Molecular correlates of response in patients with metastatic castration-resistant prostate cancer treated with olaparib with or without cediranib (NCI9984).

5080 Background: The phase 2 NCI 9984 study (NCT02893917) randomized patients with progressive metastatic castration resistant prostate cancer (mCRPC) 1:1 to the PARP inhibitor (PARPi) olaparib (O) +/- the vascular endothelial growth factor receptor inhibitor (VEGFRi) cediranib (C+O). (C+O) improved radiographic progression free survival (rPFS) compared to (O). Though further development of this combination was precluded in part by toxicity, we hypothesized that defining molecular correlates of response in this study could suggest alternative combination strategies, including alternative VEGFRi, to improve the therapeutic efficacy of PARPi in prostate cancer. Methods: Tumor biopsies were reviewed by a pathologist to ensure adequate tumor content. Whole exome DNA sequencing (WES) was performed on pre-treatment biopsies and germline DNA from blood, and RNA sequencing (RNA-seq) was performed on both pre- and on-treatment biopsies. Sequencing data was analyzed using standard computational pipelines to determine tumor-specific DNA variants and gene expression. Mutational signatures were inferred using SigMA, and clonal architecture was inferred using PhylogicNDT. Transcriptome changes were inferred using gene set enrichment analysis and DESeq2. Molecular features were associated with rPFS using descriptive statistics. Results: 23 pre-treatment biopsies– 11 (O) and 12 (C+O) – yielded WES data that passed quality control (QC). 62 RNA-seq samples passed QC – 28 (O) and 34 (C+O) – with 30 pair pre-post samples. After stratifying patients with WES data by response above or below the median rPFS for each treatment arm, no genomic alteration was significantly associated with response. We detected evidence of COSMIC mutational signature 3 (Sig3), indicative of HRR deficiency, in 9 pre-treatment samples – 4 (O) and 5 (C+O), only 2 of which contained a BRCA2 alteration. The presence of Sig3 was not associated with improved rPFS in either treatment arm. RNA-seq demonstrated decreased in HRR and double strand DNA break repair gene sets as well as an increase in interferon signaling with (C+O) treatment. Conclusions: Molecular profiling of pre- and on-treatment tumor biopsies confirms prior preclinical observations that the addition of cediranib suppresses the expression of genes associated with HRR, which may explain the improved rPFS of (C+O) treatment relative to (O). No DNA alterations were significantly associated with therapeutic response in this more limited genomic cohort, but we confirm findings from other studies that mCRPC tumors possess mutational signatures of HRR deficiency in the absence of BRCA2 alterations, which has implications for PARPi deployment in mCRPC. Additionally, our work broadly demonstrates the both the feasibility and sample attrition rate of correlative studies on clinical trials.