SARS-CoV-2 antigen exposure history shapes phenotypes and specificity of memory CD8+ T cells

Abstract

Although mRNA vaccine efficacy against severe coronavirus disease 2019 remains high, variant emergence has prompted booster immunizations. However, the effects of repeated exposures to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) antigens on memory T cells are poorly understood. Here, we utilize major histocompatibility complex multimers with single-cell RNA sequencing to profile SARS-CoV-2-responsive T cells ex vivo from humans with one, two or three antigen exposures, including vaccination, primary infection and breakthrough infection. Exposure order determined the distribution between spike-specific and non-spike-specific responses, with vaccination after infection leading to expansion of spike-specific T cells and differentiation to CCR7^-CD45RA⁺ effectors. In contrast, individuals after breakthrough infection mount vigorous non-spike-specific responses. Analysis of over 4,000 epitope-specific T cell antigen receptor (TCR) sequences demonstrates that all exposures elicit diverse repertoires characterized by shared TCR motifs, confirmed by monoclonal TCR characterization, with no evidence for repertoire narrowing from repeated exposure. Our findings suggest that breakthrough infections diversify the T cell memory repertoire and current vaccination protocols continue to expand and differentiate spike-specific memory.

Extended Data Fig. 1. Cellular and humoral response to SARS-CoV-2 infection and vaccination.

a-c. Antibody levels across study groups. Plasma was tested by ELISA for IgG antibodies specific for (a) Nucleocapsid (N), (b) the receptor-binding domain (RBD) of the spike, (c) whole spike protein of SARS-CoV-2. Normalized ODs are the percent ratio of the sample OD to the OD of the positive control samples for each plate. The black horizontal line on the plots indicates the positivity threshold, which is two times the average of the normalized ODs for all SARS-CoV-2 negative samples in the cohort. P-values for two-sided Mann-Whitney U test after Benjamini-Hochberg multiple testing correction are reported (n, number of samples is shown under x-axis labels). Donors sampled before and after mRNA vaccination are connected with a line. P-values (magenta) for paired samples were calculated with the two-sided Wilcoxon signed-rank test. Central line on violin plots shows the median, upper and lower borders show maximal and minimal values. d. Gating strategy for sorting of single live CD3+CD8+dextramer+ cells. e. Representative flow plots for donors stained with the same dextramer pools, but showing different frequencies of single live CD3+CD8+dextramer+ cells.

Extended Data Fig. 2. Dextramer assignment with feature barcodes.

Each subplot shows distribution of Log10 (# UMIs) for dextramers with certain feature barcodes in dextramer-negative (yellow) and dextramer-positive (pink) cells. Dextramer with barcode 35 B44_VEN_M did not have any specific cells.

Extended Data Fig. 3. Cross-reactivity of B15_NQK αβTCR.

a. Peptide stimulation of B15_NQK αβTCR. From left to right: unstimulated (negative control), NQKLIANQF (SARS-CoV-2) peptide stimulation, NQKLIANAF (OC43 and HKU1) peptide stimulation, PMA/Ionomycin (positive control). Top row: IFN-γ production by TCR-expressing Jurkats measured by intracellular cytokine staining. Middle row: CD69+ surface expression. Bottom row: NFAT-GFP reporter expression. b-d. Antibody titers for CCCoV spike protein and number of B15_NQK cross-reactive cells in HLA-B*15:01+ donors. Plasma collected from donors prior to infection or vaccination was tested by ELISA for IgG antibodies to the spike of b, hCoV-OC43 or c, hCoV-HKU1. The normalized ODs are the percent ratio of the sample OD to the OD of the positive control sample for each plate. The dashed line is the threshold for positivity, which is three times the average of the normalized OD for the negative control samples. d, The number of HLA-B*15:01-restricted epitope T cells after infection or vaccination (log-scale).

Extended Data Fig. 4. Composition of HLA-A*02-restricted T cell response in HLA-A*02 positive donors.

Increasing proportion of spike-targeting T cells (pink) is observed after vaccination of previously infected individuals.

Extended Data Fig. 5. Fraction of spike- and non-spike-specific T cell response after vaccination with BNT162b2.

a. Each colored ribbon represents an estimated frequency of spike- (pink) or non-spike- (blue) specific T cells for SARS-CoV-2 infected donors before and after two doses of BNT162b2. b. SARS-CoV-2 infected individuals after the first and second BNT162b2 vaccine doses (inf-vax1 and inf-vax2) have the same proportion of spike-specific T cells (p=0.9, two-sided Wilcoxon signed-rank test, n=9). Spike T cell proportion was calculated as a fraction of spike-specific T cells out of all CD8+ epitope-specific T cells of a donor in scRNAseq data. Central line on violin plots shows the median, upper and lower borders show maximal and minimal values.

Extended Data Fig. 6. GEX cluster distribution for each sample.

a. Each colored bar represents a fraction of cells from a sample in a given GEX cluster. b. UMAP visualization of cells clustered by similarity of GEX. Each subpanel shows cells from each study group. Top: cells colored by cluster. Bottom: cells colored by spike and non-spike specificity. c. Each subpanel shows cells specific for each of the tested epitopes.

