Delayed Expression of PD-1 and TIGIT on HIV-Specific CD8 T Cells in Untreated HLA-B*57:01 Individuals Followed from Early Infection

Patient characteristics and differences in CD4 T-cell count and viral load.

To define the interaction of HLA-mediated protection and CD8 T-cell exhaustion after reduction of acute-phase peak viral replication, we evaluated samples post-acute-phase infection. We refer to this time (8 to 26 wpi) as early chronic infection. Twelve HIV subtype B-positive patients were monitored from early chronic infection (median, 12 wpi) up to 7 years postinfection (median, 273.5 wpi) (Table 1). Two patients received treatment for a limited time during the study, while the other 10 patients remained untreated. Six of the patients (P1 to P6) carried the HLA-B*57:01 allele (9, 23); the other six subjects (P7 to P12) did not carry any variation of the HLA-B*57 allele (Table 2).

The initial CD4 T-cell count in the HLA-B*57:01-positive group was a median 702 cells/mm³, and that in the HLA-B*57-negative group was a median 580 cells/mm³ (Table 2). The HLA-B*57:01-positive study subjects exhibited lower baseline plasma viral loads, with a median 3,253 copies/ml, than the HLA-B*57-negative group, with a median 6,473 copies/ml (Table 2). There was no significant difference in either the viral load slope or the CD4 T-cell slope between the two groups of patients, although the HLA-B*57-negative control subjects displayed a larger range of CD4 T-cell slope (−34.0 to 1.9% change/year) (Fig. 1A) than the HLA-B*57:01-positive subjects (−15.7 to −3.6% change/year) (Fig. 1B). Over the course of the study, all patients experienced disease progression (Fig. 1A and B). In addition, several HLA-B*57-negative patients started permanent treatment after 3 to 4 years and thus became ineligible for this study. Over time, CD4 T-cell counts became more similar between the two groups, but not viral load.

Similar phylogenetic signals and sequence diversities in HIV gag p24 of HLA-B*57:01-positive and -negative subjects.

Plasma samples for phylogenetic analysis were selected from 3 to 7 time points during the study period per subject, and likelihood mapping analyses of HIV gag p24 sequences were performed (data not shown). Star-like signal was measured for all subjects, and no significant difference was observed between the two groups of patients (data not shown). The number of segregating and parsimony informative (Pi) sites (data not shown) did not reveal significant differences between the two groups of patients. Additionally, recombination analysis was performed for all 12 data sets (data not shown), and recombinant sequences were detected only in three sets (P1, P3, and P4). These sequences were excluded from subsequent phylogenetic analysis.

In each subject-specific gag p24 alignment, viral diversity and divergence were measured for sequence subsets obtained at different time points (Fig. 2A and B). As expected, both the diversity and divergence increased over time (27) for all subjects, indicating that overall, significant viral evolution could be detected in both the HLA-B*57:01-positive and -negative patients (Fig. 2B) but did not differ between the groups.

Differential frequencies of memory CD8 T-cell subsets and expression of inhibitory molecules in HLA-B*57:01-positive and HLA-B*57-negative patients.

We compared the differentiation profiles of total CD8 T cells between HLA-B*57-positive and -negative patients as well as HIV-negative donors. The cells were stained for CD27 and CD45RO to discriminate the following differentiation subsets (gating displayed in Fig. 3): naive (CD27⁺ CD45RO⁻), central/transitional memory (CM/TM; CD27⁺ CD45RO⁺), effector memory (EM; CD27⁻ CD45RO⁺), and effector memory reexpressing CD45RA/effector and effector-like (TEMRA/Eff; CD27⁻ CD45RO⁻) CD8 T cells.

A cross-sectional comparison in early chronic infection (8 to 26 wpi) revealed that the frequency of TEMRA/Eff CD8 T cells was higher among HLA-B*57-negative patients (P = 0.036) than HLA-B*57-positive patients (Fig. 4A). This difference was maintained when the available longitudinal data points up to week 150 were plotted (Fig. 4B) but not in a cross-sectional comparison during late chronic infection (155 to 309 wpi) (Fig. 5A).

