Tumor Epigenetic Signature and Survival in Resected Gastric Cancer Patients

Abstract

Precision oncology can identify patient-specific molecular signatures to better inform the prognosis and management of surgical cancer patients. Specifically, microRNAs (miRs) hold promise as prognostic biomarkers because dysregulation of individual miRs is implicated in tumorigenesis, progression, and metastases of various malignancies, including gastric adenocarcinoma (GC). To identify miRs prognostic of survival after radical gastrectomy, we studied GC patients within The Cancer Genome Atlas (TCGA) who had undergone R0 or R1 resection and had data on clinical characteristics, overall survival (OS), and tumor miR expression. The miRs expressed by at least 15% of tumors were eligible for study. From 10 replicate samples, each with 80% of patients, miRs were selected using age-adjusted proportional hazards regression with stepwise selection. Cross-validated miRs (selected by multiple replicates) were retained if they optimized an accelerated failure-time model of OS using all patients. In this GC cohort (n= 270), half (916/1,870) of miRs screened met our criteria for evaluation. Cross-validation identified 20 miRs as prognostic, of which 14 (miR-129-1, miR-373, miR-490, miR-597, miR-1185-2, miR-3943, miR-4756, miR-5683, miR-6510, miR-6733, miR-6808, miR-6855, miR-6882, miR-8072) were independently informative. The age-adjusted 14-miRNA panel remained significantly associated with OS after adjustment for pathologic prognostic factors (number of lymph nodes examined, number of positive lymph nodes) and other clinical covariates (TNM stage, residual tumor, tumor microsatellite instability, targeted molecular therapy, sex, race, ethnicity). Panel-predicted survival estimates below the upper tertile cut-off were associated with worse outcome (30% vs 74% OS at 3 years, p < 0.0001). In surgically resected GC patients, an epigenetic signature of miRs associated with survival has the potential to improve prognostication.