Multimodal Profiling of 500,000 Memory T Cells from a Tuberculosis Cohort Identifies Cell State Associations with Demographics, Environment, and Disease

Abstract

T cell phenotyping is often limited by its reliance on single classes of markers (e.g., mRNA or protein). With multiview definitions of T cell states and their associations with non-immune factors, we can more precisely identify cell states underlying disease outcomes. Here, we use an integrative, multimodal strategy to characterize the landscape of human memory T cells. We computationally integrated high-dimensional single-cell RNA and surface protein marker data to produce an atlas of 500,089 memory T cells from 259 individuals in a Peruvian tuberculosis (TB) progression cohort profiled at immune steady-state > 4 years after infection, and we defined 31 memory T cell states based on coordinated expression of relevant genes and proteins. We associated these states with 38 demographic and environmental covariates and found strong effects of age, sex, season, and ancestry on T cell composition. We also characterized a polyfunctional Th17-like effector state reduced in abundance and function in individuals who had progressed from Mycobacterium tuberculosis (M.tb) infection to active TB disease. This state — uniquely identifiable with multimodal analysis — was independently associated with TB progression and its comorbidities. Our study demonstrates the power of integrative multimodal single-cell profiling to define high-resolution cell states with functional relevance to disease and other traits.