Cytotoxic CX3CR1+ Vδ1 T cells clonally expand in an interplay of CMV, microbiota, and HIV-1 persistence in people on antiretroviral therapy

Abstract

Vδ1 γδ T cells are key players in innate and adaptive immunity, particularly at mucosal interfaces such as the gut. An increase in circulating Vδ1 cells has long been observed in people with HIV-1, but remains poorly understood. We performed a comprehensive characterization of Vδ1 T cells in blood and duodenal intra-epithelial lymphocytes, obtained from endoscopic mucosal biopsies of 15 people with HIV-1 on antiretroviral therapy and 15 HIV-seronegative controls, in a substudy of the ANRS EP61 GALT study (NCT02906137). We deciphered the phenotype, functional profile, single-cell transcriptome and repertoire of Vδ1 cells and unraveled their relationships with the possible triggers involved, in particular CMV and microbiota. We also assessed whether Vδ1 T cells may play a role in controlling the HIV-1 reservoir. Vδ1 T cells were mainly terminally differentiated effectors that clonally expanded in the blood with some trafficking with the gut of people with HIV-1. Most expressed CX3CR1 and displayed a highly cytotoxic profile, but low cytokine production, supported by a transcriptomic shift towards enhanced effector lymphocytes. This expansion was associated with CMV status and markers of occult replication, but also with changes in the duodenal and blood-translocated microbiota. Cytotoxic, but not IFN-γ-producing, Vδ1 T cells were negatively associated with cell-associated HIV-1 RNA in both the blood and duodenal compartments. The increase in Vδ1 T cells observed in people with HIV-1 has multiple triggers, particularly CMV and microbiota, and may in turn contribute to the control of the HIV-1 reservoir.

Fig 1. Increased frequency of Vδ₁ T cells in the blood of PLWH is associated with the expansion of cells with a CX3CR1⁺ TEMRA phenotype in the blood.

(A) Schematic representation of sample origin (created with Biorender). Duodenal endoscopic biopsies (n = 8) were either stored in RNA-later for subsequent sequential lysis and viral and bacterial RNA and DNA extraction, or processed and cryopreserved for subsequent analysis of IEL. (B) Flow cytometric analysis of γδ⁺ T cells among CD3⁺ T cells and Vδ₁ and Vδ₂ subsets in a PLWH blood sample. (C) Violin plots of the frequencies of γδ⁺ T cells among CD3⁺ T cells, and their Vδ₁ and Vδ₂ subsets in IEL and blood from PLWH and HIV seronegative controls (n = 15 PLWH and 15 controls, all CMV seropositive). Comparisons were made using Welch’s t-test. The graphs show the median (solid bar) and the first and third quartiles (dashed bars). (D) Proportions of naive (CD27⁺CD45RA⁺), central memory (CM; CD27⁺CD45RA^-), effector memory (EM; CD27^-CD45RA^-), and terminally differentiated effector memory (TEMRA; CD27^-CD45RA⁺) of the Vδ₁ and Vδ₂ subsets in IEL and blood of PLWH and HIV seronegative controls. (E) Representative dot plot of CX3CR1 expression among Vδ₁ T cells in PLWH blood according to their CD27 and CD45RA expression, and histogram of the median percentage of CX3CR1⁺ among circulating Vδ₁ T cells. (F) Violin plots of TEMRA CX3CR1⁺ Vδ₁ T cell counts in blood (n = 15 PLWH and 15 HIV seronegative controls, all CMV seropositive). Comparison was made using Welch’s t-test.

Fig 2. Vδ₁ T cells of PLWH present higher expression of NKG2C in both the blood and the duodenum.

(A) Opt-SNE representation of Vδ₁ T cell clustering obtained from flow cytometric data. Clustering was performed on a total of n = 15 PLWH and n = 15 HIV seronegative controls combined, all CMV seropositive. Vδ₁ T cell counts were equalized between individuals after a subsampling step. Expression of CD27, CD45, CX3CR1, NKG2C and TIM-3 in the clusters is shown with a color gradient (each marker has a different color scale to optimize the gradient visualization). (B) Violin plots of the frequencies of TIM-3⁺ and NKG2C⁺ cells among CX3CR1⁺ Vδ₁⁺ T cells in blood (n = 15 PLWH and 15 HIV seronegative controls, all CMV seropositive). Comparisons were made using the Welch’s t-test. (C) Radar plot of phenotypic markers distinguishing blood and duodenal epithelial Vδ₁ T cells. Values are median percentages of expression of each marker among Vδ₁ T cells. (D) Violin plots of the frequencies of NKG2A and NKG2C among Vδ₁⁺ T cells in the IEL (n = 15 PLWH and 15 HIV seronegative controls, all CMV seropositive). Comparisons were made using Welch’s t-test. (E) Positive correlation between blood Vδ₁/Vδ₂ ratio and NKG2C expression in IEL Vδ₁⁺ T cells (n = 15 PLWH and 15 HIV seronegative controls, all CMV seropositive). Spearman’s correlation coefficient (r_S) and P-value are shown. (F) Violin plots of blood counts of β7⁺ CD103⁺ CX3CR1⁺ Vδ₁ T cells (n = 15 PLWH and 15 HIV seronegative controls, all CMV seropositive). Comparison was made using the Welch’s t-test. See also S4 Fig for the median co-expression of activation/exhaustion markers PD-1, TIM-3 and TIGIT.

