A LYSA Phase Ib Study of Tazemetostat (EPZ-6438) plus R-CHOP in Patients with Newly Diagnosed Diffuse Large B-Cell Lymphoma (DLBCL) with Poor Prognosis Features

Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of non-Hodgkin lymphoma (NHL) accounting for about 30% of all lymphoid neoplasms. The incidence increases with age, around 50% of the patients being older than 60 years and almost one third over 75 years (1). Importantly, despite their age, the vast majority of these patients are eligible and treated with the standard-of-care immunochemotherapy regimen, namely R-CHOP [rituximab (R) in association with prednisolone, doxorubicin, cyclophosphamide, and vincristine (CHOP; refs. 1, 2)]. Age-adjusted IPI (aaIPI), a derivative of the IPI score based on lactate dehydrogenase (LDH), stage, and ECOG performance status, can stratify the patients at diagnosis and predicts outcome (3). In patients older than 60 years, DLBCL can be cured in 50% to 60% of the cases following first-line R-CHOP (4–6) and 30% to 40% will present a R-CHOP primary refractory or an early-relapse disease (7, 8). Those “poor responders” to R-CHOP have a 1-year overall survival (OS) of less than 20% and still represent a group of patients with a clear unmet medical need (7). Among all DLBCL, elderly patients with a high aaIPI represent the most challenging population with the higher primary refractory rate leading to poor OS (5, 9).

Epigenetic modulation of histones plays a critical role in oncogenic transformation in many malignancies and is now an area of intense clinical research (10). Next-generation sequencing of B-cell NHL genomes has uncovered frequent mutations affecting histone-modifying proteins especially in lymphomas derived from the germinal center (11). One of them, EZH2, is the catalytic subunit of the chromatin remodeling polycomb repressive complex 2 (PRC2), and the only human protein methyltransferase that can methylate histone 3 lysine 27 (H3K27) leading to the transcriptional repressive mark H3K27me3. During hematopoiesis, EZH2 plays an important role where it represses genes involved in cell-cycle arrest and terminal differentiation. Later during germinal center (GC) formation, EZH2 activity becomes increasingly opposed by the SWI/SNF (Switch/Sucrose Non-Fermentable) chromatin-remodeling multiprotein complex, which facilitates gene expression and terminal differentiation (12, 13). Disrupting this differentiation process through hypertrimethylation of H3K27, gain-of-function mutations in EZH2 are reported in 20% of GC B-cell lymphoma, resulting in an aberrant proliferative dependency on EZH2 activity (14–16). In normal lymph nodes, genetic deletion or pharmacologic inhibition of EZH2 suppresses GC formation in vivo, suggesting that EZH2 plays a key role in GC differentiation and that its inhibition might be active against DLBCL, even without EZH2-activating mutations (12).

Tazemetostat (EPZ-6438) is a potent, oral highly selective oral EZH2 inhibitor under clinical investigation. In a first-in-human phase I trial, tazemetostat demonstrated favorable safety with uncommon grade 3 or worse treatment-related adverse-events (AE) and one dose-limiting toxicity (DLT) of grade 4 (thrombocytopenia at 1,600 mg). Durable objective responses, including complete responses (CR), were observed in 38% of patients with B-cell NHL. On the basis of composite evaluation of AEs, pharmacokinetics, pharmacodynamics, and preliminary antitumor activity, the recommended phase II (RP2D) was determined to be 800 mg twice daily (17).

This phase Ib part of Epi-R-CHOP explored the safety and preliminary signal of activity of combining tazemetostat with R-CHOP in patients with DLBCL at diagnosis.

We performed a Phase 1b open-label, multicenter, dose-escalation study of tazemetostat (EPZ-6438) in combination with standard doses of intravenous rituximab plus CHOP in elderly patients with high-risk DLBCL (ClinicalTrials.gov identifier: NCT02889523, Eudract: 2016-001499-31). The study protocol was approved by the institutional review board and ethics committee at participating institutions in accordance with the International Conference on Harmonisation guidelines, including good clinical practice and the ethical principles originating from the Declaration of Helsinki. A written informed consent was obtained from all the included patients.

Patients were recruited in four LYSA centers and enrolled using 3 + 3 patients per dose level (starting at 400 mg twice a day) algorithm to identify the MTD and determine RP2D of tazemetostat plus R-CHOP.

