Clonotype and repertoire changes drive the functional improvement of HIV-specific CD8 T cell populations under conditions of limited antigenic stimulation

Abstract

Persistent exposure to cognate antigen leads to the functional impairment and exhaustion of HIV-specific CD8 T cells. Antigen withdrawal, due either to antiretroviral treatment or the emergence of epitope escape mutations, causes HIV-specific CD8 T cell responses to wane over time. However, this process does not continue to extinction, and residual CD8 T cells likely play an important role in the control of HIV replication. Here, we conducted a longitudinal analysis of clonality, phenotype and function to define the characteristics of HIV-specific CD8 T cell populations that persist under conditions of limited antigenic stimulation. Antigen decay was associated with dynamic changes in the TCR repertoire, increased expression of CD45RA and CD127, decreased expression of PD-1 and the emergence of poly-functional HIV-specific CD8 T cells. High definition analysis of individual clonotypes revealed that the antigen loss-induced gain of function within HIV-specific CD8 T cell populations could be attributed to two non-exclusive mechanisms: (i) functional improvement of persisting clonotypes; and, (ii) recruitment of particular clonotypes endowed with superior functional capabilities.

Introduction

Several observations suggest that antigen-specific CD8 T cells are important for the control of HIV-1 infection (1-4); it has also been demonstrated that HIV-specific CD8 T cell responses in long-term non-progressors (LTNPs) and in HLA-B*27⁺slow progressorsexhibit a wide array of effector functions (5, 6). However, chronic antigen persistence at high levels leads to dysfunction and exhaustion, and HIV-specific CD8 T cells are therefore frequently characterized by an inability to produce cytokines, compromised proliferative capacity and impaired cytotoxic activity(7-13). Despite the widespread usage of highly active antiretroviral therapy (HAART), relatively little is known about its impact on HIV-specific CD8 T cells. The frequency of such cells declines rapidly upon the initiation of HAART (14-16). Nevertheless, persisting HIV-specific CD8 T cells could still play an important role in controlling residual viral replication. Similarly, despite the fundamental significance of perpetual viral evolution in relation to disease pathogenesis, relatively little is known about the impact of immune escape on HIV-specific CD8 T cells. Of note, de novo variant-specific CD8 T cell responses can emerge, suggesting thatsome of these escape mutations can still be processed and presented to T cells (17). Furthermore, it is known that responses specific for wildtype epitopes wane over time due to diminished antigenic drive, yet this process does not lead to the extinction of CD8 T cells that recognize wild type epitopes. Thus, CD8 T cells with wildtype epitope specificity persist in some form and appear to play an important role in the maintenance of escape mutations within the viral quasispecies. Strong evidence for this assertion comes from HIV and SIV transmission studies, in which selected escape mutations rapidly revert to optimize viral fitness in the absence of the presenting major histocompatibility complex class I (MHCI) molecule and remain relatively stable in the presence of the appropriate restriction element due to the induction of wild type-specific CD8 T cell populations by viral revertants(18-21).

It was recently documented that both HAARTand viral sequence diversification lead to the emergence of poly-functional HIV-specific CD8T cells (22, 23).Rehr et al. demonstrated that, after 24 weeks of HAART, HIV-specific CD8 T cells gradually recovered their cytokine secretion capacity, displayed increased expression of CD28 and CD127, and down-regulated PD-1(22). Furthermore, Streeck et al. showed that antigen decay over time decreased the exhausted phenotype of HIV-specific CD8 T cells, while mono-functionality decreased slightly for responses directed against escaped epitopes (23). In another study, it was shown that antigen decay resulting from the emergence of escape mutations or the institution of HAART was associated with significantly decreased co-expression of CD38 and PD-1 on HIV-specific CD8 T cells, whereas a rise in viral load resulted in increased CD38/PD-1 co-expression(24). However, the characteristics of the clonal T cell receptor (TCR) repertoire under conditions of limited antigenic stimulation remain unknown.

