CD8+ T cell stemness precedes post-intervention control of HIV viraemia

Abstract

Interventions to induce lasting human immunodeficiency virus (HIV) remission are needed to obviate the requirement for lifelong antiretroviral therapy. Durable post-intervention control (PIC) of viraemia has been achieved in a subset of people following administration of broadly neutralizing anti-HIV-1 antibodies (bNAb) and analytical interruption of treatment^1-4. Previous studies support a role for CD8⁺ T cells in PIC^5-9, but the precise features of CD8⁺ T cells involved remain unclear. Here we mapped and functionally profiled CD8⁺ T cell responses to autologous HIV epitopes using longitudinal samples from four analytical treatment interruption trials in bNAb recipients. PIC was associated with superior pre-intervention HIV-specific CD8⁺ T cell proliferative capacity, stem-cell-like memory phenotype and recall cytotoxicity against autologous HIV peptide-pulsed CD4⁺ T cells. CD8⁺ T cell stemness was increased further following bNAb administration without emergence of new clonotypes targeting defined HLA-optimal epitopes. Multi-modal single-cell analyses revealed molecular features associated with PIC and HIV-specific CD8⁺ T cell stemness, including signatures of metabolic fitness and reduced T cell exhaustion. These results identify immune features that precede subsequent PIC to inform the development of combination immunotherapies that will elicit durable HIV remission.

Fig. 1. Autologous HIV epitope-specific CD8⁺ T cell responses in post-intervention controllers.

a, Study cohort overview. Longitudinal PBMC samples were used from seven controllers and five non-controllers before and after infusion of bNAbs 3BNC117 and 10-1074 in the MCA-906, MCA-965, TITAN and eCLEAR trials. b, Schematic overview of autologous HIV-specific CD8⁺ T cell response mapping and representative IFNγ ELISpot results. c,d, Summary of longitudinal and between-group differences in breadth (n = 6, 7, 5, 5 samples (c)) and magnitude (n = 23, 26, 22, 22 responses (d)) of HIV epitope-specific responses. SFC, spot-forming counts. Centre lines represent medians, boxes represent first and third quartiles and whiskers represent ranges. Symbols represent individual participants within the PIC and non-PIC groups (key in Table 1). P values reported above plots from two-sided paired (longitudinal) or unpaired (between-group) t-tests. Representative diagrams in a were modified from ref. , Spinger Nature Limited. Schematic in b was created using BioRender (https://biorender.com).

Fig. 2. HIV-specific CD8⁺ T cell stemness precedes PIC.

a,b, Schematic overview of HIV-specific CD8⁺ T cell proliferation assay (a) and representative longitudinal epitope-specific proliferation from one controller (PID 5120) and one non-controller (PID 9243 (b)). c, Summary of longitudinal and between-group differences in proliferative capacity of CD8⁺ T cell responses against each autologous HIV-1 epitope for which responses were detected by IFNγ ELISpot. Each data point represents the mean of triplicate wells for each response (n = 23, 26, 22, 22 responses). d,e, Schematic overview of expanded antigen-specific elimination assay to measure recall cytotoxicity (d) and representative results at increasing effector to target (E:T) ratios from one controller (PID 142; blue) and one non-controller (PID 109; red), including area under the curve (AUC) summaries (e). f, Correlation of proliferation and recall cytotoxicity, as measured in d and e, across responses from both pre- and post-intervention samples in controllers (blue) and non-controllers (red). Correlation (ρ) and P values calculated by Spearman correlation (n = 41 responses). g, Representative flow cytometric staining of memory subset markers CD45RA and CD62L on HIV peptide-HLA (pHLA) tetramer⁺ CD8⁺ T cells. h,i, Summary of longitudinal and between-group differences in T_SCM (h) and T_EM (i) subset frequencies among HIV pHLA tetramer⁺ (Tet⁺) CD8⁺ T cell responses from controllers (n = 9 responses) and non-controllers (n = 7 responses), and among CMV/flu Tet⁺ CD8⁺ T cells from both groups (n = 8). j, Correlation (ρ) and P values calculated by Spearman correlation between proliferative capacity and percent T_SCM cells among Tet⁺ CD8⁺ T cells in controllers (blue) and non-controllers (red), n = 16 responses. Centre lines represent medians, ticks represent means, boxes represent first and third quartiles and whiskers represent ranges. Symbols represent individual participants in PIC and no PIC groups (key in Table 1). P values reported above plots from two-sided Wilcoxon signed rank (between-group) or matched-pairs signed rank (longitudinal) tests (c), two-sided unpaired (between-group) or paired (longitudinal) t-tests (h,i) or Spearman correlation tests (f,j). Schematics in a and d were created using BioRender (https://biorender.com).