Patients received up to eight cycles of standard R-CHOP regimen every 21 days (rituximab 375 mg/m² D1, prednisolone oral 40 mg/m² D1-5, doxorubicine 50 mg/m² D1, cyclophosphamide 750 mg/m² D1, vincristine 1.4 mg/m² D1) in combination with continuous tazemetostat at 400 mg, 600 mg, or 800 mg twice a day starting on day 2 of R-CHOP cycle 1 (C1) until day 21 of cycle 8. Prevention of febrile neutropenia was allowed with G-CSF, according to the ASCO recommendations (18). Valacyclovir and cotrimoxazole prophylaxis were mandatory. Dose adaptation was allowed after the first two cycles (DLT period). A full list of the dose adaptation rules is provided in the Supplementary Appendix.

Eligible patients were aged between 60 and 80 years and had been diagnosed with an untreated high-risk DLBCL (aaIPI 2-3). Other eligible criteria wereas follows: ECOG performance status of 0–1, adequate renal function (creatinine clearance > 40 mL/min), bone marrow function (ANC ≥ 1,500/mm³, platelets ≥ 75,000/mm³, hemoglobin ≥ 9 g/dL), liver function, left ventricular ejection fraction > 50%. Patients were excluded if they presented a central nervous system or meningeal involvement, had received prior treatment with any EZH2 inhibitor, any previous lymphoma treatment, or had any conditions that might compromise their safety or the study according to the investigator. A full list of inclusion/exclusion criteria is provided in the Supplementary Appendix.

For dose-escalation purpose, toxicities assigned as DLTs were assessed during C1 and C2 (42 days) to be able to perform pharmacokinetic (PK) analysis of R-CHOP component with and without tazemetostat (cf below). The DLT set includes all patients having signed their informed consent and who completed at least C1 and C2 unless the discontinuation reason was a DLT. Consequently, patients who did not receive 85% of tazemetostat-planned doses for another reason than DLT were not evaluable and replaced.

For adverse events (AE) assessment, clinical examinations and laboratory safety tests were obtained prior to drug administration, every week during the first two cycles (and at day 2 and 4 of C1 for blood cell counts, BCC), before each cycle of treatment, and up to 28 days after the last study treatment administration. AEs/TEAEs (treatment emergent AE) type, severity, duration, and seriousness were assessed according to the National Cancer Institute Common Terminology Criteria for Adverse Events [NCI-CTCAE] v. 4.03. Laboratory abnormalities were assessed according to the NCI-CTCAE v. 4.03. A DLT was defined as an AE or clinically significant abnormal laboratory value attributable to tazemetostat or to the combination of tazemetostat and R-CHOP and unrelated to disease progression, intercurrent illness, or concomitant medication.

After completion of C2 in each cohort, all available safety data were reviewed by a Safety Review Committee (SRC) and the decision to proceed (or not) to the next dose cohort was made. If 0 of 3, or 1 of 6 patients demonstrated DLT(s), then enrollment proceeded to the next dose level. If 2 patients in any cohort demonstrated a DLT(s) during cycle 1 or 2, enrollment stopped and dose escalation was discontinued. The dose would have been reduced to the previously completed dose level for which no more than 1 DLT had been observed. The highest dose level resulting in one or less of 6 patients with DLTs was considered as the MTD/RP2D.

Tumor assessment was performed at baseline, after 4 cycles (CT scan) and after cycle 8 by PET-CT according to the Lugano classification (19).

Serial blood samples for PK analysis were collected on C1D1, before any tazemetostat intake, and C2D1, after 1 cycle of tazemetostat, for doxorubicin, its major metabolite, doxorubicinol, vincristine, and cyclophosphamide. PK results were compared using the C2/C1 geometric mean ratio (GMR) and 90% confidence interval (CI) for area under the curve (AUC) and C_max values after R-CHOP alone (C1) or R-CHOP plus tazemetostat (C2). PK analysis of tazemetostat was performed at C2D2 and compared with tazemetostat PK as single agent (E7438-G000-101).

Nucleic acid extraction was performed on FFPE samples according to standard procedure. Gene expression–based cell of origin was assessed thanks to RT-MLPA (Reverse Transcription and Multiplex Ligation-dependent Probe Amplification) and sequencing of a panel of 36 genes recurrently mutated in lymphoma (design provided in Supplementary Table), done with an amplicon-based technic as previously published (20, 21).