Extended Data Fig. 7. Number of cells in the “exhausted” cluster (cluster 7) declines over time.

a. “Exhausted” cluster 7 (circled) is enriched with cells from expanded clones. The color of each dot shows the size of the T cell clone (Log10 of number of cells) for each cell. b. UMAP visualization of cells clustered by similarity of GEX for donors sampled twice during the study (shapes connected with a line on Fig. 1b). Timepoint 1 corresponds to inf (R1–R16), inf-vax (R17-R30); timepoint 2 corresponds to inf-vax2 (R1–R30). c. Fraction of cells in cluster 7 out of all cells. Only donors with cells in cluster 7 on timepoint 1 are shown.

Extended Data Fig. 8. Peptide stimulation confirms specificity of αβTCR motifs.

Top: example of the gating strategy (B15_specific Jurkat line 1, same as Extended Data Fig. 4a). Left column: unstimulated control. Each row shows stimulation with a single peptide (middle columns), B15 specific TCRs were stimulated with both NQKLIANQF (SARS-CoV-2) peptide and NQKLIANAF (OC43 and HKU1) peptide; Right column: PMA/Ionomycin (positive control). Responsiveness of the Jurkat cell lines was determined using an endogenous NFAT-GFP reporter.

Extended Data Fig. 9. MHC-dextramer staining confirms specificity of αβTCR motifs.

Top: example of the gating strategy (B15_specific Jurkat line 1, same as Extended Data Fig. 4a). Left column: control Jurkat cell line with other known specificity. Each row shows staining with a single MHC-dextramer.

Extended Data Fig. 10. Recognition of SARS-CoV-2 mutated epitopes by αβTCR motifs.

Left column: unstimulated control. Each row shows stimulation with a single peptide (middle columns). Responsiveness of the Jurkat cell lines was determined using an endogenous NFAT-GFP reporter.

Fig 1.. Measuring CD8⁺ T cell epitope-specific responses after diverse SARS-CoV-2 exposures.

a. Study design. Selected spike and non-spike SARS-CoV-2 T cell epitopes were loaded on recombinant biotinylated MHC-monomers. Resulting peptide-MHC complexes were polymerized using fluorescently labeled and DNA-barcoded dextran backbones. Next, we stained PBMC samples with pools of MHC-multimers, isolated bound cells using FACS, and performed scRNAseq, scTCRseq, and CITEseq using the 10X Genomics platform. Figure was created with BioRender.com. b. Time of blood sampling for each donor is shown relative to the first dose of mRNA vaccine. c. Anti-RBD IgG antibody levels in previously infected individuals increase after BNT162b2 vaccination. Anti-RBD IgG levels in the plasma were determined by ELISA. The normalized OD is the percent ratio of the sample OD to the OD of the positive control for each plate. Plasma was collected from previously infected donors prior (purple, inf, n=16), after 1 vaccine dose (inf-vax1, pink, n=10), and after 2 vaccine doses (inf-vax2, blue, n=30); SARS-CoV-2 naive donors after the full vaccination (vax2, green, n=16), and donors that were infected after vaccination (breakthrough, vax2-inf, yellow, n=9). All comparisons were done with two-sided Mann-Whitney U test, p-values are reported after Benjamini-Hochberg correction. Central line on violin plots shows the median, upper and lower borders show maximal and minimal values. d. List of SARS-CoV-2 epitopes used in this study and summary statistics for resulting epitope-specific response. e. Total frequency of MHC-dextramer-positive cells is similar in all studied groups (p>0.05 for all pairwise comparisons, two-sided Mann-Whitney U test after Benjamini-Hochberg correction). Percentage of MHC-multimer-positive cells from all CD8⁺ T cells measured by flow cytometry is shown on a log₁₀-scale (inf, n=16; inf-vax1, n=10; inf-vax2, n=30; vax2, n=16; vax2-inf, n=9). Central line on violin plots shows the median, upper and lower borders show maximal and minimal values.

Figure 2.. Magnitude, dynamics, and cross-reactivity of CD8⁺ epitope-specific responses after diverse SARS-CoV-2 exposures.