AlthoughTCR repertoire studies have been performed in the context of several acute and persistent viral infections including HIV-1 (25-29), longitudinal studies that aim to characterize the evolution of the HIV-specific CD8 T cell repertoire and further couple HIV-specific CD8 T cell clonotypes to functional profiles have been limited(30). Here, we hypothesized that antigen decay would enhance the functional quality of HIV-specific CD8 T cell responses by influencing the antigen-specific CD8 T cell repertoire. Accordingly, to better define the qualitative features of HIV-specific CD8 T cells during antigen withdrawal, we undertook a comprehensive analysis of HIV-specific CD8 T cell responses in the face of antigen decay due to the initiation ofHAART or the emergence of viral epitope mutations in a cohort of 8 HIV-infected individuals. In each case, we conducted a longitudinal examination of the clonal composition, phenotypic status and functional profile of CD8 T cell populations specific either for autologous stable epitopes at time points prior to, during and after HAART or for autologous wild type epitopes and, where present, the corresponding variant epitopes before and after viral escape. For comparative purposes, CMV-specific CD8 T cell responses were studied in parallel. The data provide clear evidence for dynamic changes in the TCR repertoire associated with anantigen loss-mediated functional reconstitution within the HIV-specific CD8 T cell compartment, which can be attributed to two distinct mechanisms: (i) functional improvement of persistent clonotypes; and, (ii) recruitment of particular clonotypes endowed with superior functional capabilities.

Materials and Methods

Study participants

Eight HIV-infected subjects were enrolled from Hospital Notre Dame, Montreal, Quebec, Canada. All were male, aged between 22 and 49 years (mean, 38 years).The estimated date of infection in each case was based on clinical history, p24 ELISA and Western Blot HIV Test analysis. Longitudinal cryopreserved peripheral blood mononuclear cell (PBMC) samples were available from each subject and written informed consent for sample use was obtained in all cases. Subjects 1, 2, 3, 4 and 8 started HAART during the course of the study; subjects 5, 6 and 7 were treatment-naïveduring the course of the study.Subjects 1, 2 and 3 started HAART during early phase HIV-1 infection(1 month after infection); subjects 4 and 8 started HAART 6 months after the estimated date of infection. All 5 subjects stopped HAART voluntarily. Initial immunological characterization comprised HLA genotyping and screening for HIV epitope-specific CD8 T cell responses usingfluorescent peptide-MHCI (pMHCI) tetrameric complexes based on autologous viral sequences; CMV-specific CD8 T cells were only detected in subjects 1,2, 4 and 8.

Autologous viral sequencing

For bulk analysis of autologous viral populations, amplification and sequencing of the near complete HIV genomes was performed as described previously(31). For clonal sequencing, viral RNA was extracted using the QIAamp viral RNA minikit (Qiagen) and reverse transcribed using the SuperScript One-Step RT-PCR kit (Invitrogen). Amplification was performed using sets of outer primers specific for each region of interest. Products were then further amplified in a nested PCR using specific sets of inner primers. Amplicons were ligated into pGEM-T Easy vector (Promega) and cloning was performed by transformation of competent DH5-α E. coli. For each amplicon, 24 white colonies were picked, screened using standard M13 primers and then sequenced. All sequences were analyzed using Codon Code Aligner Version 3.0 (Codon Code Corporation).