Seventeen patients, enrolled between October 2016 and March 2018, received at least one dose of tazemetostat R-CHOP (full-analysis set). Initial characteristics of the full-analysis set are described in Table 1. The median age was 68 years (61–76). The 17 patients had a disseminated disease, 13 (76.5%) with a stage IV, and 9 (53%) with 2 or more extranodal involvement. LDH were higher than the upper normal limit (UNL) in all the cases, and all had an aaIPI score at 2. For the 15 cases with available data for RT-MLPA analysis (2 nucleic acid extraction failure), 8 were classified as GC-DLBCL NOS, 5 as ABC-DLBCL NOS, 1 as unclassified DLBCL NOS, and 1 as EBV + DLBCL

Toxicities assigned as DLT were assessed during C1 and C2. Two patients (800 mg cohort) were not evaluable for DLT and replaced (1 noncompliance and 1 lymphoma-related hepatic cholestasis) leading to a DLT set of 15 patients. One patient on cohort 1 (400 mg twice a day) experienced a DLT during C1 (grade 3 constipation). No DLT was observed on cohort 2 (600 mg twice a day). Another DLT was observed on cohort 3 (800 mg twice a day): a grade 5 pneumocystis jirovecii (PJ) in a context of documented influenza during cycle 1. This patient was a 67-year-old woman with a medical history of bronchial dilatation and tuberculosis sequela. The patient had a high tumor burden with a large mass in the cervical area (>10 cm) responsible of an upper respiratory tract compression. During cycle 1, she presented with a neutropenia (grade 3, starting D7) that resolved at D11. She was admitted on D14 to intensive care unit (ICU) due to grade 4 respiratory distress on a pneumopathy. The final diagnosis of influenzae and PJ was then performed. Despite respiratory assistance and adapted PJ treatment, the patient died in ICU on D39. No other DLT was observed in the next three consecutive patients treated at the highest planned dose (800 mg twice a day), which was considered as the RP2D.

During the DLT period, two other SAEs were reported: 1 grade 3 febrile neutropenia and 1 grade 3 hypokalemia.

Nonhematologic AEs occurring in ≥10% of the 17 patients are represented in Fig. 1A and Table 2. Gastrointestinal toxicities were the most recurrent [constipation (N = 10, 59%), nausea (N = 10, 59%), vomiting (N = 9, 53%), abdominal pain (N = 5, 29%), diarrhea (N = 4, 23%), followed by neurologic AE (N = 6, 35%), infectious (N = 6, 35%), weight loss (N = 5, 29%), musculoskeletal pain (N = 4, 23%), fatigue (N = 4, 23%), headache (N = 4, 23%) and anxiety, chest pain, cholestasis, hypokalemia, mucositis (N = 2, 12% each)]. Most of these AEs were of grade 1 or 2. Grade 3 or more toxicities observed ≥10% of the patients were constipation, (N = 4, 24%), nausea (N = 2, 12%), and hypokalemia (N = 2, 12%). Hematologic AEs occurring in ≥10% of the patients were as follows: neutropenia (N = 8, 47%), anemia (N = 7, 41%), leukopenia (N = 7, 41%), and thrombocytopenia (N = 2, 12%; Table 3). Grade 3 to 4 hematologic AEs were recorded in 8 patients (47%): neutropenia (N = 8, 47%), leukopenia (N = 5, 29%), anemia (N = 3, 18%), and thrombocytopenia (N = 2, 12%).

No organ-oriented toxicity increased with tazemetostat dosage escalation (severity or incidence). More particularly, only two grade 3–5 infectious AEs occurred including one febrile neutropenia.

Fourteen patients were treated after the DLT period. When compared with C1-C2, the incidence of hematologic AEs during C3-C8 was stable with neutropenia, anemia, and thrombocytopenia reported in 42.9%, 50%, and 21.4% of the 14 patients, respectively. Grade 3+ hematologic AEs, reported by more than 10% of the patients, were leukopenia and neutropenia (28.6%) and thrombocytopenia (21.4%). Incidence of grade 3 thrombocytopenia was higher but did not necessitate platelet transfusion. No febrile neutropenia was observed. No grade 3 or more nonhematologic AE occurred in more than 10% of the patients. One patient presented with two SAEs after the DLT period: one acute pyelonephritis with staphylococci bacteremia with a grade 1 renal failure and a postlumbar puncture syndrome.

In conclusion, no serious organ-oriented toxicity was observed after the DLT period.

