from ITP and AIHA, respectively (Figure 5C).
ated with hypogammaglobulinemia (OR 2.4; 95% CI: Mortality was linked to the disease, the treatment, or both
1.10–5.33; P=0.03) and the number of second-line treat- in eight (36%), two (9%), and twelve (55%) cases, respec-
ments (OR 1.34; 95% CI: 1.13–1.71; P=0.002). tively. The most frequent cause of death was infections
(n=12 [52%]; Online Supplementary Table S5). Four patients
Mortality (18%) died of a hemorrhage, and all were less than 13
Sixteen of the 151 patients followed for more than 5 years of age. The patients who died from a hemorrhage
years (10.6%) died, and seven other patients died before were younger than those who died from an infection
the fifth year of follow-up (23 deaths in total, 22 with (median 10 years vs. 18 years; P=0.03). All of these
available data). Patient survival at 5, 10, and 15 years after patients, except for one who died in the first month of a
the first cytopenia was 97%, 92%, and 84%, respectively cerebral hemorrhage, had at least one cIM. Eight of the
(Figure 5A). Mortality rates in patients with pES were patients (36%) had hypogammaglobulinemia.

T. Pincez et al.
A
Figure 5. Long-term survival. (A)
Survival estimate of patients in terms of
time from first cytopenia. At 10-year fol-
low-up, survival rates among patients
with chronic immune thrombocytopenic
purpura (ITP) alone, autoimmune
hemolytic anemia (AIHA) alone and
pediatric-onset Evans syndrome (pES)
were 100%, 99% and 92%, respectively.
(B) Mortality is shown in terms of time
from first cytopenia, as well as age.
Individual values are shown as dots with
medians and interquartile ranges
shown as lines. (C) Hematological sta-
tus at death. CR: complete remission;
PR: partial remission; NR: no remission.
The patients who died had received more second-line patients. Conversely, clinical and biological IM increased
treatments than the others in the cohort (median 3 vs. 2; in frequency and number with increasing patient age,
P=0.02), including splenectomy, which was more com- finally affecting almost all adult patients. The number of
mon in this subgroup (56% vs. 20%; P=0.003). Patients cIM was associated with a subsequent increase in the
who had received more than two second-line treatments number of second-line treatments received. Mortality was
had a three-fold increase in the risk of death compared to high, frequently occurred while cytopenias were in remis-
those who had received two or less (11 of 65 [16.9%] vs. sion, and most deaths concerned AYA. Two characteristics
five of 86 [5.8%], P=0.03). At death, 81% of patients were were associated to mortality: severe or recurrent infec-
receiving ongoing second-line treatment. The number of tions and the number of second-line treatments received.
second-line treatments (HR 1.3; 95% CI: 1.1–1.6; P=0.004) Overall, the age-related clinical picture showed similar
and severe/recurrent infections (HR 3.4; 95% CI: 1.2–9.7; trends for all patients, shifting from cytopenia to increased
P=0.02) were independently associated with a higher risk IM, a greater treatment burden, and an increased risk of
of mortality after 5 years of follow-up. mortality.
In setting up a nationwide cohort, the CEREVANCE
group tried to ensure unbiased patient inclusion in this
Discussion study. Omitting patients with less than 5 years of follow-
up data limited any bias due to short-term follow-up,
This large follow-up study of pES patients included which probably accounts for many of the discrepancies
more than 1,900 patient-years. Over the long term, AIHA between previous studies. Indeed, our median follow-up
and ITP were sustainably controlled in the majority of period was more than twice as long as in previous studies

(median 4.8 years [range, 3–7 years]).1,3–7 However, clusions regarding their efficacy. We were unable to inves-
although the trends reported here are clear, some factors tigate the risk associated to specific treatments given the
may also influence the estimates. The loss to follow-up high heterogeneity in second-line treatment combinations
mainly concerned AYA and few patients were followed and duration as well as the changes in management prac-
after the age of 20 years. As well, the CEREVANCE group tices since the cohort onset in 2004. The rapid initial
recommends clinical and biological follow-up at least increase in second-line treatments is partly due to the high
every 6-12 months but local practice or patients’ pheno- rate of early relapse and the current practice of administer-
type (such as the presence of cIM) may have influenced ing steroid-sparing agents to treat pES.28 However, the
biological testing. presence of cytopenia is not the only reason for using
Sustained CR was eventually achieved for both types of these drugs and first- and second-line treatments were
cytopenia in the vast majority of patients, although this also used after CR of both cytopenias. cIM are important
often took many years, especially for ITP (>10 years for in determining the number of second-line treatments
one-third of our patients). Because active treatments are used, but bIM may also play a role, particularly in patients
used to treat most AYA (notably because of cIM), hemato- with SLE biomarkers, who are frequently given hydroxy-
logical CR may be drug induced and it is impossible to chloroquine. Nevertheless, second-line treatments are
determine whether an underlying hematological autoim- rarely selected based on a single factor. Patients with pES
munity is still present. The higher rate of sustained CR in often have bIM and cIM, and the whole clinical picture
ITP among patients with pES compared to patients with needs to be assessed before selecting a treatment strategy.
cITP alone may be due to more patients with pES receiv- As previously reported,13 approximately one-third of
ing treatment.18 patients may carry alterations in genes that are potentially
One of the most striking findings in this study was the accessible to targeted therapy.29–31 Given the high burden
progressive increase in the frequency and number of IM. of second-line treatments and their association with infec-
A range of cIM, affecting almost every organ, were identi- tions and mortality, the CEREVANCE network has pro-
fied and developed independently of cytopenias. These posed implementing genetic analyses for all patients with
findings clearly show that pES is a marker for a more gen- pES to limit the use of immunosuppressive and toxic
eral tendency toward immunodeficiencies while we can- drugs.
not exclude a contribution of the second-line treatments Comprehensively, the pES clinical picture changes as
received to some IM. The underlying etiology is not com- patients age. From 10 to 20 years of age, cytopenia tends
pletely understood and may vary among patients, with to be controlled but IM are more prevalent, and active sec-
both genetic and environmental factors being important. ond-line treatments are used in more than two-thirds of
Consequently, pES may be considered a composite syn- patients during the pediatric-to-adult transition. Overall,
drome with several overlapping subgroups of secondary as patients age, the illness becomes more severe and the
pES. One of these subgroups includes patients with PID. risk of mortality increases. Both IM and treatment burden
Classically, ALPS has been associated with pES.12 In this contribute to the infection-related mortality peak
study, only 4% of patients were diagnosed with ALPS observed at the end of the second decade. The patients
based on well-defined criteria, despite evocative biological who died had received more second-line treatments,
“ALPS-like” abnormalities in a larger proportion of including splenectomy. Because these two parameters are
patients.17 This observation is consistent with our previ- correlated (r=0.60; P<0.0001), the number of deaths was
ous study,13 which showed that more immune-response too low to determine whether splenectomy alone was a
genes are potentially involved in pES than initially sus- risk factor of mortality per se or a marker of severity.
pected.13,19–21 However, pES rarely comports as a In conclusion, pES must now be considered a complex
Mendelian disease,1 and some of these variants may be multi-systemic disease in which cytopenias frequently
predisposing rather than disease-causing alleles. Even in present fewer challenges than IM and infections in long-
patients carrying a variant in a monogenic PID gene (e.g., term follow-up. Adult patients with pES form a specific
TNFRFS6 or CTLA4),13,22 the altered genes show incom- subgroup, distinct from older adults with ES.32
plete penetrance.23,24 We were unable to evaluate the pro- Multidisciplinary follow-up of patients with pES is needed
portion of patients who met common variable immunod- and must focus on IM screening, genetic diagnosis, infec-
eficiency disorders diagnostic criteria,25 as vaccine tions prevention, patient-tailored drugs development, and
responses were not evaluable in all cases due to second- AYA management. Specifically, the infection burden may
line treatments received. A second subgroup includes be reduced by ensuring up-to-date vaccinations, eradicat-
patients with SLE, although the prevalence of this sub- ing chronic infections, and using adequate antimicrobial
group is controversial.11,26,27 Our cohort suggests that SLE prophylaxis or immunoglobulin replacements. As in sever-
eventually occurs almost exclusively within the known at- al chronic pediatric diseases,33 dedicated child-to-adult
risk population of female adolescents and is frequent in transition programs are warranted to improve outcomes
this subgroup, as it developed in seven of 15 (47%) of the in patients with pES.
females >12 years old.26 Despite its heterogeneity, the
course of pES, in terms of age-related changes and trends, Disclosures
was similar for the majority of patients. The spectrum of No conflicts of interest to disclose.
IM described here is probably influenced by the underly-
ing etiology, and further analyses are needed to under- Contributions
stand the determinant of IM. The long-term follow-up of TP, HF, TL, GL and NA designed the study, analyzed the
the present study confirms that the subgroup of patients data and drafted the paper; TP and HF performed statistical
with identified PID had more cIM.13 analyses; CP and FR-L performed genetic analyses. All of the
Most patients required second-line treatments. These authors participated to prospective data collection and interpreta-
treatments reflect local practices and we cannot draw con- tion and revised the manuscript for critical content.

T. Pincez et al.
Acknowledgments Bordelaise pour l’Avancement des Sciences Pédiatriques

The list of collaborators is given in the Online Supplementary (ABASP) research charity, the Association pour la Recherche et
Appendix. The authors would like to thank all of the patients, les Maladies Hématologiques de l’Enfant (RMHE) research
families, medical and para-medical teams involved in the CERE- charity, the Association Française du Syndrome d’Evans (AFSE),
VANCE prospective cohort study from 2004 onwards. the O-CYTO patients’ association, and partially by
GlaxoSmithKline, AMGEN and Novartis. TP is a recipient of a
Funding Charles Bruneau fellowship award.
This work was supported from 2004 by the French Ministry of
Health (Programme Hospitalier de Recherche Clinique [PHRC] Data sharing statement
2005, Rare Disease Plan 2007 and 2017), the Association Data are available on request to corresponding author.
References (ALPS). Blood. 2005;105(6):2443-2448. 23. Schwab C, Gabrysch A, Olbrich P, et al.

13. Hadjadj J, Aladjidi N, Fernandes H, et al. Phenotype, penetrance, and treatment of
1. Aladjidi N, Fernandes H, Leblanc T, et al. Pediatric Evans syndrome is associated 133 CTLA-4-insufficient individuals. J.
Evans syndrome in children: long-term out- with a high frequency of potentially dam- Allergy Clin Immunol. 2018;142(6):1932-
come in a prospective French national aging variants in immune genes. Blood. 1946.
observational cohort. Front Pediatr. 2019;134(1):9-21. 24. Neven B, Magerus-Chatinet A, Florkin B, et
2015;3:79. 14. White PH, Cooley WC, Group TCRA, et al. al. A survey of 90 patients with autoim-
2. Evans RS, Takahashi K, Duane RT, Payne R, Supporting the health care transition from mune lymphoproliferative syndrome relat-
Liu C. Primary thrombocytopenic purpura adolescence to adulthood in the medical ed to TNFRSF6 mutation. Blood. 2011;
and acquired hemolytic anemia; evidence home. Pediatrics. 2018;142(5):e20182587. 118(18):4798-4807.
for a common etiology. AMA Arch Intern 15. Rodeghiero F, Stasi R, Gernsheimer T, et al. 25. Bonilla FA, Barlan I, Chapel H, et al.
Med. 1951;87(1):48-65. Standardization of terminology, definitions International Consensus Document
3. Mathew P, Chen G, Wang W. Evans syn- and outcome criteria in immune thrombo- (ICON): common variable immunodefi-
drome: results of a national survey. J cytopenic purpura of adults and children: ciency disorders. J. Allergy Clin Immunol
Pediatr. Hematol Oncol. 1997;19(5):433- report from an international working Pract. 2016;4(1):38-59.
437. group. Blood. 2009;113(11):2386-2393. 26. Tarvin SE, O’Neil KM. Systemic lupus ery-
4. Pui CH, Wilimas J, Wang W. Evans syn- 16. Petri M, Orbai A-M, Alarcón GS, et al. thematosus, Sjögren syndrome, and mixed
drome in childhood. J. Pediatr. Derivation and validation of Systemic connective tissue disease in children and
1980;97(5):754–758. Lupus International Collaborating Clinics adolescents. Pediatr Clin North Am. 2018;
5. Savaşan S, Warrier I, Ravindranath Y. The Classification Criteria for systemic lupus 65(4):711-737.
spectrum of Evans’ syndrome. Arch Dis erythematosus. Arthritis Rheum. 27. Lube GE, Ferriani MPL, Campos LMA, et
Child. 1997;77(3):245-248. 2012;64(8):2677-2686. al. Evans syndrome at childhood-onset sys-
6. Wang WC. Evans syndrome in childhood: 17. Oliveira JB, Bleesing JJ, Dianzani U, et al. temic lupus erythematosus diagnosis: a
pathophysiology, clinical course, and treat- Revised diagnostic criteria and classifica- large multicenter study. Pediatr Blood
ment. Am J Pediatr Hematol Oncol. tion for the autoimmune lymphoprolifera- Cancer. 2016;63(7):1238-1243.
1988;10(4):330-338. tive syndrome (ALPS): report from the 2009 28. Miano M. How I manage Evans Syndrome
7. Blouin P, Auvrignon A, Pagnier A, et al. NIH International Workshop. Blood. and AIHA cases in children. Br J Haematol.
Syndrome d’Evans : étude rétrospective de 2010;116(14):e35-40. 2016;172(4):524-534.
la société d’hématologie et d’immunologie 18. Ducassou S, Gourdonneau A, Fernandes H, 29. Lee S, Moon JS, Lee C-R, et al. Abatacept
pédiatrique (36 cas). Arch Pédiatrie. et al. Second-line treatment trends and alleviates severe autoimmune symptoms in
2005;12(11):1600-1607. long-term outcomes of 392 children with a patient carrying a de novo variant in
8. Aladjidi N, Leverger G, Leblanc T, et al. chronic immune thrombocytopenic purpu- CTLA-4. J. Allergy Clin Immunol.
New insights into childhood autoimmune ra: the French experience over the past 25 2016;137(1):327-330.
hemolytic anemia: a French national obser- years. Br J Haematol. 2020;189(5):931-942. 30. Lo B, Zhang K, Lu W, et al. Patients with
vational study of 265 children. 19. Rieux-Laucat F, Le Deist F, Fischer A. LRBA deficiency show CTLA4 loss and
Haematologica. 2011;96(5):655-663. Autoimmune lymphoproliferative syn- immune dysregulation responsive to abata-
9. Mannering N, Hansen DL, Frederiksen H. dromes: genetic defects of apoptosis path- cept therapy. Science. 2015;349(6246):436-
Evans syndrome in children below 13 years ways. Cell Death Differ. 2003;10(1):124- 440.
of age – a nationwide population-based 133. 31. Klemann C, Esquivel M, Magerus-Chatinet
cohort study. PoS One. 2020; 20. Schubert D, Bode C, Kenefeck R, et al. A, et al. Evolution of disease activity and
15(4):e0231284. Autosomal dominant immune dysregula- biomarkers on and off rapamycin in 28
10. Pincez T, Neven B, Le Pointe HD, et al. tion syndrome in humans with CTLA4 patients with autoimmune lymphoprolifer-
Neurological involvement in childhood mutations. Nat Med. 2014;20(12):1410- ative syndrome. Haematologica.
Evans syndrome. J Clin Immunol. 2019; 1416. 2017;102(2):e52-e56.
39(2):171-181. 21. Flanagan SE, Haapaniemi E, Russell MA, et 32. Michel M, Chanet V, Dechartres A, et al.
11. Costallat GL, Appenzeller S, Costallat LTL. al. Activating germline mutations in STAT3 The spectrum of Evans syndrome in adults:
Evans syndrome and systemic lupus ery- cause early-onset multi-organ autoimmune new insight into the disease based on the
thematosus: clinical presentation and out- disease. Nat Genet. 2014;46(8):812-814. analysis of 68 cases. Blood. 2009;
come. Joint Bone Spine. 2012;79(4):362- 22. Besnard C, Levy E, Aladjidi N, et al. 114(15):3167-3172.
364. Pediatric-onset Evans syndrome: heteroge- 33. Gabriel P, McManus M, Rogers K, White P.
12. Teachey DT, Manno CS, Axsom KM, et al. neous presentation and high frequency of Outcome evidence for structured pediatric
Unmasking Evans syndrome: T-cell pheno- monogenic disorders including LRBA and to adult health care transition interven-
type and apoptotic response reveal autoim- CTLA4 mutations. Clin Immunol. tions: a systematic review. J Pediatr. 2017;
mune lymphoproliferative syndrome 2018;188:52-57. 188:263-269.

Iron Metabolism & its Disorders ARTICLE
Risk factors for endocrine complications in Ferrata Storti Foundation

transfusion-dependent thalassemia patients
on chelation therapy with deferasirox: a risk
assessment study from a multi-center
nation-wide cohort
Maddalena Casale,1 Gian Luca Forni,2 Elena Cassinerio,3 Daniela Pasquali,4
Raffaella Origa,5 Marilena Serra,6 Saveria Campisi,7 Angelo Peluso,8 Roberta
Renni,9 Alessandro Cattoni,10 Elisa De Michele,11 Massimo Allò,12 Maurizio Haematologica 2022
Poggi,13 Francesca Ferrara,14 Rosanna Di Concilio,15 Filomena Sportelli,16 Volume 107(2):467-477
Antonella Quarta,17 Maria Caterina Putti,18 Lucia Dora Notarangelo,19 Antonella
Sau,20 Saverio Ladogana,21 Immacolata Tartaglione,1 Stefania Picariello,1 Alessia
Marcon,3 Patrizia Sturiale,22 Domenico Roberti,1 Antonio Ivan Lazzarino23 and
Silverio Perrotta1
1
Department of Woman, Child and General and Specialized Surgery, University of
Campania Luigi Vanvitelli, Naples, Italy; 2Center of Microcitemia and Congenital
Anemias, Galliera Hospital, Genoa, Italy; 3Rare Diseases Center, General Medicine Unit,
IRCCS Ca’ Granda Ospedale Maggiore Policlinico Section co, Milan, Italy; 4Endocrinology
Unit, Department of Advanced Medical and Surgical Sciences, University “ Luigi
Vanvitelli”, Naples, Italy; 5Thalassemia Center, Pediatric Hospital A CAO, AOG Brotzu,
Cagliari, Italy; 6Thalassemia Center, Department of Internal Medicine, Hospital "V. Fazzi",
Lecce, Italy; 7Thalassemia Center, Hospital Umberto I, Siracusa, Italy; 8Center of
Microcitemia, POC SS.Annunziata - ASL TA, Taranto, Italy; 9Thalassemia Center,
Department of Internal Medicine, Hospital F.Ferrari, Casarano, Italy; 10Department of
Pediatrics, Università degli Studi di Milano Bicocca, Fondazione Monza e Brianza per il
Bambino e la sua Mamma, Azienda Ospedaliera San Gerardo, Monza, Italy;
11
Immunotransfusion Medicine Unit, AOU OO.RR. S. Giovanni di Dio e Ruggi d'Aragona,
Salerno, Italy; 12Center of Microcitemia, Hospital ASL 5, Crotone, Italy; 13Department of
Endocrinology, Sant'Andrea Hospital, Rome, Italy; 14Department of Internal Medicine,
Policlinico Hospital of Modena, Modena, Italy; 15Department of Pediatrics, Hospital
Umberto I, Nocera, Italy; 16Immunotransfusion Unit, Hospital Riuniti, Foggia, Italy;
17
Center for Microcythemia, Iron Metabolism Disorders, Gaucher Disease - Hematology
and Transplantation Unit, "A. Perrino" Hospital, Brindisi, Italy; 18Department of Women's
and Child's Health (DSDB), University Hospital, Padova, Italy; 19Hematology Oncology
Unit, Children's Hospital, ASST Spedali Civili, Brescia, Italy; 20Department of Pediatric
Hematology and Oncology, Hospital “Spirito Santo”, Pescara, Italy; 21Pediatric
Oncohematology Unit, “Casa Sollievo della Sofferenza” Hospital, IRCCS, San Giovanni Correspondence:
Rotondo, Italy; 22SSD Microcitemia Center, G.O.M Reggio Calabria, Reggio Calabria, Italy MADDALENA CASALE
and 23EPISTATA – Agency for Clinical Research and Medical Statistics, London, UK maddalena.casale@unicampania.it
ABSTRACT Received: September 18, 2020.

T
ransfusion-dependent patients typically develop iron-induced car-
diomyopathy, liver disease, and endocrine complications. We aimed Pre-published: January 7, 2021.
to estimate the incidence of endocrine disorders in transfusion-
dependent thalassemia (TDT) patients during long-term iron-chelation https://doi.org/10.3324/haematol.2020.272419
therapy with deferasirox (DFX). We developed a multi-center follow-up
study of 426 TDT patients treated with once-daily DFX for a median dura-
tion of 8 years, up to 18.5 years. At baseline, 118, 121, and 187 patients had ©2022 Ferrata Storti Foundation
0, 1, or ≥2 endocrine diseases respectively. 104 additional endocrine dis- Material published in Haematologica is covered by copyright.
eases were developed during the follow-up. The overall risk of developing published material is allowed under the following terms and
a new endocrine complication within 5 years was 9.7% (95% Confidence conditions:
Interval [CI]: 6.3–13.1). Multiple Cox regression analysis identified three https://creativecommons.org/licenses/by-nc/4.0/legalcode.
Copies of published material are allowed for personal or inter-
key predictors: age showed a positive log-linear effect (adjusted hazard nal use. Sharing published material for non-commercial pur-
ratio [HR] for 50% increase 1.2, 95% CI: 1.1–1.3, P=0.005), the serum con- poses is subject to the following conditions:
centration of thyrotropin showed a positive linear effect (adjusted HR for https://creativecommons.org/licenses/by-nc/4.0/legalcode,
sect. 3. Reproducing and sharing published material for com-
1 mIU/L increase 1.3, 95% CI: 1.1–1.4, P<0.001) regardless the kind of dis- mercial purposes is not allowed without permission in writing
ease incident, while the number of previous endocrine diseases showed a from the publisher.
negative linear effect: the higher the number of diseases at baseline the
lower the chance of developing further diseasess (adjusted HR for unit

M. Casale et al.
increase 0.5, 95% CI: 0.4–0.7, P<0.001). Age and thyrotropin had similar effect sizes across the categories of
baseline diseases. The administration of levothyroxine as a covariate did not change the estimates. Although
in DFX-treated TDT patients the risk of developing an endocrine complication is generally lower than the
previously reported risk, there is considerable risk variation and the burden of these complications remains
high. We developed a simple risk score chart enabling clinicians to estimate their patients’ risk. Future
research will look at increasing the amount of variation explained from our model and testing further clinical
and laboratory predictors, including the assessment of direct endocrine magnetic resonance imaging.
Introduction We considered the following endocrine conditions:

1. Hypothyroidism (overt: thyrotropin [TSH] >10 mU/mL and
Transfusion-induced iron overload in thalassemia low free thyroxine [FT4]; subclinical: TSH 5–10 mU/mL and nor-
patients typically results in iron-induced cardiomyopathy, mal FT4).
liver disease, and endocrine complications. However, those 2. Hypoparathyroidism (low parathyroid hormone [PTH] and
three phenomena have been studied to different extents. calcium and high phosphorus).
In transfusion-dependent thalassemia (TDT) patients, 3. Hypogonadism (hypogonadotropic hypogonadism, in adult
mortality due to cardiovascular and hepatic complications female: amenorrhea, low estradiol levels and low or normal
has markedly declined during the last decades.1–3 The devel- luteinizing hormone and follicle stimulating hormone [LH/FSH]
opment of magnetic resonance imaging techniques (MRI), levels; in adult male: low testosterone levels, clinical signs or
specifically designed to quantify myocardial and hepatic symptoms consistent with hypogonadism and low/normal
iron concentration, measuring heart T2* and liver iron con- LH/FSH. Testosterone reference ranges vary according to patients’
centration (LIC), has enabled the design of clinical trials age at the time of biochemical assessment. In general, they were
evaluating the efficacy of iron chelators in targeting specific regarded as normal if >3.5 ng/mL and unequivocally pathological
iron overload.4 Moreover, new anti-hepatitis C drugs have <2.3 ng/mL. Additional data [clinical features, free testosterone]
remarkably reduced the complications linked to hepatitis C were taken into account for values between 2.3 and 3.5 ng/mL.
infection, which used to dramatically deteriorate liver iron Hypergonadotropic hypogonadism, in adult female: amenorrhea
overload.5 and raised FSH [>30 U/L] with undetectable estradiol; in adult
However, in spite of the outstanding advances in the care male: raised gonadotropins with low total testosterone and clinical
of cardiovascular and hepatic complications due to blood signs consistent with hypogonadism).
transfusions, the management of endocrine complications 4. Pubertal disturbances (delayed puberty: lack of breast bud-
has been left behind and, nowadays, they are the most fre- ding [Tanner stage 2] in girls by the age of 13 and testicular volume
quent and the most resource-draining complications in <4 mL in boys by the of 14; arrested puberty: lack of pubertal pro-
TDT patients.3 In addition, serological testing fails to iden- gression over a year or more).
tify high-risk groups and, once occurred, these complica- 5. Disorders of glucose metabolism (diabetes: fasting plasma
tions are often irreversible. While MRI imaging of glucose ≥126 mg/dL or 2-hour plasma glucose [2-h PG] value dur-
endocrine glands is promising in detecting preclinical dis- ing a 75-g oral glucose tolerance test [OGTT] >200 mg/dL;
ease, it has not reached the level of validation required for impaired fasting glucose [IFG]: fasting glucose between 100 and
routine clinical use.6 125 mg/dL; impaired glucose tolerance: 2-h PG during 75-g OGTT
The once-daily oral iron chelator deferasirox (DFX) was levels between 140 and 199 mg/dL).
shown to be effective in chelating iron from the heart and 6. Bone metabolism disorder (BMD) (osteoporosis: bone miner-
the liver, with preservation of the heart function,7–9 and with al density T score ≤-2.5 and Z score value ≤-2; osteopenia: T score
reversal of the hepatic fibrosis.10 While the effective control value 1.01/-2.5 and Z score value 1.01/-2. In childhood, osteoporo-
of heart and liver siderosis remains the primary goal in the sis was defined by either the association of at least two patholog-
management of TDT patients, observational data suggest ical fractures by the age of 10 years/ three by the age of 19 and Z
that iron loading in endocrine organs may precede myocar- score ≤-2 or by the finding of at least one vertebral crush, in the
dial involvement and there is now substantial evidence on absence of high-energy trauma or local disease, irrespectively of
the role of iron overload in endocrine morbidity.11–14 While the BMD recorded; low bone mineral density was defined as the
there have been small studies on endocrine disorders in finding of BMD Z score ≤-2, in the absence of the above-men-
TDT patients during chelation therapy with DFX,15,16 the tioned additional criteria for osteoporosis).
data are still scarce, even though DFX is nowadays the most According to standardized protocols,18 laboratory tests for
prescribed drug for iron chelation in TDT patients.17 detection of endocrine disorders were performed every year in
The aim of this study was to assess the incidence of patients with no endocrine complications and more frequently
endocrine diseases including hypothyroidism, (every 3-6 months) in patients with endocrine disorders, as per
hypoparathyroidism, glucose metabolisms disorders, consolidated clinical practice and according to the endocrinolo-
hypogonadism, and metabolic bone disease in patients suf- gists’ prescription. Routine laboratory tests, such as glycemia and
fering from TDT who are on treatment with the drug DFX. serum electrolytes, were assessed every 1-3 months in occasion of
pre-transfusion cross-match testing. Weight, height and Tanner
stage were assessed every 6 months in patients <18 years of age.
Methods The study protocol was approved by Ethical Committees and
Institutional Review Boards of all the participating centers and
In this multi-center study, TDT patients from 21 hospitals locat- was conducted in accordance with the Declaration of Helsinki and
ed in 21 cities and 19 regions of Italy were assessed for eligibility ICH guidelines for good clinical practice. All patients provided
to be recruited in the cohort. written informed consent.

Endocrine complications in iron overload
Table 1. Cross-tabulation of number of conditions at baseline and number of conditions occurred during the follow-up.
N. of endocrine diseases N. of new endocrine diseases occurred Total
at baseline during follow-up
0 1 2 3
0 75(63.6%) 32(27.1%) 10(8.5%) 1(0.9%) 118 (100%)
(23.3%) (36.4%) (66.7%) (100.0%) (27.7%)
1 87(71.9%) 30(24.8%) 4(3.3%) 0(0.0%) 121 (100%)
(27.0%) (34.1%) (26.7%) (0.0%) (28.4%)
2 86(80.4%) 21(19.6%) 0(0.0%) 0(0.0%) 107 (100%)
(26.7%) (23.9%) (0.0%) (0.0%) (25.1%)
3 59(90.8%) 5(7.7%) 1(1.5%) n/a 65 (100%)
(18.3%) (5.7%) (6.7%) (15.3%)
4 15(100.0%) 0 n/a n/a 15 (100%)
(4.7%) (0.0%) (3.5%)
Total 322(75.6%) 88(20.7%) 15(3.5%) 1(0.2%) 426 (100%)
(100%) (100%) (100%) (100%) (100%)
Values represent the total number of patients in each category, with row and column percentages in parentheses.
A detailed description of all methods used is available in the rary or permanent DFX discontinuation. The most fre-
Online Supplementary Appendix. quent AE were related to gastrointestinal intolerance (epi-
gastralgia, heartburn, abdominal pain; n=8) and increased
Statistical analysis transaminases (n=8). Increased in serum creatinine (n=1)
Data were cleaned before the analysis: we checked all variables and Lichen planus (n=1) were reported as other AE which
for missing, illogical or implausible values, also through cross- caused DFX interruption. In nine (2%) cases DFX was dis-
checks with related variables (e.g., chronologic orders). continued because of treatment failure, reported as
Continuous variables were checked for abnormal distributions increase in serum ferritin (n=6), cardiac T2* (n=2), LIC
and outliers. We used Cox-regression to fit survival analyses with (n=1). In one case treatment failure was reported along
follow-up days as the underlying time variable. Survival time was with gastrointestinal intolerance.
measured as the number of days passed from the beginning of the Table 1 shows a cross-tabulation between the number
treatment regimen with the drug DFX to the first of either the of endocrine diseases at baseline and the number of new
diagnosis of the first new endocrine disease, side effect due to endocrine diseases that occurred during the follow-up.
DFX leading to therapy suspension, death, or censoring. We cen- The 75.6% of the total sample (322 of 426) did not devel-
tered the covariates before interaction analyses. We adopted two op any new endocrine disease during the follow-up (95%
strategies for the development of the multiple models: we either Confidence Interval [CI]: 71.2–79.6). Out of the 104
started with the covariates having higher biological plausibility of (24.4%) with newly diagnosed endocrinopathies, 84.6%
effect or with those with a lower P-value from at the bivariate developed only one endocrine disease (95% CI: 76.2–
stage. The two approaches reached the same final model. The 90.9). Out of 118 patients with no endocrine diseases at
assumption of proportional hazards was checked by using the baseline, 43 (36.4%) developed at least one endocrine dis-
Schoenfeld residuals test. We computed the proportion of varia- ease during the follow-up (95% CI: 27.8–45.8). Out of
tion explained by the models (adjusted R2) using the Royston 121 patients having one endocrine disease at baseline
method with bootstrap confidence intervals (5,000 replications).19 already, 34 (28.1%) developed at least one additional
We derived the predicted probabilities of developing a new endocrine disease during the follow-up (95% CI: 20.3–
endocrine disease within 5 years and 1 year by using the margins 37.0).
command in Stata v.14. Two patients were excluded from the Among the 118 patients with no endocrine diseases at
analysis as they already had all possible five endocrine diseases at baseline, BMT disorders occurred the most (17.8% [95%
baseline. CI:11.4–25.9]), followed by hypogonadism (12.7% [95%
CI=7.3–20.1]). Those two conditions were also the most
prevalent ones in patients with one disease at a baseline
Results (80.2% [95% CI: 71.9–86.9] and 11.6% [95% CI: 6.5–
18.7] respectively) and were those that most likely
Out of 426 patients enrolled, accounting for 3,517 per- occurred as additional diseases during the follow-up.
son per years, 104 participants developed at least one Figure 1 shows the overall crude risks for all 104 first
new endocrine disease after a mean and median follow- incidents, by incident type and age group. It appears that
up time of 8 years (range, 1 month–18.5 years). The mean most of the new incidents occurred after the age of 20
iron intake at baseline was 0.28+/-0.08 mg/kg/day (range, years, with a new spike between 35 and 45 years. As for
0.14–0.49 mg/kg/day) and at the end of study was 0.26+/- pediatric patients, the increase seems to start after the age
0.12 mg/kg/day (range, 0.16–0.50 mg/kg/day). The mean of 12 years. No cases of insulin-dependent diabetes were
hemoglobin level was 9.8+/-0.68 g/dL (range, 9.4–10.6 reported in patients with no endocrine disorders at base-
g/dL) indicating the majority of patients had good control line (Figure 2). Kaplan-Meier survival probability curves
of their chronic anemia. with age as the underlying time variable are reported in
No deaths were recorded. Overall, 18 (4%) patients the Online Supplementary Appendix.
experienced adverse events (AE) that determined tempo- Tables 2A and 2B show a description of the sample by

M. Casale et al.
prevalent endocrine diseases at baseline and by incident Tables 3A and 3B show the results from the multiple
endocrine diseases during the follow-up. Age, TSH, and Cox regression models. In both models, the adjusted haz-
low BMD were associated with both prevalent and inci- ard rate of developing a new endocrine disorder
dent disorders. Among the markers of iron overload, fer- decreased by about 50% for each prevalent endocrine dis-
ritin and T2* were associated with prevalent but not with ease at baseline (P<0.001) and increased by about 25%
incident disorders, whereas LIC was not associated with for each mIU/L of TSH at baseline (P<0.001). The two
any of them. During the follow-up, iron overload test models differ in the way in which the variable age was
results either decreased by a small extent or remained sta- treated. In model 3A age was treated as a log-linear vari-
ble over time, while the standard deviation of those dif- able whereas in model 3B age was treated as a linear vari-
ferences was more than three times their means. The able, but in that case, also a binary indicator was included
number of prevalent endocrine disorders was inversely for pediatric/adult patient and an interaction term
associated with the incidence of a new disorder. Only 11 between age and the indicator as well. The latter model
patients (2.6%) had a side-effect related to DFX adminis- showed a higher adjusted R2 (0.25 vs. 0.22) although that
tration (gastrointestinal disorder). difference was not significant (95% CI: 0.19–0.42 vs. 95%
Figure 1. Overall crude risks for all 104 first

incidents (n=426), by incident type and age
group.
Figure 2. Crude risks for all 43 first incidents

in patients with no endocrinopathies at base-
line (n=118), by incident type and age group.

Table 2A. Sample description stratified by number of endocrine diseases at baseline.

Factors and categories* Number of endocrine diseases at baseline P-value
0 1 2+
N=118 N=121 N=187
Age (years) 9.1 (5.4-23.2) 28.9 (18.6-36.7) 34.7 (29.6-39.7) <0.001
Pediatric patient (<16 years) 68.6% (81/118) 17.4% (21/121) 3.7% (7/187) <0.001
Age if child (n=109) 6.9 (3.7) 10.9 (3.5) 10.9 (3.9) <0.001
Age if adult (n=317) 31.3 (11.9) 32.0 (10.5) 35.6 (7.3) 0.001
Sex male 50.8% (60/118) 49.6% (60/121) 38.5% (72/187) 0.054
Splenectomised 23.7% (28/118) 47.9% (58/121) 75.4% (141/187) <0.001
Used drugs other than DFX in the past 27.1% (32/118) 18.2% (22/121) 36.4% (68/187) 0.002
Heart disease 4.4% (5/114) 7.8% (9/116) 19.5% (34/174) <0.001
Thyroid disorder 0.0% (0/118) 6.6% (8/121) 35.8% (67/187)
Parathyroid disorder 0.0% (0/118) 1.7% (2/121) 7.0% (13/187)
Gonadal disorder 0.0% (0/118) 11.6% (14/121) 88.2% (165/187)
Glucose metabolism disorder 0.0% (0/118) 0.0% (0/121) 26.7% (50/187)
BMD 0.0% (0/118) 80.2% (97/121) 93.0% (174/187)
Ferritin (ng/mL) 1342.1 (×/2.0) 937.2 (×/2.3) 844.9 (×/2.2) <0.001
Ferritin >2,000 (ng/mL) 28.4% (29/102) 18.3% (21/115) 15.0% (26/173) 0.023
LIC (mg Fe/g dry weight) 4.9 (×/2.4) 4.0 (×/2.4) 3.9 (×/2.2) 0.18
LIC 0.34
<3 28.6% (20/70) 43.2% (41/95) 37.4% (55/147)
3- 38.6% (27/70) 27.4% (26/95) 38.1% (56/147)
7- 22.9% (16/70) 22.1% (21/95) 19.7% (29/147)
15+ 10.0% (7/70) 7.4% (7/95) 4.8% (7/147)
Heart T2* (msec) 36.0 (11.3) 35.7 (9.4) 30.5 (12.0) <0.001
EF (%) 64.0 (6.6) 62.9 (5.8) 64.4 (6.8) 0.33
TSH (mIU/L) 2.5 (×/1.5) 2.1 (×/1.6) 2.1 (×/2.1) 0.061
On levotyroxine 0.0% (0/104) 5.2% (6/115) 27.8% (49/176) <0.001
FT4 (pmol/L) 14.8 (3.3) 14.2 (2.9) 14.5 (4.6) 0.54
TSH index 2.8 (0.7) 2.6 (0.7) 2.7 (0.9) 0.39
PTH (pg/mL) 25.0 (×/1.6) 27.3 (×/1.8) 21.9 (×/2.3) 0.19
Glycemia (mg/dL) 84.5 (×/1.1) 84.2 (×/1.1) 95.1 (×/1.3) <0.001
Calcium (mg/dL) 9.4 (9.1-9.7) 9.3 (8.9-9.6) 9.3 (9.0-9.7) 0.25
Phosphorus (mg/dL) 4.3 (3.5-5.0) 4.1 (3.5-4.7) 4.0 (3.4-4.5) 0.17
BMD femur (g/cm2) 0.9 (0.8-1.0) 0.7 (0.6-0.9) 0.7 (0.6-0.8) <0.001
BMD femur (z score) -0.4 (1.5) -1.5 (1.1) -2.0 (1.0) <0.001
BMD femur (t score) -0.5 (1.5) -1.6 (0.9) -2.1 (1.0) <0.001
BMD L1-L4 (g/cm2) 1.0 (0.9-1.1) 0.8 (0.7-0.9) 0.8 (0.7-0.9) <0.001
BMD L1-L4 (z score) -0.8 (-1.4--0.3) -2.1 (-2.9--1.3) -2.5 (-3.2--1.9) <0.001
BMD L1-L4 (t score) -0.8 (-1.4--0.3) -2.2 (-3.0--1.3) -2.8 (-3.4--2.0) <0.001
New endocrine disease occurred 36.4% (43/118) 28.1% (34/121) 14.4% (27/187) <0.001
Thyroid disorder occurred 5.1% (6/118) 8.8% (10/113) 5.0% (6/120) 0.39
Parathyroid disorder occurred 0.8% (1/118) 1.7% (2/119) 4.0% (7/174) 0.18
Gonadal disorder occurred 12.7% (15/118) 12.1% (13/107) 18.2% (4/22) 0.74
Glucose metabolism disorder occurred 3.4% (4/118) 2.5% (3/121) 5.8% (8/137) 0.36
BMD 24.6% (29/118) 41.7% (10/24) 23.1% (3/13) 0.22
Side effect occurred 0.8% (1/118) 4.1% (5/121) 2.7% (5/187) 0.28
D Ferritin† -476.6 (1519.6) -310.8 (1419.9) -430.6 (1103.1) 0.62
D LIC† -1.4 (5.2) -1.6 (6.6) -1.8 (6.6) 0.93
D T2*† 0.7 (13.8) 1.9 (11.5) 5.9 (12.8) 0.008
continued on following page

M. Casale et al.
continued from previous page
D TSH† -0.1 (1.4) 0.1 (1.5) 0.5 (5.1) 0.41

D TSH index† -0.0 (0.8) 0.0 (0.8) 0.2 (0.9) 0.092
D BMD femur (g/cm2)† -0.0 (0.1) 0.0 (0.3) 0.0 (0.2) 0.74
D BMD L1-L4 (g/cm2)† -0.2 (0.7) 0.0 (0.1) 0.0 (0.2) 0.007
For normally-distributed variables, data are presented as mean (standard deviation [SD]) with P-value from ANOVA. For log-normal variables, data are presented as geometric
mean (×/geometric SD) with P-value from ANOVA on logged values. For continuous variables with other types of distributions, data are presented as median (interquartile range
[IQR]) with P-value from Kruskal-Wallis test. For categorical variables, data are presented as % (n/total) with P-value from Pearson's chi-squared test. *Measured at baseline,
unless otherwise specified. †Intra-individual difference between measures taken at the end and at the beginning of the follow-up. DFX: deferasirox ; BMD: bone metabolism dis-
order; LIC: liver iron concentration; EF: ejection fractions; FT4: free thyroxine; TSH: thyrotropin; PTH: parathyroid hormone.
Table 2B. Sample description stratified by categories of outcome measure.

Factors and categories* New incident P-value
No Yes
N=322 N=104
Age (years) 30.5 (18.2-38.3) 27.0 (11.8-35.4) 0.029
Pediatric patient (<16 years) 23.3% (75/322) 32.7% (34/104) 0.056
Age if child (n=109) 7.5 (4.0) 8.9 (3.7) 0.082
Age if adult (n=317) 34.2 (8.6) 33.2 (11.1) 0.42
Sex male 43.2% (139/322) 51.0% (53/104) 0.16
Splenectomised 54.7% (176/322) 49.0% (51/104) 0.32
Used drugs other than DFX in the past 28.9% (93/322) 27.9% (29/104) 0.84
Heart disease 10.7% (33/309) 15.8% (15/95) 0.18
Thyroid disorder 19.9% (64/322) 10.6% (11/104) 0.030
Parathyroid disorder 3.1% (10/322) 4.8% (5/104) 0.41
Gonadal disorder 46.3% (149/322) 28.8% (30/104) 0.002
Glucose metabolism disorder 14.0% (45/322) 4.8% (5/104) 0.012
Bone metabolism disorder 70.8% (228/322) 41.3% (43/104) <0.001
N. of endocrine diseases at baseline <0.001
0 23.3% (75/322) 41.3% (43/104)
1 27.0% (87/322) 32.7% (34/104)
2+ 49.7% (160/322) 26.0% (27/104)
Ferritin (ng/mL) 979.3 (×/2.2) 995.8 (×/2.4) 0.86
Ferritin >2,000 (ng/mL) 18.5% (55/297) 22.6% (21/93) 0.39
LIC (mg Fe/g dry weight) 4.1 (×/2.4) 4.2 (×/2.1) 0.85
LIC 0.51
<3 37.3% (90/241) 36.6% (26/71)
3- 33.2% (80/241) 40.8% (29/71)
7- 22.8% (55/241) 15.5% (11/71)
15+ 6.6% (16/241) 7.0% (5/71)
Heart T2* (msec) 32.9 (10.7) 34.1 (13.5) 0.46
EF (%) 64.0 (6.0) 63.2 (7.9) 0.41
TSH (mIU/L) 2.1 (×/1.9) 2.5 (×/1.6) 0.019
On levotyroxine 15.8% (47/298) 8.2% (8/97) 0.063
FT4 (pmol/L) 14.5 (4.0) 14.5 (3.2) 0.92
TSH index 2.7 (0.8) 2.8 (0.8) 0.24
PTH (pg/mL) 25.1 (×/2.0) 21.8 (×/2.1) 0.26
Glycemia (mg/dL) 89.3 (×/1.3) 86.8 (×/1.2) 0.29
Calcium (mg/dL) 9.3 (9.0-9.6) 9.3 (9.0-9.8) 0.41
Phosphorus (mg/dL) 4.1 (3.5-4.7) 4.0 (3.4-4.7) 0.54
BMD femur (g/cm2) 0.7 (0.6-0.8) 0.8 (0.7-0.9) 0.050
BMD femur (z score) -1.7 (1.1) -1.5 (1.3) 0.32
BMD femur (t score) -1.9 (1.1) -1.5 (1.3) 0.079
BMD L1-L4 (g/cm2) 0.8 (0.7-0.9) 0.9 (0.8-1.0) 0.019
BMD L1-L4 (z score) -2.3 (-3.0--1.6) -1.8 (-2.6--1.0) 0.020
BMD L1-L4 (t score) -2.6 (-3.3--1.7) -1.8 (-2.8--1.0) 0.003
continued on following page

continued from previous page
Side effect 2.8% (9/322) 1.9% (2/104) 0.63

D Ferritin† -405.3 (1263.5) -413.4 (1481.1) 0.96
D LIC† -1.7 (6.6) -1.7 (4.9) 0.97
D T2*† 3.7 (12.1) 3.4 (14.7) 0.85
D TSH† 0.2 (4.0) 0.3 (1.9) 0.89
D TSH index† 0.1 (0.8) 0.2 (0.9) 0.29
D BMD femur (g/cm2)† 0.0 (0.2) -0.0 (0.2) 0.23
D BMD L1-L4 (g/cm2)† 0.0 (0.2) -0.1 (0.5) 0.009
For normally-distributed variables, data are presented as mean (standard deviation [SD]) with P-value from ANOVA. For log-normal variables, data are presented as geometric
mean (×/geometric SD) with P-value from ANOVA on logged values. For continuous variables with other types of distributions, data are presented as median (interquartile range
[IQR]) with P-value from Kruskal-Wallis test. For categorical variables, data are presented as % (n/total) with P-value from Pearson's chi-squared test. *Measured at baseline,
unless otherwise specified. †Intra-individual difference between measures taken at the end and at the beginning of the follow-up. DFX: deferasirox ; BMD: bone metabolism dis-
order; LIC: liver iron concentration; EF: ejection fractions; FT4: free thyroxine; TSH: thyrotropin; PTH: parathyroid hormone.
Table 3A. Risk factors for developing a new endocrine disease during
CI: 0.10–0.37). According to model 3A, each 50% the follow-up: results from the simplest multiple Cox regression model.
increase in age was associated with an increase of about
18% in the hazard of an incident new disease (P=0.005) Variable at the beginning Mutually-adjusted (95% CI) P
of follow-up hazard ratio
after having adjusted for TSH and number of previous
endocrine conditions. Kaplan-Meier survival probability Diseases at baseline (1 increase) 0.53 (0.43 0.66) <0.001
curves are reported in the Online Supplementary Appendix. TSH (1 mIU/L increase) 1.25 (1.13 1.38) <0.001
Table 4A shows the 5-year risk predictions according to Age (50% increase) 1.18 (1.05 1.33) 0.005
levels of age, TSH, and number of endocrine diseases at
CI: Confidence Interval; TSH: thyrotropin.
baseline, based on estimates from model 3B. On average,
the whole cohort of patients had a risk of 9.7% (95% CI:
Table 3B. Risk factors for developing a new endocrine disease during
6.3–13.1) of developing an additional endocrine disease the follow-up: results from the multiple Cox regression model showing
within 5 years from the start of therapy with the drug the highest adjusted R2, which was used to draw the risk charts.
DFX. However, there was considerable variation accord- Variable at the beginning Mutually-adjusted (95% CI) P
ing the baseline conditions. For example, an average 14- of follow-up hazard ratio
year-old patient with a TSH of 3 mIU/L who already suf-
fered from one endocrine disorder had a risk of develop- Diseases at baseline (1 increase) 0.54 (0.43 0.67) <0.001
ing another disorder within 5 years of about a 10%, TSH (1 mIU/L increase) 1.26 (1.15 1.39) <0.001
whereas a 35-year-old patient with a TSH of 5 mIU/L and Age (5-year increase) 1.12 (1.00 1.26) 0.053
no disease at baseline had a risk of 50%. Table 4B shows
Child vs. Adult 6.70 (1.32 34.02) 0.022
the 1-year risk predictions according to levels of age, TSH
and number of endocrine diseases at baseline, based on Interaction Age*Child 1.59 (1.02 2.47) 0.041
estimates from model 3B. The overall 1-year risk was CI: Confidence Interval; TSH: thyrotropin.
1.1% (95% CI: 0.6–1.7).

Fifty-five patients were on therapy with levothyroxine ed with any marker of iron overload (Spearman's rhos
at the beginning of the follow-up. We carried out a sensi- <0.07, P>0.18).
tivity analysis by running the same analysis on patients
who were and were not on levothyroxine separately to
see if levothyroxine modified the estimates from the final Discussion
models. In patients who were (n=55) and were not
(n=371) on levothyroxine, the adjusted hazard ratio [HR] Endocrine complications remain the most common and
for 1 mIU/L increase in TSH was 1.26 (95% CI: 1.02–1.55, resource-consuming disorders secondary to iron overload in
P=0.032) and 1.29 (95% CI: 1.07–1.56, P=0.006) respec- TDT patients. In historical cohorts, disturbances of sexual
tively. The estimates from the other predictors did not axis affected 80% of patients, while BMD and short stature
change either. Therefore, TSH was a predictor of addi- were reported in up to 60% and 50% of the overall study
tional endocrine disease incidence regardless of levothy- population, respectively. Prevalence of hypothyroidism and
roxine administration. In addition, we conducted strati- diabetes ranged from 6% to 14%, while hypoparathy-
fied analyses after splitting the sample at the median fol- roidism was reported up to 25%.18
low-up time, or at the age of 16 years, or at 0/1+ preva- In our long-term cohort study of patient affected by TDT
lent endocrine diseases at baseline. The results from those treated with the iron-chelating drug DFX, the risk of devel-
subgroup analyses were similar to the main one. oping an endocrine complication is generally lower than the
Given that chronic iron overload is supposed to be the previously reported risk, but there is considerable risk vari-
main driver of endocrine complications due to blood ation, according to several parameters such as patient’s age,
transfusions, we have not only used the baseline markers number of endocrine complications already present before
of iron overload (ferritin, LIC, and T2*) in our predictive the start of the therapy, and TSH serum concentration. We
models, but we have also tested the latest available meas- developed a simple risk chart enabling clinicians to derive
ures and the difference between initial and final meas- an approximate estimate of their patients’ risk.
ures. In no cases had those markers any effect on the inci- Ferritin, LIC and cardiac T2* are considered as markers of
dence of endocrine complications. TSH was not correlat- iron overload, but the correlation between those markers

M. Casale et al.
and risk of endocrine complications is controversial,20 since heart, instead, develop pathologic iron overload exclusively
many studies have shown no correlation,21–23 confirming our through uptake of non-transferrin bound iron (NTBI). The
results. This disconnection with iron-overload parameters mechanism by which this uptake occurs is controversial,
has been observed also in chronic metabolic syndromes, too, but L-type calcium must play a role as it is present in
although substantial evidence shows that the clinical course large quantities in cardiomyocytes, pancreatic b cells, in var-
of these disorders is affected by iron overload.24 The differ- ious cell types of the anterior pituitary gland (including
ent mechanisms of iron uptake and accumulation among gonadotrophs, thyrotrophs, and corticoptrophs), and in the
different organs may be responsible of that phenomenon. parathyroid-hormone-producing cells of the parathyroid
Iron accumulates in the liver due to transferrin-mediated gland.25 Although NTBI composes a very small fraction of
mechanisms and LIC has inadequate ability to predict that body iron, it produces oxidative stress and organ damage.26
risk in extrahepatic organs.20 The endocrine glands and the While elevated LIC increases patients’ risk of iron-overload
Table 4A. Predicted risk chart for developing a new endocrine disease within 5 years, in percentages. The overall 5-year risk was 9.7%
Diseases at baseline = 0 Diseases at baseline = 1 Diseases at baseline = 2 Diseases at baseline = 3
Age TSH=1 TSH=3 TSH=5 TSH=1 TSH=3 TSH=5 TSH=1 TSH=3 TSH=5 TSH=1 TSH=3 TSH=5
1 6.3 9.9 15.2 3.4 5.4 8.5 1.9 3.0 4.7 1.0 1.6 2.5
2 6.5 10.2 15.8 3.6 5.6 8.8 1.9 3.1 4.8 1.0 1.7 2.6
4 7.1 11.1 17.1 3.9 6.1 9.6 2.1 3.3 5.3 1.1 1.8 2.9
6 7.7 12.0 18.5 4.2 6.6 10.4 2.3 3.6 5.7 1.2 2.0 3.1
8 8.4 13.1 20.1 4.6 7.3 11.4 2.5 4.0 6.3 1.4 2.2 3.4
10 9.3 14.4 21.9 5.1 8.0 12.5 2.8 4.4 6.9 1.5 2.4 3.8
12 10.3 15.8 24.0 5.7 8.9 13.8 3.1 4.9 7.6 1.7 2.6 4.2
14 11.4 17.5 26.3 6.3 9.9 15.2 3.5 5.4 8.5 1.9 3.0 4.7
16 12.8 19.4 28.9 7.1 11.0 16.9 3.9 6.1 9.6 2.1 3.3 5.3
18 14.3 21.6 31.6 8.0 12.4 18.9 4.4 6.9 10.8 2.4 3.8 6.0
20 16.1 23.9 34.5 9.1 14.0 21.0 5.1 7.9 12.2 2.8 4.3 6.8
25 21.5 30.5 41.5 12.7 18.9 27.3 7.2 11.1 16.6 4.0 6.2 9.6
30 27.8 36.9 46.9 17.7 25.0 33.8 10.5 15.6 22.4 6.0 9.2 13.8
35 33.5 41.4 50.2 23.6 31.0 38.7 15.2 21.3 28.5 9.1 13.4 19.2
40 37.2 43.9 52.7 29.1 35.1 41.5 20.8 27.1 33.2 13.5 18.9 25.0
45 39.0 45.9 55.2 32.5 37.2 43.3 26.2 31.0 35.5 19.0 24.4 29.5
50 40.5 47.9 57.9 34.0 38.6 45.2 29.5 32.8 36.8 24.2 28.3 31.6
Age in years. TSH: thyrotropin.
Table 4B. Predicted risk chart for developing a new endocrine disease within 1 year, in percentages. The overall 1-year risk was 1.1%
Diseases at baseline = 0 Diseases at baseline = 1 Diseases at baseline = 2 Diseases at baseline = 3
Age TSH=1 TSH=3 TSH=5 TSH=1 TSH=3 TSH=5 TSH=1 TSH=3 TSH=5 TSH=1 TSH=3 TSH=5
1 0.6 1.0 1.6 0.3 0.5 0.9 0.2 0.3 0.5 0.1 0.2 0.3
2 0.7 1.1 1.7 0.4 0.6 0.9 0.2 0.3 0.5 0.1 0.2 0.3
4 0.7 1.1 1.8 0.4 0.6 1.0 0.2 0.3 0.5 0.1 0.2 0.3
6 0.8 1.3 2.0 0.4 0.7 1.1 0.2 0.4 0.6 0.1 0.2 0.3
8 0.9 1.4 2.2 0.5 0.7 1.2 0.3 0.4 0.6 0.1 0.2 0.3
10 1.0 1.5 2.4 0.5 0.8 1.3 0.3 0.4 0.7 0.1 0.2 0.4
12 1.1 1.7 2.7 0.6 0.9 1.5 0.3 0.5 0.8 0.2 0.3 0.4
14 1.2 1.9 3.0 0.6 1.0 1.6 0.3 0.6 0.9 0.2 0.3 0.5
16 1.4 2.1 3.4 0.7 1.2 1.8 0.4 0.6 1.0 0.2 0.3 0.5
18 1.5 2.4 3.9 0.8 1.3 2.1 0.4 0.7 1.1 0.2 0.4 0.6
20 1.8 2.8 4.4 1.0 1.5 2.4 0.5 0.8 1.3 0.3 0.4 0.7
25 2.6 4.1 6.3 1.4 2.2 3.5 0.8 1.2 1.9 0.4 0.6 1.0
30 3.9 6.1 9.3 2.1 3.4 5.3 1.2 1.8 2.9 0.6 1.0 1.6
35 6.1 9.2 13.6 3.4 5.2 8.0 1.9 2.9 4.5 1.0 1.6 2.5
40 9.3 13.6 19.1 5.4 8.2 12.1 3.0 4.7 7.1 1.7 2.6 4.1
45 14.0 19.1 24.6 8.5 12.4 17.4 4.9 7.5 11.0 2.8 4.3 6.5
50 19.5 24.4 28.4 13.0 17.8 22.8 7.9 11.6 16.1 4.6 7.0 10.3
Age in years. TSH: thyrotropin.