Fig 3. Circulating CX3CR1⁺ TEMRA Vδ₁ T cells have a highly cytotoxic profile but produce few pro-inflammatory cytokines.

(A) Heatmaps of the mean percentage expression of Granzyme B, Perforin, IFN-γ and TNF-α among Vδ1 T cells. The red and green heatmaps are the results of 2 different flow cytometry panels. The same panels were used for both PBMC and IEL staining. Blood staining was performed on n = 15 PLWH and 15 HIV seronegative controls, all CMV seropositive. IEL staining was performed on n = 5 PLWH and n = 5 HIV seronegative controls, all CMV seropositive. Results shown are the mean of PLWH and HIV seronegative controls combined. (B) Representative dot plots of Granzyme B (GzmB) and Perforin expression in Vδ₁ T cells and CX3CR1⁺ TEMRA Vδ₁ T cells. (C) Heatmaps of the median percentage expression of the transcription factors TCF-1, Eomes, T-bet and TOX among blood Vδ1 T cells. Results are from a single flow cytometry panel. Blood staining was performed on n = 15 PLWH and n = 15 HIV seronegative controls, all CMV seropositive. (D) Violin plots of CX3CR1⁺ TEMRA T-bet⁺ Eomes^- Vδ₁ T cell percentages in blood (n = 15 PLWH and 15 HIV seronegative controls, all CMV seropositive). (E) Representative dot plots of Eomes, T-bet, TOX and TCF-1 expression in blood CX3CR1⁺ Vδ₁ T cells and naive (CD27⁺CD45RA⁺) Vδ₁ T cells. See also S5 Fig for duodenal cytotoxic function.

Fig 4. Circulating Vδ₁ T cell transcriptome shifts toward an enhanced cytotoxic effector profile.

Data shown are the result of a single-cell RNA sequencing experiment performed on n = 7,865 sorted Vδ₁ T cells from the blood of 6 PLWH and 6 HIV seronegative controls, all CMV seropositive. (A) UMAP representation of single Vδ₁ T cells according to their transcriptomic profile, which allowed to distinguish 6 clusters of interest. (B) Proportion of each cluster for each individual donor. (C) Log 2 ratio of each cluster according to its status (PLWH versus HIV seronegative control). A positive value was associated with an increased frequency in HIV seronegative controls compared to PLWH, and a negative value was associated with an increased frequency in PLWH compared to HIV seronegative controls. All log 2 ratios were significantly associated with an increased frequency of the corresponding cluster in either PLWH or HIV seronegative controls (false discovery rate (FDR) < 0.05). (D) Dot plot of the expression of genes of interest according to each cluster. Each dot size represents the percentage of cells expressing the corresponding gene within the cluster, and the color range indicates the average scaled expression level of the gene. (E) UMAP representation of the given gene expression level. The intensity of the purple color represents the level of RNA expression (quantified by the number of reads per cell for each gene) for the chemokine receptor CX3CR1, differentiation marker CD27, IL-7 receptor (CD127), cytotoxic effectors GzmB, granulysin, cytokine IFN-γ, transcription factors TCF7 (TCF-1), TOX, TBX21 (T-bet), homing integrins αE (CD103) and β7, and sphingosine-1-phosphate receptor (S1PR1). See also S6 Fig for volcano plots of the differentially expressed genes between PLWH and HIV seronegative controls.

Fig 5. Circulating Vδ₁ T cells are clonally expanded in PLWH.

Unsorted duodenal IEL and sorted Vδ₁ T cells from PBMC of the same 5 PLWH and sorted Vδ₁ T cells from PBMC of 5 HIV seronegative controls, all CMV seropositive, were used to sequence the TRDV1 chain repertoire. (A) Tree maps of unique CDR3 clonotypes. Each CDR3 color is randomly chosen and does not match between plots, except for the clonotype marked with a white asterisk and colored in red, which represents an identical clonotype. Each colored square size represents the proportion of the clonotype within the total TRDV1 chain CDR3 repertoire. (B) Diversity index (function of the frequency of each CDR3 and the total number of unique CDR3s) for the TRDV1 chain in blood and Rényi’s plot of mean Hill’s numbers of order q = 0 to q = ∞, showing lower diversity in blood from PLWH (n = 5) than in HIV seronegative controls (n = 5). P-values are the result of Welch’s t-test. (C) Diversity index for TRDV1 chain in the duodenal IEL and Rényi’s plot of mean Hill’s numbers of order q = 0 to q = ∞, showing higher diversity in duodenum (n = 5) than in blood of PLWH (n = 5). P-values are the result of paired t-test.

