Depletion of Regulatory T Cells in HIV Infection Is Associated with Immune Activation

Study population

A total of 81 HIV-1-infected Ugandan adults was enrolled at the Joint Clinical Research Centre (Table I⇓). Exclusion criteria included age <18, pregnancy, active tuberculosis (TB; defined as suspected TB or in the first 2 mo of anti-TB therapy), or moribund status. In addition, 25 healthy HIV-negative patients were enrolled as controls. All volunteers provided written informed consent. Institutional Review Board approval was obtained from the California Department of Health Services, University of California, and the Joint Clinical Research Centre.

Immunologic subsets and activation markers

Isolation of PBMC was performed by Ficoll-Hypaque (Amersham) density centrifugation. Blood was processed within 3 h of blood draw. Freshly isolated PBMC were immunophenotyped for Treg quantitation as well as CD4⁺ and CD8⁺ T cell activation. Analysis was performed by four-color flow cytometer (FACSCalibur; BD Biosciences). Treg quantitation was performed by staining with anti-CD62L FITC, anti-CD25 PE, anti-CD3 PerCP Cy5.5, and anti-CD4 allophycocyanin. Immune activation was determined by staining with anti-HLA-DR PE, anti-CD38 allophycocyanin, and either anti-CD4 or anti-CD8 PerCP Cy5.5 (all Abs and flow reagents were obtained from BD Biosciences). For validation of Treg immunophenotyping, anti-CD45 RO, anti-HLA-DR, anti-CD38, and anti-CTLA-4 were used. CTLA-4 expression was assessed by intracellular staining using cytoperm 2 reagents. Gating analysis was performed with FlowJo software (TreeStar) with the same gating applied to all samples. Absolute numbers of CD4⁺ or CD8⁺ T cells were determined by BD TruCount analysis using blood collected in EDTA. Only samples that had greater than 5000 CD4⁺ events for Treg quantitation and 1000 CD4⁺ events for CD4⁺ T cell activation were included to limit variability in flow cytometric analysis.

Validation of Treg suppressor function

Tregs and responder cells were isolated from PBMCs using a high speed MoFLo cell sorter (DakoCytomation). PBMCs were first stained with anti-CD4 APC, anti-CD25 PE, and anti-CD62L FITC. Tregs were defined as CD4⁺CD62L^highCD25^bright, and T responder cells CD4⁺CD62L^highCD25^negative. A typical sort of 10⁸ PBMCs yielded ∼500,000 Tregs. Sort purity was greater than 90%. For suppressor assays, 3000 responder cells were added to 100,000 irradiated autologous PBMCs (APCs) in a 96-well plate. Tregs were added at varying ratios (Treg:T responder) ranging from 1:1 to 1:64. Cells were stimulated with soluble anti-CD3 (OKT3; Orthobiotech) at 0.5 μg/ml and pulsed with tritium-labeled thymidine for 12 h before harvest on day 7. Background was assessed in unstimulated wells at each dilution and subtracted from stimulated samples. Background proliferation was always <100 cpm. Suppression assays were repeated in three independent experiments with two different donors, all yielding comparable results.

Treg depletions

CD25 depletions were performed using anti-CD25 beads (Dynal Biotech). Briefly, 10⁷ PBMCs were incubated with 80 μl of anti-CD25 beads for 45 min at 4°C. Beads were washed three times, and washes were pooled and centrifuged to obtain a CD25-depleted population. All depletions were verified by flow cytometry. Typically, this protocol depletes greater than 90% of Tregs, but has no discernible impact on CD8⁺ T cells. Tregs were eluted from beads and used for further immunophenotyping of surface markers.