In comparison to HIV-negative control subjects, HLA-B*57-negative patients showed higher frequencies of EM (P = 0.029) and TEMRA/Eff CD8 T cells (P = 0.029) as well as lower frequencies of naive CD8 T cells (Fig. 4A). We did not observe significant differences in the frequencies of CD8 T cells of any differentiation state between HLA-B*57-positive HIV-infected patients and HIV-negative control subjects (Fig. 4A).

The frequencies of inhibitory receptor expression on total memory CD8 T cells were then evaluated (Fig. 4C). In early chronic infection, all receptors were expressed at highly similar levels between the groups. In late chronic infection (155 to 309 wpi), HLA-B*57:01-positive patients had a higher frequency of CD160-expressing cells than HLA-B*57-negative patients (P = 0.026) (Fig. 5B). Longitudinal data of CD160 expression among memory CD8 T cells in HLA-B*57-positive patients confirmed an increasing frequency (r = 0.797, P = 0.003) over the duration of the infection, which was not observed in HLA-B*57-negative patients (Fig. 4D). Similarly, the frequency of PD-1 and TIGIT expression among CD8 memory T-cells increased in HLA-B*57-positive patients over time but not in the HLA-B*57-negative group (Fig. 5C), despite the lack of significant differences in cross-sectional comparisons during either early (Fig. 4C) or late (Fig. 5B) chronic infection.

In comparison to HIV-negative control subjects, HLA-B*57-negative patients during early chronic infection had lower frequencies of KLRG-1-expressing memory CD8 T cells (P = 0.012) (Fig. 4C) and HLA-B*57-positive patients during late chronic infection had higher proportions of CD160-expressing cells (P = 0.042) (Fig. 5B).

In summary, HLA-B*57-negative patients displayed altered patterns of CD8 T-cell differentiation in comparison to HIV-negative subjects within the first year of infection, patterns which were not apparent in HLA-B*57-positive patients. The lower frequency of the TEMRA/Eff subset distinguished HLA-B*57-positive from HLA-B*57-negative patients. Furthermore, the frequency of memory CD8 T cells expressing CD160, PD-1, and TIGIT increased longitudinally in HLA-B*57-positive but not in HLA-B*57-negative patients.

Impairment of functional features is aggravated in chronic infection (3 to 6 years after infection).

CD8 T-cell exhaustion is a phenomenon that affects HIV-specific T-cell responses more severely than the total or cytomegalovirus (CMV)-specific CD8 T-cell population of HIV-infected patients (26). Here, we aimed to compare autologous HIV epitope-specific CD8 T-cell responses restricted by HLA-B*57:01 to responses restricted by other alleles in HLA-B*57-positive and HLA-B*57-negative study subjects. To this end, we stimulated peripheral blood mononuclear cells (PBMCs) with optimal CD8 T-cell epitopes corresponding to the patient’s autologous HIV sequences and HLA alleles and identified responding CD8 memory T cells by their expression of CD107a, gamma interferon (IFN-γ), tumor necrosis factor (TNF), and/or CD107a. We then assessed the frequency of responding cells among CD8 memory T cells, as well as the proportion of functional markers and their combinations.

The HLA-restricted immune responses were characterized in a cross-sectional manner during early (8 to 26 wpi) and late (195 to 309 wpi) chronic HIV infection (Table 3), as well as longitudinally (Table 4). For cross-sectional comparisons, we selected the one or two strongest HIV epitope-specific CD8 T-cell responses per patient and time point for analysis. Responses detected against more than one variant of a specific epitope (i.e., corresponding to amino acid mutations), which could be reflective of cross-reactive T cells, were not included. For longitudinal analysis, the epitope-specific responses were selected in the same manner over all available time points but also included an additional response for patient 7 (P7) at the earliest time point (10 wpi) for improved longitudinal follow-up of a response being maintained at later time points. To avoid strong bias toward patients represented by a higher number of data points (epitope-specific responses), we used mixed-effects models for longitudinal analysis.