R-CHOP relative dose intensity (RDI) was calculated as the ratio between the total dose received by the patient and the planned dose for each started cycle. The results are presented in Table 4 and shows that the mean administration dosage for each R-CHOP component, except vincristine, was close to 100%. Indeed, 94%, 100%, and 100% of the patients received more than 90% of the expected dosage of rituximab, doxorubicine, and cyclophosphamide, respectively. However, significant reduction of vincristine dosage (i.e., 50% of dose reduction) was necessary in 7 patients (41%, due to peripheral neuropathy and/or constipation) that received less than 75% of the expected dosage.

The duration between D1 of each cycle was stable throughout the treatment and between the three cohorts (mean time between C1 and C2 was 21.4 days versus 22.8 (range 19–24) days between C7 and C8 (range 21–28).

During the DLT period (C1 and C2), 2 patients in the 800 mg cohort received less than 85% of the expected dose due to reasons not related to AE (noncompliance and lymphoma-related cholestasis) and were replaced, another patient had a grade 5 infectious DLT. The C1-C2 RDI was greater than 90% for the remaining patients.

Fourteen patients were treated after the DLT period: 1 stopped tazemetostat at C4 (investigator decision) and 13 received 8 cycles of tazemetostat-R-CHOP. For the C3 to C8 cycles, the median tazemetostat RDI remained greater than 90%, except at C7 for the 400 mg cohort (median RDI 69%), and no significant reduction in mean dosage administration could be observed within dose escalation for the treated patients.

For the full-analysis set, the average total missing daily dose of tazemetostat per patient (17 patients) was 9.4 (1–28, 25) and 13 patients had a temporary interruption with at least one caused by an AE in 7 (total missing dose 11.8 in average), whereas oversight was the only reason in 6 patients (total missing dose 7.8 in average). AEs leading to treatment-temporary interruption all recovered.

Plasma concentration–time profiles of doxorubicin, doxorubicinol, cyclophosphamide, and vincristine are presented in Supplementary Fig. S1. Tazemetostat had no significant impact on doxorubicin and doxorubicinol AUC; and although C_max of doxorubicin was lower after administration with 800 mg tazemetostat relative to administration without tazemetostat (GMR 0.68; 90%CI, 0.47–0.99), C_max of its metabolite, doxorubicinol, was similar (GMR 0.928; 90%CI, 0.691–1.25), suggesting no significant effect on the metabolism of the drug. Cyclophosphamide AUC was similar in the 400 mg cohort (GMR 1.05; 90%CI, 0.87–1.26) and slightly lower in the 800 mg cohort (GMR 0.83; 90%CI, 0.77–0.9) with a similar C_max (GMR 0.89; 90%CI, 0.74–1.07). Importantly, we did not observe any increase in hematologic toxicities associated with these slight decreases in AUC, suggesting no meaningful impact. At 800 mg, AUC (3,590 vs. 3,340 ng*h/mL) and C_max (782 vs. 829 ng/mL) of tazemetostat were similar compared with single-agent study (E7438-G000–101). Plasma vincristine concentrations were not adequate to calculate reliable PK parameters in all subjects, given that vincristine was the last component administered. However, there appeared to be no effect of tazemetostat administration on vincristine concentrations (see Supplementary Fig. S1). These data suggest no significant drug–drug interaction between doxorubicin, cyclophosphamide, vincristine, and tazemetostat.

At the time of analysis, the median follow-up was 20.6 months, 14.5 months, and 7.5 months in the 400, 600, and 800 mg cohort, respectively. Out of the 17 included, one died at cycle 1 (AE), one discontinued the treatment due to investigator decision, and two for reasons other than AE or progression (noncompliance and lymphoma-related cholestasis), and were considered as nonresponders, and 13 completed the 8 cycles of treatment and reached a metabolic CR (mCR, Lugano criteria). Therefore, the mCR rate was 76.5% (13/17). During follow-up, 2 patients presented with a relapse. One early relapse was salvaged by R-ICE regimen followed by an autologous stem cell transplant and reached second CR. One late relapse is under salvage treatment (R-DHAOX). Eleven patients remain in CR (CR duration 2–14 months).

DNA and RNA were successfully extracted for 15 patients. The most recurrently mutated genes were CREBBP (N = 8 cases); BCL2, MYC, and PIM1 (N = 4 cases each); CARD11, EZH2, FOXO1, TNFRSF14, TP53 (N = 3 cases each; Supplementary Fig. S2). The three EZH2-mutated cases had a GCB phenotype in RTMPLA. Among them, one died at C1 (infection-related death), one was excluded and not evaluable for response (lymphoma-related cholestasis), and one reached a CR. Among the two relapsing patients, none had an EZH2 mutation and both presented with a TP53 mutation (with PLCG2 and IRF4 for one and MYD88 for the other).

