A DNA Hypermethylation Profile Independently Predicts Biochemical Recurrence Following Radical Prostatectomy

Purpose: Detection of DNA hypermethylation is emerging as a novel molecular biomarker for different malignancies. We intend to define whether a hypermethylation profile of patients with prostate cancer (PCa) predicts biochemical recurrence (BCR) after radical prostatectomy (RP). Material and Methods: Genome-wide methylation analysis was performed using the GoldenGate Methylation Cancer Panel-I (II-lumina, Inc.) on 10 normal prostate tissues and 58 tumor samples from patients treated by RP followed for prostate-specific antigen (PSA) failure (>0.4 ng/ml) and disease progression. Patients were classified on the basis of D'Amico criteria according to clinical staging, PSA at diagnosis and Gleason score after pathologist review. Hypermethylation status of 1505 CpGs present in the promoter region of 807 genes was studied. Hierarchical clustering analysis was performed and relationships with outcome were investigated using log-rank analysis and Cox regression model. Results: We found 28 genes significantly hypermethylated in > 20% of the tumors analyzed. Four clusters of patients were characterized by their DNA methylation profile, one at higher risk to develop BCR (p = 0.005). Multivariate analysis revealed patients in this cluster (HR 2.56), and high-risk patients (HR 4.34) according to D'Amico classification were independent predictors of BCR after prostatectomy. From the selected genes MT1A, ALOX12, GSTM2, APC, MYCL2 and RARB hypermethylation predicted BCR and GSTM2 (HR 3.78) and MYCL2 hypermethylation (HR 2.71) did so independently. Conclusion: Epigenetic silencing of GSTM2 and MYCL2 comprise novel molecular markers to predict BCR after surgery for medium and high-risk localized PCa undergoing surgical treatment and hypermethylation of these genes could be incorporated to the clinical and pathological factors defining the patient at higher risk of PSA failure after prostatectomy. The limitation of the study is that no independent validation cohort is analysed. (C) 2016 S. Karger AG, Basel