Distinct SIV-specific CD8+ T cells in the lymph node exhibit simultaneous effector and stem-like profiles and are associated with limited SIV persistence

Abstract

Human immunodeficiency virus (HIV) cure efforts are increasingly focused on harnessing CD8⁺ T cell functions, which requires a deeper understanding of CD8⁺ T cells promoting HIV control. Here we identifiy an antigen-responsive TOX^hiTCF1⁺CD39⁺CD8⁺ T cell population with high expression of inhibitory receptors and low expression of canonical cytolytic molecules. Transcriptional analysis of simian immunodeficiency virus (SIV)-specific CD8⁺ T cells and proteomic analysis of purified CD8⁺ T cell subsets identified TOX^hiTCF1⁺CD39⁺CD8⁺ T cells as intermediate effectors that retained stem-like features with a lineage relationship with terminal effector T cells. TOX^hiTCF1⁺CD39⁺CD8⁺ T cells were found at higher frequency than TCF1^-CD39⁺CD8⁺ T cells in follicular microenvironments and were preferentially located in proximity of SIV-RNA⁺ cells. Their frequency was associated with reduced plasma viremia and lower SIV reservoir size. Highly similar TOX^hiTCF1⁺CD39⁺CD8⁺ T cells were detected in lymph nodes from antiretroviral therapy-naive and antiretroviral therapy-suppressed people living with HIV, suggesting this population of CD8⁺ T cells contributes to limiting SIV and HIV persistence.

Extended Data Fig. 1. LN TOX^hiTCF1⁺CD39⁺CD8⁺ T_M cells express high levels of inhibitory receptors after SIV infection.

a) Expression of PD-1, TIGIT, CD101 and Ki-67 on TOX⁺ and TOX⁻ LN CD8⁺ T_M cells at day 42 p.i. (n = 28 macaques). b) UMAP visualization of Phenograph clusters as in a. c) Heatmap of expression of TOX, TCF1, CD39, PD-1, TIGIT, CD101, Ki-67, GzmB and EOMES across all Phenograph clusters as in a. d) Proportion of total CD8⁺ T_M cell population made up by individual Phenograph clusters (n = 28) as in a. e) Frequency of TCF1⁺CD39⁺ cells within LN CD8⁺ T_M cells in 5 unique cohorts of macaques: uninfected (n = 10), day 21 p.i. (n = 7), day 35 p.i. (n = 20), day 42 p.i. (n = 28), late chronic infection (mean 17 months p.i. (n = 10). f-g) Expression of TOX, PD-1, TIGIT, CD101, Ki-67 and GzmB in TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of CD8⁺ T_M cells at day 35 p.i. (n = 20) (f) and late chronic infection (n = 10) (g). P values determined by Wilcoxon matched-pairs signed rank test, Kruskal-Wallis one way ANOVA with Dunn’s multiple comparison correction (e) or two-way ANOVA with Tukey’s multiple comparisons test (f-g). Box and whiskers are displayed via Tukey method. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Extended Data Fig. 2. TCF1⁺CD39⁺CD8⁺ T_M cells maintain a dual effector and stem-like proteomic profile.

a) Representative gating strategy for sorting of TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of CD8⁺ T_M at day 42 p.i. b-c) Volcano plots displaying significantly differently expressed proteins of interest between sorted LN TCF1⁺CD39⁻ and TCF1⁺CD39⁺ CD8⁺ T_M cells (b) and between sorted LN TCF1⁺CD39⁻ and TCF1⁺CD39⁺ CD8⁺ T_M cells (c) at day 42 p.i. d) Selected proteins with significantly different expression are listed (n = 4 macaques). All statistical comparisons were performed based on two-tailed Student’s t-tests using Spectronaut and custom R scripts.

Extended Data Fig. 3. Sorting of LN SIV-specific CD8⁺ T_M cells.

a) Sorting strategy for LN SIV-specific CM9⁺CD8⁺ T_M cells for downstream scRNA-seq at day 42 p.i. b) Plots and frequencies of sorted LN SIV-specific CM9⁺CD8⁺ T_M cells for all 5 macaques as in a.

