Altered differentiation is central to HIV-specific CD4⁺ T cell dysfunction in progressive disease

Abstract

Dysfunction of virus-specific CD4⁺ T cells in chronic human infections is poorly understood. We performed genome-wide transcriptional analyses and functional assays of CD4⁺ T cells specific for human immunodeficiency virus (HIV) from HIV-infected people prior and after initiation of antiretroviral therapy (ART). A follicular helper T cell (T_FH cell)–like profile characterized HIV-specific CD4⁺ T cells in viraemic infection. HIV-specific CD4⁺ T cells from people spontaneously controlling the virus (elite controllers) robustly expressed genes associated with the T_H1, T_H17 and T_H22 subsets of helper T cells. Viral suppression by ART resulted in a distinct transcriptional landscape, with a reduction in the expression of genes associated with T_FH cells but persistently low expression of genes associated with T_H1, T_H17 and T_H22 cells compared to the elite controller profile. Thus, altered differentiation is central to the impairment of HIV-specific CD4⁺ T cells and involves both gain of function and loss of function.

CD4⁺ T cell responses are critical for durable control of viral replication in chronic infections^{1, 2}. Virus-specific CD4⁺ T cell immunity is of particular interest for HIV infection, which is characterized by functional impairment and destruction of CD4⁺ T cells. While classical models divided CD4⁺ T cells into distinct lineages, studies have demonstrated the importance of CD4⁺ T cell plasticity³. Persistent antigen and inflammatory signals cause impairment of antigen-specific responses, a state called immune exhaustion. Virus-specific CD8⁺ T cell exhaustion has been extensively investigated⁴ and represents a bona fide cell differentiation program. These studies highlighted the relevance of genome-wide transcriptional studies to understand T cell impairment⁵. Compared to CD8⁺ T cells, less is known on CD4⁺ T cell dysfunction. Murine LCMV-specific CD4⁺ T cells in chronic infection, while exhibiting some characteristics shared with CD8⁺ T cells, also present distinct features^{6, 7}, including loss of a T_H1-signature⁵ and skewing towards a T follicular helper (T_FH) phenotype^{8, 9}.

Whether findings in mice can be extrapolated to human HIV infection is unclear. Some features of virus-specific CD4⁺ T cells are shared between both infections: upregulation of co-inhibitory receptors are found in HIV progressors with ongoing viremia (chronic progressors, CP) and chronic LCMV infection^{5, 10}. Rare subjects who spontaneously suppress HIV (elite controllers, EC) frequently exhibit robust virus-specific T_H1 responses¹¹ and strong proliferative capacity, similarly to mice infected with the acute strain of LCMV. However, HIV and LCMV are distinct viruses and there are notable differences between species in terms of T cell differentiation mechanisms, such as T_FH generation¹². An issue of critical clinical relevance is the lack of restoration of effective anti-HIV immunity after suppressive antiretroviral therapy (ART): viral rebound is the rule after cessation of therapy. Whether persistent HIV-specific CD4⁺ T cell dysfunction on ART contribute to this failed response is an important, yet unresolved, question. The paucity of experimental tools capable of identifying highly heterogeneous antigen-specific CD4⁺ T cells has hampered the study of HIV-specific CD4⁺ T cell help. Intracellular cytokine assays (ICS) are of limited sensitivity for many non-T_H1 effector functions, and the use of HLA Class II tetramers in humans is constrained by availability, requirement for pre-defined epitopes and genetic diversity.

To determine key pathways and molecules that link HIV-specific T cell help to viral control, we here performed genome-wide transcriptional analyses and functional assays of HIV-specific CD4⁺ T cells from HIV-infected humans with diverse viral loads prior to ART initiation and followed a subgroup of them longitudinally after viral suppression on therapy.

RESULTS

Links between HIV-specific CD4 transcriptome profiles and viremia

To define molecular features that discriminate HIV-specific CD4⁺ T cells in progressive vs. controlled infection, we performed a cross-sectional study of 38 chronically infected people who were untreated at the time of sampling. These included “elite controllers” (EC, HIV plasma viral load <50 vRNA copies/ml), “viremic controllers” (VC, viral load between 50 and 5,000 copies/ml) and “chronic progressors” (CP, viral load ≥5,000 copies/ml) (Participant characteristics: Supplementary Table 1).

We utilized an activation induced marker (AIM) assay to identify CD4⁺ T cells specific for the Gag protein (hereafter termed HIV-specific CD4⁺ T cells). Ex vivo stimulated HIV-specific CD4⁺ T cells were identified by the co-upregulation of CD40L and CD69 on their surface after a 9-h stimulation with an HIV Gag peptide pool¹³ (Fig. 1a, Supplementary Fig. 1a). Combining two markers enhanced detection of HIV-specific CD4⁺ T cells by reducing background compared to CD40L alone (Supplementary Fig. 1b,c). This AIM assay overcomes limitations of cytokine-based techniques for detection of virus-specific cells, allows live-cell sorting and captures a broader Ag-specific CD4⁺ T cell population (Supplementary Fig. 1d). There was no significant difference in the magnitude of HIV-specific CD4⁺ T cell responses amongst cohorts or correlation with viremia or CD4 count (Fig. 1b, Supplementary Fig. 1e,f).

Figure 1. Deep coverage transcriptome analysis of HIV-specific CD4⁺ T cells from untreated HIV-infected people with distinct disease status.

(a) Representative flow cytometry plots from an EC of detection of HIV-specific CD4⁺ T cells by an activation-induced marker (AIM) assay based on co-upregulation of CD40L and CD69 9h after stimulation with an HIV Gag peptide pool. (b) Quantification of (a). Comparison of net frequency (background of No Antigen condition subtracted) of CD69⁺CD40L⁺ Gag-specific CD4⁺ T cells amongst the CP, VC and EC groups by Kruskal-Wallis test. Horizontal lines and bars represent median and +/−interquartile range, respectively (n= 11CPs, 11 VCs, 13 ECs). (c) Paired volcano plot comparisons depicting differentially expressed genes (DEG) as measured by microarray analysis of sorted HIV-specific CD4⁺ T cells in the three groups. Red, pink and blue dots represent DEG with log2|FC| >1 and padj<0.05, log₂|FC|>1 or padj<0.05, respectively. A two-sided moderated t-test followed by Benjamini-Hochberg method correction was applied. (n= 11 CPs, 9 VCs, 12 ECs). (d) Euler diagrams of DEG between each pairwise comparison of the three groups of participants (log₂FC> 1, p<0.05); numbers represent DEGs. (e) Top significant enriched Gene Ontology terms from the curated C5 MSigDB via two-sided CAMERA analysis. Blue, Red and white color denote negative, positive or not significant (as defined by FDR<0.05 by Benjamini-Hochberg method), respectively.

We reasoned that the AIM assay could reveal the transcriptome modulation in response to cognate antigen-mediated TCR signaling, reflecting the functional state of CD4⁺ T cells. We thus performed microarray analysis of sorted ex vivo stimulated HIV-specific CD4⁺ T cells. Pairwise comparisons between cohorts (Fig. 1c) revealed sets of differentially expressed genes (DEGs) between all three cohorts, with extensive dissimilarities between CPs and ECs. Fewer transcriptional differences were detected between the VC and EC groups. The majority of DEGs between VCs and ECs were also observed in the CPs versus ECs comparison (Fig. 1d). CPs and VCs showed the fewest DEGs, suggesting that viremia strongly impacts the transcriptional landscape of HIV-specific CD4⁺ T cells.