complications, there is not a LIC threshold below which sample. Another reason may be that the therapy for an
cardiac and endocrine iron accumulation does not occur.27 endocrine disease ameliorates the function of other
The explanation of this paradox is that many chronically endocrine axes. This has been previously shown for BMD,
transfused patients have fully-saturated transferrin, regard- metabolic syndrome, and glucose and lipid metabolism
less of their LIC,28 and, as heart and endocrine glands exclu- disorders after treatment for hypothyroidism and hypogo-
sively accumulate NTBI, it is possible for them to be in pos- nadism.34–36 Furthermore, endocrine glands are not equally
itive iron balance even if the total body iron balance (LIC) is vulnerable to the iron toxicity, and patients with more
neutral or negative.27 Patients who miss chelator doses endocrinopathies have already wiped out the most endan-
expose their extrahepatic organs to unrestricted uptake of gered endocrine glands.
labile iron species.29 All our patients were on regular iron-chelation therapy
Previous studies reported a correlation between cardiac and had acceptable levels of iron load. Therefore, the
T2* and manifest endocrinopathies.21,22 However, those markers of iron overload were expected to be stable over
studies concerned patients with severe iron overload, with time or to have minor fluctuations. However, iron over-
T2* <20 msec, while our sample had average ferritin <1,000 load increased for same patients. If that phenomenon was
ng/mL, LIC <5 mg/dry weight (dw), and T2* >30 msec due to scarce patient compliance and if compliance was
which are considered as the acceptable target levels to reach associated with our explanatory variables, our estimates
during iron-chelating therapy.18 It has been shown that the may be biased. However, also iron overload measures
iron overload of endocrine glands preceded that of the taken contemporaneously with disease incidence have
heart, although both phenomena are mediated by NTBI.11 shown no effect, as well as their deltas. A lack of compli-
However, endocrine organs have superior reserve capacity ance could have been assumed if at least the latest tests
and the clinical manifestations concerning them may assessing iron overload were associated with higher risk.
appear years after silent iron accumulation.12 When iron Heightened TSH has been associated with endothelial
overload continues, due to the lack of patient compliance or dysfunction,37 defined as a diminished bioavailability of
due to an inadequate dose of iron-binding therapy, the nitric oxide (NO) and/or an increase in vasoconstrictive
heart starts to show signs of overload, which can be identi- factors such as endothelin (ET-1). That condition has been
fied through MRI-T2*.20 Therefore, cardiac T2* is not an well documented in thalassemia patients and is associated
early indicator of iron overload. We have not found a corre- with cardiac, hepatic and endocrine clinical
lation between T2* and endocrine complications in our complications.32,38 Endothelial dysfunction in TDT is a pro-
cohort of patients, as the vast majority of our patients had gressive process, starting from childhood, and recent stud-
an acceptable iron balance. Abnormal cardiac T2* is an ies found significantly higher plasma levels of asymmetric
excellent marker of NTBI control, but it is insensitive dimethylarginine (ADMA), a novel risk marker of cardio-
because exposure must be severe and quite prolonged. As a vascular disease implicated in the pathogenesis of
result, abnormal cardiac T2* has a very high positive predic- endothelial dysfunction, in very young TDT children.39 So,
tive value for endocrine iron deposition. However, once the increased TSH may be an early expression of systemic
heart has been successfully de-ironed, endocrine glands typ- endothelial dysfunction in TDT which is considered an
ically retain moderate iron deposition. Finally, even when independent risk factor of future complications.32 TSH
the endocrine glands have been successfully de-ironed, their appears the best marker of systemic endocrine gland dys-
functional reserve has been destroyed.4,11,12,20-22 function, as its measurement is very accurate and widely
As there is considerable variation in the risk of used in clinical practice,40 differently from the several limits
endocrinopathies in patients without signs of heart and in the assessment of other pituitary hormones, as growth
liver overload,17,21,22,30–32 and because those kinds of hormone (GH), adrenocorticotropic hormone (ACTH),
endocrinopathies, when manifest, are irreversible,17,21,22,30–32 LH/FSH.41 Furthermore, production of TSH is the last
further clinical and laboratory predictors in addition to MRI affected by the progressive damage of pituitary gland in
imaging of endocrine glands are needed to prevent TDT patients which impairs firstly GH secretion, followed
endocrine complications. by LH/FSH and ACTH.42 For all these reasons, TSH may be
We proposed our risk chart on the model 3B, in which the sentinel for endocrine gland dysfunction. Along with
the association between risk of complication and age was TSH, age is also associated with endothelial dysfunction,
considered as being linear, while adding a binary marker of which could be the main driver of endocrine and cardio-
adulthood and an interaction parameter between age and vascular risk.32 These observations pave the way for the
adulthood. We preferred this model to the simpler model early identification of clinical complications in other meta-
3A (log-linear age alone) because it had a slightly better R2 bolic diseases, which have been reported greatly affected
and mostly because adults and children affected by TDT by iron overload.24
are usually treated in separate health care centers and as a The variation in risk of complications that our best
matter of fact they define two separate categories. We model could explain was insufficient (25%). Therefore,
have developed the largest analysis on endocrine compli- there must be factors other than those we considered that
cations in TDT patients ever developed so far and this is have some effect on the incidence on endocrine complica-
the first study providing clear benchmarks for patients’ tions. These may include NTBI, transferrin saturation,
management. However, the predictive power of our risk smoking, other markers of endothelial dysfunction, pan-
chart must be improved and validated. creas and pituitary R2* which weren’t considered.
It is plausible that a diagnosis of an endocrinopathy pro- Furthermore, the different chelation history among the
duces a warning effect that is similar to that observed after study cohort (older patients treated for longer with subcu-
an abnormal cardiac T2*, which makes patients have bet- taneous DFX compared to younger patients treated for
ter compliance and clinicians increase iron-chelating longer with oral DFX) creates an inherent age effect to be
dose.33 That may explain why previous endocrinopathies taken into consideration.
were associated with lower incidence of new ones in our The apparent increase in endocrine complications after

M. Casale et al.
the age of 12 years is certainly related to different factors, Disclosures

such as the current inability to recognize hypogonadism Università degli Studi della Campania “Luigi Vanvitelli” spon-
prior to puberty; the effect of hypogonadism on bone sored the study (VALERE project) and received a partial financial
metabolism due to the impact of steroids on bone miner- support to trial costs from and Novartis Farma SpA which had no
alization; the delay between the start of iron damage in role in study design, data collection, data analysis, data interpreta-
the gland tissues and the onset of overt clinical complication, or writing of the report. The corresponding author had full
tions, e.g., diabetes.4,11-13 Furthermore, adolescence is also access to all the data in the study and had final responsibility for the
marked by less physical activity and more adverse body decision to submit for publication; MC received speaker honoraria
habitus and nutrition that worsen insulin sensitivity. For all and advisory board fees from Novartis Farma SpA; GLF received
these reasons, children are not protected by iron damage advisory board fees from Novartis Farma SpA; MP received con-
in endocrine glands and conversely, they require more sulting fees and advisory board fees from Novartis Farma SpA; SP
aggressive prophylaxis to avoid pituitary and pancreatic received grant support paid to his institution, advisory board fees
iron accumulation which will be clinically manifested only and speaker honoraria from Novartis Farma SpA. No other poten-
years later, when the functional reserve has been tial conflict of interest relevant to this article was reported. All other
destroyed. Different chelation goals (such as transferrin authors declare no conflict of interests of any kind.
desaturation and the use of direct endocrine imaging) and
alternative chelation strategies are necessary to better pro- Contributions
tect endocrine glands. MC, SP, and AIL designed the study. Each author collected the
In conclusion, although in DFX-treated TDT patients the data from his/her own center and takes responsibility for the accu-
risk of developing an endocrine complication is generally racy of the data provided. AIL carried out the statistical analysis;
lower than the previously reported risk, there is consider- MC and AIL drafted the manuscript. All authors contributed to the
able risk variation and the burden of these complications interpretation of the data and approved the manuscript. The cen-
remains high. This is the first study providing a practical ters in Naples, Genoa, Milan, and Padua are part of the European
tool for physicians to identify patients at higher risk of Reference Network on Rare Hematological Diseases (ERN-
developing endocrine complications. Future research will EuroBloodNet). All centers involved in the study are part of the
look at increasing the amount of variation explained from Italian Society for Thalassemia and Hemoglobinopathies (SITE)
our model and testing further clinical and laboratory pre- and pediatric centers are part of the Italian Association of Pediatric
dictors, including the assessment of direct endocrine MRI. Hematology and Oncology.
References stores. Blood. 2010;116(4):537-543. Thalassaemia (TDT), 3rd edn. Nicosia (CY):
10. Deugnier Y, Turlin B, Ropert M, et al. Thalassaemia International Federation, 2014
1. Modell B, Khan M, Darlison M, Westwood Improvement in liver pathology of patients http://www.ncbi.nlm.nih.gov/books/NBK2
MA, Ingram D, Pennell DJ. Improved sur- with b-thalassemia treated with deferasirox 69382/ (accessed August 15, 2020).
vival of thalassaemia major in the UK and for at least 3 years. Gastroenterology. 19. Royston P. Explained variation for survival
relation to T2* cardiovascular magnetic res- 2011;141(4):1202-1211. models. Stata J. 2006;6(1):83-96.
onance. J Cardiovasc Magn Reson. 2008;10 11. Noetzli LJ, Papudesi J, Coates TD, Wood JC. 20. Wood JC. Use of magnetic resonance imag-
(1):42. Pancreatic iron loading predicts cardiac iron ing to monitor iron overload. Hematol
2. Rund D. Thalassemia 2016: modern medi- loading in thalassemia major. Blood. Oncol Clin North Am. 2014;28(4):747-764.
cine battles an ancient disease. Am J 2009;114(19):4021-4026. 21. Au W-Y, Lam WW-M, Chu WWC, et al. A
Hematol. 2016;91(1):15-21. 12. Noetzli LJ, Panigrahy A, Mittelman SD, et al. cross-sectional magnetic resonance imaging
3. Pinto VM, Poggi M, Russo R, Giusti A, Forni Pituitary iron and volume predict hypogo- assessment of organ specific hemosiderosis
GL. Management of the aging beta-tha- nadism in transfusional iron overload. Am J in 180 thalassemia major patients in Hong
lassemia transfusion-dependent population - Hematol. 2012;87(2):167-171. Kong. Haematologica. 2008;93(5):784-786.
the Italian experience. Blood Rev. 2019;38: 13. Noetzli LJ, Mittelman SD, Watanabe RM, 22. Ang AL, Tzoulis P, Prescott E, Davis BA,
100594. Coates TD, Wood JC. Pancreatic iron and Barnard M, Shah FT. History of myocardial
4. Wood JC. Impact of iron assessment by glucose dysregulation in thalassemia major. iron loading is a strong risk factor for dia-
MRI. Hematololgy Am Soc Hematol Educ Am J Hematol. 2012;87(2):155-160. betes mellitus and hypogonadism in adults
Program. 2011;2011:443-450. 14. Belhoul KM, Bakir ML, Saned M-S, Kadhim with b thalassemia major. Eur J Haematol.
5. Maffei L, Sorrentino F, Caprari P, et al. HCV AM, Musallam KM, Taher AT. Serum fer- 2014;92(3):229-236.
infection in Thalassemia syndromes and ritin levels and endocrinopathy in medically 23. Pinto VM, Bacigalupo L, Gianesin B, et al.
hemoglobinopathies: new perspectives. treated patients with b thalassemia major. Lack of correlation between heart, liver and
Front Mol Biosci. 2020;7:7. Ann Hematol. 2012;91(7):1107-1114. pancreas MRI-R2*: results from long-term
6. El Kholy M, Elsedfy H, Soliman A, Anastasi 15. Casale M, Citarella S, Filosa A, et al. follow-up in a cohort of adult b-thalassemia
S, Raiola G, De Sanctis V. Towards an opti- Endocrine function and bone disease during major patients. Am J Hematol. 2018;93(3):
mization of the management of endocrine long-term chelation therapy with E79-82.
complications of thalassemia. J Pediatr deferasirox in patients with b-thalassemia 24. Fernández-Real JM, Manco M. Effects of
Endocrinol Metab. 2014(9-10);27:801-805. major. Am J Hematol. 2014 89(12):1102- iron overload on chronic metabolic diseases.
7. Casale M, Filosa A, Ragozzino A, et al. 1106. Lancet Diabetes Endocrinol. 2014;2(6):513-
Long-term improvement in cardiac magnet- 16. Poggi M, Sorrentino F, Pugliese P, et al. 526.
ic resonance in b-thalassemia major patients Longitudinal changes of endocrine and bone 25. Oudit GY, Trivieri MG, Khaper N, Liu PP,
treated with deferasirox extends to patients disease in adults with b-thalassemia major Backx PH. Role of L-type Ca2+ channels in
with abnormal baseline cardiac function. receiving different iron chelators over 5 iron transport and iron-overload cardiomy-
Am J Hematol. 2019;94(3):312-318. years. Ann Hematol. 2016;95(5):757-763. opathy. J Mol Med. 2006;84(5):349-364.
8. Pennell DJ, Porter JB, Cappellini MD, et al. 17. Thuret I, Pondarré C, Loundou A, et al. 26. Cabantchik ZI. Labile iron in cells and body
Deferasirox for up to 3 years leads to contin- Complications and treatment of patients fluids: physiology, pathology, and pharma-
ued improvement of myocardial T2* in with b-thalassemia in France: results of the cology. Front Pharmacol. 2014;5:45.
patients with b-thalassemia major. National Registry. Haematologica. 27. Noetzli LJ, Carson SM, Nord AS, Coates
Haematologica 2012; 97(6):842-848. 2010;95(5):724-729. TD, Wood JC. Longitudinal analysis of heart
9. Wood JC, Kang BP, Thompson A, et al. The 18. Cappellini MD, Cohen A, Porter J, Taher A, and liver iron in thalassemia major. Blood.
effect of deferasirox on cardiac iron in tha- Viprakasit V, editors. Guidelines for the 2008;112(7):2973-2978.
lassemia major: impact of total body iron management of transfusion dependent 28. Piga A, Longo F, Duca L, et al. High non-

transferrin bound iron levels and heart dis- North Am. 2018;32(2):237-245. 38. Aggeli C, Antoniades C, Cosma C, et al.
ease in thalassemia major. Am J Hematol. 33. Pennell DJ, Udelson JE, Arai AE, et al. Endothelial dysfunction and inflammatory
2009;84(1):29-33. Cardiovascular function and treatment in b- process in transfusion-dependent patients
29. Wood JC, Glynos T, Thompson A, et al. thalassemia major: a consensus statement with beta-thalassemia major. Int J Cardiol.
Relationship between labile plasma iron, from the American Heart Association. 2005;105(1):80-84.
liver iron concentration and cardiac response Circulation. 2013;128(13):281-308. 39. Gursel O, Tapan S, Sertoglu E, et al. Elevated
in a deferasirox monotherapy trial. 34. Salzano A, Marra AM, Arcopinto M, et al. plasma asymmetric dimethylarginine levels
Haematologica 2011;96(7):1055-1058. Combined effects of growth hormone and in children with beta-thalassemia major
30. Cunningham MJ, Macklin EA, Neufeld EJ, testosterone replacement treatment in heart may be an early marker for endothelial dys-
Cohen AR, Thalassemia Clinical Research failure. ESC Heart Fail. 2019;6(6):1216-1221. function. Hematology. 2018;23(5):304-308.
Network. complications of beta-thalassemia 35. Rochira V. Late-onset hypogonadism: bone 40. Clerico A, Trenti T, Aloe R, et al. A multicen-
major in North America. Blood. 2004;104 health. Andrology. 2020;8(6):1539-1550. ter study for the evaluation of the reference
(1):34-39. 36. Feingold KR, Anawalt B, Boyce A, et al., edi- interval for TSH in Italy (ELAS TSH Italian
31. Vogiatzi MG, Macklin EA, Trachtenberg FL, tors. Endotext. South Dartmouth (MA): Study). Clin Chem Lab Med. 2018;57(2):
et al. Differences in the prevalence of MDText.com, Inc., 2000 259-267.
growth, endocrine and vitamin D abnormal- http://www.ncbi.nlm.nih.gov/books/NBK2 41. Bidlingmaier M, Strasburger CJ. Growth
ities among the various thalassaemia syn- 78943/ (accessed August 15, 2020). hormone assays: current methodologies and
dromes in North America. Br J Haematol. 37. Niknam N, Khalili N, Khosravi E, their limitations. Pituitary. 2007;10(2):115-
2009;146(5):546-556. Nourbakhsh M. Endothelial dysfunction in 119.
32. Taher AT, Cappellini MD, Bou-Fakhredin R, patients with subclinical hypothyroidism 42. Jameson JL, De Groot LJ. Endocrinology-E-
Coriu D, Musallam KM. Hypercoagulability and the effects of treatment with levothy- Book: Adult and Pediatric. Elsevier Health
and vascular disease. Hematol Oncol Clin roxine. Adv Biomed Res. 2016;5:38. Sciences, 2010.

ARTICLE Iron Metabolism & its Disorders
Ferrata Storti Foundation UBA6 and NDFIP1 regulate the degradation

of ferroportin
Lisa Traeger,1 Steffen B. Wiegand,1 Andrew J. Sauer,1 Benjamin H.P. Corman,1
Kathryn M. Peneyra,1 Florian Wunderer,1,2 Anna Fischbach,1 Aranya Bagchi,1
Rajeev Malhotra,3 Warren M. Zapol1 and Donald B. Bloch1,4
1
Anesthesia Center for Critical Care Research of the Department of Anesthesia, Critical
Care and Pain Medicine, Massachusetts General Hospital and Harvard Medical School,
Boston, MA, USA; 2Department of Anesthesiology, Intensive Care Medicine and Pain
Haematologica 2022 Therapy, University Hospital Frankfurt, Goethe University, Frankfurt, Germany;
3
Volume 107(2):478-488 Cardiovascular Research Center and the Cardiology Division of the Department of
Medicine, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA
and 4Division of Rheumatology, Allergy and Immunology of the Department of Medicine,
Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA
ABSTRACT
H
epcidin regulates iron homeostasis by controlling the level of fer-
roportin, the only membrane channel that facilitates export of
iron from within cells. Binding of hepcidin to ferroportin induces
the ubiquitination of ferroportin at multiple lysine residues and subse-
quently causes the internalization and degradation of the ligand-channel
complex within lysosomes. The objective of this study was to identify
components of the ubiquitin system that are involved in ferroportin
degradation. A HepG2 cell line, which inducibly expresses ferroportin-
green fluorescent protein (FPN-GFP), was established to test the ability
of small interfering (siRNA) directed against components of the ubiquitin
system to prevent BMP6- and exogenous hepcidin-induced ferroportin
degradation. Of the 88 siRNA directed against components of the ubiq-
uitin pathway that were tested, siRNA-mediated depletion of the alter-
native E1 enzyme UBA6 as well as the adaptor protein NDFIP1 prevent-
ed BMP6- and hepcidin-induced degradation of ferroportin in vitro. A
Correspondence: third component of the ubiquitin pathway, ARIH1, indirectly inhibited
ferroportin degradation by impairing BMP6-mediated induction of hep-
DONALD B. BLOCH cidin. In mice, the AAV-mediated silencing of Ndfip1 in the murine liver
dbloch@mgh.harvard.edu
increased the level of hepatic ferroportin and increased circulating iron.
LISA TRAEGER The results suggest that the E1 enzyme UBA6 and the adaptor protein
email@lisatraeger.de NDFIP1 are involved in iron homeostasis by regulating the degradation
of ferroportin. These specific components of the ubiquitin system may
Received: February 6, 2021. be promising targets for the treatment of iron-related diseases, including
iron overload and anemia of inflammation.
Accepted: July 22, 2021.
Pre-published: July 29, 2021
Introduction
https://doi.org/10.3324/haematol.2021.278530 Iron is an essential element that is required for a spectrum of cellular and biological
processes including oxygen transport, DNA synthesis and the production of energy.
High levels of iron, in the presence of oxygen, may catalyze the production of reac-
©2022 Ferrata Storti Foundation tive oxygen species, which are free radicals that may damage cellular proteins and
Material published in Haematologica is covered by copyright. membranes. The level of iron in the body must be tightly regulated to provide suffi-
All rights are reserved to the Ferrata Storti Foundation. Use of cient levels to permit fundamental biological processes while preventing the damag-
conditions:
ing effects of excess iron.1,2
https://creativecommons.org/licenses/by-nc/4.0/legalcode. The hepatic hormone hepcidin is a critical regulator of systemic iron homeosta-
Copies of published material are allowed for personal or inter- sis.3–5 Hepcidin expression is controlled by at least three stimuli: i) increased serum
nal use. Sharing published material for non-commercial pur- and liver iron, which induce hepcidin via the bone morphogenetic protein (BMP) sig-
naling pathway; ii) increased mediators of inflammation (IL-1b and IL-6), which
sect. 3. Reproducing and sharing published material for com- increase hepcidin via the Jak/Stat pathway; and iii) the hormone erythroferrone,
mercial purposes is not allowed without permission in writing which inhibits BMP signaling by sequestering BMP6 in response to increased ery-
from the publisher. thropoietic demand.6–9 Hepcidin regulates iron homeostasis by controlling the cell
surface level of ferroportin, which is the only known membrane channel that facili-
tates export of iron from within cells.10 Ferroportin is a member of the superfamily

Regulation of ferroportin degradation
of transporters of small molecules, which includes more Methods

than 300 membrane-bound proteins.11,12 Ferroportin is pre-
dominantly expressed in tissues associated with iron trans- HepG2-FPN-GFP cell line
port including enterocytes, hepatocytes, macrophages and A plasmid encoding human ferroportin (NM_016917) fused to
erythrocytes.10,11,13 The protein has 12 membrane-spanning GFP was a gift from Tomas Ganz (David Geffen School of
domains, which create a channel through which iron is Medicine, UCLA).14 DNA encoding the fusion protein was ligated
transported. Binding of hepcidin to the main extracellular into the NheI and NotI sites of plasmid pTRE2hyg (Clontech
cavity of ferroportin causes ligation of ubiquitin molecules cat#631014). The plasmid was transfected into the HepG2 “tet-on
to multiple intracellular lysine residues.14 Polyubiquitination advanced” cell line (Clonetech cat#630932) using Effectene trans-
of ferroportin induces the internalization of the hepcidin- fection reagents (Qiagen, Germantown, MD, USA) according to
ferroportin complex followed by degradation within lyso- the manufacturer’s instructions and successfully transfected cells
somes.10,14,15 Degradation of ferroportin results in decreased were selected using hygromycin (0.4 mg/mL). Individual cell lines
serum iron, because enterocytes, hepatocytes and were established and were confirmed to express FPN-GFP in the
macrophages are no longer able to transfer intracellular iron presence of doxycycline (2 µg/mL). One colony, which expressed
to the circulation.16 In addition to the ability of hepcidin to a high level of FPN-GFP after treatment with doxycycline, was
induce degradation of ferroportin, the hormone is also able selected for further experiments. HepG2-FPN-GFP cells were
to inhibit iron export by directly occluding the iron maintained in Eagle Minimum Essential Medium, 10% FBS, L-glu-
channel.17 Occlusion of the channel may be especially tamine (2 mM), G418 (100 ng/mL), hygromycin (0.4 mg/mL),
important for cells, such as mature red blood cells, which penicillin (100 units/mL), and streptomycin (100 µg) at 37°C in 5%
lack the endocytic machinery required to degrade ferro- CO2 and 95% humidity.
portin.17
Ubiquitin is a 76 amino acid polypeptide that can be Adeno-associated virus administration
attached to lysine residues in proteins. The attachment of All experiments using mice were approved by the Partners
ubiquitin to a protein regulates the protein’s localization, Subcommittee on Research Animal Care at Massachusetts
stability and/or activity.18 The process of ubiquitination General Hospital, Boston, MA, USA (Protocol # 2007N000052).
involves the activation and transient conjugation of ubiqui- Wild-type mice on a C57BL/6J background were purchased from
tin to a carrier protein, with subsequent final ligation of the Jackson Laboratories (Bar Harbor, ME, USA). Animals were fed a
ubiquitin molecule to a substrate. In general, ubiquitination standard diet (380 ppm iron). Adeno-associated virus (AAV) short
requires three different kinds of enzymes: a ubiquitin acti- hairpin RNA (shRNA) AAV2/8-GFP-U6-m-Ndfip1-shRNA
vating enzyme (E1), a ubiquitin conjugating enzyme (E2), (AAV2/8-shNdfip1) and AAV2/8-GFP-U6-scrmb-shRNA (AAV2/8-
and a ubiquitin ligase (E3). The human ubiquitin system shControl) were obtained from Vector BioLabs (Malvern, PA, USA).
encodes two different E1 enzymes (UBA1 and UBA6), Eight-week-old male mice were injected intravenously with
approximately 50 different E2 enzymes, and more than 600 1x1011 particles of AAV2/8-shNdfip1 or AAV2/8-shControl via the
E3 enzymes.19–21 The ubiquitin E3 ligases are important for tail vein. Six weeks later, mice were anesthetized with 4% isoflu-
substrate recognition and are divided into three different rane and whole blood was collected by cardiac puncture. Liver and
classes.22 Depending on the class of ligase, the E3 enzyme spleen were harvested for further analysis.
either directly transfers ubiquitin to a substrate (“HECT”
and “RBR” ligases) or acts as an adaptor to facilitate the Statistical analysis
transfer of ubiquitin from an E2 enzyme directly to the sub- All statistical analyses were performed using GraphPad Prism
strate (“RING” E3 ligases).22,23 Binding of the E3 enzyme to 8.3.0 (GraphPad Software, San Diego, CA, USA). Data are
the substrate may also require an adaptor protein that acts expressed as mean ± standard deviation (SD). The Shapiro-Wilk
as a scaffold between the E3 enzyme and the target protein. test was performed to test for normality. Correlation analysis were
In this study, an in vitro small interfering RNA (siRNA) performed using Pearson correlation. Comparison of two groups
screen was performed to determine which proteins in the was performed using the Student’s t-test for parametric data and
ubiquitin pathway are involved in ferroportin degradation. the Mann-Whitney-U test for non-parametric data. Comparison
A previous study used a modified HEK293 cell line, in of more than two groups was performed using one-way ANOVA
which expression of ferroportin was induced by the addi- with Tukey post hoc test (for parametric data) or the Kruskal-Wallis
tion of ponasterone.24 Exogenous hepcidin and putative test with Dunn’s post hoc test (for non-parametric data). After
inhibitors of ferroportin degradation were added to this cell adjusting for multiple comparisons, a P value <0.05 was consid-
line and the level of ferroportin was then measured. To per- ered statistically significant.
mit screening for specific enzymes involved in ferroportin A detailed description of other methods can be found in the
ubiquitination without using exogenous hepcidin, we Online Supplementary Appendix.
established a HepG2 cell line that expresses the ferroportin-
green fluorescent protein (FPN-GFP) fusion protein in
response to doxycycline. In this cell line, BMP6 can be used Results
to gradually induce the expression of endogenous hepcidin.
The HepG2-FPN-GFP cell line was used to show that the Preparation and characterization of the HepG2-FPN-
alternative E1 enzyme UBA6 as well as the NEDD4 family GFP cell line
binding protein NDFIP1 are able regulate the degradation of Binding of hepcidin to ferroportin induces the polyubiq-
ferroportin in response to BMP6, as well as exogenous hep- uitination, internalization and lysosomal degradation of the
cidin. Depletion of the E3 ligase ARIH1 indirectly inhibited ligand-channel complex.10 To identify the specific enzymes
ferroportin degradation by impairing BMP6-mediated hep- that mediate ubiquitination of ferroportin, we established a
cidin induction. In vivo, depletion of Ndfip1 in the murine stable HepG2 cell line that inducibly expresses FPN-GFP
liver increased the level of hepatic ferroportin and increased (HepG2-FPN-GFP) in the presence of doxycycline.
circulating iron. Treatment of HepG2-FPN-GFP cells with 2 µg/mL of doxy-

L. Traeger et al.
cycline for 18 hours (h) induced expression of the fusion endogenous hepcidin in HepG2-FPN-GFP cells (Figure 2D).
protein, which was detected at the cell surface (Figure 1A). In contrast, siRNA directed against SMAD4 blunted BMP6-
The FPN-GFP fusion protein was able to export iron, as induced expression of hepcidin.
indirectly indicated by decreased levels of intracellular FTL To further demonstrate that depletion of UBA6 blocks
and FTH1 after FPN-GFP induction (Figure 1B; Online degradation of ferroportin, independent of an effect on
Supplementary Figure S1A). Treatment with hepcidin (50 BMP-induced expression of hepcidin, the effect of exoge-
ng/mL) for 90 minutes (min) caused FPN-GFP to localize to nous hepcidin on the cellular localization of FPN-GFP in
punctate structures in the cytoplasm (Figure 1A), and treat- HepG2 cells was investigated. To determine the amount of
ment with hepcidin for 18 h caused degradation of the hepcidin produced by HepG2 cells after treatment with
fusion protein (Figure 1C). Because BMP6 induces HepG2 BMP6, HepG2 cells were incubated with BMP6 (10 ng/mL)
cells to express hepcidin25 (Figure 1D), we were able to for 18 h and the amount of hepcidin in the tissue culture
investigate the effect of gradual induction of endogenous medium was measured by enzyme-linked immunosorbant
hepcidin on ferroportin degradation. Treatment with BMP6 assay (ELISA). Under these conditions, BMP6 induced 3.9
(10 ng/mL) for 18 h caused degradation of the FPN-GFP ng/mL (± 0.4 ng/mL) of hepcidin, and this concentration of
fusion protein as detected by indirect fluorescence and hepcidin (rather than the much higher dose of 30-50 ng/mL
immunoblotting (Figure 1A and C). Pretreatment with used in other studies10,24) was used to treat cells in subse-
chloroquine (100 µM for 2 h), an inhibitor of lysosomal quent experiments. Cells were transfected with siControl,
degradation, prevented BMP6-mediated FPN-GFP degrada- siUBA1, or siUBA6 and treated overnight with hepcidin (4
tion and caused FPN-GFP to localize to lysosomes in the ng/mL). In the presence of this low concentration of hep-
cytoplasm (Online Supplementary Figure S1B). cidin, FPN-GFP expression at the cell surface persisted in
BMP6 induces expression of hepcidin through the BMP UBA6-depleted cells but not in siControl-treated- or
receptor-SMAD 1/5/8 pathway. After binding to the BMP siUBA1-treated cells (Figure 2E). The inability of hepcidin to
receptor complex, activated BMP type I receptors phospho- degrade the FPN-GFP fusion protein in siUBA6 treated cells
rylate SMAD 1/5/8 proteins, which translocate together was confirmed by immunoblot (Online Supplementary Figure
with SMAD4 to the nucleus, and induce hepcidin expres- S2C). Taken together, these results show that UBA6 is
sion.26 The siRNA-mediated inhibition of SMAD4 in required for hepcidin induced internalization and degrada-
HepG2-FPN-GFP cells prevented BMP6-mediated degradation of FPN-GFP.
tion of FPN-GFP (Figure 1E). Taken together, these results
show that the HepG2-FPN-GFP cell line expresses The adaptor protein NDFIP1 regulates ferroportin
inducible, functional FPN-GFP. Both BMP6-induced degradation
endogenous hepcidin and exogenous hepcidin cause inter- To identify additional components of the ubiquitin path-
nalization and degradation of the FPN-GFP fusion protein. way that might be involved in ferroportin degradation,
siRNA directed against different E2 and E3 enzymes, as
E1 enzyme UBA6 is required for ubiquitination well as other known components of the ubiquitin pathway,
of ferroportin were tested for the ability to inhibit BMP6-mediated degra-
The human ubiquitin system encodes two E1 enzymes: dation of FPN-GFP (Online Supplementary Table S1, n=77). A
UBA1 (also known as in UBE1) and UBA6. To identify the commercially available library (Dharmacon, Lafayette, CO,
E1 enzyme that is involved in ferroportin degradation, USA), which contains siRNA that were previously verified
HepG2-FPN-GFP cells were transfected with siRNA that to silence the corresponding targets and to minimize off-tar-
targeted each of the E1 enzymes or with a control siRNA get effects, was used in these studies. HepG2 cells transfect-
(siControl). Twenty-four hours after transfection with ed with siRNA directed against SMAD4 were used as pos-
siRNA, cells were treated overnight with doxycycline to itive controls for inhibition of BMP6-mediated degradation
induce the expression of FPN-GFP and were then incubated of FPN-GFP; siControl was used as a negative control.
with BMP6 for 18 h. Cells that were treated with siControl Eighteen hours after treatment with BMP6, the localization
and BMP6 had decreased cell surface expression of FPN- of FPN-GFP was determined by immunofluorescence. In
GFP (Figure 2A). Depletion of UBA1 did not prevent the the first screen, we identified 23 siRNA directed against dif-
BMP6-induced localization of FPN-GFP to lysosomes and ferent E2 and E3 enzymes that appeared to block FPN-GFP
subsequent degradation. However, depletion of UBA6 pre- relocalization to the lysosome based on FPN-GFP persist-
vented BMP6-mediated degradation of ferroportin, as indi- ence at the cell surface after BMP6 treatment. In second and
cated by the persistence of the FPN-GFP fusion protein at tertiary screens, all positive candidates were re-evaluated to
the cell surface (Figure 2A). Immunoblotting confirmed that exclude false positives.
depletion of UBA6, but not UBA1, impaired degradation of Depletion of each of three E2 enzymes, UBE2R2, UBE2E2
FPN-GFP (Figure 2B; Online Supplementary Figure 2A). The and UBE2J2 partially blocked the internalization of FPN-
successful depletion of each of the E1 enzymes by the GFP (Online Supplementary Figure S3A and B), while deple-
appropriate siRNA was confirmed by quantitative poly- tion of other, individual E2 enzymes did not impair BMP6-
merase chain reaction (qPCR) (Figure 2C). Successful deple- mediated FPN-GFP degradation (data not shown).
tion of UBA6 by siUBA6 was not affected by the addition Treatment with pairwise combinations of UBE2R2,
of BMP6 (Online Supplementary Figure S2B). UBE2E2 and UBE2J2 or all three of the E2 enzymes did not
Depletion of UBA6 might block degradation of FPN-GFP further prevent the degradation of ferroportin (data not
or prevent BMP6-induced expression of hepcidin. To con- shown), suggesting that additional E2 enzymes participate
sider this latter possibility, the ability of UBA6 depletion to in FPN ubiquitination.
inhibit the BMP signal transduction pathway was investi- In an initial screen, depletion the NEDD family interact-
gated. Depletion of UBA6 had no effect on BMP6-mediated ing protein NDFIP1 and the E3 enzyme ARIH1 impaired
phosphorylation of SMAD1/5/8 (Figure 2B). In addition, BMP6-induced FPN-GFP localization to lysosomes and sub-
siRNA directed against UBA6 did not prevent expression of sequent degradation of the fusion protein (Figure 3A).

C D
Figure 1. Characterization of the HepG2-FPN-GFP cell line. (A) Images of untreated HepG2 cells, and cells treated with doxycycline (Dox; 2 µg/mL) alone, with Dox
followed by hepcidin (50 ng/mL) for 90 minutes (min), with Dox followed by BMP6 (10 ng/mL) for 18 hours (h) are shown. (B) Treatment with Dox induced the expres-
sion of the ferroportin-green fluorescent protein (FPN-GFP) fusion protein. Dox-treated cells had reduced levels of intracellular ferritin light-chain (FTL), consistent with
increased iron export in cells expressing FPN-GFP. GAPDH was used as a loading control. (C) In the absence of Dox, the fusion FPN-GFP protein was not detected by
immunoblot (lanes 1 to 3). In the presence of Dox, FPN-GFP was expressed (lane 4). Treatment with hepcidin (50 ng/mL; lane 5), or BMP6 (10 ng/mL; lane 6), for
18 h caused degradation of the FPN-GFP fusion protein. (D) BMP6 stimulation (10 ng/mL) for 18 h induced hepcidin mRNA expression in HepG2-FPN-GFP cells, as
determined by quantitative polymerase chain reaction (qPCR) (mRNA expression relative to control; **P<0.01; Mann-Whitney-U test). (E) Images of siSMAD4 trans-
fected cells treated with Dox (left panel), siSMAD4 transfected cells treated with Dox followed by BMP6 (10 ng/mL; middle panel) and siControl transfected cells treat-
ed with Dox followed by BMP6 (right panel) are shown. Small interfering RNA (siRNA)-mediated inhibition of SMAD4 prevented BMP6-mediated degradation of the
FPN-GFP fusion protein. The location of nuclei in (A) and (E) are indicated by staining with DAPI (blue). White bar indicates 100 µm.

L. Traeger et al.
siNDFIP1 successfully depleted NDFIP1 mRNA in both the ed cells were treated with exogenous hepcidin (4 ng/mL).
absence (Figure 3B) and the presence (Online Supplementary Compared to cells that were transfected with siControl,
Figure S3C) of exogenous BMP6. siRNA directed against depletion of NDFIP1 inhibited hepcidin-mediated degrada-
NDFIP1 had no effect on the ability of BMP6 to induce hep- tion of the FPN-GFP fusion protein (Figure 3D; Online
cidin expression, demonstrating that the BMP signaling Supplementary Figure S2C).
pathway was intact (Figure 3C). To confirm that depletion To investigate the possibility that NDFIP1 interacts with
of NDFIP1 blocks degradation of FPN-GFP, NDFIP1 deplet- ferroportin, HepG2 cells were incubated in the presence or
B C
D E
Figure 2. UBA6 is required for hepcidin-mediated degradation of the ferroportin-green fluorescent protein fusion protein. (A) Cells were transfected with siControl,
siUBA1 or siUBA6, and treated with doxycycline (Dox) or Dox followed by BMP6 (10 ng/mL) for 18 hours (h), as indicated. Dox-induced expression of ferroportin-green
fluorescent protein (FPN-GFP); depletion of UBA6, but not UBA1, prevented BMP6-mediated FPN degradation. (B) The levels of FPN-GFP and phosphorylated SMAD
1/5/8 in siControl-, siUBA1- and siUBA6-transfected HepG2-FPN-GFP cells in the presence (+) or absence (-) of BMP6 (10 ng/mL for 18 h) are shown. In the absence
of Dox, control cells did not express the FPN-GFP fusion protein. Treatment with BMP6 increased the level of pSMAD 1/5/8. UBA6 depletion prevented BMP6-medi-
ated degradation of FPN-GFP. GAPDH was used as a loading control. The immunoblot is representative of 3 separate experiments. (C) UBA1 and UBA6 were success-
fully depleted using the appropriate small interfering RNA (siRNA), as determined by quantitative polymerase chain reaction (qPCR) (mRNA expression relative to con-
trol; *P<0.05; **P<0.01; Mann-Whitney-U test). (D) Treatment with BMP6 (10 ng/mL for 18 h) induced the expression of hepcidin in siControl-transfected cells.
Depletion of SMAD4, but not depletion of UBA6, blunted the BMP6-mediated induction of hepcidin in HepG2-FPN-GFP cells (mRNA expression relative to control;
**P<0.01; Kruskal-Wallis test). (E) Images of cells transfected with siControl, siUBA1 and siUBA6 are shown. Cells were treated with Dox or Dox followed by hepcidin
(4 ng/mL for 18 h), as indicated. Treatment with Dox induced expression of FPN-GFP in siControl treated cells, while incubation with hepcidin caused the degradation
of the fusion protein. Depletion of UBA6, but not UBA1, prevented hepcidin-mediated degradation of the FPN-GFP fusion protein. The location of nuclei in (A) and (E)
are indicated by staining with DAPI (blue). White bar indicates 100 µm.

absence of exogenous hepcidin for 20 min, and protein FPN, the effect of ARIH2 depletion on BMP6-induced FPN-
lysates were immunoprecipitated with an antibody direct- GFP degradation was assessed. In contrast to ARIH1, deple-
ed against NDFIP1. In the absence of hepcidin, a small tion of ARIH2 had no effect on BMP6-mediated degrada-
amount of FPN-GFP was detected in the immunoprecipitat- tion of the FPN-GFP protein expression (Figure 4F; Online
ed protein lysate. Treatment with hepcidin caused an Supplementary Figure S4D).
increase in the amount of FPN-GFP that co-immunoprecip-
itated with NDFIP1 (Figure 3E). Taken together, the results Silencing of Ndfip1 stabilizes hepatic ferroportin
suggest that NDFIP1 interacts with FPN and is involved in in vivo
hepcidin-induced FPN internalization and degradation. The adaptor protein NDFIP1 was identified as a protein
A second adaptor protein (NDFIP2), which like NDFIP1 that is involved in FPN degradation in vitro. To address
facilitates ubiquitination by HECT E3 enzymes, shares whether NDFIP1 is important for FPN degradation in vivo,
79% similarity with NDFIP127. To investigate the potential mice were injected with an AAV2/8 encoding a shRNA
role of NDFIP2 in the regulation of FPN, the effect of directed against Ndfip1, under the control of a U6 promoter.
NDFIP2 depletion on BMP6-induced FPN-GFP degradation The AAV serotype 8 was used in these studies because it
was assessed. While siNDFIP1 treatment prevented degra- has a high efficiency of transduction in hepatocytes.31 In
dation of FPN-GFP, depletion of NDFIP2 had no effect on both AAV2/8-shNdfip1 and AAV2/8-shControl injected ani-
BMP6-mediated degradation of the FPN-GFP (Figure 3F; mals, GFP expression was detected in the liver, indicating
Online Supplementary Figure S3D). successful systemic administration of the virus (Figure 5A).
NDFIP1 recruits members of the NEDD4 family of E3 lig- In animals injected with AAV2/8-shNdfip1, hepatic Ndfip1
ases to target proteins.28 To investigate whether NEDD4 mRNA levels were significantly reduced compared to con-
family members (NEDD4, NEDD4L, ITCH, WWP1, trol animals (Figure 5B). Mice injected with AAV2/8-
WWP2, SMURF1, SMURF2, HECW1, HECW229) regulate shNdfip1 had a 3-fold increase in FPN protein level in the
FPN levels, the localization of FPN-GFP in cells treated with liver compared to control mice (Figure 5C and D). Hamp
siRNA directed against each of these enzymes was exam- mRNA and serum hepcidin levels were similar in both
ined. None of the siRNA directed against members of the groups, suggesting that higher FPN levels were not caused
NEDD4 family, either alone or in pair-wise combinations, by induction of hepcidin (Figure 5E and F). Increased hepat-
prevented BMP6 mediated FPN-GFP degradation (Online ic FPN was associated with a 28% increase in serum iron
Supplementary Figure S4A and B). These results indicate that levels in AAV2/8-shNdfip1, compared to AAV2/8-shControl,
either more than two of these enzymes are involved in mice (Figure 5G) and there was a correlation between
BMP6-induced FPN degradation or additional, as yet serum iron and FPN levels (Online Supplementary Figure
unidentified, enzymes are able to interact with NDFIP1 and S5A). Hepatic FTL levels were increased and TfR1 mRNA
mediate FPN degradation. was decreased in AAV2/8-shNdfip1-treated mice (Online
Supplementary Figure S5B and D). As expected because of the
ARIH1 indirectly regulates ferroportin by inhibiting targeting of AAV8 to the liver,32 splenic Ndfip1 mRNA levels
BMP6-mediated induction of hepcidin were not decreased in AAV2/8-shNdfip1 mice (Online
ARIH1 is a member of the Ariadne family of RBR E3 lig- Supplementary Figure S5E). The results show that the
ases. Treatment of HepG2-FPN-GFP cells with siRNA AAV2/8-mediated depletion of Ndfip1 increases the level of
directed against ARIH1 inhibited BMP6-mediated degrada- hepatic FPN and that Ndfip1 is required for FPN degrada-
tion of FPN-GFP (Figure 3A). ARIH1 was successfully tion in the liver.
depleted by transfection of siARIH1 in both the absence
(Figure 4A) and the presence of BMP6 (Online Supplementary
Figure S4C), as determined by qPCR. The addition of low Discussion
dose exogenous hepcidin to HepG2-FPN-GFP cells, howev-
er, reduced the level of FPN-GFP on the surface of ARIH1- This study identified components of the ubiquitin system
depleted cells (Figure 4B). The ability of exogenous hepcidin that are important for FPN degradation. A HepG2 cell line
to degrade FPN-GFP in siARIH1 treated cells was confirmed that inducibly expresses functional FPN-GFP fusion protein
by immunoblot (Online Supplementary Figure S2C). We con- was established. BMP6-induced expression of hepcidin,
sidered the possibility that depletion of ARIH1 inhibits FPN which caused the internalization and degradation of the
degradation by interfering with the ability of BMP signaling fusion protein and permitted analysis of FPN degradation
to induce hepcidin gene expression. In the absence of under conditions in which the level of hepcidin increases
BMP6, the depletion of ARIH1 reduced basal hepcidin gradually. In vitro, the alternative E1 enzyme UBA6, as well
mRNA levels (Figure 4C). Depletion of siARIH1 impaired as the adaptor protein NDFIP1, were critical for hepcidin-
BMP6-stimulated induction of hepcidin mRNA by 80% induced FPN degradation. Depletion of either UBA6 or
(Figure 4D). ARIH1 depletion also inhibited BMP6-mediat- NDFIP1 inhibited hepcidin-induced internalization and
ed induction of ID1, another target of the BMP signaling degradation of FPN-GFP. The E3 ligase ARIH1 indirectly
pathway (Figure 4E). Interestingly, BMP6-induced phos- regulated FPN stability by altering BMP6-mediated hep-
phorylation of SMAD1/5/8 proteins was not affected by cidin induction through a non-canonical pathway. In vivo,
ARIH1 depletion (Figure 4F). These results suggest that the depletion of Ndfip1 in the murine liver increased the
ARIH1 has an indirect effect on the stability of FPN by alter- level of hepatic FPN and increased circulating iron.
ing BMP6- mediated hepcidin induction through a non- In 2007, UBA6 was identified as a second ubiquitin acti-
canonical pathway. vating E1 enzyme. UBA1 and UBA6 have non-redundant
The Ariadne RBR E3 ligase ARIH2 (also known as functions and each enzyme is essential for biological
TRIAD1) is the closest relative to ARIH1 with 54% similar- processes.33,34 UBA6 is widely expressed in different tissues
ity.30 To consider the possibility that this second member of but contributes to only approximately 1% of overall cellular
the Ariadne family is involved in the indirect regulation of ubiquitination.33,35 In addition to activating ubiquitin for

L. Traeger et al.
subsequent transfer to the E2 enzyme, UBA6 also activates present study UBA6 was found to be the E1 enzyme
the ubiquitin-like protein FAT10, which plays a role in the involved in FPN regulation in vitro; depletion of UBA6, but
immune response, obesity and aging. However, Fat10-defi- not UBA1, prevented hepcidin-induced FPN degradation in
cient mice do not develop iron overload,36,37 suggesting that HepG2 cells. In contrast to UBA1, which is known to
FAT10 does not play a direct role in iron homeostasis. In the charge multiple E2 enzymes with ubiquitin, UBA6 transfers
B C
E F
Figure 3. NDFIP1 interacts with ferroportin and regulates ferroportin degradation. (A) Cells were transfected with siControl, siNDFIP1 or siARIH1 and treated with
doxycycline (Dox) (left panel) or Dox followed by BMP6 (10 ng/mL for 18 hours [h]), as indicated. Small interfering RNA (siRNA) directed against NDFIP1 or ARIH1
prevented BMP6- mediated ferroportin-green fluorescent protein (FPN-GFP) degradation. White bar indicates 100 µm. (B) Transfection with siNDFIP1 successfully
depleted Ndfip1, as determined by quantitative polymerase chain reaction (qPCR) (mRNA expression relative to control; **P<0.01; Student’s t-test). (C) BMP6 (10
ng/mL for 18 h) induced hepcidin expression in siControl transfected cells. Depletion of NDFIP1 did not impair the ability of BMP6 to induce the expression of hep-
cidin mRNA, as determined by qPCR (mRNA expression relative to control; **P<0.01; One-way ANOVA and Student’s t-test). (D) Cells were transfected with siControl
or siNDFIP1 and treated with Dox or Dox followed by hepcidin (4 ng/mL for 18 h) as indicated. In the presence of Dox, the expression of the FPN-GFP fusion protein
was induced. Treatment with hepcidin caused FPN-GFP internalization and its subsequent degradation in siControl-treated cells, but not in siNDFIP1-treated cells.
White bar indicates 100 µm. (E) A low level of FPN-GFP co-immunoprecipitated with NDFIP1 in Dox-treated HepG2-FPN-GFP cells that were not treated with hepcidin.
The level of FPN co-immunoprecipitating with NDFIP1 increased after treatment with hepcidin (50 ng/mL) for 20 minutes (min). Immunoprecipitation was performed
using rabbit anti-NDFIP1 antibody. FPN-GFP was detected using a mouse anti-GFP antibody. The immunoblot is representative of 3 separate experiments. (F) HepG2-
FPN-GFP cells were transfected with siControl-, siNDFIP1- or siNDFIP2 and incubated in the presence (+) or absence (-) of BMP6 (10 ng/mL for 18h). In the absence
of Dox, control cells did not express the FPN-GFP fusion protein. (siRNA) directed against NDFIP1, but not NDFIP2, prevented BMP6-induced degradation of the FPN-
GFP fusion protein, as determined by immunoblot. GAPDH was used as a loading control. The immunoblot is representative of 4 separate experiments.

A B
C D E
Figure 4. ARIH1 regulates BMP6-mediated induction of hepcidin. (A) Transfection with siARIH1 successfully depleted ARIH1 in HepG2-FPN-GFP cells, as determined
by quantitative polymerase chain reaction (qPCR) (mRNA expression relative to control; ***P<0.001; Student’s t-test). (B) HepG2-FPN-GFP cells were transfected
with siControl or siARIH1 and were treated with doxycycline (Dox) or Dox followed by hepcidin (4 ng/mL for 18 hours [h]) as indicated. In the presence of Dox, the
expression of the ferroportin-green fluorescent protein (FPN-GFP) fusion protein was induced. Treatment with hepcidin caused FPN-GFP localization to lysosomes and
its subsequent degradation in siControl-treated cells as well as in siARIH1-treated cells. White bar indicates 100 µm. (C) Treatment of HepG2 cells with siARIH1
reduced the basal expression of hepcidin mRNA, as determined by qPCR (mRNA expression relative to control; ***P≤0.001; Student’s t-test). (D) Pretreatment of
HepG2 cells with small interfering RNA (siRNA) directed against ARIH1 reduced BMP6-mediated hepcidin mRNA expression (relative to control), as determined by
qPCR (***=P≤0.001; One-way ANOVA). (E) BMP6 (10 ng/mL for 18 h) induced the expression of ID1 in siControl transfected cells. Pretreatment of HepG2 cells with
siRNA directed against ARIH1 blunted BMP6 induced expression of ID1, as determined by qPCR (mRNA expression relative to control; **P<0.01; ***P<0.001; One-
way ANOVA). (F) Immunoblot showing levels of FPN-GFP and phosphorylated SMAD1/5/8 in siARIH1- or siARIH2-transfected cells in the presence (+) or absence (-)
of BMP6 (10 ng/mL for 18 h). The level of pSMAD 1/5/8 was increased in all BMP6-treated cells. siRNA directed against ARIH1, but not ARIH2, prevented BMP6-
induced degradation of the FPN-GFP fusion protein. GAPDH was used as a loading control. The immunoblot is representative of 4 separate experiments.

L. Traeger et al.
A B
D E
F G
Figure 5. Ndfip1 regulates ferroportin in vivo. (A) Green fluorescent protein (GFP) was detected in the liver of AAV2/8-shControl- and AAV2/8-shNdfip1-injected mice.
GAPDH was used as a loading control. (B) AAV2/8-shNdfip1 reduced the level of Ndfip1 mRNA in the liver, as determined by quantitative polymerase chain reaction
(qPCR) (mRNA expression relative to control; **P<0.01; Student’s t-test). (C) The level of ferroportin (FPN) in the liver of AAV2/8-shNdfip1-treated mice was increased
compared to AAV2/8-shControl-treated mice, as determined by immunoblot. GAPDH was used as a loading control. (D) Densitometric analysis of immunoblot in (C)
(***P<0.001; Student’s t-test). (E) HAMP mRNA expression was similar in the liver of AAV2/8-shNdfip1-treated mice compared to AAV2/8-shControl-treated mice as
determined by qPCR (mRNA expression relative to control). (F) Serum hepcidin levels were similar in AAV2/8-shNdfip1- and AAV2/8-shControl-treated mice as deter-
mined by enzyme-linked immunosorbant assay. (G) Serum iron levels were increased in mice treated with AAV2/8-shNdfip1 (*P<0.05; Student’s t-test).

ubiquitin to a small number of E2 enzymes.35 Although homeostasis.41–43 NDFIP1 binds to divalent metal trans-
some E2 enzymes interact with both UBA1 and UBA6, one porter 1 (DMT1), the major iron transporter for non-heme
E2 enzyme (USE1, also known as UBE2Z) is exclusively iron import.44 NDFIP1 recruits the NEDD4 family member
charged by UBA635,38. In this study, we found that inhibi- WWP2 to ubiquitinate DMT143. In vivo, Ndfip1 is involved
tion of USE1 did not interfere with hepcidin-induced FPN in the regulation of DMT1 in enterocytes.41 The expres-
degradation (data not shown), indicating that an E2 enzyme sion of Dmt1 in enterocytes of Ndfip1 deficient mice is
other than (or in addition to) USE1 is involved in FPN regu- increased under normal iron conditions as well as during
lation. In UBA6-depleted cells, UBA1 failed to induce hep- iron deficiency. The increased level of Dmt1 leads to
cidin-mediated FPN degradation, indicating non-redundant increased iron absorption, and under normal dietary iron
functions of UBA1 and UBA6 in FPN regulation. The results conditions Ndfip1-deficient mice develop a phenotype
suggest that an as yet unidentified E2 enzyme, exclusively resembling classic hereditary hemochromatosis, with
charged by UBA6, plays a role in FPN degradation. increased hepatic, duodenal and serum iron levels.43,45 In
Members of the NEDD4 family of HECT-type E3 ligases this study we show that depletion of Ndfip1 in the liver
contain a “WW” domain that interacts with a proline rich increased the level of FPN. Ndfip1 appears to regulate two
PPXY (PY) motif in the target protein. However, some target steps in iron metabolism: iron import by DMT1 in entero-
proteins lack a PY domain and ubiquitination of these pro- cytes and iron export by FPN in the liver. Ndfip1-deficient
teins requires the presence of adaptor proteins NDFIP1 or mice were not used in this study, because Ndfip1 deficien-
NDFIP2 to act as a scaffold between the two proteins. cy results in a severe inflammatory phenotype caused by
NDFIP proteins contain three transmembrane domains as hyperactivation of T cells.45,46
well as two PY motifs, which interact with the WW domain In summary, this study demonstrated that the E1
of several members of the NEDD4 family of E3 ligases.28 In enzyme UBA6 and the adaptor protein NDFIP1 are impor-
this study, NDFIP1 was shown to interact with FPN in tant for iron homeostasis, regulating the degradation of
HepG2 cells in vitro and regulates the level of FPN in the liver hepatic FPN. In the future, it may be possible to target spe-
in vivo. None of the WW domain-containing NEDD4 family cific components of the ubiquitin pathway with small
members that were tested individually or in pair-wise com- molecules;47 the results of this study may offer novel
bination prevented BMP6-induced FPN degradation. The approaches to treating disorders of iron metabolism.
results suggest that several NEDD4 family members may
have a redundant role in FPN degradation. Another possibil-
ity is that an as yet unknown E3 ligase interacts with the Disclosures
adaptor protein NDFIP1 to ubiquitinate FPN. No conflicts of interest to disclose.
ARIH1 is a member of the Ariadne family of E3 RBR lig-
ase. ARIH1 is highly expressed in the nucleus, where it Contributions
interacts with Cajal and PML nuclear bodies.39 ARIH1 LT, FW and DBB designed and conceived the study; LT and
associates with neddylated Cullin-RING E3 ligases (CRL) DBB wrote and edited the manuscript; LT, SBW, AJS, BHPC, KP,
and monoubiquitinates CRL targets.40 In this study, AF and DBB performed experiments; LT, RM, AB and DBB ana-
ARIH1 was shown to indirectly regulate FPN stability by lyzed and interpreted the data; DBB and WMZ supervised the
altering BMP6-mediated hepcidin induction through a study. All authors approved the final version for submission.
non-canonical pathway. Depletion of ARIH1 blunted
basal, as well as BMP6-mediated, hepcidin and ID1 Funding
mRNA expression without altering the phosphorylation This study was supported by Luisa Hunnewell and Larry
of SMAD 1/5/8 proteins in response to BMP6. Further Newman (DBB), the German Research Foundation (TR 1642/1-
studies are needed to elucidate the mechanism as to how 1 to LT, WI 5162/2-1 to SBW, Wu 841/1-1 to FW, FI 2429/1-1
ARIH1 regulates hepcidin expression in response to to AF), NIH (R01HL142809 to RM, R01DK125786 to AB) and
BMP6. the American Heart Association (18TPA34230025 to RM)
NDFIP1 was previously shown to have a role in iron 20IOA35360009 to AB).
References synthesized in the liver. J Biol Chem. 2093.

