Our findings confirm and extend prior work regarding the prevalence of EZH2
mutations in B cell lymphomas and the molecular function of these mutations. Clinical detection of oncogenic point mutations conferring enzymatic gain-of-function has been a fruitful strategy for triaging patients to molecularly targeted therapy in several cancer types, including lung adenocarcinoma (EGFR
and melanoma (BRAF
V600E mutations). 
While specific small-molecule inhibitors of EZH2 are not yet clinically available, there is ongoing interest in the development of such agents. 
Our SNaPshot-based assay reproducibly detected EZH2
mutations in a substantial proportion of germinal center lymphomas from routine FFPE clinical specimens. The absence of EZH2
mutations in BL highlights both the biological specificity of the EZH2
codon 641 mutation for certain lymphoma subtypes, and the technical specificity of our assay.
In comparison to prior studies, we detected a relatively high rate of EZH2
mutations in FL (22% overall and 27% in the common grade 1–2 subtype) compared to prior reports (7% 
and 12% 
overall). This may be due, in part, to the higher sensitivity of SNaPshot compared to Sanger (chain terminator) sequencing for detection of mutations in heterogeneous tumor tissue. 
Our results are consistent with those of a recent study which used deep-sequencing techniques (genome, exome, and RNA sequencing) on a small number of follicular lymphomas and reported EZH2
mutations in 4/13 cases (31%). 
Thus, the true rate of EZH2
mutations in follicular lymphoma may be higher than originally reported. EZH2
mutations were not enriched in more proliferative subgroups of FL (low grade/high PI FL and grade 3 FL), suggesting that EZH2
mutations are not associated with known aggressive subtypes of FL. Larger studies with clinical follow-up will be required to determine whether this mutation carries independent prognostic significance in FL.
In contrast to the findings in FL, we detected mutations in a relatively low proportion of GCB-DLBCL (14%) compared to the original report (22%) 
. Our use of an immunophenotypic surrogate for the GCB-DLBCL gene expression profile could partly explain this discrepancy, although our findings are similar to those of another recent study, which found EZH2
mutations in 14% of a cohort of 63 microarray-defined GCB-DLBCL. 
Interestingly four of the five EZH2
mutations detected in the GCB-DLBCL cohort occurred in a subgroup of nine cases in which the diagnostic biopsy showed a large cell lymphoma with areas of both follicular and diffuse growth, resulting in concurrent diagnoses of DLBCL and grade 3 FL by current WHO criteria. 
None of these nine patients had a prior history of follicular lymphoma, and these cases appear to represent de novo
high-grade germinal center phenotype lymphomas with partial residual folliculotropism. This finding raises the possibility that EZH2
-mutated DLBCL may have characteristic morphological and biological features, and merits further investigation.
codon 641 mutation was first identified through whole-genome sequencing of a follicular lymphoma that lacked a t(14;18) IGH
However, we found frequent, though not universal, co-occurrence of EZH2
codon 641 mutations with rearrangements of BCL2
. This implies that EZH2
mutations do not substitute for BCL2
rearrangement, but rather represent a functionally distinct oncogenic “hit” in germinal center B cell lymphomagenesis. Targeted inhibitors of BCL2 have been developed, with some advancing into clinical trials. 
Our finding that many lymphomas show oncogenic activation of both BCL2 and EZH2 raises the possibility that combined BCL2 and EZH2 inhibition may represent a potential therapeutic strategy for many patients.
BL is a prototypical germinal center-derived lymphoma, but shows substantial biologic, pathogenic, and epidemiologic differences from FL and DLBCL. 
Our findings suggest that EZH2
mutations are rare or absent in this lymphoma type. We did detect EZH2
mutations in MYC
-rearranged “double hit” high grade B cell lymphomas, which often show morphologic, genetic, and immunophenotypic features intermediate between DLBCL and BL, and are associated with a poor prognosis 
. Large-scale integrative studies including gene expression profiling have suggested that the relative paucity of large-scale genomic imbalances (“genetic complexity”) can be helpful in distinguishing “true” BL from other high-grade B cell lymphomas. 
Our results suggest that EZH2
mutation status could also potentially contribute to this distinction.
The results of our mutant EZH2 expression experiments indicate that the mutant protein conferred a gain-of-function phenotype as evidenced by increased generation of the H3K27me3 histone mark, the mechanism of which has been demonstrated by others. 
Interestingly, the mutation that was most common in our series, Y641N, was estimated to be the most potent of the four common mutations at generating H3K27me3 in the heterozygous state, based on in-vivo
enzyme kinetics, 
while Y641H, the least common mutation in our series, was predicted to be least potent. These results suggest that more potent mutations may undergo stronger oncogenic selection. Our novel demonstration of a similar gain-of-function activity for the Y641C mutant in an overexpression assay is consistent with the reported occurrence of this mutation in B cell lymphoma, 
in which most EZH2
mutations are thought to confer gain-of-function. However, the Y641C mutation has also been reported in an acute myeloid leukemia cell line, despite the fact that many reported EZH2
mutations in myeloid tumors are clearly inactivating. 
These findings suggest the possibility of a more complex function for some myeloid EZH2
mutations, and highlight the need for further functional investigation of these mutant proteins. While the H3K27me3 modification is known to have a repressive effect on gene expression, the precise genomic targets that are affected by the mutant protein during lymphomagenesis and confer selective advantage have not been identified.
For our transgenic work, we used a cDNA clone for EZH2
with a coding sequence matching that of the original reported EZH2
coding sequence 
Refseq isoform C (NM_001203247). Our results suggest that this is the predominant long splice form of EZH2
in lymphoma and other tissues, and for this reason we prefer to designate the EZH2
codon affected by recurrent mutations in B cell lymphoma as Y641, rather than Y646. Shorter EZH2
transcripts produced by variant splicing of exons 3 and 4 are also well supported by ESTs and other public data, and are of uncertain significance.
In conclusion, we have developed a clinically applicable assay for sensitive detection of EZH2 codon 641 mutations in FFPE tissue, and demonstrate frequent occurrence of these gain-of-function mutations in the full biological spectrum of BCL2-rearranged germinal center phenotype B cell lymphomas, but not in Burkitt lymphoma.