We next used CAMERA (Correlation Adjusted Mean Rank gene set test)¹⁴ enrichment analysis to investigate molecular mechanisms underlying HIV-specific T help alterations associated with disease status. We compared the identified DEGs to the Molecular Signatures Database (MSigDB) C5 collection of curated gene sets, which comprises the Gene Ontology (GO) terms associated with biological processes, molecular functions and cellular components (Fig. 1e). The “Positive regulation of T helper cell differentiation” signature was negatively associated with viremia. Consistently, the “Positive regulation of CD4⁺ αβ T cell activation” and “Regulation of leukocyte migration” gene sets were significantly underrepresented in CP compared to both VCs and ECs, contrasting with upregulation of the MHC class II protein complex in CPs. Expression of genes associated with protein transport and/or processing was reduced in CPs. The pathways enriched with viremia included processes associated with chromatin modification, mitotic cell cycle, DNA packaging, chromosome segregation, response to stress, and apoptosis.

This approach identifies potential pathways associated with HIV-specific CD4⁺ T cell dysfunction, and points at a central role for altered T cell differentiation, regulation and activation in this impairment.

HIV-specific CD4 polarization is associated with disease status

Given the potential role for altered T cell differentiation in HIV-specific CD4⁺ T cell impairment, we determined features of CD4⁺ T cell differentiation associated with disease status. We compared by CAMERA our sets of DEGs with human T helper polarization signatures from the C7 collection MSigDB and publicly available microarray datasets. While the different conditions used in the generation of these signatures (e.g activated vs unactivated cells) may complicate the interpretation of such associations, we observed notable differences amongst cohorts. Compared to CPs, ECs and VCs displayed enriched T_H1 and T_H17 signatures^{15, 16, 17} (Fig. 2a,b,e,f; Supplementary Fig. 2a; Supplementary Tables 2–4). Cardinal genes associated with T_H17 cell differentiation and function were upregulated in ECs (Fig. 2f). In viremic people (CPs, VCs), we noted a striking enrichment of a T_FH signature¹⁸ (Fig. 2c,g, Supplementary Fig. 2a) compared to ECs. There was congruent upregulation of several key T_FH-associated genes compared to EC with the notable exception of Bcl6 that was instead upregulated in ECs (Fig. 2g, Supplementary Table 5). Although we did not observe a statistically significant enrichment of an LCMV-specific CD4⁺ T cell exhaustion signature⁵ in CPs compared to ECs (Fig. 2d) (which might be due at least in part to the methodological differences), several of the DEGs enriched in CP were well-known exhaustion-associated genes (Fig. 2h, Supplementary Table 6). Several Interferon-stimulated genes (ISGs) upregulated in the LCMV model of exhaustion, such as IFH1, IFIT3 and IFI44, were upregulated in ECs.

Figure 2. HIV-specific CD4⁺ T cell polarization is associated with level of viral control.

(a-d) Barcode plots of enriched Thelper polarization and exhaustion gene sets by CAMERA in CP (n=11) vs EC (n=12) people. Red and blue lines denote positive and negative enrichment of CP genes vs. EC, respectively. (A) T_H1 signature: GSE59295 (T_H1 vs T_H2), (B) T_H17 signature: GSE49703 (T_H17 vs T_H1), (C) T_FH signature: GSE50391 (CXCR5high CD45RO vs CXCR5-tonsil samples), (d) LCMV exhaustion signature GSE41866 (LCMV Clone 13 D30 vs Armstrong D30-Exhausted vs Memory). Two sided p-value by CAMERA and FDR by Benjamini-Hochberg method. The arrows beneath the barcode plots depict the orientation of the enrichment between CP and EC. (e-h) Heatmaps of mRNA expression of the top 50 significant genes of each CD4⁺ T cell signature in CPs and ECs. Red and blue font of gene names denote upregulated and downregulated mRNA expression in CPs vs ECs, respectively. (i) PCA representation of CPs (n=11), VCs (n=9) and ECs (n=12) based on transcriptional expression of 90 genes associated with CD4⁺ T cell differentiation, effector functions and exhaustion. Transcriptional expression (-ΔCt values) was evaluated by qRT-PCR (Fluidigm). The top 35 contributing genes are displayed and color-coded according to their contribution. Bigger dots represent the mean PC coordinates for each group. The numbers in parenthesis stand for the percentage of variance that the PC accounts for. (j) Correlation of PC1 values for each individual with viral load by two-tailed Spearman correlation (n= 11CPs, 9 VCs, 12 ECs).

We next used qRT-PCR to further quantify levels of key DEGs identified by microarrays and selected additional CD4⁺ T cell genes. Principal component analysis discriminated CPs from ECs, whereas there was more overlap between CPs and VCs (Fig 2i). PC1 values significantly inversely correlated with viremia, showing that changes in the CD4⁺ T cell profile relate to antigen load (Fig 2j). The qRT-PCR data were consistent with the microarray results (Supplementary Table 7). They confirmed that genes associated with T_FH functions and B cell help (CXCL13, SLAMF6, CD84) and exhaustion (TIGIT, CD200, LAG3, TOX) were enriched in CPs. This contrasted with genes related to T_H17 and T_H22 functions (IL17F, IL22) and the interferon-induced chemokines (CXCL9, CXCL10, CXCL11), which were enriched in ECs (Fig 2i). Expression of selected transcription factors was consistent with these observations (Supplementary Fig 2b): ECs expressed significantly more TBX21 and AHR, genes respectively coding for T_H1 and T_H22 master transcription factors, and there was in ECs a trend for enhanced expression of RORC, a master transcription factor of T_H17 differentiation. The behavior of the master T_FH transcription factor BCL6 contrasted to most T_FH associated genes: BCL6 was significantly upregulated in ECs compared to CPs in contrast to MAF, another transcription factor critical for T_FH differentiation, which was highly expressed in CPs. Helios (IKZF2), a transcription factor overexpressed in LCMV-specific CD4⁺ T cells in chronic infection, was also upregulated in CPs compared to ECs. Examining correlations between gene expression revealed positive and negative associations of genes, many not anticipated by classical dichotomic models of T_H differentiation. These results are consistent with a complex spectrum of differentiation and functional profiles of HIV-specific CD4⁺ T cells (Supplementary Fig 2c,d).

These data show higher expression of a T_FH-like signature in HIV-specific CD4⁺ T cells in progressive infection, contrasting with enriched T_H1 and T_H17 signatures in subjects achieving spontaneous viral control.

Upregulation of classical T_FH-associated genes in CXCR5_neg HIV-specific CD4 of CPs

We next investigated whether the enrichment of the T_FH signature in HIV-specific CD4⁺ T cells of CPs compared to ECs was due to a higher fraction of HIV-specific circulating T_FH cells (cT_FH). We used the CD69/CD40L AIM assay to assess the percentages of HIV-specific CXCR5⁺CD45RA^– memory CD4⁺ T cells (CXCR5_mem) in both cohorts (Fig. 3a and Supplementary Fig. 3a,b). As PD-1 expression on HIV-specific CD4⁺ T cells correlates with viremia¹⁹, PD-1 was omitted as a T_FH marker to avoid bias towards CPs. We observed only a trend for higher fractions of HIV-specific CXCR5_mem in CPs than ECs, with considerable overlap between groups (Fig. 3b). In contrast, single subject GSEA (ssGSEA) scores showed a significant enrichment of the T_FH signature in CPs compared to ECs (Fig. 3c), consistent with the cohort-based CAMERA analyses of Fig. 2c,g. The T_FH signature enrichment by ssGSEA correlated directly with viremia (Fig. 3d). These data suggest that differences in percentage of CXCR5_mem alone are not sufficient to explain the markedly enriched T_FH signature observed in HIV-specific CD4⁺ T cells of CPs.

Figure 3. An atypical T_FH-like transcriptional signature in CXCR5_neg HIV-specific CD4⁺ T cells discriminates chronic progressors from elite controllers.

