In-depth analysis of human virus-specific CD8+ T cells delineates unique phenotypic signatures for T cell specificity prediction

Following viral infection, the human immune system generates CD8+ T cell responses to virus antigens that differ in specificity, abundance, and phenotype. A characterization of virus-specific T cell responses allows one to assess infection history and to understand its contribution to protective immunity. Here, we perform in-depth profiling of CD8+ T cells binding to CMV-, EBV-, influenza-, and SARS-CoV-2-derived antigens in peripheral blood samples from 114 healthy donors and 55 cancer patients using high-dimensional mass cytometry and single-cell RNA sequencing. We analyze over 500 antigen-specific T cell responses across six different HLA alleles and observed unique phenotypes of T cells specific for antigens from different virus categories. Using machine learning, we extract phenotypic signatures of antigen-specific T cells, predict virus specificity for bulk CD8+ T cells, and validate these predictions, suggesting that machine learning can be used to accurately predict antigen specificity from T cell phenotypes. [Display omitted] •In-depth profiling of CMV-, EBV-, influenza-, and SARS-CoV-2-specific CD8+ T cells•Identification of unique phenotypes of T cells specific for different virus antigens•Machine learning infers phenotypic signatures from virus-specific T cells•Machine learning-based prediction of T cell specificity and functional validation Schmidt et al. demonstrate that machine learning can be used to predict antigen specificity from T cell phenotypes. Unprecedented in-depth profiling of hundreds of virus-specific T cell responses provides a comprehensive phenotypic atlas of human peripheral CD8+ T cells specific for antigens from different virus categories. Machine learning can be used to extract unique phenotypic features of antigen-specific T cells and to accurately predict virus specificity for bulk T cells.