Repertoire size and predictability of influenza A virus matrix protein (M158–66)-specific CD8+ TCR

Abstract

Abstract Influenza A virus (IAV) is a major human pathogen that causes seasonal infections worldwide. IAV-specific CD8+ T-cell response provides an essential part of immune protection against IAV infections. In HLA-A2+ humans, the dominant epitope on IAV is matrix protein M1 residues 58–66 (GILGFVFTL, referred to as GIL). Here, we investigated the TCR repertoire size, predictability, and signaling strength of GIL-specific CD8+ T cells healthy adults. We analyzed over 40,000 GIL-tetramer positive CD8+ T cells by scRNAseq with TCR sequences and identified over 10,000 unique abTCRs sequences. Next, we developed a machine learning algorithm (an ensemble random forest models) to further predict the probability of GIL-binding TCR (ranging from 0.1 to 1 with 1 being highest binding confidence) using reported GIL-TCRs. To confirm the predicting accuracy, we expressed 72 TCR with different RF predicting score and showed that TCR with a RF score greater than 0.8 showed 100% GIL-binding (30 different TCRs). Using RF model and filtering, we identified over 4,000 unique GIL-specific TCRs with less than 800 were previously reported. Overall, our study revealed TCR repertoire size of GIL-specific CD8+ TCRs and indicated the features of GIL-binding TCRs are identified by our ML model to accurately predict this TCR specificity, which paces way to investigate frequency and quality of antigen specific TCR repertoire size and function in the general T cell population.