DNA methylation signature predictive of benefit from neoadjuvant chemotherapy in esophageal adenocarcinoma: Results from the MRC OEO2 phase III trial.

Abstract

43 Background: Platinum and 5-Fluorouracil (5FU) neoadjuvant chemotherapy followed by surgery is one of the standard approaches for patients with resectable EAC. To date, there are no predictive biomarkers of chemotherapy benefit. We hypothesize that DNA methylation of genes in key biologic and oncogenic pathways predict for chemotherapy benefit in EAC. Methods: In the OE02 trial, 802 patients with resectable esophageal carcinoma were randomised to surgery alone (S) versus two cycles of cisplatin and 5FU chemotherapy followed by surgery (CS). DNA was extracted from 213 EAC resection specimens (110 from the (CS) arm, 103 from the (S) arm). DNA methylation was analyzed at 1505 CpG sites within 807 genes using the Illumina GoldenGate platform. Cox proportional hazard analysis was performed to identify predictive markers of survival in (CS) arm; non-negative matrix factorization (NMF) was used to delineate methylation signatures. Results: Methylation status of 1505 CpG sites had no statistical difference between the (CS) and (S) arms. In the (CS) arm, 87 (5.7%) CpG sites were initially identified as promising candidates in univariate analysis (p < 0.05 cutoff). NMF generated a 4 CpG site signature which divided patients into poor risk and good risk. Genes involved in the signature include RUNX1T1, CCND2, MST1R and MMP14. Survival was significantly different between poor risk and good risk in (CS) arm (HR 0.32, 95% CI: 0.21 to 0.52, p < 0.0001). No difference in survival was detected in the surgery arm (HR 1.12, 95% CI: 0.76 to 1.80, p = 0.48), suggesting the signature served as a predictive and not prognostic biomarker. Methylation signature remained an independent predictor of survival in multivariate analysis with clinicopathologic factors (along with age and vascular invasion). Conclusions: Chemotherapy does not appear to change methylation status of EAC. Hypermethylation of RUNX1T1, CCND2 and hypomethylation of MST1R and MMP14 leads to significantly decreased benefit from chemotherapy in EA. We describe an epigenetic signature which may serve as a predictive biomarker for chemotherapy benefit using data form the largest bank of DNA methylation in EA reported to date.