Abstract 847: Molecular characterization of breast tumor T-cell infiltration in exome datasets

Abstract

Abstract Tumors are heterogeneous, containing tumor-infiltrating lymphocytes (TILs) which presence has been associated with favorable prognosis in multiple tumor types. Importantly, most of the T-cell specificity for tumor neo-antigen lies in a unique sequence of DNA (CDR3) resulting from the recombination of the V, D, and J segments of the T-cell receptor beta (TCRB clonotype). The Cancer Genome Atlas (TCGA) is the most comprehensive molecular dataset of tumors but does not include detailed information about TILs or tumor immunity in general. In order to characterize the T-cell infiltration in public cancer genomic datasets, we hypothesize that such a unique CDR3 DNA fragment can be recovered from bulk tumor genome-wide DNA or RNA sequencing and that the information can be used for clinical association studies. We first establish the feasibility of the approach by comparing the deep CDR3 sequencing (ImmunoSeq) and CDR3 reads recovered in tumor exome data. The clonotypes were detected using three CDR3 read detection tools and results were compared to the ImmunoSeq results. Each tool identifies between 10 and 38 reads. As expected, most of the CDR3 reads are misaligned to the naïve reference genome (60% clipped). Across all three tools, we identify 26 unique clonotypes, 15 of which are detected in the ImmunoSeq experiment. IMSEQ shows the strongest overlap with the other tools (13/14 clonotypes) and with ImmunoSeq (9/14 clonotypes) showing that our strategy identifies bona fide CDR3 reads from exome datasets. We applied the approach to a cohort of 1,078 TCGA samples. We identified CDR3 reads in 473 samples which are more likely to have a higher fraction of TILs as detected by histological review (p = 1.14×10-8). Using the RNA-seq datasets, 906 samples had CDR3 reads and also associated with higher TIL content (p = 4.41×10-6). Tumors with CDR3 reads in both DNA and RNA datasets tend have a higher TIL fraction. A set of 66 tumors shares a unique TCRâ clonotype with however no association with tumor subtypes or HLA haplotype. When restricted to Her2+ tumors, high CDR3 read counts, but not high fraction of TILs, was associated with better overall survival (p = 0.016), suggesting that CDR3 reads are predictive of anti-Her2 response. In contrast, the presence of CDR3 reads in Her2- tumors is not prognostic. A detailed analysis of 22 immune meta-gene expression signatures in these samples complements CDR3 reads information and suggests mechanisms of immune-tolerance. Our results demonstrate the feasibility of identifying genetic evidence of T-cells in bulk tumor samples using deep sequencing, providing information about the T-cell abundance and clonal identity. This novel type of tumor immune biomarker enriches the comprehensive dataset collected through TCGA, and reveals correlations with other molecular features and clinical outcomes. Citation Format: Olivier Harismendy, Eric D. Levy, Valentine Garate-Calderon, Brian Woo, Ricardo Armisen. Molecular characterization of breast tumor T-cell infiltration in exome datasets. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 847.