(a) Representative flow cytometry plots from a CP illustrating the enrichment of cT_FH (CXCR5⁺ CD45RA-) in CD69⁺CD40L⁺ Gag-specific CD4⁺ T cells vs. total population in a CP subject, assessed 9h after stimulation with a Gag peptide pool. (b,c) Quantification from (a). Comparison between CPs and ECs of (b) the percentage of Gag-specific CXCR5_mem assessed by flow cytometry (two-sided Mann-Whitney test) (n=10 CPs and 10 ECs); and (c) the enrichment score of a T_FH tonsil signature (GSE50391) by ssGSEA (two-sided Mann-Whitney test) (n_CP=11, n_EC=12). Bars represent the median +/−interquartile range. (d) Correlation of single-sample enrichment score for the T_FH tonsil signature with viral load by two-tailed Spearman test (n= 11 CPs, 9 VCs, 12 ECs). (e) Volcano plots depicting differential gene expression between CP and EC, amongst their live-sorted Gag-specific CXCR5_mem or CXCR5_neg subsets, by microarray analysis. Red and pink dots represent DEG with log₂|FC|>1 and padj<0.05 or log₂|FC| >1, respectively, and the number of DEGs are displayed in the small tables. Two-sided moderated t-test followed by Benjamini-Hochberg correction method (padj) (n= 5 CPs and 3 ECs). (f) CAMERA enrichment analysis of the tonsil T_FH signature of CXCR5_mem or CXCR5_neg Gag-specific CD4 T cells of CPs compared to ECs (n= 5 CPs and 3 ECs). Two sided p-value by CAMERA and FDR by Benjamini-Hochberg method. (g) Comparison of expression of T_FH surface markers on CXCR5_mem and CXCR5_neg HIV-specific CD4⁺ T cells of CPs and ECs by flow cytometry (n= 10 CPs and 10 ECs). (h) mRNA expression of T_FH-related surface markers or (i) transcription factors by qRT-PCR (Fluidigm) (n= 6 CPs and 6 ECs) on sorted CXCR5_mem and CXCR5_neg HIV-specific CD4⁺ T cells of CPs and ECs. (g,h,i) Two-tailed Mann-Whitney test or for paired comparisons by two-tailed Wilcoxon matched-pairs signed-ranked test were used to verify significance. Only p˂0.05 are displayed for clarity. Bars represent the median ⁺/− interquartile range.

To define the cell subsets that contributed to the T_FH transcriptional signature observed in CPs, we applied microarray analysis to paired live-sorted subsets of CXCR5_mem and CXCR5_neg HIV-specific CD4⁺ T cells from CPs and ECs (Fig.3e; gating strategy Supplementary Fig. 3a). CXCR5 mRNA expression was consistent with protein expression (Fig. 3h). Differences in gene expression between CPs and ECs were more significant in the CXCR5_neg than in the CXCR5_mem subsets of HIV specific CD4⁺ T cells (Fig. 3e). CPs displayed a marked enrichment of the T_FH signature not only in the CXCR5_mem compartment, but also in the CXCR5_neg HIV-specific CD4⁺ T cell subset (Fig. 3f). We confirmed expression of several markers classically associated with T_FH cells in CXCR5_mem and CXCR5_neg HIV-specific CD4⁺ T cells at the mRNA level by qRT-PCR and at the protein level by flow cytometry. Consistent with the microarray results, upregulation of the TCR co-receptors PD1, TIGIT, CD200 and ICOS on CXCR5_neg HIV-specific CD4⁺ T cells clearly differentiated CPs from ECs (Fig. 3g and Supplementary Fig. 3c). Surface markers SLAMF6 and CD84 (both associated with help to B cells) and BTLA (highly expressed in T_FH cells; Fig. 3h) were significantly higher in CXCR5_neg CD4⁺ T cells of CPs compared to ECs, while the mRNA expression of BCL6 and TCF7 did not differ in CXCR5_neg CD4⁺ T cells between the two groups. Transcriptional expression of PRDM1 and MAF was significantly higher in CXCR5_neg CD4⁺ T cells of CPs compared to ECs (Fig. 3i).

In contrast to ECs, CXCR5 expression in CPs did not clearly delineate cells with a T_FH transcriptional profile: ICOS, SLAMF6, BTLA, CD84, TIGIT, CD200 and the T_FH-associated transcription factor cMAF mRNA levels in CXCR5_neg cells were equal or even superior to their CXCR5_mem autologous counterparts (Fig. 3ghi, Supplementary Fig. 3c). These data demonstrate the atypical upregulation of “T_FH-like” signature in the CXCR5_neg HIV-specific CD4⁺ T cells of CPs compared to ECs.

Robust expression of T_FH cytokines in CXCR5_neg HIV-specific CD4 of CP

We next sought to elucidate if the enriched T_FH signature in CXCR5_mem and CXCR5_neg HIV-specific CD4⁺ T cells of CPs compared to ECs is congruent with the enhancement of some T_FH functions and determine the phenotype of these effector cells. We assessed mRNA expression of the T_FH cytokines CXCL13 and IL-21 at single-cell resolution using a sensitive flow cytometric RNA Fluorescent In Situ Hybridization (RNA Flow FISH), as previously described^{20, 21}. This assay is superior to standard ICS for detection of IL-21-producing cells²⁰. There were increased frequencies of CXCL13 mRNA⁺ and IL21 mRNA⁺ HIV-specific CD4⁺ T cells in CPs compared to ECs (Fig. 4a,b). This was a distinct feature of HIV-specific responses, as the expression after stimulation with the superantigen SEB was similar in both cohorts (Fig. 4c) and CMV pp65-specific CD4⁺ T cells expressed less CXCL13 and IL21 than HIV-specific CD4⁺ T cells in paired comparisons within the same humans (Supplementary Fig. 4a). As CD40L is not expressed on all Ag-specific CD4⁺ T cells upon activation, we confirmed that the differential expression of CXCL13 and IL21 between CPs and ECs hold true without combining the RNA Flow FISH and AIM assays and without pre-gating on CD40L⁺ cells (Supplementary Fig. S4b–i). Largely distinct CD4⁺ T cell populations produced CXCL13 and IL21 mRNAs at the time point examined (Fig. 4d,e). We confirmed the increased CXCL13 production in CPs compared to ECs at the protein level, by measuring CXCL13 concentration by ELISA in the supernatant of CD8-depleted PMBCs after 48h stimulation with HIV Gag (Supplementary Fig. 5a). CXCL13⁺ and IL21⁺ cells expressed high levels of CXCR5 protein to compared to total HIV-specific CD4⁺ T cells in ECs. In contrast, the levels of CXCR5 expression on CXCL13⁺ and IL21⁺ cell were comparable to the total population in CPs (Fig 4f). This gain of T_FH-like functions by CXCR5_neg HIV-specific CD4⁺ T cell in CPs was further apparent when we compared the ratios of IL21⁺ and CXCL13⁺ responses between the CXCR5_mem and CXCR5_neg subsets (Fig. 4g). As the inhibitory receptors PD-1 and TIGIT are markers of T_FH cells¹² and may also be associated with exhaustion, we examined their expression on CXCL13 mRNA⁺ and IL21 mRNA⁺ HIV-specific CD4⁺ T cells. The majority of these cytokine-expressing cells were PD-1⁺/TIGIT⁺, particularly in CPs (Fig. 4h,i, Supplementary Fig. 5b,c.). Thus, the expression pattern of the inhibitory co-receptors PD-1 and TIGIT was consistent with T_FH skewing and may not imply exhaustion of these responses.

Figure 4. T_FH cytokine expression by CXCR5_neg HIV-specific CD4⁺ T cells of chronic progressors.

