EZH2 mutation status predicts inferior outcomes after bendamustine-based immunochemotherapy in frontline follicular lymphoma: A GELTAMO multicenter validation study

Abstract

Follicular lymphoma (FL) is the most common indolent lymphoma,^{1, 2} a biologically and clinically heterogeneous entity with a diverse genetic landscape and characterized by slowly progressive but relapsing clinical course.³ The t(14;18) translocation driving BCL-2 overexpression is present in 85%–90% of patients⁴ although additional genetic events are required for lymphomagenesis. Next-generation sequencing (NGS) studies have identified recurrent mutations in chromatin-modifying genes (KMT2D, CREBBP, and EZH2).^{5, 6} EZH2 encodes the catalytic subunit of the Polycomb repressor complex 2 and mediates H3K27 methylation. Missense mutations in EZH2, present in 20%–25% of patients, impair transcriptional programs regulating cell cycle and plasma-cell differentiation, promoting oncogenic transformation.^7-11

These insights led to the development of the m7-FLIPI risk model, which includes the mutational status of seven genes.¹² EZH2 mutations carry the strongest weight and are associated with a more favorable prognosis in patients receiving chemoimmunotherapy. However, m7-FLIPI's prognostic performance depends on treatment regimen, performing less well with rituximab (R)-bendamustine or chemo-free approaches.^{13, 14}

Frontline treatment of newly diagnosed FL commonly involves R-bendamustine or R plus cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP). R-bendamustine is non-inferior in terms of efficacy and has less toxicity, but none of them has demonstrated benefits in overall survival (OS).^15-17

A previous retrospective single-center study by Martínez-Laperche et al.¹⁰ suggested that EZH2 status may influence treatment response and guide regimen selection: patients with EZH2-mutated FL achieved improved progression-free survival (PFS) and OS when treated with R-CHOP compared to R-bendamustine (4-year PFS 92% vs. 40%, OS 100% vs. 78% P = 0.039), whereas patients with wild-type EZH2 could obtain a greater benefit from R-bendamustine. These observations have not yet been confirmed, and clinical guidelines do not recommend routine use of EZH2 status to guide first-line therapy.

The aim of this study was to evaluate the predictive value of EZH2 mutation status for frontline treatment efficacy in FL and to validate whether EZH2 mutations identify patients who have a differential benefit from R-CHOP versus R-bendamustine in a multicenter real-world cohort.

We performed a retrospective multicenter analysis of 324 patients diagnosed with FL between 2002 and 2020 at six Spanish academic hospitals. Only patients who completed or were planned to receive six full-dose cycles of R-CHOP or R-bendamustine were included. Patients with histological transformation or any diffuse large B-cell lymphoma component were excluded from the analysis. All patients included in the study had newly diagnosed, previously untreated FL. Patients who had received prior systemic therapy for FL were excluded. Tumor DNA was analyzed for EZH2 mutations using either targeted NGS panels covering the complete EZH2 gene (n = 162, 50%) or Sanger sequencing restricted to exons 16 and 18, which include the canonical EZH2 mutational hotspots described in FL, including codon Y646. EZH2-mutated cases were defined as those harboring EZH2 mutations detected by the sequencing method used. All detected EZH2 mutations corresponded to canonical hotspot mutations, with the majority involving codon Y646.

Study protocol was approved by the local Ethics Committee. All patients provided informed consent.

Univariate analyses (Kaplan–Meier, log-rank) were performed, and significant variables included in Cox models for PFS and OS.

Patients' characteristics are presented in Table 1. In total, 82 (25%) presented EZH2 variants: 14/70 in the R-bendamustine group (20%) and 68/254 in the R-CHOP group (27%) (P = 0.28). Variants were detected by SS in 36 (22%) and by NGS in 46 (28%) patients (P = 0.20). Patients treated with R-bendamustine were significantly older (median 67 vs. 58, P < 0.001). FL grade was categorized as Grades 1–2 or Grade 3A. No patients with grade 3B FL were identified. The distribution of FL grade was similar between treatment groups (P = 0.2). The percentage of patients who received R maintenance was similar between both groups (79% vs. 77%, P = 0.75).

Median follow-up was 6.3 years (95% confidence interval [CI] 5.1–7.5 years). Median PFS was 8.5 years (95% CI 6.2–10.8) and median OS was not reached. Six-year PFS was 59% (95% CI 53–65), and 6-year OS was 83% (95% CI 78–88). FLIPI 3–5 (hazard ratio [HR] 1.9 [95% CI 1.2–2.8]) and male sex (HR for female 0.6 [95% CI 0.4–0.9]) were independently associated with worse PFS. Only FLIPI 3–5 was independently associated with worse OS (HR 1.9, 95% CI 1.2–3.1). FL grade was not significantly associated with either PFS or OS. EZH2 variants did not have an impact on PFS or OS in the entire cohort.