We found no significant differences in quantity of HIV-specific cells as a percentage of CD8 memory T cells responding to peptide stimulation (data not shown). When the data on individual functional markers were compared in a cross-sectional manner during early or late chronic infection, a high variability between the individual responses dominated over any differences between the patient groups (Fig. 6A). Differences, as seen in the higher frequency of IL-2-, GrzA-, or TNF-producing HLA-B*57-restricted cells and CD107a-producing cells restricted by other alleles, did not reach significance due to the small sample size. However, longitudinal analyses revealed sustained higher frequency of CD107a-producing CD8 T cells among HLA-B*57-negative subjects as well as an increase in the proportion of cells expressing CD107a among both HLA-B*57-restricted (estimated unit change per additional wpi, 0.064; P = 0.008) and responses restricted by other alleles (estimated unit change per additional wpi, 0.100; P < 0.0001) (Fig. 6B; Table 5). Among non-HLA-B*57-restricted responses, the frequency of GrzA-expressing cells decreased (estimated unit change per additional wpi, −0.058; P = 0.010) (Fig. 6C; Table 5), while the respective counterparts showed no significant change with duration of infection.

Polyfunctionality, defined as the simultaneous expression of multiple effector molecules, was evaluated during early and late chronic infection. The cross-sectional comparison of HLA-B*57:01-restricted to non-HLA-B*57-restricted HIV-specific CD8 T-cell responses revealed no differences in the permutation test (data not shown). We did, however, find differences in specific populations among responding cells. In early chronic infection, five cell populations with distinct functional assets were found at significantly different frequencies between the patient groups (Fig. 6D and E). Three cell populations, expressing two to four effector molecules simultaneously (GrzA/GrzB/perforin/TNF, GrzA/perforin/TNF, and perforin/TNF) were more frequently detected in responses to HLA-B*57:01-restricted epitopes, while those more frequent in non-HLA-B*57-restricted responses were singly positive for either CD107a or IFN-γ (Fig. 6D). Longitudinal analysis of those cell populations revealed decreasing frequencies of GrzA/perforin/TNF triple-positive cells (percent change per additional wpi, −1.6%; P < 0.0001) as well as IFN-γ single-positive cells (percent change per additional wpi, −0.9%; P < 0.0001) among non-HLA-B*57-restricted responses (Fig. 6F and G; Table 5) but not among HLA-B*57-restricted responses. This difference in association with duration of HIV infection is significantly different between the groups (P = 0.002 for GrzA/perforin/TNF triple-positive cells; P = 0.043 for the IFN-γ single-positive cell population) (Table 5).

In late chronic infection, HLA-B*57:01-restricted responses harbored a higher proportion of cell populations positive for three or five markers (CD107a/GrzA/IFN-γ/IL-2/TNF and IFN-γ/IL-2/TNF), while responses restricted by other alleles were higher in a cell population positive for CD107a, GrzA, and GrzB (Fig. 6F).

In summary, our results indicate differences in the polyfunctionality pattern of responding HIV epitope-specific CD8 T cells restricted by HLA-B*57:01 or other HLA alleles. The HLA-B*57-restricted responses displayed higher frequencies of granzyme A, and of perforin together with TNF, but constantly fewer CD107a-positive cells than the non-HLA-B*57-restricted responses.

Differential TIGIT and PD-1 expression of HIV-specific CD8 T-cell responses is apparent within a few months after infection.

To gain insight into the exhaustion of HIV epitope-specific CD8 T-cell responses restricted by either HLA-B*57:01 or other HLA class I alleles, we identified responding CD8 memory T cells by their expression of CD107a and/or IFN-γ and assessed their expression of the inhibitory receptors 2B4, CD160, KLRG-1, PD-1, and TIGIT.

