Landscape of homologous recombination reversion mutations in pancreaticobiliary malignancies.

4156 Background: Homologous recombination (HR) pathway reversion mutations (REV) are emerging treatment resistance biomarkers for platinum and PARP inhibitor therapy. The wide diversity of REV presents a diagnostic challenge. We implemented an automated computational approach to detect REV and examined the genomic features of REV-positive pancreaticobiliary cancers. Methods: Retrospective study of tissue (n = 31,124) and liquid biopsy (n = 3,870) samples from patients undergoing hybrid-capture comprehensive genomic profiling during routine clinical care 11/2012-03/2021. A proprietary algorithm tested samples for seven distinct REV mechanisms in BRCA1, BRCA2, or PALB2. For subjects with multiple samples, the earliest REV-positive sample was used for downstream analyses. Results: 5.3% (1,866/34,994) of biliary and pancreatic primary tumors harbored one or more pathogenic variants (PV) in BRCA1 (1.4%, n = 499), BRCA2 (3.2%, n = 1,119), or PALB2 (0.81%, n = 282). Among patients with at least one PV in these genes, 2.4% (45/1,866) had REV. The majority of REV involved BRCA2 (82%, 37/45), significantly higher than the proportion of BRCA2 PV in the cohort overall (59%, 1,119/1,900; p = 0.022). The remainder of REV involved either BRCA1 (8.9%, 4/45) or PALB2 (8.9%, 4/45). A total of 63 REV pairs were identified. The most frequent REV mechanism was an exonic non-frameshift deletion completely encompassing a PV (36%, 23/63), followed by a frameshift restoring the reading frame of a pathogenic frameshift (23%, 15/63), a benign missense replacing a PV at the same codon (17%, 11/63), and deletion with intronic breakpoints causing loss of a PV (17%, 11/63). REV were approximately equally prevalent in liquid biopsy (2.8%, 5/178) and tissue samples (2.4% 40/1,688; p = 0.72). A range of 1-9 REV were found per sample (liquid biopsy = 1-9, tissue = 1-2), and liquid biopsy samples had a higher median REV per sample compared to tissue (2.0 vs. 1.0; p < 0.001), as well as a higher prevalence of multiple REV per sample (60% [3/5] vs 7.5% [3/40]; p = 0.013). These differences likely reflect stage of disease and sampling of multiple subclones. Conclusions: Although REV are uncommonly detected in pancreaticobiliary carcinomas, the presence of REV in BRCA2, BRCA1, and PALB2 in patients with pancreaticobiliary neoplasms supports HR as a therapeutic target in these cancers. Diverse REV mechanisms highlight a need for robust detection to incorporate REV in identifying treatment resistance and guiding therapy selection.