In patients treated with R-CHOP, median follow-up was 6.75 years (95% CI 5.2–8.1). Median PFS was 9.4 (95% CI 6.25–12.6) years, and median OS was not reached. Six-year PFS was 60% (95% CI 53–67) and 6-year OS was 85% (95% CI 80–90). Multivariate analysis identified male sex (HR 1.78, 95% CI 1.17–2.72; P = 0.007) and FLIPI 3–5 (HR 2.18, 95% CI 1.38–3.42; P < 0.001) as independent predictors of worse PFS. Age >60 was the only variable significantly associated with worse OS (HR 4.0, 95% CI 1.8–8.5; P < 0.001). EZH2 mutation status was not associated with differences in PFS (6-year PFS 60% for both mutated and wild-type; P = 0.87) or OS.

In patients treated with R-bendamustine, median follow-up was 3.1 years (95% CI 2.2–4). Median PFS was 7.8 (95% CI 0–15.8) and median OS was 8.64 (95% CI 7.5–9.7) years. Mutated EZH2 was associated with worse PFS both in univariate and multivariate analyses (HR 2.59 [95% CI 1.12–6]). Six-year PFS was 37% in EZH2 mutated versus 59% in EZH2 wild-type (P = 0.021) (Figure 1A). A multivariate analysis for OS was performed including the only variable that was significant in the univariate analysis (age >60 years). Age >60 years remained associated with inferior OS in the multivariate model, although the association did not reach statistical significance (HR 7.2, 95% CI 0.95–54.8; P = 0.056).

In EZH2 wild-type patients (n = 241), no significant difference in 6-year PFS was found between those treated with R-bendamustine versus R-CHOP. Worse PFS was associated with age >60 (73.9, 95% CI 51.4–96.4 vs. 141 months, 95% CI 101–180; P = 0.01) and FLIPI > 2 (68.7f, 95% CI 41.1–96.3 vs. 21 months, 95% CI 82.2–160.9; P = 0.003). In univariate analysis, worse OS was associated with age >60, presence of B symptoms, and R-bendamustine treatment.

In patients with mutated EZH2 (n = 82), 6-year PFS was significantly worse when treated with R-bendamustine compared with R-CHOP (37% vs. 60%; P = 0.026) (Figure 1B).

Fifty-seven patients experienced disease progression within 24 months (POD24). Median OS was not reached in this group, while median PFS was 12.2 months (95% CI, 11.8–12.6) and was significantly shorter in EZH2-mutated compared with EZH2 wild-type patients (9.9 months [95% CI, 9.0–10.7] vs. 12.5 months [95% CI, 11.1–13.8], respectively; P = 0.048).

We additionally compared response quality and clinically relevant outcome measures according EZH2 mutational status and treatment, including complete response (CR) rate, relapse after remission, progression of disease within 24 months (POD24), and histologic transformation. EZH2-mutated patients treated with R-bendamustine showed numerically lower CR rates (69% vs. 75%), a higher proportion of progressive disease (15% vs. 5%), more POD24 events (36% vs. 19%), and more histologic transformations (17% vs. 5%) than those treated with R-CHOP, although none of these differences reached statistical significance. Conversely, among EZH2 wild-type patients, relapse after remission was numerically less frequent with R-bendamustine than with R-CHOP (25% vs. 37%), without statistically significant differences.

In this multicenter validation cohort, we extended our previous single-center observations confirming the treatment-specific predictive value of EZH2 mutation status in frontline FL. Patients with FL harboring EZH2 variants experience significantly shorter PFS when treated with R-bendamustine versus R-CHOP. In contrast, EZH2 variants did not negatively impact outcomes in patients receiving R-CHOP. These findings are consistent with the biological role of EZH2 in FL, where gain-of-function mutations promote aberrant germinal center maintenance and treatment resistance.¹⁸

Earlier clinical observations, including those described by Martínez-Laperche et al.¹⁰ in our original single-center study and those reported by Jurinovic et al.¹⁹ in the GALLIUM trial, suggested that EZH2-mutated FL may derive greater benefit from anthracycline-containing regimens than from R-bendamustine.