Fig. 3. Molecular signatures associated with PIC.

a, Schematic overview of processing, isolation and multiomics sequencing of HIV and CMV epitope-specific CD8⁺ T cells. b, Multi-modal clustering by weighted nearest-neighbours plotted using uniform manifold approximation and projection (UMAP) for dimension reduction. c, Left, cluster frequencies among HIV-specific CD8⁺ T cells from both pre- and post-intervention samples in controllers and non-controllers and among CMV-specific CD8⁺ T cells and with cluster annotations based on differential expression of genes, gene sets and surface markers shown in d; right, breakdown of participant phenotype (PIC or no PIC) and pathogen specificities (HIV, CMV) on UMAP plot as shown in b. P values reported above plots from χ² tests. d, Bubble plot comparing Z-scaled mean normalized expression and detection rates for curated surface markers, transcripts (italics) and gene signatures supporting cluster annotations, as detailed in Methods. e,f, Volcano plots summarizing differentially expressed genes (e) and surface proteins (f) among HIV-specific CD8⁺ T cells from controllers (blue) and non-controllers (red). g, Summary of top ten most significantly upregulated and downregulated gene set subnets from GSNA of HIV-specific CD8⁺ T cells from controllers versus non-controllers. Schematic in a was created using BioRender (https://biorender.com). HTO, hashtag oligonucleotide antibodies; FACS, fluorescence-activated cell sorting; 10X, single-cell encapsulation via 10X Genomics platform; NGS, next-generation sequencing; GEX, ADT, and VDJ represent gene expression, antibody-derived tags (CITE-seq) and TCR libraries, respectively.

Fig. 4. Augmented CD8⁺ T cell stemness following bNAb administration is associated with pre-existing clonotypes.

a, Longitudinal TCR clonal diversification summarized as one minus Morisita-Horn similarity index (MHSI) among HIV-specific responses from controllers (blue, n = 6 responses) and non-controllers (red, n = 7 responses) or CMV-specific responses (n = 4). b, Longitudinal TCRβ CDR3 clonotypic frequencies and MHSI of HIV (n = 13) and CMV (n = 4) epitope-specific CD8⁺ T cell responses (paired columns) at pre- and post-bNAb time points from sorted pHLA tetramer⁺ cells, ordered and coloured by within-response rank for all responses with at least ten cells and longitudinal sampling and all clonotypes that occurred more than once in the data set; full data in Supplementary Data 3. c–f, Summaries of epitope-specific frequencies measured by pHLA tetramer (tet) staining among total CD8⁺ T cells (c), activation measured by surface CD38 and HLA-DR co-expression (d), proliferation measured by intranuclear Ki-67 (e) and cytotoxic effector differentiation measured by intracellular perforin and granzyme B co-expression (f) among HIV pHLA tet⁺CD8⁺ T cell responses from controllers (n = 9 responses) and non-controllers (n = 7 responses), and among CMV/flu tet⁺CD8⁺ T cell responses (n = 8). g,h, Volcano plots summarizing longitudinal changes among HIV-specific CD8⁺ T cell responses from all participants with longitudinal sampling in gene (g) and surface protein (h) expression before (pre, gold) and after (post, magenta) intervention via CITE-seq analyses. i, Summary of top ten most significantly upregulated and downregulated gene set subnets from GSNA among HIV-specific CD8⁺ T cells from post- versus pre-intervention. j, Longitudinal cluster frequencies among HIV- and CMV-specific CD8⁺ T cells from controllers and non-controllers. k, Violin plot of single-cell AUCell expression levels of a gene signature associated with lymph node follicular CD8⁺ T cells across clusters. Centre lines represent medians, boxes first and third quartiles, and whiskers ranges. Colour–symbol combinations represent participants (key in Table 1). P values reported above plots from two-sided Wilcoxon signed rank (a and k), two-sided paired (longitudinal) or unpaired (between-group) t-tests (b–e) and χ² tests (j).

Extended Data Fig. 1. Flow cytometric CD8⁺ T cell profiling.

(a-c) Representative gating schema for measurement of epitope-specific proliferation (a), elimination of peptide-pulsed (CellTrace Far Red⁺) CD4⁺ T cell targets by peptide-expanded CD8⁺ T cell effectors (b), and phenotypic profiling of pHLA tetramer⁺ (Tet⁺) cells (c) by flow cytometry. Panel a also includes representative proliferation histogram overlays for HIV epitope-specific responses from PIC 5120 (blue) and PINC 9243 (red) relative to unstimulated controls (gray). (d) Memory subset frequencies among HIV Tet⁺ CD8⁺ T cell responses from controllers (PIC, n = 9 responses) and non-controllers (PINC, n = 7 responses), and among CMV/flu Tet⁺ CD8⁺ T cell responses from both groups (n = 8). Center lines represent medians, ticks represent means, boxes represent first and third quartiles, and whiskers represent ranges. Color-symbol combinations represent participants (key in Table 1). P-values reported above plots from two-sided paired (longitudinal) or unpaired (between-group) t-tests.

Extended Data Fig. 2. Differential expression between clusters.

(a) Feature plots of expression levels of selected differentially expressed surface proteins and transcripts (italics) projected onto UMAP plots, supporting cluster annotations in Fig. 3. (b) Bubble plots of z-scaled mean normalized expression and detection rates for top differentially expressed transcripts (left) and surface proteins (right) upregulated in each cluster, ranked by adjusted p value.

Extended Data Fig. 3. Multimodal clustering and TCR clonotypes.

(a-d) UMAP of HIV and CMV epitope-specific CD8⁺ T cells colored by WNN cluster (a), participant (b), TRB CDR3 clonotype (c), or TRB CDR3 clonotype separated by participant and response (d). Gray points represent singlets, whereas colored points are clonally expanded.