2001;276(11):7806-7810. 11. Drakesmith H, Nemeth E, Ganz T. Ironing
1. Salahudeen AA, Bruick RK. Maintaining 6. Dev S, Babitt JL. Overview of iron metabo- out ferroportin. Cell Metab. 2015;22(5):777-
mammalian iron and oxygen homeostasis: lism in health and disease. Hemodial Int. 787.
sensors, regulation, and cross-talk. Ann N Y 2017;21(Suppl 1):S6-S20. 12. Lin L, Yee SW, Kim RB, Giacomini KM. SLC
Acad Sci. 2009;1177:30-38. 7. Sebastiani G, Wilkinson N, Pantopoulos K. transporters as therapeutic targets: emerging
2. Pantopoulos K, Porwal SK, Tartakoff A, Pharmacological targeting of the opportunities. Nat Rev Drug Discov.
Devireddy L. Mechanisms of mammalian hepcidin/ferroportin axis. Front Pharmacol. 2015;14(8):543-560.
iron homeostasis. Biochemistry. 2016;7:160. 13. Zhang D-L, Ghosh MC, Ollivierre H, Li Y,
2012;51(29):5705-5724. 8. Wang C-Y, Xu Y, Traeger L, et al. Rouault TA. Ferroportin deficiency in ery-
3. Pigeon C, Ilyin G, Courselaud B, et al. A new Erythroferrone lowers hepcidin by seques- throid cells causes serum iron deficiency and
mouse liver-specific gene, encoding a protein tering BMP2/6 heterodimer from binding to promotes hemolysis due to oxidative stress.
homologous to human antimicrobial peptide the BMP type I receptor ALK3. Blood. Blood. 2018;132(19):2078-2087.
hepcidin, is overexpressed during iron over- 2020;135(6):453-456. 14. Qiao B, Sugianto P, Fung E, et al. Hepcidin-
load. J Biol Chem. 2001;276(11):7811-7819. 9. Wunderer F, Traeger L, Sigurslid HH, induced endocytosis of ferroportin is
4. Nicolas G, Viatte L, Lou D-Q, et al. Meybohm P, Bloch DB, Malhotra R. The role dependent on ferroportin ubiquitination.
Constitutive hepcidin expression prevents of hepcidin and iron homeostasis in athero- Cell Metab. 2012;15(6):918-924.
iron overload in a mouse model of sclerosis. Pharmacol Res. 2020;153:104664. 15. Ross SL, Tran L, Winters A, et al. Molecular
hemochromatosis. Nat Genet. 2003;34(1):97- 10. Nemeth E, Tuttle MS, Powelson J, et al. mechanism of hepcidin-mediated ferro-
101. Hepcidin regulates cellular iron efflux by portin internalization requires ferroportin
5. Park CH, Valore EV, Waring AJ, Ganz T. binding to ferroportin and inducing its inter- lysines, not tyrosines or JAK-STAT. Cell
Hepcidin, a urinary antimicrobial peptide nalization. Science. 2004;306(5704):2090- Metab. 2012;15(6):905-917.

L. Traeger et al.
16. Camaschella C, Nai A, Silvestri L. Iron potential role in protein trafficking. J Cell Sci. Sci U S A. 2014;111(14):5313-5318.
metabolism and iron disorders revisited in 2004;117(Pt 16):3679-3689. 38. Wang F, Zhao B. UBA6 and its bispecific
the hepcidin era. Haematologica. 2020;105 28. Harvey KF, Shearwin-Whyatt LM, Fotia A, pathways for ubiquitin and FAT10. Int J Mol
(2):260-272. Parton RG, Kumar S. N4WBP5, a potential Sci. 2019;20(9):2250.
17. Aschemeyer S, Qiao B, Stefanova D, et al. target for ubiquitination by the Nedd4 fami- 39. Elmehdawi F, Wheway G, Szymanska K, et
Structure-function analysis of ferroportin ly of proteins, is a novel Golgi-associated al. Human homolog of drosophila Ariadne
defines the binding site and an alternative protein. J Biol Chem. 2002;277(11):9307- (HHARI) is a marker of cellular proliferation
mechanism of action of hepcidin. Blood. 9317. associated with nuclear bodies. Exp Cell Res.
2018;131(8):899-910. 29. Ingham RJ, Gish G, Pawson T. The Nedd4 2013;319(3):161-172.
18. Pickart CM, Eddins MJ. Ubiquitin: structures, family of E3 ubiquitin ligases: functional 40. Scott DC, Rhee DY, Duda DM, et al. Two
functions, mechanisms. Biochim Biophys diversity within a common modular archi- distinct types of E3 ligases work in unison to
Acta. 2004;1695(1-3):55-72. tecture. Oncogene. 2004;23(11):1972-1984. regulate substrate ubiquitylation. Cell.
19. Morreale FE, Walden H. Types of ubiquitin 30. Kelsall IR, Duda DM, Olszewski JL, et al. 2016;166(5):1198-1214.
ligases. Cell. 2016;165(1):248-248. TRIAD1 and HHARI bind to and are activat- 41. Foot NJ, Leong YA, Dorstyn LE, et al. Ndfip1-
20. hen M, Schmitt S, Buac D, Dou QP. ed by distinct neddylated Cullin-RING ligase deficient mice have impaired DMT1 regula-
Targeting the ubiquitin-proteasome system complexes. EMBO J. 2013;32(21):2848-2860. tion and iron homeostasis. Blood.
for cancer therapy. Expert Opin Ther Targets. 31. Nam H-J, Lane MD, Padron E, et al. Structure 2011;117(2):638-646.
2013;17(9):1091-1108. of adeno-associated virus serotype 8, a gene 42. Foot NJ, Gembus KM, Mackenzie K, Kumar
21. Stewart MD, Ritterhoff T, Klevit RE, Brzovic therapy vector. J Virol. 2007;81(22):12260- S. Ndfip2 is a potential regulator of the iron
PS. E2 enzymes: more than just middle men. 12271. transporter DMT1 in the liver. Sci Rep.
Cell Res. 2016;26(4):423-440. 32. Snyder RO, Miao CH, Patijn GA, et al. 2016;6:24045.
22. Ardley HC, Robinson PA. E3 ubiquitin ligas- Persistent and therapeutic concentrations of 43. Foot NJ, Dalton HE, Shearwin-Whyatt LM,
es. Essays Biochem. 2005;41:15-30. human factor IX in mice after hepatic gene et al. Regulation of the divalent metal ion
23. Weber J, Polo S, Maspero E. HECT E3 ligas- transfer of recombinant AAV vectors. Nat transporter DMT1 and iron homeostasis by
es: a tale with multiple facets. Front Physiol. Genet 1997;16(3):270-276. a ubiquitin-dependent mechanism involving
2019;10:370. 33. Barghout SH, Schimmer AD. E1 Enzymes as Ndfips and WWP2. Blood. 2008;112(10):
24. Fung E, Sugianto P, Hsu J, Damoiseaux R, Therapeutic Targets in Cancer. Pharmacol 4268-4275.
Ganz T, Nemeth E. High-throughput screen- Rev. 2021;73(1):1-56. 44. Yanatori I, Kishi F. DMT1 and iron transport.
ing of small molecules identifies hepcidin 34. Groettrup M, Pelzer C, Schmidtke G, Free Radic Biol Med. 2019;133:55-63.
antagonists. Mol Pharmacol. 2013;83(3):681- Hofmann K. Activating the ubiquitin family: 45. Oliver PM, Cao X, Worthen GS, et al. Ndfip1
690. UBA6 challenges the field. Trends Biochem protein promotes the function of itch ubiqui-
25. Meynard D, Kautz L, Darnaud V, Canonne- Sci. 2008;33(5):230-237. tin ligase to prevent T cell activation and T
Hergaux F, Coppin H, Roth M-P. Lack of the 35. Jin J, Li X, Gygi SP, Harper JW. Dual E1 acti- helper 2 cell-mediated inflammation.
bone morphogenetic protein BMP6 induces vation systems for ubiquitin differentially Immunity. 2006;25(6):929-940.
massive iron overload. Nat Genet. regulate E2 enzyme charging. Nature. 46. Nemeth E, Valore EV, Territo M, Schiller G,
2009;41(4):478-481. 2007;447(7148):1135-1138. Lichtenstein A, Ganz T. Hepcidin, a putative
26. Xiao X, Alfaro-Magallanes VM, Babitt JL. 36. Canaan A, Yu X, Booth CJ, et al. FAT10/diu- mediator of anemia of inflammation, is a
Bone morphogenic proteins in iron home- biquitin-like protein-deficient mice exhibit type II acute-phase protein. Blood.
ostasis. Bone. 2020;138:115495. minimal phenotypic differences. Mol Cell 2003;101(7):2461-2463.
27. Shearwin-Whyatt LM, Brown DL, Wylie FG, Biol. 2006;26(13):5180-5189. 47. Deng L, Meng T, Chen L, Wei W, Wang P.
Stow JL, Kumar S. N4WBP5A (Ndfip2), a 37. Canaan A, DeFuria J, Perelman E, et al. The role of ubiquitination in tumorigenesis
Nedd4-interacting protein, localizes to multi- Extended lifespan and reduced adiposity in and targeted drug discovery. Signal
vesicular bodies and the Golgi, and has a mice lacking the FAT10 gene. Proc Natl Acad Transduct Target Ther. 2020;5(1):11.

Non-Hodgkin Lymphoma ARTICLE
Early detection of T-cell lymphoma with Ferrata Storti Foundation

T follicular helper phenotype by RHOA
mutation analysis
Rachel Dobson,1 Peter Y. Du,1 Lívia Rásó-Barnett,2 Wen-Qing Yao,1 Zi Chen,1
Calogero Casa,2 Hesham EI-Daly,2 Lorant Farkas,2,3 Elizabeth Soilleux,1,4
Penny Wright,4 John W. Grant,4 Manuel Rodriguez-Justo,5 George A. Follows,6
Hala Rashed,7 Margarete Fabre,6,8 E. Joanna Baxter,6 George Vassiliou,6,8
Andrew Wotherspoon,9 Ayoma D. Attygalle,9 Hongxiang Liu2
and Ming-Qing Du1,4
1
Haematologica 2022
Division of Cellular and Molecular Pathology, Department of Pathology, University of Volume 107(2):489-499
Cambridge, Cambridge, UK; 2The Haematopathology and Oncology Diagnostic Service,
Addenbrooke's Hospital, Cambridge University Hospitals NHS Foundation Trust,
Cambridge, UK; 3Department of Pathology, Akershus University Hospital, Lorenskog,
Norway; 4Department of Histopathology, Addenbrooke's Hospital, Cambridge University
Hospitals NHS Foundation Trust, Cambridge, UK; 5Department of Pathology, Royal Free
and University College Medical School, London, UK; 6Department of Haematology,
Addenbrooke's Hospital, Cambridge University Hospitals NHS Foundation Trust,
Cambridge, UK; 7Department of Cellular Pathology, University Hospitals of Leicester,
East Midlands Pathology Services, Leicester, UK; 8Wellcome-MRC Cambridge Stem Cell
Institute, Jeffrey Cheah Biomedical Centre, University of Cambridge, Cambridge, UK
and 9Histopathology Department, The Royal Marsden Hospital, London, UK
ABSTRACT
A
ngioimmunoblastic T-cell lymphoma (AITL) and peripheral T-cell
lymphoma with T follicular helper phenotype (PTCL-TFH) are a
group of complex clinicopathological entities that originate from T
follicular helper cells and share a similar mutation profile. Their diagnosis
is often a challenge, particularly at an early stage, because of a lack of spe-
cific histological and immunophenotypic features, paucity of neoplastic T
cells and prominent polymorphous infiltrate. We investigated whether the
lymphoma-associated RHOA Gly17Val (c.50G>T) mutation, occurring in
60% of cases, is present in the early “reactive” lesions, and whether muta-
tion analysis could help to advance the early diagnosis of lymphoma. The
RHOA mutation was detected by quantitative polymerase chain reaction Correspondence:
with a locked nucleic acid probe specific to the mutation, and a further MING-QING DU
peptide nucleic acid clamp oligonucleotide to suppress the amplification of mqd20@cam.ac.uk
the wild-type allele. The quantitative polymerase chain reaction assay was
highly sensitive and specific, detecting RHOA Gly17Val at an allele fre- Received: July 9, 2020.
quency of 0.03%, but not other changes in Gly17, nor in 61 controls. Accepted: January 22, 2021.
Among the 37 cases of AITL and PTCL-TFH investigated, RHOA Gly17Val
was detected in 62.2% (23/37) of which 19 had multiple biopsies including Pre-published: February 11, 2021.
preceding biopsies in ten and follow-up biopsies in 11 cases. RHOA
Gly17Val was present in each of these preceding or follow-up biopsies https://doi.org/10.3324/haematol.2020.265991
including 18 specimens that showed no evidence of lymphoma by com-
bined histological, immunophenotypic and clonality analyses. The muta-
tion was seen in biopsies 0-26.5 months (mean 7.87 months) prior to the ©2022 Ferrata Storti Foundation
lymphoma diagnosis. Our results show that RHOA Gly17Val mutation Material published in Haematologica is covered by copyright.
analysis is valuable in the early detection of AITL and PTCL-TFH. published material is allowed under the following terms and
conditions:
https://creativecommons.org/licenses/by-nc/4.0/legalcode.
Introduction Copies of published material are allowed for personal or inter-
nal use. Sharing published material for non-commercial pur-
Angioimmunoblastic T-cell lymphoma (AITL), peripheral T-cell lymphoma with https://creativecommons.org/licenses/by-nc/4.0/legalcode,
T follicular helper phenotype (PTCL-TFH) and follicular T-cell lymphoma are a sect. 3. Reproducing and sharing published material for com-
group of complex clinicopathological entities that originate from T follicular helper mercial purposes is not allowed without permission in writing
(TFH) cells. These lymphomas are the most common among various T-cell malig- from the publisher.
nancies in Western countries. Patients with these lymphomas commonly show an
aggressive clinical course, with a median 3-year survival rate of only 30%

R. Dobson et al.
(https://www.hmrn.org/ statistics/disorders/34). The poor clini- nine cases of AITL or PTCL-TFH with known RHOA mutations
cal outcome is largely a consequence of the lack of under- (Gly17Val, c.50G>T, n=6 with variant allele frequency [VAF] rang-
standing of their molecular mechanism and targeted ther- ing from 2-32%; Gly17Leu, c.49-50GG>TT, n=2; Gly17Glu,
apy. c.50G>A, n=1; Ser26Arg, c.76A>C, n=1) were available from our
The diagnosis of these lymphoma entities, particularly previous study,2 while crude DNA samples were available from
at an early stage, is a challenge because of the lack of spe- routine clonality analysis or similarly prepared. The histology of
cific clinical and histological features, low tumor cell con- these specimens was reviewed by specialist hematopathologists.
tent and the presence of prominent polymorphous inflam-
matory infiltrates that often mask the neoplastic cells. To Quantitative polymerase chain reaction with peptide
help the diagnosis, T-cell clonality analysis is commonly nucleic acid clamp and locked nucleic acid probe
used, but this is frequently not helpful because of the The qPCR with peptide nucleic acid (PNA) clamp and locked
paucity of malignant T cells. The diagnostic difficulty is nucleic acid (LNA) probe was performed as previously described.16
further exacerbated by the increasing use of needle core The qPCR assay contained two probes and a PNA clamp (Figure
biopsies to avoid more invasive surgical excision. When a 1A, Online Supplementary Table S1). The total probe served as an
lymphoproliferative lesion is suspected but uncertain on internal control to monitor the PCR performance. The PNA clamp
histological diagnosis, patients are commonly subjected to specifically and strongly binds to the wild-type DNA sequence,
a “watch and wait” approach, and further biopsied when resisting the 5’ nuclease activity of Taq DNA polymerase, and thus
showing signs of disease progression. blocks the wild-type allele from PCR amplification. This results in
There is a wide range of genetic changes in T-cell lym- preferential amplification of the mutant allele which is detected
phomas with TFH phenotype, including early TET2 and specifically by the LNA mutant probe.
DNMT3A mutations, which occur in hematopoietic stem PCR primers with various probes were first tested using puri-
cells, and late genetic changes, which are specifically seen fied DNA samples, with the optimized conditions outlined below
in lymphoma cells.1-4 Among the latter, RHOA mutation (Online Supplementary Table S1). Briefly, the PNA-LNA PCR was
is the most frequent,2,5-9 occurring in 60-70% of cases, carried out in a 20 µL reaction containing 10 mL Premix Ex Taq
with Gly17Val (c.50G>T) accounting for 95% of the (Probe qPCR) Master Mix (Takara, Shiga, Japan), 0.2 mM of each
changes. Interestingly, studies of mouse models have sug- forward and reverse primer, 0.1 µM of each total and mutant
gested that RHOA Gly17Val mutation induces TFH cell probe, 0.05 mM of the PNA clamp probe, and 2 mL of crude DNA
differentiation and, together with loss-of-function TET2 or 25 ng of purified DNA. Real-time PCR was carried out in tripli-
mutations, can promote the development of AITL-like cate using Quantstudio 6 (Thermo Fisher Scientific, Waltham,
lymphomas.10-12 Importantly, RHOA mutation is preferen- MA, USA) with denaturation at 95°C for 30 s followed by
tially associated with the above lymphoma entities and 45 cycles at 95°C for 3 s, and at 62°C for 30 s.
has, so far, been shown only in the neoplastic T-cell pop-
ulation, but not in reactive B and T cells in these malignant Targeted sequencing using the Fluidigm Access Array
conditions.2,5,13,14 Studies to date suggest that the RHOA and Illumina MiSeq
Gly17Val mutation could be used as a marker for the diag- Targeted sequencing was performed on a selected case with
nosis of AITL and PTCL-TFH.6,15 However, it is unclear consecutive tissue biopsies as described in our recent study.2 Each
whether the mutation is present in tissue biopsies prior to of the DNA samples was investigated in duplicate for mutations
the diagnosis of AITL and whether mutation analysis in TET2, DNMT3A, IDH2, RHOA, PLCG1, CCND3, CD28 and
could help in the early detection of these lymphomas. To TNFRSF21 by Fluidigm PCR and Illumina MiSeq sequencing.
investigate these issues, we established a highly sensitive Sequence reads alignment, variant calling, filtering to eliminate
quantitative polymerase chain reaction (qPCR) assay to false positive and benign changes were carried out according to
detect the RHOA mutation, and screened a large cohort our previously established protocols.2 Only the reproducible vari-
(n=37 cases) of AITL and PTCL-TFH with multiple ants that appeared in both replicates were regarded as true
sequential biopsies together with 61 controls. changes.
BaseScope in situ hybridization

Methods This was performed on a selected case. The sequence of the
clonal TRB rearrangement in a case of AITL was available from a
Tissue materials and DNA extraction recent study.2 Based on the unique VDJ junctional sequence,
The use of archival tissues for research was approved by the unique BaseScope probes were designed and used to identify the
ethics committees of the institutions involved. In total, 78 tissue lymphoma T cells by in situ hybridization. The BaseScope in situ
specimens from 37 patients (multiple consecutive specimens in 29 hybridization was carried out according to the manufacturer's
cases) with AITL (n=35) or PTCL-TFH (n=2), together with 61 instructions (Advanced Cell Diagnostics, Newark, CA, USA),2
controls (13 with reactive lymph nodes with paracortical expan- with the conditions optimized using the AITL specimen from
sion of T cells, 10 with classic Hodgkin lymphoma and 1 with which the clonal TRB rearrangement was sequenced. Tonsils were
marginal zone lymphoma with a prominent background of T used as a negative control.
cells) and peripheral blood DNA samples from 16 individuals
(TET2 and DNMT3A mutations in 2, TET2 mutation in 4,
DNMT3A mutation in 5) with clonal hematopoiesis of indetermi- Results
nate potential (CHIP) (1 individual classed as having high-risk
CHIP) were successfully investigated. DNA samples were pre- Peptide nucleic acid – locked nucleic acid quantitative
pared from a whole tissue section of each specimen. Purified DNA polymerase chain reaction is highly sensitive and
was obtained using a QIAamp DNA-Micro kit (Qiagen), while specific for the detection of RHOA Gly17Val (c.50G>T)
crude DNA was obtained by digesting samples at 37°C overnight Upon the optimization of PCR conditions for RHOA
with proteinase K and NP-40 buffer. Purified DNA samples from mutation detection, we first determined the sensitivity of

RHOA mutation analysis in early lymphoma diagnosis
Figure 1. Detection of RHOA p.Gly17Val (c.50G>T) by real time polymerase chain reaction with peptide nucleic acid/locked nucleic acid probes. (A) Schematic
illustration of the real time polymerase chain reaction (PCR) design. The total probe is used as an internal control to monitor the PCR performance. The peptide
nucleic acid (PNA) clamp binds to the wild-type sequence, resists the 5’ nuclease activity of Taq DNA polymerase and thus blocks the wild-type allele from PCR ampli-
fication. This results in preferential amplification of the mutant allele which is detected specifically by the locked nucleic acid (LNA) mutant probe. (B) The PCR assay
is highly sensitive, capable of detecting the RHOA mutation at a variant allele frequency (VAF) of 0.03% based on serial dilutions of an angioimmunoblastic T-cell
lymphoma sample with known mutation allele frequency by next-generation sequencing (left panel). For simplicity, only the LNA mutant but not total probe signals
are shown. The PCR assay shows a linear correlation among the serial dilutions (right panel).
the qPCR assay using serial dilutions of three purified essentially the poor quality of the DNA, as evidenced by
DNA samples with known VAF of Gly17Val (c.50G>T) quality control PCR.
into tonsil DNA. The PNA-LNA qPCR was highly sensi-
tive, capable of detecting the mutation at a VAF of 0.032% RHOA Gly17Val (c.50G>T) is detected in initial
(Figure 1B). As expected, the qPCR was highly specific to biopsies not diagnostic for lymphoma by conventional
Gly17Val (c.50G>T) and showed no detectable signal of approaches
the LNA mutant probe with the DNA samples harboring To confirm that the RHOA mutation was lymphoma-
other RHOA mutations including Gly17Leu specific, not associated with reactive conditions prolifer-
(c.49-50GG>TT) and Gly17Glu (c.50G>A), or tonsillar ations or other lymphoproliferative disorders with
DNA (Online Supplementary Figure S1A). enriched T cells, we investigated 45 tissue biopsies
As crude DNA preparations were routinely used for including 27 from lymph nodes, for which T-cell clonali-
clonality analysis in our clinical diagnostic laboratories, ty analysis was requested during routine histological
we tested whether such crude DNA samples were diagnosis, but showed no evidence of a T-cell lymphoma.
amenable to the above PNA-LNA qPCR (Online These included 13 specimens showing paracortical T-cell
Supplementary Figure S1C). Of the ten crude DNA sam- expansion or enriched CD4+ T cells, and ten classic
ples initially tested, nine yielded excellent amplification Hodgkin lymphomas with a prominent background of T
comparable with the results from purified DNA samples. cells. The qPCR was successful for all these specimens, as
We then used crude DNA preparations for the qPCR in indicated by the total probe control, but they were all
the remaining investigations, with high quality results negative for the RHOA mutation. Similarly, we also
from 139 of the 144 samples investigated. The reason showed absence of the RHOA mutation in peripheral
that the other five samples failed to support qPCR was blood lymphocytes (Online Supplementary Figure S1D)

R. Dobson et al.
Figure 2. Analysis of RHOA (c.50G>T; p.Gly17Val) mutation by quantitative polymerase chain reaction in the RHOA-positive cases with longitudinal biopsies in
patients with angioimmunoblastic T-cell lymphoma (only including cases with at least one biopsy classed as not diagnostic [“group B” or “group C”]). The diagnoses
of these specimens were reviewed, and categorized into three groups: group A (filled circles): lymphoma diagnosed by histology and immunophenotype, further sup-
ported by clonal TCR gene rearrangement; group B (half-filled circles): lymphoma not diagnostic by histology and immunophenotype alone, but ascertained by clonal
TCR gene rearrangement; group C (open circles): lymphoma not diagnosed by combined analyses. Red (regardless of the symbol: including outline only, half-filled,
and completely filled symbols) indicates RHOA p.Gly17Val mutant positive biopsy. Gray indicates RHOA p.Gly17Val status unknown. E denotes extranodal biopsies.
*denotes lymph node excision biopsies and all others are core biopsy specimens. An open square denotes a diagnosis of classic Hodgkin lymphoma with no apparent
evidence of angioimmunoblastic T-cell lymphoma (AITL). Case 38 lacks a final diagnosis as the patient died, indicated by a dashed line. Cases with only group A mul-
tiple biopsies are not included in this figure.
from 16 individuals with CHIP (mutational information and clonality analyses (Online Supplementary Table S3).
provided in Online Supplementary Table S2), in keeping Interestingly, RHOA mutation was detected in each of
with the findings from whole exome or panel these specimens, in the positive cases, including those not
sequencing.17,18 diagnostic for lymphoma by conventional integrated
We then investigated 37 cases of AITL (n=35) or nodal diagnostic investigations (Figure 2).
PTCL-TFH (n=2) with unknown RHOA mutation status,
of which 29 had multiple longitudinal biopsies (2-5) avail- Characteristics of biopsies not diagnostic for
able for RHOA analysis. RHOA mutation was seen in 23 lymphoma but positive for RHOA mutation
cases, but was negative in the remaining 14 cases. Among Of the 23 cases of AITL or PTCL-TFH with RHOA
the 23 cases with RHOA mutation, 19 cases had multiple mutation, the initial biopsy was not diagnostic in ten
biopsies. We reviewed the original histological diagnosis cases including two with an excisional lymph node spec-
in each of these specimens and categorized them into imen. The time from the initial non-diagnostic biopsy to
three groups. Group A specimens (n=26) showed clear that establishing AITL or PTCL-TFH ranged from 0 to
evidence of lymphoma by histology and immunopheno- 26.5 months, with an average of 7.87 months. All these
type, further supported by clonal TCR rearrangement. initial non-diagnostic biopsies showed polyclonal (n=4),
Group B (n=6) had suspicious histological and weak clonal (n=5) or oligoclonal (n=2) TCR gene
immunophenotypic findings, although not entirely diag- rearrangements or failed (n=1) by BIOMED-2 TRB and
nostic for lymphoma, which was ascertained by detec- TRG clonality analysis. Seven of the initial non-diagnostic
tion of strong clonal TCR gene rearrangements. Finally, biopsies were also subjected to BIOMED-2 B-cell clonali-
group C mutations (n=18) were not diagnostic for lym- ty analysis and six showed a clonal IG gene rearrange-
phoma by combined histological, immunophenotypic ment. Epstein-Barr virus (EBV)-encoded small RNA

Figure 3. Histological and immunophenotypic findings in case 30. The first lymph node excision biopsy shows partial effacement of the lymph node architecture by
polymorphous infiltrates, particularly B cells with plasmacytoid differentiation. A large proportion of B cells were EBER-positive and showed IG κ light chain restriction
(not shown). The lymphoid follicles appear to be reactive and shows no apparent expansion of T follicular helper (TFH) cells, with only a few CD10-positive cells spilling
out of the germinal center, consistent with pattern-1 histology of angioimmunoblastic T-cell lymphoma (AITL). The second lymph node biopsy shows effacement of
the lymph node architecture by medium-sized atypical lymphoid cells with regressed follicles. There is a prominent proliferation of follicular dendritic cell meshworks
and high endothelial venules with atypical lymphoid cells clustered in their vicinity. The atypical lymphoid cells are T cells expressing TFH markers, spilling out of the
germinal center to the interfollicular region. EBER in situ hybridization shows only scattered positive cells.

R. Dobson et al.
CD3
CD4 PD1 BCL6
CD20 CD30 EBER
CD30
CD20 OCT2 CD4
Figure 4. Histological and immunophenotypic findings in case 7. The first biopsy shows a subcutaneous perivascular infiltrate of CD3+ T cells with vasculitic features.
The third biopsy shows cardinal features of angioimmunoblastic T-cell lymphoma (AITL) and also a prominent pleomorphic infiltrate including an EBER-positive B-cell
population with Hodgkin and Reed/Sternberg (HRS)-like morphology and immunophenotype. The fifth biopsy shows no apparent evidence of AITL, but a polymor-
phous infiltrate with a more prominent EBER-positive B-cell population that has HRS cell morphology and immunophenotype, rosetting by CD4+ T cells.

Table 1. Summary of histological, immunohistochemical, clonality analysis and genetic findings in case 7.
First biopsy Second biopsy Third biopsy Fourth biopsy Fifth biopsy
Time after 11 months later 11.5 months later 33 months later 35 months later
the 1st biopsy
Clinical Skin rash, fevers; Episode of non-specific symptoms; Slowing enlarging lymph node
presentation All self-resolved lymphadenopathy revealed by CT, with increasing symptoms.
with no therapy but relatively low PET signal; Treated with 6 cycles of R-GCVP*
opted for close surveillance. and achieved CR;
Ongoing remission in May 2020
Biopsy site Skin punch Lymph node core Left axillary lymph Right axillary lymph Right axillary lymph
biopsy biopsy node core biopsy node core biopsy node core biopsy
Histology Mild perivascular Predominant infiltrate Very similar Lymph node Normal structure
and infiltrate, predominantly of small to medium to the 2nd biopsy. structure effaced by largely effaced
Immuno-phenotype CD3+ T cells, admixed sized T cells Predominantly polymorphous infiltrate by polymorphous
histiocytes and B cells with vasculocentric small to medium with scattered infiltrate with
pattern, CD4+, sized T cells large atypical cells prominent large
some CD10+ and BCL6+ associated with expressing CD30, atypical cells
T cells possibly HEV proliferation, CD20 (weak), expressing CD30,
spilled outside B-cell CD4+, PD1+, CD79, PAX5, BCL6, CD20, CD79a,
follicle, expanded BCL6 variable+; MUM1, OCT2, BOB1, PAX5, MUM1,
FDC meshworks; Expanded FDC and EBER+, BCL6(weak),
scattered large and pleomorphic but not CD15. and EBER+, but
EBER+ B cells: infiltrate, CD3 T cells: not CD15, CD10,
CD30+, CD15 weak+ scattered large largely CD4 ,+
CYCLIN D1, ALK
B cells with HRS some ICOS+, and CD25.
morphology CD30+, occasional PD1+, but No detectable
CD15 weak+ CD10– and CXCL13– T-cell abnormalities
Diagnosis Panniculitis AITL AITL Classic Hodgkin Classic Hodgkin
lymphoma lymphoma
T-cell clonality
TRG-A n/a 217 bp 217 bp Poly Poly
TRG-B n/a Poly Poly Poly Poly
TRB-A n/a 245 bp 245 bp Poly Poly
TRB-B n/a Poly Poly Poly Poly
TRB-C n/a 302 bp Poly Poly 187bp
B-cell clonality n/a n/a n/a Poly Poly
BaseScope-ISH Few positive cells Diffuse positive Diffuse positive n/a Negative
for TRB-V5-
Targeted sequencing n/a DNMT3A (VAF: 34%) n/a n/a DNMT3A (VAF:20%)
(c.920C>T; p.P307L) (c.920C>T; p.P307L)
TET2 (VAF: 8%) TET2 (VAF: 4%)
(c.3646C>T; p.R1216X) (c.3646C>T; p.R1216X)
TET2 (VAF: 36%) TET2 (VAF: 20%)
(c.3781C>A; p.R1261S) (c.3781C>A, pR1261S)
TET2 (VAF: 3%) TET2 (VAF: 8%)
(c.3866G>T; p.C1289F) (c.3866G>T; p.C1289F)
TET2 (VAF: 18%)
(c.4947T>A; p.Y1649X)
RHOA (VAF: 20%) RHOA (VAF: 1%)
(c. 50G>T; p.G17V) (c.50G>T; p.G17V)
RHOA c.50G>T Weak positive Strong positive Strong positive Weak positive Weak positive
n/a: not available; R-GCVD: rituximab, gemcitabine, cyclophosphamide, vincristine, prednisolone; CR: complete remission; VAF: variant allele frequency.

R. Dobson et al.
(EBER) in situ hybridization was carried out in eight cases Retrospective analysis showed the presence of RHOA
during the routine diagnostic workup, and four showed Gly17Val mutation in the above two specimens by qPCR
variable positivity from a few scattered to numerous and targeted sequencing with 3% and 23% VAF in the
EBER-positive cells, of which three displayed clonal IG first and second biopsy, respectively. The targeted
gene rearrangement. In four cases, EBV-driven prolifera- sequencing also revealed a pathogenic nonsense substitu-
tion or classic Hodgkin lymphoma were considered in the tion in DNMT3A (c.2311C>T, p.R771*) in the first and
initial diagnosis. second biopsy with 17% and 26% VAF, respectively.
Six follow-up biopsies were not diagnostic of involve- Case 7. An 82-year-old man, with remission of a previ-
ment by AITL or PTCL-TFH by routine diagnostic inves- ous mantle cell lymphoma, presented with a skin rash on
tigations; these included two bone marrow and two skin the right calf. A punch biopsy showed a mild perivascu-
specimens. In each specimen, a CD4+ T-cell infiltrate was lar infiltrate by CD3+ T cells, with a histological diagnosis
noted, but a definite aberrant immunophenotype and of panniculitis (Figure 4, Table 1). To investigate potential
expression of TFH markers could not be ascertained. T- lymphoma relapse, lymph node core biopsies were taken
cell clonality analysis showed weak polyclonality in two at 11 and 12 months of follow-up: neither specimen
and a weak clonal or oligoclonal pattern in two. showed evidence of mantle cell lymphoma, but both
revealed cardinal features of AITL and also a prominent
Representative cases pleomorphic infiltrate including an EBER-positive B-cell
Case 30. A 79-year-old man presented with bilateral population with Hodgkin and Reed/Sternberg (HRS)-like
tender neck lymph nodes, and had mild sweats, but no morphology and immunophenotype (Figure 4, Table 1).
weight loss or fever. Clinical examination revealed multi- T-cell clonality analysis demonstrated clonal rearrange-
ple bilaterally enlarged neck and groin lymph nodes (up ments by TRG-A and TRB-A with identical sized ampli-
to 1.5 cm diameter), and palpable liver and spleen. Right fied products between the two biopsies, and an addition-
level II neck lymph node excision biopsy showed partial al clonal rearrangement by TRB-C in the second biopsy.
effacement of the lymph node architecture and expansion Further follow-up biopsies were taken at 25 and 26
of the interfollicular area by a polymorphous population months, and neither showed apparent evidence of AITL,
of lymphoid cells, including B cells with plasmacytoid dif- but both had polymorphous infiltrates with a more
ferentiation, and scattered large cells (Figure 3). These B prominent EBER-positive B-cell population that had HRS-
cells expressed pan B-cell markers (CD20, CD79a, CD19), cell morphology and immunophenotype (Figure 4, Table
and MUM1, but were negative for CD10 and BCL6 (data 1). Neither specimen showed any evidence of the clonal
not shown). A high proportion of the B cells were EBER- TRG/TRB rearrangements seen in the early lymph node
positive and showed IG k light chain restriction. The lym- biopsies, although a fifth biopsy displayed an isolated
phoid follicles appeared to be reactive and showed no clonal rearrangement by TRB-C. B-cell clonality analyses
apparent expansion of TFH cells, with only a few CD10- demonstrated polyclonal IG gene rearrangements in both
positive cells spilling out of the germinal centers (Figure specimens (Figure 4, Table 1). A classic Hodgkin lym-
3). BIOMED-2 clonality analyses showed clonal IGH and phoma arising from the EBV-positive B-cell component of
IGK gene rearrangements, but a weak oligoclonal pattern the AITL was considered.
with TRG and TRB. The histological diagnosis was In a retrospective study, the TRB-A and B PCR products
uncertain: a clonal EBV-positive polymorphous lympho- from the second biopsy were sequenced using an Illumina
proliferation was considered. Close follow-up with a low MiSeq platform and a dominant TRBV5-J2 rearrangement
threshold for re-biopsy was recommended. (86%) was identified.2 Based on the unique VDJ junctional
Two months later, the patient presented with increasing sequence, we designed unique BaseScope probes to iden-
fatigue, fever, maculopapular chest rash and increased size tify the lymphoma T cells by in situ hybridization (Online
of peripheral lymphadenopathy. Positron-emission Supplementary Figure S2). As expected, both the second
tomography (PET) scan revealed extensive bilateral cervi- and third biopsies with an AITL diagnosis showed diffuse
cal, mediastinal, bilateral iliac and groin lymphadenopa- positivity with the lymphoma clone-specific probe.
thy. Left level V neck lymph node excision biopsy showed Interestingly, the initial skin biopsy also displayed isolated
partial effacement of the lymph node architecture by an positive cells, while the fifth biopsy with a diagnosis of
infiltrate of medium-sized atypical lymphoid cells with EBV lymphoproliferative disease (LPD) gave a negative
regressed follicles (Figure 3). There was hyperplasia of fol- result. Both the second (AITL) and fifth (EBV-LPD) biop-
licular dendritic cell meshworks and high endothelial sies were investigated by panel sequencing for recurrent
venules with the atypical lymphoid cells clustered in their somatic mutations, and this identified five shared muta-
vicinity. The atypical lymphoid cells were positive for tions – one DNMT3A, three TET2 and one RHOA changes
CD3, CD5 and TFH markers (PD1, CD10, ICOS, BCL6) – between the two specimens, and one further TET2
(Figure 3). EBER in situ hybridization revealed only scat- mutation only in the second specimen (Table 1). In gener-
tered positive cells. BIOMED-2 clonality analyses showed al, the mutation load in each of the above shared changes
clonal TRB and TRG gene rearrangements, and also weak was much higher in the second biopsy (AITL) than in the
clonal IGH and IGK gene rearrangements which were dif- fifth biopsy (EBV-LPD). This was particularly striking for
ferent from those of the previous biopsy in the size of the RHOA Gly17Val mutation, with a VAF of 20% in the
their amplified IG products. A diagnosis of AITL was second biopsy but of only 1% in the fifth biopsy. As
made. The patient was initially treated with six cycles of expected, qPCR for the RHOA mutation demonstrated its
CHOP (cyclophosphamide, doxorubicin, vincristine, and strong positivity in both the second and third biopsies
prednisone), then avelumab (cycle 12 at the most recent showing AITL, but a weak positive signal in the initial
follow-up) under the AVAIL-T trial, and was well, show- skin biopsy, and the fourth and fifth biopsies displaying
ing no constitutional symptoms or palpable cervical classic Hodgkin lymphoma-like EBV-LPD (Figure 4, Online
lymph nodes, 20 months following the AITL diagnosis. Supplementary Figure S2).

Discussion context of histological and immunophenotypic findings. If

a histological diagnosis of AITL/PTCL-TFH cannot be
By using a highly sensitive qPCR assay, we confirmed made, an excision biopsy or a low threshold for an early
that RHOA Gly17Val (c.50G>T) mutation is specifically follow up biopsy, when appropriate, is indicated.
associated with AITL or PTCL-TFH, but is not found in In a few case studies, RHOA mutation was detected in
other lymphoproliferative conditions including those with circulating free DNA or peripheral blood lymphocytes
florid infiltration of T-helper cells. More importantly, from patients with AITL/PTCL-TFH,15 and the mutation
detection of RHOA Gly17Val (c.50G>T) mutation could burden appeared to correlate with the treatment
help early detection of AITL and PTCL-TFH, and also outcome.23,24 It remains to be investigated whether the
diagnosis of their extranodal involvement. mutation is detectable in circulating free DNA or periph-
The difficulty in making a histological diagnosis of AITL eral blood lymphocytes at the time of initial non-diagnos-
or PTCL-TFH is well recognized, particularly when the tic biopsies. Given that the RHOA mutation burden in cir-
neoplastic cell content is low, inconspicuous by histologi- culating free DNA and peripheral blood lymphocytes
cal and immunophenotypic assessment and undetectable reflects, at least theoretically, the overall lymphoma load,
by analysis of TCR gene rearrangements. Apart from the simultaneously analyzing blood samples, in addition to
paucity of neoplastic T cells, the polymorphous infiltrate, tissue biopsy, could add further value in lymphoma diag-
particularly the presence of numerous EBV-positive B cells, nosis, particularly when a specimen is not representative.
including HRS-like cells (EBV-positive or -negative), may Apart from the initial biopsies, the diagnosis of extra-
lead to diagnostic consideration of an EBV-driven B-cell nodal involvement by AITL/PTCL-TFH such as skin and
proliferation.19-21 Such misleading diagnostic features may bone marrow is also difficult for reasons similar to those
also be reinforced by the frequent demonstration of clonal discussed above, including low tumor cell content and
IG gene rearrangement. As shown in this study, EBV-dri- polymorphous infiltrate.25 In particular, extranodal infil-
ven B-cell proliferation was considered as a diagnosis in trates may not harbor the cardinal features of AITL such
four of 11 of the initial biopsies. Indeed, it has been report- as TFH marker expression in the neoplastic T cells and
ed by several independent studies that EBV-associated follicular dendritic cell meshworks. The DNA sample
lymphoproliferation is a frequent pitfall in the diagnosis of from these extranodal sites is often not informative for
AITL.19-22 clonality analysis because of insufficient lymphoid cells
In the present study, the majority of the initial speci- and/or poor DNA quality. As shown in our study, RHOA
mens that were not diagnostic were core biopsies. It is mutation analysis is highly valuable for confirming
possible that these core biopsies were not totally represen- AITL/PTCL-TFH involvement in follow-up biopsies.
tative, with characteristic lymphoma components missed Similarly, RHOA mutation analysis is valuable in the
because of sampling errors. Nevertheless, two initial non- diagnosis of AITL/PTCL-TFH from cytological speci-
diagnostic biopsies were excisional lymph node specimens.9
mens, indicating that an absence of diagnostic features of Aside from the RHOA mutation, there are several other
AITL/PTCL-TFH in the initial biopsies was also a real genetic changes including IDH2, CD28, PLCG1, VAV1 and
issue. Interestingly, the time interval between the initial TNFRSF21 mutations, and VAV1-STAP2, CTLA4-CD28
non-diagnostic biopsies and the follow-up biopsies that and ITK-SYK fusions, which occur at variable frequencies
established the diagnosis varied considerably, ranging in AITL and PTCL-TFH.2,26-28 Of note, mutations in VAV1,
from 0 to 26.5 months. In the cases with a long interval, a signaling molecule downstream of TCR, occurs in 8.2%
it is likely that the initial biopsy represented an early pre- of AITL and appears to be mutually exclusive of the
malignant lesion, while the follow-up biopsy reflected RHOA mutation.28 Detection of these additional lym-
more progressed disease, thus having more cardinal fea- phoma-associated genetic changes could also help the
tures for making the AITL/PTCL-TFH diagnosis. While in early detection of AITL/PTCL-TFH, together with RHOA
the cases with a short interval, it is possible that the mutation detection, potentially being valuable for diagno-
enlarged lymph nodes were variably involved by sis in up to 90% of these T-cell lymphomas. These diverse
AITL/PTCL-TFH and the initial non-diagnostic biopsies genetic changes could be readily investigated by targeted
represented early involvement by the lymphoma. sequencing either alone or as part of a comprehensive
Both the above possibilities may exist, without exclud- lymphoma panel.
ing the other. It is impossible to distinguish the two sce- The detailed analyses of multiple biopsies in case 7 also
narios based on the analysis of a single biopsy, including revealed the mutation burden in non-malignant T cells.
RHOA mutation analysis. Nonetheless, detection of The fifth lymph node biopsy showed little involvement
RHOA Gly17Val (c.50G>T) mutation can certainly raise by AITL as the lymphoma clone was undetectable by
the alarm to perform more in-depth histological and BaseScope in situ hybridization and the VAF of the RHOA
immunophenotypic investigations, for example a more mutation was only 1%, but the specimen had a high TET2
careful search for evidence of TFH cell expansion, as mutation burden (20% VAF). This implies that the TET2
shown in the first biopsy in case 30. It is important to mutation must have been present in a large proportion of
emphasize that detection of a RHOA mutation is not reactive B and T cells,1,5,29-31 probably up to 40% of the
equivalent to a diagnosis of lymphoma because mutation- total cell population, which is well above the EBER-posi-
al analysis by qPCR or targeted sequencing is highly sen- tive cell fraction (~5%). Hence, this enlarged lymph node
sitive, and the mutation is seen in biopsies without histo- was essentially caused by lymphoid proliferation driven
logical evidence of AITL. As discussed above, the RHOA by the TET2 mutation and/or EBV infection. These find-
mutation-positive non-diagnostic biopsy may represent a ings further highlight the markedly variable histological
premalignant lesion or early involvement by the lym- presentation of enlarged lymph nodes in patients with
phoma. Therefore, RHOA mutation analysis should be AITL and, hence, the danger of potential sampling errors
used as an auxiliary tool with the results interpreted in the in the diagnosis of AITL. In this context, it is pertinent to

R. Dobson et al.
investigate RHOA and other lymphoma-associated muta- Disclosures

tions in patients with advanced age and biopsies showing No conflicts of interest to disclose.
EBV-positive polymorphous infiltrates, to explore how
such mutation analyses can help to circumvent this diag- Contributions
nostic pitfall. At the genetic level, the high number of RD, PD, WQ, and CC designed the experiments and collected
TET2 mutations and their frequent presence in non-neo- and analyzed data; WQ and ZC performed the Illumina
plastic T cells suggest that the patient most likely had an sequencing analysis and variant calling; LRB, HED, LF, ES,
underlying CHIP with TET2 and DNMT3A mutations that PW, JWG, MRJ, GAF, HR, AW, ADA, HL, MF, EJB, and GV
occurred in hematopoietic stem cells, consequently contributed cases and pathology review; MQD and RD wrote
extending to the progenies of these cells. and prepared the manuscript; MQD was responsible for
In general, patients with AITL typically have an aggres- research funding, study design and coordination. All authors
sive clinical course and respond poorly to currently avail- commented on the manuscript and approved its submission for
able therapies. Case 7 appeared to be an exception to this publication.
rule, showing slow disease progression in the absence of
any treatment, although it is not possible to rule out a Acknowledgments
response to the R-GCVP (rituximab, gemcitabine, The authors thank Shubha Anand and Yuanxue Huang for
cyclophosphamide, vincristine, prednisolone) that was their assistance with using TapeStation, Graeme Clark and
aimed at treating the EBV-associated proliferations (Table Ezequiel Martin for their assistance with Illumina sequencing
1). An indolent clinical course has been previously report- and Fangtian Wu for carrying out DNA extraction on a control
ed for some patients with AITL showing pattern-1 histol- case.
ogy, including those treated only with steroids.22,32 It
remains to be investigated how to identify such indolent Funding
cases at the time of diagnosis and stratify their clinical The research in MQD's laboratory was supported by grants
management accordingly, particularly in view of the from Blood Cancer UK (13 006, 15 019), CRUK
advance in early detection of AITL. (C8333/A29707), and the Kay Kendall Leukaemia Fund
In summary, we have shown that RHOA mutation is (KKL582) UK. WY was supported by a research fellowship
specifically associated with AITL, PTCL-TFH and their from the China Scholarship Council, and an International
related lesions, and investigation of the mutation is highly Collaborative Award from the Pathological Society of Great
valuable in early detection of these T-cell lymphomas and Britain and Ireland, UK. The Human Research Tissue Bank is
their extranodal involvement. supported by the NIHR Cambridge Biomedical Research Centre.
References G17V mutation in angioimmunoblastic T- munoblastic T-cell lymphoma. Cancer Sci.

cell lymphoma: a potential biomarker for 2018;109(5):1682-1689.
1. Nguyen TB, Sakata-Yanagimoto M, Asabe cytological assessment. Exp Mol Pathol. 17. Jaiswal S, Fontanillas P, Flannick J, et al.
Y, et al. Identification of cell-type-specific 2019;110:104294. Age-related clonal hematopoiesis associat-
mutations in nodal T-cell lymphomas. 10. Zang S, Li J, Yang H, et al. Mutations in 5- ed with adverse outcomes. N Engl J Med.
Blood Cancer J. 2017;7(1):e516. methylcytosine oxidase TET2 and RhoA 2014;371(26):2488-2498.
2. Yao WQ, Wu F, Zhang W, et al. cooperatively disrupt T cell homeostasis. J 18. Lewis NE, Petrova-Drus K, Huet S, et al.
Angioimmunoblastic T-cell lymphoma Clin Invest. 2017;127(8):2998-3012. Clonal hematopoiesis in angioim-
contains multiple clonal T-cell populations 11. Cortes JR, Ambesi-Impiombato A, munoblastic T-cell lymphoma with diver-
derived from a common TET2 mutant pro- Couronné L, et al. RHOA G17V induces T gent evolution to myeloid neoplasms.
genitor cell. J Pathol. 2020;250(3):346-357. follicular helper cell specification and pro- Blood Adv. 2020;4(10):2261-2271.
3. Iqbal J, Amador C, McKeithan TW, Chan motes lymphomagenesis. Cancer Cell. 19. Steciuk MR, Massengill S, Banks PM. In
WC. Molecular and genomic landscape of 2018;33(2):259-273. immunocompromised patients, Epstein-
peripheral T-cell lymphoma. Cancer Treat 12. Ng SY, Brown L, Stevenson K, et al. RhoA Barr virus lymphadenitis can mimic
Res. 2019;176:31-68. G17V is sufficient to induce autoimmunity angioimmunoblastic T-cell lymphoma mor-
4. Xie M, Lu C, Wang J, et al. Age-related and promotes T-cell lymphomagenesis in phologically, immunophenotypically, and
mutations associated with clonal mice. Blood. 2018;132(9):935-947. genetically: a case report and review of the
hematopoietic expansion and malignan- 13. Dobay MP, Lemonnier F, Missiaglia E, et al. literature. Hum Pathol. 2012;43(1):127-133.
cies. Nat Med. 2014;20(12):1472-1478. Integrative clinicopathological and molecu- 20. Nicolae A, Pittaluga S, Venkataraman G, et
5. Sakata-Yanagimoto M, Enami T, Yoshida K, lar analyses of angioimmunoblastic T-cell al. Peripheral T-cell lymphomas of follicular
et al. Somatic RHOA mutation in angioim- lymphoma and other nodal lymphomas of T-helper cell derivation with Hodgkin/
munoblastic T cell lymphoma. Nat Genet. follicular helper T-cell origin. Reed-Sternberg cells of B-cell lineage: both
2014;46(2):171-175. Haematologica. 2017;102(4):e148-e151. EBV-positive and EBV-negative variants
6. Nakamoto-Matsubara R, Sakata- 14. Yoo HY, Sung MK, Lee SH, et al. A recur- exist. Am J Surg Pathol. 2013;37(6):816-826.
Yanagimoto M, Enami T, et al. Detection of rent inactivating mutation in RHOA 21. Laforga JB, Gasent JM, Vaquero M.
the G17V RHOA mutation in angioim- GTPase in angioimmunoblastic T cell lym- Potential misdiagnosis of angioim-
munoblastic T-cell lymphoma and related phoma. Nat Genet. 2014;46(4):371-375. munoblastic T-cell lymphoma with
lymphomas using quantitative allele-specif- 15. Hayashida M, Maekawa F, Chagi Y, et al. Hodgkin's lymphoma: a case report. Acta
ic PCR. PLoS One. 2014;9(10):e109714. Combination of multicolor flow cytometry Cytol. 2010;54(5 Suppl):840-844.
7. Palomero T, Couronne L, Khiabanian H, et for circulating lymphoma cells and tests for 22. Tan LH, Tan SY, Tang T, et al.
al. Recurrent mutations in epigenetic regu- the RHOA(G17V) and IDH2(R172) hot- Angioimmunoblastic T-cell lymphoma
lators, RHOA and FYN kinase in peripheral spot mutations in plasma cell-free DNA as with hyperplastic germinal centres (pattern
T cell lymphomas. Nat Genet. 2014; liquid biopsy for the diagnosis of angioim- 1) shows superior survival to patterns 2 and
46(2):166-170. munoblastic T-cell lymphoma. Leuk 3: a meta-analysis of 56 cases.
8. Ondrejka SL, Grzywacz B, Bodo J, et al. Lymphoma. 2020;61(10):2389-2398. Histopathology. 2012;60(4):570-585.
Angioimmunoblastic T-cell lymphomas 16. Tanzima Nuhat S, Sakata-Yanagimoto M, 23. Sakata-Yanagimoto M, Nakamoto-
with the RHOA p.Gly17Val mutation have Komori D, et al. Droplet digital polymerase Matsubara R, Komori D, et al. Detection of
classic clinical and pathologic features. Am chain reaction assay and peptide nucleic the circulating tumor DNAs in angioim-
J Surg Pathol. 2016;40(3):335-341. acid-locked nucleic acid clamp method for munoblastic T- cell lymphoma. Ann
9. Lee PH, Weng SW, Liu TT, et al. RHOA RHOA mutation detection in angioim- Hematol. 2017;96(9):1471-1475.

24. Nguyen TB, Sakata-Yanagimoto M, phomas. Blood. 2016;128(11):1490-1502. al. TET2 inactivation results in pleiotropic
Fujisawa M, et al. Dasatinib Is an effective 27. Rohr J, Guo S, Huo J, et al. Recurrent acti- hematopoietic abnormalities in mouse and
treatment for angioimmunoblastic T-cell vating mutations of CD28 in peripheral T- is a recurrent event during human lym-
lymphoma. Cancer Res. 2020;80(9):1875- cell lymphomas. Leukemia. 2016; phomagenesis. Cancer Cell. 2011;20(1):25-
1884. 30(5):1062-1070. 38.
25. Attygalle AD, Diss TC, Munson P, Isaacson 28. Fujisawa M, Sakata-Yanagimoto M, 31. Schwartz FH, Cai Q, Fellmann E, et al.
PG, Du MQ, Dogan A. CD10 expression in Nishizawa S, et al. Activation of RHOA- TET2 mutations in B cells of patients affect-
extranodal dissemination of angioim- VAV1 signaling in angioimmunoblastic T- ed by angioimmunoblastic T-cell lym-
munoblastic T-cell lymphoma. Am J Surg cell lymphoma. Leukemia. 2018;32(3):694- phoma. J Pathol. 2017;242(2):129-133.
Pathol. 2004;28(1):54-61. 702. 32. Ch'ang HJ, Su IJ, Chen CL, et al.
26. Vallois D, Dobay MP, Morin RD, et al. 29. Couronné L, Bastard C, Bernard OA. TET2 Angioimmunoblastic lymphadenopathy
Activating mutations in genes related to and DNMT3A mutations in human T-cell with dysproteinemia--lack of a prognostic
TCR signaling in angioimmunoblastic and lymphoma. N Engl J Med. 2012;366(1):95-96. value of clear cell morphology. Oncology.
other follicular helper T-cell-derived lym- 30. Quivoron C, Couronné L, Della Valle V, et 1997;54(3):193-198.