Tetrameric pMHCI complexes

Soluble biotinylated pMHCI monomers were manufactured by the CANVAC tetramer core facility (Montreal, Canada) as described previously(32) and tetramerized with fluorochrome-conjugated streptavidin at a 4:1 molar ratio. The following pMHCI tetramers were produced: HLA-A*0201-FLGKIWPSHK (HIV Gag p2p7p1p6 FL10, residues 70-79) and HLA-A*0201-NLVPMVATV (CMV pp65 NV9, residues 495-503) forsubject 1, HLA-A*0301-RLRPGGKKR (HIV Gag p17 RR9, residues 20-28) and HLA-B*0702-TPRVTGGGAM (CMV pp65 TM10, residues 417-426) for subject 2, HLA-B*0801-FLKEKGGL (HIV Nef FL8, residues 90-97) forsubject 3, HLA-B*0702-TPGPGVRYPL (HIV Nef TL10, residues 128-137) and HLA-B*0702-TPRVTGGGAM (CMV pp65 TM10, residues 417-426) for subject 4, HLA-B*0702-FPQGEAREL (HIV Pol FL9, residues 8-16), a novel epitope predicted on the basis of binding algorithms and autologous viral sequence data, for subject 5, HLA-A*0301-RLRPGGKKK (HIV Gag p17 RK9, residues 20-28) for subjects 6 and 7, and HLA-B*0801-GEIYKRWII (HIV Gag p24 GI9, residues 127-135) and HLA-B*0702-TPRVTGGGAM (CMV p65 TM10, residues 417-426) for subject 8. Tetrameric complexes with the variant epitope HLA-A*0301-RLRPGGRKR were also produced for experiments conducted with samples from subject 7.

Phenotypic analysis of antigen-specific CD8 T cells

Thawed PBMC were stained with PE- or APC-conjugated pMHCI tetramers for 15 min at 37°C, then washed and surface stained with panels constructed from combinations ofthe following monoclonal antibodies (mAbs): (i) αCD3-Pacific Blue, αCD8-PerCPCy5.5, αCCR7-FITC,αCCR7-PECy7, αCD27-Alexa700, αCD27-Qdot605,αCD28-FITC, αCD45RA-APCCy7, αCD127-Pacific Blue, αCD127-PECy5, αPD-1-APC, αPD-1-PECy7and αIFN-γ-APC(BD Biosciences); and, (ii)αCD8-ECD and αVβ6-2-PE (Beckman Coulter). Live/dead fixable Aqua (Invitrogen) was used to exclude dead cells from the analysis.Data were collected using an LSRII flow cytometer (BD Biosciences) and analyzed with FlowJo software (version 8.7.3; TreeStar Inc.); compensation was performed electronically andthe Boolean platform was used to create arrays of different marker combinations. Subsequent analysis was performed using SPICE software (version 5.1; http://exon.niaid.nih.gov/spice/)(33).

Functional analysis of antigen-specific CD8 T cells

Cryopreserved PBMC were thawed and rested for 1 hr in R10 (RPMI 1640 medium supplemented with 10% fetal calf serum, antibiotics and L-glutamine) prior to stimulation with cognate peptide. Stimulated samples were pre-stained with the corresponding PE-conjugated pMHCI tetramers for 15 min at 37°C. After addition of the co-stimulatory mAbs αCD28 and αCD49d (1 μg/ml each; BD Biosciences), monensin (0.7 μg/ml; BD Biosciences), brefeldin A (10 μg/ml; Sigma-Aldrich) and αCD107a-Alexa680, cells were stimulated with the relevant autologous peptides at a concentration of 5 μg/ml for 6 hr at 37°C. Co-stimulation alone and staphylococcal enterotoxin B (1 μg/ml; Sigma-Aldrich) were used as negative and positive controls, respectively; in all cases, cells in the negative control tubes were stained with pMHCI tetramers at the end of the stimulation period. After a single wash, the cells were stained with αCD3-Pacific Blue andαCD8-ECD,then washed again prior to fixation/permeabilization (2% paraformaldehyde / 0.05% saponin) and intracellular staining with αIL-2-FITC, αTNF-Alexa700 and αIFN-γ-APC (BD Biosciences). Live/dead fixable Aqua (Invitrogen) was used to exclude dead cells from the analysis. Antigen sensitivity was assessed using a similar protocol for the measurement of intracellular IFN-γ after stimulation with a ten-fold dilutional series of cognate peptide from 10 μg/ml to 0.0001 μg/ml; experiments were performed in triplicate and the EC50 was defined as the peptide concentration that yielded 50% of the maximum IFN-γ response. For all experiments, data were acquired immediately using an LSRII flow cytometer (BD Biosciences) after a final wash step and fixation with 2% paraformaldehyde. All live lymphocyte events were collected and files were analyzed with FlowJo software (version 8.7.3; Tree Star Inc.) after electronic compensation. Functional capacity was determined after Boolean gating and subsequent analyses were performed using SPICE software (version 5. 1; http://exon.niaid.nih.gov/spice/)(33). Values used for the analysis of proportionate response representation were background subtracted.

