Study participants (n = 81) were recruited from the Ubuntu Clinic, Khayelitsha (Cape Town, South Africa). Participants were divided into four groups based on their HIV-1 and TB status: HIV−/LTBI (n = 20), HIV+/LTBI (n = 21), HIV−/aTB (n = 20), and HIV+/aTB (n = 20). LTBI was diagnosed based on a positive IFNγ release assay (QuantiFERON^®-TB Gold In-Tube), no symptoms of aTB and a negative Mtb-sputum (GeneXpert). Diagnosis of aTB was based on clinical symptoms and positive Mtb-sputum. All aTB cases were fully drug sensitive and TB treatment-naïve. All HIV-infected individuals were antiretroviral treatment (ART)-naive. The study was approved by the University of Cape Town Human Research Ethics Committee (No. 896/2014).

PBMC were isolated by Ficoll-Hypaque density gradient centrifugation (GE Health care), cryopreserved and stored until needed. Cryopreserved PBMC were thawed and rested in RPMI 1640 containing 10% heat-inactivated FCS prior to antigen stimulation. PBMC were stimulated using Mtb cell lysate (strain H37Rv, BEI resources) or early secretory antigenic target (ESAT)-6/culture filtrate protein (CFP)-10 peptide pool consisting of 17 and 16 peptides covering the entire 6-kDa ESAT-6 and 10-kDa CFP-10, respectively (2 µg/ml, Peptide Synthetics) or a pool of cytomegalovirus virus (CMV) peptides consisting of 138 peptides covering the entire pp65 protein (2 µg/ml). Brefeldin A (10 µg/ml, Sigma) was added 3 h after Mtb lysate stimulation or at the time of peptides addition. All stimulations were performed in the presence of co-stimulatory antibodies, anti-CD28 and anti-CD49d (both at 1 µg/ml; BD) for 16 h.

After stimulation, cells were washed, stained with LIVE/DEAD^® Fixable Near-IR Stain (Invitrogen) and, subsequently, surface stained with the following antibodies: CD14-APC/Alexa Fluor 750 (Invitrogen) and CD19-APC/Alexa Fluor 750 (Invitrogen), CD4-FITC (BD), CD27-BV711 (BD), CCR4-BV510 (Biolegend), CCR6-BV605 (Biolegend), CXCR3-PE-Cy7 (BD Biosciences), KLRG1-PerCP/eFluor 710 (eBioscience), and HLA-DR-PE (BD). Cells were then fixed and permeabilized using Cytofix/Cytoperm buffer (BD) and stained with CD3-BV650 (BD), IL-2-PE/Dazzle™ (Biolegend), TNFα-eFluor 450 (eBioscience), and IFNγ-Alexa Fluor 700 (BD). Finally, cells were washed and fixed in 1% formaldehyde in PBS. Samples were acquired on a LSR-II (BD) and analyses were performed using FlowJo (Treestar). A positive IFNγ response was defined as at least twice the background measured in the presence of co-stimulatory antibodies without antigen. Cell polyfunctionality was analyzed using Pestle and Spice software (19). The gating strategy is presented in Figure S1 in Supplementary Material.

Statistical analyses were performed using GraphPad Prism. The Mann–Whitney U test and Wilcoxon-matched pairs test were used for unmatched and paired samples, respectively, and the Kruskal–Wallis ANOVA using Dunn’s test for multiple comparisons. Correlations were performed using the Spearman rank test. Logistic regression followed by receiver operating characteristic (ROC) curve analysis, which plots sensitivity (true-positive rates) vs 1-specificity (false-positive rates), was used to evaluate the performances of each marker to identify aTB and to establish probability cutoff values.

To better understand the profile of Mtb-specific T cell responses, we first defined the frequency of Mtb-specific CD4+ T cells in 81 participants grouped according to HIV-1 and TB status (Table 1). We measured IFNγ production in response to ESAT-6/CFP-10 peptide pool and Mtb cell lysate. In parallel, we also monitored T cell responses to CMV, used as an unrelated pathogen for comparison (Figure 1A).

For all tested pathogens, the frequency of IFNγ responding CD4+ T cells was comparable in HIV-uninfected individuals, irrespective of their TB status (Figure 1B).

Conversely, in HIV-infected individuals, Mtb lysate and ESAT-6/CFP-10 responses were significantly elevated in participants with aTB disease when compared to LTBI (median: 0.64 vs 0.24% for Mtb lysate and 0.13 vs 0.03% for ESAT-6/CFP-10, respectively). Such differences were not observed for CMV-specific responses (Figure 1B). This surprising elevation of the frequency of Mtb-specific CD4+ T cells in HIV+/aTB individuals, despite the severe lymphopenia characterizing this group (median CD4 count: 149 cells/mm³), suggests that active Mtb replication promotes Mtb-specific CD4+ T cell expansion in ART naïve HIV+/aTB patients. Moreover, the elevated frequency of ESAT-6/CFP-10-specific CD4+ T cells observed in the HIV+/LTBI group could appears to differ from a previous study showing a rapid depletion of Mtb-specific CD4+ T cells in subjects with LTBI within 1 year after HIV infection (20). However, in our study cohort, recruited from a highly TB endemic area, recurrent Mtb exposure, and the relatively well-preserved CD4+ T cell count in HIV-infected individuals with LTBI (median: 563 cells/mm³) could account for the conservation of Mtb-specific CD4+ T cells.

