EZH2 gene mutations and copy number variation as prognostic markers in advanced follicular lymphoma

Introduction. Among the most common genetic alterations in follicular lymphoma (FL) there are the gain-of-function mutations in the Enhancer of Zeste Homolog 2 (EZH2) gene, encoding for a histone methyltransferase, which can also be over-expressed by copy number variations (CNV). The clinical impact of EZH2 aberrations in FL is debated, with some studies suggesting a favourable outcome, especially when patients received R-CHOP/CVP. Aims. In the “FOLL-EZ” study we aimed to explore the role of EZH2 aberrations in modifying the treatment effect of R- CHOP vs bendamustine-rituximab (BR) in the prospective trial “FOLL12” of the Fondazione Italiana Linfomi (FIL, EudraCT 2012-003170-60), enrolling advanced FL patients treated with frontline R-CHOP or BR, according to physician’s choice. Methods. Leftover DNA samples from unsorted bone marrow (BM) aspirates at baseline, centralized at the FIL MRD Network labs, were tested by a multiplex droplet digital PCR (ddPCR) for the detection of the recurrent EZH2 mutations and by a ddPCR CNV assay for the EZH2 region (chr 7q36.1). Results. Overall, 409 BM samples were tested for mutations and 359 for CNV (337 for both). An EZH2 mutation was found in 137 cases (33%), concordantly with literature, being more frequent in patients with a documented histological infiltration at the BM biopsy (BM+ n=124/309, 40%) than in BM- (n=13/100, 13%). In detail 68 cases were mutated in Y646 (50%), 41 in A682G (30%) and 73 in A692V (53%). 102 cases (74%) showed one single mutation: 52 in Y646, 13 in A682G, 37 in A692V. 6 cases showed all 3 mutations, 21 cases showed mutation in both 682 and 692 and 9 samples in 646 and 682 or 692. The most common mutation in Y646 was the Y646N 38/68 (56%) with a variant allele frequency (VAF) of 0.25% (0.03-32%), the less frequent the Y646S 8/68 (12%), VAF 0.59% (0.03- 21%). Notably 8/68 Y646 cases (12%) carried 2 distinct mutations and only one sample showed 3 distinct mutations. Interestingly, the 137 EZH2 mutated patients showed less favourable clinical features than the WT, in terms of higher B2M (>ULN 72% vs 53%, p<0.001), lower Hb values (Hb<12, 26% vs 15%, p=0.015), more frequent BM+ (91% vs 68%, p<0.001) and higher FLIPI-2 (61% HR vs 40%, p<0.001); also baseline PET parameters were more unfavorable among EZH2 mutated cases, namely metabolic tumor volume (MTV) >200 ml (71% vs 57%, p=0.011) and maximum tumor dissemination (Dmax) >400 mm (34% vs 21%, p=0.013): nonetheless, this did not translate into a statistically significant worse outcome (5-year PFS 59% vs 64%, p=0.324). Overall, no difference in PFS was observed between R-CHOP (n=239) and BR-treated (n=170) patients (5-y PFS 60% vs 67%, p=0.652): EZH2 mutations had no statistically significant impact in modifying the treatment effect of either therapeutic schedule (HR R-CHOP vs BR: 1.77 (95CI 0.96-3.29) p=0.067 for mutated and 0.89 (95CI 0.60-1.32), p=0.564 for WT patients, Figure 1A). In a subgroup analysis, the 30 EZH2 mutated patients who received R-CHOP not followed by R maintenance experienced the worst outcome (5-y PFS 24%, p=0.007). Regarding EZH2 CNV, a CN gain was found in 50 cases (14%) and a CN loss in 9 (3%): among the 337 patients analysed for both mutations and CNV, 13 carried both EZH2 mutation and gain (4%), 106 had mutations only (31%) and 31 had CN gain only (9%). The 50 patients with CN gain had overall less BM involvement (62% vs 79%, p=0.018) but no significant differences in other baseline clinical features nor in outcome; interestingly, among the 199 patients treated with R-CHOP those with CN gain only (n=20) showed a more favorable outcome than the others (5-y PFS: 75% vs 59%, p=0.053, Figure 1B). Conclusions. EZH2 mutations and CNV were tested by ddPCR in 409 BM samples from the largest, prospective series of FL: they had no clear impact in modifying the treatment effect (R-CHOP/BR), so a modulation of the first line treatment choice in FL based on EZH2 does not seem appropriate, so far. To better address this issue, mutational and CNV testing ofdiagnostic lymph node samples is currently ongoing.