ARTICLE Non-Hodgkin Lymphoma
Ferrata Storti Foundation Efficacy and safety assessment of prolonged

maintenance with subcutaneous rituximab in
patients with relapsed or refractory indolent
non-Hodgkin lymphoma: results of the phase
III MabCute study
Simon Rule,1 Wolney Gois Barreto,2 Javier Briones,3 Angelo M. Carella,4 Olivier
Casasnovas,5 Chris Pocock,6 Clemens-Martin Wendtner,7 Francesco Zaja,8
Haematologica 2022 Susan Robson,9 Lachlan MacGregor,9 Roger R. Tschopp,9 Sonja Nick9 and
Volume 107(2):500-509 Martin Dreyling10
1
Derriford Hospital and Plymouth University Medical School, Plymouth, UK; 2Hemocentro
Ribeirão Preto, University of São Paulo, São Paulo, Brazil; 3Hospital de la Santa Creu i
Sant Pau, Barcelona, Spain; 4IRCCS AOU San Martino-IST, Genova, Italy; 5Centre
Hospitalier Universitaire François Mitterand, Dijon, France; 6Kent & Canterbury Hospital,
Canterbury, UK; 7Munich Clinic Schwabing, Academic Teaching Hospital, Ludwig-
Maximilians University, Munich, Germany; 8DSM, University of Trieste, Trieste, Italy;
9
F. Hoffmann-La Roche Ltd, Basel, Switzerland and 10Klinikum der Universität München,
Munich, Germany
ABSTRACT
R
ituximab plus chemotherapy induction followed by rituximab
maintenance for up to 2 years confers a long-term benefit in terms
of progression-free survival in patients with indolent non-Hodgkin
lymphoma. It is not known whether further prolonged maintenance with
rituximab provides additional benefit. The phase III MabCute study
enrolled 692 patients with relapsed or refractory indolent non-Hodgkin
lymphoma. Patients who responded to induction with rituximab plus
chemotherapy and were still responding after up to 2 years’ initial main-
tenance with subcutaneous rituximab were randomized to extended
maintenance with subcutaneous rituximab (n=138) or observation only
(n=138). The primary endpoint of investigator-assessed progression-free
survival in the randomized population was un-addressed by the end of
Correspondence: study because of an insufficient number of events (129 events were need-
ed for 80% power at 5% significance if approximately 330 patients were
MARTIN DREYLING randomized). In total, there were 46 progression-free survival events, 19
martin.dreyling@med.uni-muenchen.de
and 27 in the rituximab and observation arms, respectively (P=0.410 by
stratified log-rank test; hazard ratio 0.76 [95% confidence interval: 0.37–
Received: November 10, 2020. 1.53]). The median progression-free survival was not reached in either
Accepted: May 3, 2021. randomized arm. There were no new safety signals; however, adverse
events were seen slightly more frequently with rituximab than with
Pre-published: June 17, 2021.
observation during extended maintenance. Maintenance for up to 2 years
with rituximab after response to initial induction therefore remains the
https://doi.org/10.3324/haematol.2020.274803 standard of care in patients with relapsed or refractory indolent non-
Hodgkin lymphoma. (Clinicaltrials.gov identifier: NCT01461928)
All rights are reserved to the Ferrata Storti Foundation. Use of Introduction
conditions:
https://creativecommons.org/licenses/by-nc/4.0/legalcode. Non-Hodgkin lymphoma (NHL) accounts for approximately 85% of lymphomas.1
Copies of published material are allowed for personal or inter- Indolent forms include follicular lymphoma (FL), Waldenström macroglobulinemia/
nal use. Sharing published material for non-commercial pur- lymphoplasmacytic lymphoma and marginal zone lymphoma. Of these, FL is the
most common,1,2 accounting for 5/100,000 cases in Western Europe.3 Indolent NHL
sect. 3. Reproducing and sharing published material for com- usually develops slowly (and may not need immediate treatment), follows a relaps-
mercial purposes is not allowed without permission in writing ing-remitting course, and is often incurable.1
from the publisher. Chemoimmunotherapy based on the human/murine chimeric anti-CD20 mono-
clonal antibody rituximab is standard treatment for a range of B-cell malignancies,
including indolent and aggressive forms of NHL.3-7 Intravenously administered ritux-

Prolonged maintenance with rituximab in indolent NHL
imab prolongs time to disease progression and increases venous (375 mg/m2) and seven SC (1,400 mg fixed-dose) with six
overall survival (OS),8 but is associated with infusion reac- to eight chemotherapy cycles as induction (Figure 1; further infor-
tions, which can be severe.9,10 Thus, a slow infusion is mation is available in the Online Supplementary Appendix). Patients
required during the first antibody administration, which with complete or partial response received 2 years’ maintenance
generally takes at least 3.5–4 h.9-11 Faster infusion rates are with rituximab SC (Maintenance I). Patients with continuing
used for subsequent infusions;10,11 nevertheless, infusion response at the end of Maintenance I were randomized to pro-
duration remains a challenge for patients and healthcare longed maintenance with rituximab SC or to observation
providers, particularly when multi-agent chemotherapy is (Maintenance II).
being used.11
A subcutaneous (SC) formulation of rituximab and Study endpoints and procedures
recombinant human hyaluronidase has been developed to The primary endpoint was PFS from the time of randomization
address this concern.12 At fixed doses, rituximab SC has to extended maintenance with rituximab SC or observation in
shown comparable efficacy and safety to intravenous ritux- Maintenance II (PFS in the randomized intent-to-treat [ITT ]
rand rand
imab in patients with NHL or chronic lymphocytic population). Secondary endpoints included OS from the time of
leukemia, with non-inferior serum trough rituximab con- randomization in Maintenance II (OS ), overall response rate
rand
centrations.12-16 Additionally, patients’ preference/satisfac- (Cheson criteria29) at end of Induction, and partial response to
tion and time and motion data (active healthcare practition- complete response conversion rate at the end of Maintenance I.
er time and chair time for patients) favor the use of the SC An exploratory analysis of PFS and OS from enrollment to end of
formulation,17,18 which is currently approved in Europe, the Maintenance I (i.e., the non-randomized part of the study; PFS , reg
USA and numerous other countries for multiple indications OS ) according to induction chemotherapy was also performed.
reg
(chronic lymphocytic leukemia, diffuse large B-cell lym- Safety was assessed in all patients who received at least one
phoma and FL).9,19 Dosing advantages over intravenous dose of study medication and included adverse events (using
treatment include administration over 5–7 minutes, with a National Cancer Institute Common Toxicity Criteria Version 4.0
requirement for only 15 minutes of monitoring.9,19 and coded with Medical Dictionary for Regulatory Activities ver-
Rituximab plus chemotherapy induction followed by rit- sion 2.0), laboratory tests and vital signs.
uximab maintenance is an approved treatment in FL,9,19 and
has shown long-term progression-free survival (PFS) benefit Analytical plan
in patients with indolent NHL.20-27 Tumor response and sur- Sample size was based on a phase III randomized study of 465
vival data show improvements in outcomes that persist R/R FL patients. Overall, 129 PFS events were required to achieve
rand
over the longer term when rituximab maintenance therapy 80% power for the log-rank-test at a two-sided significance level
is given for up to 2 years.27,28 Whether further and prolonged of 5%; therefore, approximately 700 patients needed to be
maintenance therapy (beyond 2 years) would benefit enrolled to randomize 330 patients (allowing for a 10% dropout)
patients with relapsed/refractory (R/R) indolent NHL who after the 2.5-year Induction plus Maintenance I. Randomization to
have maintained their response to treatment remains Maintenance II was 1:1, stratified by indolent NHL subtype and
unknown. MabCute (NCT01469128) is a phase III trial in Follicular Lymphoma International Prognostic Index (FLIPI) cate-
which patients with R/R indolent NHL were randomized to gory.30 The end of study was defined as the time when all patients
prolonged rituximab SC maintenance or observation after randomized into Maintenance II had been followed up for ≥15
completing rituximab SC-based induction and 2 years months, or earlier if at least 129 PFS events had been observed.
maintenance therapy, provided that they were in response PFS , OS , PFS and OS were reported with medians, 95%
rand rand reg reg
and willing to continue treatment. confidence intervals (95% CI), and Kaplan-Meier estimates and
their 95% CI. The randomized treatment arms (prolonged ritux-
imab maintenance vs. observation in Maintenance II) were com-
Methods pared using log-rank testing stratified according in indolent NHL
subtype and FLIPI category. Cox regression was used to estimate
Study design hazard ratios (HR).
This was a phase III, open-label, multicenter, international, ran-
domized interventional study enrolling patients from 141 centers
worldwide (mostly in Europe). MabCute was divided into Results
Induction (6–8 months), Maintenance I (24 months) and
Maintenance II (minimum 15 months) phases (Figure 1). Patients
The study was carried out in accordance with the Declaration of In total, 692 patients received rituximab plus chemother-
Helsinki and Good Clinical Practice, local legislation and the apy as induction (ITT population for Induction); 60.5% of
approval of institutional review boards. Written informed consent patients received bendamustine, 12.4% received
was obtained from participants. cyclophosphamide, doxorubicin, vincristine and pred-
nisone (CHOP) and 11.8% received cyclophosphamide,
Study population vincristine and prednisone (CVP) (Online Supplementary
Adults aged ≥18 years with R/R CD20+ grade 1, 2 or 3a FL or Table S1); very small numbers received fludarabine,
other CD20+ indolent NHL (Waldenström macroglobulinemia/lym- cyclophosphamide and mitoxantrone (FCM) or mitox-
phoplasmacytic lymphoma or marginal zone lymphoma), and antrone, chlorambucil and prednisone (MCP). The distribu-
Eastern Cooperative Oncology Group (ECOG) performance status tion of patients who received each induction regimen was
≤2 were recruited. Details of the baseline assessments are provided maintained out to Maintenance II (Online Supplementary
in the Online Supplementary Appendix. Table S1). Of the patients who received induction therapy,
148 discontinued treatment because of adverse events (70
Study treatments patients; 10.1%), disease progression (29; 4.2%), patients’
Eligible patients received eight rituximab cycles, one intra- request (16; 2.3%), investigators’ request (7; 1.0%), loss to

S. Rule et al.
Figure 1. Study design. aWaldenström macroglobulinemia/lymphoplasmacytic lymphoma or marginal zone lymphoma. bChemotherapy options included bendamus-
tine, CHOP, CVP, FCM, MCP, CHVP-IFN, chlorambucil, fludarabine-containing regimen or GIFOX. cMaintenance started within 8–12 weeks of completion of induction.
R/R: relapsed or refractory; FL: follicular lymphoma; Gr: grade; NHL: non-Hodgkin lymphoma; R: rituximab; Cs: cycles; PD: disease progression; CR: complete
response; PR: partial response; SD: stable disease; SC: subcutaneous; IV: intravenous; CHOP: cyclophosphamide, doxorubicin, vincristine and prednisone; CVP:
cyclophosphamide, vincristine and prednisone; FCM: fludarabine, cyclophosphamide and mitoxantrone; MCP: mitoxantrone, chlorambucil and prednisone; CHVP-IFN:
cyclophosphamide, doxorubicin, etoposide and prednisone + interferon-α; GIFOX: gemcitabine, ifosfamide and oxaliplatin.
follow-up (4; 0.6%), death (2; 0.3%) or other reasons (20; lation for Induction (n=692), patients receiving bendamus-
2.9%) (Online Supplementary Figure S1A, B, Online tine were older than those receiving CHOP or CVP, and a
Supplementary Table S2). A further 39 patients withdrew greater proportion had a high FLIPI score and Ann Arbor
after Induction and before Maintenance I because of disease stage III/IV disease at screening (Online Supplementary Table
progression (16; 2.3%), adverse events (6; 0.9%), patients’ S3). More patients receiving bendamustine and CHOP had
request (2; 0.3%), investigators’ request (2; 0.3%), death (1; FL compared with those receiving CVP.
0.1%), loss to follow-up (1; 0.1%) or other reasons (11; Six of 138 patients in the Maintenance II rituximab arm
1.6%, all with stable disease). discontinued before the start of treatment (Figure 2).
Of the 505 patients who continued to Maintenance I, 494 Maintenance II was completed thereafter by 109 patients
were treated (ITT population for Maintenance I; treatment randomized to rituximab and 111 randomized to observa-
was not given because of adverse events in 5 patients, dis- tion; 23 patients (16.7%) randomized to rituximab and 27
ease progression in 2, investigator’s request or death in 1 (19.6%) randomized to observation discontinued during
each, and other reasons in 2). During Maintenance I, 188 Maintenance II (Figure 2). The median follow-up time was
patients (38.1%) discontinued study treatment because of 28.1 (range, 0–46) months. A single patient had progressive
disease progression (82; 16.6%), adverse events (66; disease at the end of Maintenance I but was randomized to
13.4%), patients’ request (20; 4.0%), investigators’ request rituximab in error. This was subsequently recorded as a
(8; 1.6%), death (3; 0.6%), loss to follow-up (2; 0.4%) or protocol violation.
other reasons (7; 1.4%: 2 with stable disease) (Online
Supplementary Figure S1C). A further 28 patients (5.7%) Rituximab exposure
completed Maintenance I but discontinued before The median duration of exposure to rituximab during
Maintenance II. Reasons were patients’ request (12; 2.4%), Induction was 6.4 (range, 0–11) months. The median num-
disease progression (9; 1.8%), adverse events (4; 0.8%), ber of rituximab cycles was 8.0 (range, 1–9); 522 patients
investigator’s request (1; 0.2%) and other reasons (2; 0.4%). (75.4%) received the planned eight cycles.
Two other patients (0.4%) failed to meet randomization cri- In Maintenance I, the median duration of exposure to rit-
teria; the remaining 276 patients were randomized to uximab was 20.3 (range, 0–28) months, with a median of
Maintenance II (Figure 2). 12.0 (range, 1–12) cycles being given. Of the 494 patients,
The median durations of the Induction and Maintenance 295 (59.7%) received the planned maximum 12 injections
I periods were 8.2 (range 0–18) months and 22.1 (range 0– every 8 weeks for 24 months.
31) months, respectively. The median duration of exposure during Maintenance II
The primary ITT population included 276 patients who
rand treatment was 24.8 (range, 0–43) months, and the median
were randomized into Maintenance II (138 each in the rit- number of rituximab cycles was 14.0 (range, 1–24).
uximab and observation arms). Two further patients were Three patients received the highest number of rituximab
initially planned for randomization but were subsequently treatments (44 cycles across the entire study), while two
found to be ineligible: one had disease progression and one received the lowest (21 cycles).
had stable disease. Just over half of all patients were male
and approximately two-thirds had Ann Arbor stage IV dis- Safety and tolerability
ease (Table 1). Patients were evenly distributed across FLIPI Induction
score categories in both arms. Approximately 40% of Treatment-emergent adverse events and serious adverse
patients had bone marrow involvement, and just over half events were reported during Induction in 89.0% (616/692)
of all patients had FL. Nearly 60% overall had received rit- and 30.1% (208/692) of patients, respectively. Half of all
uximab plus bendamustine at Induction. Of the ITT popu- patients experienced at least one treatment-emergent

Figure 2. Patients’ disposition during Maintenance II. aTwo additional patients originally intended for randomization failed to meet continuation criteria and were con-
sequently not treated in Maintenance II. bDerived by subtracting patients who discontinued from treated patients.
adverse event of grade ≥3 intensity (n=344; 49.7%), most of neutropenia reported as an adverse event were similar
commonly neutropenia (n=160; 23.1%). Febrile neutrope- across induction chemotherapy regimens.
nia (n=31; 4.5%), pneumonia (n=28; 4.0%) and neutropenia
(n=16; 2.3%) were the most commonly reported serious Maintenance I
adverse events (occurring in >2% of patients). Treatment-emergent adverse events and serious adverse
Infusion/administration-related reactions were reported in events were reported in 380/494 (76.9%) and 134/494
330 patients (47.7%) during Induction; 54 patients (7.8%) (27.1%) patients, respectively, during Maintenance I. At
had a grade ≥3 event. The most common infusion/adminis- least one treatment-emergent adverse event of grade ≥3
tration-related reaction of any grade during Induction was intensity was reported in 163 patients (33.0%), most com-
nausea (n=57; 8.2%); neutropenia was the most common monly neutropenia (n=59; 11.9%). Pneumonia was the
grade ≥3 infusion/administration-related reaction (n=15; most commonly reported serious adverse event affecting
2.2%). At least one treatment-emergent adverse event lead- >2% of patients during Maintenance I (n=17; 3.4%).
ing to rituximab discontinuation was reported in 66 Infusion/administration-related reactions were reported in
patients (9.5%) during Induction, most frequently neu- 75 patients (15.2%), with 20 (4.0%) experiencing at least
tropenia (7 patients; 1.0%). At least one treatment-emer- one grade ≥3 event. The most common infusion/adminis-
gent adverse event leading to death was reported in 12 tration-related reaction of any grade was decreased neu-
patients (1.7%). trophil count (n=14; 2.8%). Neutropenia was the most
Similar incidences of treatment-emergent adverse events commonly reported grade ≥3 infusion/administration-relat-
were seen across induction chemotherapy regimens (Online ed reaction during Maintenance I (9 patients; 1.8%).
Supplementary Table S4). Patients receiving bendamustine Rituximab discontinuation due to a treatment-emergent
experienced more general disorders and administration site adverse event was reported in 28 patients (5.7%) during
conditions overall than those in other groups. Frequencies Maintenance I. Of these, only neutropenia and pneumonia

S. Rule et al.
Table 1. Patient and disease characteristics at the start of Table 2. Summary of adverse events occurring during extended main-
Maintenance II. tenance.
Number of patients (%) Patients with ≥1 event, n (%) R-SC Observation
Characteristic R-SC Observation n=138 n=138
n=138 n=138
≥1 AE 111 (80.4) 80 (58.0)
Median age, years (range) 64 (26-89) 65 (34-86) Grade ≥3 AE affecting ≥1% patients 48 (34.8) 40 (29.0)
Male, n (%) 74 (53.6) 68 (49.3) in either arm
Ann Arbor stage at diagnosis, n/N (%) Neutropenia 12 (8.7) 8 (5.8)
I 13/134 (9.7) 8/135 (5.9) Pneumonia 7 (5.1) 4 (2.9)
II 12/134 (9.0) 19/135 (14.1) Hypertension 3 (2.2) 0
III 21/134 (15.7) 30/135 (22.2) Neutrophil count decreased 3 (2.2) 0
IV 88/134 (65.7) 78/135 (57.8) Acute kidney injury 0 2 (1.4)
Febrile neutropenia 2 (1.4) 0
FLIPI score, n (%) Leukopenia 0 2 (1.4)
Low 25 (34.2) 28 (36.4) Myelodysplastic syndrome 1 (0.7) 2 (1.4)
Intermediate 22 (30.1) 27 (35.1) Upper respiratory tract infection 0 2 (1.4)
High 26 (35.6) 22 (28.6) Sepsis 2 (1.4) 2 (1.4)
Bone marrow involvement, n (%) 60 (43.5) 59 (42.8) Thrombocytopenia 1 (0.7) 2 (1.4)
Median lactate dehydrogenase, 3.26 (1.30-11.77) 3.32 (1.40-9.15) Vomiting 2 (1.4) 0
ukat/L (range) Serious AE affecting ≥1% patients 31 (22.5) 32 (23.2)
Type of NHL at screening, n (%) in either arm
FL 73 (52.9) 77 (55.8) Pneumonia 8 (5.8) 4 (2.9)
WM/LPL 28 (20.3) 25 (18.1) Acute kidney injury 0 2 (1.4)
MZL 36 (26.1) 35 (25.4) Appendicitis 2 (1.4) 0
Induction chemotherapy regimen Bronchitis 0 2 (1.4)
Bendamustine 80 (58.0) 79 (57.2) Fall 0 2 (1.4)
CHOP 20 (14.5) 19 (13.8) Febrile neutropenia 2 (1.4) 0
CVP 26 (18.8) 22 (15.9) Myelodysplastic syndrome 1 (0.7) 2 (1.4)
Other 12 (8.6) 18 (13.0) Neutropenia 0 2 (1.4)
Sepsis 2 (1.4) 2 (1.4)
R-SC: subcutaneous rituximab; FLIPI: Follicular Lymphoma International Prognostic
Index; NHL: non-Hodgkin lymphoma; FL: follicular lymphoma; WM/LPL: Waldenström Squamous cell carcinoma of skin 1 (0.7) 2 (1.4)
macroglobulinemia/lymphoplasmacytic lymphoma; MZL: marginal zone lymphoma; Grade 5 (fatal) AE 5 (3.6) 5 (3.6)
CHOP: cyclophosphamide, doxorubicin, vincristine and prednisone; CVP: cyclophos-
phamide, vincristine and prednisone. AE leading to treatment discontinuation 10 (7.2) 0
R-SC: subcutaneous rituximab; AE: adverse event.
were seen in more than one patient (2 patients each). At
least one treatment-emergent adverse event leading to ous adverse events was similar for both arms (22.5% with
death was reported in eight patients (1.6%). rituximab and 23.2% for observation) (Table 2), with pneu-
Adverse events were the most common reason for death monia (5.8% and 2.9%, respectively) and sepsis (1.4% for
during Induction and Maintenance I (40/692; 5.8% and both arms) being most commonly reported. All fatal
32/494; 6.5%, respectively). Sepsis was the most frequent adverse events (5 in each arm) were considered unrelated
event leading to death during these phases (7 patients to study treatment by the investigators. These events were
[1.0%] and 2 patients [0.4%], respectively). Rituximab- pneumonia, septic shock, acute myocardial infarction,
related sepsis was associated with death in four patients Crohn disease, abdominal infection and diverticulitis
(0.6%) during Induction and one patient (0.2%) during (same patient) in the rituximab arm, and acute myeloid
Maintenance I. leukemia, cardiopulmonary failure, ventricular tachycar-
dia, pneumonia and lung disorder in the observation arm.
Maintenance II Five further deaths in the rituximab arm and two in the
The original wording of the study protocol led to differ- observation arm were due to disease progression; a single
ences in adverse event reporting between the rituximab additional death with unknown cause was recorded in the
and observation arms in Maintenance II (see the Online observation arm.
Supplementary Appendix for details). After a protocol There were no safety concerns or new signals related to
amendment to permit retrospective collection of adverse hematology, biochemistry or immunological parameters in
events of grade ≥3 during this phase (allowing adverse any phase of the study, and no meaningful changes from
event reporting to be consistent between the rituximab baseline in vital signs. There were also no unexpected
and observation arms), neutropenia and pneumonia were changes from baseline in worst-on-treatment ECOG
the most frequently reported grade ≥3 adverse events in scores, and no noteworthy differences in score shifts
both the rituximab arm (8.7% and 5.1%, respectively) and between the rituximab and observation arms in
the observation arm (5.8% and 2.9%, respectively) (Table Maintenance II.
2). However, when looking at median neutrophil counts
(based on laboratory data), similar values were observed in Efficacy
both treatment arms in Maintenance II. There were three The overall response rate at the end of Induction was
grade ≥3 infusion/administration-related reactions (1 each 84.7% (95% CI: 81.1–87.3), and was similar across differ-
of lymphopenia, urinary tract infection and hypertensive ent chemotherapies: 86.4% (95% CI: 82.7–89.5) for ben-
crisis). There were no reports of grade ≥3 rash, erythema or damustine; 87.2% (95% CI: 78.3–93.4) for CHOP; 84.1%
skin reaction during Maintenance II. The incidence of seri- (95% CI: 74.4–91.3) for CVP; and 76.9% (95% CI: 67.6–

Figure 3. Survival outcomes during the randomized Maintenance II period. Kaplan-Meier analysis of progression-free (A) and overall survival (B) during the random-
ized Maintenance II period. R-SC: subcutaneous rituximab.
84.6) for other regimens (including FCM and MCP). All but FLIPI risk category and NHL subtype as stratification fac-
one patient per arm among the 276 who were randomized tors. PFS rates at 6, 9, 12, 15 and 18 months (Kaplan-Meier
rand
in Maintenance II were responders after Induction (Online estimates) were similar for both arms (between 0.97 at 6
Supplementary Table S5). Proportions of patients in com- months and 0.88 at 18 months for rituximab, and between
plete response or partial response at the end of 0.96 at 6 months and 0.87 at 18 months for observation).
Maintenance I were also comparable between arms among The median PFS was not reached in either arm (Figure 3A).
rand
the 276 patients who were randomized (Online One patient, randomized to observation, discontinued
Supplementary Table S5). from the study and subsequently died 2 months later. This
Of the 357 patients who achieved a partial response at event was not taken into consideration in the primary
the end of Induction, 77 achieved a complete response by analysis due to a recording issue. It had no effect on the
the end of Maintenance I, providing a conversion rate of overall results or conclusions of the study.
21.6% (95% CI: 17.4–26.2). The median PFS (from enrollment to end of
reg
The MabCute study was unable to address its primary Maintenance I) (Figure 4A) was 46.32 months (95% CI:
endpoint (investigator-assessed PFS ) because the number
rand 42.87–60.02) in patients receiving bendamustine, 39.62
of events reported was insufficient: 129 PFS events were rand months (95% CI: 27.86–not reached) in patients receiving
needed for 80% power at 5% significance, with approxi- CHOP, and 37.03 months (95% CI: 33.87–74.12) in
mately 700 patients needed initially to yield the 330 patients receiving CVP. Three-year PFS estimates for
required for randomization. There were 46 PFS events at rand patients receiving bendamustine, CHOP, and CVP were
the end of study: 19 and 27 in the rituximab and observa- 0.63 (95% CI: 0.57–0.69), 0.58 (0.46–0.68), and 0.59 (0.28–
tion arms, respectively: P=0.410 by log-rank test stratified 0.80), respectively. The median OS (from enrollment to
reg
by FLIPI risk category and NHL subtype; HR 0.76 (95% CI: end of Maintenance I) (Figure 4B) was not reached in
0.37–1.53), estimated using a Cox regression model with patients receiving bendamustine (95% CI: 66.86–not

S. Rule et al.
Figure 4. Survival outcomes from enrollment to end of Maintenance I, according to induction chemotherapy. Kaplan-Meier analysis of progression-free (A) and over-
all survival (B) from enrollment to end of Maintenance I, according to induction chemotherapy received (bendamustine vs. CHOP and CVP). aIntent to treat population
for Induction. Time to event calculated from first induction therapy up to the earliest date of event until randomization; data censored after randomization. CHOP:
cyclophosphamide, doxorubicin, vincristine and prednisone; CVP: cyclophosphamide, vincristine and prednisone.
reached) or CVP (95% CI: 46.82–not reached). The median induction therapy (Online Supplementary Table S7).
OS was 58.84 months (95% CI: 42.22-not reached) in
reg There were 18 deaths (OS events) in total, ten in the rit-
rand
patients receiving CHOP induction. Three-year OS esti- uximab arm and eight in the observation arm (not includ-
mates for patients receiving bendamustine, CHOP, and ing the patient with a retrospective record of death). The
CVP were 0.83 (95% CI: 0.78–0.87), 0.70 (95% CI: 0.53– median OS was not reached (Figure 3B).
rand
0.82), and 0.82 (95% CI: 0.59–0.93), respectively. PFS and reg
OS by NHL subtype are available in Online Supplementary

reg
Table S6. Unfortunately, due to the low numbers no con- Discussion

clusion can be drawn from these data.
Response rates at the end of Induction by both The benefit of 2 years of maintenance therapy with rit-
chemotherapy regimen and patients remaining at the start uximab after response to frontline induction in patients
of each subsequent study phase showed that 56.9% of with indolent NHL is well established in terms of signifi-
patients (157/276) who responded and who ultimately cantly improved PFS. Early trials of rituximab mainte-
entered Maintenance II had received bendamustine as nance in patients with R/R indolent NHL indicated effica-

cy in this setting too.23,24,31,32 The findings prompt the ques- in patients’ selection; the investigator could decide what
tion of whether further and prolonged maintenance ther- regimens to give to which patients – most patients were
apy (beyond 2 years) would benefit patients with R/R treated with bendamustine in the Induction period, and
indolent NHL who have maintained their response to the size of the subgroups is very different. Therefore, a
treatment. direct comparison between treatment regimens is not
The overall tumor response rate after induction (~85%) appropriate. In MabCute, approximately 60% of patients
in MabCute was consistent with rates observed in previ- received bendamustine at Induction, and this proportion
ous studies in R/R indolent NHL (75-95%).22-24,31,32 These of patients was maintained out to the Maintenance II
trials showed significant improvements in response dura- phase. There are few data available on the use of ben-
tion and median PFS when rituximab maintenance thera- damustine in R/R NHL. A study by Sakai et al. recently
py was given for up to 2 years compared with observation reported 3-year PFS and OS rates of 71% and 89%, respec-
alone, and are supported by a meta-analysis of 2,586 tively, in a population of patients with R/R FL,37 while the
patients participating in nine randomized trials which STIL group reported a 1-year PFS of 76% and median OS
showed a significant improvement of median OS with rit- of 109.7 months in patients with R/R indolent NHL or
uximab maintenance therapy versus observation only in mantle cell lymphoma.34 However, comparison between
patients with R/R FL (HR 0.72, 95% CI: 0.57–0.91).28 these trials is difficult; the PFS survival data from the cur-
reg
Maintenance with rituximab for 2 years following the end rent study were censored after Maintenance I, and are
of Induction in the current study was associated with a therefore not comparable with general PFS data.
rate of partial response to complete response conversion No unexpected toxicities were reported during
similar to that observed in previous studies.21 Maintenance II, and good tolerability and safety were
Although an OS benefit has been observed following maintained throughout follow-up. The proportion of
rituximab maintenance in the R/R setting, it has not been patients who experienced adverse events during long-
demonstrated in the frontline setting. A 10-year follow- term maintenance was slightly greater in the rituximab
up of the PRIMA study in 1,018 patients with high tumor arm than in the observation arm. These observations were
burden, previously untreated FL showed a significant as expected, given the known profile of rituximab SC.12
long-term PFS benefit of rituximab maintenance over Rituximab is always given by intravenous infusion for the
observation for 2 years after response to induction with first cycle, when the risk of infusion-related reactions is
rituximab and chemotherapy.27 Although there was no sig- greatest, to allow slowing or stopping of the infusion (as a
nificant OS benefit, the authors noted that over half of preventative measure). The incidence of infusion-related
patients in the rituximab arm had not had disease progres- reactions decreases with subsequent infusions. The over-
sion over the 10 years, and had not required new anti- all safety profile of rituximab SC is similar to that of the
lymphoma treatment. Similar findings (significant PFS intravenous formulation, but with a greater incidence of
improvement but no significant effect on OS) were report- mostly mild-to-moderate infusion/administration-related
ed by the ECOG-ACRIN group after a median 11.5 years reactions, primarily injection-site reactions, which
of follow-up of 387 patients who attained at least stable decrease in frequency over time.14,15,17,38 This pattern was
disease after CVP induction.33 In addition, a prior study by observed in the current study (i.e. from 47.7% of patients
the German STIL group confirmed the benefit of ritux- during Induction to 15.2% during Maintenance I and
imab maintenance in R/R indolent NHL after a bendamus- 10.1% of rituximab patients in Maintenance II).
tine or fludarabine salvage therapy.34 Interestingly, in line with prior publications on frontline
The key benefits of SC rituximab, with its short admin- therapy of FL, a bendamustine-based induction resulted in
istration time, are linked to reductions in healthcare more frequent pyrexia and neutropenia (Online
resource utilization18,35 and patients’ preference15,36 relative Supplementary Table S5).
to the intravenous formulation, particularly for long-term In conclusion, the MabCute study was unable to
therapy. address the question of whether prolonged (beyond 2
Unfortunately, MabCute was unable to address its pri- years) maintenance therapy with rituximab adds any clear
mary endpoint of investigator-assessed PFS in the ran- benefit compared with observation only in patients with
domized population. This was due to a much lower than R/R indolent NHL (who have responded to induction ther-
anticipated number of PFS events, representing only a
rand apy with rituximab plus chemotherapy), due to a low
third of the required events to have a power of 80% with number of PFS events. Extension of treatment was not
a hazard ratio of 0.605. The reason for the low rate of associated with any important additional toxicity (in par-
PFS events was not clear, but may have been related to
rand ticular no additional neutropenia or infection), and no new
the effectiveness of supportive care and treatment delivery safety signals were observed. Two years of maintenance
under the study protocol. with rituximab after response to initial induction therapy
There were 18 deaths in total, ten in the rituximab arm therefore remains the standard of care in these patients.
and eight in the observation arm (not including the patient
with a retrospective record of death). This study was not Disclosures
powered to evaluate survival, and follow-up was relative- SR declares a consultation or advisory role for Janssen,
ly short at the time of the analysis. AstraZeneca, F. Hoffmann-La Roche Ltd, Sunesis,
The exploratory analysis of PFS and OS from enroll-
reg reg Pharmacyclics, Celgene, Celltrion, Kite; speakers bureau for
ment to end of Maintenance I (i.e. the non-randomized Janssen; and research funding from Janssen. WGB declares no
part of the study) showed 3-year PFS and OS rates of
reg reg conflict of interest. JB declares honoraria from F. Hoffmann-La
63% and 83%, respectively in patients treated with ben- Roche Ltd, Takeda, Celgene, Novartis, and Gilead; consultation
damustine, 58% and 70%, respectively, in those treated or advisory role for Takeda, Janssen, Celgene, and Gilead;
with CHOP, and 59% and 82%, respectively, in those research funding from F. Hoffmann-La Roche Ltd; and travel
treated with CVP. It should be noted that there was a bias and/or accommodation expenses from F. Hoffmann La-Roche

S. Rule et al.
Ltd, Takeda, Celgene, Janssen, and Gilead. AMC does not research funding from Celgene, Janssen, Mundipharma, and F.
declare any conflict of interest. OC declares honoraria from F. Hoffmann-La Roche Ltd.
Hoffmann-La Roche Ltd, Takeda, BMS, Merck, Gilead, and
Janssen; consultation or advisory role for F. Hoffmann-La Roche Contributions
Ltd, Takeda, BMS, Merck, Gilead, and Janssen; research fund- SR, WGB, JB, AMC, OC, CP, C-MW, FZ and MD were
ing from F. Hoffmann-La Roche Ltd, Takeda, Gilead, and involved in accrual and treatment of patients. SR analyzed data.
AbbVie; and travel and/or accommodation expenses from F. All authors were involved in interpreting the data, critically
Hoffmann-La Roche Ltd, Takeda, and Janssen. CP declares reviewing the manuscript, approved the manuscript for submis-
honoraria from Janssen and Gilead; consultancy or advisory role sion and agree to be accountable for the accuracy and integrity of
for Takeda and Celgene and travel and/or accommodation the study.
expenses from Gilead. C-MW declares honoraria from F.
Hoffmann-La Roche Ltd, Janssen-Cilag, Gilead, and AbbVie; Funding
consultation or advisory role for F. Hoffmann-La Roche Ltd, MabCute was sponsored by F. Hoffmann-La Roche Ltd.
Janssen-Cilag, Gilead, and AbbVie; research funding from F. Third-party medical writing assistance, under the direction of
Hoffmann-La Roche Ltd, Janssen-Cilag, Gilead, and AbbVie; Simon Rule and Martin Dreyling, was provided by Christopher
and travel and/or accommodation expenses from F. Hoffmann-La Dunn and Scott Malkin of Ashfield MedComms, an Ashfield
Roche Ltd, Janssen-Cilag, Gilead, and AbbVie. FZ declares Health company, and was funded by F. Hoffmann-La Roche Ltd.
honoraria from F. Hoffmann-La Roche Ltd, Janssen-Cilag,
Gilead, Celgene, AbbVie, Takeda, and Novartis; consultation or Data-sharing statement
advisory role for Sandoz, F. Hoffmann-La Roche Ltd, Janssen- Qualified researchers may request access to individual patient
Cilag, Gilead, Celgene, AbbVie, Takeda, and Novartis; research level data through the clinical study data request platform
funding from Celgene and Novartis; and travel and/or accommo- (https://vivli.org/). Further details on Roche's criteria for eligible
dation expenses from F. Hoffmann-La Roche Ltd, Celgene, studies are available here (https://vivli.org/members/ourmem-
AbbVie, Takeda, and Novartis. SR is employed by F. Hoffmann- bers/). For further details on Roche's Global Policy on the
La Roche Ltd. LMacG, RRT, and SN are employed by F. Sharing of Clinical Information and how to request access to
Hoffmann-La Roche Ltd. MD declares honoraria from Bayer, related clinical study documents, see here
Celgene, Gilead, Janssen, and F. Hoffmann-La Roche Ltd; con- (https://www.roche.com/research_and_development/who_we_a
sultation or advisory role for Acerta, Bayer, Celgene, Gilead, re_how_we_work/clinical_trials/our_commitment_to_data_shar
Janssen, Novartis, F. Hoffmann La Roche Ltd, and Sandoz; and ing.htm
References ments/product-information/mabthera-epar- lymphocytic leukaemia (SAWYER): a phase

product-information_en.pdf). 1b, open-label, randomised controlled non-
10. Highlights of prescribing information. inferiority trial. Lancet Haematol. 2016;3(3):
1. Ninkovic S, Lambert J. Non-Hodgkin lym- Rituxan® (rituximab) injection, for intra- e128-138.
phoma. Medicine. 2017;45(5):297-304. venous use. South San Francisco, CA: 17. Rummel M, Kim TM, Aversa F, et al.
2. Shankland KR, Armitage JO, Hancock BW. Biogen Idec, Inc. and Genentech USA, Inc.; Preference for subcutaneous or intravenous
Non-Hodgkin lymphoma. Lancet. 2012;380 2019 September. administration of rituximab among patients
(9844):848-857. 11. Yelvington BJ. Subcutaneous rituximab in with untreated CD20+ diffuse large B-cell
3. Dreyling M, Ghielmini M, Rule S, et al. follicular lymphoma, chronic lymphocytic lymphoma or follicular lymphoma: results
Newly diagnosed and relapsed follicular leukemia, and diffuse large B-cell lym- from a prospective, randomized, open-label,
lymphoma: ESMO Clinical Practice phoma. J Adv Pract Oncol. 2018;9(5):530- crossover study (PrefMab). Ann Oncol.
Guidelines for diagnosis, treatment and fol- 534. 2017;28(4):836-842.
low-up. Ann Oncol. 2021;32(3):298-308. 12. Davies A, Berge C, Boehnke A, et al. 18. De Cock E, Kritikou P, Sandoval M, et al.
4. Kastritis E, Leblond V, Dimopoulos MA, et Subcutaneous rituximab for the treatment of Time savings with rituximab subcutaneous
al. Waldenström’s macroglobulinaemia: B-cell hematologic malignancies: a review of injection versus rituximab intravenous infu-
ESMO Clinical Practice Guidelines for diag- the scientific rationale and clinical develop- sion: a time and motion study in eight coun-
nosis, treatment and follow-up†. Ann ment. Adv Ther. 2017;34(10):2210-2231. tries. PLoS One. 2016;11(6):e0157957.
Oncol. 2018;29(Suppl 4):iv41-iv50. 13. Davies A, Merli F, Mihaljevic B, et al. 19. Highlights of prescribing information:
5. Tilly H, Gomes da Silva M, Vitolo U, et al. Pharmacokinetics and safety of subcuta- Rituxan Hycela™ (rituximab and
Diffuse large B-cell lymphoma (DLBCL): neous rituximab in follicular lymphoma hyaluronidase human) injection, for subcu-
ESMO Clinical Practice Guidelines for diag- (SABRINA): stage 1 analysis of a randomised taneous use. San Francisco, CA: Genentech,
nosis, treatment and follow-up. Ann Oncol. phase 3 study. Lancet Oncol. 2014;15(3):343- Inc. 2017.
2015;26(Suppl 5):v116-125. 352. 20. Schneider T, Rosta A, Losonczy H, et al.
6. Zelenetz AD, Gordon LI, Abramson JS, et al. 14. Davies A, Merli F, Mihaljevic B, et al. Efficacy and tolerability of a 2-year ritux-
NCCN guidelines insights: B-cell lym- Efficacy and safety of subcutaneous rituximab maintenance therapy in patients with
phomas, version 3.2019. J Natl Compr Canc imab versus intravenous rituximab for advanced follicular lymphoma after induc-
Netw. 2019;17(6):650-661. first-line treatment of follicular lymphoma tion of response with rituximab-containing
7. Dreyling M, Campo E, Hermine O, et al. (SABRINA): a randomised, open-label, first line-regimens (HUSOM Study). Pathol
Newly diagnosed and relapsed mantle cell phase 3 trial. Lancet Haematol. 2017;4(6): Oncol Res. 2018;24(2):199-205.
lymphoma: ESMO Clinical Practice e272-e282. 21. Salles G, Seymour JF, Offner F, et al.
Guidelines for diagnosis, treatment and fol- 15. Lugtenburg P, Avivi I, Berenschot H, et al. Rituximab maintenance for 2 years in
low-up. Ann Oncol. 2017;28(Suppl 4):iv62- Efficacy and safety of subcutaneous and patients with high tumour burden follicular
iv71. intravenous rituximab plus cyclophos- lymphoma responding to rituximab plus
8. Salles G, Barrett M, Foa R, et al. Rituximab phamide, doxorubicin, vincristine, and pred- chemotherapy (PRIMA): a phase 3, ran-
in B-cell hematologic malignancies: a review nisone in first-line diffuse large B-cell lym- domised controlled trial. Lancet. 2011;377
of 20 years of clinical experience. Adv Ther. phoma: the randomized MabEase study. (9759):42-51.
2017;34(10):2232-2273. Haematologica. 2017;102(11):1913-1922. 22. Forstpointner R, Dreyling M, Repp R, et al.
9. MabThera Summary of Product 16. Assouline S, Buccheri V, Delmer A, et al. The addition of rituximab to a combination
Characteristics. 2020. (Accessed 27 April, Pharmacokinetics, safety, and efficacy of of fludarabine, cyclophosphamide, mitox-
2021, at subcutaneous versus intravenous rituximab antrone (FCM) significantly increases the
https://www.ema.europa.eu/en/docu- plus chemotherapy as treatment for chronic response rate and prolongs survival as com-

pared with FCM alone in patients with Sustained progression-free survival benefit Randomized phase 3 study in low-grade
relapsed and refractory follicular and mantle of rituximab maintenance in patients with lymphoma comparing maintenance anti-
cell lymphomas: results of a prospective ran- follicular lymphoma: long-term results of CD20 antibody with observation after
domized study of the German Low-Grade the PRIMA study. J Clin Oncol. 2019;37(31): induction therapy: a trial of the ECOG-
Lymphoma Study Group. Blood. 2004;104 2815-2824. ACRIN Cancer Research Group (E1496).
(10):3064-3071. 28. Vidal L, Gafter-Gvili A, Salles G, et al. Cancer. 2016;122(19):2996-3004.
23. van Oers MH, Klasa R, Marcus RE, et al. Rituximab maintenance for the treatment of 34. Rummel M, Kaiser U, Balser C, et al.
Rituximab maintenance improves clinical patients with follicular lymphoma: an Bendamustine plus rituximab versus flu-
outcome of relapsed/resistant follicular non- updated systematic review and meta-analy- darabine plus rituximab for patients with
Hodgkin lymphoma in patients both with sis of randomized trials. J Natl Cancer Inst. relapsed indolent and mantle-cell lym-
and without rituximab during induction: 2011;103(23):1799-1806. phomas: a multicentre, randomised, open-
results of a prospective randomized phase 3 29. Cheson BD, Pfistner B, Juweid ME, et al. label, non-inferiority phase 3 trial. Lancet
intergroup trial. Blood. 2006;108(10):3295- Revised response criteria for malignant lym- Oncol. 2016;17(1):57-66.
3301. phoma. J Clin Oncol. 2007;25(5):579-586. 35. Rule S, Collins GP, Samanta K.
24. van Oers MH, Van Glabbeke M, Giurgea L, 30. Solal-Celigny P, Roy P, Colombat P, et al. Subcutaneous vs intravenous rituximab in
et al. Rituximab maintenance treatment of Follicular lymphoma international prognos- patients with non-Hodgkin lymphoma: a
relapsed/resistant follicular non-Hodgkin's tic index. Blood. 2004;104(5):1258-1265. time and motion study in the United
lymphoma: long-term outcome of the 31. Forstpointner R, Unterhalt M, Dreyling M, Kingdom. J Med Econ. 2014;17(7):459-468.
EORTC 20981 phase III randomized inter- et al. Maintenance therapy with rituximab 36. Rule S, Briones J, Smith R, et al. Preference
group study. J Clin Oncol. 2010;28(17):2853- leads to a significant prolongation of for rituximab subcutaneous (SC) and intra-
2858. response duration after salvage therapy with venous (IV) among patients with CD20+
25. Oh SY, Kim WS, Kim JS, et al. Phase II study a combination of rituximab, fludarabine, non-Hodgkin's lymphoma (NHL) complet-
of R-CVP followed by rituximab mainte- cyclophosphamide, and mitoxantrone (R- ing the RASQ measure In randomized phase
nance therapy for patients with advanced FCM) in patients with recurring and refrac- III studies Prefmab and Mabcute. Value
marginal zone lymphoma: consortium for tory follicular and mantle cell lymphomas: Health. 2014;17(7):A537.
improving survival of lymphoma (CISL) Results of a prospective randomized study 37. Sakai R, Ohmachi K, Sano F, et al.
study. Cancer Commun (Lond) 2019;39 of the German Low Grade Lymphoma Bendamustine-120 plus rituximab therapy
(1):58. Study Group (GLSG). Blood. 2006;108(13): for relapsed or refractory follicular lym-
26. Ferrario A, Pulsoni A, Olivero B, et al. 4003-4008. phoma: a multicenter phase II study. Ann
Fludarabine, cyclophosphamide, and ritux- 32. Martinelli G, Schmitz SF, Utiger U, et al. Hematol. 2019;98(9):2131-2138.
imab in patients with advanced, untreated, Long-term follow-up of patients with follic- 38. Salar A, Avivi I, Bittner B, et al. Comparison
indolent B-cell nonfollicular lymphomas: ular lymphoma receiving single-agent ritux- of subcutaneous versus intravenous admin-
phase 2 study of the Italian Lymphoma imab at two different schedules in trial istration of rituximab as maintenance treat-
Foundation. Cancer. 2012;118(16):3954- SAKK 35/98. J Clin Oncol. 2010;28(29): ment for follicular lymphoma: results from a
3961. 4480-4484. two-stage, phase IB study. J Clin Oncol.
27. Bachy E, Seymour JF, Feugier P, et al. 33. Barta SK, Li H, Hochster HS, et al. 2014;32(17):1782-1791.

ARTICLE Myeloid Biology
Ferrata Storti Foundation

Unique ethnic features of DDX41 mutations in
patients with idiopathic cytopenia of
undetermined significance, myelodysplastic
syndrome, or acute myeloid leukemia
Eun-Ji Choi,1* Young-Uk Cho,2* Eun-Hye Hur,1 Seongsoo Jang,2 Nayoung Kim,3
Han-Seung Park,1 Jung-Hee Lee,1 Kyoo-Hyung Lee,1 Si-Hwan Kim,2
Sang-Hyun Hwang,2 Eul-Ju Seo,2 Chan-Jeoung Park2 and Je-Hwan Lee1
1
Haematologica 2022 Department of Hematology, Asan Medical Center, University of Ulsan College of
Volume 107(2):510-518 Medicine; 2Department of Laboratory Medicine, Asan Medical Center, University of Ulsan
College of Medicine and 3Asan Institution for Life Sciences and Department of
Convergence Medicine, Asan Medical Center, University of Ulsan College of Medicine,
Seoul, Korea
*E-JG and Y-UC contributed equally as co-first authors.
ABSTRACT
D
DX41 mutations are associated with hematologic malignancies
including myelodysplastic syndrome (MDS) and acute myeloid
leukemia (AML), but the incidence in idiopathic cytopenia of
undetermined significance (ICUS) is unknown. We investigated the inci-
dence, genetic characteristics, and clinical features of DDX41 mutations
in Korean patients with ICUS, MDS, or AML. We performed targeted
deep sequencing of 61 genes including DDX41 in 457 patients with ICUS
(n=75), MDS (n=210), or AML (n=172). Germline DDX41 mutations
with causality were identified in 28 (6.1%) patients, of whom 27
(96.4%) had somatic mutations in the other position of DDX41.
Germline origins of the DDX41 mutations were confirmed in all of the
11 patients in whom germline-based testing was performed. Of the
germline DDX41 mutations, p.V152G (n=10) was most common, fol-
lowed by p.Y259C (n=8), p.A500fs (n=6), and p.E7* (n=3). Compared
with non-mutated patients, patients with a DDX41 mutation were more
frequently male, older, had a normal karyotype, low leukocyte count,
Correspondence: and hypocellular marrow at diagnosis. Three of the four ICUS patients
JE-HWAN LEE with germline DDX41 mutations progressed to MDS. The incidence of
jhlee3@amc.seoul.kr DDX41 mutations in Korean patients was high and there was a distinct
mutation pattern, in that p.V152G was a unique germline variant. ICUS
Received: August 27, 2020. harboring germline DDX41 mutations may be regarded as a hereditary
Accepted: February 2, 2021. myeloid neoplasm. Germline DDX41 mutations are not uncommon and
should be explored when treating patients with myeloid malignancies.
Pre-published: February 25, 2021.
https://doi.org/10.3324/haematol.2020.270553 Introduction
Inherited hematologic malignancies have been established in well-defined

©2022 Ferrata Storti Foundation hereditary syndromes, which exhibit a particular phenotype, often present in
Material published in Haematologica is covered by copyright. childhood, or have a strong family history.1 There is also an increasing awareness
All rights are reserved to the Ferrata Storti Foundation. Use of of additional autosomal dominant genetic aberrations with predisposition to
conditions:
myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML), which are
https://creativecommons.org/licenses/by-nc/4.0/legalcode. primarily sporadic diseases and typically present in older adults. The recently
Copies of published material are allowed for personal or inter- updated World Health Organization (WHO) classification defined myeloid neo-
nal use. Sharing published material for non-commercial pur- plasms with germline predisposition2 and categorized familial myeloid neoplasms
into three groups: those with an absence of pre-existing disorder or organ dysfunc-
sect. 3. Reproducing and sharing published material for com- tion (e.g., CEBPA or DDX41 mutations), those with a pre-existing platelet disorder
mercial purposes is not allowed without permission in writing (e.g., RUNX1 mutations), and those with dysfunction of other organs (e.g., GATA2
from the publisher. mutations). DDX41 mutations have recently joined the growing list of genetic
alterations in familial myeloid malignancies.3,4 MDS or AML with germline
DDX41 mutations usually occurs in the sixth decade of life or beyond, whereas

Ethnic features of DDX41 mutations in ICUS/MDS/AML
most other cases of cancers with germline predisposing The Institutional Review Board of Asan Medical Center
mutations typically develop in adolescence or early adult- approved the protocols of this study (2018-0042 and 2018-0048
hood.5-11 [for prospective and retrospective analysis of patients with lower-
The DDX41 gene is located at 5q35.3 and encodes a risk MDS or ICUS], 2019-0794 [for sequencing the DDX41 gene in
DEAD-box RNA helicase, which is involved in pre-mRNA DDX41-mutated patients and their family members], and 2020-
splicing, RNA processing, and ribosome biogenesis.12 0131 [for retrospective analysis of data from patients with higher-
Several mechanisms have been proposed to explain the risk MDS or AML]), which was carried out in accordance with the
contributions of DDX41 mutations to the development of 2008 Declaration of Helsinki.
hematologic malignancies. DDX41 mutations can: (i)
cause aberrant mRNA splicing leading to exon retention or Mutational and cytogenetic analysis
exon skipping, (ii) disrupt the STING-interferon pathway; For rhe NGS assay, we prepared the sequencing libraries from
and (iii) induce aberrant pre-rRNA trimming and ribosome genomic DNA using customized probes (Integrated DNA
biogenesis.3 DDX41 mutations include both germline and Technologies, Inc., Coralville, IA, USA) to capture and enrich the
somatic mutations, with the latter being found in over half entire coding regions of 61 target genes (HEMEaccuTest DNA
of the patients with germline mutations in the other allele Target Enrichment kit; NGeneBio, Seoul, Korea) (Online
of DDX41.5 In recent studies, germline DDX41 variants Supplementary Table S1). We carried out sequencing on the
were found in 2.4% of 1,385 patients with MDS or MiSeqDx (Illumina, San Diego, CA, USA) with 2×150 bp, paired-
AML,11 and germline or somatic DDX41 variants were end reads according to the manufacturer’s instructions. Initial read
found in 3.4% of 1,002 patients with myeloid neo- mapping was carried out against the human reference genome
plasms.10 (hg19/GRCh37). We subsequently analyzed the sequencing data
Following advances in genetic testing, clinical next-gen- for variant calling using commercial software (CLC Genomics
eration sequencing (NGS)-based leukemia panels are Workbench; QIAGEN Bioinformatics, Redwood City, CA, USA).
being increasingly used to identify somatic mutations to We retained the potentially pathogenic variants by filtering out
facilitate the diagnosis, improve prognostication, and common polymorphisms (minor allele frequency in the popula-
select optimal treatment strategies in patients with hema- tion ≥1%) and sequencing/mapping errors, and by filtering in the
tologic malignancies. Some variants found in the panels known oncogenic variants based on the available population or
can also be germline mutations in genes associated with cancer mutation-specific databases. We set the minimum cutoff of
hereditary hematopoietic malignancies.13-15 The myeloid variant allele frequency at 2.0% for reporting. We performed the
leukemia panel used at our institute includes the DDX41 cytogenetic analysis using conventional G-banding techniques
gene, and the frequencies of DDX41 mutations in Korean based on the analysis of 20 or more metaphase cells.
patients with MDS or AML appeared to be higher than
the previously reported incidences. Importantly, DDX41 Germline variant confirmation and determination of
mutations have not been evaluated in patients with idio- causality
pathic cytopenia of undetermined significance (ICUS), Variants with allele frequencies between 40% and 60% were
which is a known precursor lesion of MDS. In this study, considered to be probable germline mutations. We performed
we investigated the incidence, genetic characteristics, and germline-based testing in 11 of 34 patients with probable germline
clinical features of DDX41 mutations in Korean patients DDX41 mutations using sorted blood T cells. This strategy of
with ICUS, MDS, or AML. using sorted T cells was based on recent work confirming that
T cells yield sufficient DNA and high rates of somatic variant calls
in MDS. It was suggested that, given the challenge of obtaining
Methods skin biopsies, T cells would be preferential germline tissues for
MDS genomic studies.18 Peripheral blood mononuclear cells were
Patients harvested by standard Ficoll (GE Healthcare, Sweden) density gra-
We included patients with ICUS, MDS, or AML whose bone dient centrifugation, and T cells were isolated using the Pan T Cell
marrow samples were collected between 2009 and 2019 at Asan Isolation Kit, human (MACS Miltenyi Biotec, Auburn, CA, USA)
Medical Center (Seoul, Korea). Patients with ICUS or lower-risk according to the manufacturer’s instructions. The isolated T cells
MDS were either prospectively enrolled (since January 2018) or were analyzed with CD3-FITC using a FACSCalibur (Becton
retrospectively analyzed, while those with higher-risk MDS or Dickinson, Franklin Lakes, NJ, USA), and genomic DNA was puri-
AML were retrospectively analyzed. All patients in the study fied by the QIAamp DNA mini kit (Qiagen,
cohort were unrelated individuals, not including an index case and QIAGEN GmbH, Germany). The pathogenicity of probable
his or her family members. Diagnoses of MDS and AML were germline DDX41 mutation was determined according to the
based on the WHO 2016 Classification.2 ICUS was defined by the guideline from the American College of Medical Genetics and
proposed criteria of the 2007 Consensus Group:16 cytopenia in one Genomics (ACMG).19 The concurrence of a somatic DDX41 muta-
or more of cell lineages for ≥6 months (hemoglobin <11 g/dL, neu- tion was considered as strong evidence for causality. Thus, we
trophils <1.5x109/L, and platelets <100x109/L) while excluding classified germline DDX41 variants as “causal” if they were either
other causes of cytopenia such as a history of pelvic irradiation or pathogenic (or likely pathogenic) by the ACMG guideline or
cytotoxic chemotherapy, splenomegaly, heart failure or liver cir- accompanied by a somatic DDX41 mutation regardless of the
rhosis with portal hypertension, active viral infections, and a his- ACMG interpretation.
tory of blood or bone marrow diseases. Clonal cytopenia of unde-
termined significance was defined as ICUS with myeloid neo- Statistical analysis
plasm-related somatic mutations with a variant allele frequency ≥ Categorical variables were compared using the χ2 test or Fisher
2%, or clonal karyotypic abnormalities. Myeloid neoplasm-relat- exact test, and continuous variables were compared using the
ed somatic mutations were based on those specified in the updat- Mann-Whitney U-test or the Student t-test, as appropriate.
ed National Comprehensive Cancer Network guideline for Survival was calculated by the Kaplan-Meier method and the
MDS.17 resulting survival curves were compared using the log-rank test

E.-J. Choi et al.
Table 1. Patients’ characteristics at diagnosis.

