Prevalence of human leukocyte antigen-B27 supertype in the context of positively charged neoepitopes and association with PD-L1 as an immune escape mechanism.

3083 Background: Recent evidence suggests efficacy of immune checkpoint blockade may be influenced by human leukocyte antigen (HLA)-B. HLA-B27 supertype has an electronegative binding pocket which favorably binds and displays neoepitopes harboring positively charged amino acids (AAs). Based on immune surveillance, we postulate that B27 tumors that have favorable neoepitopes should face negative selective pressure, and B27 tumors with favorable neoepitopes that develop could be more likely to upregulate immune escape mechanisms. Here we evaluate the relationship between prevalence of B27 and positively charged neoepitopes and assess association between positively charged neoepitopes and expression of PD-L1. Methods: TCGA datasets from head and neck squamous cell (HNSC), lung squamous cell (LUSC), and melanoma (SKCM) patients were evaluated. HLA alleles were determined with OptiType and supertype was based on 2008 criteria. Nonsynonymous mutations were annotated with Ensembl VEP and VAtools. pVAC-Seq using NetMHCpan algorithm predicted neoepitopes 9 AAs in length. Favorable B27 neoepitopes were defined as those having new positively charged AA substitutions (H/K/R) from negative or uncharged wildtype AAs. RNA-seq data for the PD-L1 gene were normalized on transcripts per million and log2 transformed. Linear regression tests were performed between PD-L1 gene expression values and fraction of nonsynonymous mutations resulting in neoepitopes with new positively charged AAs in patients with B27. Results: Data from 497 HNSC, 494 LUSC, and 468 SKCM patients were analyzed. B27 was observed in 20.1%, 23.2%, and 26.5% of HNSC, LUSC, and SKCM patients, respectively, with a significant difference seen between HNSC and SKCM by chi-square test (χ² = 5.14, p = .023). Of new charged AAs resulting from nonsynonymous mutations, 76.3% in HNSC, 74.0% in LUSC, and 72.0% in SKCM were positively charged (p < .05 between all histologies, paired t-tests). In B27 patients, association between PD-L1 gene expression and fraction of neoepitopes with new positively charged AAs was seen in HNSC (r = 0.25 p = .036) and SKCM (r = 0.30 p = .007), but not LUSC (r = -0.12 p = .296). Conclusions: With increasing fraction of positively charged neoepitopes, a decrease in prevalence of B27 was observed, suggesting improved binding and immune elimination of tumors with favorable neoepitopes. In B27 tumors that develop despite having favorable neoepitopes, upregulation of PD-L1 could be a putative mechanism to evade immune detection.