Assessment of clade A and D Gag-specific CD4⁺ and CD8⁺ T cell responses by IFN-γ intracellular production and ELISPOT assays

Treg-depleted PBMCs were assayed in parallel with undepleted specimens for intracellular HIV-specific IFN-γ production, as previously described (29). Briefly, freshly isolated PBMCs (depleted and undepleted) were rested overnight in complete RPMI with 10% FCS and then stimulated for 6 h in the presence of brefeldin, anti-CD28, and anti-CD49d Abs. Stimulations were performed with pools of synthetic 15-mer peptides (11-aa overlap) spanning the entire consensus Gag protein for both clade A and D viruses (circulating strains in Uganda). Intracellular IFN-γ was measured using anti-IFN-γ FITC, anti-CD3 PerCP, and anti-CD8 allophycocyanin. Staphylococcus enterotoxin B (1 μg/ml; Sigma-Aldrich) was used as a positive control, and medium alone was used to establish background for every sample. Analysis was performed using FLOWJO software, and results were expressed as a percentage of responding CD8⁺ T cells after subtracting background. Responses were considered positive if more than twice background and at least 0.1% of CD8⁺ cells. Negative responses were assigned a value of zero.

For ELISPOT assay, 96-well nitrocellulose plates (Millititer; Millipore) were coated with mAb 1-D1K (Mabtech) overnight at 4°C using previously described methods (30, 31, 32). PBMC were added at 10⁵ cells/well and stimulated with pools of overlapping Gag A and D peptides (40 peptides per pool at final 2 μg/ml per peptide). Negative and positive controls used were unstimulated (no peptide) PBMC as well as PHA-stimulated (10 μg/ml; Sigma-Aldrich) PBMC, respectively. Plates were incubated overnight at 37°C in 5% CO₂, and washed with PBS before adding biotinylated anti-IFN-γ mAb (7-B6-1 biotin; Mabtech). Streptavidin-conjugated HRP (BD Biosciences) was subsequently added (1 μg/ml), followed by 3-amino-9-ethylcarbazole peroxidase substrate coloring solution (100 μl/well; Vector Laboratories). All plates were analyzed using an ELISPOT C.T.L. reader (C.T.L.). Results are expressed as the number of spot-forming cells (SFC) per 10⁶ PBMC after subtraction of the background. Positive responses were defined as >100 SFC per million PBMC and more than twice the background values. Background values were <10 SFC per million PBMC.

Plasma HIV-1 RNA viral load

HIV-1 RNA level was determined from plasma using Roche Amplicor 1.5 assay (Roche). All undetectable values (below 400 copies) were assigned a value of 399. In addition, all samples with values greater than 750,000 were serially diluted and quantitated, as per the manufacturer’s recommendation.

Statistical analysis

Linear least-square regression model was used for statistical analysis. With nonlinear distribution, values were log transformed to permit the use of the linear model. All models were verified by viewing residual plots to ensure that the basic assumptions of linear regression were not violated and to assure the logarithmic transformation was appropriate. Statistical significance was defined as p < 0.05. Standardized coefficients were used to provide meaningful comparisons among the independent variables in a multivariate regression analysis. Standardized coefficients are measured by SD, and thus, the magnitudes of these coefficients are independent measures and provide a meaningful comparison among the regression coefficients. This approach allows the evaluation of the contribution of independent variables with differing units in a multivariate analysis. All dependent variables in our multivariate analysis were analyzed in separate models to evaluate correlation with independent variables in Table II⇓. None of the interactions between different Treg representations was statistically significant.

Demographic and clinical characteristics of patients

Patients attending the outpatient HIV clinic at the Joint Clinical Research Centre were recruited to a cross-sectional study to assess T cell function, activation, and Tregs. Clinical characteristics for the HIV-positive study participants are summarized in Table I⇑. Median CD4 count was 265 with a range of 56–1050 (interquartile range 190–429). Thirty-four percent of the participants were World Health Organization stage 3 or 4. All participants were antiretroviral drug (ARV) naive. In addition, 25 healthy HIV-negative Ugandans were enrolled as controls. Median CD4 count in the HIV-negative volunteers assayed (n = 11) was 704 (range 547-1618), similar to previous reports from Uganda (33).

