Figure 4: Inferred spike-specific T cells in SARS-CoV-2 infected samples.

(A) UMAP visualization from (Adamo et al.), representing 6,070 CD8+ T cells collected during acute COVID-19 disease. Cells with positive dextramer staining are highlighted in red. (B) Left: Violin plots showing distribution of gene module score, comparing dextramer-positive versus -negative cells. Right: ROC curve assessing the ability of the gene module score to correctly predict dextramer staining labels in single cells. (C) The vaccine-induced gene expression module is conserved in SARS-CoV-2 infected samples. Shown is the expression of the top 50 vaccine-induced marker genes for dextramer-positive and -negative cells. For visualization purposes, a randomly selected subset of dextramer-negative cells are presented. (D) WNN UMAP visualization of 65,889 single cells from the COMBAT dataset. WNN was performed based on RNA and protein modalities, and identifies cell populations that we infer are specific to SARS-CoV-2 antigens. (E) Milo analysis of differential abundance changes, comparing healthy versus SARS-CoV-2 infected groups, as in Figure 1 D-E. (F) Amongst all CD8+ T cells, donor fraction of antigen, antigen_proflif, and CD38+KLRG1+ cells, grouped by disease state. Boxplots show variation across n=71 donors, and p-values from two-tailed Wilcoxon signed-rank test. (G) The fraction of inferred antigen-specific cells correlates with clinical outcome. Same as in (F), but restricted to patients exhibiting severe symptoms, and grouped by their clinical outcome. (H) UpSet plot visualizing the overlap of antigen-specific TCR sequences across distinct molecular states of CD8+ T cells. (I) Fraction of TCR clonotypes identified in either antigen cells (right) or antigen_prolif cells (left) that are also observed in Cytotoxic TEM cells. Boxplots show variation across diseased donors. (J) Density plots showing the abundance distribution of all cells harboring expanded antigen-specific TCR sequences.