In early chronic infection, we found lower percentages of cells expressing PD-1 (P = 0.038) and a trend toward lower frequencies of TIGIT expression (P = 0.067) among HLA-B*57:01-restricted responses than among the responses restricted by other HLA alleles (Fig. 7A). This difference was not apparent in late infection (Fig. 8A). In agreement with the blunted difference over time, the frequency of 2B4 (P < 0.0001), CD160 (P < 0.0001), KLRG-1 (P < 0.0001), PD-1 (P < 0.001), and TIGIT (P < 0.0001) expression among HLA-B*57-restricted CD8 T-cell responses increased over time, while only 2B4 (P < 0.0001) and KLRG-1 (P < 0.0001) expression significantly correlated with time in responses restricted by other alleles (Fig. 8B; Table 5). The frequency of responding HLA-B*57-restricted CD8 T cells expressing CD160 (P = 0.002) and TIGIT (P = 0.035) increased more rapidly over time than the level of expression on responding T cells restricted by other alleles (Table 5).

The permutation test comparing HLA-B*57:01-restricted HIV-specific CD8 T-cell responses to responses restricted by other alleles revealed a trend implying differences in early (P = 0.054) but not late (P = 0.750) chronic infection (Fig. 7B). In early chronic infection, all cell populations differing between the groups coexpressed TIGIT and PD-1 together with one to three additional inhibitory receptors or did not express TIGIT or PD-1. Consequently, we analyzed the frequencies of TIGIT and/or PD-1 expression on HIV-specific responses in detail (Fig. 7C). Cross-sectional comparisons of data from early chronic infection confirmed a significantly lower frequency of cells coexpressing TIGIT and PD-1 (P = 0.019) and a higher frequency of TIGIT and PD-1 double-negative cells (P = 0.019) among HLA-B*57:01-restricted responses (Fig. 7D). The frequencies of the single-positive cell populations (TIGIT⁺ PD-1⁻ or TIGIT⁻ PD-1⁺) were similar between the groups (Fig. 7D). In agreement with the permutation test, the expression of TIGIT and PD-1 double-positive cells did not differ significantly between the patient groups at late chronic infection (Fig. 8C). Longitudinally, only HLA-B*57-restricted responses expressing both TIGIT and PD-1 significantly correlated with time of infection (unit change per additional wpi, 0.108; P < 0.001), but the difference between the groups did not reach significance (P = 0.091) (Fig. 7E; Table 5).

Given the close relationship of CD8 T-cell exhaustion and antigen levels, we investigated the correlation between the frequency of TIGIT and PD-1 double-positive cells and clinical parameters. While we found that TIGIT and PD-1 double-positive cells correlated with viral load for HIV-specific HLA-B*57-restricted responses (estimated unit change for each 10-fold difference in viral load, 19.504; P < 0.001), responses restricted by other alleles showed no significant correlation (Fig. 7F; Table 5). The correlation with viral load was significantly different between the groups (P = 0.031). Only HLA-B*57-restricted responses showed significant correlation between the frequency of TIGIT and PD-1 double-positive cells and CD4 count (estimated unit change per 2-fold increase in CD4 count, −17.109; P = 0.002) (Fig. 7G; Table 5). Both groups showed inverse correlations with the percentage of CD4 T-cells among lymphocytes (CD4%; P < 0.001 for B*57-restricted responses, P = 0.046 for other responses) and CD4/CD8 ratio (P < 0.0001 for B*57-restricted responses, P = 0.025 for other responses) (Fig. 7H and I; Table 5).

In summary, the lower frequency of cells coexpressing TIGIT and PD-1 among HLA-restricted HIV-specific responses within the first year of HIV infection distinguished HLA-B*57-positive from HLA-B*57-negative patients.