(a) Representative RNA flow FISH plots from an CP of CXCL13 and IL21 mRNA in CD69⁺CD40L⁺ CD4⁺ T cells after a 15h stimulation with a Gag peptide pool. Numbers represent the frequency of mRNA⁺ cells among (red) CD69⁺CD40L⁺ and (grey) total CD4⁺ T cells (n= 6 CPs and 5 ECs) (b,c) Quantification of (a). Comparison of frequencies of CXCL13 or IL21 mRNA⁺ CD4⁺ T cells between CPs (n=6) and ECs (n=5) obtained as in a after Gag or SEB stimulation. Bars represent median frequencies +/−interquartile range (two-tailed Mann-Whitney test, **p<0.01). (d,e) RNA flow FISH analysis of co-expression of CXCL13 and IL21 mRNA by Gag-specific CD4⁺ T cells with (d) representative plots from CP and EC people and (e) related statistical analysis by permutation test (10000 permutations) in SPICE (n= 6 CPs and 5 ECs). Pie slices represent median frequency of CXCL13/IL21 mRNA⁺ subpopulations. (f) Representative histograms from CP and EC people of CXCR5 expression on CXCL13 or IL21 mRNA⁺ HIV-specific CD4⁺ T cells overlaid on total HIV-specific CD4⁺ T cells (grey). Numbers represent the percentages of CXCR5⁺ cells within each cytokine⁺ population (n= 6 CPs and 5 ECs). (g) Statistical analysis from (f) on 6 CPs and 5 ECs of the ratios between the CXCR5_mem to the CXCR5_neg subsets of frequencies of cytokine or chemokine mRNA⁺ HIV-specific CD4⁺ T cells (CXCL13, IL21, IL2 and IFNγ) as assessed by RNA flow cytometry. Bars represent median frequencies +/−interquartile range (two-tailed Mann-Whitney test). (h,i) Co-expression patterns of PD-1 and TIGIT (surface protein Ab labeling) on CXCL13 mRNA⁺ or IL21 mRNA⁺ Gag-specific CD4⁺ T cells identified by RNA flow FISH (n= 6 CPs and 5 ECs). (h) Representative examples of CP and EC people; cytokine/chemokine mRNA⁺ HIV-specific CD4⁺ T cells (red dots for CPs, blue dots for ECs) are overlaid on total HIV-specific CD4⁺ T cell subpopulations (grey dots). Numbers represent the percentages of CXCL13 RNA⁺ or IL21 mRNA⁺ HIV-specific CD4⁺ T cells located in each quadrant. (i) SPICE analysis of PD-1 and TIGIT expression from (h) on CXCL13 mRNA⁺ or IL21 mRNA⁺ HIV-specific CD4⁺ T cells. (*p<0.05, **p<0.01 by permutation test, 10000 permutations). Slices represent median frequency of PD-1 and TIGIT subpopulations. (j) Comparison of p24-specific antibodies levels in plasma between CPs and ECs as assessed by ELISA. (two-tailed Mann-Whitney test (n= 12 CPs and 13). Bars represent medians +/− interquartile range. (k,l,m) Correlation between p24-specific antibodies levels measured by ELISA and frequencies of k) total HIV-specific CD4⁺ T cells; l) CXCR5_mem HIV-specific CD4⁺ T cells and m) CXCR5_neg HIV-specific CD4⁺ T cells assessed by the CD69/CD40L AIM assay after a 9h stimulation with a Gag peptide pool (two-tailed Spearman test, n=12 CPs and 13 ECs).

Finally, we explored possible links between HIV-specific CD4⁺ T cell help and antibody (Ab) responses. We assessed plasma levels of p24-specific Abs by ELISA and Biolayer Interferometry (BLI) binding analysis. The results obtained by the two methods were highly correlated (Supplementary Fig. 5d). Although p24-specific Ab concentrations in the CP and EC groups were overall similar (Fig. 4j), we observed a significant correlation between the magnitude of HIV-specific CD4⁺ T cell responses and p24-specific Abs, which was maintained when only the CP group was considered (Fig 4k). There was no significant association between CXCR5_mem responses and p24-specific antibodies (Fig. 4l). However, there was a significant correlation between CXCR5_neg HIV-specific CD4⁺ T cell frequencies and Ab responses in CPs, but not ECs (Fig. 4m), which might suggest that the CXCR5_neg subset has T helper function to B cells in vivo.

These data demonstrate that dysregulated CXCR5_neg HIV-specific CD4⁺ T cells in progressive infection gain functions and phenotypic features classically attributed to T_FH cells.

Mucosal immunity-related cytokines in HIV-specific CD4 of ECs

As we observed an enriched T_H17 signature in HIV-specific CD4⁺ T cells in ECs compared to CPs, we characterized the expression of IL-17F and IL-22 at the single-cell level. RNA Flow FISH enabled superior detection of these cytokines in HIV-specific CD4⁺ T cells compared to ICS (Fig. 5a,b). We observed more robust IL17F and IL22 HIV-specific CD4⁺ T cell responses in ECs compared to CPs (Fig 5c). In contrast, IL17F and IL22 expression patterns in SEB- or CMV activated CD4⁺ T cells did not differ between the groups (Fig. 5d, Supplementary Fig. S4a). We confirmed the strong IL-17F, IL-22 and IL-17A secretion by HIV-specific CD4⁺ T cells in ECs at the protein level, in contrast to low or undetectable production in CPs (Fig. 5e, Supplementary Fig. 5e,f,g).

Figure 5. Increased expression of cytokines related to mucosal immunity in HIV-specific CD4⁺ T cells of elite controllers compared to chronic progressors.

(a,b). Representative flow cytometry plots depicting detection of IL22 and IL17F in CD69⁺CD40L⁺ CD4⁺ T cells in one EC after stimulation with a Gag peptide pool by delayed ICS for cytokine protein and RNA flow FISH for cytokine mRNA, respectively. Numbers represent frequencies of mRNA⁺ HIV-specific CD4⁺ T cells among (blue) CD69⁺CD40L⁺ and (grey) total CD4⁺ T cells (n= 8 CPs and 8 ECs). (c,d) Statistical comparison from (a) of frequencies of cytokine mRNA⁺ antigen-specific CD4⁺ T cells between CPs and EC after (c) a Gag peptide pool (n= 8 CPs and 8 ECs) or (d) SEB stimulation (n=6 CPs and 8 ECs) (two-tailed Mann-Whitney test). (e) Statistical comparison of IL-22 and IL-17F protein levels detected by Luminex beads array in the supernatant of CD8-depleted PBMCs 48h after stimulation with a Gag peptide pool (n= 8 CPs and 8 ECs). (c,d,e) Bars represent median with +/−interquartile range. *p<0.05, **p<0.01 by two-tailed Mann-Whitney test. (f) Representative flow cytometry plots of co-expression patterns of IL22 mRNA and IL17F mRNA in HIV-specific CD4⁺ T cells in CP and EC people (n=6 CPs and ^ ECs). (g) Statistical analysis from (f) by SPICE (*p<0.05 by permutation test, 10000 permutations). Slices represent median frequency of IL22/IL17F mRNA⁺ subpopulations of HIV-specific CD4⁺ T cells (n= 6 CPs and 6 ECs). (h,i) Comparison of the frequencies of CCR6 and CXCR3 expression on IL22 mRNA⁺ and IL17F mRNA⁺ HIV-specific CD4⁺ T cells compared to total HIV-specific CD4⁺ T cells in CPs (n=6) and ECs (n=6) by two-tailed Mann-Whitney test and by two-tailed Wilcoxon matched-pairs signed-ranked test. Only p˂0.05 are displayed for clarity. Bars represent median with interquartile range. (j,k) Correlation of (j) IL22 and (k) IL-17F mRNA levels assessed by qRT-PCR on sorted HIV Gag-specific CD4⁺ T cells and CD4⁺ T cell activation measured by HLA-DR and CD38 co-expression (two-tailed Spearman test; n= 10 CPs, 6 VCs, 12 ECs). (l,m) Statistical comparisons of microbial translocation in plasma in cohorts of CP, EC, and uninfected control donors (UD) (l) quantitation of bacterial RNA reads in plasma (transcripts per million, TPM); (m) bacterial RNA species diversity (Shannon entropy score) (Kruskal Wallis test; n= 10 CPs, 8 ECs and 6 UDs). Bars represent median +/−interquartile range. (n,o) Correlation between bacterial RNA species diversity in plasma and frequencies of (n) HIV-specific IL22⁺ CD4⁺ T cells and (o) HIV-specific IL17⁺ CD4⁺ T cells (two-tailed Spearman test; n=10 CPs and 8 ECs). (p,q) Correlation between abundance of Proteobacteria translocation and frequencies of (p) HIV-specific IL22⁺ CD4⁺ T cells and (q) HIV-specific IL17⁺ CD4⁺ T cells (two-tailed Spearman test; n=10 CPs and 8 ECs).

