Abstract:Background: Currently there are no biomarkers to identify resistance to androgen-deprivation therapy (ADT) in men with hormone-naive prostate cancer. 5-hydroxymethylcytosines (5hmC) in the gene body are associated with gene activation and are critical for epigenomic regulation of cancer progression. Objective: To evaluate whether 5hmC signature in cell-free DNA (cfDNA) predicts early ADT resistance. Design, Setting, and Participants: Serial plasma samples from 55 prostate cancer patients receiving ADT were collected at three timepoints including baseline (prior to initiating ADT, N=55), 3-month (after initiating ADT, N=55), and disease progression (N=15) within 24 months or 24-month if no progression was detected (N=14). 20 of the 55 patients showed disease progression during the 24-month follow-up. The remaining 35 patients showed no progression in the same follow-up period. Outcome Measurements and Statistical Analysis: cfDNA (5-10ng) was used for selective chemical labeling (hMe-Seal) sequencing to map 5hmC abundance across the genome. Read counts in gene bodies were normalized with DESeq2. Differential methylation and gene set enrichment analyses were performed to identify the 5hmC-enriched genes and biological processes that were associated with disease progression. Kaplan-Meir analysis was utilized to determine the association of 5hmC signatures with progression-free survival. Results and Limitations: 5hmC-sequencing generated an average of 18.6 (range 6.03 to 42.43) million reads per sample with 98% (95-99%) mappable rate. Baseline sample comparisons identified significant 5hmC difference in 1,642 of 23,433 genes between 20 patients with progression and 35 patients without progression (false discovery rate, FDR<0.1). Patients with progression showed significant enrichments in multiple hallmark gene sets with androgen responses as the top enriched gene set (FDR=1.19E-13). Interestingly, this enrichment was driven by a subgroup of patients with disease progression featuring a significant 5hmC hypermethylation of the gene sets involving AR, FOXA1 and GRHL2. To quantify overall activities of these gene sets, we developed a gene set activity score algorithm using a mean value of log2 ratios of gene read counts in an entire gene set. We found that the activity scores in these gene sets were significantly higher in this subgroup of patients with progression than in the remaining patients regardless of the progression status. Furthermore, the high activity scores in these gene sets were associated with poor progression-free survival (p <0.05). Longitudinal analysis showed that activity scores in this subgroup with progression were significantly reduced after 3-month ADT but returned to high levels when the disease was progressed. Conclusions: 5hmC-sequencing in cfDNA identified a subgroup of prostate cancer patients with preexisting activation (5hmC hypermethylation) of gene sets involving AR, FOXA1 and GRHL2 before initiating ADT. Activity scores in these gene sets may serve as sensitive biomarkers to determine treatment resistance, monitor disease progression and potentially identify patients who would benefit from upfront treatment intensification. More studies are needed to validate this initial finding.

5-hydroxymethylcytosine sequencing in plasma cell-free DNA identifies unique epigenomic features in prostate cancer patients resistant to androgen deprivation therapy