We simultaneously analyzed cytokine production and effector/memory markers of CMV- and HIV-responsive T cells using the gating strategy shown in Figure 1, in a cohort of 19 HIV-positive subjects with progressive disease and 20 healthy controls (Table 1). Since activation of T cells was required to detect cytokine production, we wanted to ensure that the phenotypic markers examined did not modulate during short-term stimulation. Using an MHC-peptide tetramer to isolate CMV-specific T cells, we demonstrated that 6-hour stimulation had minimal effect on the distribution of CD27, CD28, and CD45RA on these cells (Figure 2). Therefore, our analysis closely reflected the in vivo state of differentiation of these cells, rather than the effects of in vitro stimulation.

The magnitude of the T cell IFNγ responses to CMV pp65+IE-1 and to HIV Gag+Env, as well as the IL-2+ component of those responses, is shown in Figure 3 for 20 healthy HIV-negative subjects and 19 HIV-positive progressors. There were significant differences in the magnitude of CMV responses of HIV-negative versus HIV-positive subjects. These were attributable to differences in CD4 and CD8 counts between the groups (see next paragraph). However, the only significant difference in the magnitude of CMV versus HIV responses of HIV-positive subjects was a reduced frequency of HIV-specific CD8+ T cells producing both IFNγ and IL-2 (p = 0.0005).

To determine whether this reduction in CD8+ T cell IL-2 production among HIV-specific T cells was consistent when data were analyzed on a proportional basis, we examined the ratio of IL-2+/IFNγ+ T cells responding to HIV versus CMV (Figure 4). This analysis clearly showed that the proportion of HIV-specific CD8+ T cells that could produce IL-2 was significantly reduced compared to that of CMV-specific CD8+ T cells in HIV-positive subjects (p = 0.0005). No other significant differences in IL-2+/IFNγ+ T cell ratios were found.

We next analyzed the differentiation profile of HIV-specific and CMV-specific T cells in healthy subjects versus HIV-positive progressors. Several trends were observed.

The response of healthy subjects to CMV (Figure 5A, top panel) was quite heterogeneous, involving most of the possible phenotypes of both CD4+ and CD8+ T cells. However, the CD8+ response was higher than the CD4+ response, and the fraction of IFNγ+ IL-2+ T cells was small compared to total IFNγ+ T cells. Among the CD8+ T cell phenotypes, terminally-differentiated effector cells predominated, followed by CD27- CD28- CD45RA- cells. Thus, the healthy donor response to CMV was highly skewed towards effector cell phenotypes.

The CMV-specific response of HIV-positive progressors was similarly skewed towards effector CD8+ IFNγ+ T cells (Figure 5A, middle panel). The number of CMV-specific CD4+ T cells was significantly lower in the HIV-positive subjects, although the distribution of phenotypes was similar to that of HIV-negative subjects. These differences were largely a result of reduced CD4+ T cell counts in the HIV-positive subjects, since the differences were not statistically significant on a percentage basis (Table 2).

The HIV response of HIV-positive progressors (Figure 5A, bottom panel) was different from the CMV responses in that it was dominated by CD8+ IFNγ+ T cells of intermediate differentiation (CD27+ CD28- CD45RA-). This phenotype was rare within the healthy donor CMV response. Also, the HIV responses contained hardly any CD4+ or CD8+ IL-2-producing cells, and only very few CD4+ IFNγ+ cells. These differences in CD4 and CD8 compartments were significant by ANOVA (Table 2), with the most highly significant differences being in the CD8+ IFNγ+ subset (p ≤ 0.00008). These differences were significant even when subjects receiving ART were excluded (Table 2, bottom half).

It should be noted that the distribution of CD8+ T cell phenotypes seen in HIV-responsive cells was not reflected in the overall CD8+ T cell compartment (Figure 5B). The total CD8+ T cell pool was quite heterogeneous, and included a large cohort of effector-like cells (CD27- CD28- CD45RA+ and CD27- CD28- CD45RA-).

