Preliminary evidence of a molecular predictor of tazemetostat response, beyond ezh2 mutation, in nhl patients via characterization of archive tumor and circulating tumor dna

Introduction: B-cell malignancies may depend on the histone methyl transferase EZH2 to perpetuate a less differentiated state, with activating mutations (MT's) of EZH2 being potential oncogenic drivers. Tazemetostat, a potent, selective EZH2 inhibitor, is in phase 2 clinical development in relapsed or refractory (RR) NHL. Objective responses were observed in patients (pts) with MT or wild type (WT) EZH2 tumors in phase 1. The ongoing phase 2 study enrolls pts with MT or WT EZH2 RR DLBCL or FL to determine efficacy and safety (reported separately). The primary endpoint is overall response rate (ORR). Here, we report results of a molecular analysis of patient tumor material and associations with preliminary response data, including the discovery of a novel candidate molecular predictor of tazemetostat response. Methods: Archive tumor and/or plasma derived circulating DNA (ctDNA) samples were obtained during screening in the phase 2 trial of tazemetostat in NHL (NCT01897571). Prospectively, archive tumor was analyzed for EZH2 hot spot MT's Y646X, A682G, and A692V using a cobas EZH2 mutation test (Roche Molecular Systems, in development). Retrospectively, next generation sequencing (NGS) was performed on archive tumor DNA (target coverage of 1500X) and ctDNA (20 000X for somatic MT's and 5000X for structural alterations) to identify somatic MT's, amplifications and translocations in a panel of 62 genes commonly mutated in NHL. Best ORR on January 26, 2017, was used to generate 2 groups, responders (CR + PR n = 25), and nonresponders (PD n = 43). Gene alterations associated with either group were identified using Fischer's exact test, and logistic regression modelling was used to identify multi gene predictors of response. Results: Regardless of sample type or technology, the concordance rate for detection of EZH2 status was >95%. NGS and cobas testing of archive tumor samples was 100% concordant (n = 92) with 11 EZH2 MT cases detected. Concordance of EZH2 status between archive tumor and ctDNA samples was 97% (n = 125). EZH2 mutation detection rates in archive tumor were 9% in DLBCL and 21% in FL, consistent with previous reports. MT's in EZH2 or MYD88 in WT EZH2 pts were associated with response (P < .1) whereas MT's in HIST1H1E or MYC (P < .08) were associated with nonresponse. Pts matching a multi gene predictor consisting of WT MYC and/or HIST1H1E but with MT STAT6 and/or MYD88 in archive tumor had ORR = 53% (10/19) whereas pts who did not match this profile had ORR = 16% (12/73) indicating potential for these genes to predict response to tazemetostat. Conclusions: Molecular genetic profiling of NHL pts identified potential predictors of response beyond EZH2 MT and offered new insights into mechanisms of response in EZH2 WT pts. Plasma-based ctDNA screening may be a viable method to identify NHL pts with EZH2 MT's in the absence of archive tumor samples. Keywords: diffuse large B-cell lymphoma (DLBCL); EZH2; follicular lymphoma (FL)