Similar transcription factor profiles in both patient groups.

The transcription factors eomesodermin (Eomes) and T-bet are fundamental for CD8 T cells: T-bet correlates with effector functions, and Eomes is associated with the differentiation and long-term survival of memory cells (28–30). Their expression levels seem to some degree mutually exclusive, as CD8 T cells high in one of the transcription factors show only intermediate expression of the other, leading to the distinction of T-bet^hi Eomes^dim and T-bet^dim Eomes^hi CD8 T-cell populations.

Cross-sectional comparisons of the transcription factor profiles did not reveal any significant differences between the patient groups in early or late chronic infection (Fig. 9A). However, consistent with previous findings (25), we found declining percentages of the T-bet^hi Eomes^dim CD8 T-cell population (percent change per additional wpi, −0.4%; P < 0.001) and a corresponding increase in T-bet^dim Eomes^hi cells over time (percent change per additional wpi, 0.2%; P = 0.007) in our HLA-B*57-positive study subjects (Fig. 9B and C; Table 5). Similar changes in the HLA-B*57-negative patients were not statistically significant, consistent with the greater individual differences and genetic heterogeneity in this group. A third cell population negative for both T-bet and Eomes showed a decrease over the study duration in HLA-B*57-positive patients (percent change per additional wpi, −0.5%; P < 0.001) (Fig. 9D; Table 5) and no change over time among HLA-B*57-negative patients (Fig. 9D; Table 5). This discrepancy between the groups reached statistical significance (P = 0.002).

Study subjects.

Twelve HIV-infected subjects (P1 to P12) were selected from the San Francisco-based cohort OPTIONS at the University of California (47) (Tables 1 and 2). Patients were monitored for up to 7 years from early infection, where the time of infection was estimated as the midpoint between the last reported negative test and the first positive test (9). Six subjects (P1 to P6) carried the HLA B*57:01 allele (9). The other six subjects (P7 to P12), not carrying the HLA allele B*57, have been described in an article linking early HIV-specific responses against HIV Gag to slower disease progression (12). Healthy control samples (n = 9) were selected based on the presence of CMV- or Epstein-Barr virus (EBV)-specific responses, as well as age (median, 33 years; interquartile range [IQR], 26 to 33 years) and gender (89% males). The University of California, San Francisco (UCSF), Committee on Human Research (IRB no. 10-00301) and the Regional Ethical Council in Stockholm, Sweden (2008/1099-31, 2016/1314-32), approved this study, and all patients provided written informed consent.

RNA extraction, cDNA synthesis, PCR amplification, and sequencing.

Longitudinal plasma samples were stored at –80°C and subsequently processed. HIV gag p24 single genome sequences for the HLA-B*5701 subjects (P1 to P6) were obtained as previously described (9). For the control subjects (P7 to P12), viral RNA was extracted from 1 ml plasma with a QIAamp viral RNA minikit, and the entire viral RNA extraction was used for cDNA synthesis. cDNA synthesis was performed using a Superscript III system (Invitrogen) with gene-specific primers using a limiting-dilution digital nested-PCR, single-genome sequencing (9). Sequencing of the PCR products was performed at the sequencing core facility of the University of Florida on an ABI3700 system, under IRB approval no. 258-2012. Sequences were assembled with Geneious 5.6 software created by Biomatters (Auckland, New Zealand). Chromatograms were manually examined for the presence of double peaks indicative of at least two templates per sequencing reaction, and these genomes were discarded.

Sequence analysis.

Subject-specific alignments were manually obtained with the freely available BioEdit software developed by Tom Hall. For each data set, mean nucleotide divergence and diversity were estimated at each time point, and standard errors were calculated by bootstrapping (500 replicates), with standard errors inferred from 500 bootstrap replicates. Calculations were carried out with MEGA5 software (48). The phylogenetic signal in each data set was investigated by likelihood mapping (49). Likelihood mapping analyses were performed with the program TREE-PUZZLE (50) for each data set by analyzing 10,000 randomly chosen quartets. The presence of potential intrahost recombinant sequences was investigated with a previously published PHI test-based algorithm (51). Calculations were performed with the SplitsTree package version 4.8.