The IL17F mRNA⁺ and IL22 mRNA⁺ subsets constituted two partially overlapping cell populations (Fig. 5f,g). In both CPs and ECs, these cells preferentially expressed CCR6, a chemokine receptor critical for recruitment of T_H17 and T_H22 cells to mucosal tissues (Fig. 5h). Expression of CXCR3, a marker of T_H1 cells also expressed on T_H1-T_H17 cells of mixed properties that are expanded in some inflammatory diseases¹⁷, was low on IL17F and IL22 mRNA⁺ of both CPs and ECs (Fig 5i, Supplementary Fig. 5h). Performing the RNA Flow FISH without the AIM assay and CD40L⁺ pre-gating showed again contrasting patterns of IL17F and IL22 expression between CPs and ECs (Supplementary Figure S4b–i).

Systemic translocation of microbial products play a critical role in immune activation related to HIV pathogenesis²². Loss of gut mucosal integrity and T_H17 cell depletion have been associated with immune activation and disease progression²³. We thus investigated possible links between HIV-specific T_H17 and T_H22 responses and systemic immune activation. We observed an inverse correlation between IL22 mRNA expression by HIV-specific CD4⁺ T cells and co-expression of the activation markers HLA-DR and CD38 on total CD4⁺ and CD8⁺ T cells (Fig. 5j, Supplementary Fig. 5i). We identified similar associations for IL17F (Fig. 5k, Supplementary Fig. 5j). We next investigated the relationship between preservation of IL17⁺ and/or IL22⁺ HIV-specific CD4⁺ T cells and microbial translocation by deep sequencing of cell-free RNA fragments in plasma — a reflection of the part of the body microbiome that can access the circulation, the majority of which resides in the gut. The total levels of bacterial RNA reads in plasma did not significantly differ between the groups of CPs, ECs and uninfected control donors (UDs) (Fig. 5l). However, we observed a lower plasma bacterial RNA species diversity in CPs compared to the EC and UD groups (Fig. 5m). These observations may reflect low gut microbiome diversity, a state associated with systemic inflammation in several diseases²⁴ (Supplementary Fig. 5k,l). IL22 and IL17F transcription in HIV-specific CD4⁺ T cells was significantly correlated with bacterial RNA species diversity (Fig. 5n,o). CPs and ECs qualitatively differed in the bacterial RNA profile, forming largely distinct clusters (Supplementary Fig. S5m). IL22 and IL17F mRNA levels in HIV-specific CD4⁺ T cells inversely correlated with the prevalence of Proteobacteria, a phylum of Gram-negative bacteria that is increased in the gut microbiome in HIV-infected people²⁵ and has been proposed as a marker of dysbiosis²⁶ (Fig. 5p,q).

Thus, robust IL22⁺ and IL17F⁺ HIV-specific CD4⁺ T cells can be detected in the blood of ECs, while these responses are severely impaired in CPs. Associations of this deficiency with systemic immune activation, reduced plasma bacterial RNA species diversity with a skewing towards greater Proteobacteria abundance suggest that these findings in blood reflect perturbed gut homeostasis and loss of protective mucosal immunity.

ART results in a distinct transcriptome of HIV-specific CD4

Finally, we sought to determine whether the observed HIV-specific CD4⁺ T cell profiles were the consequences of ongoing viral antigen exposure, or the results of durable cell-fate decision programs that would persist after viral suppression on ART. We longitudinally followed a subgroup of 8 CPs and analyzed a second sample obtained while on ART (participant characteristics: Supplementary Table 8). Pairwise comparisons between the EC, CP pre-ART (CPpre) and CP post-ART (CPpost) cohorts (Fig 6a,b) showed sets of DEGs between all three groups, with fewer differences in the CPpre-CPpost and CPpost-EC than the CPpre-EC comparison, as may be expected. We next focused on selected GO terms and gene signatures identified as differentially expressed between CPs and ECs in Figs 1 and 2. While initiation of ART in CP modulated the activity of several apoptotic pathways regulating apoptosis and gene expression towards the levels observed in EC, differences in the “Positive regulation of T helper cell differentiation”, “Positive regulation of CD4⁺ αβ T cell activation” and “Regulation of leukocyte migration” persisted (Fig 6c). We therefore examined how ART affected Thelper polarization signatures. Expression of the T_FH signature in HIV-specific CD4⁺ T cells was markedly reduced by ART, to levels similar to that found in EC (Fig 6d), while expression of the T_H1, T_H17 and LCMV exhaustion signatures was little affected by therapy: the differences clearly persisted between CPpost and EC (Fig 6e,f, Supplementary Fig. S6a). Expression of selected individual genes of interest was also consistent with these findings when expression levels in EC were considered as references (Fig 6g, Supplementary Fig. S6b and Supplementary Tables 9 and 10). While the results should be considered as trends given the limited size of the cohorts, expression of some genes appeared corrected by ART (e.g, CXCL13), whereas others are not restored at all (e.g. IL22).

Figure 6. Distinct transcriptional imprint with reduction in T_FH-associated gene expression but poor correction of T_H1, T_H17 and T_H22 gene levels after suppression of viremia by ART.

(a) Paired volcano plot comparisons depicting DEGs as measured by microarray analysis of sorted HIV-specific CD4⁺ T cells in 8 CPs before and after ART and 12 ECs. Red, pink and blue dots represent DEG with log₂|FC| >1 and padj<0.05, log₂FC>1 or padj<0.05, respectively. Moderated two-sided t test followed by Benjamini-Hochberg method. (b) Euler diagrams of DEG between each pairwise comparison of the three groups of participants (log₂|FC|> 1, p<0.05); numbers represent DEGs. (c) Significant enriched Gene Ontology terms from the curated C5 MSigDB via CAMERA analysis. Blue, Red and white color denote negative, positive or not significant (as defined by FDR<0.05, respectively. Two sided p-value by CAMERA and FDR by Benjamini-Hochberg method. (d-f) Barcode plots of enriched Thelper polarization gene sets by CAMERA in paired comparisons between the three groups (n= 8 CPpre 8 CPpost and 12 EC). Red and blue lines denote positive and negative enrichment, respectively. T_H1 signature: GSE59295 (T_H1 vs T_H2), T_H17 signature: GSE49703 (T_H17 vs T_H1), T_FH signature: GSE50391 (CXCR5high CD45RO vs CXCR5-tonsil samples). Two sided p-value by CAMERA and FDR by Benjamini-Hochberg method. (g) Differential expression of genes associated with CD4 T cell differentiation and function in HIV-specific CD4⁺ T cells of CPs before and after treatment compared to ECs. Red and purple denote logarithmic fold change for comparisons CPpre. vs EC and CPpost vs EC, respectively, by microarray analysis in 8 CP before/after ART and 12 ECs.

