Epigenomic alterations associated with androgen receptor pathway inhibitor response in metastatic castration-resistant prostate cancer.

Abstract

220 Background: Androgen receptor pathway inhibitors (ARPIs) improve survival in patients with metastatic castration-resistant prostate cancer (mCRPC). However, some patients fail to respond, and resistance frequently emerges. While AR amplification and mutation are established genetic contributors to heterogeneous therapy response, the role of epigenetic mechanisms in modulating ARPI response remains poorly understood. In this study, we evaluated methylation features associated with duration of ARPI response in mCRPC and developed a predictive classifier to identify patients likely to progress within one year of treatment. Methods: We analyzed matched DNA methylation, genomic, and transcriptomic profiles from 117 mCRPC biopsies in the West Coast Dream Team cohort. Global methylation patterns were assessed by identifying highly variable recurrent hypomethylated regions, followed by unsupervised hierarchical clustering. In ARPI-naïve, AR -positive tumors (n = 36), differential methylation analysis was performed between patients in the top and bottom quartiles of radiographic progression-free survival (rPFS; n = 18 total) to maximize contrast in treatment response. Matched transcriptomic data were integrated to identify methylation changes correlated with gene expression. We then trained a random forest classifier using methylation profiles from all 36 samples to predict patients who would experience radiographic progression within one year of ARPI initiation. Results: Hierarchical clustering revealed global methylation patterns segregated by molecular subtype, with neuroendocrine tumors and those exhibiting a CpG methylator phenotype forming a distinct cluster. In multivariable Cox models adjusted for AR and AR upstream enhancer copy number, methylation of the downstream regulatory region was associated with longer rPFS (per 1%-unit increase: HR = 0.96, 95% CI: 0.93–0.99, p = 0.024). When dichotomized at the median, patients with high methylation showed significantly improved rPFS compared to those with low methylation (HR = 0.18, 95% CI: 0.07–0.44, p = 0.00023). Importantly, we have previously shown this region undergoes DNA amplification upon ARPI progression (Zhu et al., JCI 2024). Our random forest classifier distinguished good vs. poor outcomes on ARPI, achieving an F1 score of 0.86. Conclusions: We identified methylation changes and genomic regions, including a putative enhancer downstream of AR , associated with ARPI response. We also developed a predictive classifier that accurately identified patients who progressed more rapidly on ARPI therapy. Validation experiments for this classifier and association analyses are ongoing. These findings highlight a novel link between epigenomic alterations and ARPI response and support the development of methylation-based biomarkers for patient stratification in mCRPC.