Characteristic Total (n = 457) ICUS (n = 75) MDS (n = 210) AML (n = 172)
Sex, n(%)
Male 272 (59.5) 37 (49.3) 134 (63.8) 101 (58.7)
Female 185 (40.5) 38 (50.7) 76 (36.2) 71 (41.3)
Median age (range), years 59 (16-89) 54 (19-89) 61 (18-87) 57 (16-81)
ICUS, n(%)
CCUS* 42 (56.0)
Non-CCUS 33 (44.0)
WHO classification
MDS, n(%)
MDS with SLD/RS-SLD 41 (19.5)
MDS with MLD/RS-MLD 73 (34.8)
MDS with EB-1 45 (21.4)
MDS with EB-2 18 (8.6)
MDS, unclassifiable 26 (12.4)
MDS with isolated del(5q) 2 (1.0)
Unknown 2 (1.0)
AML, n(%)
AML with RGA 84 (48.8)
AML with MRC 40 (23.3)
Therapy-related 7 (4.1)
AML, NOS 40 (23.3)
MPAL 1 (0.6)
Risk stratification, n(%)
MDS
IPSS-R score ≤ 3.5 75 (35.7)
IPSS-R score > 3.5 135 (64.3)
AML#
Favorable 63 (36.6)
Intermediate 35 (20.3)
Adverse 73 (42.4)
Unknown 1 (0.6)
Karyotype, n(%)
Normal 230 (50.3) 69 (92.0) 92 (43.8) 69 (40.1)
Abnormal 227 (49.7) 6 (8.0) 118 (56.2) 103 (59.9)
ICUS: idiopathic cytopenia of undetermined significance; CCUS: clonal cytopenia of undetermined significance; WHO: World Health Organization; MDS: myelodysplastic syn-
drome; SLD: single lineage dysplasia; RS: ring sideroblasts; MLD: multilineage dysplasia; EB: excess blasts; AML: acute myeloid leukemia; RGA: recurrent genetic abnormalities;
MRC: myelodysplasia-related changes; NOS, not otherwise specified; MPAL, mixed phenotype acute leukemia; IPSS-R, International Prognostic Scoring System-Revised. *CCUS
was defined as ICUS with myeloid neoplasm-related somatic mutations of variant allele frequency ≥2%, or clonal karyotypic abnormalities. #Risk stratification of AML according
to the 2017 European LeukemiaNet risk stratification.
(univariate analysis). The Kaplan-Meier survival curves were ren- risk in 75 (35.7%) and higher-risk in 135 (64.3%) accord-
dered as a graph using Prism version 5.0 (GraphPad Software, Inc., ing to the Revised International Prognostic Scoring System
La Jolla, CA, USA). In all analyses, the P-values were two-tailed (IPSS-R).20 Of the AML patients, 63 (36.6%), 35 (20.3%),
and those <0.05 were considered statistically significant. and 73 (42.4%) were classified into favorable, intermedi-
ate, and adverse genetic risk categories, respectively,
according to the 2017 European LeukemiaNet risk stratifi-
Results cation.21
Patients’ characteristics Frequency and genetic characteristics of DDX41

The clinical characteristics of the 457 included patients mutations
at diagnosis are shown in Table 1. There were 75 patients We detected genetic DDX41 mutations in 39 (8.5%)
(16%) with ICUS, 210 (46%) with MDS and 172 (38% patients. Thirty-four (7.4%) patients had germline muta-
with AML. The median age at diagnosis was 59 years tions, of whom 27 (79.4%) also had somatic mutations at
(range, 16-89), and 60% were men. Forty-two (56.0%) of the other position of DDX41. Five (1.1%) patients had
the ICUS patients had clonal cytopenia of undetermined somatic DDX41 mutations only. In 28 patients, the
significance. Disease risk of the MDS patients was lower- germline DDX41 mutations were considered causal and

Figure 1. Frequency of DDX41

mutations according to the type
of hematologic malignancy.
ICUS: idiopathic cytopenia of
undetermined significance; MDS:
myelodysplastic syndrome; AML:
acute myeloid leukemia.
only these patients were included in further analyses. The less frequently associated with the germline p.V152G vari-
frequency of the causal germline DDX41 mutations was ant (3 of 10) than with p.Y259C (6 of 8) or p.A500fs (4 of
6.1% (28 of 457); 6.7% (5 of 75) in ICUS, 9.0% (19 of 210) 6), whereas the somatic p.T227M variant tended to be
in MDS, and 2.3% (4 of 172) in AML (Figure 1). Detailed more frequently associated with p.V152G (4 of 10) than
information on the DDX41 variants, concurrent mutations with p.Y259C (1 of 8) or p.A500fs (0 of 6) (Online
of other genes, and karyotypes in the 28 patients are pro- Supplementary Table S6).
vided in Online Supplementary Table S2. Germline origins of Twenty-two (78.6%) of the 28 patients with mutations
the DDX41 mutations were confirmed in all of the 11 in DDX41 had concurrent mutations in other genes.
patients who underwent germline-based testing (p.V152G Genes mutated in over 10% of the patients were PHF6
in 5, p.Y259C in 3, p.A500fs in 2, and p.L328R in 1). and ASXL1 (5 patients [17.9%] each), followed by CBL
Of the 55 DDX41 mutations detected in this study, 28 and NF1 (4 patients [14.3%] each), and DNMT3A and
were germline and the other 27 appeared to be somatic. TP53 (3 patients [10.7%] each) (Figure 2B; Online
All of the somatic mutations were missense, whereas Supplementary Table S2). We observed six variants of the
germline mutations were missense in 19 (67.9%) cases, PHF6 gene in five patients with DDX41 germline muta-
frameshift in six (21.4%), and nonsense in three (10.7%). tions: p.M1T and p.R116* in one patient, and p.G248D,
The majority of somatic mutations were located in the p.C20F, p.M1T and p.M1V in one patient each.
helicase C or C-terminal domain (n=18, 66.7%), whereas Interestingly, PHF6 p.M1T/V variants were detected in
the majority of germline mutations were in the helicase only three patients harboring DDX41 germline mutations
ATP-binding or N-terminal domain (n=22, 78.6%; among the whole study population of 457 patients.
P=0.001) (Figure 2A). Of the germline DDX41 mutations,
p.V152G (n=10, 35.7%) was the most common, followed Clinical features and outcomes of the patients with
by p.Y259C (n=8, 28.6%), p.A500fs (n=6, 21.4%), p.E7* DDX41 mutations
(n=3, 10.7%), and p.L328R (n=1, 3.6%). There was a male predominance among the
Two germline variants (p.A500fs and p.E7*) were classi- DDX41-mutated patients (96.4% vs. 57.1%; P<0.001), and
fied as pathogenic according to the ACMG guideline. The the patients with this mutation tended to be older (median
other germline variants (p.V152G, p.Y259C, and p.L328R) 66 vs. 57 years; P<0.001), and were more likely to have a
were classified as being of uncertain significance, but were normal karyotype (75.0% vs. 48.7%; P=0.007), lower
considered causal when accompanied by somatic DDX41 white blood cell count (median 1.8 vs. 3.7×109/L; P=0.047),
mutations (Online Supplementary Table S3). Notably, four and lower marrow cellularity (median 30% vs. 60%;
mutations (p.V152G, p.Y259C, p.A500fs and p.E7*) were P<0.001) at diagnosis compared with the non-mutated
found at a significantly higher frequency in the study patients (Table 2). Among patients with MDS, the DDX41
patients than in healthy Koreans, as shown by high odds mutations were significantly more frequent in the MDS
ratios (38.5, 17.3, 49.6 and 26.5, respectively) (Online subtypes with excess blasts (EB)-1 and EB-2, compared to
Supplementary Table S4). Two germline mutations other categories with bone marrow blasts <5%, although
(p.V152G and p.Y259C) were only detected in ICUS/MDS the mutation frequencies were not significantly different
(75.0%) and not in AML (0%), whereas p.A500fs and between patients with lower risk or higher risk according
p.E7* were detected in both ICUS/MDS and AML groups to the IPSS-R (Table 2). Of 23 MDS or AML patients with
(Online Supplementary Table S5). Of the 27 somatic DDX41 causal germline DDX41 mutations, data regarding blood
mutations, p.R525H (n=14, 51.9%) was the most com- counts before diagnosis were available for 16 patients, and
mon, followed by p.T227M (n=5, 18.5%), and the remain- all 16 patients had a history of cytopenia at least 1 year
ing eight somatic DDX41 mutations were detected in one prior to diagnosis.
(3.7%) patient each. The somatic p.R525H variant was During the median follow-up of 25.5 months, 116

E.-J. Choi et al.
Table 2. Comparison of clinical features according to the presence of germline DDX41 mutations.
DDX41 mutations (+) DDX41 mutations (-) P
(n=28) (n=429)
Sex, n(%)
Male 27 (96.4) 245 (57.1) < 0.001a
Female 1 (3.6) 184 (42.9)
Median age (range), years 66 (41-79) 57 (16-89) < 0.001c
Chromosome, n(%)
Normal 21 (75.0) 209 (48.7) 0.007a
Abnormal 7 (25.0) 220 (51.3)
WBC, × 109/ L, median (range) 1.8 (1.0-3.3) 3.7 (0.7-313.1) 0.047c
Hb, g/dL, median (range) 10.1 (5.2-13.2) 9.1 (2.3-16.4) 0.113c
Platelets, × 109/ L, median (range) 90 (13-174) 68 (3-638) 0.689c
BM cellularity, %, median (range) 30 (5-60) 60 (3-100) < 0.001c
BM blasts, %, median (range) 6.2 (0.8-65.2) 5.2 (0-98.8) 0.016c
N. of mutated genes, median (range) 3 (2-6) 2 (0-12) 0.036c
MDS, n(%)
MDS with SLD/MLD/del(5q)/U 7 (4.8) 138 (95.2) 0.001aa*
MDS with EB-1/EB-2 12 (19.0) 51 (81.0)
Unknown 0 2 (100)
Risk stratification, n(%)
MDS 0.693a
IPSS-R ≤ 3.5 (%) 6 (31.6) 69 (36.1)
IPSS-R > 3.5 (%) 13 (68.4) 122 (63.9)
AML 0.215b
Favorable 0 (0) 63 (37.5)
Intermediate 2 (50.0) 33 (19.6)
Adverse 2 (50.0) 71 (42.3)
Unknown 0 (0) 1 (0.6)
WBC: white blood cells, Hb: hemoglobin; BM, bone marrow; MDS: myelodysplastic syndrome; SLD: single lineage dysplasia; MLD: multilineage dysplasia; EB: excess blasts; IPSS-
R, International Prognostic Scoring System-Revised; AML: acute myeloid leukemia. aby the χ2 test; bby the Fisher exact test; cby t-test; *SLD/MLD/del(5q)/U vs. EB-1/EB-2.
patients (7 ICUS, 55 MDS, and 54 AML) died. The 5-year AML).5,6,10,11,22 In a study comparing the clinical and genet-
overall survival rate was 60.8% in the overall population ic characteristics of DDX41 mutations in AML and MDS
and 84.6%, 62.2%, and 38.9% in patients with ICUS, patients between two ethnically distinct populations,
MDS, and AML, respectively. There was no significant germline DDX41 mutations were found in 3.9% of a
correlation between overall survival and the presence of Japanese cohort and in 0.8% of a Caucasian cohort.22
DDX41 mutations in each disease category of ICUS, Therefore, there seems to be an ethnic difference in the
MDS, and AML (Figure 3) as well as in the total study pop- incidence of DDX41 mutations in patients with myeloid
ulation (Online Supplementary Figure S1). neoplasms between Asian and Western patients. In con-
Online Supplementary Table S7 shows the clinical course trast, the clinical features of our DDX41 -mutated patients,
of each patient with a DDX41 mutation. Clinical courses such as male predominance, old age at presenta-
could be followed up in four of the five ICUS patients tion,5,6,10,11,23 hypocellular marrow,3,4,6 leukopenia,6 and fre-
with probable germline DDX41 mutations, and notably, quent normal cytogenetics3-6,11 were similar to those
three of these four patients showed disease progression to reported in other ethnic populations. The DDX41 muta-
MDS EB-1 (n=2; 77.9 and 17.6 months after ICUS diagno- tions did not show significant associations with survival
sis) or MDS EB-2 with a gain of PTPN11 mutation (n=1; 9 outcomes.
months after ICUS diagnosis) during the follow-up. There are several noteworthy findings in our study
Another ICUS patient with a germline DDX41 mutation regarding the genetic characteristics of DDX41 muta-
had a son with Hodgkin lymphoma. tions. First, the germline mutations were mostly N-termi-
nal variants (78.6%), whereas somatic mutations were
mostly C-terminal variants (66.7%). This finding is con-
Discussion sistent with the observations in two recent studies.10,11
The N-terminal region of DDX41 has the helicase ATP
In our cohort of 457 patients with ICUS, MDS, or AML, binding domain,24,25 and the structural rearrangement in
6.1% of the patients carried causal germline DDX41 muta- the N-terminal region may change the conformation of
tions, which is a higher incidence than those found in pre- the ATP-binding site and eventually decrease ATP-bind-
vious studies which ranged between 0.8% and 3.9% in ing ability.24 In contrast, the helicase C-terminal domain is
patients with myeloid malignancies (mostly MDS and involved in ATP hydrolysis.24,25 Therefore, genetic alter-

Figure 2. Distribution of DDX41 mutations and concurrent mutations in other genes. (A) Distribution of DDX41 mutations detected in the current study and two pre-
vious studies (Quesada et al.10 and Sebert et al.11). This figure shows the differences in positional distribution (N-terminal skewed vs. C-terminal skewed) and muta-
tional effects (variable vs. missense-dominated) between germline and somatic mutations. The protein structure of DDX41 was based on the RefSeq accession num-
ber of NM_016222.3 and the UniProtKB entry of Q9UJV9: the 622 amino acid long protein comprises the helicase ATP-binding domain (position 212-396), the heli-
case C-terminal domain (position 407-567), and a zinc finger domain (position 580-597). Different colors indicate different effects of mutations: light blue, missense
mutation; light green, inframe indel; purple, nonsense mutation; brown, splicing mutation; red, frameshift mutation; black, start codon loss. Different shapes repre-
sent the three studies: square, Sebert et al.11 diamond, Quesada et al.10 circle, current study. (B) Concurrent mutations of other genes identified in bone marrow sam-
ples from DDX41-mutated patients. The types of genetic alterations and diseases are presented in the legend.

E.-J. Choi et al.
ations in the N-terminal region may cause greater suscep- pathways of downstream oncogenes and tumor suppres-
tibility to protein hypofunction than those in the C-termi- sor genes.3
nal region. Second, the patterns of germline DDX41 Donor-derived leukemia has been reported in several
mutations in our Korean population were distinct from families with germline DDX41 mutations; in all such
those in Western populations10,11 or even other Asian pop- cases, donors had the same type of germline DDX41 vari-
ulations.22,23 The germline DDX41 mutations (p.V152G, ants as the respective recipients.31,32 In our study, DDX41
p.Y259C, p.A500fs, p.E7*, and p.L328R) in our study are mutations (germline p.E7* and somatic p.G228C) were
totally different from those reported in Western popula- found in a 60-year-old man with high-risk MDS (#12). No
tions (p.M1I, p.D140fs, p.G173R, and Q41*). Korean and
Japanese patients shared three major germline DDX41
variants (p.Y259C, p.A500fs, p.E7*),22 but p.V152G was A
only found in Koreans and not in Japanese or other ethnic
populations. Third, we observed the exclusive presence
of PHF6 p.M1T/V variants in three patients with proba-
ble germline DDX41 mutations. These variants potential-
ly cause a complete lack of protein production as a conse-
quence of start codon loss and are causative germline
mutations of the Börjeson-Forssman-Lehmann
syndrome.26-28 Thus, our findings suggest that the same
genetic mutation can induce both hereditary diseases and
sporadic cancer, as exemplified by mutations in ETV6.29
The possible association between PHF6 p.M1T/V vari-
ants and germline DDX41 mutations should be investi-
gated further.
Germline mutations that predispose an individual to
MDS or AML may also contribute to the development of
ICUS, but the genetic predisposition to ICUS has not been
systematically investigated. In a recent study of germline B
DDX41 mutations in adult patients with MDS or AML,
45.5% of patients with pathogenic germline DDX41
mutations had a previous history of cytopenia before the
diagnosis of MDS or AML, and the preexisting cytopenia
might indicate the presence of ICUS in these patients.11
We also observed similar findings. Furthermore, five
(6.7%) of 75 ICUS patients had causal germline DDX41
mutations, three of whom progressed to MDS. Our study
shows that germline DDX41 mutations are not uncom-
mon in ICUS patients. Our findings do not indicate that
the germline DDX41 mutations contribute to the progres-
sion of ICUS to MDS, but instead do suggest that ICUS
patients harboring such variants may be considered as
having a hereditary myeloid neoplasm.
Our observations highlight the potential oncogenic role
of germline DDX41 mutations in the pathogenesis of
ICUS/MDS in comparison with AML. First, the patients C
with ICUS/MDS carried germline DDX41 mutations more
frequently than did AML patients and germline missense
mutations were highly enriched in ICUS/MDS rather than
in AML. These findings might support the notion that less
disruptive variants are associated with a milder phenotype
in the disease spectrum. Second, only one (3.6%) of the 28
patients with germline DDX41 mutations carried a muta-
tion in the splicing factor gene. This finding is in line with
previous observations that splicing factor gene mutations
were largely mutually exclusive with DDX41 muta-
tions.5,10 DDX41 interacts with core splicing proteins such
as SF3B, U2 complex, PRPF8 scaffold protein, and U5 com-
plex, indicating that spliceosomal proteins are the top
functional group associated with DDX41.5,25 Genetic alter-
ations of the splicing components affect the 3’-splice site
recognition during pre-mRNA processing and are involved
Figure 3. Overall survival of patients with different hematologic disorders
in the pathogenesis of myelodysplasia.30 This indicates according to DDX41 mutation status. (A-C) Overall survival of patients with idio-
that mutant DDX41 can have an oncogenic role in MDS pathic cytopenia of undetermined significance (A), myelodysplastic syndrome (B)
via aberrant mRNA splicing with the assumption that or acute myeloid leukemia (C) according to whether they had DDX41 mutations
(red) or not (blue).
mutations in these splicing factors have an impact on the

HLA-matched sibling or unrelated donor was available, studies.5,6,10,11 This also has provided another illustration
and his two adult offspring had the same DDX41 muta- that germline alterations predispose to the acquisition of
tion (p.E7*). Fortunately, the patient’s somatic mutations in the same genes which act as a sec-
HLA-haploidentical brother did not carry the DDX41 ond hit being associated with cancer development as
mutation, and the patient could undergo haploidentical demonstrated by JAK2, CEBPA, and RUNX1 muta-
hematopoietic stem cell transplantation from him. tions.34,35 Therefore, in cases harboring a germline DDX41
Considering that the risk of malignancy in DDX41 carriers mutation, the acquisition of a somatic DDX41 mutation
is yet to be determined, we calculated the odds ratios of should be considered as strong evidence for causality.
major germline DDX41 variants detected in our study Lastly, our data may not reflect the whole Korean popula-
(Online Supplementary Table S4). Nevertheless, an extensive tion, although study patients included in this study come
population-based study is needed to obtain more reliable from all across the Korean peninsula.
data that may be useful in establishing genetic counseling In conclusion, our results delineate the unique ethnic
guidelines for germline DDX41 variants, which are cur- features of DDX41 mutations in Korean patients, such as
rently available only for donor selection in allogeneic higher incidence and different patterns, compared with
hematopoietic stem cell transplantation. patients from Western countries or other Asian countries.
NGS-based targeted genotyping for somatic mutations Specifically, the most common germline mutation in our
can identify patients who are at risk of hereditary cohort was p.V152G, which was not found in previous
hematopoietic malignancies. In a recent study, of 25 studies in other ethnicities. Our results suggest that ICUS
pathogenic or likely pathogenic variants with variant harboring germline DDX41 mutations may be regarded as
allele frequency >40% in 24 patients with germline tis- a hereditary myeloid neoplasm. Germline DDX41 muta-
sues available, six variants (24%) were of germline origin tions may be predicted with a high probability by using
– three DDX41 variants, two GATA2 variants, and one clinical NGS-based leukemia panels based on variant allele
TP53 variant; DDX41 had a 100% diagnostic yield for frequency levels and public databases. Germline DDX41
pathogenic germline variants in that study.33 In another mutations are not uncommon and should be explored
study, targeted NGS showed that 17 patients had puta- when treating patients with myeloid malignancies.
tive germline DDX41 variants with a variant allele fre-
quency >40%, all of which were of germline origin.11 We Disclosures
were also able to confirm germline origin in all of the 11 No conflicts of interest to disclose.
patients with probable germline DDX41 mutations. In
cases in which germline samples are not available, NGS- Contributions
based leukemia panels seem to predict germline DDX41 E-JC and Y-UC analyzed and interpreted the data; E-JC, Y-
variants with high probability. However, it is worth UC and J-HL contributed to the manuscript; E-HH performed
mentioning that NGS-based panels may fail to detect experiments. All authors provided patients’ data, reviewed, and
deletions or gene rearrangements that are responsible for approved the final manuscript.
the predisposition syndrome.
Our study has some limitations. The number of patients Acknowledgments
included in the study was relatively small, and this might We thank Dr. Joon Seo Lim from the Scientific Publications
have had an impact on the analysis for clinical associations Team at Asan Medical Center for his editorial assistance in
of DDX41 mutations with clinical outcomes. Family his- preparing this manuscript.
tory was not systematically collected in this study,
although such information is helpful in pinpointing the Funding
pathogenicity of sequence variants. We did not perform This research was supported by a Basic Science Research
functional studies to demonstrate that sequence variants Program through the National Research Foundation of Korea
detected in this study had a deleterious effect in vivo. (NRF) funded by the Ministry of Science, ICT and Future
Experimental data can be useful to support pathogenicity, Planning (NRF-2017R1E1A1A01074383). The biospecimens
particularly for missense variants of uncertain significance. and data used in this study were provided by Asan Bio-Resource
We acknowledge that these limitations may hamper the Center, Korea Biobank Network (2018-08).
precise variant classification based on the ACMG guide-
line. However, the concurrence of germline and somatic Data-sharing statement
DDX41 mutations was a recurrent finding across recent For original data, please contact imeunjeee@gmail.com.
References Muller-Tidow C. DDX41-related myeloid acute myeloid leukemia. Leukemia. 2016;

neoplasia. Semin Hematol. 2017;54(2):94- 30(10):2083.
1. Rafei H, DiNardo CD. Hereditary myeloid 97. 8. Li R, Sobreira N, Witmer PD, Pratz KW,
malignancies. Best Pract Res Clin Haematol. 5. Polprasert C, Schulze I, Sekeres MA, et al. Braunstein EM. Two novel germline DDX41
2019;32(2):163-176. Inherited and somatic defects in DDX41 in mutations in a family with inherited
2. Arber DA, Orazi A, Hasserjian R, et al. The myeloid neoplasms. Cancer Cell. 2015; myelodysplasia/acute myeloid leukemia.
2016 revision to the World Health 27(5):658-670. Haematologica. 2016;101(6): e228.
Organization classification of myeloid neo- 6. Lewinsohn M, Brown AL, Weinel LM, et al. 9. Vairo FPE, Ferrer A, Cathcart-Rake E, et al.
plasms and acute leukemia. Blood. 2016; Novel germ line DDX41 mutations define Novel germline missense DDX41 variant in
127(20):2391-2405. families with a lower age of MDS/AML a patient with an adult-onset myeloid neo-
3. Cheah JJC, Hahn CN, Hiwase DK, Scott HS, onset and lymphoid malignancies. Blood. plasm with excess blasts without dysplasia.
Brown AL. Myeloid neoplasms with 2016;127(8):1017-1023. Leuk Lymphoma. 2019;60(5):1337-1339.
germline DDX41 mutation. Int J Hematol. 7. Cardoso SR, Ryan G, Walne AJ, et al. 10. Quesada AE, Routbort MJ, DiNardo CD, et
2017;106(2):163-174. Germline heterozygous DDX41 variants in al. DDX41 mutations in myeloid neoplasms
4. Maciejewski JP, Padgett RA, Brown AL, a subset of familial myelodysplasia and are associated with male gender, TP53

E.-J. Choi et al.
mutations and high-risk disease. Am J Association for Molecular Pathology. Genet Forsman-Lehmann syndrome in a family
Hematol. 2019;94(7):757-766. Med. 2015;17(5):405-423. with four affected young boys. Mol
11. Sébert M, Passet M, Raimbault A, et al. 19. Padron E, Ball MC, Teer JK, et al. Germ line Syndromol. 2015;6(4):181-186.
Germline DDX41 mutations define a signif- tissues for optimal detection of somatic vari- 28. Todd MA, Ivanochko D, Picketts DJ. PHF6
icant entity within adult MDS/AML ants in myelodysplastic syndromes. Blood. degrees of separation: the multifaceted roles
patients. Blood. 2019;134(17):1441-1444. 2018;131(21):2402-2405. of a chromatin adaptor protein. Genes.
12. Linder P. Dead-box proteins: a family affair- 20. Greenberg PL, Tuechler H, Schanz J, et al. 2015;6(2):325-352.
-active and passive players in RNP-remodel- Revised international prognostic scoring sys- 29. Feurstein S, Godley LA. Germline ETV6
ing. Nucleic Acids Res. 2006;34(15):4168- tem for myelodysplastic syndromes. Blood. mutations and predisposition to hematolog-
4180. 2012;120(12):2454-2465. ical malignancies. Int J Hematol. 2017;
13. DiNardo CD, Routbort MJ, Bannon SA, et 21. Döhner H, Estey E, Grimwade D, et al. 106(2):189-195.
al. Improving the detection of patients with Diagnosis and management of AML in 30. Yoshida K, Sanada M, Shiraishi Y, et al.
inherited predispositions to hematologic adults: 2017 ELN recommendations from an Frequent pathway mutations of splicing
malignancies using next-generation international expert panel. Blood. 2017;129 machinery in myelodysplasia. Nature.
sequencing-based leukemia prognostication (4):424-447. 2011;478(7367):64-69.
panels. Cancer. 2018;124(13):2704-2713. 22. Takeda J, Yoshida K, Makishima H, et al. 31. Kobayashi S, Kobayashi A, Osawa Y, et al.
14. Drazer MW, Kadri S, Sukhanova M, et al. Genetic predispositions to sporadic myeloid Donor cell leukemia arising from
Prognostic tumor sequencing panels fre- neoplasms caused by germline DDX41 preleukemic clones with a novel germline
quently identify germ line variants associat- mutations in Asian and Caucasian popula- DDX41 mutation after allogenic hematopoi-
ed with hereditary hematopoietic malignan- tions. Haematologica. 2016; 101 (s1):66-67. etic stem cell transplantation. Leukemia.
cies. Blood Adv. 2018;2(2):146-150. 23. Polprasert C, Takeda J, Niparuck P, et al. 2017;31(4):1020.
15. Guidugli L, Johnson AK, Alkorta-Aranburu Novel DDX41 variants in Thai patients with 32. Berger G, van den Berg E, Sikkema-Raddatz
G, et al. Clinical utility of gene panel-based myeloid neoplasms. Int J Hematol. B, et al. Re-emergence of acute myeloid
testing for hereditary myelodysplastic syn- 2020;111(2):241-246. leukemia in donor cells following allogeneic
drome/acute leukemia predisposition syn- 24. Omura H, Oikawa D, Nakane T, et al. transplantation in a family with a germline
dromes. Leukemia. 2017;31(5):1226-1229. Structural and functional analysis of DDX41 mutation. Leukemia. 2017;31(2):
16. Valent P, Horny H-P, Bennett JM, et al. DDX41: a bispecific immune receptor for 520.
Definitions and standards in the diagnosis DNA and cyclic dinucleotide. Sci Rep. 33. Drazer MW, Kadri S, Sukhanova M, et al.
and treatment of the myelodysplastic syn- 2016;6(1):1-11. Prognostic tumor sequencing panels fre-
dromes: consensus statements and report 25. Jiang Y, Zhu Y, Liu Z-J, Ouyang S. The quently identify germ line variants associat-
from a working conference. Leuk Res. emerging roles of the DDX41 protein in ed with hereditary hematopoietic malignan-
2007;31(6):727-736. immunity and diseases. Protein Cell. 2017; cies. Blood Adv. 2018;2(2):146-150.
17. NCC Network. Myelodysplastic syndrome 8(2):83-89. 34. Kilpivaara O, Mukherjee S, Schram AM, et
(Version 2.2020). https://www.nccn.org/ 26. Crawford J, Lower KM, Hennekam RC, et al. A germline JAK2 SNP is associated with
professionals/physician_gls/PDF/mds.pdf al. Mutation screening in Börjeson- predisposition to the development of JAK2
Accessed May 10, 2020. Forssman-Lehmann syndrome: identifica- V617F-positive myeloproliferative neo-
18. Richards S, Aziz N, Bale S, et al. Standards tion of a novel de novo PHF6 mutation in a plasms. Nat Genet. 2009;41(4):455-459.
and guidelines for the interpretation of female patient. J Med Genet. 2006; 35. Brown AL, Hahn CN, Scott HS. Secondary
sequence variants: a joint consensus recom- 43(3):238-243. leukemia in patients with germline tran-
mendation of the American College of 27. Ernst A, Le VQ, Højland AT, et al. The PHF6 scription factor mutations (RUNX1, GATA2,
Medical Genetics and Genomics and the mutation c. 1A> G; PM1V causes Börjeson- CEBPA). Blood. 2020;136(1):24-35.

Platelet Biology & its Disorders ARTICLE
Sequence-specific 2'-O-methoxyethyl Ferrata Storti Foundation

antisense oligonucleotides activate human
platelets through glycoprotein VI, triggering
formation of platelet-leukocyte aggregates
Martina H. Lundberg Slingsby,1,2 Prakrith Vijey,2 I-Ting Tsai,1,2 Harvey Roweth,2
Genevieve Couldwell,2 Adrian R. Wilkie,1,2 Hans Gaus,3 Jazana M. Goolsby,2
Ross Okazaki,2 Brooke E. Terkovich,2 John W. Semple,4,5 Jonathan N. Thon,2
Scott P. Henry,3 Padmakumar Narayanan3 and Joseph E. Italiano Jr.1,2
1
Vascular Biology Program, Department of Surgery, Boston Children’s Hospital, Boston,
Haematologica 2022
MA, USA; 2Division of Hematology, Brigham and Women's Hospital, Harvard Medical Volume 107(2):519-531
School, Boston, MA, USA; 3Nonclinical Development, Ionis Pharmaceuticals Inc.,
Carlsbad, CA, USA; 4Departments of Pharmacology and Medicine, University of Toronto,
Toronto, Ontario, Canada and 5Division of Hematology and Transfusion Medicine, Lund
University, Lund, Sweden
ABSTRACT
A
ntisense oligonucleotides (ASO) are DNA-based, disease-modifying
drugs. Clinical trials with 2'-O-methoxyethyl (2’MOE) ASO have
shown dose- and sequence-specific lowering of platelet counts
according to two phenotypes. Phenotype 1 is a moderate (but not clinically
severe) drop in platelet count. Phenotype 2 is rare, severe thrombocytope-
nia. This article focuses on the underlying cause of the more common phe-
notype 1, investigating the effects of ASO on platelet production and
platelet function. Five phosphorothioate ASO were studied: three 2’MOE
sequences; 487660 (no effects on platelet count), 104838 (associated with
phenotype 1), and 501861 (effects unknown) and two CpG sequences;
120704 and ODN 2395 (known to activate platelets). Human cord blood-
derived megakaryocytes were treated with these ASO to study their effects
on proplatelet production. Platelet activation (determined by surface P-
selectin) and platelet-leukocyte aggregates were analyzed in ASO-treated
blood from healthy human volunteers. None of the ASO inhibited pro-
platelet production by human megakaryocytes. All the ASO were shown
to bind to the platelet receptor glycoprotein VI (KD ~0.2-1.5 mM). CpG ASO Correspondence:
had the highest affinity to glycoprotein VI, the most potent platelet-activat- MARTINA LUNDBERG SLINGSBY
ing effects and led to the greatest formation of platelet-leukocyte aggre- martina.slingsby@gmail.com
gates. 2’MOE ASO 487660 had no detectable platelet effects, while 2’MOE
ASOs 104838 and 501861 triggered moderate platelet activation and SYK-
Received: June 22, 2020.
dependent formation of platelet-leukocyte aggregates. Donors with higher
platelet glycoprotein VI levels had greater ASO-induced platelet activation. Accepted: February 2, 2021.
Sequence-dependent ASO-induced platelet activation and platelet-leuko- Pre-published: February 11, 2021.
cyte aggregates may explain phenotype 1 (moderate drops in platelet
count). Platelet glycoprotein VI levels could be useful as a screening tool to
identify patients at higher risk of ASO-induced platelet side effects. https://doi.org/10.3324/haematol.2020.260059

Introduction
Antisense oligonucleotides (ASO) are short, synthetic, single-stranded DNA mole- All rights are reserved to the Ferrata Storti Foundation. Use of
cules (between 8 and 50 nucleotides) that bind to specific mRNA segments through conditions:
Watson-Crick base pairing.1 They are designed to modulate targeted mRNA by inter- https://creativecommons.org/licenses/by-nc/4.0/legalcode.
fering with its function or promoting its degradation, ultimately altering protein Copies of published material are allowed for personal or inter-
expression.1,2 ASO offer therapeutic opportunities for treating rare genetic diseases nal use. Sharing published material for non-commercial pur-
such as spinal muscular atrophy (nusinersen),3,4 hereditary transthyretin amyloidosis https://creativecommons.org/licenses/by-nc/4.0/legalcode,
(inotersen)5 and homozygous familial hypercholesterolemia (mipomersen).6 There sect. 3. Reproducing and sharing published material for com-
has been a surge in ASO entering clinical trials for a wide range of diseases, which mercial purposes is not allowed without permission in writing
has been attributed to improvements in chemical modifications of ASO.7 from the publisher.
Incorporating a phosphorothioate (PS)-containing backbone to the ASO increased the
stability and resistance of the oligonucleotides to nucleolytic degradation.1 Second-

M.H. Lundberg Slingsby et al.
generation ASO have the PS backbone and often include 2’- extent 2’MOE-modified PS-ASO share the platelet-activat-
O-methoxyethyl (MOE) modifications, which can further ing effects of PS-ODN 2395 on human platelets. In the cur-
reduce ASO degradation and increase affinity for the target rent study, we therefore focused on investigating the in vitro
mRNA.1 effects of 2’MOE ASO on proplatelet production from
While several trials with ASO drugs have shown 2’MOE- human cord blood-derived megakaryocytes and studied the
containing ASO drugs to be well tolerated, there have been direct effects of 2’MOE ASO on human platelets as well as
reports of adverse events including dose-dependent throm- interactions with immune cells. Specifically, we tested
bocytopenia, with two phenotypes having been 2’MOE ASO 104838, which is known to cause phenotype
described.8,9 Phenotype 1 is a moderate dose-dependent 1 platelet count reductions.10,16 We also included 2’MOE
drop in platelet counts that is reversible with cessation of ASO 501861, with unknown effects on human platelets,
drug treatment.8,10 The Ionis integrated safety database con- and 2’MOE ASO 487660, which has not been associated
cluded that certain sequences (3 out of 16 2’MOE ASO) with reductions in platelet counts in monkeys. CpG ASO
were associated with phenotype 1, with dose-dependent 2395 and another CpG 2’MOE ASO 120704 were included
moderate (>30%) declines in platelet count without an for comparison for their known platelet-activating
effect on bleeding risk.8 Phenotype 1 has also been observed effects.10,14
in monkeys treated with 40% of the evaluated 2’MOE
ASO, in which it was manifested as a consistent, repro-
ducible decline of platelets over 4-6 weeks, with the platelet Methods
counts decreasing moderately (by 30-50%) and then
remaining steady.11 Phenotype 2 is a sporadic severe drop in Human and mouse megakaryocyte cultures and
platelet count (clinical thrombocytopenia) and is often not proplatelet analysis
reproducible (for the same 2’MOE ASO), but appears to be Human cord blood-derived primary CD34+ cells were cultured
dose-dependent.11 Treatment with the 2’MOE ASO inot- and mature megakaryocytes were purified by magnetic bead sep-
ersen induced phenotype 2 in a few individuals who were aration on day 11 as previously described.17 Mouse megakaryo-
predisposed, in relation to their underlying disease, and was cytes were derived from fetal liver cultures extracted from CD-1
shown to be dependent on platelet antibodies.12 pregnant mice at day 13.5 of gestation.18 Human-derived, or
This paper will focus on the underlying cause of the more mouse-derived megakaryocytes were plated on a 96-well half-area
prevalent phenotype 1. plate (Greiner Bio one 675101), followed by addition of ASO (5
Drug-induced thrombocytopenia can stem from mM) and imaged at hourly intervals for 24 h using an IncuCyte Live
decreased production of platelets in the bone marrow Cell Analysis System (IncuCyte Zoom). These images were ana-
and/or increased destruction and clearance of platelets from lyzed for the percentage of megakaryocytes producing pro-
peripheral blood.13 To examine the in vivo effects on platelet platelets, as well as the area containing proplatelets using Ilastik
counts, monkeys were treated with the 2’MOE ASO (version 1.3.0) and Cell Profiler (version 3.0.0) as described previ-
104838 for 12 weeks.10 Platelet counts decreased ~50% in ously.19
all monkeys by day 30, and four out of five monkeys had
decreases in platelet counts consistent with phenotype 1. Blood collection
The drops in platelet counts were attributed to a 60-80% Blood was collected from healthy male and female human
increase in platelet sequestration in liver and spleen, caused donors, after informed consent and institutional review board
by either increased phagocytosis of platelets or trapping of approval (2012P001526), in accordance with the Declaration of
platelets on the reticuloendothelial surface of these organs.10 Helsinki. Washed platelets, platelet-rich plasma or whole blood
Thrombopoietin levels were not altered in the monkeys was treated for 30 min (platelet studies) or 6 h (cytokine release
and bone marrow megakaryocyte morphology, cell density experiments) with 1, 5 or 10 mM ASO. Previous studies have
and maturation appeared normal. These findings suggested shown that ASO-induced activation of platelets is concentration-
that the mild thrombocytopenia was due to increased dependent.10,14,15 A therapeutically relevant dose of ASO, i.e., 300
platelet destruction or splenic sequestration and not to mg, administered subcutaneously results in a Cmax (maximum
diminished platelet production).10 concentration that a drug achieves after dosing) of 1-2 mM.20
There have been few studies investigating the direct
effects of ASO on human platelets. Flierl et al. showed that Antisense oligonucleotides
incubating human platelets with an oligonucleotide PS- ASO were synthesized at Ionis Pharmaceuticals, Inc. (Carlsbad,
ODN 2395 (without 2’MOE modifications), led to platelet CA, USA), purified by reverse-phase high-performance liquid
activation (increased platelet P-selectin surface expres- chromatography and formulated in 10 mM HEPES. Five ASO (all
sion).14 The same study established that the PS backbone with a PS backbone) were included: three with 2’-MOE sequences;
modification of ODN 2395 is a significant driver of this 487660 (no effects on platelet count), 104838 (associated with phe-
drug’s effects on platelets, through binding and activation of notype 1),8,10 501861 (effects unknown) and two with CpG
the platelet receptor glycoprotein VI (GPVI), by enhanced sequences; 120704 and ODN 2395 (known to activate platelets)10,14
GPVI receptor clustering/dimerization.14 Sewing et al. sup- (Table 1). More details of the Methods can be found in the Online
ported these findings and demonstrated that ASO contain- Supplementary Material.
ing locked nucleic acid modifications had reduced binding
to GPVI and platelet activation.15 These studies did not
include 2’MOE-modified ASO. Considering that most ASO Results
in clinical use and in the drug development pipeline are
2’MOE ASO, it is imperative to understand how 2’MOE Proplatelet formation from human cord blood or
modifications affect platelet responses in humans. murine fetal liver-derived megakaryocytes is not
Furthermore, it remains unknown whether ASO affect pro- inhibited by 2’MOE or CpG ASO treatment
platelet production from megakaryocytes and to what To investigate whether the ASO induced thrombocy-

Effects of 2'MOE ASO on human platelets
Table 1. Antisense oligonucleotides included in this study.

Ionis Back bone Length PS load* Platelet side effects
ASO
487660 2’MOE PS 20 19 No reported drops
in platelet count
487660
Sequence MOE C*MOEC*MOEA*MOEG*MOEC*T*C*A*A*C*C*C*T*T*CMOET*MOET*MOET*MOEA*MOEA*
(5’3’)
Phenotype 1-moderate
104838 2’MOE PS 20 19 drops in platelet
count8,10
104838
Sequence MOE G*MOEC*MOET*MOEG*MOEA*T*T*A*G*A*G*A*G*A*G*MOEG*MOET*MOEC*MOEC*MOEC*
(5’3’)
501861 2’MOE PS 20 19 Unknown
501861
Sequence MOE T*MOEC*MOEA*MOEC*MOEA*G*A*A*T*T*A*T*C*A*G*MOEC*MOEA*MOEG*MOET*MOEA*
(5’3’)
120704 CpG PS 24 23 Platelet activation10
120704
Sequence T*C*pG*T*C*pG*T*T*T*T*G*T*C*pG*T*T*T*T*G*T*C*pG*T*T*
(5’3’)
818290 (ODN 2395) CpG PS 22 21 Platelet activation14
ODN 2395
Sequence T*C*pG*T*C*pG*T*T*T*T*C*pG*G*C*G*C*G*C*G*C*C*pG*
(5’3’)
All antisense oligonucleotides (ASO) were phosphorothioate (PS)-modified (locations indicated with *). MOE indicates the position of the 2'-O-methoxyethyl (2’MOE)-modified
sugar residues with 2ʹ-deoxynucleotides in between. All cytosine residues were methylated at the five position. 818290 is ODN 2395 and is referred to as ODN 2395 in the paper,
for ease of comparison to previous reports.14ODN 2395 and 120704 are non-MOE ASO with unmethylated CpG dinucleotide-rich motifs.
topenia by affecting proplatelet production from hydrophilic, poly-anionic molecules with a high degree of
megakaryocytes, human cord blood-derived mature plasma protein binding (typically >90%), and circulate
megakaryocytes were incubated with 5 mM ASO for 24 h. transiently in the blood before interacting with cell-sur-
Puromycin, an inhibitor of protein synthesis known to face proteins and typically gaining entry into cells by
impair proplatelet production,21,22 had an inhibitory effect endocytosis.23 The plasma-free conditions in the experi-
on both human- and murine-derived megakaryocytes ments shown in Figure 2 were intended to maximize visu-
(Figure 1A, B). Compared to the vehicle, none of the ASO alization of ASO binding to platelets. Electron micro-
tested lowered proplatelet counts from the human-derived graphs revealed that ASO immunogold staining was
megakaryocytes (Figure 1A). Similarly, there was no either localized to the platelet plasma membrane or, when
decrease in percent proplatelet producing murine fetal liver- internalized, to the cytoplasm with sporadic staining of
derived megakaryocytes following incubation with any of internal membranes and granules (Figure 2B-F). The
the ASO (Figure 1B). The CpG ASO 120704 slightly 2’MOE ASO 487660 appeared to stain less than the other
increased proplatelet counts in the human-derived ASO, especially at the plasma membrane (Figure 2B-F).
megakaryocytes after 22 h (117±5 vs. vehicle 79±7) (Figure To exclude any direct platelet cytotoxicity of ASO, we
1A). There was also a small increase in proplatelet-produc- performed a lactate dehydrogenase leakage assay in washed
ing murine megakaryocytes after 24 h (Figure 1B); vehicle platelets and confirmed that none of the ASO was cytotoxic
(17±1%), 487660 (25±1%), 104838 (23±1%), 501861 at doses of 1 or 5 mM (Online Supplementary Figure S1).
(25±2%), 120704 (23±1%), and ODN 2395 (26±1%).
Representative images of proplatelet production at 0, 8, 16 2’MOE ASO (104838 and 501861) and CpG ASO
and 24 h by murine megakaryocytes treated with 2’MOE increase platelet P-selectin expression in platelet-rich
ASO 104838 or CpG ASO ODN 2395 appeared compara- plasma and whole blood
ble to those of megakaryocytes treated with the vehicle To establish whether the ASO activate human platelets
(Figure 1C). directly, we investigated the effects on platelet surface P-
selectin levels using flow cytometry (Figure 3). Platelet-
Human platelets internalize both 2’MOE and CpG ASO rich plasma treated with the positive control thrombin
Given that the ASO did not appear to inhibit proplatelet receptor activating peptide (TRAP) showed a significant
production, we focused on examining the direct effects of increase in P-selectin surface expression, while the 2’MOE
ASO on human platelets. We used immune-electron ASO 487660 (which does not affect platelet count) did not
microscopy to visualize how ASO (at 5 mM) interact with increase platelet P-selectin compared to vehicle (Figure
human washed platelets. ASO with a PS backbone are 3A). Platelet activation nearly doubled after treatment

Figure 1. The effect of antisense oligonucleotides on proplatelet production by megakaryocytes derived from human cord blood and mouse fetal liver. Human and
mouse megakaryocytes (MK) were treated with vehicle (HEPES, 10 mM), puromycin (100 nM, a known inhibitor of proplatelet production), or 5 mM of the 2’MOE anti-
sense oligonucleotides (ASO) 104838 or 501861 or the CpG ASO 120704 or ODN 2395 for 24 h using the IncuCyte Live Cell Analysis System and analyzed with Ilastik
and Cell Profiler. (A) Proplatelet count: individual count of proplatelet protrusions that emanate from the human cord blood-derived MK body, over the course of the 24
h incubation. (B) Percent of proplatelet-producing mouse fetal liver-derived MK treated with the above treatments as well as the 2’MOE ASO 487660. *P<0.05 by two-
way analysis of variance with Bonferroni post-test (n=4 repeat runs). (C) Representative brightfield images at 20x magnification, of proplatelet-producing mouse fetal
liver-derived MK at 0 h, 8 h, 16 h and 24 h treated with vehicle (HEPES), the 2’MOE ASO 104838 or the CpG ASO ODN 2395. Scale bars indicate 50 mm.

with 5 mM of the 2’MOE ASO 104838 and 501861 stronger effects (comparable to the effect of TRAP) (Figure
(19±3% and 20±3%, respectively), compared to vehicle 3B). Pre-treatment with the SYK inhibitor blocked the
(10±2%); the CpG ASO 120704 and ODN 2395 activated ASO-induced P-selectin expression (Figure 3B).
the platelets more potently (38±3% and 38±4%, respec-
tively) (Figure 3A). Responsiveness to ASO treatment is strongly
We also studied platelet activation in whole blood, in correlated to individual GPVI levels
which platelet P-selectin was elevated by the platelet ago- Since we noticed donor-to-donor variability in the
nists TRAP and collagen (Figure 3B). Pre-treatment with a responsiveness to ASO treatment (Figure 3B), we investi-
Spleen tyrosine kinase (SYK) inhibitor (PRT-060318) gated whether this could be related to differential platelet
decreased collagen, but not TRAP activation since SYK is surface expression of GPVI receptors, which we measured
downstream of the collagen receptor GPVI signaling in by flow cytometry. Basal platelet GPVI levels varied
platelets (Figure 3B). The selective inhibitory effect on col- between donors and activating the platelet-rich plasma
lagen signaling is demonstrated more clearly in Figure 6A. with TRAP reduced the platelet GPVI levels (Online
The 2’MOE ASO 104838 and 501861 had mild platelet- Supplementary Figure S2), consistent with reports of GPVI
activating effects in whole blood (similar to that of colla- shedding upon platelet activation.24 Pearson correlation
gen) whereas the CpG ASO 120704 and ODN 2395 had analysis showed a strong positive correlation (correlation
A B
C D
E F
Figure 2. Uptake and localization of antisense oligonucleotides in human washed platelets. Washed platelets were incubated with a therapeutically relevant con-
centration of the antisense oligonucleotides (ASO) (5 mM), then fixed and stained with an anti-ASO antibody and labeled with protein A-gold. Representative electron
microscopy images at 15000x of ASO localization, using anti-ASO immunogold labeling (shown as black dots) of human washed platelets treated for 30 min with (A)
vehicle (HEPES, 10 mM), (B) 2’MOE ASO 487660, (C) 2’MOE ASO 104838, (D) 2’MOE ASO 501861, (E) CpG ASO 120704, or (F) CpG ASO ODN 2395 (all at 5 mM).
Black arrows indicate ASO localizing on the platelet membrane and white arrows indicate internalized ASO.

C D
E F
Figure 3. Effect of antisense oligonucleotides on platelet activation marker P-selectin and stromal cell derived factor 1a release in human platelet-rich plasma
and whole blood and correlation with individual platelet glycoprotein VI receptor levels. Platelet activation was identified through increases in P-selectin expression
on the platelet surface after 30 min treatment with vehicle (HEPES, 10 mM) or thrombin receptor activating peptide (TRAP, 25 mM, to activate the platelets) or 5 mM
of the antisense oligonucleotides (ASO): 104838 (2’MOE ASO), 501861 (2’MOE ASO), 120704 (CpG ASO) and ODN 2395 (CpG ASO), assessed by flow cytometry in
(A) platelet-rich plasma (PRP) (from 9 human donors, the 2’MOE ASO 487660 was included in 3 of the experiments) and (B) whole blood (WB) (from 8 human donors)
pretreated or not with the spleen tyrosine kinase (SYK) inhibitor PRT-060318 (10 mM) before addition of vehicle, TRAP (25 µM) or ASO (5 µM) that increased P-selectin
in PRP (i.e., 104838, 501861, 120704 and ODN 2395). *P<0.05 compared to vehicle by one-way analysis of variance (ANOVA), with the Dunnett post-test. #P<0.05
paired Student t-test for the effect of the SYK inhibitor. (C, D) Individual donor platelet glycoprotein (GP)VI receptor levels (median fluorescence intensity, MFI) were
correlated to the same individual platelet P-selectin levels after treatment with: (C) 2’MOE ASO 104838 or 501861, and (D) CpG ASO 120704 or ODN 2395, (7
human donors). P<0.05 by Pearson correlation analysis. (E, F) Stromal cell derived factor 1α (SDF1a), released from platelets upon activation, was measured by
Mesoscale U-plex multiplex assay in (E) PRP and (F) WB blood treated for 30 min with vehicle (HEPES), TRAP (25 mM) or 1, 5 or 10 mM of the ASO in blood from four
human donors. *P<0.05 compared to vehicle by one-way ANOVA, with Dunnett post-test.

coefficients: 104838 r=0.97, 501861 r=0.96, 120704 r=0.85, vated platelets, we measured secreted plasma levels of
ODN 2395 r=0.86) between donors’ platelet GPVI levels stromal cell derived factor 1a (SDF1a) in ASO-treated
and their platelet activation in response ASO treatment, platelet-rich plasma and whole blood. TRAP stimulated
with the highest ASO responders also having the highest SDF1a release in both platelet-rich plasma and whole
platelet GPVI levels (Figure 3 C, D). blood (Figure 3 E, F). Consistent with the pattern of the P-
selectin effects, the CpG ASO 120704 and ODN 2395 trig-
2’MOE ASO (104838 and 501861) and CpG ASO gered robust release of SDF1a in platelet-rich plasma and
trigger SDF1a release from whole blood whole blood in a concentration-dependent manner: thus
As an additional marker of a-granule release from acti- the effect could be seen with doses of 5 and 10 mM, but
A B
C D
E F
Figure 4. Platelet aggregation in human platelet-rich plasma and whole blood treated with antisense oligonucleotides. (A, B) Ninety-six-well platelet aggregometry
was used to generate full concentration-response curves to the platelet agonist collagen after incubating platelet-rich plasma (PRP) with vehicle or (A) 5 mM of the
2’MOE antisense oligonucleotides (ASO) 487660, 104838 and 501861 or (B) the CpG ASO 120704 or ODN 2395. *P<0.05 by two-way analysis of variance (ANOVA)
with Bonferroni post-test, compared to vehicle (HEPES, 10 mM), n=6-8 human donors. (C) Platelet aggregation was also assessed using traditional light transmission
aggregometry (LTA) by incubating PRP (n=8 human donors) with the ASO for 30 min at 1200 rpm stirring speed without stimulation (to detect spontaneous aggre-
gation) or after stimulation with the platelet agonist thrombin receptor activating peptide (TRAP, 25 mM). (D) Impedance aggregometry was used to analyze platelet
aggregation by incubating whole blood (WB) (n=7 human donors) with the ASO (at 1 and 5 mM) for 30 min at 1200 rpm stirring speed, without stimulation, or TRAP
(25 mM) stimulated aggregation. (E) Platelet-platelet aggregates in WB (n=7 human donors) treated with vehicle (HEPES, 10 mM), collagen (20 mg/mL), TRAP
(25 mM) or ASO (5 mM) were analyzed using flow cytometry. (F) Platelet-leukocyte aggregates (platelet marker CD41/61+ leukocyte marker CD14+) were analyzed in
WB (n=6 human donors) treated with vehicle (HEPES, 10 mM), collagen (20 mg/mL), TRAP (25 mM), or ASO (5 mM), using flow cytometry. *P<0.05 compared to vehi-
cle by one-way ANOVA, Dunnett post-test (C, D, E, F).

