Background Classical human leukocyte antigen (HLA) class I molecules are expressed by most somatic cells and present peptides to cytotoxic T cells. The HLA-genotype of an individual contains up to six different HLA-I molecules and defines the repertoire of peptides that can be presented to cytotoxic T cells. Homozygosity for one or more HLA-loci could translate in a smaller repertoire of tumour neoantigens possibly presented to cytotoxic T cells in an individual and potentially predispose such individuals with a disadvantage to fight a nascent tumour.

Material and Methods High-resolution HLA-genotyping from germline normal DNA of 80 esophago-gastric adenocarcinoma (EGA) patients was performed with the NGS method by Illumina. Whole exome sequencing (WES) was performed on tumor tissue and normal peripheral blood cells (n=39). The data were processed, and non-synonymous mutations were called. The amount of potential high-affinity binders derived from 10 cancer testis antigens (CTAs) frequently expressed in EGA and non-synonymous mutations obtained from WES data were determined using an in-silico approach for MHC-binding (IEDB.org). RNA-extraction and gene expression profiling were performed using the NanoString technology.

Results We compared the frequency of HLA homozygosity in EGA patients to an HLA-matched reference population derived from a large cohort of bone marrow donors (n=7.615 out of 615.017 donors). We demonstrate that EGA patients are more likely to be homozygous for at least one HLA-I gene than the control population. In EGA patients, 35% of HLA-A, -B, and -C alleles were homozygous in comparison with 19% of HLA alleles among the HLA-matched general population. This difference corresponded to an odds ratio (OR) for homozygosity of 2.282 (95% confidence interval (CI) 1.442-3.615, p<0.001). The odds ratios for homozygosity at HLA-A (OR=1.885, CI=1.111-3.236, p<0.05), HLA-B (OR=3.045, CI=1.346-6.499, p<0.05) and HLA-C (OR=2.170, CI=1.445-3.579, p<0.05) were significantly different. We then aimed to estimate the influence of HLA-homozygosity in the context of tumour immune surveillance. Predictions by IEDB analysis resource tool indeed showed a reduced repertoire of high and moderate-affinity MHC-binders (both CTA-derived and mutation-derived peptides) in the homozygous cohort. Our findings demonstrate a reduced amount of potentially immunogenic peptides in EGA patients with HLA-homozygosity for at least one locus, which may result in impaired cancer immunosurveillance. In line with this observation, we also found increased levels of CTA expression in homozygous compared to heterozygous patients. After artificial modification of the genotype of homozygous patients to a heterozygous genotype, the set of predicted good-binding peptides was comparable to the heterozygous cohort.

Conclusion Our results highlight the effect of HLA-I homozygosity on the immunopeptidome as important prerequisite of anti-tumor immunity. The high frequency of genomic HLA-I homozygosity observed in the EGA cohort may reflect an increased cancer risk for these patients. Together with previous reports demonstrating reduced survival after checkpoint therapy, our study suggests consideration of germ-line HLA-homozygosity for the design and interpretation of immunotherapeutic trials.

Disclosure Information M.A. Garcia-Marquez: None. M. Thelen: None. E. Bauer: None. K. Wennhold: None. J. Lehmann: None. D. Keller: None. B. Gathof: None. L. Maas: None. J. George: None. C. Bruns: None. A. Quaas: None. M. von Bergwelt-Baildon: C. Other Research Support (supplies, equipment, receipt of drugs or other in-kind support); Modest; Astellas, Roche, MSD. D. Speakers Bureau/Honoraria (speakers bureau, symposia, and expert witness); Modest; BMS. M. Peifer: None. H.A. Schlößer: B. Research Grant (principal investigator, collaborator or consultant and pending grants as well as grants already received); Significant; Astra Zeneca. D. Speakers Bureau/Honoraria (speakers bureau, symposia, and expert witness); Modest; BMS.