Biomarkers of response to immunotherapy in pancreatic ductal adenocarcinoma (PDAC) with homologous recombination deficiency (HRD).

4176 Background: PDAC is associated with a paucity of immune effector cells, low antigenicity, and immunosuppressive factors in the tumor microenvironment (TME); consequently, treatment of unselected PDAC patients with immune checkpoint inhibitors (ICIs) has been ineffective. In HRD-PDAC, dual PD-1/CTLA-4 ICI therapy has a response rate of 14-42%. In this study, we investigated biomarkers of immunotherapy response in PDAC tumors with HRD. Methods: We generated a murine model of HRD-PDAC and performed transcriptomic analysis of mouse responders versus non-responders to ICI therapy to identify biomarkers of response. DNA and RNA sequencing was also performed for patient tumor samples submitted to Caris Life Sciences. Samples harboring pathogenic or likely pathogenic (P/LP) BRCA1 , BRCA2 or PALB2 mutations were classified HRD, the remaining samples non-HRD. Microsatellite instability-high neoplasms were excluded. TMB-High (TMB-H) was defined as ≥10 mutations/Mb. PD-L1 positivity was determined by IHC (SP142, ≥2+, ≥5%). High genomic loss of heterozygosity (gLOH-H) was defined as LOH at ≥16% of segments analyzed (up to 552). Immune cell infiltration was estimated by RNA deconvolution using quanTIseq. Mann-Whitney U, Fisher’s Exact, or Chi-squared tests were used to determine statistical significance (p), with multiple comparisons corrections as appropriate (q). Results: In murine HRD-PDAC tumors, chronic platinum exposure was a biomarker for response to ICI. Responding tumors had differential enrichment of cytosolic DNA sensing and cGAS-STING-related pathways, and secretomes significantly enriched for the T-cell attractant chemokines CXCL9 and CXCL10. Of 6396 human PDAC samples, 4.2% were HRD (2.7% BRCA2 , 0.9% BRCA1 , 0.6% PALB2 P/LP), and 95.8% non-HRD. Compared to the non-HRD cohort, the HRD cohort was younger (median age: 66 vs 68 years, p = 0.0006), had lower prevalence of TP53 (57.4% vs 79.0%, q < 0.0001), CDKN2A (14.1% vs 24.7%, q = 0.0062), and RNF43 (0.8% vs 5.8%, q = 0.0322) mutations, and was more frequently PD-L1+ (21.7% vs 14.0%, q = 0.0460), TMB-H (6.4% vs 1.9%, q = 0.0001), and gLOH-H (41.2% vs 9.5%, q < 0.0001). Median infiltration of M1 macrophages was higher in the HRD cohort (6.2% vs 5.3%, p = 0.0028), while that of M2 macrophages was lower (2.9% vs 3.3%, p = 0.0081).TheHRD cohort demonstrated higher median expression measured in transcripts per million of CGAS (6.3 vs 5.4 TPM, p = 0.0002), CXCL9 (2.19 vs 1.60 TPM, p = 0.0015), and CXCL10 (4.65 vs 3.65 TPM, p = 0.0308), phenocopying observations in the murine model. Conclusions: PDAC tumors associated with canonical HRD variants ( BRCA1/2 , PALB2 ) have distinct genomic, transcriptomic, and TME features which are immune-permissive and explain the sensitivity of this subgroup of patients to ICI therapy. Understanding the underlying mechanisms could inform strategies to broaden the impact of ICI in this population.