Methylation-guided stratification of patients with high-risk prostate cancer for the prediction of clinical failure.

Abstract

155 Background: To explore differential DNA methylation of APC, CCDN2, GSTP1, PTGS2, and RARBas a tool to predict clinical failure (CF) in surgically treated patients with high-risk prostate cancer (PCa), as defined by a clinical stage T3a or higher, a biopsy Gleason score of 8 to 10 and/or pre-operative prostatic serum antigen levels more than 20 ng/ml. Methods: A quantitative multiplex, methylation-specific PCR assay was developed to assess promoter methylation of the marker genes. Formalin-fixed, paraffin-embedded radical prostatectomy samples from 218 patients with high-risk PCa, classified in a training and validation cohort, were analyzed. Kaplan-Meier analysis and Cox multivariate models were used. Results: First, high-risk PCa patients were ranked according to their GSTP1 methylation level, which varied from 0 to 80% in both cohorts. Next, we categorizedthe tumors into three groups based upon the GSTP1 methylation level: low methylation (LM, methylation less than 15%), moderate methylation ([MM], methylation 15 to 50%) and high methylation (HM, methylation more than 50%). When all five marker genes were combined, the median methylation values in the LM groups did not exceed 6% in both cohorts. In contrast, the median methylation values for the marker genes in the HM groups reached 45% in the training cohort and 50% in the validation cohort. Finally, the median methylation values in the MM groups amounted to 18% and 28% for the training and validation cohorts, respectively. Remarkably, patients belonging to the LM, HM, and the combined LM+HM groups had a 2- to 11-fold higher risk of CF as compared to those of the MM group, when correcting for established clinico-pathologic prognostic factors. In addition, the combined LM+HM group showed a significant association with cancer-related death in a multivariate analysis (Hazard ratio, 3.59) compared to the MM group, when both cohorts were combined for analysis. Conclusions: We found that differential methylation of GSTP1 allows the stratification of high-risk PCa patients into three molecular episubtypes, i.e. low, moderate and high DNA methylation level groups. Surprisingly, our data revealed that both low and high methylation levels of GSTP1 are independent predictors of CF.