These data demonstrate that suppressive ART results in a transcriptional landscape of HIV-specific CD4⁺ T cells distinct from the profiles of both untreated CPs and ECs. We identified gene modules susceptible to manipulation of antigen load, while others were unresponsive to antiviral therapy alone.

Discussion

The molecular features of virus-specific CD4⁺ T cells associated with pathogen containment are poorly understood in chronic human infections. We used genome-wide transcriptional analyses following ex vivo stimulation with HIV antigen to uncover differences in HIV-specific CD4⁺ T cell differentiation and function in chronically infected humans with various viremia levels. An atypical enrichment of a T_FH-like signature in CXCR5_neg HIV-specific CD4⁺ T cells characterized progressive infection. In contrast, elite control of viral replication was associated with robust HIV-specific T_H22 and T_H17 responses, which correlated negatively with immune activation and correlated positively with a greater diversity of bacterial RNA species in plasma and lower Proteobacteria abundance. None of these differences were detected in SEB or CMVpp65-reactive cells, revealing antigen-specific TCR signaling-driven processes. Suppressive ART in chronic progressors elicited a distinct HIV-specific CD4⁺ T cell transcriptome, revealing in comparison to elite controllers, the presence of viremia sensitive (e.g, T_FH signature) and viremia unresponsive (e.g, T_H1 and T_H17 and T_H22 signatures) gene modules. These data suggest that HIV-specific CD4⁺ T cell dysregulation does not represent a mere loss of Thelper functions, as classically defined for immune exhaustion, but also an adaptation to environmental cues combined with persistent changes, likely resulting from durable cell-fate decision programs.

At first sight, the marked T_FH signature¹⁸ identified in CPs contrasts with the murine LCMV model, in which T_FH responses in chronic infection are critical for delayed viral control⁸. The extraordinary ability of HIV to generate escape mutations and elude autologous Ab neutralization contribute to these opposite associations between T_FH differentiation and disease outcome. A study showed that T_FH cells in chronic LCMV infection are capable of supporting generation of neutralizing Ab effective against the contemporaneous virus⁹. In contrast, the immune system lags behind the evolution of autologous HIV strains²⁷. Kinetics of viral evolution in HIV infection may thus overcome the ability of adaptive T_FH skewing to contribute to delayed viral control. While technical limitations have to date hampered investigation of HIV-specific GC T_FH in lymph nodes, robust expansion of total GC T_FH populations was demonstrated in viremic HIV-infected people and SIV-infected macaques, albeit with qualitative defects^{28, 29, 30, 31}.

Genes classically associated with T_FH cells that are enriched in the CXCR5_neg HIV-specific CD4⁺ T cells of CPs include IL21 and CXCL13. HIV p24-specific Abs titers in plasma correlated with frequencies of CXCR5_neg HIV-specific CD4⁺ T cells, suggesting that these responses may provide help to B cells in vivo. Studies in murine models have shown that CXCR5−/− CD4⁺ T cells are able to take over some B helper functions^{32, 33}, resulting in the generation of low-affinity binding antibodies⁹. Acquisition of T_FH-like properties by CXCR5_neg CD4⁺ T cells has also been reported in rheumatoid arthritis³⁴ and cancer³⁵ and combines B help function with migration to inflamed tissues. These analogies suggest shared pathways of CD4⁺ T cell differentiation between diverse diseases associated with chronic inflammation.

Damage to the gut mucosa with associated microbial translocation²² and increased inflammatory properties of the enteric microbiota³⁶ are major drivers of HIV-associated chronic inflammation. We observed that HIV-specific T_H17 and T_H22 responses were abundant in ECs and rare in CPs. IL-17F and IL-17A are closely related cytokines recruiting neutrophils to sites of inflammation. IL-22 promotes intestinal barrier function³⁷, and supports complement-mediated elimination of pathobionts³⁸. Lack of HIV-specific T_H17 and T_H22 responses was associated with CD4⁺ and CD8⁺ T cell activation and a reduced diversity of translocated bacterial RNA in plasma, with overrepresentation of Proteobacteria. These results are consistent with studies in SIV infection showing preferential translocation of Proteobacteria compared to other phylums from the gut lumen into tissues, resulting in enhanced immune activation³⁹. Our data do not establish causal links between features of HIV-specific CD4⁺ T cells and microbial translocation. It is more likely that lack of HIV-specific T_H-17/T_H22 responses and altered translocated microbiome both result from persistent insults to the gastrointestinal tract. The paucity of HIV-specific IL17⁺ and IL22⁺ CD4⁺ T cells observed in CPs may result from high susceptibility of T_H17 and T_H22 subsets to HIV infection^{37, 40} and/or alteration of CD4⁺ T cell differentiation⁴¹. Together with the analysis of bacterial RNA fragments in plasma, these CD4⁺ T cell data suggest that studies of peripheral blood can provide useful insight into the systemic consequences of perturbed gut homeostasis. Probiotic and IL-21 therapy in ART-treated SIV-infected macaques was associated with enhanced polyfunctional T_H17 expansion and reduced markers of microbial translocation and dysbiosis compared to ART alone⁴². It will be important to determine if such interventions restore HIV-specific CD4⁺ T cell profiles.

This work raises questions that will need to be addressed in further studies. AIM assays require activation to enable cell identification. While we believe that analyzing the transcriptional profile modulated by TCR signaling is highly relevant for analysis of T cell responses, it will be important to investigate other populations of HIV-specific CD4⁺ T cells that do not co-upregulate CD69 and CD40L, for example by other AIM markers or by tetramers. Experiments in animal models will be required to define the ontogeny of CXCR5_neg CD4⁺ T cells with T_FH-like features and whether they transitioned through a bona fide T_FH stage during their development. Longitudinal studies from the time of acute infection may identify events in the genesis or persistence of HIV-specific CD4⁺ T cell dysregulation. Investigations of human gut and lymph node biopsies, while challenging, would define the relationships between features of HIV-specific CD4⁺ T cells in blood and other anatomical compartments.

In summary, this work identifies extensive alterations in HIV-specific CD4⁺ T cell transcriptional profiles and demonstrates that skewed lineage differentiation is a key feature of altered CD4⁺ T cell help in the setting of high viral load, in which atypical gain of function is combined with classical exhaustion features. The data also reveal distinct features of HIV-specific CD4⁺ T cells after successful ART. The identified molecular features may therefore help direct selective interventions to improve immune function. Given the paucity of transcriptome datasets available on pathogen-specific CD4⁺ T cells in humans, our results will also provide useful benchmark comparisons for other infectious diseases and vaccine-induced responses.

ONLINE METHODS

Participants and Samples

Leukaphereses were obtained from study participants at the Montreal General Hospital, Montreal, Canada; at the Centre Hospitalier de l’Université de Montréal (CHUM) in Montreal, Canada; and at the Massachusetts General Hospital (MGH) in Boston, USA. The study was approved by the respective IRBs, written informed consent obtained from all participants prior to enrolment and PBMC samples obtained by leukapheresis. Samples were collected between years 2013 and 2018. Subject characteristics are summarized in Supplementary Table 1. “Chronic progressors” (CP) were defined as people with a viral load of more than 5,000 vRNA copies per ml of plasma; “viremic controllers” (VC) were defined as people with more than 50 and less than 5,000 vRNA per ml plasma; and “elite controllers” (EC) were defined as subjects who spontaneously controlled viremia to below 50 RNA copies per ml plasma in the absence of therapy. CPs, VCs, and ECs.were either treatment naive or untreated for at least 3 months. Eight CPs were followed longitudinally and a second sample was obtained at least 6 months after initiation of ART therapy and suppression of viral load below 50 copies mL (Supplementary Table 8). PBMC were isolated by the Ficoll density gradient method and stored in liquid nitrogen.