Fig 6. CMV is associated with Vδ₁ T cells expansion in PLWH on ART.

(A) Violin plots of the number of γδ⁺ T cells and of their Vδ₁ and Vδ₂ subsets, and the frequencies of Vδ₁ and Vδ₂ subsets among γδ⁺ T cells in the blood according to HIV and CMV infection status (n = 15 HIV⁺ CMV⁺/ n = 15 HIV^- CMV⁺/ n = 12 HIV⁺ CMV^-/ n = 12 HIV^- CMV^- individuals). Comparisons were made using Welch’s t-test. The violin plots show the median (solid bar) and the first and third quartiles (dashed bars). (B) Vδ₁/Vδ₂ subset ratio among γδ⁺ T cells in the blood according to HIV and CMV infection status. (C) Violin plots of CX3CR1⁺ cell frequencies among Vδ₁ T cells in the blood according to HIV and CMV infection status. Comparisons were made using Welch’s t-test. (D) Violin plots of CMV IgG index in HIV⁺ CMV⁺ (n = 15) and HIV^- CMV⁺ individuals (n = 15). Comparisons were made using Welch’s t-test. See also S8 Fig for plasma inflammatory biomarkers comparisons between the same groups.

Fig 7. PLWH show changes in duodenal mucosal and blood translocated microbiota.

Cladogram of phylogenetic relationships of bacterial lineages associated with HIV status and histogram of the linear discriminant analysis (LDA) scores for differentially abundant bacterial phyla, classes, orders, families and genera between PLWH and HIV seronegative controls. The taxonomic level in the cladograms is represented by phylum in the inner ring and genus in the outer ring, with each circle representing a taxon within that level. LDA scores represent the size and ranking of each differentially abundant taxon. Positive values (in green) represent higher abundance in HIV seronegative controls and negative values (in red) represent higher abundance in PLWH. (A) Duodenal biopsies (n = 24 PLWH and n = 25 HIV seronegative controls). LDA scores shown are *P < 0.05; (B) Blood samples (n = 33 PLWH and n = 39 HIV seronegative controls). LDA scores shown have threshold > 3 and **P < 0.01. See also S7 Fig for principal component analysis and diversity indices.

Fig 8. Cytotoxic CX3CR1⁺ TEMRA Vδ₁ T cell expansion is associated with the interplay of CMV, microbiota changes and HIV-1 persistence.

(A) Chord diagrams of the associations in PLWH between (i) CMV IgG index, changes in (ii) duodenal microbiota (relative abundance of some Proteobacteria: Haemophilus; Bacteroidetes: Porphyromonas; and Actinobacteria: Propionibacterium), (iii) the blood microbiota (relative abundance of some Proteobacteria: Candidatus pelagibacter, Arcobacter, Pelomonas; and Actinobacteria: Propionibacterium), and the phenotype of circulating and duodenal Vδ₁ T cells (CX3CR1⁺ TEMRA Vδ₁ T cells and their cytotoxic or activating profile). Color scale indicates Spearman’s correlation coefficient (P < 0.05). Positive correlations are shown in red and negative correlations are shown in blue. (B) Chord diagrams of the associations between the phenotype of circulating and duodenal Vδ₁ T cells (CX3CR1⁺ TEMRA Vδ₁ T cells and their cytotoxic, activating or exhausted profile) and their clonal expansion (inverse of TCR diversity), virological parameters of the HIV-1 reservoir (frequency of intact proviral DNA among total HIV-1 DNA in (i) PMBC and (ii) gut;cell-associated HIV-1 RNA in PBMC and duodenal mucosa) and (iii) clinico-biological parameters of PLWH (percentage of CD4⁺ and CD8⁺ T cells, CD4/CD8 ratio, and nadir of CD4⁺ T cell count in blood; duration between HIV-1 diagnosis and enrollment; duration of consecutive undetectable HIV viral load prior to enrollment). Color scale indicates Spearman’s correlation coefficient (P < 0.05). Positive correlations are shown in red and negative correlations are shown in blue. (C) Dot plots of correlations between the frequency of circulating cytotoxic (GzmB⁺Perforin⁺) Vδ1 cells and (first panel) the CMV IgG index (n = 15 PLWH), and (second panel) Porphyromonas relative abundance in the duodenal mucosa (n = 11 PLWH); correlations between the frequency of circulating CX3CR1⁺ Vδ₁ T cells and (third panel) cell-associated HIV-1 RNA in PBMC (copies per million of PBMC, n = 15), and (fourth panel) Porphyromonas relative abundance in the duodenal mucosa (n = 11 PLWH). Spearman’s correlation coefficients (ρ) and P-values are shown. See also S9 Fig for correlations with plasma inflammatory biomarkers.