A B
C D
E F
Figure 5. Immunostimulatory effects after 30 minutes and 6 hours in antisense oligonucleotide-treated whole blood. Whole blood (WB) from four human donors
was incubated with vehicle (HEPES, 10 mM), thrombin receptor activating peptide (TRAP, 25 mM), lipopolysaccharide (LPS 0.01 mg/mL) or 1, 5 or 10 mM of the anti-
sense oligonucleotides (ASO). Interleukin 8 (IL-8) and monocyte chemotactic protein-1 (MCP-1) release was measured in plasma (pg/mL), using an MSD U-plex assay,
after incubation for 30 min and 6 h at 37°C in 5% CO2. (A) IL-8 after 30 min. (B) MCP-1 after 30 min. (C) IL-8 after 6 h. (D) MCP-1 after 6 h. *P<0.05 compared to
vehicle by one-way analysis of variance, Dunnett post-test. Concentration-response curves for (E) IL-8 release after 6 h and (F) MCP-1 release after 6 h.
not at a dose of 1 mM (Figure 3E, F). The 2’MOE ASO donors with high platelet GPVI levels (Online
104838 and 501861 triggered a small release at 5 mM in Supplementary Figure S2). However, no significant sponta-
whole blood only (Figure 3F). The 2’MOE ASO 487660 neous aggregation was detected by light transmission
did not trigger SDF1a release at any concentration tested, aggregometry in any of the ASO-treated platelet-rich plas-
in either platelet-rich plasma or whole blood (Figure 3E, F). ma samples and stimulating the 5 mM ASO-treated sam-
ples with TRAP resulted in a similar light transmission
2’MOE ASO 104838 increases platelet reactivity aggregometry-measured aggregation response to that fol-
to collagen lowing stimulation with the vehicle (Figure 4C).
To investigate whether the platelet activation docu- Differences in the mechanical shear stress environment in
mented in Figure 3 would translate into an effect on 96-well aggregometry versus light transmission aggregom-
platelet aggregation, we performed 96-well aggregometry etry25 may explain the discrepancy in results between
and traditional light transmission aggregometry under these assays.
both unstimulated and stimulated conditions (Figure 4A-
D). The 2’MOE ASO 104838 triggered a small potentiat- 2’MOE ASO (104838 and 501861) and CpG ASO poten-
ing effect on collagen-induced platelet aggregation, but tiate platelet reactivity to TRAP in whole
this effect was not seen after treatment with the other two blood impedance aggregometry
2’MOE ASO (Figure 4A). The CpG ASO 120704 and ODN When whole blood was incubated with 5 mM of the
2395 also increased platelet reactivity to collagen (Figure CpG ASO ODN 2395 in an impedance aggregometer,
4B), including ~40% spontaneous aggregation in two spontaneous aggregation was detected over the course of

Figure 6. SYK inhibition of the formation of platelet-leukocyte aggregates in whole blood treated with antisense oligonucleotides. Whole blood (WB) from four
human donors was pretreated or not with a spleen tyrosine kinase (SYK) inhibitor (PRT-060318, 10 mM) followed by vehicle (10 mM HEPES), thrombin receptor acti-
vating peptide (TRAP, 25 mM), collagen (20 mg/mL), or the 2’MOE non CpG antisense oligonucleotides (ASO): 104838 or 501861, or the CpG ASO: 120704 or ODN
2395 (all at 5 mM) and analyzed by flow cytometry for: (A) platelet-leukocyte aggregates (PLA, platelet marker CD41/61+, leukocyte marker CD14+). *P<0.05 com-
pared to vehicle by one-way analysis of variance (ANOVA), Dunnett post-test, #P<0.05 paired Student t-test for the effect of the SYK inhibitor. (B) Representative con-
focal images of platelet-leukocyte aggregates in fixed and fluorescently labeled WB. Red represents CD45 (leukocyte marker), green represents CD41/61 (platelet
marker). The white boxes indicate the location of the zoomed-in part of image shown to the right. Scale bar = 10 mm.
the 30 min incubation (12±6 Ω vs. 0.4±0.3 Ω with vehicle) was perhaps not solely driven by homogenous platelet-
(Figure 4D). When platelet-platelet interactions were stud- platelet aggregates but could also contain heterogeneous
ied in ASO-treated (but otherwise unstimulated) whole platelet-leukocyte aggregates.26 We, therefore, analyzed
blood samples using flow cytometry, only the CpG ASO ASO-treated (but otherwise unstimulated) whole blood
120704 and ODN 2395 had significantly more platelet- for the formation of platelet-leukocyte aggregates by flow
platelet aggregates (13±6 and 14±6%, respectively, vs. cytometry (Figure 4F). TRAP and collagen triggered sub-
0.2±0.04% with vehicle) (Figure 4E), consistent with the stantial formation of platelet-leukocyte aggregates, where-
impedance aggregometry results. After stimulation with as the 2’MOE ASO 487660 evoked a similar response to
TRAP, whole blood aggregation was potentiated to a sim- that produced by the vehicle (Figure 4F). However, whole
ilar level in both the 2’MOE ASO (104838 and 501861) blood treated with 2’MOE ASO 104838 and 501861 and
and the CpG (120704 and ODN 2395)-treated samples at the CpG 120704 and ODN 2395 had more platelet-leuko-
both 1 and 5 mM (Figure 4D). cyte aggregates (33±8, 37±9, 69±4, and 46±4%, respec-
tively) than vehicle-treated whole blood (12±1%) (Figure
2’MOE ASO (104838 and 501861) and CpG ASO 4F).
trigger formation of platelet-leukocyte aggregates
in unstimulated whole blood Brief ASO treatment does not induce neutrophil or
The impedance aggregometry results led us to hypothe- monocyte activation
size that the enhancement of whole blood aggregation To further explore the interaction of ASO with mono-

A B
C D
E F
Figure 7. The effect of antisense oligonucleotides on platelet-neutrophil, platelet-monocyte and glycoprotein VI interactions. Whole blood (WB) was incubated with
vehicle (10 mM HEPES), thrombin receptor activating peptide (TRAP, 25 mM), or the 2’MOE antisense oligonucleotides (ASO): 104838 or 501861, or the CpG ASO:
120704 or ODN 2395 (all at 5 mM) and analyzed by flow cytometry for (A) platelet-neutrophil aggregates (platelet marker CD41/61+, neutrophil marker CD66b+), (B)
platelet-monocyte aggregates (platelet marker CD41/61+, monocyte marker CD14+). (C) Surface P-selectin-positive (CD62p+) platelet-neutrophil aggregates
(CD41/61+, CD66b+, CD62p+) (median fluorescence intensity, MFI). (D) Surface P-selectin-positive platelet-monocyte aggregates (CD41/61+, CD14+, CD62p+) (MFI)
in WB from five to nine human donors. (E) Individual donor platelet glycoprotein (GP)VI levels were correlated to platelet-neutrophil aggregate formation after treat-
ment with the ASO in six human donors. *P<0.05 by Pearson correlation analysis. (F) A fluorescence polarization assay was used to measure binding affinity of
Alexa647-labeled ASO to human GPVI. Bmax is the total density of receptors in a sample and KD is the equilibrium dissociation constant. The smaller the KD, the
greater the binding affinity of the ASO to human GPVI.
cytes and neutrophils, we assessed whether ASO Supplementary Figure S4A,B). In support of these data, the
increased surface expression of CD11b, a broad immune proinflammatory chemokines interleukin-8 (IL-8) and
cell activation marker. Lipopolysaccharide was included monocyte chemoattractant protein-1 (MCP-1) were not
as a positive control to increase expression of CD11b on released from whole blood incubated with ASO (1, 5 and
the surface of neutrophils (Online Supplementary Figure 10 mM) for 30 min (Figure 5A, B).
S4A) or monocytes (Online Supplementary Figure S4B).
SYK is also involved in leukocyte intracellular signal- Treatment with CpG ASO (but not 2’MOE ASO) leads
ing,27,28 and there was an inhibitory effect on CD11b sur- to IL-8 and MCP-1 release
face expression in SYK-treated samples exposed to vehi- CpG motifs have been shown to be immunostimulato-
cle and lipopolysaccharide (Online Supplementary Figure S4 ry15 and proinflammatory effects of the CpG ASO 120704
A, B). None of the ASO tested had any effect on CD11b and ODN 2395 were apparent after 6 h of incubation of
expression on either neutrophils or monocytes (after 30 whole blood, with 5 and 10 mM (but not 1 mM) resulting
min incubation); hence the ASO did not appear to acti- in robust IL-8 and MCP-1 release (Figure 5C, D). None of
vate these cells directly within this timeframe (Online the 2’MOE ASO evoked a proinflammatory effect at any

of the concentrations (Figure 5C-D). The concentration- Discussion

dependent effects of 120704 and ODN 2395 are shown in
Figure 5E, F. The main findings from the current study are: (i) none of
the ASO sequences investigated had an inhibitory effect
104838, 501861, 120704 and ODN 2395-induced on proplatelet production by either human- or mouse-
formation of platelet-leukocyte aggregrates in derived megakaryocytes; (ii) all ASO demonstrated uptake
unstimulated whole blood is blocked by SYK into human platelets (2’MOE 487660 less than the others);
pretreatment (iii) a subset of 2’MOE ASO (104838 and 501861) and
When whole blood was pretreated with a SYK inhibitor, CpG ASO (120704 and ODN 2395) activated human
collagen-induced platelet-leukocyte aggregates were platelets, triggering P-selectin and SDF1a release from
markedly reduced, while TRAP-induced platelet-leuko- platelet a-granules and platelet-leukocyte aggregate for-
cyte aggregates were maintained, confirming the selectiv- mation; (iv) the ASO-induced platelet activation and
ity of the SYK inhibitor (Figure 6A). SYK pre-treatment platelet-leukocyte aggregate formation were fully reversed
completely reversed the ASO-induced formation of by pre-treatment with a SYK inhibitor; (v) ASO-induced
platelet-leukocyte aggregates (Figure 5A). Confocal imag- platelet effects appeared to be sequence-dependent, rather
ing of the aggregates using a leukocyte marker (CD45) and than 2’MOE-dependent, since the 2’MOE ASO 487660
a platelet marker (CD41/61), confirmed the presence of did not affect platelet function; (vi) all the ASO were
platelet-leukocyte aggregates (Figure 6B) but showed shown to bind to human GPVI (CpG ASO had the
greater platelet-platelet aggregation in the platelet-leuko- strongest affinity); (vii) only the CpG ASO had a concen-
cyte aggregates in whole blood treated with the CpG ASO tration-dependent proinflammatory effect, triggering IL-8
(which was more similar to TRAP-treated samples) com- and MCP-1 release in whole blood after incubation for 6
pared to the 2’MOE ASO, which mainly involved single h; (viii) the responsiveness to the ASO-induced platelet-
platelets bound to the leukocytes (Figure 6B). activating and platelet-leukocyte aggregate-forming
effects varied between donors and showed a strong posi-
ASO increase levels of P-selectin-enriched tive correlation to individual platelet GPVI surface expres-
platelet-neutrophil and platelet-monocyte aggregates, sion.
through a SYK-dependent mechanism, correlating The novelty of this study concerns the effect of 2’MOE
to GPVI levels ASO on human platelets. Our focus was the 2’MOE ASO
Further investigation into the types of immune cells 104838, which has been shown to be representative of a
that were driving the ASO-induced formation of platelet- subset of ASO sequences that produce phenotype 1 reduc-
leukocyte aggregates revealed an increase in both tions in platelet count in monkeys and humans.8,10 We
platelet-neutrophil aggregates and platelet-monocyte identified that the 2’MOE ASO 104838 binds to GPVI
aggregates (Figure 7A, B). The CpG ASO 120704 and receptors on human platelets, increasing platelet surface
ODN 2395 once again produced a more robust response P-selectin and prompting the formation of platelet-neu-
than the 2’MOE ASO 104838 and 501861 (65±7 and 47±7 trophil aggregates and platelet-monocyte aggregates, most
vs. 24±10 and 27±11 vs. 6±1 for vehicle). The platelet-neu- likely through an interaction between platelet P-selectin
trophil and platelet-monocyte aggregates that formed in and leukocyte P-selectin glycoprotein ligand 1 (PSGL-1).29
the ASO-treated samples were enriched in P-selectin This appears to be a platelet-driven interaction (at least
(Figure 7C, D) and the greatest increase was observed in when tested in vitro), as there were no signs of leukocyte
the samples treated with 120704 and ODN 2395. activation (CD11b or IL-8/MCP-1 release) even after 10
However, platelet-neutrophil aggregates and platelet- µM of the 2’MOE ASO 104838.
monocyte aggregates were not enriched in CD11b Mechanistically, we suggest that 2’MOE ASO 104838
(Online Supplementary Figure S4C, D), implying that the lowers platelet count, producting a phenotype 1, by acti-
initial formation of these aggregates was driven by activating platelets, an effect which triggers platelet-leukocyte
vated platelets (not activated leukocytes). Pearson correla- aggregates and subsequent clearance of platelets by leuko-
tion analysis between platelet GPVI levels and platelet cytes. Maugeri et al. showed that platelet-neutrophil inter-
neutrophil aggregates showed a strong positive correla- actions might contribute to platelet clearance through
tion for all the ASO tested; higher platelet GPVI expres- active phagocytosis of P-selectin-positive platelets by neu-
sion was associated with a stronger platelet-neutrophil trophils.30 The internalization of platelets by neutrophils
aggregate response (Figure 7E). has been observed in patients with viral infection (associ-
ated with reductions in platelet count).31 Increased levels
CpG ASO bind to GPVI with a higher affinity compared of P-selectin can also enhance sequestration of platelets on
to 2’MOE ASO vascular endothelial surfaces.32 When monkeys were
Fluorescence polarization experiments revealed that all treated with 2’MOE ASO 104838, there was increased
ASO bound to human GPVI, with equilibrium dissocia- platelet sequestration in the densely vascularized liver and
tion constant (KD) values in the low micromolar range spleen.10 However, in vitro incubation of platelet-rich plas-
(Figure 7F): the smaller the KD, the greater the binding ma from treatment-naïve monkeys with 2’MOE ASO
affinity. All the ASO bound to GPVI with stronger affinity 104838 did not lead to platelet activation10 (as we
than to control human serum albumin (Online observed in human platelet-rich plasma and whole blood).
Supplementary Figure S5). The 2’MOE 487660 showed This may be because the sample size used (n=3) did not
stronger affinity (KD 12.9 mM) to human serum albumin capture the variability seen in ASO responsiveness.
than the other two 2’MOE ASO (104838 KD 24 mM and Monkeys treated with 2’MOE ASO 104838 also had high-
501861 KD 26 mM). The CpG ASO displayed higher affin- er plasma levels of IL-8 and MCP-1,10 which we did not
ity (lower KD) for GPVI compared to the 2’MOE ASO see in our in vitro assay with 104838 in human blood (using
(Figure 7F). doses up to 10 mM). This may reflect in vitro rather than in

vivo conditions or be due to species differences. Further There is a degree of subject variability in platelet count
investigation is warranted to see if in vivo treatment, or reductions following ASO treatment in monkeys and
indeed clinical use of certain 2’MOE ASO sequences leads humans.8,10,16 We also noticed a high degree of variability
to increased platelet activation and formation of platelet- in responsiveness to ASO in our in vitro studies of blood
leukocyte aggregates and to what degree this is paired from healthy human donors. Although the sample was
with clinical reductions in platelet count. small (n=7), there was a strong positive correlation
Our data showing that pre-treatment with a SYK between an individual’s platelet GPVI levels and their
inhibitor was able to fully reverse ASO-induced platelet platelet responsiveness to both 2’MOE (104838 and
activation and platelet-leukocyte aggregate formation, 501861) and CpG ASO-induced activation. Platelet GPVI
speculatively highlights the potential of using a clinically levels have been shown to vary in healthy individuals35
available SYK inhibitor, for instance fostamatinib,33 to and to be increased in different disease states such as obe-
treat ASO-induced platelet side effects. This would need sity.36 Platelet GPVI levels could potentially be useful as a
to be investigated further. screening tool (used before commencement of treatment)
Flierl et al. were the first to identify that ODN 2395 to identify at-risk patients who may be more susceptible
binds to and activates GPVI on platelets.14 In the current to platelet side effects of some sequence-specific GPVI-
study, we confirmed these findings and added the novel activating ASO.
finding that 2’MOE ASO also bind to GPVI. Overall, the We have shown that 2’MOE ASO that have been asso-
2’MOE ASO had lower affinity to GPVI compared to the ciated with phenotype 1 may not simply be reducing
CpG ASO, which is consistent with their weaker platelet platelet counts, but can also have direct effects on
activating effects. In general it has been shown that the platelets, triggering interactions between platelets and
stronger affinity with which PS ASO bind, the larger the immune cells.
conformational change of the protein.34 Hence, CpG In summary, we have defined new mechanisms by
ASO, with their greater affinity than 2’MOE ASO for which 2’MOE ASO-based drugs affect human platelets,
GPVI, may induce a more substantial conformational which may yield new strategies to avoid ASO sequences
change in the GPVI protein, affecting for instance GPVI with unfavorable platelet effects.
clustering/dimerization,14 explaining the greater potency
of CpG ASO to activate platelets. Disclosures
Interestingly, the 2’MOE ASO 487660, which has not JEI has financial interest in and is a founder of Platelet
been shown to lower platelet counts in monkeys, did not BioGenesis, a company that aims to produce donor-independent
affect any of the platelet function tests we performed in human platelets from human-induced pluripotent stem cells at
this study. Hence, the platelet effects observed with scale. The interests of JEI were reviewed and are managed by the
104838 and 501861 appear to be sequence-dependent Brigham and Women’s Hospital and Partners HealthCare. The
rather than due to their 2’MOE-backbone chemistry. remaining authors declare that they have no conflicts of interest.
487660 did however still bind to human GPVI with a sim-
ilar affinity as that of the other 2’MOE ASO, 104838 and Contributions
501861. This finding highlights the importance of func- MHLS supervised the study, performed research, analyzed
tional in vitro platelet studies beyond ASO-GPVI binding data and wrote the manuscript; PV, IT, HR, GC, AW, HG, JM,
assays in cell-free environments. The discrepancy by RO, BE, and JT performed research, analyzed data and edited
which 2’MOE ASO 487660 can bind GPVI but does not the manuscript; JWS, SPH and PN analyzed data and edited
activate platelets may be explained by its greater affinity the manuscript; JEI supervised the study, analyzed data and edit-
to bind human serum albumin, compared to the other ed the manuscript.
2’MOE ASO investigated, which may keep 487660 more
bound to plasma proteins. 2’MOE ASO 487660 may also Funding
interact differently with the GPVI receptor as there This work was funded by a corporate sponsored research
appeared to be less surface and internalized 2’MOE ASO agreement between Ionis Pharmaceuticals Inc. and Brigham and
487660 staining in platelets in the electron-microscopy Women’s Hospital and Partners HealthCare in accordance with
images. their conflict-of-interest policies.
References 5. Mathew V, Wang AK. Inotersen: new 2017;27(3):121-129.

promise for the treatment of hereditary 9. Chi X, Gatti P, Papoian T. Safety of anti-
1. Bennett CF, Baker BF, Pham N, Swayze E, transthyretin amyloidosis. Drug Des Devel sense oligonucleotide and siRNA-based
Geary RS. Pharmacology of antisense Ther. 2019;13:1515-1525. therapeutics. Drug Discov Today. 2017;
drugs. Annu Rev Pharmacol Toxicol. 6. Geary RS, Baker BF, Crooke ST. Clinical 22(5):823-833.
2017;57:81-105. and preclinical pharmacokinetics and phar- 10. Narayanan P, Shen L, Curtis BR, et al.
2. Bennett CF, Swayze EE. RNA targeting macodynamics of mipomersen Investigation into the mechanism(s) that
therapeutics: molecular mechanisms of (Kynamro®): a second-generation anti- leads to platelet decreases in Cynomolgus
antisense oligonucleotides as a therapeutic sense oligonucleotide inhibitor of monkeys during administration of ISIS
platform. Annu Rev Pharmacol Toxicol. apolipoprotein B. Clin Pharmacokinet. 104838, a 2'-MOE-modified antisense
2010;50:259-293. 2015;54(2):133-146. oligonucleotide. Toxicol Sci. 2018;164(2):
3. Finkel RS, Mercuri E, Darras BT, et al. 7. Schoch KM, Miller TM. Antisense oligonu- 613-626.
Nusinersen versus sham control in infan- cleotides: translation from mouse models 11. Henry SP, Narayanan P, Shen L, Bhanot S,
tile-onset spinal muscular atrophy. N Engl J to human neurodegenerative diseases. Younis HS, Burel SA. Assessment of the
Med. 2017;377(18):1723-1732. Neuron. 2017;94(6):1056-1070. effects of 2'-methoxyethyl antisense
4. Mercuri E, Darras BT, Chiriboga CA, et al. 8. Crooke ST, Baker BF, Witztum JL, et al. The oligonucleotides on platelet count in
Nusinersen versus sham control in later- effects of 2'-O-methoxyethyl containing Cynomolgus nonhuman primates. Nucleic
onset spinal muscular atrophy. N Engl J antisense oligonucleotides on platelets in Acid Ther. 2017;27(4):197-208.
Med. 2018;378(7):625-635. human clinical trials. Nucleic Acid Ther. 12. Narayanan P, Curtis BR, Shen L, et al.

Underlying immune disorder may predis- Metab Toxicol. 2013;9(2):169-182. LDL. Br J Pharmacol. 2012;167(5):990-999.
pose some transthyretin amyloidosis sub- 21. Thon JN, Devine MT, Jurak Begonja A, 29. Kappelmayer J, Nagy B, Jr. The Interaction
jects to inotersen-mediated thrombocy- Tibbitts J, Italiano JE Jr. High-content live- of selectins and PSGL-1 as a key compo-
topenia. Nucleic Acid Ther. 2020;30(2):94- cell imaging assay used to establish mecha- nent in thrombus formation and cancer
103. nism of trastuzumab emtansine (T-DM1)-- progression. Biomed Res Int. 2017;2017:
13. Aster RH, Bougie DW. Drug-induced mediated inhibition of platelet production. 6138145.
immune thrombocytopenia. N Engl J Med. Blood. 2012;120(10):1975-1984. 30. Maugeri N, Rovere-Querini P, Evangelista
2007;357(6):580-587. 22. Machlus KR, Wu SK, Stumpo DJ, et al. V, et al. Neutrophils phagocytose activated
14. Flierl U, Nero TL, Lim B, et al. Synthesis and dephosphorylation of MAR- platelets in vivo: a phosphatidylserine, P-
Phosphorothioate backbone modifications CKS in the late stages of megakaryocyte selectin, and b2 integrin-dependent cell
of nucleotide-based drugs are potent maturation drive proplatelet formation. clearance program. Blood. 2009;113(21):
platelet activators. J Exp Med. 2015; Blood. 2016;127(11):1468-1480. 5254-5265.
212(2):129-137. 23. Gaus HJ, Gupta R, Chappell AE, 31. Koupenova M, Vitseva O, MacKay CR, et
15. Sewing S, Roth AB, Winter M, et al. Ostergaard ME, Swayze EE, Seth PP. al. Platelet-TLR7 mediates host survival and
Assessing single-stranded oligonucleotide Characterization of the interactions of platelet count during viral infection in the
drug-induced effects in vitro reveals key chemically-modified therapeutic nucleic absence of platelet-dependent thrombosis.
risk factors for thrombocytopenia. PLoS acids with plasma proteins using a fluores- Blood. 2014;124(5):791-802.
One. 2017;12(11):e0187574. cence polarization assay. Nucleic Acids 32. Ed Rainger G, Chimen M, Harrison MJ, et
16. Sewell KL, Geary RS, Baker BF, et al. Phase Res. 2019;47(3):1110-1122. al. The role of platelets in the recruitment
I trial of ISIS 104838, a 2'-methoxyethyl 24. Rayes J, Watson SP, Nieswandt B. of leukocytes during vascular disease.
modified antisense oligonucleotide target- Functional significance of the platelet Platelets. 2015;26(6):507-520.
ing tumor necrosis factor-alpha. J immune receptors GPVI and CLEC-2. J Clin 33. Connell NT, Berliner N. Fostamatinib for
Pharmacol Exp Ther. 2002;303(3):1334- Invest. 2019;129(1):12-23. the treatment of chronic immune thrombo-
1343. 25. Chan MV, Leadbeater PD, Watson SP, cytopenia. Blood. 2019;133(19):2027-2030.
17. Ferrer-Marin F, Stanworth S, Josephson C, Warner TD. Not all light transmission 34. Crooke ST, Liang XH, Crooke RM, Baker
Sola-Visner M. Distinct differences in aggregation assays are created equal: quali- BF, Geary RS. Antisense drug discovery and
platelet production and function between tative differences between light transmis- development technology considered in a
neonates and adults: implications for sion and 96-well plate aggregometry. pharmacological context. Biochem
platelet transfusion practice. Transfusion. Platelets. 2018;29(7):686-689. Pharmacol. 2020;114196.
2013;53(11):2814-2821. 26. Russell-Smith NC, Flower RJ, Cardinal DC. 35. Furihata K, Clemetson KJ, Deguchi H,
18. Vijey P, Posorske B, Machlus KR. In vitro Measuring platelet and leucocyte aggrega- Kunicki TJ. Variation in human platelet gly-
culture of murine megakaryocytes from tion/adhesion responses in very small vol- coprotein VI content modulates glycopro-
fetal liver-derived hematopoietic stem cells. umes of whole blood. J Pharmacol tein VI-specific prothrombinase activity.
Platelets. 2018;29(6):583-588. Methods. 1981;6(4):315-333. Arterioscler Thromb Vasc Biol. 2001;
19. French SL, Vijey P, Karhohs KW, et al. High 27. Berton G, Mocsai A, Lowell CA. Src and 21(11):1857-1863.
Content, Label-free analysis of proplatelet Syk kinases: key regulators of phagocytic 36. Barrachina MN, Sueiro AM, Izquierdo I, et
production from megakaryocytes. J cell activation. Trends Immunol. 2005; al. GPVI surface expression and signalling
Thromb Haemost. 2020;18(10):2701-2711. 26(4):208-214. pathway activation are increased in
20. Yu RZ, Grundy JS, Geary RS. Clinical phar- 28. Miller YI, Choi SH, Wiesner P, Bae YS. The platelets from obese patients: elucidating
macokinetics of second generation anti- SYK side of TLR4: signalling mechanisms potential anti-atherothrombotic targets in
sense oligonucleotides. Expert Opin Drug in response to LPS and minimally oxidized obesity. Atherosclerosis. 2019;281:62-70.

ARTICLE Red Cell Biology & its Disorders
Ferrata Storti Foundation Sulfated non-anticoagulant heparin derivative

modifies intracellular hemoglobin, inhibits cell
sickling in vitro, and prolongs survival of
sickle cell mice under hypoxia
Osheiza Abdulmalik,1* Noureldien H. E. Darwish,2,3*
Vandhana Muralidharan-Chari,2° Maii Abu Taleb2 and Shaker A. Mousa2,4
1
Division of Hematology, the Children’s Hospital of Philadelphia, Philadelphia, PA, USA;
Haematologica 2022 2
The Pharmaceutical Research Institute, Albany College of Pharmacy and Health
Volume 107(2):532-540 Sciences, Rensselaer, NY, USA; 3Clinical Pathology (Hematology Section), Faculty of
Medicine, Mansoura University, Mansoura, Egypt and 4Vascular Vison Pharmaceuticals
Co., Rensselaer, NY, USA
*
OA and NHED contributed equally as co-first authors.
°
Current address: College of Nanoscale Science and Engineering, SUNY Polytechnic Institute,
Albany, NY, USA.
ABSTRACT
S
ickle cell disease (SCD) is an autosomal recessive genetic disease
caused by a single point mutation, resulting in abnormal sickle
hemoglobin (HbS). During hypoxia or dehydration, HbS polymer-
izes to form insoluble aggregates and induces sickling of red blood cells,
which increases the adhesiveness of the cells, thereby altering the rheo-
logical properties of the blood, and triggers inflammatory responses, lead-
ing to hemolysis and vaso-occlusive crises. Unfractionated heparin and
low-molecular weight heparins have been suggested as treatments to
relieve coagulation complications in SCD. However, they are associated
with bleeding complications after repeated dosing. An alternative sulfat-
ed non-anticoagulant heparin derivative (S-NACH) was previously
reported to have no to low systemic anticoagulant activity and no bleed-
ing side effects, and it interfered with P-selectin-dependent binding of
sickle cells to endothelial cells, with concomitant decrease in the levels of
adhesion biomarkers in SCD mice. S-NACH has been further engineered
Correspondence: and structurally enhanced to bind with and modify HbS to inhibit sick-
SHAKER A. MOUSA ling directly, thus employing a multimodal approach. Here, we show that
shaker.mousa@acphs.edu S-NACH can: (i) directly engage in Schiff-base reactions with HbS to
decrease red blood cell sickling under both normoxia and hypoxia in vitro,
(ii) prolong the survival of SCD mice under hypoxia, and (iii) regulate the
Received: September 15, 2020.
altered steady state levels of pro- and anti-inflammatory cytokines. Thus,
Accepted: January 25, 2021. our proof-of-concept, in vitro and in vivo preclinical studies demonstrate
Pre-published: February 11, 2021. that the multimodal S-NACH is a highly promising candidate for devel-
opment into an improved and optimized alternative to low-molecular
weight heparins for the treatment of patients with SCD.
©2022 Ferrata Storti Foundation Introduction

All rights are reserved to the Ferrata Storti Foundation. Use of Sickle cell disease (SCD) is a hemoglobinopathy resulting from a mutation replac-
conditions:
ing the glutamic acid amino acid with the less polar valine amino acid at the sixth
https://creativecommons.org/licenses/by-nc/4.0/legalcode. position of the b chain, converting normal adult hemoglobin (HbA) to sickle hemo-
Copies of published material are allowed for personal or inter- globin (HbS).1 Deoxygenated HbS polymerizes into long, rigid fibers, causing sick-
nal use. Sharing published material for non-commercial pur- ling of red blood cells (RBC).2 These characteristic sickled RBC impair blood flow
through the microvasculature, leading to hemolysis, episodes of vaso-occlusion,
sect. 3. Reproducing and sharing published material for com- and multi-organ damage.3-7 The loss of membrane phospholipid asymmetry on
mercial purposes is not allowed without permission in writing sickled RBC exposes phosphatidylserines8 that increase the adhesion of sickled
from the publisher. RBC to neutrophils, monocytes, platelets, and endothelial cells to activate coagula-
tion and inflammatory pathways,9-12 culminating in a ‘hypercoagulable’ state.13
Currently, four drugs have been approved by the Food and Drug Administration

Sulfated non-anticoagulant heparin in SCD treatment
for the treatment of SCD in the USA. L-glutamine with informed consent. None of the subjects had been transfused
(Endari), approved in 2017, increases the amount of the within 4 months prior to their blood samples being used, and four
reduced form of NADH in erythrocytes, which allows of the five donors were on hydroxyurea therapy.
sickle RBC to maintain homeostasis more appropriately
during oxidative stress, ultimately resulting in fewer Anti-sickling, oxygen equilibrium and hemoglobin
painful vaso-occlusive crises and adverse events.14 modification studies using human sickle blood
Crizanlizumab (Adakveo)15 and voxelotor16 (Oxbryta, The morphology of hypoxic sickled RBC was evaluated using
GBT440) were approved in 2019. Crizanlizumab is a a previously reported method.22,23 Blood samples from individ-
monoclonal antibody that targets P-selectin to prevent ual donors with SCD (n=5) were diluted using HEMOX buffer
pathological endothelial adhesion of sickle erythrocytes supplemented with glucose (10 mM) and bovine serum albumin
and leukocytes, leading to a reduction in the frequency of (0.2%) to adjust the hematocrit of the suspensions to about
painful vaso-occlusive crises.17,18 The anti-sickling agent 20%. We used standardized hematocrit for anti-sickling assays
voxelotor is the first of a new class of aromatic aldehydes to normalize the ratio of RBC to drug for assay consistency and
that target HbS polymerization by increasing Hb O2 affin- reproducibility. The suspensions were pre-incubated under air in
ity.19-21 Finally, hydroxyurea, which works by inducing the the absence or presence of three concentrations (0.5, 1, and 2
expression of fetal Hb (HbF), is the most proven therapeu- mM) of S-NACH at 37°C for 1 h. The suspensions were subse-
tic approach for SCD,22,23 as evidenced by its sustained quently incubated under a 2.5% O2/97.5% N2 gas mixture at
clinical use for over two decades. However, a reported 37°C for 2 h. Aliquots (5–20 mL) of each sample were collected
lack of response to hydroxyurea in up to 30% of patients, without exposure to air into 2% glutaraldehyde solution for
and supposed poor compliance, tend to limit its use.22 The immediate fixation. Fixed cell suspensions were thereafter intro-
reported limitations of hydroxyurea led to investigation of duced into glass microslides (Fiber Optic Center, New Bedford,
other modes of therapy, including the three more recently MA, USA)34 and subjected to microscopic morphological analy-
approved drugs. However, their true benefits will only sis of bright field images (at 40x magnification) of single layer
manifest over time. Additionally, the inherently complex cells on an Olympus BX40 microscope fitted with an Infinity 2
nature of SCD dictates the urgent need for a multimodal camera (Olympus, Waltham, MA, USA), with the coupled
form of therapy. Image Capture software. The percentage of sickled cells for each
Antiplatelet molecules, anticoagulants, and heparin condition was determined using blood with a computer-assisted
have been investigated to mitigate vaso-occlusive crises.24 image analysis system, as described previously.33,35 Untreated
Although heparin is beneficial, the associated risks of samples, as well as samples treated with GBT440/voxelotor,
internal bleeding preclude its utility as a drug25 and the were used as controls. The residual samples were washed in
need for alternatives remains critical. We developed a sul- phosphate-buffered saline (PBS) and hemolysed in hypotonic
fated non-anticoagulant heparin (S-NACH) with no to low lysis buffer for subsequent analyses.
systemic anticoagulant activity that can be safely adminis- For the oxygen equilibrium study, approximately 100 mL
tered in mice (at doses >300 mg/kg daily for 10 days; aliquot samples from clarified lysates obtained from the anti-
unpublished data) without causing internal bleeding.26,27 sickling studies were mixed with 3 mL of 0.1 M potassium phos-
S-NACH does not bind antithrombin and thus does not phate buffer, pH 7.0, in cuvettes, and subjected to hemoximetry
inhibit systemic antithrombin-dependent clotting factors analysis using a Hemox™ Analyzer (TCS Scientific Corp., New
(activated factors X and II). Sulfation on S-NACH increas- Hope, PA, USA) to assess P50 shifts.36-38 Degree of P50 shift
es the drug’s affinity for endothelium to cause the release (DP50) was expressed as percentage fractions of control
of endothelial tissue factor pathway inhibitor (TFPI).26,28 dimethylsulfoxide-treated samples.
Furthermore, S-NACH interferes with P-selectin-depen- Finally, for the Hb adduct formation study, clarified lysates, also
dent binding of cancer cells29 and RBC30 to endothelial from the above anti-sickling study, were subjected to cation-
cells and regulates plasma levels of adhesion biomarkers in exchange high performance liquid chromatography (Hitachi
SCD mice.30 Finally, S-NACH was further optimized to D-7000 Series, Hitachi Instruments, Inc., San Jose, CA, USA),
interact directly with HbS to exert desirable therapeutic using a weak cation-exchange column (Poly CAT A: 30 mm x 4.6
benefits. mm, Poly LC, Inc., Columbia, MD, USA). Hemoglobin isotype
In this study we tested our hypothesis that S-NACH can peaks were eluted with a linear gradient of mobile phase B from
bind to HbS and directly prevent sickling and decrease 0% to 80% at A410nm (mobile phase A: 40 mM Bis-Tris, 5 mM
inflammation in SCD due to the bidirectional relationship EDTA, pH 6.5; mobile phase B: 40 mM Bis-Tris, 5 mM EDTA,
between inflammation and coagulation31 and investigated 0.15 M sodium chloride, pH 7.5).33,36 A commercial standard con-
the effects of S-NACH on RBC morphology. sisting of approximately equal amounts of composite HbF, HbA,
HbS, and HbC (Helena Laboratories, Beaumont, TX, USA), was
used as the reference isotypes. The areas of new peaks, represent-
Methods ing HbS adducts, were obtained, calculated as percentage fractions
of total Hb area, and reported as levels of modified Hb. All assays
Reagents were conducted in five biological replicates on different days.
S-NACH (average molecular weight 4,000 Da) was synthesized
by Suzhou Ronnsi Pharma Co. Ltd. (Jiangsu Province, China). 5- Animal studies
hydroxymethyl-2-furfural (5-MF) and other common reagents C57/B mice aged 5-6 weeks were purchased from Harlan
were purchased from Sigma (St. Louis, MO, USA). Laboratories (Indianapolis, IN, USA) and acclimatized for 5 days
before initiating TFPI measurements after administration of S-
Sickle blood samples NACH. Townes SCD mice (stock # 013071) were purchased from
Leftover blood samples from individuals with homozygous SS The Jackson Laboratory (Bar Harbor, ME, USA), bred, genotyped,
(SCD) were obtained and used, based on an approved Institutional and used in experiments between 10 and 12 weeks of age. Animal
Review Board protocol at the Children’s Hospital of Philadelphia, studies were conducted at the animal research facility, Albany

O. Abdulmalik et al.
A B
Figure 1. S-NACH binds to intracellular HbS (and HbF) and inhibits sickling of SS red blood cells under hypoxia. In this experiment, SS red blood cells (RBC) were
incubated with or without sulfated non-anticoagulant heparin derivative S-NACH and subjected to hypoxia. (A) Cation-exchange high performance liquid chromatog-
raphy analyses of aliquot samples demonstrated a concentration-dependent modification of intracellular HbS to the high-affinity adduct form. (B) Fixed SS RBC
aliquots were subjected to microscopic image analysis and demonstrated a corresponding dose-dependent inhibition of sickling.
Table 1. Hemoglobin adduct formation, oxygen equilibrium, and anti-sickling studies using homozygous sickle red blood cells with a sulfated non-
anticoagulant heparin derivative.
S-NACH Modified Hb (%)a DP50 (%)b Sickling inhibition (%)c
0.5 mM 20.5±8.2 21.2±10.6 33.1±5.3
1.0 mM 44.5±13.0 57.6±9.0 58.6±14.3
2.0 mM 69.7.5±5.5 65.7±3.2 85.8±4.7
1.0 mM GBT440 ND ND 92.7±4.7
All studies were conducted with SS cell suspensions (20% hematocrit) incubated with 0.5, 1, of 2 mM of sulfated non-anticoagulant heparin derivative (S-NACH). The results are
mean values ± standard deviation for five separate experiments (biological replicates).aHbS adduct values obtained from high performance liquid chromatography elution pat-
terns of hemolysate after incubation of compounds with SS cells. bP50 is the oxygen pressure at which hemolysates are 50% saturated with oxygen. DP50 (%) was determined as:
DP50 (%) = P50 of lysates from untreated cells - P50 of lysates from treated cells x 100
P50 of lysates from untreated cells
c
Anti-sickling studies with SS cells were conducted under hypoxia (2.5% O2/97.5% N2 gas mixture).
College of Pharmacy and Health Sciences (ACPHS; Albany, NY, Tissue factor pathway inhibitor and cytokine assays
USA) in accordance with and approved by the ACPHS Plasma TFPI was measured using a kit from Neoscientific
Institutional Animal Care and Use Committee following institu- (Woburn, MA, USA). Cytokines in blood plasma were measured
tional guidelines for humane animal treatment. Animals were using commercial Bio-Plex beads in a Bio-Plex 200 system (Bio-
maintained under standard climatic and light conditions with Rad Laboratories, Hercules, CA, USA), and analyzed using Bio-
ad libitum access to food and water. For TFPI analysis, plasma was Plex manager software.
obtained from three groups of four C57/B mice each, via retro-
orbital bleeding 2 h after subcutaneous injections of PBS or Morphological analysis
S-NACH (100 mg/kg or 300 mg/kg). For normoxic studies, SCD Total blood was harvested from SCD animals in the presence of
mice were grouped into six groups of six mice each. Blood smears EDTA, treated with PBS, 5-HMF, or S-NACH and incubated under
were made from tail snips before and after subcutaneous injection either normoxia or hypoxia (2% O2) at 37°C for 1 h. A blood
of S-NACH at various time points. Total plasma was harvested for smear from each sample was stained with Leishman stain and
cytokine analysis. Blood smears and plasma were obtained after 2 analyzed under an oil immersion light microscope.39
h from six untreated and 5-HMF-treated animals. For survival
studies, SCD mice were treated with physiological PBS (n=6) or S- Statistical analyses
NACH (n=8) by subcutaneous injection and subjected to hypoxia Results are presented as the means ± standard deviation com-
(5% O2) 30 min after the treatments and observed for 1.5 h. paring experimental and control groups. A t-test was used for sta-

A B
Figure 2. Effect of S-NACH on HbS oxygen binding affinity. (A) The sulfated non-anticoagulant heparin derivative (S-NACH) increases hemoglobin oxygen affinity.
Aliquots of hemolysates from the sickling assay were subjected to p50 analyses using the Hemox Analyzer. (B) Representative curves show a dose-dependent left
shift, indicating an increase in oxygen affinity. Summarized data for biological replicates (n=5) are indicated in the graph. The findings confirm the primary direct
anti-sickling mechanism of S-NACH.
tistical analyses, and results are considered statistically significant mg/kg. 5-HMF (10 mg/kg) was used as a positive control
if P<0.05. because it decreases RBC sickling.33 Both S-NACH and
5-HMF moderately decreased the sickling of RBC by
35-50% (data not shown). Townes SCD mice treated with
Results S-NACH showed a significant (*P<0.05) decrease in the
percentage of circulating sickled RBC for up to 4 h, with a
S-NACH modified intracellular HbS and reduced maximum decrease at 2 h after administration (50% to
sickling of SS cells 35%) (Figure 3B, C) (n=6). Thus, S-NACH can decrease
S-NACH was engineered to have an aldehyde moiety, sickling of RBC under normoxia. Plasma samples (untreat-
which confers anti-sickling properties primarily due to ed, 5-HMF, 2 and 6 h after S-NACH treatment) were quan-
specific interactions with HbS to increase its affinity for titatively analyzed for various pro-inflammatory media-
oxygen. We therefore tested the ability of S-NACH to tors (interleukin [IL] 1b, IL-6, tumor necrosis factor-a
modify HbS, increase oxygen affinity of HbS, and prevent [TNF-a]), anti-inflammatory mediators (IL-10, interferon g
RBC sickling. [IFN-g], monocyte chemoattractant protein 1 [MCP-1]),
Our in vitro sickling assay under hypoxic conditions and growth factors (monocyte colony-stimulating factor
demonstrated that S-NACH, in a dose-dependent manner, [M-CSF], vascular endothelial growth factor [VEGF])
significantly modified intracellular Hb (Figure 1A) and (Figure 4). Plasma levels of IL-1b, IL-6, IFN-g, MCP-1,
reduced the sickling of SS cells, with the maximum effect TNF-a, M-CSF, and VEGF were increased in SCD untreat-
at the concentration of 2 mM, comparable to that of 1 mM ed samples, whereas they were significantly decreased
GBT440 (Figure 1B; Table 1). This supports our hypothesis (P<0.0005) in S-NACH-treated samples, at both 2 and 6 h.
considering that two molecules were designed to target Furthermore, S-NACH was able to increase the decreased
both N-terminal valine residues of the a globin in a levels of 1L-10. The regulatory effect of S-NACH was
tetrameric Hb molecule. most effective at 2 h, similar to its effectiveness on sickled
RBC morphology.
Levels of modified intracellular HbS translated into a
dose-dependent increase in Hb oxygen affinity S-NACH decreases red blood cell sickling and prolongs
When aliquots of HbS-complex solution from the same survival of sickle cell disease mice under hypoxia
studies were investigated in the oxygen equilibrium assay, In the Townes SCD mouse model, hypoxia increases
we observed a similar concentration-dependent effect on RBC sickling, causing death within 15 min due to pul-
increasing HbS affinity for oxygen, with a reduction in P50 monary sequestration of sickled RBC.26 We investigated
values of about 55% at the highest concentration the effect of S-NACH on RBC sickling and survival under
(65.7±3.2 at 2 mM) (Figure 2A; Table 1). These findings hypoxia. Ex vivo deoxygenation was associated with
correlated linearly with the anti-sickling effects and increasing RBC sickling of up to 70%. In the presence of
degrees of HbS modifications, thus confirming this target- S-NACH, sickling was significantly (P<0.05) decreased to
ed mechanism of action (Figure 2B; Table 1). 30% (Figure 5).
In the survival study, while all the untreated mice died
S-NACH decreases in vivo red blood cell sickling and within the first 15 min under hypoxia (5% O2), 50% of the
regulates inflammatory cytokines under normoxia S-NACH-treated mice were alive at 30 min, which was 1
When administered to C57/B mice, S-NACH caused an h after S-NACH administration (Figure 6), a typical times-
approximately 3-fold increase in plasma TFPI after 2 h of pan used for investigating survival.33,40,41 Because S-NACH
treatment (Figure 3A) at both doses tested. To determine exhibited maximal effectiveness at 2 h, mice under hypox-
the effect of S-NACH on RBC sickling, total blood from ia were observed for an additional 1 h, during which
SCD mice was incubated at normoxia with S-NACH. 37.5% of S-NACH-treated animals survived. Thus, S-
Based on the lower effective dose with respect to TFPI NACH increased the survival of SCD animals under
release, the S-NACH dose for animal studies was set at 10 hypoxia for up to 3 h.

Figure 3. Effects of S-NACH under normoxia. (A) A sulfated non-anticoagulant heparin derivative (S-NACH) increases the release of endogenous tissue factor pathway
inhibitor (TFPI). C57/B mice were treated with 100 and 300 mg/kg of free S-NACH, and plasma was obtained after 2 h. TFPI in plasma was measured in duplicate.
TFPI levels were compared between S-NACH-treated samples and phosphate-buffered saline (PBS)-treated control samples, (n=4) (*P<0.05). (B) S-NACH treatment
decreases sickling of red blood cells (RBC) in Townes sickle cell disease (SCD) mice. Blood smears were made from tail snips before and after subcutaneous injection
of S-NACH (10 mg/kg) at the time points shown. 5-hydroxymethyl-2-furfural (5-MF) was used as a positive control (*P<0.05). (C) Morphology of the RBC from Townes
SCD mice was examined in stained blood smears and expressed in percentage. RBC from four different fields or 120 cells were analyzed to calculate the percentage
of sickled RBC. Blood from untreated samples contained higher percentages of sickled and distorted RBC (shown by arrows). S-NACH treatment decreased the pres-
ence of sickled RBC for up to 4 h with the greatest decrease seen at 2 h (n=6) (*P<0.05). SD: standard deviation.
Discussion
that occur in untreated samples from individuals with
We designed S-NACH, a modified low molecular SCD. Although there was no prior evidence to indicate a
weight heparin, to be devoid of anticoagulant properties, relationship between RBC morphology and inflammatory
while acquiring new direct anti-sickling properties. In vitro, mediators, the effect of both on decreased sickling and
S-NACH directly modified intracellular HbS, increased blood viscosity cannot be ruled out.6 Based on our previ-
oxygen affinity, and inhibited sickling of RBC under ous studies, the observed reduction in the levels of pro-
hypoxia. Additionally, S-NACH reduced the levels of cir- inflammatory cytokines was not unexpected. For exam-
culating sickled cells in Townes SCD mice. We confirmed ple, in one previous study in an asthma-induced mouse
the in vitro release of endothelial-TFPI by S-NACH30 in model, S-NACH caused a robust reduction in airway
C57/B mice, as demonstrated by a significant increase in eosinophilia, mucus production, and airway hyperrespon-
plasma TFPI. Based on this, we speculate that S-NACH siveness even after chronic repeated challenges with aller-
might exert local antithrombotic activity by increasing the gen (ovalbumin).44 These effects were linked to suppres-
concentration of endothelial TFPI in the vascular area. An sion of Th2 cytokines IL-4/IL-5/IL-13/GM-CSF and upreg-
in vivo increase in plasma TFPI levels after S-NACH ulation of IL-10. The levels of these inflammatory
administration confirms our earlier reported findings.26 cytokines increased around 2- to 8-fold (in both serum and
According to Kemme et al., TFPI release increased by bronchoalveolar lavage fluid) after induction with the
3-fold (from 62.9 ng/mL to 237 ng/mL) after infusion of allergen and decreased again to baseline after treatment
heparin.42 Kouta et al. reported a marked increase of TFPI with S-NACH. Similar observations were made for total
release (~2.5-fold) within 20 min after intravenous admin- white blood cell count, as well as eosinophil, macrophage,
istration of different types of heparins (bovine, ovine, and and lymphocyte counts, which were markedly elevated in
porcine) to non-human primates.43 Additionally, our the asthma-induced mouse model (6-, 4-, 1.5-, 1.5-, and
observation with different species, including mice, rats, 4-fold, respectively) after exposure to an allergen. S-
and rabbits (unpublished data) are consistent with these NACH also reduced lung fibrosis in mice that were chron-
results. ically exposed to the allergen. As we showed in that study,
We demonstrated in vitro that S-NACH permeated RBC the protective effects of S-NACH were attributable to
membranes to modify HbS and exert a significant anti- modulation of the IL-4/JAK1 signal transduction pathway
sickling effect by maintaining normal RBC morphology, through inhibition of STAT6 phosphorylation and subse-
protecting against the typical changes in RBC morphology quent inhibition of GATA-3 and inducible nitric oxide syn-

Figure 4. S-NACH treatment regulates the levels of inflammatory mediators. Total plasma from sickle cell disease (SCD) mice that were untreated, treated with sul-
fated non-anticoagulant heparin derivative (S-NACH; 10 mg/kg) or treated with 5-hydroxymethyl-2-furfural (5-MF) was harvested and frozen. Cytokines in blood plas-
ma were measured in triplicate. S-NACH treatment significantly changed the plasma levels of the analytes (*P<0.0005). For most analytes, the effects of 5-HMF
were comparable to those of S-NACH at 6 h. SD: standard deviation; IL: interleukin; TNF: tumor necrosis factor; IFN: interferon; MCP-1: monocyte chemoattractant
protein 1; M-CSF: monocyte colony-stimulating factor; VEGF: vascular endothelial growth factor
thase expression. The protective effects of S-NACH treat- acid (glutamic acid) in position 6 of the b chain of Hb with
ment were associated with reductions of the basal expres- a non-polar, hydrophobic valine amino acid and this
sion of the two isoforms of arginase, ARG1 and ARG2, in change causes a disturbance in Hb structure. We therefore
lung epithelial cells.44 speculate that S-NACH (polyanionic glycosaminoglycan)
In another previous study, we measured the different may reverse the polarity to make the HbS more soluble.
biomarkers of inflammation in patients with SCD (35 This mechanism of action remains under investigation.
patients with painful crises and 30 patients in steady state) We further speculate that S-NACH antagonizes hep-
in and 35 healthy donors. Plasma levels of several cidin and might provide additional benefits in SCD by
chemokines and cytokines including TNF-a, IL-1b, IL-6, improving iron hemostasis, as suggested in a recent
IL-8, MCP-1, macrophage inflammatory protein 1a report.49 Additionally, our findings demonstrate that
(MIP1a), and IFN-g in patients with SCD were distinctly S-NACH plays a role similar to that of the non-anticoagu-
and statistically significantly higher during painful crises lant heparin fractions from enoxaparin, which were
and at steady state than in healthy donors (2- to 10-fold shown to have an effect on inflammatory mediators.26,50
increases).45 Furthermore, our thromboelastography assay results with
The observed anti-sickling properties are in concor- S-NACH did not show any changes in platelet functions
dance with our expectations when compared to those (data not shown). Indeed, S-NACH retains all the multi-
produced by GBT440 (voxelotor, Oxybryta), which was modal actions of the low molecular weight heparin tinza-
recently approved for use by the Food and Drug parin but without systemic anticoagulation and its associ-
Administration. There are some concerns that oxygen ated bleeding side effects. Tinzaparin demonstrated signif-
affinity-shifting strategies may be associated with differ- icant effects on the resolution of acute pain crises in
ent cerebrovascular risks,46 although this was adequately patients with SCD in double-blind, randomized clinical
addressed by Estepp.47 Nonetheless, definitive reports on trials.51
long-term use will provide conclusive information on this Overall, the effects of S-NACH on RBC sickling mor-
issue. With this in mind, our multimodal approach also phology, RBC adhesion, and regulation of inflammation
incorporates polyanionic glycosaminoglycans such as resulted in increased survival of SCD mice under hypoxia.
heparins, which can be introduced into sickle RBC HbS by This study serves as a proof-of-concept that S-NACH is
synthetic lipid vesicles. Once introduced, they would safe with respect to bleeding tendencies and argues for
block sickling and also modulate ATPase activity and the further detailed safety and efficacy studies in preclinical
charge of the RBC membrane in hypoxia.48 SCD occurs models of toxicity, the results of which would help guide
due to the replacement of an acidic, hydrophilic amino and inform future human studies.

Figure 5. S-NACH treatment decreases sickling of red blood cells ex vivo under hypoxia. Total blood harvested from sickle cell disease mice (n=8) was mixed with
a sulfated non-anticoagulant heparin derivative (S-NACH) at the dose of 1, 5, or 10 mg/mL and incubated in 2% O2 at 37°C for 1 h. Blood smears were made, stained,
and the morphology of red blood cells (RBC) was analyzed. Hypoxia increased the percentage of sickled RBC. S-NACH treatment decreased the sickling of RBC in a
dose-dependent manner. Phosphate-buffered saline (PBS) was used as a negative control and 5-hydroxymethyl-2-furfural (5-MF) was used as a positive control.
*P<0.05. SD: standard deviation
Figure 6. S-NACH treatment increases the survival of sickle cell disease mice under hypoxia. Sickle cell disease (SCD) mice were treated with phosphate-buffered
saline (PBS, n=6) or 10 mg/kg sulfated non-anticoagulant heparin derivative (S-NACH, n=8). After 30 min, mice were incubated in a hypoxia chamber (5% O2), and
the survival of animals was observed for 1.5 h. Surviving mice were euthanized, as per the guidelines. S-NACH treatment was associated with increased survival of
mice.
S-NACH increased the levels of TFPI in plasma, mary, our data demonstrate direct and support pleiotropic
decreased RBC sickling under normoxia and hypoxia, and effects of S-NACH in ameliorating the complex patho-
reduced the levels of the pro-inflammatory mediators physiological mechanisms involved in SCD. Development
IL-1, IL-6, IFN-g, MCP-1, TNF-a, M-CSF, and VEGF while into an effective drug would lead to improved outcome in
increasing anti-inflammatory factors such as IL-10, further patients globally with SCD, considering the current limit-
establishing it as a promising bona fide multimodal candi- ed therapeutic options, especially for the vast majority of
date drug worthy of additional investigations for acute patients with SCD who reside in underdeveloped areas of
and chronic disease management in SCD patients. In sum- the world.52

Disclosures Acknowledgments
SAM holds a US patent on S-NACH28 and other related US We appreciate Dr. Kelly A. Keating, Pharmaceutical Research
patents. None of the authors has any conflicts of interest. Institute (PRI), for her excellent editing of this manuscript.
Contributions Funding
SAM designed the study and is the Principal Investigator; OA, This project was funded by Vascular Vison Pharmaceuticals
NHED, VM-C and MAT conducted the experiment; OA did the Co. to PRI and an NIH STTR Phase 1 grant (1R41HL147737-
data analysis; OA and NHED contributed equally to the manu- 01-A1, NIH/NHLB “Multi-modal Mechanisms of Novel
script write up and data interpretation.All authors have approved Sulfated Non-Anticoagulant Heparin (S-NACH) in Sickle Cell
the final version of the manuscript. Disease Management”) subaward to Children’s Hospital of
Philadelphia (CHOP), university of Pennsylvania.
References ment-sickle-cell-disease Haemost. 2006;96(6):816-821.

15. Lee JO, Lee JY, Chun EJ, et al. Incidence and 27. Alyahya R, Sudha T, Racz M, Stain SC,
1. Ilesanmi OO. Pathological basis of symp- predictors of venous thromboembolism in Mousa SA. Anti-metastasis efficacy and
toms and crises in sickle cell disorder: medically ill hospitalized elderly cancer safety of non-anticoagulant heparin deriva-
implications for counseling and psy- patients: a prospective observational study. tive versus low molecular weight heparin
chotherapy. Hematol Rep. 2010;2(1):e2. Support Care Cancer. 2019;27(7):2507- in surgical pancreatic cancer models. Int J
2. Lu L, Li X, Vekilov PG, Karniadakis GE. 2515. Oncol. 2015;46(3):1225-1231.
Probing the twisted structure of sickle 16. FDA approves voxelotor for sickle cell dis- 28. Mousa SA, inventor Oxidized heparin frac-
hemoglobin fibers via particle simulations. ease. 2019 [accessed November 25, 2019]; tions and their use in inhibiting angiogene-
Biophys J. 2016;110(9):2085-2093. Available from: https://www.fda.gov/ sis. US patent no. 8,071,569. 2011 Dec 6.
3. Telen MJ. Beyond hydroxyurea: new and drugs/resources-information-approved- 29. Sudha T, Phillips P, Kanaan C, Linhardt RJ,
old drugs in the pipeline for sickle cell dis- drugs/fda-approves-voxelotor-sickle-cell- Borsig L, Mousa SA. Inhibitory effect of
ease. Blood. 2016;127(7):810-819. disease non-anticoagulant heparin (S-NACH) on
4. Connes P, Alexy T, Detterich J, Romana M, 17. Matte A, Zorzi F, Mazzi F, Federti E, pancreatic cancer cell adhesion and metas-
Hardy-Dessources MD, Ballas SK. The role Olivieri O, De Franceschi L. New therapeu- tasis in human umbilical cord vessel seg-
of blood rheology in sickle cell disease. tic options for the treatment of sickle cell ment and in mouse model. Clin Exp
Blood Rev. 2016;30(2):111-118. disease. Mediterr J Hematol Infect Dis. Metastastis. 2012;29(5):431-439.
5. Hebbel RP, Eaton JW, Steinberg MH, White 2019;11(1):e2019002. 30. Alshaiban A, Muralidharan-Chari V, Nepo
JG. Erythrocyte/endothelial interactions 18. Ataga KI, Kutlar A, Kanter J, et al. A, Mousa SA. Modulation of sickle red
and the vasocclusive severity of sickle cell Crizanlizumab for the prevention of pain blood cell adhesion and its associated
disease. Progr Clin Biol Res. 1981;55:145- crises in sickle cell disease. N Engl J Med. changes in biomarkers by sulfated nonanti-
162. 2017;376(5):429-439. coagulant heparin derivative. Clin Applied
6. Kaul DK, Fabry ME, Costantini F, Rubin 19. Oksenberg D, Dufu K, Patel MP, et al. GBT Thromb Hemost. 2016;22(3):230-238.
EM, Nagel RL. In vivo demonstration of red 440 increases haemoglobin oxygen affinity, 31. Petaja J. Inflammation and coagulation. An
cell-endothelial interaction, sickling and reduces sickling and prolongs RBC half-life overview. Thromb Res. 2011;127(Suppl
altered microvascular response to oxygen in a murine model of sickle cell disease. Br 2):S34-37.
in the sickle transgenic mouse. J Clin J Haematol. 2016;175(1):141-153. 32. Hijiya N, Horiuchi K, Asakura T.
Invest. 1995;96(6):2845-2853. 20. Hutchaleelaha A, Patel M, Washington C, Morphology of sickle cells produced in
7. Noguchi CT, Schechter AN, Rodgers GP. et al. Pharmacokinetics and pharmacody- solutions of varying osmolarities. J Lab Clin
Sickle cell disease pathophysiology. namics of voxelotor (GBT440) in healthy Med. 1991;117(1):60-66.
Baillieres Clin Haematol. 1993;6(1):57-91. adults and patients with sickle cell disease. 33. Abdulmalik O, Safo MK, Chen Q, et al. 5-
8. de Jong K, Larkin SK, Styles LA, Bookchin Br J Clin Pharmacol. 2019;85(6):1290-1302. hydroxymethyl-2-furfural modifies intra-
RM, Kuypers FA. Characterization of the 21. Metcalf B, Chuang C, Dufu K, et al. cellular sickle haemoglobin and inhibits
phosphatidylserine-exposing subpopula- Discovery of GBT440, an orally bioavail- sickling of red blood cells. Br J Haematol.
tion of sickle cells. Blood. 2001;98(3):860- able R-state stabilizer of sickle cell hemo- 2005;128(4):552-561.
867. globin. ACS Med Chem Lett. 2017; 34. Asakura T, Mayberry J. Relationship
9. Wautier MP, Heron E, Picot J, Colin Y, 8(3):321-326. between morphologic characteristics of
Hermine O, Wautier JL. Red blood cell phos- 22. Khandros E, Huang P, Peslak SA, et al. sickle cells and method of deoxygenation. J
phatidylserine exposure is responsible for Understanding heterogeneity of fetal Lab Clin Med. 1984;104(6):987-994.
increased erythrocyte adhesion to endothe- hemoglobin induction through compara- 35. Horiuchi K, Ohata J, Hirano Y, Asakura T.
lium in central retinal vein occlusion. J tive analysis of F and A erythroblasts. Morphologic studies of sickle erythrocytes
Thromb Haemost. 2011;9(5):1049-1055. Blood. 2020;135(22):1957-1968. by image analysis. J Lab Clin Med.
10. Setty BNY, Kulkarni S, Stuart MJ. Role of 23. Green NS, Barral S. Emerging science of 1990;115(5):613-620.
erythrocyte phosphatidylserine in sickle hydroxyurea therapy for pediatric sickle 36. Abdulmalik O, Ghatge MS, Musayev FN,
red cell–endothelial adhesion. Blood. cell disease. Pediatr Res. 2014;75(1-2):196- et al. Crystallographic analysis of human
2002;99(5):1564-1571. 204. hemoglobin elucidates the structural basis
11. Proenca-Ferreira R, Brugnerotto AF, Garrido 24. Charneski L, Congdon HB. Effects of of the potent and dual antisickling activity
VT, et al. Endothelial activation by platelets antiplatelet and anticoagulant medications of pyridyl derivatives of vanillin. Acta
from sickle cell anemia patients. PLoS One. on the vasoocclusive and thrombotic com- Crystallogr D Biol Crystallogr. 2011;67(Pt
2014;9(2):e89012. plications of sickle cell disease: a review of 11):920-928.
12. Belcher JD, Marker PH, Weber JP, Hebbel the literature. Am J Health Syst Pharm. 37. Abdulmalik O, Safo MK, Lerner NB, et al.
RP, Vercellotti GM. Activated monocytes in 2010;67(11):895-900. Characterization of hemoglobin bassett
sickle cell disease: potential role in the acti- 25. Garcia DA, Baglin TP, Weitz JI, Samama (a94Asp→ Ala), a variant with very low
vation of vascular endothelium and vaso- MM. Parenteral anticoagulants: antithrom- oxygen affinity. Am J Hematol. 2004;
occlusion. Blood. 2000;96(7):2451-2459. botic therapy and prevention of thrombo- 77(3):268-276.
13. Ataga KI, Key NS. Hypercoagulability in sis, 9th ed: American College of Chest 38. Abdulmalik O, Safo MK, Seeholzer SH,
sickle cell disease: new approaches to an Physicians Evidence-Based Clinical Practice Asakura T, Hasbrouck NC, Russell JE. Hb
old problem. Hematology Am Soc Hematol Guidelines. Chest. 2012;141(2 Suppl):e24S- Baden: structural and functional characteri-
Educ Program. 2007;91-96. 43S. zation. Am J Hematol. 2010;85(11):848-
14. FDA approves new treatment for sickle cell 26. Mousa SA, Linhardt R, Francis JL, 852.
disease. 2017 [accessed September 11, Amirkhosravi A. Anti-metastatic effect of a 39. Leishman WB. Note on a simple and rapid
2020]; Available from: https:// non-anticoagulant low-molecular-weight method of producing Romanowsky stain-
w w w. f d a . g o v / n e w s - e v e n t s / p r e s s - heparin versus the standard low-molecular- ing in malarial and other blood films. Br
announcements/fda-approves-new-treat- weight heparin, enoxaparin. Thromb Med J. 1901;2(2125):757-758.