Development and validation of a Treg immunophenotyping strategy

Human Treg suppressor activity is primarily confined to CD4⁺CD25^bright cells (23, 25), and almost every Treg surface marker described, including CD25, is also expressed on activated CD4⁺ cells. Therefore, the large number of activated CD4⁺ T cells associated with HIV infection makes discrimination of Tregs from activated cells problematic. One of the only known Treg surface markers that is not also an activation marker is CD62L (23), which is expressed on greater than 95% of CD4⁺CD25^bright cells. This marker is frequently overlooked in Treg analysis because it cannot be analyzed on frozen specimens (34). To avoid potential misidentification of activated cells as Tregs, we applied the stringent criteria that Tregs must be CD4⁺CD3⁺CD25^brightCD62L^high. The addition of CD62L as a Treg marker provides better distinction between activated CD4⁺ T cells and Tregs, and allows the visualization of a more readily definable population regardless of HIV infection status (Fig. 1⇓A). This gating consistency is demonstrated by the very small variation in Treg number seen in HIV-negative samples assayed over the course of this study. We additionally found that the gated population was enriched in its expression of CD45RO, CTLA-4, HLA-DR, and CD38 (data not shown), consistent with the known immunophenotype of Tregs (23, 25).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Validation of a Treg immunophenotyping strategy. Freshly isolated PBMCs from HIV-negative and HIV-positive Ugandan volunteers were stained with anti-CD62L FITC, anti-CD25 PE, anti-CD3 PerCP Cy5.5, and anti-CD4 allophycocyanin. After gating on CD3⁺CD4⁺ cells, Tregs were defined as CD25^bright and CD62L^high (a). Treg function was validated by sorting this population and performing suppressor assays on responder cells (CD4⁺CD25^negCD62L^high) (Tresp) to show dose-dependent inhibition of anti-CD3-induced proliferation (b). Tregs were added at varying ratios to responder cells (Treg:Tresp) ranging from 1:1 to 1:64. Cells were stimulated with soluble anti-CD3, and proliferation was measured by tritium incorporation. Background proliferation was always <100 cpm and subtracted from net values.

Validation of our Treg immunophenotyping strategy was achieved by sorting the CD25^brightCD62L^high population and demonstrating that this population potently suppresses anti-CD3-induced proliferation, the gold standard for assaying Treg activity (Fig. 1⇑B). Our observed degree of suppression (greater than 40% at 1:1.32) is among the strongest reported for human Tregs (reviewed in Ref. 19). The high suppressor capacity of this CD25^bright CD62L^high population is in agreement with reports in murine models (35).

Tregs are depleted in HIV-positive volunteers

To our knowledge, Tregs have never been quantitated in an African population. As such, we included both HIV-positive and HIV-negative volunteers to control for any potential effect of endemic pathogens or genetic background. In HIV-negative Ugandans, 2.8% (SD = 1.1; data not shown) of CD4⁺ T cells were CD25^brightCD62L^high and were thus defined as Tregs. This number closely parallels reports from previous studies in humans (2–4% of CD4⁺ cells) (23, 24, 25). We found little variability in Tregs (expressed as a percentage of CD4 cells) and no significant association between Tregs and CD4 count in our cohort of HIV-negative volunteers (Fig. 2⇓A). HIV-infected ARV naive Ugandans demonstrate a wide range of Treg levels particularly at CD4 counts below 400 (Fig. 2⇓B). The observed variation was not attributable to an assay technique, as HIV-negative samples were evaluated concurrently with HIV-positive samples using preset gating.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Tregs as compared with CD4 count in HIV-positive and HIV-negative volunteers. Greater variability in Tregs (represented as a percentage of CD4⁺ cells) is seen in HIV-positive (b) compared with HIV-negative volunteers (a). This variability is especially apparent in individuals at late stages of disease. Absolute Treg number is correlated with CD4 count in HIV-positive volunteers (c).

Traditionally, quantitation of Tregs has been expressed as a percentage of CD4⁺ T cells. However, CD4 counts decline in advanced HIV infection, and expressing Tregs as a percentage of CD4⁺ T cells will not adequately reflect the loss of Tregs or their stoichiometry with CD8⁺ cells. Therefore, we adopted the unique approach of expressing Tregs as an absolute number and as ratios of CD4-, CD8-, or CD3-positive cells. For this analysis, Tregs were measured independently as both percentage of CD4 and percentage of CD3-positive cells by flow cytometry, while the absolute Treg count and Treg:CD8 ratios were calculated using the absolute CD4 and CD8 counts. This approach allows the comprehensive representation of the interaction of Tregs with their target cells yielding insight into their mechanism of action.