In addition, we also detected IFNγ production in response to ESAT-6/CFP-10 stimulation originating in the CD3+ CD4− compartment. Such responses were predominantly observed in aTB patients [75% (15/20) for the HIV− group and 60% (12/20) for the HIV+ group]. On the contrary, only one-fifth of latently infected persons exhibited detectable ESAT/CFP-specific IFNγ+ CD4− CD3+ T cell responses [20% (4/20) for HIV− and 19% (4/21) for HIV+], as previously reported (21) (data not shown).

We next compared the phenotype of Mtb- and CMV-responding IFNγ+ CD4+ T cells during latent and aTB disease. We focused on the memory differentiation profile (CD27), the activation status (HLA-DR and KLRG1) and the homing potential (CCR4, CCR6, CD161, and CXCR3) of these cells, as these parameters are likely to be important features regulating T cell function (Figure 2A).

In the context of aTB disease (irrespective of HIV infection), ESAT-6/CFP-10-specific IFNγ+ CD4+ T cells were characterized by significantly lower expression of CD27 (median: 26% for HIV− and 20% for HIV+) when compared to persons with LTBI (70% for HIV− and 72% for HIV+). Concomitantly, HLA-DR expression on these cells was significantly elevated in persons with aTB (median: 57% for HIV− and 79% for HIV+) compared to LTBI (5% for HIV− and 13% for HIV+) (Figure 2B). No differences in the expression of KLRG1 or any of the chemokine receptors tested were observed among Mtb-specific CD4+ T cells in the four clinical groups (Figure 2B; Figure S3 in Supplementary Material). HIV infection per se did not significantly alter the phenotype of ESAT-6/CFP-10-specific IFNγ+ CD4+ T cells. However, in LTBI, there was a trend toward increased HLA-DR expression on these cells in the HIV-infected persons compared HIV-uninfected participants. Notably, in the LTBI/HIV+ group, HLA-DR expression on Mtb-specific IFNγ+ CD4+ T cells mirrors HLA-DR expression in the whole CD4 compartment (p = 0.03, r = 0.49, data not shown), suggesting that during LTBI, elevated activation of Mtb-specific CD4+ T cells in probably driven by HIV-induced systemic immune activation. Similar results were obtained for Mtb lysate-specific IFNγ+ CD4+ T cells (Figure S2A in Supplementary Material).

Changes observed during aTB were restricted to CD4+ T cells targeting Mtb antigens, as the phenotypic profile of CMV-specific IFNγ+ CD4+ T cells was comparable between the LTBI and aTB groups (Figure 2C; Figure S3 in Supplementary Material). However, unlike the Mtb-specific response, HIV altered CMV-specific CD4+ T cell profile differently, inducing memory cell maturation. Indeed, CD27 expression was reduced on CMV-specific IFNγ+ CD4+ T cells in HIV-infected persons (median: 11% in LTBI and 6% in aTB) when compared to HIV-uninfected (49% in LTBI and 26% in aTB) (Figure 2C).

Overall, these results show that aTB disease associates with the activation and memory differentiation of Mtb-specific CD4+ T cells; these cellular features would be consistent with ongoing bacterial replication.

To define whether the phenotypic changes observed in the context of aTB disease (or HIV infection) affect cell functions, we next compared the polyfunctional potential of CD4+ T cells in response to ESAT-6/CFP-10 and CMV antigens in the four clinical groups. We measured the expression of IL-2, IFNγ, and TNFα (Figure 3A).

While the majority of ESAT-6/CFP-10-responding cells were polyfunctional (co-expressing IL-2, IFNγ, and TNFα) in all study groups, Mtb-specific CD4+ T cells in persons with aTB disease (irrespective of HIV status) were characterized by a significant elevated proportion of IL-2−IFNγ+TNFα+ cells, with a concomitant decrease in the proportion of IL-2+IFNγ−TNFα+ cells (Figure 3B). Similar functional profiles were obtained in response to Mtb lysate (Figure S2B in Supplementary Material). For CMV-specific CD4+ T cells, no functional differences were observed between the HIV−/LTBI and HIV−/aTB groups. HIV infection, however, significantly skewed CMV-specific CD4+ response profiles, inducing the expansion of IL-2−IFNγ+TNFα+ cells with a concomitant contraction of IL-2+IFNγ+TNFα+ and/or IL-2+IFNγ+TNFα− cells (Figure 3C).