Overall, the data of Table 2 confirm that the differences in phenotypic patterns observed in the CMV and HIV responses of HIV-positive progressors were statistically significant. The most significant difference (p ≤ 0.00008) was seen in CD8+ IFNγ+ cells. This was true when data were analyzed as percentages or as absolute counts, and whether or not subjects on ART were included.

The above data demonstrate that HIV-specific CD8+ T cells in HIV-positive progressors show two major differences compared to CMV-specific CD8+ T cells: (1) a lower proportion of IL-2-producing cells, and (2) a less differentiated phenotype. We tested two potential hypotheses that might explain the coexistence of these two phenomena.

The first hypothesis, that the less differentiated CD8+ T cells tend not to produce IL-2, can be easily dismissed by examination of IL-2 production as a function of phenotype (Figure 6). While IL-2-producing T cells could be found amongst all phenotypes of CD8+ IFNγ+ T cells specific for CMV, proportionally fewer cells produced IL-2 upon differentiation to effector phenotypes (Figure 6, top panel). This pattern was somewhat disrupted in the CMV response of HIV-positive subjects (middle panel), but the most pronounced difference was observed in the HIV response (bottom panel). HIV-responsive CD8+ T cells of all phenotypes essentially did not make IL-2. Thus, there is a pervasive breakdown in IL-2 production from all HIV-specific CD8+ T cells, and not simply a replacement of cells that normally make IL-2 with cells that normally do not.

A second hypothesis for the co-existence of fewer CD8+ IL-2+ T cells and altered differentiation in HIV responses was that IL-2 production is required to drive effector T cell differentiation. If this were true, one would expect to see a quantitative correlation between CMV-specific or HIV-specific T cells and differentiation state. Figure 7A shows that the number of HIV-specific IFNγ+ IL-2+ CD8+ T cells is significantly correlated with the number of HIV-specific CD8+ terminal effector T cells (CD27- CD28- CD45RA+) (p = 0.0004, top). The correlation remained significant when CMV-specific responses were also included (p = 0.0073, bottom). Finally, the correlations remained significant (p = 0.03) when frequencies were reported as a percentage of CD8+ T cells, rather than as absolute counts (data not shown). Therefore, a greater number of CD8+ IL-2-producing cells correlated with more terminally differentiated CD8+ effector cells in the antigen-specific response.

We also looked at the correlation of CD4+ IL-2-producing cells with CD8+ effector cells. These were not as well correlated for either HIV-specific or pooled HIV-specific and CMV-specific responses (p = 0.0743 and p = 0.0410, respectively; data not shown). However, when the frequency of all IFNγ+ IL-2+ T cells (CD4+ and CD8+) were plotted against the frequency of CD8+ terminal effector cells (Figure 7B), the correlation was highly significant, both for the HIV-specific stimulation (p = 0.0006, top) and for pooled HIV-specific and CMV-specific results (p = 0.0034, bottom). This argues strongly for a role of IL-2 in promoting CD8+ effector T cell differentiation.

Individual peptides of 15 amino acid residues, overlapping by 11 amino acids each, were designed to span the sequences of CMV pp65 (138 peptides) and IE-1 (Immediate Early-1; 120 peptides), and HIV p55 Gag (SF2 strain, 127 peptides) and env (MN strain, 204 peptides). Peptide mixes (SynPep, Dublin, CA) were dissolved in DMSO at stock concentrations of 0.7–1.0 mg/ml per peptide. The mixes were used at a final concentration of 1.2–2 μg/ml per peptide 29.

PBMC were isolated from heparinized whole blood within eight hours of collection, by centrifugation in CPT tubes (BD Vacutainer, Franklin Lakes, NJ). PBMC were washed and resuspended in RPMI with 10% fetal bovine serum (cRPMI) at 5 × 10⁶ – 1 × 10⁷ cells/ml. Five hundred microliters of PBMC in cRPMI were aliquoted into wells of a 24-well deep-well plate (Qiagen, Valencia, CA). CMV pp65+IE-1 or HIV p55 Gag+Env peptide mixes were added to the appropriate wells, plus brefeldin A (BD Biosciences) at a final concentration of 10 μg/ml; an additional well received only brefeldin A as a negative control. The plate was incubated for six hours at 37°C, then held overnight at 18°C.