Extended Data Fig. 4. Transcriptional clustering of LN SIV-specific CD8⁺ T_M cells overlaps with established human CD8 T cell signatures.

a) Heatmap of top 20 differentially expressed genes across the 5 clusters of LN SIV-specific CM9⁺CD8⁺ T_M cells at day 42 p.i. b) UMAP projection of cells matching established human CD8⁺ T cell signatures (Supplementary Data File 1) of cell cycling, type 1 interferon, human stem-like CD8 and human terminally differentiated (TD) CD8. c) Distribution of LN SIV-specific CM9⁺CD8⁺ T_M cells across the 5 clusters for all macaques as in a. d) Scatter plot of expression levels and thresholds (dotted lines) for identifying TCF7⁺ENTPD1⁺ CD8⁺ T_M cells as in a.

Extended Data Fig. 5. TCR analysis of expanded TCF7⁺ENTPD1⁺CD8⁺ T_M cells demonstrates lineage relationship with terminally differentiated CD8⁺ T_EFF cells.

a-e) UMAP projection of the top 3 clonotypes within TCF7⁺ENTPD1⁺ CD8⁺ T_M cells and the top 5 clonotypes of TCF7⁺ENTPD1⁺ CD8⁺ T_M cells population from each macaque and the cluster distribution frequency of those clones across all SIV-specific CM9⁺CD8⁺ T_M cells at day 42 p.i. Number of total cells expressing each TCR clonotype is shown below each bar.

Extended Data Fig. 6. Response of TCF1⁺CD39⁺CD8⁺ T_M cells to dual IL-10 + PD-1 mAb blockade.

a) Study schematic demonstrating dosing regimen and timing of lymph node biopsies (week −1 and week 12 post-treatment) for flow cytometry analysis placebo-treated control (n = 6) and treated (n = 9) macaques. b) Fold change of the frequency of TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of LN CD8⁺ T_M cells, calculated using frequencies at week 12 post-treatment compared to baseline frequencies at week −1 (control (n = 6) and treated (n = 9)). c) Fold change of the frequency of LN Ki-67⁺ cells within TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of CD8⁺ T_M cells (control (n = 6) and treated (n = 9)). P values determined by multiple Mann-Whitney tests with Holm-Sidak correction for multiple comparison. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Extended Data Fig. 7. TCF1⁺CD39⁺CD8⁺ T_M cells are associated with reduced viral burden.

a-b) Association of the frequency of TOX⁺ (a) and TCF1⁺ CD39⁺ (b, red) cells within LN CD8⁺ T_M cells with cell-associated SIV RNA levels in LN at day 42 p.i. (n = 26). c-h) Association of the frequency of TCF1⁺ CD39⁺ (c-e, red) and TOX⁺ (f-h) cells within LN CD8⁺ T_M cells with plasma viral load (c,f), total SIV DNA (d,g) and cell associated RNA (e,h) at day 35 p.i. (n = 20). All correlations were determined using two-tailed Spearman analysis.

Extended Data Fig. 8. Expression of CXCR5 by TOX⁺TCF1⁺CD39⁺CD8⁺ T_M cells is associated with viral burden.