Antibodies

For details of antibodies used, see Supplementary Tables 11–20 and descriptions of use below. In all cases, antibodies are monoclonal and raised in mice unless otherwise stated. All antibodies are commercially available, were validated by the suppliers and titrated using biological and/or isotype controls.

CD40L/CD69 AIM assay for live-cell sorting of HIV Gag-specific CD4⁺ T cells

Cryopreserved PBMCs were thawed and rested at 37°C for 3 hours. Cells were aliquoted into wells of a 24-well plate, at a total of 15 × 10⁶ cells per well in RPMI medium supplemented with 10% Human AB (HAB) serum and PenStrep. For each assay, we used a no exogenous stimulation (No Ag) condition as control and an HIV-specific stimulation with an overlapping peptide pool corresponding to HIV Gag (JPT, PM-HIV-Gag ULTRA), or HCMVA pp65 (JPT, PM-PP65) at a final concentration of 0.5μg/ml/peptide. A CD40 blocking antibody (Miltenyi Biotec, clone HB14) was added to each well 15 minutes prior to stimulation at a final concentration of 0.5 μg/ml. Following a 9 hr stimulation, the cells were stained for 30 min at 4°C before live cell sorting on a FACS Aria cell sorter (BD BioSciences) equipped for handling of biohazardous material. Two panels of antibodies were used to isolate live cells subsets according to combinations of surface markers. In the first panel for isolation of total CD69⁺CD40L⁺ Gag-specific CD4⁺ T cells, PBMCs were stained as indicated in Supplementary Table 11. The gating strategy for cell sorting is illustrated in Supplemental Figure 1. In the second panel used for isolation of CXCR5mem and CXCR5-subsets of CD69⁺CD40L⁺ Gag-specific CD4⁺ T cells, PBMCs were stained as indicated in Supplementary Table 12. The gating strategy for cell sorting is illustrated in Supplementary Fig. 3. The FACS Aria was operated at 70 pounds per square inch with a 70-μm nozzle. For all populations, 5,000 cells were collected directly into RLT lysis buffer (Qiagen).

Phenotyping of Gag-specific CD4⁺ T cells identified by the CD69/CD40L AIM assay.

PBMCs were thawed and plated in 24-well plates at 10 × 10⁶/ml in RPMI ⁺ 10% HAB Serum (HAB) and rested for 3 hours at 37°C. Prior to stimulation, CD40 blocking antibody was added directly into culture at a final concentration of 0.5μg/ml, as well as antibodies against CXCR5, CCR6 and CXCR3. Cells were incubated for 15 min, then left unstimulated or stimulated either with an overlapping peptide pool corresponding to HIV Gag (JPT, PM-HIV-Gag ULTRA), at a final concentration of 0.5μg/ml/peptide, or with Staphylococcal Enterotoxin B (SEB) as a positive control (1 μg/mL). Cells were stimulated for 9h, collected and stained for 30 min at 4°C with the antibodies in Supplementary Table 13. Cells were washed, fixed for 20 min in 2% paraformaldehyde, washed, and resuspended in 1% FCS/PBS for flow acquisition on a 5-laser LSRII (BD BioSciences).

Phenotyping of immune activation markers

PBMCs were thawed and rested overnight at 37°C. Next day, cells were washed and stained with viability dye (20 min at 4 C°) and surface markers (20 min at 4 C°) (Supplementary Table 14). Fluorescence minus one staining was used to define the cutoff for positivity. Cells were washed, fixed for 1h in 2% paraformaldehyde, washed, and resuspended in 1% FCS/PBS for flow acquisition on a 5-laser LSRII (BD BioSciences).

Standard intracellular cytokine staining (ICS) for cytokine measurements

PBMCs were incubated for 6 hours with an HIV Gag peptide pool at a concentration of 0.5μg/ml/peptide in the presence of 2.5 μg/ml brefeldin A (BD GolgiPlug) and (0.3μl/ml) monensin (BD GolgiStop). For delayed ICS assays, 2.5 μg/ml brefeldin A (BD GolgiPlug) and (0.3μl/ml) monensin (BD GolgiStop) were added 9h after stimulation with the Gag peptide pool and cells were cultured for 12h. Unstimulated cells were used as negative control and SEB (0.5) μg/ml as positive control. Cells were then stained with viability dye (20 min at 4 C°), surface markers (20 min at 4 C°), fixed with Fixation Solution (eBioscience), and stained for intracellular proteins in permeabilization 1X buffer (eBioscience) (30 min at 4 C°) as described in Supplementary Tables 15 and 16. Cells were acquired on an LSR II (BD Biosciences).

Gene Expression Analysis

Whole Transcriptome Analysis by microarrays

Total RNA was purified using the RNeasy Plus Micro Kit (Qiagen). RNA integrity was assessed using a 2100 Bioanalyzer (Agilent Technologies). Sense-strand cDNA was synthesized from total RNA using a fixed volume of 3μL. Fragmentation and labeling were performed to produce ss-DNA with the Applied Biosystems™ GeneChip® WT Pico Terminal Labeling Kit according to manufacturer’s instructions (Applied Biosystems™). After fragmentation and labeling, 5 μg DNA target was hybridized on GeneChip® Clariom™ D, human (ThermoFisher Scientific) and incubated at 450C in the Genechip® Hybridization oven 640 (Affymetrix) for 17 hours at 60 rpm. GeneChips were then washed in a GeneChips® Fluidics Station 450 (Affymetrix) using Applied Biosystems™ Hybridization Wash and Stain kit according to the manufacturer’s instructions (Applied Biosystems™). The microarrays were finally scanned on a GeneChip® scanner 3000 (Affymetrix).

Validation of transcriptional expression by Fluidigm

Total RNA was purified using the RNeasy Plus Micro Kit (Qiagen). cDNA was synthesized using all RNA available (or 1–5 ng) with the High-Capacity Reverse Transcription Kit with RNase Inhibitor (Life Technologies) (25° C for 10 min, 37° C for 120 min, 85° C for 5 min). cDNA equivalent to 1000 sorted cells was subjected to gene-specific preamplification using Taqman Preamp MasterMix (Applied Biosystems) and 96 pooled TaqMan Assays (Applied Biosystems) (Supplementary Tables 21–23) at final concentration 0.2X (95°C for 10 min, followed by 16 cycles of 95°C for 15 s and 60°C for 4 min). The preamplified cDNA was diluted 5-fold in DNA suspension buffer (Teknova) and was mixed with TaqMan Universal PCR Master mix (Life Technologies) and 20X GE sample loading reagent (Fluidigm). 20X Taqman assays were diluted 1:1 with 2X assay loading buffer (Fluidigm). Taqman assays mixtures were loaded onto a primed 96.96 Dynamic Array chip (Fluidigm). The chip was loaded into the IFC Controller, where each sample was mixed with each assay in every possible combination. The chip was transferred in a Biomark (Fluidigm) for real-time PCR amplification and fluorescence acquisition using single probe (FAM-MGB, reference: ROX) settings and the default hot-start protocol with 40 cycles. Cycle thresholds (Ct) were calculated using the Fluidigm BioMark software v1.4.2.

