Interrogating real world tumor-infiltrating T-cell repertoires to identify antigen enriched TCRs in a large pan-cancer clinical cohort

Authors Paul Fields, Fu Luo, Taylor Harding, Michelle M Stein, Catherine Igartua and Justin Guinney

Background TCR repertoire profiling can provide a useful window into the complex interactions between tumor cells and infiltrating lymphocytes. Despite recent advances in repertoire sequencing methods, the characterization of tumor-infiltrating T-cell repertoires has been limited by access to large, real-world patient cohorts with genomic, TCR, and clinical profiling. In this study, we analyzed a large, real-world, clinicogenomic database of over 130k patients – with T-cell repertoire data covering a diverse landscape of HLA genotypes and tumor neoantigens – to identify public TCRs associated with HLA-specific neoantigens and viral epitopes.

Methods TCR sequences were detected using the Tempus xR assay, a hybrid-capture whole exome RNA assay optimized with probes for capture of TCR genes. Repertoire profiling reads were aligned, assembled, and annotated against IMGT reference sequences. Neoantigens were identified from paired samples using the Tempus xT targeted DNA panel covering 648 cancer-associated genes. HPV status was determined using a panel of probes to E6 and E7 genes in HPV16, 18 and 33 and captured by both the RNA and DNA assays.

Results Investigation of intermediate public TCRs – defined as TCRs shared among 10 < N < 1000 patients – revealed TCRs that were co-occurring with multiple neoantigens, including KRASTP53, and EGFR (Fisher adjusted p<0.001). KRAS Q61H-associated TCRs were enriched within subjects with A*01:01, consistent with peptides predicted to bind and present by MHC class I allele. Expanding beyond neoantigens to viral epitopes, integration of HPV status in parallel with repertoire data revealed TCRs that were significantly enriched within HPV positive tumors in HNSCC and CESC (P < 0.05, Fisher Exact Test), but not the inverse, consistent with antigen selection. Similar to KRAS neoantigen enriched TCRs, these TCRs were associated with specific HLA class I alleles.

Conclusions By incorporating routine TCR repertoire profiling into a high-volume clinical genomic sequencing program, we have developed a rich, multi-modal resource for studying the complex tumor-immune interaction. Analysis of this resource – encompassing a diverse collection of cancer types, HLA genotypes, and mutational/viral contexts – revealed a subset of TCRs enriched with allele-specific neo-antigens. This dataset is a valuable resource for TCR therapeutic discovery, and can help identify naturally occuring TCRs that may minimize on-target, off-tumor toxicity.