40. Iyamu EW, Turner EA, Asakura T. Niprisan 44. Ghonim MA, Wang J, Ibba SV, et al. of hemoglobin S with anionic polysaccha-
(Nix-0699) improves the survival rates of Sulfated non-anticoagulant heparin blocks rides. Am J Pediatr Hematol Oncol. Spring
transgenic sickle cell mice under acute Th2-induced asthma by modulating the IL- 1984;6(1):77-81.
severe hypoxic conditions. Br J Haematol. 4/signal transducer and activator of tran- 49. Mohanty P, Jena RK, Sethy S. Variability of
2003;122(6):1001-1008. scription 6/Janus kinase 1 pathway. J Trans iron load in patients of sickle cell anaemia
41. Zhang C, Li X, Lian L, et al. Anti-sickling Med. 2018;16(1):243. (HbSS): a study from Eastern India. J Clin
effect of MX-1520, a prodrug of vanillin: an 45. Qari MH, Dier U, Mousa SA. Biomarkers Diagn Res. 2017;11(3):Ec19-ec22.
in vivo study using rodents. Br J Haematol. of inflammation, growth factor, and coagu- 50. Shastri MD, Stewart N, Horne J, et al. Non-
2004;125(6):788-795. lation activation in patients with sickle cell anticoagulant fractions of enoxaparin sup-
42. Kemme MJ, Burggraaf J, Schoemaker RC, disease. Clin Appl Thromb Hemost. press inflammatory cytokine release from
Kluft C, Cohen AF. Quantification of 2012;18(2):195-200. peripheral blood mononuclear cells of aller-
heparin-induced TFPI release: a maximum 46. Hebbel RP, Hedlund BE. Sickle hemoglobin gic asthmatic individuals. PLoS One.
release at low heparin dose. Br J Clin oxygen affinity-shifting strategies have 2015;10(6):e0128803.
Pharmacol. 2002;54(6):627-634. unequal cerebrovascular risks. Am J 51. Qari MH, Aljaouni SK, Alardawi MS, et al.
43. Kouta A, Hoppensteadt D, Bontekoe E, et Hematol. 2018;93(3):321-325. Reduction of painful vaso-occlusive crisis
al. Studies on tissue factor pathway 47. Estepp JH. Voxelotor (GBT440), a first-in- of sickle cell anaemia by tinzaparin in a
inhibitor antigen release by bovine, ovine class hemoglobin oxygen-affinity modula- double-blind randomized trial. Thromb
and porcine heparins following intravenous tor, has promising and reassuring preclini- Haemost. 2007;98(2):392-396.
administration to non-human primates. cal and clinical data. Am J Hematol. 52. Sankaran VG, Weiss MJ. Anemia: progress
Clin Appl Thromb Hemost. 2020; 2018;93(3):326-329. in molecular mechanisms and therapies.
26:1076029620951851. 48. Winter WP, Seale WR, Yodh J. Interaction Nat Med. 2015;21(3):221-230.

LETTERS TO THE EDITOR
centile 82, 75th percentile 101) and platelets) 35x109/L
SARS-CoV-2 infection in aplastic anemia (range, 2-121; IQR, 42). None developed evidence of sec-
ondary hemophagocytic lymphohistiocytosis.
Coronavirus disease 2019 (COVID-19), caused by Upon review of blood results prior to the SARS-CoV-2
severe acute respiratory syndrome coronavirus-2 (SARS- infection, it was possible to appreciate a progressive
CoV-2), was declared a pandemic by the World Health decline in all hematologic indices consistent with overt
Organization in March 2020. Compared to patients with relapse (confirmed by bone marrow hypocellularity
non-hematologic cancers, patients affected by hemato- meeting diagnostic criteria) in two patients and, although
logic disorders have increased mortality and more pro- not meeting relapse criteria, requiring treatment, intense
longed viral RNA persistence.1-3 Since the early phase of monitoring, and transfusion support in 15 patients.
the pandemic, several groups have described thrombocy- Interestingly, three cases (12.5%) of idiopathic AA
topenia or secondary hemophagocytic lymphohistiocyto- were diagnosed a few weeks after documented SARS-
sis in patients infected by SARS-CoV-2, problems likely CoV-2 infection. Blood counts performed in the immedi-
due to a cytokine storm and the potential cytotoxicity of ate past for other medical reasons showed normal para
the virus.4,5 meters in all these three patients, who developed severe
Aplastic anemia (AA), a rare autoimmune disease with or very severe AA with heavy transfusion dependency,
an incidence of two cases per million population, is char- and eventually required treatment with immunosuppres-
acterized by cytopenia and bone marrow sive therapy or hematopoietic stem cell transplantation
hypocellularity.6 It has been proposed that, in acquired (Table 1). At the time of reporting, all three patients are
AA, an initiating event provokes an aberrant immune in remission with good hematologic response.
response, triggering oligoclonal expansion of cytotoxic T Figure 1 shows the median values of white blood cell
cells that destroy hematopoietic stem cells.
The consequences of SARS-CoV-2 infection in known
cases of AA are not clear.7 Additionally, it is unknown Table 1. Characteristics of the patients with aplastic anemia at the
whether this virus can trigger an aberrant immune time of infection with severe acute respiratory syndrome coron-
response leading to depletion of the stem cell compart- avirus-2.
ment and inducing bone marrow failure. Here we AA disease characteristics
describe the features and clinical outcome of a group of at SARS-Cov-2 infection N (%) or median [range]
patients affected with AA and SARS-CoV-2 infection Number of patients (n, %) 23 (100)
between April 2020 and January 2021. Female 16 (70)
A national survey was launched in April 2020 to assess Male 7 (30)
the clinical features and outcome of patients with pre-
existing AA and new onset AA after SARS-CoV-2 infec- Age in years, median [range] 49 [20-77]
tion. The criteria for diagnosing AA and classifying its Disease category, n (%)
severity were described previously by Camitta et al.8 The Very severe AA 7 (30)
diagnosis of SARS-CoV-2 infection was confirmed by Severe AA 6 (26)
nasopharyngeal swab9 at the onset of symptoms or at Non-severe AA 10 (43)
access to the hematology department. Disease status (n, %)
The study population consisted of 23 patients with AA New onset/diagnosis 3 (13)
(30% with very severe AA, 26% with severe AA and In remission 14 (60)
43% with non-severe AA) with a median age of 49 years On treatment 7 (30)
(range, 20–77); there were seven females and 16 males. On CSA after hATG 6 (26)
All cases were acquired, except one with Fanconi anemia. Eltrompopag 1 (4)
A subclinical paroxysmal nocturnal hemoglobinuria Others* 6 (26)
clone was present in five cases. None of the patients was
Post-HSCT** on IST 7 (30)
vaccinated against SARS-CoV-2.
At the onset of SARS-CoV-2 infection, 60% (14/23) of SARS-CoV2 features N (%)
the patients were on active immunosuppressive therapy Severity
– six on high-dose cyclosporine maintenance treatment Mild 13 (57)
after horse antithymocyte globulin, one on eltrombopag Moderate 7 (30)
and cyclosporine and seven on a combination of Severe 3 (13)
cyclosporine and mycophenolate mofetil – as part of Oxygen supplementation 3 (13)
graft-versus-host disease prophylaxis after a reduced- Intensive care admission 2 (8)
intensity allograft. Table 1 summarizes the populations' AA status after SARS-CoV2
demographic and allogeneic stem cell transplant details. New onset AA 3 (13)
The most common symptoms were fatigue, general Relapse of AA 1 (4)
malaise, fever, dry cough, shortness of breath, loss of Decline in hematologic indices 15 (65)
smell, and diarrhea; 29% (7/23) of patients who devel-
Outcome of AA
oped a COVID-19-defining event (6 pneumonia and 1
Death 1 (4)
hepatitis) were hospitalized (median 5 days; range, 3-12).
New treatment 4 (17)
Within this subgroup, three patients required oxygen
IST 3 (13)
supplementation, of whom two needed escalation to
HSCT 1 (4)
intensive care unit admission for high-flow oxygen and
*Others: included patients who never required treatment for aplastic anemia (AA)
monitoring, but none required mechanical ventilation. and also patients whose cyclosporine was successfully withdrawn after they had
At diagnosis of the infection, median blood count achieved remission of their AA. **Matched unrelated (n=3), matched sibling (n=2),
parameters showed pancytopenia: white blood cells mismatched unrelated (n=1), and haploidentical (n=1); this group includes
2.3x109/L (range, 0.42–5.1; interquartile range [IQR], patients who underwent transplantation either upfront or at failure of immunosup-
pressive therapy. AA: aplastic anemia; SARS-CoV-2: severe acute respiratory syn-
0.97), neutrophils 1.08x109/L (range, 0.14–2.56; IQR, drome coronavirus-2; CSA: cyclosporine A; hATG: horse antithymocyte globulin;
0.68), hemoglobin 93.5 g/L (range, 74–139; 25th per- HSCT: hematopoietic stem cell transplant, IST: immunosuppressive therapy.

542
Figure 1. Hematologic parameters in patients with aplastic anemia and SARS-CoV-2 infection. Median values of white blood cell count (x109/L), hemoglobin (g/L), and platelet count (x109/L) at three different time-points (before, during
and after SARS-CoV2 infection) in the four groups of the study population: (newly diagnosed aplastic anemia [AA], AA on active immunosuppressive therapy [IST], AA off IST and after hematopoietic stem cell transplantation [post-HSCT]).
Note complete data for some of the blood parameters were not available in four cases, which were, therefore, excluded from the graphical illustration.

count, hemoglobin concentration and platelet count at important to prevent cytopathic effects of the virus with
three different time-points (pre-infection, at infection, a successful AA vaccination program, although close
and post-infection) in the four groups of the study popu- monitoring is required as vaccination-induced AA has
lation: newly diagnosed, on active immunosuppressive been reported in the literature.
therapy, off immunosuppressive therapy and after
hematopoietic stem cell transplantation. Daniele Avenoso,1 Judith C.W. Marsh,1 Victoria Potter,1
For those patients affected by transfusion-independent Antonio Pagliuca,1 Simon Slade,1 Fiona Dignan,2 Eleni
non-severe AA (10/23), there was a new requirement of Tholouli,2 Sajjan Mittal,3 Bernard Davis,4 Sudhir Tauro,5
transfusion support in seven patients, but no cases of Rachel Kesse-Adu,6 Morag Griffin,7 Elspeth Payne,8
transition to severe/very severe AA were recorded. Shreyans Gandhi1 and Austin G. Kulasekararaj1
Despite profound neutropenia and being on immuno- 1
suppressive therapy, only 13% of patients (3/23) devel- King’s College Hospital NHS Foundation Trust, London; 2Royal
oped COVID-19. This may reflect some specific, favor- Manchester Infirmary, Manchester; 3Northampton General Hospital,
able host factors (such as young age) or might be second- Northampton; 4Whittington Hospital, London; 5University of Dundee,
ary to protective immune dysregulation known to be Dundee; 6Guys and St Thomas Hospital NHS Foundation Trust,
present in AA.10 Indeed, hypotheses on the role of hyper- London; 7St James University Hospitals, Leeds and 8University College
inflammation resulting in a more severe disease pheno- London, London, UK
type have resulted in proposals of trials to investigate the Correspondence:
use of agents blocking these pathways for the treatment
DANIELE AVENOSO - d.avenoso@nhs.net
of severe SARS-CoV-2 infection in non-AA patients.11
Despite the lack of cytokine studies or viral polymerase doi:10.3324/haematol.2021.279928
chain reaction analysis of bone marrow aspirates in our Received: September 2, 2021.
study, it is reasonable to speculate a potential myelosup- Accepted: October 13, 2021.
pressive effect of SARS-CoV-2: as demonstrated in Figure
1, patients had a clear decline in hematopoiesis, causing Pre-published: October 21, 2021.
worsening of blood parameters and relapse of AA. Disclosures: no conflicts of interest to disclose.
However, the study does not clarify whether the virus Contributions: all the authors were involved in the care of the
has a direct cytotoxic effect on hematopoietic stem cells patients and contributed equally to writing this manuscript; DA col-
or acts through the cytokine storm or aberrant immune
lected and analyzed the data; AK supervised the study.
dysregulation following the infection and might have a
bias due to non-reporting of milder cases. Acknowledgments: the authors thank all NHS staff for fighting
We demonstrate that SARS-CoV-2 infection is another the COVID-19 pandemic
factor that can jeopardize residual hematopoiesis during
AA, as previously described for other viral infections
(e.g., hepatitis). The kinetics of the deterioration in blood
References
counts after SARS-CoV-2 infection mirrors the previously 1. Zhu N, Zhang D, Wang W, et al. A novel coronavirus from patients
reported kinetics of AA diagnosis or relapse in pregnancy. with pneumonia in China, 2019 N Engl J Med. 2020;382(8):727-733.
Although a clear correlation between pregnancy and the 2. Yang X, Yu Y, Xu J, et al. Clinical course and outcomes of critically
onset or relapse of AA has never been demonstrated, sev- ill patients with SARS-CoV-2 pneumonia in Wuhan, China: a single-
eral groups have described worsening of hematologic centered, retrospective, observational study. Lancet Respir Med.
indices at the onset of pregnancy and subsequent recov- 2020;8(5):475-481.
3. Shah V, Ko Ko T, Zuckerman M, et al. Poor outcome and prolonged
ery in the post-partum period.12,13 persistence of SARS-CoV-2 RNA in COVID-19 patients with
Our study does not enable clear conclusions to be haematological malignancies; King’s College Hospital experience. Br
drawn about the severity and long-term prognosis of J Haematol. 2020;190(5):e279-e282.
SARS-CoV-2 infection in AA; despite the lack of COVID- 4. Dewaele K, Claeys R. Hemophagocytic lymphohistiocytosis in
19 deaths, the viral infection was a risk factor for the SARS-CoV-2 infection. Blood. 2020;135(25):2323.
onset of AA and for worsening of blood parameters in 5. Hersby DS, Do TH, Gang AO, Nielsen TH. COVID-19-associated
patients already with AA. pancytopenia can be self-limiting and does not necessarily warrant
This is the first report describing the outcome of AA bone marrow biopsy for the purposes of SARS-CoV-2 diagnostics.
Ann Oncol. 2021;32(1):121-123.
following SARS-CoV-2 infection, and while it is encour- 6. Young NS. Aplastic anemia. N Engl J Med. 2018;379(17):1643-1656.
aging to note that most patients (including transplanted 7. Paton C, Mathews L, Groarke EM, et al. COVID-19 infection in
cases) made a full recovery without the development of patients with severe aplastic anaemia. Br J Haematol.
significant symptoms, this population needs to be consid- 2021;193(5):902-905.
ered at risk of complications of worsening cytopenias fol- 8. Camitta BM, Storb R, Thomas ED. Aplastic anemia: pathogenesis,
lowing COVID-19. Indeed, one patient died as a conse- diagnosis, treatment, and prognosis. N Engl J Med.
quence of infectious complications due to relapsed AA. 1982;306(11):645-652.
9. Corman VM, Landt O, Kaiser M, et al. Detection of 2019 novel
A possible temporal relationship between SARS-CoV2
coronavirus (2019-nCoV) by real-time RT-PCR. Eurosurveillance.
infection and AA can be suggested in three cases in our 2020;25(3):2000045.
series. Are these cases a casual association of SARS-CoV- 10. Dufour C, Ferretti E, Bagnasco F, et al. Changes in cytokine profile
2 infection and AA or are they cases of secondary AA as pre- and post-immunosuppression in acquired aplastic anemia.
a result of the viral insult? The detection of three new Haematologica. 2009;94(12):1743-1747.
cases of AA within a total of 4.5 million cases of SARS- 11. Biran N, Ip A, Ahn J, et al. Tocilizumab among patients with
CoV-2 infection in the UK is intriguing. Further studies COVID-19 in the intensive care unit: a multicentre observational
that include measurement of cytokines and other factors study. Lancet Rheumatol. 2020;2(10):e603-e612.
12. Aitchison RGM, Marsh JCW, Hows JM, Russell NH, Gordon-Smith
such as regulatory T-cell subsets are needed to character-
EC. Pregnancy associated aplastic anaemia: a report of five cases and
ize the immune and inflammatory environment follow- review of current management. Br J Haematol. 1989;73(4):541-545.
ing SARS-CoV-2 infections in AA patients to help predict 13. Oosterkamp HM, Brand A, Kluin-Nelemans JC, Vandenbroucke JR.
outcomes and prognosis. Furthermore, considering the Pregnancy and severe aplastic anaemia: causal relation or coinci-
availability of vaccines against SARS-CoV2 infection, it is dence? Br J Haematol. 1998;103(2):315-316.

to etoposide, which did cause significant cell death at

The insecticides permethrin and chlorpyrifos show doses as low as 1 µM, neither permethrin nor chlorpyri-
limited genotoxicity and no leukemogenic poten- fos induced cell death/apoptosis at doses 50-fold higher
tial in human and murine hematopoietic stem pro- (Online Supplementary Figure S1A), supporting the con-
genitor cells cept that the absence of high-grade DNA fragmentation
may represent a chromatin physical impediment for
Epidemiological and clinical studies have revealed that MLL DSB to fuse in-frame with a partner gene and
maternal exposure to pesticides-insecticides during preg- encode an oncogenic fusion protein. Inverse polymerase
nancy is associated with an increased risk of infant or chain reaction assays confirmed the absence of in-frame
childhood acute leukemia.1-3 The household insecticides MLL fusions (data not shown). Overall, these data indicate
permethrin and chlorpyrifos, which are members of the that chronic exposure to insecticides neither enriches for
pyrethroid and organophosphate families of pesticides, MLL breaks nor generates MLL fusion oncogenes or
respectively, have been associated with both the induc- gross genomic instability.
tion of MLL rearrangements (MLLr) and the develop- To test the bona fide ability of permethrin and chlor-
ment of infant acute leukemia.4-9 Despite the epidemio- pyrifos to function as topoisomerase II poisons, and their
logical association between insecticides and increased ability to generate DNA-DSB we performed an in vivo
risk of leukemia, whether such insecticides act as topo complex of enzyme assays to analyze covalent genomic
isomerase II DNA-damaging poisons remains unknown DNA/topoisomerase II complexes in live cells.12
and studies testing the biological plausibility of such an Treatment of hESC with permethrin and chlorpyrifos at
association are lacking. Here, we assessed the genotoxi- 10–500 µM induced only minimal poisoning of the
city, induction of MLLr and leukemogenic potential of topoisomerase II isoforms (a and b), whereas as little as
permethrin and chlorpyrifos by taking advantage of cut- 1 µM etoposide induced significant poisoning of both
ting-edge in vitro and in vivo models using prenatal, isoforms (Online Supplementary Figure S1B). Similarly,
neonatal and adult hematopoietic stem and progenitor monitoring of g-H2AX by western blotting (Online
cells (HSPC). Our data suggest that the insecticides test- Supplementary Figure S1C) or fluorescent activated cell
ed show no cytotoxicity, limited genotoxicity and no sorting (FACS) analysis (Online Supplementary Figure
leukemogenic potential in human and murine HSPC in S1D-F) revealed that the pesticides were unable to
vitro and in vivo. induce DSB in hESC, neonatal or adult CD34+ cells at
We initially assessed whether acute exposure to etopo- relatively high concentrations, whereas etoposide
side (1 µM), permethrin or chlorpyrifos (10 µM or 50 potently induced DSB that were slightly repaired to
µM) induces MLL breaks in undifferentiated human some extent over time in somatic CD34+ cells but not in
embryonic stem cells (hESC) and CD34+ HSPC derived hESC (Online Supplementary Figure S1F). Overall, the data
from hESC, cord blood or adult peripheral blood (Figure indicate that permethrin and chlorpyrifos are not topoi-
1A). Acute exposure (24 h) to either permethrin or chlor- somerase II poisons and do not generate DNA-DSB in
pyrifos consistently induced MLL breaks in 3–7% of embryonic or somatic CD34+ cells.
embryonic, neonatal and adult CD34+ cells in a dose- We next sought to determine whether prolonged in
independent manner (Figure 1B, C). Embryonic and vivo exposure of human CD34+ HSPC to pesticides
somatic CD34+ cells were sensitive to the three treat- induces MLL breaks or initiates leukemia (Figure 2A).
ments (Figure 1C). Of note, whereas embryonic and Immunodeficient (NSG) mice xenotransplanted with
neonatal CD34+ cells were slightly more sensitive to cord blood-derived human CD34+ HSPC were exposed
etoposide than to insecticides (Figure 1C), no differences for 12 weeks to permethrin or chlorpyrifos in the drink-
were found for adult CD34+ cells, suggesting that the ing water to mimic the environmental exposure in
potential genotoxicity of the insecticides may have a humans. Our target concentration for pesticide exposure
more relevant etiological impact in adult MLLr acute was 10 mg/kg/day, and 5 mg/kg/day for etoposide. As
leukemia. the actual concentration to which mice are exposed will
Chronic exposure to low doses of etoposide has been depend on their weight and their consumption of drink-
reported to induce apoptosis, MLL breaks and major ing water over the experimental period, we measured
chromosomal abnormalities in hESC.10 To test whether both parameters weekly. Results revealed an actual
this occurred after exposure to insecticides, we treated intake of permethrin, chlorpyrifos and etoposide only
hESC for 24 h with 10 µM of either permethrin or chlor- 10–20% lower than the theoretical concentrations (8.5,
pyrifos followed by a daily “booster” dose (2 µM) for 40 9 and 4.2 mg/kg/day, respectively) (Online Supplementary
days (Figure 1D). After 5 days of recovery (without treat- Figure S2A-C). Importantly, the permethrin and chlor-
ments), MLLr breaks and gross genomic abnormalities pyrifos metabolites 3-BPA and TCPy were readily
were assayed by interphase fluorescence in situ detected by gas chromatography-mass spectrometry in
hybridization (iFISH), G-banding and comparative serum and urine as soon as 48 h after pesticide exposure,
genomic hybridization (Figure 1E-G). In contrast to the confirming consistent exposure through administration
frequency of MLL breaks (3–7%) observed after acute of drinking water (Online Supplementary Figure S2D).
exposure (Figure 1C), MLL breaks were scarcely iFISH analysis at sacrifice revealed a small but significant
detectable upon chronic exposure to permethrin or increase in the frequency of bone marrow-engrafted
chlorpyrifos, suggesting a legitimate repair of the DNA human CD45+ cells harboring MLL breaks in etoposide-
damage/double strand breaks (DSB) at the MLL locus treated mice, but not in permethrin- or chlorpyrifos-
(Figure 1E). Likewise, karyotyping and comparative treated mice (Figure 2B). FACS analysis at sacrifice
genomic hybridization analysis revealed no numerical or revealed similar levels of human graft and normal multi-
structural chromosomal alterations (Figure 1F) or DNA lineage (immature, myeloid and B-cell lymphoid)
gains or losses after chronic exposure (Figure 1G). It has engraftment in the bone marriow and peripheral blood
been previously suggested that the chromosomal topol- of etoposide-, permethrin- and chlorpyrifos-treated mice
ogy and chromatin structure resulting from early apop- (Figure 2C, Online Supplementary Figure S2E), with no
tosis may represent the underlying substrate for MLL evidence of splenomegaly (Figure 2D). These findings
chromosomal translocations to occur.11 Here, in contrast indicate that chronic in vivo exposure to the indicated

A B
E F
Figure 1. In vitro induction of MLL rearrangements in embryonic, neonatal and adult human CD34+ hematopoietic stem and progenitor cells following acute
and chronic exposure to etoposide, permethrin and chlorpyrifos. (A) Experimental design to assess the induction of MLL breaks in human undifferentiated
embryonic stem cells (hESC) and CD34+ hematopoietic stem and progenitor cells (HSPC) after 24 h single-pulse exposure to the indicated treatments (etoposide,
ETO; permethrin, PER; and chlorpyrifos, CPF). (B) Representative interphase fluorescence in situ hybridization (iFISH) images showing MLL germline and
rearranged (MLLr) human CD34+ cells. (C) Frequency of MLL breaks in undifferentiated hESC and embryonic, neonatal and adult CD34+ HSPC after 24 h sin-
gle-pulse exposure to the indicated treatments (n=3 independent experiments for each cell type). Asterisks indicate statistically significant differences of a given
treatment as compared with dimethylsulfoxide (DMSO, vehicle treatment) *P<0.05, **P<0.01. Dotted lines in the graphs show the percentage of MLL breaks
in the DMSO-treated control groups. A minimum of 500 nuclei were analyzed except in some samples for which 80-400 nuclei were analyzed. (D) Experimental
design to analyze the frequency of MLL breaks and gross chromosomal damage in hESC after continuous exposure to PER and CPF. (E) Frequency of MLL breaks
detected by iFISH 45 days after chronic treatment with either PER or CPF (n=3 independent experiments). A minimum of 400 nuclei was analyzed per experi-
ment. (F) Representative image of a G-banding karyotype 45 days after chronic treatment with either PER, CPF or DMSO (n=3). (G) Representative image of DNA
copy number variation profiling by comparative genomic hybridization array analysis 45 days after chronic treatment with either PER or CPF. CGH: comparative
genomic hybridization; AMP: amplification; DEL: deletion; LOH: loss of heterozygosity.

doses of etoposide or the pesticides fails to induce MLL cides (20 mg/kg/day) in drinking water from gestational
breaks or to initiate leukemia in NSG-reconstituting day 0.5 to day 21 (Figure 3A). No significant differences
human CD34+ HSPC. were found between treatment groups for the number of
We next assessed whether chronic exposure to pesti- pups per litter at birth (range, 11–18) or sex distribution
cides during embryonic development induces Mll breaks (Online Supplementary Figure S3A, B). The mothers and
in developing murine HSPC. To do this, CD1 male and one-half of the litter were euthanized at weaning to ana-
female mice were mated and pregnant CD1 females lyze the impact of etoposide and pesticides on the Mll
were exposed to etoposide (10 mg/kg/day) or to pesti- locus and on hematopoietic homeostasis, and the
Figure 2. Continuous exposure to permethrin or chlorpyrifos fails to induce MLL breaks or leukemia in NSG-reconstituting human CD34+ hematopoietic stem
and progenitor cells. (A) Experimental design to determine whether prolonged in vivo exposure of human CD34+ hematopoietic stem and progenitor cells (HSPC)
to insecticides induces MLL breaks or can initiate leukemia. In brief, 32 NSG mice were irradiated and cord blood-derived human CD34+ cells were transplanted
into the bone marrow 6–8 h later. Four days later, mice were divided into four groups (8 mice/group) to initiate exposure in drinking water to etoposide (ETO, 5
mg/kg/day), permethrin (PER, 10 mg/kg/day), and chlorpyrifos (CPF, 10 mg/kg/day) or 0.1% dimethylsulfoxide (DMSO). The presence of the PER and CPF
metabolites 3-BPA and TCPy in urine and serum was analyzed 48 h later by gas chromatography-mass spectometry. Mice were sacrificed for interphase fluores-
cence in situ hybridization (iFISH) and fluorescence activated cell sorting (FACS) analysis after 12 weeks of continuous treatment. (B) Left, scheme depicting
the human chromosome 11 and the 11q23 region where the MLL probe hybridizes. Middle, representative iFISH image showing human cells with germline MLL
or MLL rearrangement (MLLr). Right, percentage of human CD45+ cells harboring MLL breaks detected by iFISH at sacrifice. A minimum of 500 nuclei per sam-
ple was analyzed, except in one mouse from the DMSO group, for which only 216 nuclei could be analyzed. (C) Upper panels show the percentage of human
engraftment (CD45+HLA-ABC+ cells) in bone marrow and peripheral blood. Lower panels show the relative proportion of immature (CD34+), myeloid (CD33+) and
B-cell (CD19+) populations within human engraftment in bone marrow and peripheral blood. (D) Representative macroscopic images of spleens at sacrifice for
each experimental group. GC-MS: gas chromatography-mass spectrometry.

remaining half of the litter was maintained for 32 weeks progenitors in the mothers and adult offspring, but not
for analysis at adulthood (Figure 3A). iFISH analysis in the pups (Figure 3B, C, Online Supplementary Figure
revealed that exposure to permethrin or chlorpyrifos S3C).
during pregnancy failed to induce Mll breaks in bone Finally, we analyzed whether prenatal exposure to
marrow Lin-Kit+ progenitors from mothers, pups or adult etoposide or pesticides affects the hematopoietic home-
offspring, whereas etoposide exposure induced Mll ostasis of mothers and pups (at weaning) and adult off-
breaks in a small but significant proportion of Lin-Kit+ spring. Analysis at sacrifice revealed no significant differ-
B C
D E
Figure 3. Legend on following page

Figure 3. Continuous exposure to permethrin or chlorpyrifos during pregnancy fails to induce Mll breaks in bone marrow progenitors or alterations in the
hematopoietic homeostasis of mothers, pups or adult offspring in mice. (A) Experimental design to determine whether prenatal exposure to etoposide (ETO),
permethrin (PER) or chlorpyrifos (CPF) induces Mll breaks or hematopoietic alterations in mothers, pups or adult offspring in mice. In brief, pregnant CD1 female
mice were exposed to PER (20 mg/kg/day), CPF (20 mg/kg/day), ETO (10mg/kg/day) or 0.1 % dimethylsulfoxide (DMSO) from day 0 to day 21 of gestation. The
number of pups per litter and their sex distribution were analyzed at birth. The mothers and one-half of the litter were analyzed at weaning while the remaining
one-half of the offspring were maintained for 32 weeks for analysis in adulthood. (B) Upper panel, scheme depicting mouse chromosome 9 and the region where
mouse Mll probes hybridize (UCSC, GRCm38/mm10). Lower left panel, a representative fluorescence in situ hybridization (FISH) image of a mouse metaphase
and presence of fluorescence signals in both Mll alleles. Lower right panel, a zoom image of mouse chromosome 9 by DAPI banding, with and without the BAC
fluorescence signal revealing the Mll gene localization. (C) Percentage of mouse bone marrow (BM) LK cells with Mll breaks detected by interphase FISH at sac-
rifice. DMSO and ETO were used as negative and positive controls, respectively. The numbers in bars indicate the number of mice analyzed. (D) Fluorescence
activated cells sorting (FACS) BM analysis of the Lin- Sca-1+ Kit+ (LSK) subpopulation (upper panels), hematopoietic stem and progenitor cell subsets (middle
panels), and mature cells (lower panels) in mothers, pups and adult offspring. (E) Representative macroscopic images of spleens (upper panel) and livers (lower
panel) at sacrifice of mothers, pups and adult offspring exposed to the indicated treatments. HSC: hematopoietic stem cells; MPP: multipotent progenitors; HPC:
hematopoietic progenitor cells.
ences in the proportions of mature cells (myeloid, T and mal translocations encoding MLL fusions. This explains
B) in peripheral blood or total Lin-Sca+Kit+ progenitors, their limited genotoxicity and no leukemogenic potential
hematopoietic stem cells, multipotent progenitors or and reinforces why these compounds are still considered
hematopoietic progenitor cells in the bone marrow as non-classifiable carcinogens. Linking environmental
between treatments in any group (Figure 3D, Online or genotoxic exposure to causal and/or functional MLL
Supplementary Figure S3D). Similarly, prenatal exposure chromosomal translocations has long been
to etoposide or pesticides did not affect the hematopoi- controversial,11 and the cutting-edge in vitro and in vivo
etic homeostasis in peripheral blood, as determined by cellular models employed in the present study also have
absolute numbers of white blood cells, red blood cells obvious limitations. For instance, site-specific cleavage
and platelets (Online Supplementary Figure S3E). Lastly, within the MLL break cluster region (bcr) has been
no evidence of splenomegaly or hepatomegaly was shown to be induced by either topoisomerase II poisons
observed in mothers, pups or adult offspring (Figure 3E). but also genotoxic chemotherapeutic agents which do
Our results thus suggest that chronic exposure to perme- not target topoisomerase II and even by non-genotoxic
thrin or chlorpyrifos during pregnancy does not induce stimuli of apoptotic cell death. In addition, MLL chromo-
Mll breaks in bone marrow progenitors or alterations in somal translocations have been linked to higher-order
the hematopoietic homeostasis of mothers, pups or adult chromatin fragmentation that occurs during the initial
offspring. stages of apoptosis, suggesting that the generation of
A unique strength of the present study is the cutting- MLL chromosomal translocations (and likely others) are
edge in vivo models employed to assess the genotoxicity part of a generalized acute apoptotic response-mediated
and leukemogenesis potential of etoposide, permethrin higher-order chromatin fragmentation which ultimately
or chlorpyrifos. The NSG mice model was established to renders a chromosome topology and chromatin struc-
mimic the adult exposure associated with occupational ture prone to chromosomal DNA exchanges.14 This is
risk, whereas the CD1 mice model attempted to mimic further supported by the ambiguity of MLL transloca-
prenatal exposure to topoisomerase II poisons and insec- tions partnering with a large number of different chro-
ticides suggested to be involved in the etiology of infant mosomal loci.
leukemia. Continuous exposure to permethrin or chlor-
pyrifos in both models failed to induce MLL breaks or Virginia C. Rodriguez-Cortez,1 María Pilar Navarrete-
alterations in hematopoietic homeostasis, confirming Meneses,2* Oscar Molina,1* Talia Velasco-Hernandez,1
the in vitro results of limited genotoxicity and no leuke- Jessica Gonzalez,3 Paola Romecin,1 Francisco Gutierrez-
mogenic potential of permethrin or chlorpyrifos in Agüera,1 Heleia Roca-Ho,1 Meritxell Vinyoles,1 Eric
human and murine HSPC after chronic exposure. The Kowarz,4 Pedro Marin,5 Sandra Rodriguez-Perales,6 Carlos
fact that MLL breaks are acutely induced by permethrin Gomez-Marin,7 Patricia Perez-Vera,2 Felipe Cortes-
or chlorpyrifos but are not sustained upon long-term Ledesma,7 Anna Bigas,1,3,8 Andrea Terron,9 Clara Bueno1,8
chronic exposure in vitro or in vivo indicates a legitimate and Pablo Menendez1,8,10
repair of the DNA damage/DSB in the MLL locus. Of 1
note, although long-term in vivo exposure to etoposide Josep Carreras Leukemia Research Institute. Department of
did induce MLLr in some hematopoietic progenitors, it Biomedicine. School of Medicine, University of Barcelona. Barcelona,
failed to initiate leukemia in either in vivo models, in line Spain; 2Laboratorio de Genética y Cáncer, Departamento de Genética
with a previous study confirming that in utero exposure Humana, Instituto Nacional de Pediatría, Ciudad de México, México;
3
to etoposide did not trigger the development of leukemia Cancer Research Program, Institut Hospital del Mar d'Investigacions
in either Atm+/+ or Atm-/- mice.13 The eventual develop- Mèdiques, Hospital del Mar, Barcelona, Spain; 4Institute of
ment of overt leukemia might depend on the survival Pharmaceutical Biology/DCAL, Goethe-University of Frankfurt,
and proliferative advantage of minor MLLr pre-leukemic Frankfurt/Main, Germany; 5Hematology Department. Hospital Clínic
clones, targeting the right cell-of-origin, on stromal bone de Barcelona, Barcelona, Spain; 6Molecular Cytogenetics Group,
marrow interactions and also on the acquisition of sec- Human Cancer Genetics Program, Spanish National Cancer Research
ondary cooperating oncogenic alterations. The clearance Center (CNIO), Madrid, Spain; 7Topology and DNA Breaks Group,
and lack of selection of MLLr clones is consistent with Spanish National Cancer Center (CNIO), Madrid, Spain; 8Centro de
the development of MLLr treatment-related acute Investigación Biomedica en Red-Oncología (CIBERONC), Madrid,
leukemia in adults or infant leukemia in only a rare Spain; 9European Food and Safety Authority. Parma. Italy and
10
subset of patients exposed to topoisomerase II poisons. Instituciò Catalana de Recerca i Estudis Avançats (ICREA),
Our results clearly suggest that permethrin and chlor- Barcelona, Spain.
pyrifos induce MLL breaks in human HSPC across *MPNM and OM contributed equally.
ontogeny. However, such insecticide-induced DNA-DSB Correspondence:
are successfully repaired, and do not involve chromoso- PABLO MENÉNDEZ - pmenendez@carrerasresearch.org

VIRGINIA RODRÍGUEZ-CORTEZ - vrodriguez@carrerasre- EXPLORA Project BIO2017-91272-EXP and AECC_Lab_2020

search.org Project. PM is an investigator of the Spanish Cell Therapy cooper-
doi:10.3324/haematol.2021.279047 ative network (TERCEL).
Received: April 21, 2021.
Accepted: October 19, 2021. References
Pre-published: October 28, 2021. 1. Infante-Rivard C, Labuda D, Krajinovic M, Sinnett D. Risk of child-
Disclosures: no conflicts of interest to disclose. hood leukemia associated with exposure to pesticides and with
gene polymorphisms. Epidemiology. 1999;10(5):481-487.
Contributions:VR-C and MPN-M: conceptualization, methodol- 2. Ma X, Buffler PA, Gunier RB, et al. Critical windows of exposure to
ogy, formal analysis and investigation, writing and preparation of household pesticides and risk of childhood leukemia. Environ
the original draft; OM and TV-H: methodology, formal analysis Health Perspect. 2002;110(9):955-960.
and investigation; JG, PR, FG-A, HR-H, EK, MV, SR-P and 3. Turner MC, Wigle DT, Krewski D. Residential pesticides and child-
CG-M: methodology; PMa, AB, AT, PP-V and FC-L: resourses; hood leukemia: a systematic review and meta-analysis. Environ
Health Perspect. 2010;118(1):33-41.
CB and Pme: conceptualization, writing, review and editing, super-
4. Navarrete-Meneses MP, Pedraza-Meléndez AI, Salas-Labadía C,
vision. Moreno-Lorenzana D, Pérez-Vera P. Low concentrations of perme-
Acknowledgments:the authors would like to thank the staff of the thrin and malathion induce numerical and structural abnormalities
mass spectrometry unit of the scientific and technological services of in KMT2A and IGH genes in vitro. J Appl Toxicol. 2018;38(9):1262-
the University of Barcelona for pesticide metabolite analysis. We are 1270.
5. Navarrete-Meneses MDP, Pérez-Vera P. Pyrethroid pesticide expo-
also grateful to the staff at the animal facility of the scientific park of sure and hematological cancer: epidemiological, biological and
Barcelona for urine collection in metabolic cages. molecular evidence. Rev Environ Health. 2019;34(2):197-210.
Funding: this work was supported by CERCA/Generalitat de 6. Lu C, Liu X, Liu C, et al. Chlorpyrifos induces MLL translocations
Catalunya and Fundació Josep Carreras-Obra Social la Caixa through caspase 3-dependent genomic instability and topoiso-
merase II inhibition in human fetal liver hematopoietic stem cells.
through institutional support. Pablo Menendez’s laboratory was
Toxicol Sci. 2015;147(2):588-606.
supported by the Spanish Ministry of Science and Innovation 7. Ferreira JD, Couto AC, Pombo-de-Oliveira MS, Koifman S. In utero
(PID2019-108160RB-I00 / AEI / 10.13039/501100011033), pesticide exposure and leukemia in Brazilian children < 2 years of
Retos Colaboración with Banc de Sang i Teixits (RTC-2017-6367- age. Environ Health Perspect. 2013;121(2):269-275.
1), the European Food and Safety Authority 8. Ding G, Shi R, Gao Y, et al. Pyrethroid pesticide exposure and risk
(EFSA.PRAS.2018.04-CT1), and the European Research Council. of childhood acute lymphocytic leukemia in Shanghai. Environ Sci
VR-C and MV were supported by Juan de la Cierva fellowships Technol. 2012;46(24):13480-13487.
9. Borkhardt A, Wilda M, Fuchs U, Gortner L, Reiss I. Congenital
from the Spanish Ministry of Science and Innovation (FJCI-2015- leukaemia after heavy abuse of permethrin during pregnancy. Arch
24303 and IJCI-2017-33172). CB was supported by the Health Dis Child Fetal Neonatal Ed. 2003;88(5):436-437.
Institute Carlos III (ISCIII/FEDER PI17/01028 and PI20/00822). 10. Bueno C, Catalina P, Melen GJ, et al. Etoposide induces MLL
PP-V and PN-M were supported by Consejo Nacional de Ciencia rearrangements and other chromosomal abnormalities in human
y Tecnología (CB-2012-01-183467) and by Fondos del embryonic stem cells. Carcinogenesis. 2009;30(9):1628-1637.
Presupuesto Federal para la Investigación (project 001/2013, 11. Stanulla M, Wang J, Chervinsky DS, Thandla S, Aplan PD. DNA
Instituto Nacional de Pediatría). JG was supported by Centro de cleavage within the MLL breakpoint cluster region is a specific event
which occurs as part of higher-order chromatin fragmentation dur-
Investigación Biomédica en Red en Cáncer (PID2019-104695RB- ing the initial stages of apoptosis. Mol Cell Biol. 1997;17(7):4070.
I00). TV-H was supported by a Maria Sklodowska-Marie Curie 12. Schellenberg MJ, Lieberman JA, Herrero-Ruiz A, et al. ZATT
fellowship (792923). OM was supported by a Beatriu de Pinós (ZNF451)–mediated resolution of topoisomerase 2 DNA-protein
postdoctoral fellowship (BP2016-00048) from the Generalitat de cross-links. Science. 2017;357(6358):1412-1416.
Catalunya, a Lady Tata award from the Lady Tata Memorial 13. Nanya M, Sato M, Tanimoto K, Tozuka M, Mizutani S, Takagi M.
Trust, and an investigator fellowship from the Spanish Cancer Dysregulation of the DNA damage response and KMT2A
rearrangement in fetal liver hematopoietic cells. PLoS One.
Research Association (AECC INVES211226MOLI). SR-P is sup-
2015;10(12):1-18.
ported by grants from the Spanish National Research and 14. Aplan PD. Chromosomal translocations involving the MLL gene:
Development Plan, Instituto de Salud Carlos III, and FEDER molecular mechanisms. DNA Repair (Amst). 2006;5(9-10):1265-
(PI17/02303, PI20/01837 and DTS19/00111); AEI/MICIU 1272.

CD4+ T cells, indicating that the mutations were specific

Somatic STAT3 mutations in CD8+ T cells of healthy for the CD8+ T-cell subset. Among the four STAT3
blood donors carrying human T-cell leukemia virus mutations detected, three were missense STAT3 muta-
type 2 tions (Y640F, N647I and D661Y) and one was a non-
frameshift insertion (Y657_K658insY), with VAF of
Chronic viral antigen stimulation may underlie CD8+ 11.9%, 0.5%, 4.9%, and 1.2%, respectively (Figure 1B).
T-cell expansion and T-cell large granular lymphocyte All the mutations identified in CD8+ T cells were locat-
leukemia (T-LGLL).1 In T-LGLL, CD8+ T-cell expansions ed in the SH2 domain of STAT3 and have been previous-
are associated with somatic mutations that solidify clon- ly reported in T-LGLL (Online Supplementary Figure S1A).2
al dominance, such as in the case of activating STAT3 The proportion of differentiated, putatively cytotoxic
mutations which are found in 40% of patients.2 CD57+, CD16+ CD8+ T cells were higher in STAT3
However, whether chronic exposure to viral antigens are mutated compared to STAT3 unmuted HTLV-2 positive
associated with somatic mutations in expanding CD8+ T individuals (Figure 1C). In addition, higher level of the
cells among individuals without clinically detectable cytotoxic marker perforin was noted in CD8+ T cells of
lymphoproliferations is currently not known. STAT3 mutated, HTLV-2 positive individuals (Online
Human T-cell leukemia virus type 2 (HTLV-2) prefer- Supplementary Figure S1B and C). TCRb deep sequencing9
entially targets CD8+ T cells, causing strong expansion of sorted CD8+ T cells revealed higher clonality index in
of the infected CD8+ T-cell clones.3 Whereas HTLV-1 is the STAT3 mutated compared to STAT3 unmuted indi-
the causative agent of adult T-cell leukemia/ lym- viduals (Figure 1D). The VAF of the STAT3 Y640F muta-
phomas,4 the etiologic role of HTLV-2 in lymphoprolifer- tion was consistent with clonal event in the largest
ative diseases is less clear. In 1992 an incidental case of TCRBV03-01 clone; the other three STAT3 mutations
LGL leukemia with HTLV-2 seropositivity was with smaller VAF likely occurred as subclonal events or
described.5 Later, Thomas et al.6 reported anti-HTLV in the smaller CD8+ T-cell clones (Figure 1E).
antibody positivity in 44% of T-LGLL patients. While The age distribution was similar between HTLV-2 neg-
some cross-reactivity cannot be excluded, 7.5% (4 of 53) ative subjects (median age 52 year), HTLV-2 positive
of T-LGLL patients tested positive for HTLV-2 by both subjects without STAT3 mutations (median age 53
western blotting and polymerase chain reaction (PCR),6 years), and HTLV-2 positive subjects with STAT3 muta-
suggesting that HTLV-2 may participate in T-LGLL tions (median age 58.5 years) (P-value between HTLV-2
pathogenesis in a minority of cases. In this study, we positive subjects with and without STAT3 mutations,
examined whether CD8+ T cells from healthy, asympto- P=0.50; Mann–Whitney U test) (Figure 1F). No statistical
matic blood donors with chronic HTLV-2 infection har- difference in viral load was detected between STAT3
bor somatic mutations in STAT3 or other immune-asso- mutated (median =0.149) and unmuted HTLV-2 positive
ciated genes, potentially identifying invidiuals at risk of subjects (median =0.0005) (P=0.5; Mann–Whitney U
subsequent lymphoproliferative diseases. test) (Figure 1G). No serial HTLV-2 viral load measure-
This study was conducted with samples from the ments were available; however, HTLV-2 viral load has
HTLV Outcomes Study (HOST) which includes samples been reported to be stable over time.10 There was no dif-
from subjects recruited from five major US blood dona- ference in total white blood cell and lymphocyte counts
tion centers.7 HTLV status was analyzed with an ezyme- between STAT3 mutated and unmuted cases (Figure 1H
linked immunosorbant (ELISA) assay, followed by west- and I).
ern blot confirmation and HTLV-1 versus HTLV-2 typing In order to characterize a larger spectrum of somatic
by either real-time PCR or a type specific ELISA.7 At each variants in genes linked to immune regulation, we ana-
visit, cohort participants were interviewed in detail for lyzed CD8+ T cells from 28 HTLV-2 positive subjects
symptoms, followed by a physical and neurological using a custom next generation sequencing panel cover-
examination. Informed consent was obtained from all ing the coding regions of 2,533 immune-related genes.11
participants. The study and sample collection were Samples were sequenced with Illumina HiSeq or
approved by the University of California San Francisco NovaSeq 6000 system (Online Supplementary Figure S2),
committee on human research and other Institutional and somatic variant calling followed a previously
Review Boards. We obtained frozen peripheral blood described approach.11 Variants were filtered using popu-
mononuclear cells (PBMC) of 30 HTLV-2 infected and 35 lation based filtering, MuTect2 filters and against CD4+
HTLV-2 uninfected blood donors from University of and CD8+ panels of normals from 21 healthy controls
California San Francisco and Vitalant Research Institute (Online Supplementary Figure S2). Variants only found in
(CA, USA). The PBMC samples collected between 2000 CD8+ T cells, using sorted CD4+ T cells as matched nor-
and 2008 were randomly selected from HOST. mals, were evaluated further. Sequencing coverages are
We separated CD4+ and CD8+ T cells from PBMC presented in the Online Supplementary Figure S2. Two
samples of HTLV-2 positive (n=30) and negative (n=35) STAT3 mutations (Y640F and D661Y) were detected in
healthy blood donors. The presence of STAT3 mutations two of four STAT3 mutant cases. N647I and
in the sorted fractions was analyzed by ultra-deep tar- Y657_K658insY variants did not pass filtering due to
geted amplicon sequencing, covering hotspot regions of lower sequencing coverage compared to amplicon
STAT3 gene (median coverage =7,472, sensitivity =0.5% sequencing; their presence was confirmed with visual
variant allele frequency [VAF]). The sequencing was per- inspection of the sequencing data using Integrative
formed with Illumina Miseq System (Online Genomics Viewer.12 In addition to STAT3 mutations, we
Supplementary Figure S1A), and variant calling was per- identified a total of 66 coding somatic variants in 61
formed as previously reported.8 Somatic nonsynony- genes in CD8+ T cells (Figure 2A). Nineteen (68%) of the
mous STAT3 mutations were discovered in CD8+ T cells subjects had at least one variant, and the median number
from four of 30 (13.3%) HTLV-2 positive subjects, of variants was one per subject.
whereas no STAT3 mutations were discovered in HTLV- Eight subjects (29%) harbored variants in the genes
2 negative subjects using deep amplicon sequencing of previously discovered in LGLL (STAT3, KMT2D, TYRO3,
STAT3 (Figure 1A; Fisher’s exact test P=0.04). DIDO1, BCL11B, CACNB2, KRAS, LRBA and FANCA),13
Furthermore, no STAT3 mutations were discovered in and five subjects (18%) harbored genes involved in JAK-

A B
C D E
F G H I
Figure 1. STAT3 mutations discovered in CD8+ T cells of HTLV-2 positive subjects. (A) Prevalence of STAT3 mutations in CD4+ and CD8+ T cells of human T-
cell leukemia virus type 2 (HTLV-2) positive subjects (n=30) and HTLV-2 negative subjects (n=35). Four (13.3%; Fisher’s exact test P=0.04) of 30 HTLV-2 positive
individuals had STAT3 mutations in CD8+ T cells. (B) STAT3 mutations found in HTLV-2 positive subjects by amplicon sequencing. STAT3 mutations (insertion,
N647I, Y640F and D661Y) were discovered in CD8+ T cells from 4 HTLV-2 positive subjects. (C) Flow cytometry based immunophenotyping was performed to
identify the proportion of differentiated, putatively cytotoxic (CD56+, CD57+, and CD16+) CD8+ T cells in 9 HTLV-2 positive blood donors. For the immunophe-
notyping, anti-CD3 APC, -CD45 V500, -CD4 APC-H7, -CD8 PE-Cy7, -CD16 PerCP-Cy5.5, -CD56 FITC, and -CD57 PE, were used. Each dot represents one individual,
and horizontal lines indicate median values. Statistically significant difference was evaluated using Mann-Whitney U test. (D) The CD8+ T-cell clonality index by
STAT3 mutation status in 8 HTLV-2 positive blood donors. The clonality index was calculated using ImmnoSEQ Analyzer software (Adaptive Biotechnologies, WA,
USA) as 1 minus Shannon entropy normalized by the logarithm of the number of productive T-cell receptor (TCR) sequences. Each dot represents 1 individual,
and horizontal lines indicate median values. P-values were evaluated using Mann-Whitney U test. (E) CD8+ T-cell repertoire analyzed with TCRb deep sequencing
(Adaptive Biotechnologies). Sorted CD8+ T cells of HTLV-2 positive cases bearing STAT3 mutations (n=4) and without STAT3 mutations (n=4) were used. Variant
allele frequency (VAF) was analyzed by amplicon sequencing. The graph shows top 3 TCR clones in each sample. (F-I) (F) Age distribution, (G) HTLV-2 proviral
load in copies per peripheral blood mononuclear cells (PBMC), (H) white blood cells count and (I) lymphocytes count within HTLV-2 negative subjects (HTLV2-),
HTLV-2 positive subjects without STAT3 mutations (No STAT3Mut) and HTLV-2 positive subjects harboring STAT3 mutations (STAT3Mut). Each dot represents one
individual. P-values were calculated using Mann–Whitney U test (No STAT3Mut vs. STAT3Mut). Horizontal lines indicate median values. Ref: reference base; Var:
variant base.

Table 1. Clinical characteristics of HTLV-2 positive subjects analyzed in immunogene panel sequencing.
Mutation Number
Total 0 1-2 >3 P-value
n = 28 (%) n = 10 (36) n = 8 (29) n = 10 (36)
Age
Median (range) 53 (43 - 70) 52 (46 - 70) 54 (43 - 70) 60 (49 - 68) 0.04†
Sex
Female 19 (68) 7 (70) 6 (75) 6 (60) 0.63^
Male 9 (32) 3 (30) 2 (25) 4 (40)
Race
Black 9 (32) 1 (10) 4 (50) 4 (40) 0.41^
Hispanic 5 (18) 2 (20) 1 (12) 2 (20)
White 14 (50) 7 (70) 3 (38) 4 (40)
White blood cells (109/L) 6.2 6.2 6.6 5.8 0.85†
Median (range) (3.5-10) (3.5-8.2) (6.0-9.1) (3.9-10)
Hemoglobin (g/dL) 13.9 13.4 13.6 14.2 0.27†
Median (range) (4.8-17.4) (4.8-16.7) (12.5-17.4) (12.8-16.2)
Hematocrit (%) 40.8 40.3 39.9 42.7 0.15†
Median (range) (14-50.5) (14-49.2) (38.7-50.5) (38.1-48.9)
Platelets (109/L) 231 222 239 240 0.80†
Median (range) (101-435) (144-435) (200-340) (101-350)
Lymphocytes (109/L) 2.378 1.906 2.907 2.378 0.24†
Median (range) (0.984-4.122) (0.984-3.854) (1.365-4.122) (1.566-3.610)
Eosinophils (109/L) 0.130 0.117 0.185 0.078 0.21†
Median (range) (0.028-0.319) (0.028-0.246) (0.130-0.319) (0.057-0.200)
Neutrophils (109/L) 3.207 3.418 3.393 2.944 0.51†
Median (range) (1.404-6.232) (1.582-6.232) (2.500-4.940) (1.404-5.700)
Monocytes (109/L) 0.410 0.378 0.434 0.480 0.24†
Median (range) (0.042-1.140) (0.042-0.711) (0.195-0.919) (0.235-1.140)
Basophils (109/L) 0.0 0.0 0.0 0.0 0.99†
Median (range) (0.0-0.2) (0.0-0.2) (0.0-0.1) (0.0-0.2)
The data are at sample collection. P-values are calculated by Cuzick’s trend test (†), or Kruskal-Wallis trend test (^).
STAT signaling pathway (NFKBIA, PIK3R5, MAPK14, revealed by mutational signature analysis (Figure 2D), fur-
EP300, MPL, IFNAR1, IL6ST and IL20RA) according to ther supported age-dependent accumulation of somatic
Uniprot identifier. Three genes had more than one vari- variants in CD8+ T cells among HTLV-2 positive subjects.
ant: VWF (3 mutations), SMAD7 and MXRA5 (2 muta- No association between peripheral blood counts and num-
tions each) (median VAF: 7%, 2% and 4%, respectively). ber of variants was observed (Table 1).
Subject 13 who harbored a STAT3 Y640F mutation also In summary, our results highlight the presence of STAT3
had mutations in KMT2D, NFKBIA, PIK3R5, CTCF, and mutations in CD8+ T cells of healthy blood donors harbor-
VWF. In this subject with multiple somatic mutations, ing HTLV-2 without clinical history of lymphoproliferative
STAT3 had the highest VAF (16.2%, Online disease. HTLV-2 positive subjects with STAT3 mutations
Supplementary Table S1), suggesting that other variants showed variable clonal expansion of CD8+ T cells, sug-
may be subclonal. Subject 12 with a STAT3 N647I muta- gesting that HTLV-2 infection may promote lymphoprolif-
tion also harbored variants in INPP5D, FSCN1 and eration and STAT3 mutagenesis outside the clinical context
PLA2R1, CD248, and P4HTM. Subject 11 with STAT3 of T-LGLL. We identified additional mutations in CD8+ T
Y657_K658insY harbored variants in MTA1, NCOR2, cells of HTLV-2 positive subjects in genes involved in JAK-
BCL6, ADCY8, RPS6KA3. Subject 27 with a STAT3 STAT signaling, immune regulation and lymphoprolifera-
D661Y mutation had no additional variants. The overall tion. In addition to T-LGLL, somatic STAT3 mutations
number of mutations was higher in HTLV-2 positive have been detected in LGLL associated diseases such as
blood donors harboring STAT3 mutations (median =6) aplastic anemia, hypoplastic myelodysplastic syndrome
compared to HTLV-2 positive blood donors without and Felty’s syndrome, and in some patients with multiple
STAT3 mutations (median =1; P=0.061; Mann–Whitney sclerosis and rheumatoid arthritis.15 Although CD8+ T cell
U test), although the difference was not statistically sig- expansions can also be detected in other diseases such as
nificant. The complete list of variants and VAF can be in rheumatoid arthritis, somatic STAT3 mutations are not
found in the Online Supplementary Table S1. common in these conditions.8 STAT3 mutations and a his-
Somatic mutations can accumulate in tissues with tory of HTLV-2 infection may highlight a subset of blood
aging.14 Accordingly, the total number of coding variants donors who are at risk of subsequent diagnosis of lympho-
was associated with older age among HTLV-2 positive proliferative diseases. However, no clinical follow-up
blood donors (no variants, median age 51 years; 1-2 vari- information is available from our study participants, and
ants, 53 years; 3 or more variants, 59 years; P=0.04) (Table future studies are needed to elucidate whether HTLV-2
1; Figure 2B). The most frequent single nucleotide transi- positive subjects carrying STAT3 and other mutations are
tion was C>T involved in age-associated mutational signa- at increased risk of T-LGLL or other lymphoproliferative
ture 1 (Figure 2C). The high prevalence of signature 1, diseases.