Combined cytokine/chemokine mRNA Flow FISH and protein staining assays

PBMCs were either left unstimulated or were stimulated with an HIV Gag peptide pool (JPT) or SEB for 15 hours after 15h min incubation with a CD40 blocking antibody (Miltenyi Biotec, clone HB14) at a final concentration of 0.5 μg/ml. For experiments for evaluation of mRNA expression without CD40 blocking antibody PBMCs were stimulated for 12h. After stimulation, cells were stained with a viability dye (20 min, 4°C, Fixable LiveDead, eBioscience) and then surface markers were labelled (30 min 4°C) with antibodies (See panels in Supplementary Tables 17–20). Samples were next subjected to the PrimeFlow RNA assay (Affymetrix/eBioscience) for specific mRNA detection as per manufacturer’s instructions. All buffers and fixation reagents were provided with the kit, with the exception of flow cytometry staining buffer (2% FBS/PBS). After fixation and permeabilization, cytokine/chemokine mRNAs were labeled with sets of probe pairs, as listed in Supplementary Tables 17–20. The probes were diluted 1:5 in diluent and hybridized to the target mRNA for 2 hr at 40 °C. Samples were washed to remove excess probes and stored overnight in the presence of RNAsin. Signal amplification was achieved as previously described²⁰ by sequential 1.5 hr, 40 °C incubations with the pre-amplification and amplification mix. Amplified mRNA was labeled with fluorescently-tagged probes for 1 hr at 40 °C. A negative control probe and a positive control probe (against house-keeping gene RPL13A) were included in each experiment. Gates were set on the unstimulated control where appropriate or scrambled probes. Samples were acquired on an LSRII (BD Bioscience) or a BD LSRFortessa™ (experiments without CD40 blocking antibody). Analysis was performed using FlowJo (Treestar, V10) and Spice 5.35 software.

Measurements of cytokine secretion by beads arrays and ELISA

One million CD8 T cell-depleted PBMCs (Dynabeads CD8, StemCell) were incubated with an HIV Gag peptide pool (1μg/ml/peptide) or left unstimulated and incubated at 37°C in RPMI supplemented with 10% HAB and PenStrep. Cytokine secretion in supernatants was measured 48 hours after stimulation on a Luminex beads array platform, as previously described^{19, 20}. Measurements were performed in duplicates using the Human High Sensitivity Cytokine Premixed Kit B (R&D Systems) for IL-2, IFN-γ, TNF-α, IL-17A, IL-22 and IL-31 or for IL-17F on a Bio-plex 200 array system (Bio-Rad Laboratories) per manufacturer’ s instructions. CXCL13 was measured separately with the Human CXCL13 Quantikine ELISA Kit (R&D Systems), as per manufacturer’s instructions.

Measurement of CD4i p24-specific antibodies by ELISA

The anti-p24 ELISA was performed as follows: 96 well plates (Thermo Scientific Nunc, FluoroNunc/LumiNunc, MaxiSorp Surface) were coated with 0.1μg/ml of recombinant p24 (NIH #12028) or BSA (Bioshop #ALB001.100) in PBS over night at 4°C. Plates were blocked for 90 min at room temperature (RT) with blocking buffer (TBS, Tween 0.1%, BSA 2%) and then washed 4 times with washing buffer (TBS, Tween 0.1%). Dilutions of human sera (1/1000, 1/3000, 1/6000) or rabbit anti-HIV p24 antiserum (NIH #4250) in washing buffer containing 0,1% of BSA were incubated for 2 h at RT. Plates were washed 4 times with washing buffer before incubation for 90 min at RT with HRP-conjugated secondary Abs: goat anti-human IgG HRP (Invitrogen #31413) or anti IgG rabbit HRP (Invitrogen #31462). Plates were then washed 4 times with washing buffer before revealing with standard ECL (Perkin Elmer #NEL105001EA) with a TriStar luminometer (LB 941, Berthold Technologies).

Plasma microbiome sequencing

Circulating total RNA and DNA were isolated from frozen plasma using RNAzolBD (MRCgene) according to the manufacturer’s recommendations. After RNA fragmentation and reverse transcription using random hexamers, obtained transcripts as well as the isolated plasma DNA were individually incorporated into barcoded cDNA libraries based on the Illumina TruSeq platform. Libraries were validated by microelectrophoresis, quantified, pooled and clustered on Illumina patterned flow cells. Clustered flow cells were sequenced on an Illumina HiSeq 4000 in 75-base paired reads.

Gene expression data analysis

Microarrays:

The preprocessing of the microarray data was done in R⁴³ v3.3.3 using the oligo package⁴⁴ v1.38.0. Specifically, the CEL files were loaded in R and the rma function was applied for background subtraction, normalization and summarization. Quality control was performed by inspecting various diagnostic plots of the intensity distribution and correlation structure of control and regular probes. After review of the QC plots, 3 outliers were removed. Combat (sva package 3.30.1) has been applied for batch normalization of samples using 4 replicates as reference for the normalization. The normalized data were used for subsequent subanalyses CPpre vs EC, CPpre vs CPpost, CPpost vs EC.

Differential expression:

Differential analysis of gene expression was done by the R package limma45. Using the lmFit function, a linear model was fit to each gene separately. Moderated t-tests were performed between each group. P-values were adjusted using Benjamini-Hochberg method for multiple test correction.

CAMERA

T_H1 signature (GSE59295; T_H1 vs T_H2) and T_H17 signature (GSE49703; T_H17 vs T_H1) were available in C7 collection of MSigDB. The T_FH signature (GSE50391; CXCR5high CD45RO vs CXCR5-tonsil samples) and LCMV exhaustion signature (GSE41866; LCMV Clone 13 D30 vs Armstrong D30-Exhausted vs Memory) were created using the GEO2R online tool (https://www.ncbi.nlm.nih.gov/geo/geo2r/). For each signature, a gene set test was performed using the camera function from the limma package, accounting for inter-gene correlation.

Venn diagrams were created using the online tool Eulerr.co, (R package version 4.1.0, https://cran.r-project.org/package=eulerr)46. Unsupervised hierarchical clustering was performed using the in heatmap2 function in R⁴³ on relative signal intensity values applying the Euclidean distance metric and complete linkage clustering method. Barcode plots were generated with the barcodeplot function from the limma package.

qRT-PCR results

Analysis of the qRT-PCR data obtained on the microfluidic platform was carried out using GenEx software (MultiD Analyses, URL: http://www.multid.se). Five endogenous control genes were included in the Fluidigm run and the stability of endogenous control genes across all experimental samples was evaluated applying the NormFinder algorithm⁴⁷ in GenEx. The mean expression of the most stable endogenous control genes was used for normalization and calculation of -ΔCt values. Principal component analysis, biplots and correlograms were created using the prcomp, fviz_pca_biplot and corrplot functions, respectively, in R programming language.

Statistical analyses of flow cytometry, beads array and qRT-PCR results

Statistical analyses were performed using Prism V6.0 (GraphPad) using non-parametric tests. Two-group comparisons were performed using the Mann-Whitney test and pairwise comparisons were performed using the Wilcoxon matched-pair test. For comparisons between three groups, Kruskal-Wallis or Friedman one-way ANOVA with Dunn’s post-test was used. Permutation test (10000 permutations) was applied for pie-chart comparison using the SPICE software. For correlations, Spearman’s R correlation coefficient was applied. Statistical tests were two-sided and p<0.05 were considered significant.

Bioinformatic analysis of deep sequencing results of cell-free RNA fragments in plasma

Raw, paired-end Illumina reads were first trimmed for low quality bases and adaptor contamination with Trimmomatic⁴⁸. As a first pass to remove host-derived sequences, trimmed reads were first aligned against the HG38 genome with Bowtie 2⁴⁹ and then aligned to HG38 transcriptome with the splice aware aligner, STAR⁵⁰. Unaligned paired-end reads were then assembled into contigs using Trinity⁵¹. Trinity contigs were then aligned against HG38 again to further remove host-derived sequences with BLAST⁺⁵². Remaining contigs were then aligned against the NCBI nt database⁵³ and quantified with Salmon⁵⁴. Taxonomic information was assigned for each contig and TPM values were collapsed at the species level for each sample for downstream analysis. TPM values for each sample were collapsed on the Genus taxonomic level and Morisita-Horn Dissimilarity indexes were calculated on a pairwise basis using the R package, vegan⁵⁵. The pairwise dissimilarity matrix was then clustered and the heatmap was produced with gplots⁵⁶.