TCR clonotype analysis

For RNA-based clonotype analysis, HIV-specific CD8 T cells were stained with cognate pMHCI tetramer and sorted viably to >98% purity by flow cytometry directly into 1.5 ml Sarstedt tubes containing 100 μl RNAlater (Applied BioSystems). mRNA was extracted using the Oligotex mRNA mini kit (Qiagen) and subjected to a template-switch anchored RT-PCR using a 3′ TRB constant region primer as described previously(34). Amplicons were ligated into pGEM-T Easy vector (Promega) and cloned by transformation of competent DH5-α E. coli. At least 50 white colonies were amplified by PCR for each sorted population using standard M13 primers and then sequenced. For DNA-based clonotype analysis, distinct functional CD8 T cell subsets were sorted to >98% purity by flow cytometry after intracellular cytokine staining. DNA was extracted by lysis of sorted T cells in 100 μg/ml proteinase K (Boehringer) for 1 hr at 56°C and then 10 min at 95°C. A hemi-nested multiplex touchdown PCR was then performed using previously described TRBV/TRBJ primer combinations and PCR conditions(35). Amplicons were then subcloned using the pGEM-T vector system and sequenced.All sequences were analyzed using Sequencher (Gene Codes Corp.); non-functional sequences were discarded from the analysis. The ImMunoGeneTics (IMGT) nomenclature is used throughout the manuscript.

Statistical analyses

Statistical analysis was performed using the two-tailed Student’s paired T-test for the data shown in Figures 2 and 3; the Wilcoxon rank sum test was used for the analyses shown in Figure 4. P < 0.05 was considered significant. The similarity of the TCR repertoire between time pointsand estimation of the reference similarity were assessed using the Morisita-Horn coefficient, calculated according to the method developed by Venturi et al.(36).The size of all samples was reduced to the size of the smallest sample, and the Morisita-Horn similarity indices were computed as if a standard number of 26 TCR sequences had been obtained in all samples. The randomization step consisted of randomly drawing a sample size of 26 in the antigen high (Ag), antigen low (NoAg) and antigen rebound (ReAg) states, and calculating the similarity measures for each sample pair of Ag/NoAg and Ag/ReAg. The distribution median resulting from 10,000 random samplings of the two reduced sample pairs was used to represent the Morisita-Horn similarity indices between the Ag/NoAg and between the Ag/ReAg subsets.We also determined whether the similarities for the observed TCR sample pairs were lower than expected by chance, if the sample pairs had been randomly assigned from the same TCR population. The reference similarity was generated under the null hypothesis that the Ag and NoAg sets were randomly assigned from the same TCR population. The TCR sequences of Ag and NoAg states were first pooled, then two sets of 26 TCR sequences were randomly drawn from the pooled populations to calculate the similarity measures. The median of this distribution provided a reference similarity for the two TCR repertoire sets, namely HIV Ref and CMV Ref.

Figure 2. Phenotypic changes in HIV-specific CD8 T cell populations under conditions of reduced antigen load.

(A&C) Maturation status of HIV-specific CD8 T cells at each time point, based on the expression of CCR7, CD27 and CD45RA. After the gates for the 3 markers were created on tetramer⁺cells, the Boolean gate platform was used to create 8 different combinations. The colored slices in each pie represent different phenotypic combinations;CD45RA⁺with other combinations (dark blue),CD45RA⁻ with other positive combinations (medium blue), and CCR7⁻CD27⁻CD45RA⁻ (light blue). (B&D)Exhaustion and survival capacity of HIV-specific CD8 T cells at each time point, based on the expression of CD28, CD127 and PD-1. The colored slices in each pie represent different phenotypic combinations; CD127⁺ with other combinations (purple), CD127⁻PD-1⁺ with other combinations (medium pink) and CD127⁻PD-1⁻ with other combinations (light pink). In (A&B), which depict representative analyses for subject 1, the bars on the x-axis demonstrate the frequencies of cells belonging to a particular combination. The numbers under the pies indicatethe time points studied (m=month). Black, light grey and dark grey indicate pre-HAART or pre-escape, HAART or escape, and post-HAART time points, respectively. (E-G) Percentage of tetramer⁺ cells expressing CD45RA (E), CD127 (F) and PD-1 (G) at the first time point studied (high antigen load) and the second time point obtained after antigen decay (low antigen load). Ag=antigen.