a. Antigen specificity of each T cell inferred from dextramer-barcode UMI counts. Representative distribution of the number of UMIs in cells called dextramer-positive (pink) and dextramer-negative (yellow). b. T cells within a clone have largely consistent specificity assignments, except T cells that cross-react with common cold coronavirus epitopes (B15_NQK_A/B15_NQK_Q pair). Each bar shows a fraction of cells of a given clonotype attributed to different dextramers. The 43 most abundant clones (more than 20 cells) are shown. c. The correlation between the number of UMIs for B15_NQK_Q (SARS-CoV-2) and B15_NQK_A (OC43 and HKU1) dextramers (Spearman rank correlation ρ=0.8, two-sided test p<2.2·10⁻¹⁶). d. Cross-reactivity between HLA-B*15:01-NQK epitope variants confirmed in vitro. Jurkat cell line expressing αβTCR identified from scTCRseq data binds pMHC multimers loaded with both SARS-CoV-2 and CCCoV variants of the epitope. e. The magnitude of epitope-specific CD8⁺ T cell responses. Each point depicts an estimated frequency of epitope-specific T cells in a sample (n, number of samples is shown under x-axis labels). Estimated frequency was calculated as a fraction of dextramer-specific T cells in scRNAseq results multiplied by bulk frequency of dextramer-stained CD8⁺ cells of all CD8⁺cells measured by flow cytometry. Central line on boxplot shows the median. Epitopes from spike protein are in bold font. Boxes represent the median, 25th to 75th percentiles, whiskers are minimum to maximum but no further than 1.5 IQR. f. Composition of HLA-A*01-restricted T cell response in HLA-A*01 positive donors. Increasing proportion of spike-targeting T cells (pink) is observed after vaccination of infected individuals. g. Boosting of spike-specific epitope fraction after vaccination (donor R6). h. Previously infected individuals have a higher proportion of spike-specific T cells after vaccination than before vaccination (p=0.025, one-sided Wilcoxon signed-rank test). Spike T cell proportion (shown on a log₁₀-scale) was calculated as a fraction of spike-specific T cells out of all CD8⁺ epitope-specific T cells of a donor in scRNAseq data (inf, n=14; inf-vax2, n=14; vax2-inf, n=7). Central line on violin plots shows the median, upper and lower borders show maximal and minimal values.

Figure 3.. Phenotypic diversity of epitope-specific CD8⁺ T cells after diverse SARS-CoV-2 exposures.

a. UMAP (Uniform manifold approximation and projection) of all SARS-CoV-2 epitope-specific CD8 T cells based on gene expression (GEX). Color shows results of graph-based unsupervised clustering performed with the Seurat package. b. Density plot of CCR7 and CD45RA surface expression (measured by CITE-seq) in GEX clusters. c. Bubble plot of representative differentially expressed genes for each cluster. Size of the circle shows percentage of cells in a cluster expressing a certain gene, color scale shows gene expression level. d. Distribution of epitope-specific T cells in gene expression clusters between study groups. e. Proportion of spike-specific T cells is significantly increased in cluster 1 after vaccination of previously infected individuals, compared to the pre-vaccination timepoint (p<0.0001, two-sided Fisher exact test). f. Proportion of spike-specific cells in EMRA (cluster 1) across study groups for samples with more than ten spike-specific cells (Kruskal-Wallis H test p=0.028; inf, n=8; inf-vax1, n=8; inf-vax2, n=13; vax2, n=10; vax2-inf, n=4). Boxes represent the median, 25th to 75th percentiles, whiskers are minimum to maximum but no further than 1.5 IQR. g. Expression of classical cytotoxic and memory markers across study groups and T cell specificities. Size of the circle shows percentage of cells in a cluster expressing a certain gene, color scale shows gene expression level. h. Clone size distribution within GEX clusters. Fractions of cells from 10 most abundant clonotypes in each cluster are shown with colors, all other clonotypes are shown in grey. i. Number of cells in cluster 7 (Exhausted) and cluster 10 (Cycling) in samples are strongly correlated (Spearman rank correlation ρ=0.79, two-sided test p<2.2·10⁻¹⁶). Line shows linear fit. Shaded area shows 95% confidence interval for linear fit. j-k. T cell repertoire diversity of spike (j) and non-spike specific repertoires across study groups (p=0.63 for spike, p=0.17 for non-spike, Kruskal-Wallis H test). Normalized Shannon entropy of TCRβ is plotted for samples with more than 3 unique TCRβ clonotypes (for spike: inf, n=10; inf-vax1, n=9; inf-vax2, n=21; vax2, n=13; vax2-inf, n=7 and for non-spike: inf, n=13; inf-vax1, n=9; inf-vax2, n=18; vax2-inf, n=6). Boxes represent the median, 25th to 75th percentiles, whiskers are minimum to maximum but no further than 1.5 IQR.

Figure 4.. Diverse polyclonal repertoires of epitope-specific T cells after diverse SARS-CoV-2 exposures

a. SARS-CoV-2 epitope-specific αβTCR amino acid clonotypes feature clusters of highly similar sequences with the same epitope specificity. Each node on a similarity network is a unique paired αβTCR amino acid sequence, and an edge connects αβTCRs with TCRdist less than 110. Each color represents a certain epitope specificity. Only clusters with more than two members are shown. Spike-derived epitopes are in bold font. b. TCR amino acid sequence motifs of α and β chains (TCRdist logos) for the largest clusters of highly similar TCRs for each epitope (circled with dashed line on a). c. TCRs with the same sequence motifs are found across all study groups in a matching HLA-background. Occurrence of TCR motifs on the left is shown for all HLA matching samples (rectangles on the plot). Grey rectangles represent samples lacking the TCR motif. The color of the rectangle that has a TCR motif corresponds to the sample group.