To examine the relationship between overall CD4 cell count and Tregs, we first expressed Tregs as either an absolute number or as a percentage of CD4⁺ T cells and performed linear regression against absolute CD4 cell count. We found a significant correlation between the decline in absolute CD4⁺ T cells and the decline in absolute Tregs (Fig. 2⇑C; r² = 0.39, p < 0.0001), thereby demonstrating that Tregs are depleted in the course of HIV disease. However, the strength of the correlation coefficient suggests that the decline of these two populations of CD4⁺ T cells may have distinct kinetics. Thus, CD4 count alone is not necessarily predictive of Treg count, and these populations can be analyzed independently. Furthermore, we observed increased numbers of Tregs (expressed as percentage of CD4⁺ T cells) in many HIV-positive volunteers with low CD4 cell count, suggesting a slower decline of this population; however, this trend did not reach statistical significance (r² = 0.04 and p = 0.068) (Fig. 2⇑B).

Treg depletion is associated with immune activation

Because Tregs decrease as HIV disease progresses, we hypothesized that the depletion of cells with suppressive activity might contribute to the immune hyperactivation associated with advanced HIV disease. We defined immune activation for both CD4⁺ and CD8⁺ T cells as the simultaneous expression of both HLA-DR and CD38. This immunophenotyping approach has been extensively validated in prior studies and demonstrates T cell activation to be a strong prognostic indicator for progression to AIDS (1, 2, 4, 5, 36, 37, 38, 39, 40, 41, 42). In our study, HIV-negative Ugandans had levels of CD4⁺ and CD8⁺ T cell activation comparable to that reported for HIV-negative Ethiopians (5 vs 7–8% for CD4⁺, and 13 vs 12–15% for CD8⁺ T cells, respectively) (41). In contrast, antiretroviral naive HIV-positive individuals showed a striking increase in both mean CD4⁺ and CD8⁺ T cell activation (31% for CD4⁺ and 63% for CD8⁺ T cells, respectively), exceeding levels previously reported for HIV infection in developed countries (41) (Fig. 3⇓A). CD4⁺ and CD8⁺ T cell activation were negatively associated with CD4 count (r² = 0.34, p < 0.001 and r² = 0.09, p = 0.0011, respectively) and positively associated with viral load (r² = 0.146, p < 0.0001 and r² = 0.193, p < 0.0001, respectively) in this population (Fig. 3⇓, B and C, and data not shown). No significant association between CD4 count and CD4⁺ or CD8⁺ T cell activation was observed in HIV-negative Ugandans (p = 0.335 and p = 0.861, respectively; data not shown).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

HIV is associated with high CD4⁺ and CD8⁺ T cell activation. CD4⁺ and CD8⁺ T cell activation were measured in HIV-negative and HIV-positive ARV naive volunteers by flow cytometry using coexpression of HLA-DR and CD38 (a). CD4⁺ T cell activation is inversely associated with CD4 cell count and directly associated with viral load (b and c). Distribution of Treg (expressed as percentage of CD3⁺ T cells) and CD4⁺ T cells with an activated phenotype in HIV-negative individuals (d).

The relationships between Tregs and immune activation in HIV-positive ARV naive volunteers were next examined using all four representations of Tregs (Table II⇑). Univariate analysis revealed a negative correlation between Tregs and both CD4⁺ and CD8⁺ T cell activation. In multivariate regression analysis controlling for the effect of overall CD4⁺ T cell number and viral load on immune activation, this relationship remained highly significant for all four representations of Treg number, except between CD8⁺ T cell activation and Tregs expressed as absolute number or as a percentage of CD4⁺ T cells. Interestingly, Tregs expressed as a percentage of CD3⁺ T cells appeared to correlate best with overall higher r² values. All associations remained significant whether CD4 count was represented as an absolute number, CD4:CD8 ratio, or a percentage of T cells (data not shown). Finally, we found no correlation between Tregs and sex or viral load in multivariate analysis; however, a negative association was apparent with age (Table II⇑). In contrast, similar associations between Tregs and CD4⁺ and CD8⁺ T cells expressing activation markers were not observed for any of the four representations of Treg number (Fig. 3⇑D and data not shown) in HIV-negative Ugandan volunteers.