Figure 3. Functional changes in HIV-specific CD8 T cell populations under conditions of reduced antigen load.

(A)Representative example of simultaneous multi-functional assessment of HIV-specific and CMV-specific CD8 T cells by multi-parametric flow cytometry at a given time point. Cells were stimulated for 6 hours with the corresponding cognate peptide before intracellular staining; αCD107a was present thoughout the assay to capture degranulating cells as described in the Materials and Methods. Percentages of function⁺tetramer⁺ cells are shown. Plots are gated on CD3⁺CD8⁺ cells. Dead cells were excluded from the analysis using a viability dye.(B) Multi-functional assessment of HIV-specific CD8 T cell responses by multi-parametric flow cytometry performed longitudinally for subject 1. After the gates for 4 functions were created (CD107a, IFN-γ, TNF and IL-2), the Boolean gate platform was used to create an array of 15 different positive combinations. The pies represent the functional profiles of HIV-specific CD8 T cell populations. The slices within each pie represent different functional combinations; 4⁺ (red), 3⁺ (orange), 2⁺ (yellow) and 1⁺ (green). The bars on the x-axis represent the response frequency for each combination. Numbers under the pies indicate the time points studied (m=month). Black, light grey and dark grey indicate pre-HAART or pre-escape, HAART or escape, and post-HAART time points, respectively.(C) Longitudinal flow cytometric analysis of HIV-specific CD8 T cell function for all 8 subjects. (D) Percentage of tetramer⁺ cellsexpressing CD107a, IFN-γ and TNF under conditions of high and low antigen load; the first time point before antigen decay and one time point after antigen decay, respectively, are shown. Ag=antigen.

Figure 4. Evolution of the TCR repertoire in the context of antigen decay.

(A) Longitudinal clonotype frequencies for HIV-specific and CMV-specific CD8 T cell responses. The major persistent clonotypes are represented for each subject, color-coded to match those shown in Supplementary Table 1. Under the x-axis, the black, diagonal line and grey bars indicate pre-HAART or pre-escape, HAART or escape, and post-HAART time points, respectively. (B) Morisita-Horn coefficients for HIV-specific and CMV-specific CD8 T cell populations before and after antigen decay (Ag and NoAg, respectively). The reference similarity for HIV-specific and CMV-specific CD8 T cell repertoireswas generated under the null hypothesis that the Ag and NoAg sets were randomly assigned from the same TCR population (HIV Ref and CMV Ref respectively). The similarity between Ag and NoAg time points was significantly lower than the reference similarities for HIV-specific but not for CMV-specific CD8 T cell repertoires.(C)Morisita-Horn coefficients for HIV-specific CD8 T cell responses comparing the time point before antigen decay (Ag) to either time points with low antigen load (NoAg) or time points after antigen rebound due to cessation of HAART (ReAg) for subjects 1, 2, 3 and 4. The similarity between time points before antigen decay and antigen rebound (Ag/ReAg) was significantly lower than between time points before and after antigen load decrease (Ag/NoAg).