a-c) Association of the frequency of CXCR5⁺ cells within LN CD8⁺ T_M cells with plasma viral load (a), total SIV DNA (b) and cell associated RNA (c) at day 42 p.i.(a n = 27 macaques, b-c n = 25). d-f) Association of frequency of CXCR5⁺ cells within TCF1⁺CD39⁻(d, burgundy), TCF1⁺CD39⁺ (e, red) and TCF1⁻CD39⁺ (f, pink) subsets of LN CD8⁺ T_M cells with cell-associated SIV RNA levels in sorted LN memory CD4 T cells as in a (n = 25). g) Frequency of CXCR5⁺ cells within LN GzmB⁻GzmK⁺, GzmB⁺GzmK⁺ and GzmB⁺GzmK⁻ CD8⁺ T_M cells LN memory CD8 T cells as in a (n = 27). h) Representative image of immunofluorescence staining of CD39 (green) TCF1 (blue) and TOX (red). i) Representative image of immunofluorescence analysis identifying TCF1⁺ and TOX⁺ CD8⁺ T cells in the B cell follicle and T cell zone. Arrows designate examples of TCF1⁺ TOX⁺ CD8⁺ T cells. j-k) Representative image of immunofluorescence analysis identifying SIV⁻infected CD4⁺ T cells and TCF1⁺TOX⁺CD8⁺ T cells in the T cell zone (j) and the dark zone of the b cell follicle (k). The perimeter of analysis is defined by the red circle. Blue arrows indicate uninfected CD4⁺ T cells not surrounded by TCF1⁺TOX⁺CD8⁺ T cells. The dotted line delimits the edge of the b cell follicle. Number of events analyzed per LN area: B cell follicle SIV⁺ = 175, B cell follicle SIV⁻ = 178, T cell zone SIV⁺ = 501, T cell zone SIV⁻ = 475, dark zone B cell follicle SIV⁺ = 217, dark zone bell cell follicle SIV⁻ = 153. All analyses were performed on samples from animals at early chronic infection (D42 p.i.). Correlations represent Spearman analysis. P values were determined using Friedman test one-way ANOVA with Dunn’s multiple comparison test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Extended Data Fig. 9. TCF1⁺CD39⁺CD8⁺ T_M cells express high levels of inhibitory receptors and low cytolytic markers in ART-suppressed PLWH.

a-d) Expression frequency of TOX, PD-1, TIGIT (a), Ki-67 (c), GzmB and GzmK (c-d) in TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of LN CD8⁺ T_M cells in ART-suppressed PLWH (n = 10). P values determined by two-way ANOVA with Tukey’s multiple comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Extended Data Fig. 10. TCF1⁺CD39⁺CD8⁺ T_M cells are not associated with viral burden in ART-naïve PLWH.

a-c) Associations of the frequency of TCF1⁺CD39⁻ (burgundy), TCF1⁺CD39⁺ (red), and TCF1⁻CD39⁺ (pink) subsets within LN CD8⁺ T_M cells from ART-naïve PLWH with plasma viral load (n = 18) (a), intact reservoir size as measured by IPDA (n = 10) (b) and total HIV DNA (n = 10) (c). d-e) Associations of the frequency of TCF1⁺CD39⁻ (burgundy), TCF1⁺CD39⁺ (red), and TCF1⁻CD39⁺ (pink) subsets within LN CD8⁺ T_M cells from ART-Suppressed PLWH with intact reservoir size as measured by IPDA (n = 10) (d) and total HIV DNA (n = 10) (e). Correlation evaluated by two-tailed Spearman analysis.

Figure 1.. TOX is upregulated on CD8⁺ T cells after SIV infection.

a) Representative flow cytometry staining of TOX in total CD8⁺ T cells in lymph nodes (LN) isolated from macaques infected with SIV at day 42 and late chronic infection (mean 17 months p.i.) or left uninfected. b) Frequency of TOX⁺ cells in total CD8⁺ T cells in the LN of uninfected macaques (n=10), or in SIV-infected macaques at day 42 p.i. (n=28.) and late chronic infection(mean 17 months p.i.) (n=10). c) Representative flow cytometry staining of TOX in CD95⁻CD28⁺ naïve CD8⁺ T cells (CD8⁺ T_N cells), CD95⁺ memory CD8⁺ T cells (CD8⁺ T_M cells) and SIV-specific CM9⁺CD8⁺ T cells at day 42 p.i. d) Frequency of TOX⁺ cell with CD8⁺ T_N cells, CD8⁺ T_M cells and CM9⁺CD8⁺ T cells in the LN of A*01⁺ macaques (n=7) at day 42 p.i. e) UMAP plot of flow cytometry data showing expression patterns TOX, TCF1, CD39, PD-1, TIGIT, CD101, Ki-67, GzmB and EOMES in CD8⁺ T_M cells at day 42p.i. f) Heatmap showing the expression of markers of stemness, differentiation and effector capacity within a subset of clusters that represent CD95⁺GzmB⁺CD8⁺ T_EFF cells and TCF1^hiCD39^hi CD8⁺ T cells, as identified through Phenograph analysis of flow cytometry data. P values were determined using Kruskal-Wallis one-way ANOVA with Dunn’s multiple comparisons (b) or Friedman’s one-way ANOVA with Dunn’s multiple comparisons (d). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 2.. TCF1⁺CD39⁺CD8⁺ T cells expand after SIV infection and are a unique phenotypic population.