Because Tregs are a subset of CD4⁺ T cells, we next asked whether the previously described associations between CD4 count and T cell activation were actually attributable specifically to Tregs rather than to CD4⁺ T cells in general. We focused our analysis on Tregs as expressed as a percentage of CD3⁺ T cells, as this variable had the strongest correlation with immune activation (shown above) and also most accurately represents Treg interactions with both CD4⁺ and CD8⁺ T cells. Using standardized coefficients in multivariate analysis to compare the strength of the associations between immune activation and Tregs, CD4 count, and viral load, we found that both CD4⁺ and CD8⁺ T cell activation correlate more strongly with Tregs than with CD4 count (standardized coefficient 0.40 vs 0.29 and 0.29 vs 0.19 for CD4⁺ and CD8⁺ T cell activation, respectively) (Table III⇓). The increased relative strength of this relationship between immune activation and Tregs held true for all representations of Tregs, except for Tregs represented as a percentage of CD4⁺ T cells (data not shown). This suggests that Treg depletion in HIV infection plays a critical role in immune activation that has previously been attributed only to CD4⁺ T cell depletion. Viral load continued to have a significant, positive association with CD4⁺ and CD8⁺ T cell activation in our analysis. Compared with viral load, Tregs appear to have a stronger association with CD4⁺ immune activation (standardized coefficient 0.40 vs 0.28), but a weaker association with CD8⁺ T cell activation (standardized coefficient 0.29 vs 0.42) (Table III⇓).

Tregs and CTL dysfunction

Next, we investigated whether Tregs contribute to T cell effector dysfunction as measured by IFN-γ assays. Impaired CD8⁺ T cell function in HIV infection has previously been attributed to loss of CD4⁺ T cell help (8). We hypothesized that CD8⁺ T cell dysfunction may alternatively be caused by the observed increasing ratio of Tregs to T helpers in advanced diseases. To test this principle, we first depleted Tregs using anti-CD25 magnetic beads and examined the effect of in vitro depletion on Gag-specific CD8⁺ T cell responses. On average, our depletions achieved greater than 10-fold reduction in the CD25^brightCD62L^high Treg population. We found that Treg depletion in our study population resulted in an increase in CD8⁺ HIV-specific responses in six of six randomly selected ARV naive individuals assayed (Fig. 4⇓A), thereby confirming that Tregs appear to have the ability to suppress HIV-specific responses at least in vitro, similar to previous reports (26, 27).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

Depletion of Tregs in vitro increases HIV-specific CD8⁺ T cell responses. In vitro depletion of Tregs in all six HIV⁺, ARV naive individuals results in an increase in the percentage of Gag-responding CD8⁺ T cells (as measured by intracellular IFN-γ) (a). Correlation of Gag-specific responses (as measured by IFN-γ ELISPOT), with CD4 count (b) and Tregs expressed as a percentage of CD3⁺ T cells (c).

We next examined whether we could detect an ex vivo influence of Tregs on HIV-specific responsiveness. ELISPOT assays were used instead of intracellular flow cytometry to accommodate the large number of samples. For this analysis, Tregs were represented as a percentage of CD3⁺ cells and as a percentage of CD4⁺ cells to assess their effects on all responding cells, and the impact of increasing regulator to help ratios. HIV-specific effector function did not correlate with absolute CD4 count or HIV viral load (p = 0.09 and p = 0.82, respectively) (Fig. 4⇑B and data not shown) similar to previous studies (7, 43). No significant association was seen between Gag-specific responses and Tregs expressed either as a percentage of CD3⁺ (p = 0.6) (Fig. 4⇑C) or as a percentage of CD4⁺ cells (p = 0.9) (data not shown) in univariate or multivariate analysis controlling for CD4 count and viral load (data not shown).