Recently, Passerini et al.²⁰ evaluated the predictive value of EZH2 and CREBBP mutations in patients with FL treated within the GALLIUM trial and in an independent validation cohort. The authors identified a molecular subgroup defined by CREBBP mutation and EZH2 wild-type status that appeared to derive greater benefit from bendamustine-based immunochemotherapy than from CHOP/CVP-based treatment, whereas no significant difference in outcome was observed in the other molecular subgroups. These findings, together with our results showing superior outcomes with R-CHOP compared with R-bendamustine among EZH2-mutated patients, support the potential role of epigenetic modifiers as biomarkers for treatment selection in FL. However, further prospective validation is needed before these biomarkers can be used to guide frontline treatment decisions.

Although the mechanisms by which EZH2 influences treatment response remain incompletely understood, preclinical data suggest that activating EZH2 mutations may contribute to immune evasion by repressing antigen-presentation pathways, including MHC Class I and II expression.²¹ In lymphoma models, CHOP-like chemotherapy, but not bendamustine, was able to restore the loss of MHC-I/MHC-II expression associated with the EZH2 Y641 mutation.²² These findings provide a plausible biological rationale for the distinct outcomes observed in EZH2-mutated patients treated with R-CHOP versus R-bendamustine, although this interpretation remains hypothesis-generating and requires further validation. Our results reinforce the concept that treatment intensity or chemotherapy backbone may interact with EZH2-driven epigenetic alterations, supporting the potential clinical utility of EZH2 status in regimen selection.

The present study shows some limitations. BCL2 rearrangement status was not systematically available, as FISH testing was performed according to local practice, precluding potential meaningful analyses. Molecular profiling was restricted to EZH2 status, without assessment of other potentially relevant mutations. Its retrospective design may introduce selection bias, and the limited number of patients EZH2-mutated reduced the power of subgroup analyses. The smaller number of patients treated with R-bendamustine reflects real-world practice in Spain and may also have limited statistical power.

Patients were monitored according to the standard clinical practice of each participating center, and imaging studies were performed at the discretion of the treating physician. Data on the type, timing, and frequency of imaging assessments were not systematically collected. Therefore, variability in follow-up and imaging schedules may have influenced the timing of progression detection, representing an inherent limitation of the retrospective multicenter design of the study.

However, the consistency of results across univariate and multivariate analyses supports the robustness of the association between EZH2 status and PFS in this subgroup. Future studies incorporating liquid biopsy may help to non-invasively assess EZH2 mutational status over time.

In conclusion, EZH2 mutation status is a clinically relevant, treatment-specific, and available predictive biomarker in frontline FL. While it did not affect outcomes with R-CHOP, EZH2 mutations were associated with inferior PFS in patients treated with R-bendamustine. These findings suggest that EZH2-mutated patients may preferentially benefit from anthracycline-containing regimens and, pending prospective validation, support a role for EZH2 status in informing frontline treatment selection.

Carolina Martínez-Laperche: Conceptualization (equal); investigation (equal); methodology (equal); writing—review and editing (equal); supervision (equal); project administration (equal). Pilar Lancho-Lavilla: Writing—original draft (equal); writing—review and editing (equal); supervision (equal); investigation (equal). Antonio Gutiérrez: Data curation (equal); formal analysis (lead); writing—review and editing (equal); investigation (equal). Antonio Salar: Writing—review and editing (equal); investigation (equal). Javier López-Jiménez: Investigation (equal); writing—review and editing (equal). Blanca Ferrer-Lores: Investigation (equal); writing—review and editing (equal). Alicia Serrano: Investigation (equal); writing—review and editing (equal). Ana Jiménez-Ubieto: Investigation (equal); writing—review and editing (equal). María Poza: Investigation (equal); writing—review and editing (equal). Francisco Javier Díaz-Crespo: Investigation (equal); writing—review and editing (equal). Paula Fernández-Caldas: Investigation (equal); writing—review and editing (equal). Laura Sanz-Villanueva: Investigation (equal). Javier Menárguez-Palanca: Investigation (equal); writing—review and editing (equal). Ismael Buño: Investigation (equal). María García-Álvarez: Investigation (equal); writing—review and editing (equal). Paula Muñiz: Investigation (equal). Pablo Silva-de-Tena: Investigation (equal); writing—review and editing (equal). Ramón García-Sanz: Investigation (equal); writing—review and editing (equal). Miguel Alcoceba: Investigation (equal); writing—review and editing (equal). Mariana Bastos-Oreiro: Investigation (equal); conceptualization (equal); writing—review and editing (equal); methodology (equal); project administration (equal).

The authors declare no conflicts of interest.

This research received no external funding.

The data that support the findings of this study are available from the corresponding author upon reasonable request.