a,b) Representative staining of TCF1 and CD39 (a) and frequency of TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ cells within CD8⁺ T_M cells (b) in the LN of SIV-infected macaques at day 42 p.i. (n=28) and late chronic infection (mean 17 months p.i.) (n=10) or uninfected (n=10). c) Expression of TOX, PD-1, TIGIT, CD101 and Ki-67 in TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of CD8⁺ T_M cells at day 42 p.i. in the LN of SIV infected macaques (n=28 macaques). d-f) Representative staining of GzmB and GzmK (d) and frequency of GzmB⁺ (e), GzmB⁻GzmK⁺, GzmB⁺GzmK⁺ and GzmB⁺GzmK⁻ cells in TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets (f) within LN CD8⁺ T_M cells as in c ((e) n=28, (f) n=27). h) Frequency of CD107a⁺ and CD107a⁺IFNγ⁺ cells within TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of CD8⁺ T_M cells as in c in response to SIV Gag peptide stimulation (n=26 macaques). Background subtracted values from only animals exhibiting significant responses (>0.01% of cells responding) are shown. P values were determined using two-way ANOVA with Tukey’s multiple comparisons test (b,f), Friedman test one way-ANOVA with Dunn’s multiple comparisons (c,e) or mixed effects two-way ANOVA with Holm-Šídák multiple comparison. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 3.. LN SIV-specific TCF1⁺CD39⁺CD8⁺ T cells maintain a dual effector and stem-like transcriptional profile.

a) Representative flow cytometry staining of SIV-specific CM9⁺CD8⁺ T_M cells is shown. b) UMAP cluster analyses of LN SIV-specific CM9⁺CD8⁺ T_M cells from 5 macaques at day 42 p.i. c) UMAP plots showing expression of GZMA, GZMB, GZMK, PRF1, IFNG, PDCD1, TIGIT, LAG3, CTLA4, TCF7, SELL, JUN, FOS, IL7R, MKI67, BRCA1, IRF7, MX1, ENTPD1 and TOX in LN SIV-specific CM9⁺CD8⁺ T_M cells as in b. d) UMAP projection of TCF7⁺ENTPD1⁺ CM9⁺CD8⁺ T_M cells as in b. E) Distribution of all of LN SIV-specific CM9⁺CD8⁺ T_M cells and CM9⁺TCF7⁺ENTPD1⁺ CD8⁺ T_M cells among the 5 identified clusters. f-g) Volcano plot displaying differentially expressed genes between cluster 2 (stem) and TCF7⁺ENTPD1⁺ CD8⁺ T_M cells (f) and cluster 1 (effector) and TCF7⁺ENTPD1⁺ CD8⁺ T_M cells as in b. Selected genes with significantly different expression are listed. H) Top 3 clonotypes within TCF7⁺ENTPD1⁺ CD8⁺ T_M cells in one representative macaque. Each color represents a unique TCR clonotype. I) Top 5 clonotypes of TCF7⁺ENTPD1⁺ CD8⁺ T_M cells from one representative macaque and the cluster distribution frequency of clones across all SIV-specific CM9⁺CD8⁺ T_M cells. Number of total cells expressing each TCR clonotype is denoted below each bar.

Figure 4.. TOX⁺ and TCF1⁺CD39⁺CD8⁺ T_M cell frequencies are associated with reduced viral burden.

a) Association of frequency of TOX⁺ cells within LN CD8⁺ T_M cells and plasma viral load at day 42 p.i.(n=28). b) Association of frequency of TOX⁺ cells within LN SIV-specific CM9⁺CD8⁺ T_M cells and plasma viral load as in a (n=8). c) Association of frequency of TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ cells within LN CD8⁺ T_M cells and plasma viral load as in a (n=28). d-e) Association of frequency of TOX⁺ (d) or TCF1⁺CD39⁺ (e) cells within LN CD8⁺ T_M cells and total SIV DNA in LN CD4⁺ T_M cells as in a (n=26). f-g) Association of frequency of TOX⁺ (f) or TCF1⁺CD39⁺ (g) cells within LN CD8⁺ T_M cells and memory CD4⁺ T cell counts in peripheral blood (n=28). All correlations were evaluated using two-tailed Spearman analysis.

