Abstract A053: Epigenomic analyses of circulating tumor DNA in patients with metastatic castration resistant prostate cancer (mCRPC) treated with immune checkpoint blockade and PARP inhibition

Background: Molecular characterization of mCRPC has revealed disruption of DNA damage repair genes that may improve sensitivity to PARP inhibitors such as olaparib (O), which in turn may further complement the activity of the PD-L1 immune checkpoint inhibitor, durvalumab (D), through mechanisms such as activation of the STING pathway. We hypothesized that (epi)genomic analyses of circulating tumor DNA (ctDNA) would identify mechanisms of treatment response in patients with mCRPC treated with O plus D (NCT02484404). Methods: Eligible patients for this study had mCRPC with prior treatment by abiraterone and/or enzalutamide. Study participants were administered 300 mg of O twice daily and 1500 mg of D every 4 weeks. We obtained blood samples in Streck tubes from 38 (out of 60) study participants. Plasma samples were acquired at baseline and at progression. Buffy coat was isolated from each sample. Genomic DNA and cell-free DNA (cfDNA) were separately isolated and assembled into sequencing libraries. Low-pass whole-genome sequencing (3-4 × coverage) was performed for somatic copy number and fragmentomic analyses. In a subset of samples with the greatest cfDNA content, chromatin immunoprecipitation sequencing (ChIP-seq) with antibodies against H3K4me2 and H3K4me3 was performed directly in additional aliquots of plasma. Results: Amongst the 38 patients with at least one cfDNA sample, 5 patients had a radiographic partial response (PR) per RECIST v1.1, 28 patients had stable disease (SD), and 6 patients had progressive disease (PD) during the first two months of treatment. At baseline, ctDNA content in plasma was significantly lower in patients who went on to have PR or SD versus PD (p=0.02). However, upon progression, ctDNA content in plasma was not significantly different between groups. Losses to PTEN and NKX3-1 were significantly enriched in ctDNA from patients with PD (p=0.041 and p=0.047, respectively). By contrast, biallelic combined losses to BRCA2, RB1, BRCA1 and TP53 were observed in all PRs at baseline but not in any patient with PD (p=0.048). In patients with SD, longitudinal fragmentomic analysis inferred that the activity of the mitogenic KLF4 transcription factor (TF) was the most increased upon progression, while activity of the protective HIF2A TF was the most decreased. However, the TF binding motifs predicted by H3K4me2 and H3K4me3 cfDNA ChIP-seq did not significantly change in patients with PD. Conclusions: Distinctive genomic and epigenomic p