Epi-R-CHOP is the first trial to evaluate the PK and safety of the combination of the EZH2 inhibitor tazemetostat, and R-CHOP, in elderly patients with previously untreated high-risk DLBCL. This phase Ib study with a prolonged DLT period of 42 days was unable to identify a MTD and 800 mg twice a day was considered as the RP2D, identical to tazemetostat dosage in monotherapy.

In this high-risk elderly DLBCL population, 1 patient discontinued the treatment due to DLT at C1 (grade 5), 1 due to investigator decision at C4 (DLT experienced at C1). Two other patients discontinued the treatment during C1 due to reasons other than DLT (noncompliance and cholestasis related to the lymphoma).

The most recurrent AEs were cytopenias and gastrointestinal toxicities. Importantly, the incidence and severity of neutropenia and anemia did not increase with time. However, grade 3–4 thrombocytopenia was slightly more frequent during C3-C8 (21.4%) versus C1-C2 (11.8%), but no platelet transfusion was necessary. The incidence of neutropenia is in line with other R-CHOP studies such as the recent GOYA trial that reported 38.1% of grade 3 to 5 neutropenia (6). In the GOYA trial, 7.5% of the R-CHOP–treated patients presented with a grade 3–5 anemia, which is similar to the incidence reported in Epi-R-CHOP between C3 and C8 (7%), whereas the incidence during C1-C2 (17.6%, 3 patients) was slightly higher in Epi-R-CHOP. This might, in part, be related to the initial characteristics of the population in EpiRCHOP as 2 of these 3 patients presented with an initial grade 2 anemia. Of note, 100% of the patients had a high intermediate or high IPI score in Epi-R-CHOP versus 42% in GOYA. Importantly, this did not result in a high rate of red blood cell transfusion (only 10 RBC transfusions for the 110 cycles, including two at diagnosis) and the dose intensity of the R-CHOP backbone was maintained. Finally (Supplementary Table S1) the incidence of cytopenia, and more particularly of febrile neutropenia, does not exceed what is reported in phase Ib trials combining R-CHOP with other targeted therapies (22–27) in this elderly population. In Epi-R-CHOP, G-CSF prophylaxis might explain the low incidence of febrile neutropenia and severe infectious events. We observed one grade 5 infectious event due to a PJ infection in a context of documented influenzae 10 days after the treatment initiation in a patient presenting with a background of pulmonary fibrosis. Although the event was regarded as a DLT, the link with the study drug is very unlikely. The global incidence of 17.5% of grade 3–5 infections (N = 3), including 11.7% (N = 2) of febrile neutropenia, does not exceed what is reported in the literature for R-CHOP or R-CHOP combination trials. Indeed, in the recent GOYA trial, 19.2% of the patients presented with a grade 3–5 infectious and 17.5% a grade 3–5 febrile neutropenia. Furthermore, when treated with venetoclax, vorinostat, ibrutinib, lenalidomide, bortezomib, or polatuzumab vedotin plus R-CHOP 33.3%, 38%, 18%, 9.5%, 13.5%, and 18% of the patients presented with a febrile neutropenia, respectively (22–27, 29).

Constipation, nausea, and vomiting were the most frequent nonhematologic AEs in this elderly population (59%, 59%, and 53%, respectively) with 24% and 12% of grade 3 constipation and nausea, respectively. The severity and incidence of constipation seem to be higher when compared with other trials (42% in ibrutinib-RCHOP, 33% in venetoclax RCHOP, 38% in vorinostat-RCHOP, 26% in polatuzumab-RCHOP) and led to the first DLT (stercoral stasis). Consequently, as opposed to the other R-CHOP components, vincristine dosage reduction was applied in few patients. The higher incidence of GI toxicities might, in part, be explained by the study design that collected AEs every week during C1 and C2 but also by the older age of the patients, as the elderly population is more subject to constipation with vincristine-based chemotherapy (28). On the basis of PK analysis, there appeared to be no effect of tazemetostat administration on vincristine concentrations to explain the incidence of constipation. No significant drug–drug interaction between doxorubicin, cyclophosphamide, and tazemetostat was found either. These GI AEs were not reported in tazemetostat single agent (3% of constipation, grade 1–2 only; ref. 17). The GI AEs are probably explained by the exhaustive weekly collection of the AEs in this elderly population that received 8 cycles of R-CHOP. A reduction to 6 cycles of RCHOP, that is, the recently accepted new standard of care in this population (6), should be considered in the phase II part of the trial. Improvement in concomitant medications and antiemetic prophylactic measures are clearly warranted in this elderly population treated with chemotherapy (28).

