Résumé
Introduction: B-cell malignancies may depend on the histone methyltransferase EZH2 to perpetuate a less differentiated state, with activating mutations of EZH2 being potential oncogenic drivers. Tazemetostat, a potent, selective EZH2 inhibitor, is in phase 2 clinical development in relapsed or refractory (R/R) non-Hodgkin lymphoma (NHL). Objective responses were observed in patients (pts) with EZH2 mutant or wild type tumors in the phase 1 part of the phase 1/2 study. The ongoing phase 2 study is enrolling pts with mutant or wild type EZH2 having R/R diffuse large B-cell lymphoma (DLBCL) or follicular lymphoma (FL) to determine efficacy and safety. The primary endpoint is overall response rate. Here we report results of an updated molecular analysis of archived tumor and circulating tumor DNA (ctDNA) collected from pts plasma and associations with preliminary response data, including the discovery of novel candidate molecular predictors of tazemetostat response.
Methods: Archived tumor and/or plasma-derived ctDNA samples were obtained during screening from R/R DLBCL and FL pts enrolled to the phase 2 trial of tazemetostat in NHL (NCT01897571). Next generation sequencing was performed retrospectively on archived tumor DNA (target coverage of 1,500X) and ctDNA (20,000X for somatic mutations and 5,000X for structural alterations). DNA was isolated from archived tumor and was tested across a panel of approximately 200 genes, including 62 genes commonly altered in NHL to identify somatic mutations, amplifications, and translocations. The ctDNA sequencing was restricted to the 62 gene NHL panel. Best objective overall response data (Cheson 2007) as of May 1, 2018 were used to generate two groups: responders (R = CR + PR), and non-responders (NR = progressive disease, stable disease, or unknown clinical response). Fisher's exact test (two tailed) was performed to identify DNA variants associated with either the R or NR groups for FL+DLBCL combined and for each NHL sub-type independently (Table 1). All gene associations met an unadjusted P-value threshold of ≤ 0.1.
Results: In the combined analysis for all NHL pts, EZH2, BRCA2, ETV6, IKZF3 and TNFRSF14 mutations were associated with clinical response. Eleven genes, including BCL6, BCL2, TP53, PIM1, HIST1H1B and HIST1H1E, mutations were associated with a lack of response to tazemetostat. Association of detection of mutations in EZH2, BCL6 and HIST1H1E and response were detected in both ctDNA and archived tumor analysis, with all other genes achieving P ≤ 0.1 in either the ctDNA or the archived tumor analysis. Analysis of the NHL subcategories revealed in FL that mutations in EZH2 and STAT6 were associated with clinical response while mutations in BCL2, TNFAIP3, FOXO1 and MYD88 were identified as negative predictors. EZH2 was the only gene identified in both the ctDNA and archived tumor analyses in FL. In DLBCL, while EZH2 was not detected as a predictor of response to tazemetostat at P ≤ 0.1, the number of pts with mutant rather than wild type EZH2 was higher in responders than non-responders in both analyses. Positive predictors of tazemetostat response were detected in DLBCL: MYD88, MEF2B, ETV6, MLH1, RECQL4, and RNF43. Negative predictors of response were also identified in DLBCL: BCL2, PDL1, PDL2 and SOCS1. When the genes identified as associated with tazemetostat response in DLBCL were compared between the archived tumor and ctDNA analyses, no genes in common were identified.
Conclusions: These analyses confirm genes previously described as potential predictors of tazemetostat response in analyses using smaller patient numbers and less mature clinical response data such as STAT6, EZH2, MYD88, TP53,BCL2, BCL6, PIM1 and HIST1H1E. In addition, identification of PDL1/PDL2 and RECQL4/MLH1/BRCA2 as novel predictors points to potential relevance of immune checkpoint and DNA damage repair pathways as additional pathways that may be of relevance to tazemetostat response in NHL. Subanalyses performed independently on each NHL subtype revealed disease specific biomarkers of response that may point to differences in the factors that can influence response to tazemetostat in DLBCL and FL.