Results

Virological and immunological features of the study population

The clonal structures, functional profiles and phenotypic properties of HIV-specific CD8 T cell populations were studied longitudinally in 8HIV-infected individuals under conditions of bothhigh and low antigen load, as inferred from measurements of plasma viremia and autologous viral epitope sequencing, with low levels of antigenemia reflecting HAART-mediated suppression of HIV replication and/or immune-driven mutational escape. Where feasible, CMV-specific CD8 T cell responses were studied in parallel. Figure 1A depicts the time points studied in each subject and the autologous viral epitope sequences for the specificities studied at each time point. Periods of reduced antigen load, due to effective HAART and/or viral escape, are indicated by diagonal hatching. Supplementary Figure 1 displays the viral load trajectories for each subject together with concomitant CD4 and CD8T cell counts, and all viral epitope sequences. For subjects 1-4, there was no evidence of mutation within any of the targeted epitopes throughout the study period (Figure 1A & Supplementary Figure 1B). For subjects 5-8, autologous plasma viral sequencing revealeddirectional changes in antigenic sequences over time (Figure 1A & Supplementary Figure 1B), consistent with epitope escape through mutation at MHCI anchor positions and/or putative TCR contact residues (subjects 5, 7 and 8). In subject 5, we observed E8K and L9F mutations in the Pol FL9 epitope. In subject 6, 100% of analyzed sequences acquired a K9Q mutation in the Gag p17 RK9 epitope over the course of the study. In subject 7, a dual K7R and K9R mutation was observedin the Gag p17 RK9 epitope. However, at the final time point of the study (26 months), the majority of viral sequences contained the K9R mutation either alone or in combination with additional mutations; this indicates replacement of the intermediate viral quasispecies with an optimal immune escape virus. In subject 8, I3V and R6K mutations emerged in the Gag p24 GI9 epitope at 17 months. Of note, subject 8 initiated HAART after the first time point of study and exhibited an undetectable plasma viral load at 7 months. Representative dualpMHCI tetramer stainings are shown in Figure 1B. Consistent with antigen load decay, we observed a progressive decline of HIV-specific CD8 T cell frequencies in all 8 subjectsafter initiation of HAART and/or the emergence of targeted epitope escape mutations (Figure 1C). In contrast, the frequencies of CMV-specific CD8 T cells remained largely constant(37). Of note, the frequencies of HIV-specific CD8 T cells all increased with antigen rebound after cessation of HAART in subjects 1-4.

Figure 1. Virological and immunological characteristics of the study cohort.

(A) Schematic representation of viremia and antigen sequence variation over time.A color code is assigned for each of the 3 main antigen load categories studied; high antigen load pre-HAART or pre-escape (black), low antigen load on HAART or after viral escape (diagonal lines) and antigen load rebound after cessation of HAART (grey).For subjects 1-4, the diagonal lines represent the administration ofHAART. For subjects 5-7, the diagonal lines represent epitope escape. For subject 8, the diagonal lines represent the combination of HAART and epitope escape.Month 0 indicates the estimated date of infection. Orange circles indicate the time pointsat which blood samples were analyzed for HIV-specific and CMV-specific CD8 T cell responses in each subject. Black arrows indicate the time points from which viral sequences were obtained. The targeted epitopes are depicted in each case; longitudinal epitope sequence variation is also displayed. (B) Representative tetramer co-staining showing the frequencies of HIV-specific and CMV-specific CD8 T cells longitudinally in subject 2. (C) Longitudinal frequencies of HIV-specific and CMV-specific CD8 T cells in all 8 subjects. Bars represent the following epitope-specific CD8 T cell populations: HLA-A*0201-FLGKIWPSHK (HIV Gag p2p7p1p6) and HLA-A*0201-NLVPMVATV (CMV pp65) in subject 1, HLA-A*0301-RLRPGGKKR (HIV Gag p17) and HLA-B*0702-TPRVTGGGAM (CMV pp65) in subject 2, HLA-B*0801-FLKEKGGL (HIV Nef) in subject 3, HLA-B*0702-TPGPGVRYPL (HIV Nef) and HLA-B*0702-TPRVTGGGAM (CMV pp65) in subject 4, HLA-B*0702-FPQGEAREL (HIV Pol) in subject 5, HLA-A*0301-RLRPGGKKK (HIV Gag p17) in subjects 6 and 7 (KKK and RKR represent wild type and variant epitopes in subject 7, respectively), and HLA-B*0801-GEIYKRWII (HIV Gag p24) and HLA-B*0702-TPRVTGGGAM (CMV pp65) in subject 8. The black, diagonal line and grey bars indicate pre-HAART or pre-escape, HAART or escape, and post-HAART time points, respectively.