A B
Figure 2. Somatic mutations identified in CD8+ T cells of HTLV-2 positive subjects. (A) Mutation landscape in CD8+ T cells of human T-cell leukemia virus type
2 (HTLV-2) positive subjects (n=28). Coding variants identified by immunogene panel sequencing are presented, together with age, sex, number of mutations
and mutation types. * (white asterisk), STAT3 variants detected in the deep amplicon sequencing but filtered out in the immunogene panel sequencing due to
lower coverage but confirmed with visual inspection with the Integrative Genomics Viewer (Broad Institute, USA). The average expression of mutated genes in
healthy controls are shown on the right, presented as counts per millions reads mapped (CPM) with mean ± standard deviation (n =5). The complete list of vari-
ants and variant allele frequencies (VAF) can be found in the Online Supplementary Table S1. (B) Correlation plot of age vs. mutation number in HTLV-2 positive
blood donors. (C) Percentages of somatic base substitutions and indel identified by immunogene panel sequencing in CD8+ T cells of HTLV-2 positive blood
donors. (D) Normalized weights of COSMIC signatures contributions. Signature 1 (weight: 0.362) was highly related signature.
Daehong Kim,1,2 Mikko Myllymäki,1,2 Matti Kankainen1-4 Precision Cancer Medicine Flagship, University of Helsinki and
Timo Järvinen,1,2 Giljun Park,1,2 Roberta Bruhn,5,6 Edward L. Helsinki University Hospital, Helsinki, Finland; 5Vitalant Research
Murphy5,6 and Satu Mustjoki1,2,4 Institute, San Francisco, CA, USA and 6University of California San
1
Hematology Research Unit Helsinki, Helsinki University Hospital Francisco, San Francisco, USA
Comprehensive Cancer Center, Helsinki, Finland; 2Translational Correspondence:
Immunology Research Program and Department of Clinical Chemistry
SATU MUSTJOKI - satu.mustjoki@helsinki.fi
and Hematology, University of Helsinki, Helsinki, Finland;
3
Department of Medical and Clinical Genetics, University of Helsinki doi:10.3324/haematol.2021.279140
and Helsinki University Hospital, Helsinki, Finland; 4iCAN Digital Received: May 10, 2021.

Accepted: October 20, 2021. 2. Koskela HL, Eldfors S, Ellonen P, et al. Somatic STAT3 mutations
in large granular lymphocytic leukemia. N Engl J Med.
Pre-published: October 28, 2021. 2012;366(20):1905-1913.
Disclosures: SM has received honoraria and research funding 3. Melamed A, Witkover AD, Laydon DJ, et al. Clonality of HTLV-2
from Novartis, Pfizer and Bristol-Myers Squibb (not related to this in natural infection. PLoS Pathog. 2014;10(3):e1004006.
study). MM has received honoraria from Celgene and Sanofi. All 4. Kataoka K, Nagata Y, Kitanaka A, et al. Integrated molecular
other authors declare have no conflicts of interest to disclose. analysis of adult T cell leukemia/lymphoma. Nat Genet.
2015;47(11):1304-1315.
Contributions: DK, MM and SM designed the study and 5. Loughran TP, Jr., Coyle T, Sherman MP, et al. Detection of human
experiments; DK performed sample preparation and analyzed T-cell leukemia/lymphoma virus, type II, in a patient with large
data; MK performed immunogene panel sequencing analysis; TJ granular lymphocyte leukemia. Blood. 1992;80(5):1116-1119.
performed coverage analysis of immunogene panel sequencing; GP 6. Thomas A, Perzova R, Abbott L, et al. LGL leukemia and HTLV.
performed sample preparation; RB and ELM provided the clinical AIDS Res Hum Retroviruses. 2010;26(1):33-40.
7. Lee TH, Chafets DM, Busch MP, Murphy EL. Quantitation of
data and study materials; DK, MM and SM wrote the manuscript
HTLV-I and II proviral load using real-time quantitative PCR with
with the input of other authors; SM supervised the study. All SYBR Green chemistry. J Clin Virol. 2004;31(4):275-282.
authors read and approved the final manuscript. 8. Savola P, Kelkka T, Rajala HL, et al. Somatic mutations in clonally
Acknowledgment: Amplicon sequencing and Immunogene panel expanded cytotoxic T lymphocytes in patients with newly diag-
sequencing were performed at the Institute for Molecular Medicine nosed rheumatoid arthritis. Nat Commun. 2017;8:15869.
Finland (FIMM) Technology Center, which is supported by 9. Kim D, Park G, Huuhtanen J, et al. Somatic mTOR mutation in
clonally expanded T lymphocytes associated with chronic graft
Biocenter Finland. The authors also acknowledge IT Center for
versus host disease. Nat Commun. 2020;11(1):2246.
Science Ltd for computational resources. 10. Kwaan N, Lee TH, Chafets DM, et al. Long-term variations in
Funding: the research was funded by European Research Council human T lymphotropic virus (HTLV)-I and HTLV-II proviral loads
(M-IMM project), ERAPerMed consortium JAKSTAT-TARGET, and association with clinical data. J Infect Dis. 2006;194(11):1557-
Academy of Finland, Finnish special governmental subsidy for 1564.
health sciences, research and training, Sigrid Juselius Foundation, 11. Savola P, Martelius T, Kankainen M, et al. Somatic mutations and
T-cell clonality in patients with immunodeficiency.
Helsinki Institute of Life Sciences Fellow funding, and Cancer Haematologica. 2020;105(12):2757-2768.
Foundation Finland. The HOST study was supported by the US 12. Robinson JT, Thorvaldsdottir H, Winckler W, et al. Integrative
National Heart, Lung and Blood Institute research grant 2R01-HL- genomics viewer. Nat Biotechnol. 2011;29(1):24-26.
62235 and Signe and Ane Gyllenberg Fundation. 13. Coppe A, Andersson EI, Binatti A, et al. Genomic landscape char-
Data sharing statement: all data are available in the Online acterization of large granular lymphocyte leukemia with a sys-
tems genetics approach. Leukemia. 2017;31(5):1243-1246.
Supplementary Appendix.
14. Blokzijl F, de Ligt J, Jager M, et al. Tissue-specific mutation accu-
References mulation in human adult stem cells during life. Nature.
2016;538(7624):260-264.
1. Lamy T, Moignet A, Loughran TP, Jr. LGL leukemia: from patho- 15. Mustjoki S, Young NS. Somatic mutations in "benign" disease. N
genesis to treatment. Blood. 2017;129(9):1082-1094. Engl J Med. 2021;384(21):2039-2052.

center study was to formally investigate the incidence of

Incidence and outcome of SARS-CoV-2 infection in SARS-CoV-2 infection, as well as the characteristics and
patients with monoclonal gammopathy of undeter- the clinical outcome of COVID-19 in a larger cohort of
mined significance: a case-control study MGUS patients. The study was conducted within the
context of the clinicaltrials gov. Identifier:
In March 2020, "Coronavirus Disease 2019" (COVID- NCT04352556.
19) outbreak due to SARS-CoV-2 was declared a pan- Between March 1, 2020 and April 30, 2021, we col-
demic by the World Health Organization. Clinical mani- lected, among 1.454 MGUS patients followed at our cen-
festations of COVID-19 are variable, ranging from com- ter, clinical data from 91 patients with SARS-CoV2 infec-
plete absence of symptoms, to severe pneumonia, multi- tion, diagnosed by RT-PCR on nasopharyngeal swabs.
organ failure, and death. Main risk factors for poor out- Data were mainly extracted from “GIAVA-COVID-19”, a
come of COVID-19 are advanced age and comorbidities, regional platform where authorized medical health
conditions that often recur in patients with monoclonal workers can view the results of the nasopharyngeal
gammopathies. In this setting, several papers have swabs for SARS-CoV-2 performed, along with other
reported more frequent and severe COVID-19, as well as information. In MGUS patients a review of medical
higher fatality rate, in patients with multiple myeloma records was also carried out. Clinical data collected
(MM), particularly in those older than 60 years, with regarded age, cardiovascular, pulmonary, diabetic and
high risk, active/progressive disease and/or renal failure, neoplastic comorbidities, presence of symptoms (in
respect to general population.1-3 By contrast, very few detail: fever or chills, cough, shortness of breath or diffi-
data are available about patients with monoclonal gam- culty breathing, fatigue, muscle or body aches,
mopathies of undetermined significance (MGUS).4 In our headache, loss of taste or smell, sore throat, congestion
retrospective study, we investigated the incidence of or runny nose, nausea or vomiting, diarrhea), hospital-
SARS-CoV-2 infection and COVID-19 outcomes in ization, hospitalization in ICU, and outcome
MGUS patients. Overall, we found that subjects with (alive/dead). Patients with monoclonal gammopathies of
MGUS neither have an increased risk of contracting clinical (renal, dermatological or neurological) signifi-
SARS-CoV-2, nor show poorer COVID-19 outcomes cance (MGCS) were excluded.
compared to controls. The characteristics of COVID-19 in MGUS patients
Patients with MGUS are frequently asymptomatic and were compared with those of 182 age- and sex-matched
diagnosed by chance; therefore, differently from MM, normal controls infected by SARS-CoV-2. Furthermore,
one would not expect an increased incidence of SARS- the incidence of SARS-CoV-2 infection in MGUS
CoV-2 infection or a worse outcome of COVID-19 patients was compared to that of the entire Apulian pop-
respect to the normal population in these subjects. ulation (Apulia is a region of Southern Italy with about 4
Notwithstanding, according to well recognized risk fac- million inhabitants). Wilcoxon test, chi-square tests and
tors, they may have different risks of developing MM, as multivariate logistic regression were applied, as appro-
well as clinical findings, including older age, presence of priate, by using STATA software MP17.
medical comorbidities, and impaired humoral/cellular Table 1 summarizes the main characteristics of SARS-
immunity, which could still play a role when assessing CoV-2 infected MGUS patients and controls. The two
their risk during the COVID-19 pandemic.5,6 Notably, in groups were comparable for age, sex, and presence of
epidemiological studies, people with MGUS were comorbidities. Among the MGUS group, 68 patients
shown to have an increased risk of developing both showed a non-IgM subtype and nine an IgM subtype:
venous and arterial thrombosis, bacterial and viral infec- this information was missing in five patients. Nine
tions, as well as an excess mortality risk due to bacterial patients had a double M-component. Immunoparesis (at
infections as compared to age and sex-matched healthy least one uninvolved immunoglobulin below reference
controls.7,8 On this basis, the presence of MGUS could levels) was present in 19% of 68 evaluable patients.
possibly increase susceptibility to SARS-CoV-2 infection Regarding MGUS risk-stratification according to Mayo
and severity of COVID-19, and might theoretically Clinic model, most (94%) of 47 patients with complete
account for the increased mortality due to COVID-19 available dataset scored as low or low/intermediate risk.
observed in the elderly population.9 However, a retro- Sixty-two patients showed the presence of at least one
spective chart review aiming to study the vulnerability co-existing, potentially clinically relevant comorbidity
of 228 patients with MGUS (3 of whom resulted infected (cardiovascular disease 40.6%, diabetes 11%, non-
by SARS-CoV2, with 1 death) and their clinical out- hematological cancer 8.8%, pulmonary disease 6.6%).
comes during the COVID-19 pandemic, concluded that As shown in Table 1, rates of COVID-19-related
there were neither significant differences in the mean symptoms, hospitalization, hospitalization in ICU and
age or survival of the MGUS patients not infected by deaths due to COVID-19 were slightly higher in the
SARS-COV-2 who died before versus after the pandemic MGUS group than in the control group, but these differ-
onset, nor an increase in venous thrombotic events.10 ences were not statistically significant. In MGUS
Furthermore, in a small case series of seven MGUS patients, sex, immunoparesis, presence/number of
patients experiencing COVID-19, 71% were hospital- comorbidities and IgM versus non-IgM isotype did not
ized, but none of the patients required mechanical ven- significantly influence the risk of death, while a statisti-
tilation or ICU (intensive care unit) management.4 cally significant association was observed with older
Patients had an age range between 59 and 92 years and age; importantly, the risk of death was not correlated
all had underlying high-risk comorbidities. One patient with the presence of MGUS (Table 2). Lastly, incidence
with acute kidney injury recovered after hemodialysis. of SARS-CoV2 infection in MGUS patients (91/1.454,
The only death was a male patient with advanced age, 6.2%) was not statistically different from that observed
nursing home residency, multiple comorbidities and ele- in the entire population of the Apulia region
vated D-dimer. This small case series would suggest that (227.761/3.926.931, 5.8%) during the same period
MGUS does not pose additional risks for poor outcome (Table 1).
in COVID-19 patients. Thus, in our study, patients with MGUS, contrarily to
The aim of our observational, retrospective, single what is seen in MM, did not show an increased incidence

of SARS-CoV-2 infection compared to the general popula- there could be unintentional patient selection bias:
tion. Furthermore, although rates of symptoms, hospital- MGUS may be misclassified and, conversely, individuals
ization, hospitalization in ICU and deaths were slightly may be unaware of its presence. Second, not all MGUS
higher than in controls, MGUS did not appear to repre- patients had a complete dataset and some laboratory
sent a risk for a poorer COVID-19 outcome. The only fac- findings were lacking. Finally, long-term outcomes of
tor associated with an increased risk of death was older MGUS and COVID-19 infection should also be explored.
age; however this was likely not related to the presence of Therefore, further data are needed to achieve greater
MGUS, but rather to the well known predictive power of generalizability of our findings.
this clinical parameter for a worse outcome in the general Other questions, particularly the possibility of a sub-
population infected by SARS-CoV-2.9 optimal response in people with MGUS to anti-SARS-
Although, to the best of our knowledge, we conducted CoV-2 vaccine (as observed in MM), possibly due to age
the largest study of SARS-CoV-2 infection in patients and MGUS-related immune dysfunction, will be proba-
with MGUS, several limitations are present in our analy- bly soon addressed by ongoing studies. In this setting,
sis. First, as with any observational retrospective study, preliminary data show that MGUS patients receiving
Table 1. Characteristics of patients with monoclonal gammopathies of undetermined significance versus controls and COVID-19 outcome.
MGUS Controls P-value
Total, n. 91 182
Mean Age, years +/- SD (range) 65.6 +/- 13.3 (29-89) 65.2 +/- 13.4 (29-89) 0.734
Sex, n. (%) 0.796
Male 42 (46.2) 80 (44)
Female 49(53.8) 102 (56)
Comorbidities, n. (%) *
no 29 (31.9) 50 (27.5) 0.684
1 30 (32.9) 59 (32.5) 1.000
2 19 (20.9) 64 (35) 0.125
≥3 13 (14.3) 9 (5) 0.148
Mean number, n. +/- SD (range) 1.3 +/-1.3 (0-5) 1.2 +/- 0.9 (0-3) 0.844
Incidence SARS-CoV-2 infection (%) 6.2 ** 5.8 § 0.454
COVID-19 outcome
Presence of symptoms, n. (%; 95% CI) 54 (59.3; 48.5-69.5) 102 (56.0; 48.5-63.4) 0.604
Hospitalization, n. (%; 95% CI) 19 (20.9; 13.1-30.7) 26 (14.3; 9.5-20.2) 0.166
Hospitalization in ICU, n. (%; 95% CI) 10 (11.0; 5.4-19.3) 16 (8.8; 5.1-13.9) 0.560
Death due to COVID-19, n. (%; 95% CI) 8 (8.8; 3.9-16.6) 10 (5.5; 2.7-10.0) 0.301
MGUS subtype (available in 86 patients), n. (%) NA
IgG 63 (73.2)
IgA 5 (6.0)
IgM 9 (10.4)
Biclonal 8 (9.3)
LC only 1 (1.1)
Immunoparesis (available in 68 patients) NA
0/1/2, n. 55/10/3
≥1 subclass (%) 13/68 (19.1)
MGUS risk (available in 47 patients)°, n. (%) NA
0 Low 22 (46.8)
1 Low-intermediate 22 (46.8)
2 High-intermediate 3 (6.4)
*Comorbidities evaluated included cardiovascular disease, pulmonary disease, diabetes, and non-hematological cancers; ** Incidence among 1,454 MGUS followed at our
Institution; § Incidence among the entire population of the Apulia region (227.761 cases of SARS-CoV-2 infection in 3.926.931 inhabitants); ° Rajkumar et al. Blood
2005;106(3):812-7. MGUS: monoclonal gammopathies of undetermined significance; NA: not applicable; ICU: intensive care unit; LC: light chain; CI: confidence interval; SD:
standard deviation
Table 2. Risk for death in COVID-19 monoclonal gammopathies of undetermined significance patients, adjusted for different clinical param-
eters.
Determinants aOR 95% CI P-value
Group (MGUS vs. controls) 0.76 0.20 - 3.00 0.700
Age (years) 0.88 0.78 - 0.99 0.035
Gender (Male vs. Female) 0.27 0.04 - 1.91 0.193
Presence of comorbidities (Yes vs. No) 2.16 0.26 - 17.7 0.474
Number of comorbidities (1-2 vs. ≥3) 2.01 0.11 - 37.8 0.641
IgM vs non-IgM 0.28 0.31 - 2.56 0.262
Immunoparesis (0 vs. ≥1) 0.45 0.05 - 3.98 0.473
MGUS: monoclonal gammopathies of undetermined significance; aOR: adjusted odds ratio; CI: confidence interval; IgM: immunoglobulin M.

anti-SARS-CoV-2 vaccines could have a better antibody factors of inpatient mortality. Blood Cancer J. 2020;10(10):103.
response than those with MM.11 Anyway, in patients 3. Engelhardt M, Shoumariyeh K, Rösner A, et al. Clinical characteris-
with MGUS, as in those with MM, full vaccination tics and outcome of multiple myeloma patients with concomitant
against SARS-CoV-2 should be strongly recommended COVID-19 at Comprehensive Cancer Centers in Germany.
Haematologica. 2020;105(12):2872-2878.
to reduce the possibility of infection rate and severity of
4. Gonzalez-Lugo JD, Bachier-Rodriguez L, Goldfinger M, et al. A case
the illness.12-14 Last, but not least, the psychosocial
series of monoclonal gammopathy of undetermined significance
impact of the pandemic on MGUS patients and their and COVID-19. Br J Haematol. 2020;190(3):e130-e133.
management in the long term follow-up also would war- 5. Kyle RA, Larson DR, Therneau TM, et al. Long-term follow-up of
rant to be further and specifically investigated.15 monoclonal gammopathy of undetermined significance. N Engl J
Med. 2018;378(3):241-249.
Nicola Sgherza,1 Paola Curci,1 Rita Rizzi,1,2 Vanda 6. van de Donk NW, Palumbo A, Johnsen HE, et al. The clinical rele-
Strafella,2 Daniela Di Gennaro,2 Angelantonio Vitucci,1 vance and management of monoclonal gammopathy of undeter-
Antonio Palma,1 Antonella Vita Russo Rossi,1 Francesco mined significance and related disorders: recommendations from
Albano,2 Pasquale Stefanizzi,3 Silvio Tafuri3 and Pellegrino the European Myeloma Network. Haematologica. 2014;99(6):984-
Musto1,2 996.
1 7. Kristinsson SY, Björkholm M, Andersson TM, et al. Patterns of sur-
Hematology and Bone Marrow Transplantation Unit, AOUC
vival and causes of death following a diagnosis of monoclonal gam-
Policlinico di Bari; 2Department of Emergency and Organ mopathy of undetermined significance: a population-based study.
Transplantation, “Aldo Moro” University School of Medicine and Haematologica. 2009;94(12):1714-1720.
3
Department of Biomedical Science and Human Oncology, “Aldo 8. Kristinsson SY,Tang M, Pfeiffer RM, et al. Monoclonal gammopathy
Moro” University School of Medicine, Bari, Italy of undetermined significance and risk of infections: a population-
Correspondence: based study. Haematologica. 2012;97(6):854-858.
9. G. Onder, G. Rezza, S. Brusaferro. Case-fatality rate and character-
PELLEGRINO MUSTO- pellegrino.musto@uniba.it istics of patients dying in relation to COVID-19 in Italy. JAMA.
doi:10.3324/haematol.2021.279895 2020;323(18):1775-1776.
Received: September 1, 2021. 10. Lee H, Tay J, Street L, Duggan P, Jiménez-Zepeda VH. Monoclonal
gammopathy of undetermined significance clinic during the coron-
Accepted: October 28, 2021. avirus disease-19 pandemic: caring for the vulnerable in an academic
Pre-published: November 4, 2021. medical center. Rev Invest Clin. 2021;73(4):259-264.
11. Terpos E, Gavriatopoulou M, Ntanasis-Stathopoulos I, et al. The
Disclosures: no conflicts of interest to disclose.
neutralizing antibody response post COVID-19 vaccination in
Contributions: PM and NS conceived and led the project; NS patients with myeloma is highly dependent on the type of anti-
conducted database building, extraction and coding; NS, PM, PS myeloma treatment. Blood Cancer J. 2021;11(8):138.
and ST queried and analyzed the data; PM and NS wrote the 12. https://cms.cws.net/content/beta.myelomasociety.org/files/PM%2
main manuscript text and created all tables. All authors made a 0COVID%20vaccination%20in%20MM%20guidelines%20The%
substantial intellectual contribution to the study, interpreted the data, 20Final.pdf
discussed the results and reviewed, edited and approved the final 13. Ludwig H, Meckl A, Engelhardt M. Compliance with vaccination
version of the manuscript. recommendations among patients with multiple myeloma: a real
world experience. Hemasphere. 2021;5(7):e597.
14. Gavriatopoulou M, Ntanasis-Stathopoulos I, Korompoki E, Terpos
References E, Dimopoulos MA. SARS-CoV-2 vaccines in patients with multiple
myeloma. Hemasphere. 2021;5(3):e547.
1. Chari A, Samur MK, Martinez-Lopez J, et al. Clinical features asso-
ciated with COVID-19 outcome in multiple myeloma: first results 15. Quinn SJ, Anderson LA, Lohfeld L, McShane CM. The psychosocial
from the International Myeloma Society data set. Blood. impact of the COVID-19 pandemic on patients with monoclonal
2020;136(26):3033-3040. gammopathy of undermined significance, smouldering and active
2. Martínez-López J, Mateos MV, Encinas C, et al. Multiple myeloma myeloma: findings from an international survey. Br J Haematol.
and SARS-CoV-2 infection: clinical characteristics and prognostic 2021;194(2):294-297.

normal and tumor cell survival and death.2 MK express

Interferon a-induced SAMHD1 regulates human the IFN-1 receptor that signals through Janus
cultured megakaryocyte apoptosis and proplatelet kinases/signal transducer and activator of transcription
formation proteins (JAK/STAT) pathway in response to IFN-1
cytokines.3 IFNa, an IFN-1 cytokine, has been effectively
Megakaryocyte (MK) growth, differentiation and mat-
utilized in the treatment of myeloproliferative neo-
uration are required for thrombopoiesis and platelet pro-
plasms and viral hepatitis. Thrombocytopenia is a com-
duction. Most studies of megakaryocytopoiesis have uti-
lized in vitro culture systems expected to model a healthy mon adverse effect of IFNa therapy that can require dose
human condition. However, consistent with the ability reduction. Although there are inconsistent reports
of MK to respond to inflammatory mediators, chronic regarding IFNa suppression of colony forming units of
inflammatory conditions often induce thrombocytosis, megakaryocyte progenitors (CFU-MK) in cultures of
whereas acute inflammation can result in thrombocy- human CD34+ cells, there are consistent findings to sup-
topenia. Furthermore, there is an increasing awareness port a mechanism of decreased platelet production
of the role MK play in innate and adaptive immunity.1 rather than reduced platelet life span.4-7 However, the
Type 1 interferons (IFN-1), including IFNa, IFNb and molecular mechanisms regulating IFNa-induced
INFω are a family of cytokines that bind to the IFN-1 decrease in platelet production and peripheral blood
receptor and trigger transcription of diverse genes. IFN- thrombocytopenia are poorly understood. The major
inducible genes regulate resistance to viral infections, conclusions in this report are (i) using genome-wide gene
enhance innate and adaptive immunity, and modulate expression profiling we show that IFNa upregulates the
A B C
D E F
Figure 1. Interferona regulates platelet production. (A to E) CD34+ hematopoietic stem cells and progenitor cells were isolated by immunomagnetic separation
from human umbilical cord blood. Cells were cultured for 13 days in stem cells expansion media supplemented with thrombopoietin that promotes megakary-
ocyte (MK) differentiation. 1,000 units/milliliter (U/mL) of human interferon α (INFα) and phosphate-buffered saline (PBS) (used as a negative control) was
added at day 9 and further incubated until day 13. All assays mentioned in panels A to E were performed at day 13. (A) MK proplatelet formation (PPF) was
counted blinded as to the IFNa treatment. At least 200 cells were counted per culture (n=4). (B) Representative images of day 13 cultured MK treated with
PBS or IFNa. Treated MK were plated on fibrinogen on day 12 overnight, and fixed with 4% paraformaldehyde, stained with Alexa Fluor 488 Phalloidin (green)
and a nuclear stain, DAPI (blue). Images were taken by a confocal microscope at 40X oil objective lens. (C) Platelet-like particles (PLP) were collected from IFNa
or PBS-treated MK cultures and stained with APC labeled anti-CD41a antibody at 37°C for 10 minute and measured by flow cytometry (n=3). PLP were gated
based on human peripheral blood platelets. (D) MK were stained with APC-labeled CD41a and PE-labeled CD42a antibodies, and CD41a+ CD42a+ MK (a marker
for MK maturation) were assessed by flow cytometry (n=3). (E) Cultured MK treated with IFNa or PBS were stained with APC-labeled CD41a and propidium
iodide, and ploidy was assessed by flow cytometry (n=3). The quantification of the ploidy distribution is shown on the y-axis by calculating the percentage of
cells with 2n, 4n, 8n and 16n. Apoptotic population were gated out. Statistical significance was determined by two-tailed paired t-test (A to E). Error bars repre-
sent mean ± standard error of mean. (F) 25,000 units of murine IFNa or PBS (negative control) were administered intraperitoneally in wild-type mice for con-
secutive three days (n=5 per group). On day 4, mice blood was harvested by cardiac puncture and platelet count was measured by Hemavet. Statistical signifi-
cance was determined by two-tailed unpaired t-test with Welch’s correction. Error bars represent mean ± standard error of mean.

expression of MK sterile a motif (SAM) and histidine- (Figure 2A). Figure 2B and C shows that IFNa treatment
aspartate (HD) domain containing deoxynucleoside of cultured MK greatly increased the abundance of
triphosphate triphophohydrolase 1 (SAMHD1) and (ii) SAMHD1 mRNA and protein (n=3 independent biologi-
SAMHD1 expression inhibits cultured human MK pro- cal replicates). SAMHD1 is a hydrolase, the activated
platelet formation (PPF) and promotes apoptosis. This is form of which degrades the intracellular pool of
the first identification of SAMHD1 in human MK and deoxynucleoside triphosphates (dNTPase) into deoxynu-
report of a dNTP hydrolase regulating platelet produc- cleosides and inorganic triphosphates, and is known to
tion. restrict viral replication of the human immunodeficiency
In order to pursue studies on the effects of inflamma- virus type-1.8 In addition to viral restriction, SAMHD1 is
tion on megakaryocytopoiesis, we used CD34+ required for cellular functions including replication fork
hematopoietic stem cells derived from human umbilical progression, cell proliferation, apoptosis and DNA dam-
vein cord blood. IFNa significantly decreased day 13 PPF age repair.9 IFNa stimulation induces SAMHD1 expres-
and platelet-like particles (Figure 1A to C), but did not sion in human monocytes,10 astrocytes, microglia,11
affect the percentages of MK or polyploidy (Figure 1D to HEK293T and HeLa cells,12 but there are no prior reports
E). Importantly, we also showed that exogenous IFNa of SAMHD1 expression and/or function in MK or
induces thrombocytopenia in wild-type mice (Figure 1F), platelets.
consistent with studies in immunodeficient mice.4 In Platelet RNA And eXpression 1 (PRAX1) data13
order to begin to understand how IFNa regulates late- demonstrated that SAMHD1 transcript levels are nega-
stage megakaryopoiesis and platelet production, we tively associated with platelet count in healthy human
used an unbiased, transcriptome-wide approach and per- subjects (Figure 3A), suggesting a possible inhibitory role
formed RNA sequencing (RNA-seq) on CD61-purified, of SAMHD1 in platelet production. Since SAMHD1
day 13 cultured MK stimulated with IFNa. Our analyses modulates the intracellular levels of dNTP, we hypothe-
identified 201 transcripts that were differentially sized that an increase in the abundance SAMHD1 upon
expressed at a nominal significance threshold (P<0.05). IFNa stimulation leads to decreased MK proliferation,
Adjusting for multiple comparisons and setting a false maturation and DNA synthesis (MK polyploidy).
discovery rate (FDR) threshold of <0.05, we found that However, deletion of SAMHD1 by CRISPR/Cas9 gene
66 of the 201 transcripts were upregulated by IFNa editing in cultures promoting unilineage MK differentia-
(Online Supplementary Table S1). Increased mRNA tion (Figure 3B) did not affect MK maturation (Online
expression in response to IFNa was validated by real- Supplementary Figure 2A and B) or ploidy (Online
time polymerase chain reaction (PCR) analysis for all five Supplementary Figure 2C). This suggests SAMHD1 effects
genes tested (SAMHD1, PHF11, ISG20, IFITM3 and thrombopoiesis rather than megakaryocytopoiesis.
TAP2) (Online Supplementary Figure S1). Gene ontology Similar to Figure 1, IFNa stimulation caused a significant
analysis indicated that the differentially expressed genes decrease in MK PPF MK without CRISPR modification
were associated with the type 1 interferon signaling (Figure 3C, first 2 bars). The effect of IFNa on PPF was
pathway, defense response to virus, and negative regula- abolished when SAMHD1 was deleted (Figure 3C, sec-
tion of viral genome replication. Subsequent studies ond 2 bars). Lastly, IFNa is well-established as pro-apop-
focused on SAMHD1, whose abundance increased more totic.2 MK must restrain apoptosis to survive and
than 16-fold with IFNa induction (FDR=2.0x10-18) progress safely through PPF and platelet shedding.14,15
A B C
Figure 2. Interferon α upregulates. dNTP hydrolase enzyme SAMHD1 in cultured megakaryocytes. CD34+ hematopoietic stem cells and progenitor cells were
isolated by immunomagnetic separation from human umbilical cord blood. Cells were cultured in megakaryocyte (MK) promoting conditions by supplementation
with thrombopoietin for 13 days. 1,000 units per milliliter (U/mL) of interferon α (INFα) and phosphate-buffered saline (PBS) (negative control) was added in 3
independent cultures at day 13 for 8 hours. (Dose and time course studies were performed in Meg-01 to select optimal concentrations and times, data not
shown). CD61+ MK were separated by immunomagnetic beads and total RNA was obtained by mirVana kit. RNA samples with RIN score of >8 were used for
RNA sequencing. The library preparation was done using TruSeq standard total RNA sample preparation kit with RNA depletion (Illumina Inc.). RNA sequensing
was performed using NextSeq500 instrument with read length of 50 basepair single-end read. On average 68.9 million reads were obtained from each sample
and genome mapping was on average 94.5% for all samples. (A) Volcano plot shows the relationship between the P-values and the fold change in normalized
expression for IFNa or PBS treated cultured human MK. Differentially expressed genes (false discovery rate [FDR] <0.05) on IFNa treatment was plotted.
SAMHD1 is shown in red. (B) Real-time polymerase chain reaction analysis validates the upregulation of SAMHD1 in IFNa-treated CD61+ MK compared to PBS
control. Actin was used as housekeeping gene. Log2 fold change of SAMHD1 mRNA levels is plotted for IFNa vs. PBS control (n=3). Statistical significance was
determined by one sample t-test. (C) Representative immunoblot shows increase in SAMHD1 protein in IFNa-treated MK compared to PBS control. Actin is used
as loading control.

A B
C D
Figure 3. Interferon α-induced SAMHD1 expression regulates human megakaryocyte (MK) proplatelet formation and MK apoptosis. (A) Plot of Pearson corre-
lation (R) between platelet count and SAMHD1 mRNA levels in 154 healthy donors. Dotted lines represent 95% confidence intervals. (B to D) CRISPR/Cas9
knock-down of SAMHD1 in CD34+ derived human umbilical cord cells were performed at day 3. Cells were cultured in MK differentiation promoting conditions,
and treated with 1,000 U/mL interferon α (INFα) and phosphate-buffered saline (PBS) control. On day 13, MK treated with IFNa or PBS were assayed as men-
tioned below. (B) Representative immunoblot of SAMHD1 after CRISPR-Cas9 knock-down (denoted as “SAMHD1 cr” throughout the figure) in day 13 human
CD34+ derived cultured MK with or without IFNa. Guide RNA not targeting known genes are used as negative control with or without IFNa (“neg cr” throughout
the figure). Fold changes of densitometric quantification of SAMHD1 immunoblots, normalized to actin is shown on right (n=5). (C) The percentage of proplatelet
formation (PPF) MK, scored blinded, for SAMHD1 cr vs neg cr, with or without IFNa is shown (n=6). (D) The mean fluorescence intensity (MFI) of annexin V bind-
ing (a marker of apoptosis-induced phosphatidylserine expression) on IFNa stimulation in SAMHD1 cr vs. neg cr MK, with or without IFNa is plotted (n=5).
Statistical significance was determined by paired t-test (B to D). Error bars represent mean ± standard error of mean.
Although viral infections induce MK apoptosis,15 we are stage platelet production by MK. Prior work has shown
not aware of in vitro studies assessing IFNa-induced MK that expression of three candidate MK transcription fac-
apoptosis. Under the culture conditions described in tors is inhibited by IFNa4, and perhaps SAMHD1 is also
Figure 3C, we assessed annexin V binding as a measure regulated at a transcriptional level in MK. Post-transcrip-
of MK apoptosis, and observed a significant increase in tional mechanisms may also be at play, since the enzy-
response to IFNa stimulation (Figure 3D, first 2 bars). matic ability of SAMHD1 to maintain dNTP homeosta-
Next, because SAMHD1 also promotes apoptosis,9 we sis in other cells requires protein phosphorylation.
tested the effects of SAMHD1 deletion on IFNa-induced Future studies in MK will be need to address these
MK apoptosis. Deletion of SAMHD1 caused a modest issues.
reduction in annexin V binding (Figure 3D, bar 1 vs. bar
3), and significantly reduced IFNa-induced MK annexin Seema Bhatlekar,1 Shancy Jacob,1 Emilia Tugolukova,1
V binding (Figure 3D, compare bars 2 and 4), supporting Bhanu K. Manne,1 Yasuhiro Kosaka,1 Phillipe Loher,2 Ryan
a role for SAMHD1 as a mediator of IFNa-induced MK M. O’Connell,3 Vicente Planelles,4 Matthew T. Rondina,1,5
apoptosis. Isidore Rigoutsos5 and Paul F. Bray1,6
1
In summary, our study indicates that IFNa leads to Program in Molecular Medicine and Department of Internal
reduced platelet production and thrombocytopenia Medicine, University of Utah, Salt Lake City, UT; 2Computational
through apoptosis, and that IFNa-induced SAMHD1 is Medicine Center, Thomas Jefferson University, Philadelphia, PA;
3
at least partially responsible for these effects on late- Division of Microbiology and Immunology, Department of Pathology,

and Huntsman Cancer Institute, University of Utah Health Sciences 2. Borden EC, Sen GC, Uze G, et al. Interferons at age 50: past, cur-
Center, University of Utah, Salt Lake City, UT; 4Division of rent and future impact on biomedicine. Nat Rev Drug Discov.
2007;6(12):975-990.
Microbiology & Immunology, Department of Pathology, University of 3. Negrotto S, De Giusti CJ, Lapponi MJ, et al. Expression and func-
Utah, Salt Lake City, UT; 5Department of Pathology, University of tionality of type I interferon receptor in the megakaryocytic line-
Utah, Salt Lake City, UT and the George E. Wahlen VAMC age. J Thromb Haemost. 2011;9(12):2477-2485.
Department of Medicine and George E. Wahlen VAMC GRECC, 4. Yamane A, Nakamura T, Suzuki H, Ito M, Ohnishi Y, Ikeda Y,
Salt Lake City, UT and 6Division of Hematology and Hematologic Miyakawa Y. Interferon-alpha 2b-induced thrombocytopenia is
Malignancies, Department of Internal Medicine, University of Utah, caused by inhibition of platelet production but not proliferation
and endomitosis in human megakaryocytes. Blood. 2008;112(3):
Salt Lake City, UT, USA. 542-550.
Correspondence: 5. Ganser A, Carlo-Stella C, Greher J, Volkers B, Hoelzer D. Effect of
recombinant interferons alpha and gamma on human bone mar-
PAUL F. BRAY- Paul.bray@hsc.utah.edu row-derived megakaryocytic progenitor cells. Blood. 1987;70(4):
doi:10.3324/haematol.2021.279864 1173-1179.
6. Mazur EM, Richtsmeier WJ, South K. Alpha-interferon: differen-
Received: August 24, 2021 tial suppression of colony growth from human erythroid,
Accepted: October 28, 2021. myeloid, and megakaryocytic hematopoietic progenitor cells. J
Interferon Res. 1986;6(3):199-206.
Pre-published: November 11, 2021. 7. Wadenvik H, Kutti J, Ridell B, et al. The effect of alpha-interferon
Disclosures: no conflicts of interest to disclose. on bone marrow megakaryocytes and platelet production rate in
essential thrombocythemia. Blood. 1991;77(10):2103-2108.
Contributions: SB, SJ, ET, BK, YK performed research; PL, IR 8. Laguette N, Sobhian B, Casartelli N, et al. SAMHD1 is the den-
performed RNA-Seq analysis; SB, PB wrote the manuscript; RO, dritic- and myeloid-cell-specific HIV-1 restriction factor counter-
VP, MR, IR, PF provided scientific expertise. acted by Vpx. Nature. 2011;474(7353):654-657.
9. Coggins SA, Mahboubi B, Schinazi RF, Kim B. SAMHD1 functions
Funding: this study was supported by grants from the National
and human diseases. Viruses. 2020;12(4):382.
Institutes of Health National Heart, Lung and Blood Institute 10. Berger A, Sommer AF, Zwarg J, et al. SAMHD1-deficient CD14+
(HL116713, HL142804, HL145237, and HL130541), the cells from individuals with Aicardi-Goutieres syndrome are high-
National Institute of Allergy, Immunology and Infectious Diseases ly susceptible to HIV-1 infection. PLoS Pathog. 2011;7(12):
(AI143567), and the Division of Hematology and Hematologic e1002425.
Malignancies at the University of Utah. This work was also sup- 11. Jin C, Peng X, Liu F, et al. Interferon-induced sterile alpha motif
and histidine/aspartic acid domain-containing protein 1 expres-
ported by Merit Review Award Number I01 CX001696 from the sion in astrocytes and microglia is mediated by microRNA-181a.
United States (U.S.) Department of Veterans Affairs Clinical AIDS. 2016;30(13):2053-2064.
Sciences R&D (CSRD). This material is the result of work sup- 12. St Gelais C, de Silva S, Amie SM, et al. SAMHD1 restricts HIV-1
ported with resources at the George E. Wahlen VA Medical Center, infection in dendritic cells (DCs) by dNTP depletion, but its
Salt Lake City, Utah. The contents do not represent the views of expression in DCs and primary CD4+ T-lymphocytes cannot be
the U.S. Department of Veterans Affairs or the U.S. Government. upregulated by interferons. Retrovirology. 2012;9:105.
13. Edelstein LC, Simon LM, Montoya RT, et al. Racial differences in
The authors thank the University of Utah Flow Cytometry Facility human platelet PAR4 reactivity reflect expression of PCTP and
in addition to the National Cancer Institute through Award Number miR-376c. Nat Med. 2013;19(12):1609-1616.
5P30CA042014-24. 14. Bhatlekar S, Basak I, Edelstein LC, et al. Anti-apoptotic BCL2L2
increases megakaryocyte proplatelet formation in cultures of
human cord blood. Haematologica. 2019;104(10):2075-2083.
References 15. Josefsson EC, James C, Henley KJ, et al. Megakaryocytes possess
a functional intrinsic apoptosis pathway that must be restrained
1. Cunin P, Nigrovic PA. Megakaryocytes as immune cells. J Leukoc to survive and produce platelets. J Exp Med. 2011;208(10):2017-
Biol. 2019;105(6):1111-1121. 2031.

transcript (4/56 [7%]) and one was found in a patient

Detection of ABL1 kinase domain mutations in with M-bcr (1/35 [3%]). The mutations are present in
therapy-naïve BCR-ABL1-positive acute lym- subclones (<10% of BCR-ABL1-positive cells) in patients
phoblastic leukemia #2, #4 and #5 (Table 1).
For patient #1, the Q252H (cag>cat) mutation had
Mutations in the ABL1 kinase domain are the main expanded to 97% of BCR-ABL1 transcripts at relapse
mechanism of resistance to tyrosine kinase inhibitors after 5 months of treatment with the GMALL elderly
(TKI) in Philadelphia chromosome-positive (Ph+) protocol including imatinib, to which the mutation caus-
leukemia. In acute lymphoblastic leukemia (ALL), a very es resistance (Figure 1). Patients #2 and #3 were both
early acquisition of mutations can be observed and transplanted as part of standard of care at the respective
reports have described that mutations may exist already time. Patient #3 (T315I) had achieved complete remis-
before TKI treatment.1-4 Initial data came from cloning or sion; however, the T315I mutation would be particularly
allele-specific polymerase chain reaction assays,5 which severe because it causes resistance to four TKI (imatinib,
are labor-intensive, do not detect all possible mutations nilotinib, dasatinib and bosutinib).6 Follow-up data for
and especially come with the doubt of detecting amplifi- patients #4 and #5 were not available.
cation artefacts. Despite the first description over 18 Next, we focused on patients who had a molecular fol-
years ago, ABL1 mutation screening is not a standard test low-up of at least 6 months (n=35) for longitudinal
for ALL patients before initiation of TKI treatment. The mutation testing. This excluded patients who died with-
current broad availability of highly sensitive next-gener- in the first 6 months or those from whom no sample was
ation sequencing approaches enables the initial limita- available for BCR-ABL1 expression and ABL1 mutation
tions to be overcome. In this study, we used next-gener- testing. In 19 of 35 cases, mutation analysis was consid-
ation sequencing to screen 91 BCR-ABL1-positive ALL ered relevant during follow-up due to failure of primary
patients for mutations before therapy and identified five therapy or relapse (defined as a BCR-ABL1/ABL1 ratio of
(5.5%) with ABL1 kinase domain mutations at initial at least ~1%) (Online Supplementary Table S1). Mutation
diagnosis. testing was performed between 4 and 71 months after
We studied the diagnostic samples of 39 female and 52 diagnosis (median: 12 months). One or more known
male ALL patients diagnosed between April 2007 and resistance mutations were found in 15 of 19 (79%) Ph+
November 2016. The median age was 61 years (range, ALL cases with relapse or increased BCR-ABL1/ABL1
18-84), and treatment included first- to third-generation expression despite treatment (Online Supplementary Table
TKI (Online Supplementary Table S1). Thirty-five patients S1). This is in line with the previously published ABL1
had the M-bcr (p210) transcript and 56 the m-bcr (p190) mutation frequency of 70-80% for relapsed Ph+ ALL.7
transcript. The study was approved by the Internal Except for the Q252H mutation in patient #1 (Table 1),
Review Board and by the Bavarian Ethics Commitee, the we did not detect any of the other resistance mutations
Bavarian State Medical Association (Bayerische from the relapse sample at the diagnostic time point
Landesärztekammer) with number 05117. The study with a 1% sensitivity cutoff. For comparison, Soverini et
adhered to the tenets of the Declaration of Helsinki. al. used a relapse-to-diagnosis backtracking approach to
RNA for BCR-ABL1 detection and mutation analysis study resistance mutations and found one out of 34
was isolated from bone marrow (n=56) or peripheral patients with a mutation (Y253H) already at the initial
blood (n=35) with standard protocols and cDNA synthe- time point.8
sis was performed using SuperScript™ IV VILO™ While selective pressure of TKI treatment makes the
(Thermo Fisher Scientific, Waltham, MA, USA). Six outgrowth of clones with resistance mutations very
amplicons for sequencing on the MiSeq (Illumina, San plausible, the outgrowth in therapy-naïve cells requires a
Diego, CA, USA) covered amino acids 184-510 of ABL1 fitness advantage. In previously published biochemical
and were generated from pre-amplified BCR-ABL1. A or cell culture studies, the effect of mutations without
1% detection limit was used for analysis with SeqNext TKI was evaluated. D276G increased catalytic activity,9
(JSI Medical Systems, Kippenheim, Germany). and E255K exhibited increased transformation potency.10
Using next-generation sequencing, we found well- Although T315I showed reduced kinase activity com-
characterized ABL1 kinase domain mutations in five of pared to unmutated controls, cell culture assays suggest-
the 91 (5.5%) diagnostic samples (Table 1, Online ed increased oncogenic potency via different pathways
Supplementary Table S1). Mutations were confirmed in a (e.g., phosphorylation).11 In a recent study of pediatric
second independent sequencing run to exclude potential Ph-like ALL (with EBF1-PDGFRB rearrangement), Tran
artefacts from random polymerase chain reaction errors. and colleagues back-tracked the gatekeeper T681I muta-
Four patients with a mutation at diagnosis had the m-bcr tion in PDGFRB, which is the analog of the T315I muta-
Table 1. Characteristics of patients with mutations at initial diagnosis.

Patient #1 #2 #3 #4 #5
Gender male female female male male
Age 72 years 67 years 52 years 79 years 76 years
Blasts 91% 88% 12% 85% 85%
Transcript p190 p190 p190 p210 p190
Mutation Q252H (G>T): 46% D276G: 3% T315I: 14% M244V: 2% E255K: 1%
Q252H (G>C): 8%
Treatment GMALL elderly alloSCT alloSCT NA NA
Outcome remission followed by relapse NA CR NA NA
GMALL: protocol of the German Multicenter Study Group on Adult Acute Lymphoblastic Leukemia; alloSCT: allogeneic stem cell transplantation; CR: complete remission;
NA: data not available.

A Constance Baer, Manja Meggendorfer, Claudia

Haferlach, Wolfgang Kern and Torsten Haferlach
MLL Munich Leukemia Laboratory, Munich, Germany
Corresponding author:
CONSTANCE BAER - constance.baer@mll.com
doi:10.3324/haematol.2021.279807
Accepted: November 3, 2021.
Pre-published: November 11, 2021.
B
Disclosures: CB and MM are employees in MLL Munich
Leukemia Laboratory; WK, CH and TH own equity in MLL Munich
Leukemia Laboratory.
Contributions: CB and MM analyzed the data; CB, CH, WK and
TH designed the study; CB and TH wrote the manuscript.
References
1. Soverini S, Vitale A, Poerio A, et al. Philadelphia-positive acute lym-
phoblastic leukemia patients already harbor BCR-ABL kinase
Figure 1. Molecular monitoring of BCR-ABL1 transcripts and ABL1 point domain mutations at low levels at the time of diagnosis.
mutations in a patient with Philadelphia chromosome-positive acute lym- Haematologica. 2011;96(4):552-557.
phoblastic leukemia. (A) Patient #1 was treated according to the GMALL eld- 2. Pfeifer H, Wassmann B, Pavlova A, et al. Kinase domain mutations of
erly protocol including imatinib. BCR-ABL1 expression (BCR-ABL/ABL1 in %) BCR-ABL frequently precede imatinib-based therapy and give rise to
was measured by quantitative polymerase chain reaction. (B) The variant relapse in patients with de novo Philadelphia-positive acute lym-
allele frequency (VAF in %) is given for two follow-up time-points. phoblastic leukemia (Ph+ ALL). Blood. 2007;110(2):727-734.
3. Hofmann WK, Komor M, Wassmann B, et al. Presence of the BCR-
ABL mutation Glu255Lys prior to STI571 (imatinib) treatment in
tion in ABL1. They identified the mutation in three of 23 patients with Ph+ acute lymphoblastic leukemia. Blood.
patients prior to treatment.12 We speculate that muta- 2003;102(2):659-661.
4. Hato T, Yamanouchi J, Tamura T, et al. Existence of leukemic clones
tions which confer resistance but reduce kinase activity
resistant to both imatinib mesylate and rituximab before drug thera-
or transformative capacity would not outgrow before pies in a patient with Philadelphia chromosome-positive acute lym-
TKI treatment and therefore only a subset of resistance phocytic leukemia. Int J Hematol. 2004;80(1):62-66.
mutations can exist in therapy-naïve patients. The high 5. Willis SG, Lange T, Demehri S, et al. High-sensitivity detection of
turnover rate of ALL cells should allow a much faster BCR-ABL kinase domain mutations in imatinib-naive patients: corre-
selection of mutated clones with a relative growth lation with clonal cytogenetic evolution but not response to therapy.
Blood. 2005;106(6):2128-2137.
advantage if compared to the situation in chronic 6. Cortes JE, Kantarjian H, Shah NP, et al. Ponatinib in refractory
myeloid leukemia. In an initial dataset of treatment- Philadelphia chromosome-positive leukemias. N Engl J Med.
naïve patients with chronic phase chronic myeloid 2012;367(22):2075-2088.
leukemia, only minor subclones (<1%) were identified, 7. Soverini S, De Benedittis C, Papayannidis C, et al. Drug resistance
and showed no correlation with endpoints.13 and BCR-ABL kinase domain mutations in Philadelphia chromo-
In summary, known resistance mutations in the ABL1 some-positive acute lymphoblastic leukemia from the imatinib to
the second-generation tyrosine kinase inhibitor era: the main
kinase domain were detected in five of 91 (5.5%) thera- changes are in the type of mutations, but not in the frequency of
py-naïve patients with BCR-ABL1-positive ALL. For mutation involvement. Cancer. 2014;120(7):1002-1009.
patient #1 (Figure 1) we can show that the mutated clone 8. Soverini S, De Benedittis C, Papayannidis C, et al. Clinical impact of
expands rapidly if treatment with an insensitive TKI is low-burden BCR-ABL1 mutations detectable by amplicon deep
chosen. At present, we see no significant difference in sequencing in Philadelphia-positive acute lymphoblastic leukemia
patients. Leukemia. 2016;30(7):1615-1619.
outcome for the five patients with mutations, but the
9. Piazza RG, Magistroni V, Gasser M, et al. Evidence for D276G and
heterogeneity of the cohort and the small number of L364I Bcr-Abl mutations in Ph+ leukaemic cells obtained from
cases with mutations need to be considered. Further patients resistant to imatinib. Leukemia. 2005;19(1):132-134.
studies will be necessary, especially involving the full 10. Griswold IJ, MacPartlin M, Bumm T, et al. Kinase domain mutants of
spectrum of TKI, which are now increasingly a backbone Bcr-Abl exhibit altered transformation potency, kinase activity, and
and guarantor of success of improving outcomes in Ph+ substrate utilization, irrespective of sensitivity to imatinib. Mol Cell
Biol. 2006;26(16):6082-6093.
disease. Next-generation sequencing is a direct and sen- 11. Skaggs BJ, Gorre ME, Ryvkin A, et al. Phosphorylation of the ATP-
sitive (1%) strategy to identify patients at risk of resist- binding loop directs oncogenicity of drug-resistant BCR-ABL
ance before any TKI therapy is started. However, 93% mutants. Proc Natl Acad Sci U S A. 2006;103(51):19466-19471.
(14/15 patients) of mutations found in relapse samples 12. Tran TH, Nguyen JV, Stecula A, et al. The EBF1-PDGFRB T681I
were most likely acquired under the selective pressure of mutation is highly resistant to imatinib and dasatinib in vitro and
(TKI) treatment. Therefore, testing at initial diagnosis detectable in clinical samples prior to treatment. Haematologica.
2021;106(8):2242-2245.
should only be considered in addition to established 13. Franke G-N, Maier J, Wildenberger K, et al. Incidence of low level
mutation testing in refractory/relapsed disease. mutations in newly diagnosed CML patients: a substudy of the
German Tiger Trial. Blood. 2017;130(Suppl 1):252

2022 02 Versione Digitale

Uploaded by

Copyright:

Available Formats

2022 02 Versione Digitale

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2022 02 Versione Digitale

Uploaded by

Copyright:

Available Formats

haematologica

Journal of the Ferrata Storti Foundation

Brief information on Haematologica

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

Subscription. Detailed information about subscriptions is available at www.haematologica.org. Haematologica is an open

Associated with USPI, Unione Stampa Periodica Italiana.

Landmark Papers in Hematology

Acute Myeloid Leukemia

Haematologica 2022; vol. 107 no. 2 - February 2022

Cell Therapy & Immunotherapy

Iron Metabolism & its Disorders

478 UBA6 and NDFIP1 regulate the degradation of ferroportin

Platelet Biology & its Disorders

Haematologica 2022; vol. 107 no. 2 - February 2022

Red Cell Biology & its Disorders

Letters to the Editor

Haematologica 2022; vol. 107 no. 2 - February 2022

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

Scientific Latin haematologicus (adjective) = related to blood

Scientific Latin haematologica (adjective, plural and neuter,

Modern English The oldest hematology journal,

haematologica | 2022; 107(2) 351

TITLE Aggregation of blood platelets by adenosine diphosphate and its reversal

AUTHORS Gustav Victor Rudolph Born

JOURNAL Nature. 1962;194:927-929. PMID: 13871375

Born's platelet aggregometer. (A) The

352 haematologica | 2022; 107(2)

haematologica | 2022; 107(2) 353

354 haematologica | 2022; 107(2)

haematologica | 2022; 107(2) 355

356 haematologica | 2022; 107(2)

haematologica | 2022; 107(2) 357

Ferrata Storti Foundation Integrating genetic and epigenetic factors in

358 haematologica | 2022; 107(2)

haematologica | 2022; 107(2) 359

360 haematologica | 2022; 107(2)

haematologica | 2022; 107(2)

Single-cell-based gene expression analysis Gene expression signatures associated with

362 haematologica | 2022; 107(2)

haematologica | 2022; 107(2) 363

364 haematologica | 2022; 107(2)

haematologica | 2022; 107(2)

366 haematologica | 2022; 107(2)

haematologica | 2022; 107(2) 367

368 haematologica | 2022; 107(2)

References genome-wide cDNA microarray analysis. 2017;23(6):692-702.

haematologica | 2022; 107(2) 369

associated DNMT3A mutations reduce evolutionary dynamics of clonal Genet. 2018;56(3):149-175.

370 haematologica | 2022; 107(2)

Comprehensive analysis of dose intensity of Ferrata Storti Foundation

©2022 Ferrata Storti Foundation

haematologica | 2022; 107(2) 371

372 haematologica | 2022; 107(2)

ified conversions (i.e., pegaspargase 2,500 units/m2 were consid- Results

haematologica | 2022; 107(2) 373

Figure 3. Thiopurine dose intensity

374 haematologica | 2022; 107(2)

haematologica | 2022; 107(2) 375

376 haematologica | 2022; 107(2)

haematologica | 2022; 107(2) 377