Figure 5.. TCF1⁺CD39⁺CD8⁺ T_M cells preferentially penetrate the follicular environment.

a-b) Representative flow cytometry staining (a) and Frequency of CXCR5⁺ cells within TCF1⁺CD39⁻, TCF1⁺CD39⁺, TCF1⁻CD39⁺, TOX⁺ and TOX⁻ subsets of CD8⁺ T_M cells in LN at day 42 p.i. (b) (n=27 macaques). c) Association of CXCR5 expression frequency within TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of CD8⁺ T_M cellsand plasma viral load as in b (n=27). d) Association of CXCR5 expression frequency within TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of CD8⁺ T_M cells and SIV DNA levels in LN CD4⁺ T_M cells as in b (n=25). e) Quantification of the proportion of total CD8⁺ T cells expressing TOX and TCF1 within B cell follicles or T cell zones as in b (n=8). f) RNAscope with PhenoCycler-Fusion immunostaining-based proximity scoring analysis of TOX⁺TCF1⁺CD8⁺ T cells with SIV-infected CD4⁺ cells in the T cell zone and B cell follicle as in b (n=11–21 animals). P values were determined using Friedman test one-way ANOVA with Dunn’s multiple comparisons (b, within TCF1/CD39 subsets), two-tailed Wilcoxon (b, within TOX⁺/TOX⁻) or two-way ANOVA with Holm-Šídák multiple comparison. Correlations were deteremined using two-tailed Spearman analysis. Differences in proximity scores were assessed on mean values of proximity scores for infected cells versus noninfected ones within the different tissue compartments. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 6.. TCF1⁺CD39⁺CD8⁺ T_M cells contract during long-term ART but remain associated with viral burden.

a-b) Representative flow cytometry staining (a) and frequency of TOX⁺ cells within within LN CD8⁺ T_M cells from longitudinally matched macaques at day 42 p.i. and week 64 post-ART initiation (b) (n=28 day 42 p.i., n=25 ART). c-d) Representative flow cytometry staining (c) and frequency of TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of LN CD8⁺ T_M cells as in b (n=25). e) Expression of TOX, PD-1, TIGIT, CD101, Ki-67 and GzmB within TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of LN CD8⁺ T_M cells as in b (n=25). F) Association between frequency of LN TCF1⁺CD39⁺ CD8⁺ T_M cells and levels of intact SIV DNA in total LN CD4⁺ T cells at week 64 post-ART initiation (n=24). P values were determined using two-tailed Wilcoxon (b), multiple two-tailed Wilcoxon (d) or Friedman test one way-ANOVA with Dunn’s multiple comparisons. Correlation evaluated using two-tailed Spearman analysis. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 7.. HIV infection results in expansion of LN HIV-specific TCF1⁺CD39⁺CD8⁺ T_M cells.

a-b) Representative flow cytometry staining (a) and (b) proportion of LN TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of LN CD8⁺ T_M cells in uninfected individuals (n=4), ART-naïve PLWH (n=18) and ART-suppressed PLWH (n=10). c-d) Expression of TOX, PD-1, TIGIT and Ki-67 (c), GzmB (d) within TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of LN CD8⁺ T_M cells ART-naïve PLWH (n=18 participants). e) Proportion of cells within TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of LN CD8⁺ T_M cells expressing one or both of Gzmb and Gzmk as in c (n=18 participants). f) Proportion of cells within TCF1⁺CD39⁻, TCF1⁺CD39⁺ and TCF1⁻CD39⁺ subsets of LN CD8⁺ T_M cells expressing CD107a and CD107a in combination with production of IFNy in response to HIV Gag peptide stimulation as in c (n=8 participants). Background substracted values and only samples exhibiting significant responses (>0.01% of cells responding) are shown. P values were determined using two-way ANOVA with Holm-Šídák multiple comparison (b), Friedman test one-wayANOVA with Dunn’s multiple comparisons (c,d,f) or two-way ANOVA with Tukey’s multiple comparison. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.