The PK-related risks of the study included the risk that tazemetostat would induce doxorubicin and cyclophosphamide metabolism to the more active species and increase toxicity. All in all, the data presented here indicated that the toxicity of RCHOP was globally not increased with tazemetostat. The PK analysis of the main R-CHOP drugs suggest a similar metabolism of the anthracycline (with identical doxorubicinol PK in C1 vs. C2) and modest modification regarding cyclophosphamide [similar C_max and lower AUC for at 800 mg (MGR 0.83)] that might warrant further investigation. However, the absence of increase in the hematologic toxicities is reassuring and suggests no relevant consequences on cyclophosphamide metabolism when combined with tazemetostat. Furthermore, vincristine is a CYP3A substrate, with a nonactive primary metabolite, and tazemetostat is a weak CYP3A inducer (30), suggesting that vincristine drug exposure should not be increased with the association.

Preliminary antitumor activity data showed an mCR rate of 76.5% (13/17 of the included patients, and all the 13 evaluable patients that completed the 8 cycles of tazemetostat-R-CHOP were in mCR). Two relapses were observed in the EZH2 wild-type population, both presenting with a TP53 mutation. Although exploratory and reflecting only preliminary data, this rate of mCR is encouraging (Supplementary Table S2) and supports the necessity of a phase II trial.

In conclusion, the phase Ib Epi-R-CHOP study showed that the combination of R-CHOP plus tazemetostat is generally well tolerated and the addition of tazemetostat to R-CHOP does not appear to substantially change the expected R-CHOP toxicities. The MTD was not reached and the RP2D of tazemetostat combined with R-CHOP is 800 mg twice a day. On the basis of these results, further investigation is warranted and the expansion phase II Epi-R-CHOP study for 60- to 80-year-old patients with newly diagnosed aaIPI 1 or higher DLBCL will be enrolling shortly.

F. Morschhauser is a paid advisory board member for Epizyme and Roche, and reports receiving speakers bureau honoraria from Roche and Celgene. J. Michot is an unpaid consultant/advisory board member for Celgene, Iqone, Amgen, and Merck. B. Suttle is an employee of Epizyme. G. Salles is a paid consultant for Epizyme. V. Ribrag reports receiving other commercial research support from Epizyme and Argen-X, and is an unpaid consultant/advisory board member for Infinity, Bristol-Myers Squibb, Parmamar, Gilead, Incyte, MSD, Servier, Nanostring, Roche, and Immune Design. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Sarkozy, B. Suttle, G. Salles, V. Ribrag

Development of methodology: C. Sarkozy, S. Dubois, F. Jardin, G. Salles, V. Ribrag

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Sarkozy, F. Morschhauser, T. Molina, J.-M. Michot, P. Cullières-Dartigues, L. Karlin, S. Le Gouill, J.-M. Picquenot, R. Dubois, H. Tilly, C. Herbaux, F. Jardin, G. Salles, V. Ribrag

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Sarkozy, F. Morschhauser, J.-M. Michot, B. Suttle, H. Tilly, F. Jardin, G. Salles, V. Ribrag

Writing, review, and/or revision of the manuscript: C. Sarkozy, F. Morschhauser, S. Dubois, T. Molina, J.-M. Michot, B. Suttle, S. Le Gouill, F. Jardin, G. Salles, V. Ribrag

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Salles

Study supervision: C. Sarkozy, V. Ribrag

The authors thank the patients and their families. The authors thank LYSARC and Alexia Schwartzmann for managing the study; all the LYSA investigators and centers for including patients; The LYSA-P and Nadine Vailhen for managing the central review and biological studies; Elodie Bohers (NGS), Pascaline Etancelin (banking), and Pierre Julien Viailly (bioinfo) from the U1245 Centre H.Bequerel for the sequencing analysis. This study was funded by Epizyme and sponsored by the LYSARC (The Lymphoma Academic Research Organisation).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.