Flow cytometry.

Flow cytometry was performed as previously described (26). Briefly, peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll-Paque density gradient separation and stored in liquid nitrogen. The PBMCs were thawed, washed, and left to rest for 6 h. The cells were stimulated with autologous HLA-restricted HIV Gag peptides or optimal CMV or EBV peptides (JPT Innovative Peptide Solutions) in the presence of monensin (BD Bioscience), brefeldin A (BD Bioscience), and CD107a antibody for 10 h. Thereafter, the cells were washed, extracellularly stained (exhaustion panel), permeabilized (human FoxP3 buffer set; eBioscience), and intracellularly stained before analysis on the flow cytometer (LSR Fortessa; BD). For lineage identification and determination of differentiation phenotypes, CD14/CD19 V500 (M5E2/HIB19; BD Bioscience), CD27 BV785 (O323; Biolegend), CD3 APC-H7 (SK7; BD Bioscience), CD45RO BV605 (UCHL1; BD Bioscience), CD4 BV711 (OKT4; Biolegend), or PE-Cy5.5 (S3.5; Invitrogen) and CD8 Qd565 (3B5; Life Technologies) were used in both panels. We stained for T-bet phycoerythrin (PE) (4B10; Biolegend) and eomesodermin (Eomes) PE-eF610 (WD1928, eBioscience) in both panels due to their central role in effector function and survival of memory T cells, as well as CD107a PE-Cy5 or PE-Cy7 (H4A3; BD Bioscience) and IFN-γ AF700 (B27; BD Bioscience) for identification of CD8 T-cell responses after peptide stimulation. A panel focused on inhibitory receptors (exhaustion panel) included staining of KLRG-1 AF647, CD160 AF488 (BY55; BD Bioscience), PD-1 BV421 (EH12.2H7; Biolegend), 2B4 PE-Cy5 (C1.7; Beckman Coulter) and TIGIT PE-eF710 (MBSA43; eBioscience). The second panel (function panel) included GrzB PE-Cy5.5 (GB11; BD Bioscience), IL-2 APC (MQ1-17H12; BD Bioscience), GrzA AF488 (CB9; Biolegend), perforin BV421 (B-D48; Biolegend), and TNF PE-Cy7 (MAb11; BD Bioscience) for a detailed analysis of effector functions, including degranulation capacity, loading of cytotoxic granules, and cytokine production.

Data analysis and statistics.

Data were analyzed with FlowJo LLC software (10.1r1), SPICE (52), GraphPad PRISM (version 6.0d), and R scripts (custom tailored in collaboration with the research group of Ole Lund, DTU, Copenhagen). To compare the combined expression of inhibitory receptors on epitope-specific CD8 T cells between the patient groups, we used permutation tests and Student´s t test provided in SPICE. Cross-sectional comparisons in GraphPad PRISM were made with Mann-Whitney tests; for correlations of total (memory) CD8 T cells with duration of infection and clinical parameters, Spearman´s rank correlations were determined. To avoid bias toward patients with multiple available responses, mixed-effects models were calculated (Table 5) with Stata v16.1 software (StataCorp LLC, College Station, TX) when comparing HIV epitope-specific responses. If not indicated otherwise, linear mixed-effects models were used. Due to the scarcity of the cell population depicted, a zero-inflated negative binomial (ZINB) regression was used for the statistical analysis shown in Fig. 6F. All results of mixed-effects models are shown in Table 5.

Data availability.

GenBank accession numbers for the sequences for subjects P1 to P6 are JX234575 to JX235332. GenBank accession numbers for the sequences for subjects P7 to P12 are MT274756 to MT275450.