Cell-surface markers such as CD3, CD4, and CD8 can be stained before or after cell fixation and permeabilization. We stained for these markers before fixation, since this enhanced resolution of positive and negative populations. Although some down-modulation of these markers occurred on activated cells, we were still able to include down-modulated cells within the positive gate.

Details of the cytokine flow cytometry protocol can be found on the Maecker Lab weblog 30. Briefly, cells were stained for 60 minutes with CD28 PerCP-Cy5.5, CD45RA PE-Cy7, CD27 APC, CD8 APC-Cy7, CD3 Pacific Blue, and CD4 AmCyan (all from BD Biosciences). Samples were treated with BD FACS Lysing Solution followed by BD FACS Permeabilizing Solution 2, washed, then stained with IFNγ FITC and IL-2 PE (BD Biosciences) for 60 minutes. They were then washed and resuspended in BD Stabilizing Fixative (BD Biosciences) and held at 4°C until sample acquisition.

For each donor, 100 μl of PBMC remained unstained, for use when establishing PMT voltage settings. Eight tubes containing BD CompBeads Anti-Mouse Ig,κ and Negative Control Compensation Particles (BD Biosciences) were each stained with one of the eight mAbs listed above. The beads were treated identically to the PBMC samples, except that all stains (including cytokines) were added during the "surface staining" step.

Flow cytometry was performed using an LSRII (BD Biosciences) equipped with blue (488 nm), red (633 nm), and violet (405 nm) lasers. Compensation settings were established using the AutoComp option of FACS DiVa software (BD Biosciences). An average of 2 × 10⁶ CD3+ cells were collected for HIV-negative subject samples, and an average of 4 × 10⁵ CD3+ cells were collected for HIV-positive subject samples.

Data were analyzed with FACS DiVa software. Where appropriate, data were displayed with a transformation that allows for improved visualization of events at the lower end of the log scale (BiExponential analysis). A gating hierarchy was designed as shown in Figure 1. The result of this gating strategy was 32 possible phenotypes of antigen-responsive T cells.

Dead cells were excluded by use of a side scatter gate and by gating on single-stained cells for either CD4 or CD8. Non-viable cells generally bound both of these antibodies non-specifically.

Cytokine production in the absence of stimulation was low, and largely restricted to CD8+ terminally differentiated effector cells (CD27- CD28- CD45RA+ 121531323334353637; Fig 5A, black bars). As such, backgrounds were not subtracted from the responses shown in the grey bars. Responses ≥ 100 cells/ml were significantly higher than background for all phenotypes except the terminal effector cells mentioned above.

Data were batch-analyzed to ensure uniform gating, and statistics files were uploaded to a database created for this study. This database enabled selection of particular data subsets of interest, which were then downloaded into Microsoft Excel and GraphPad Prism (GraphPad, San Diego, CA) for statistical analysis and graphing. Statistical comparisons between HIV-positive and HIV-negative subjects were performed using a Mann Whitney test. Statistical comparisons between different stimulations of HIV-positive samples were performed using a Wilcoxon signed rank test. Values of p < 0.025 were considered statistically significant, to correct for the two sets of comparisons listed above.

ANOVA (Table 2) were conducted using the Open Source statistics package R 38. For each of the four cell populations (CD4+ IFNγ+, CD4+ IFNγ+ IL-2+, CD8+ IFNγ+, or CD8+ IFNγ+ IL-2+), a separate model was fit in which the response was the absolute cell count for each subject and the main effect was the eight combinations of the markers CD27, CD28, and CD45RA. Two contrasts were added to test specifically whether or not the average numbers for each phenotype followed the same pattern for HIV-positive CMV versus HIV-negative CMV, and for HIV-positive CMV versus HIV-positive HIV responses. The p-values reported were for the interaction term between each of these contrasts and the main effect for phenotype. A p-value < 0.05 indicates that the two subject groups had significantly different patterns in the numbers of cells of each phenotype.