We then assessed whether these functional differences were related to specific cell phenotypic characteristics. Figures 3D,E show that in HIV-uninfected individuals, CD27 expression on antigen-specific IFNγ+ CD4+ T cells correlated negatively with the proportion of IL-2−IFNγ+TNFα+ cells for both Mtb and CMV responses (p < 0.0001). Similar associations were observed for HIV-infected persons (data not shown) and for Mtb lysate-responding CD4+ T cells (Figure S2C in Supplementary Material). This indicates that the differentiation of antigen-specific CD4+ T cells into effector memory or effector cells (i.e., cells lacking CD27 expression), as a result of Mtb replication for Mtb-specific cells or HIV infection for CMV-specific CD4+ T cells, partially attenuated their capacity to secrete IL-2.

Finally, further analysis of the functional profile of Mtb-specific cells revealed that the median fluorescence intensity (MFI) of TNFα in ESAT-6/CFP-10-specific IFNγ+ CD4+ T cells was significantly higher in aTB (regardless of HIV infection), when compared to LTBI (median MFI: 24,920 vs 14,180 in HIV− and 27,720 vs 14,980 in HIV+, respectively) (Figures 4A,B). A similar pattern of TNFα expression level was also observed in response to Mtb lysate between the LTBI and aTB groups (Figure S2D in Supplementary Material), but no such differences were detected for CMV-specific cells (Figure 4B). Interestingly, MFI levels of TNFα in Mtb- and CMV-specific IFNγ+ CD4+ T cells in HIV-uninfected individuals were positively associated with the expression of KLRG1 on these cells (Figure 4C). Of note, similar results were obtained when these analyses were restricted to polyfunctional cells (i.e., cells co-expressing IL-2, IFNγ, and TNFα) (Figure S4 in Supplementary Material). This shows that the elevated TNFα MFI within Mtb-specific CD4+ T cells in individuals with aTB was not simply related to an overall increase in the proportion of cell producing TNFα, but rather reflects an intrinsic increase in the pro-inflammatory potential of these cells.

These data suggest that specific functional changes occur in Mtb-specific cells in aTB compared to LTBI, and these can be linked to phenotypic changes. Moreover, during aTB disease, Mtb-specific responses become highly pro-inflammatory and such characteristics appear to be inherently linked to KLRG1 expression on these cells.

We next assessed which parameter best distinguished latent from aTB by generating ROC curves from the data obtained in ESAT-6/CFP-10 responders. HLA-DR expression on ESAT-6/CFP-10-specific IFNγ+ CD4+ T cells showed the best capability to discriminate latent from aTB, irrespective of HIV status [p < 0.0001, area-under-the-curve (AUC) = 0.92, 95% CI: 0.82–1.01, specificity: 82%, sensitivity: 84% for HIV− and p < 0.0001, AUC = 0.99, 95% CI: 0.98–1.01, specificity: 94%, sensitivity: 93% for HIV+]. However, the optimum cutoff values discriminating LTBI from aTB were distinct for HIV-uninfected and HIV-infected individuals (12 vs 54%, respectively) (Figure 5A). Our data were comparable to Adekambi et al.’s (17), despite the disparity in the cutoff value for these markers, which could be explained by flow cytometry technical differences. CD27 expression on Mtb-specific IFNγ+ cells also distinguished between LTBI and aTB irrespective of HIV infection (p = 0.0003, AUC = 0.85, 95% CI: 0.73–0.98, specificity: 77%, sensitivity: 79% for HIV− and p < 0.0001, AUC = 0.93, 95% CI: 0.84–1.02, specificity: 89%, sensitivity: 87% for HIV+) with comparable optimum cutoff values for HIV-uninfected and HIV-infected individuals (50 vs 43%, respectively, Figure 5B). Portevin et al. reported that CD27 expression defined as a ratio (CD27 MFI on total CD4+ cells/CD27 MFI on ESAT-6/CFP-10-specific IFNγ+ CD4+ cells) could be used to diagnose TB in children (13). We, thus, compared the ability of CD27 expressed as a percentage or as a ratio to distinguish latent from aTB in our study cohort. Figure S5 in Supplementary Material shows that both approaches had comparable sensitivity and specificity. Moreover, analysis combining HLA-DR and CD27 expression on Mtb-specific IFNγ+ CD4+ T cells did not improve the sensitivity or the specificity compared to HLA-DR alone (Figure S6 in Supplementary Material). Finally, we assessed the use of TNFα expression levels to delineate aTB from LTBI (Figure 5C). While the TNFα MFI on ESAT-6/CFP-10-specific IFNγ+ CD4+ T cells could distinguish aTB cases in both HIV-uninfected and HIV-infected individuals (p = 0.002, AUC = 0.79, and p = 0.007, AUC = 0.77, respectively); this measurement was less robust than HLA-DR or CD27. All the markers described above exhibited comparable ability to differentiate LTBI from aTB when measured on Mtb lysate-specific IFNγ+ CD4+ T cells (Figure S7 in Supplementary Material).