High Incidence of EZH2 Mutations with Variable Mutation Load in Follicular Lymphoma and Its Consequences for EZH2 Targeted Therapy

Abstract 545

The tumor genetics of follicular lymphoma (FL) is unique in the field of cancer with almost all cases having one or more mutations in components of the epigenome. Our previous studies demonstrated that DNA hypermethylation occurs at 7% of loci in FL and these are specifically enriched for targets of the Polycomb Repressive Complex 2 (PRC2). This complex includes the histone methyltransferase EZH2, responsible for transcriptional repression via trimethylation of lysine 27 on histone H3 (H3K27me3). This link between DNA and histone methylation was confirmed by identification of gain of function mutations in EZH2 (Morin et al, Nature Genetics 2010), which leads to elevated levels of H3K27me3. As EZH2 inhibitors are currently in development, we assessed EZH2 mutation status and H3K27me3 deregulation in a large series of 367 FL cases (238 diagnosis, 129 relapse) as a first step in determining which patients may be best suited to targeted EZH2/H3K27me3 therapies.

Next-generation sequencing (NGS) was initially performed for all 27 EZH2 exons of 50 FL cases and this confirmed mutations at the 3 codons Y646, A682, and A692 previously reported (Morin et al, Nature 2011). Targeted resequencing was therefore restricted to these 3 codons and performed on whole tumor DNA from 367 FL cases by both conventional Sanger and NGS using the Roche 454 platform (at least 200-fold coverage). Sixty-three EZH2 mutations (17%) were detected using both approaches with 51 mutations targeting codon Y646, 7 mutations at codon A682 and 5 mutations at codon A692 at a mean mutation load of 29.8% (range: 4–61%). Critically, deep-sequencing enabled detection of an additional 43 mutations with an average mutational load of 10.2% (range: 2–31%) increasing the overall mutation rate to 29%. Multiple EZH2 mutations were detected in 6 cases; these typically had low mutation loads e.g. Case 1: Y646F (10%) and Y646N (5%); Case 2: Y646N (3.5%), Y646S (4.5%) and A692V (2%) and were present in both diagnostic and relapsed cases. There was no difference in the mutation frequency at diagnosis and relapse. Of note, we also detected additional rare novel heterozygous variants K634E and V679M within the SET domain, both of unknown significance.

It is plausible that the reported mutation load may reflect variations in the contamination of tumor by infiltrating normal cells, or alternatively true genetic heterogeneity with EZH2 mutation confined to a minor population of tumor cells. We made use of our previously generated methylation data from 164 FL patients (115 wt; 49 mutated), to rank biopsies based on their similarity to benign lymph node controls as an estimate of tumor content. The majority of cases with low EZH2 mutation load corresponded to cases with low tumor burden in the biopsy sample, therefore representing genuine clonal events. However, there are also a small but significant number of cases where the mutations represent examples of genetic heterogeneity. This is also consistent with mutation profiling in this series where 7 EZH2 mutated cases with a low mutation load (12%) harbor clonal mutations in TNFRSF14, another gene frequently mutated in FL, suggesting that, at least in these cases, EZH2 is not a tumor initiating event, but acquired during the evolution of the disease

We next performed immunohistochemistry on tissue microarrays generated from a subset of 120 FL cases (86 wt; 34 mutated) to determine the suitability of H3K27me3 staining as a surrogate biomarker for EZH2 mutation. Overall, 45% of cases demonstrated high H3K27me3 levels that included both wild type and EZH2 mutated biopsy samples. This suggests that raised H3K27me3 levels are not restricted to mutated cases and therefore EZH2-directed treatments may also be a rational approach in certain populations of EZH2 wild type patients.

In summary, our experiments demonstrate higher prevalence of EZH2 mutations in FL then previously reported and highlight some of the challenges faced in developing the appropriate companion diagnostics tools to parallel the testing of novel therapies targeting EZH2/H3K27me3.