Antigen load decay leads toincreased expression of CD45RA and CD127and decreased expression of PD-1 on HIV-specific CD8 T cells

Flow cytometric analysis using a panel of cell surface markers allowed us to assess the maturation status (monitored by the expression of CCR7, CD27 and CD45RA; Figure 2A&C), exhaustion profile and survival capacity (quantified by theexpression of CD28, CD127 and PD-1; Figure 2B&D) of HIV-specific CD8 T cell populations longitudinally in 7subjects. The full dataset is shown in Supplementary Figure 2. In agreement with previous results, HIV-specific CD8 T cells generally expressed either a CCR7⁻CD27⁺CD45RA⁻ or a CCR7⁻ CD27⁻CD45RA⁻ phenotype (38). Frequencies of CCR7⁻CD27⁺CD45RA⁻ cells (medium blue) declined as levels of antigen decreased, with the exception of subject 4, while the proportion of CD45RA⁺ cells (dark blue) increased for all subjects.This increase in CD45RA expressioncomprisedboth CCR7⁻CD45RA⁺ cells,reflecting differentiation within the memory compartment, andCCR7⁺CD45RA⁺ cells, a phenotype that typifies naïve or stem-cell memory(Tscm) cells(39). Figure 2E represents the significant increase of HIV-specific CD8 T cells expressing CD45RA between the first time point, where antigen load was high, and the second time point obtained after antigen decline for all subjects studied (P=0.005). Similarly, and concomitantly with the decay in antigen level, the frequency of CD28^+/−CD127⁻PD-1⁺ CD8 T cells (medium pink) decreased in all subjects, while frequencies of CD28^+/−CD127⁺PD-1^+/−CD8 T cells (purple)increased in all subjects except subject 4. The significant increase in CD127 expression and decrease in PD-1 expressionon HIV-specific CD8 T cells between the first time point studied and the second time point after antigen decline is shown in Figure 2F&G (P=0.006 and P=0.011, respectively). In contrast, CMV-specific CD8 T cells displayed a more terminally differentiated phenotype (CCR7⁻CD27⁻CD45RA^−/+); these cells were mostly CD28⁻CD127^+/−PD-1⁻ and did not show consistent phenotypic changes during the study course (Supplementary Figure 2B&D). Collectively, these results indicate that HIV-specific CD8 T cells lose expression of the exhaustion marker PD-1 and up-regulate CD127 expression levels under conditions of reduced antigen load, which could allow them to persist within the memory compartment (40); moreover, the increase in CD45RA expression indicates that decreased levels of antigen lead to enhanced differentiation within the specific CD8 T cell compartment as a whole.

Reduced antigenemia improves the functionality of residual HIV-specific CD8 T cells

To determine if changes in antigen levels, mediated either by intervention with HAART or immune-driven viral escape, led to significant changes in the functional profile of HIV-specific CD8 T cells, we undertook a comprehensive flow cytometric analysis of cytokine production (IFN-γ, TNF and IL-2) and degranulation (CD107a) in all 8subjects (Figure 3 and Supplementary Figure 3). Weused Boolean analysis to create an array of different combinations of T cell functions that allowed the enumeration of cells with various degrees of functionality. Figure 3A shows representative flow cytometric data for CD107a mobilization and IFN-γ, TNF and IL-2 production by HIV-specific or CMV-specific CD8 T cells in the presence of cognate peptide stimulation. We observed dynamic changes in the functionality of HIV-specific CD8 T cells in all subjects studied as levels of cognate antigen decayed (Figure 3B&C); the full dataset is shown in Supplementary Figure 3A. For all 8 subjects, HIV-specific CD8 T cell responses at the early time points of high viremia were predominantly mono- or bi-functional, mainly comprising IFN-γ⁺ and/or CD107a⁺ cells. Interestingly, the functional profile of HIV-specific CD8 T cells for each epitope was characterized by the acquisition of several functions following antigen load decay as shown by a significant increase in the frequencies of cells exhibiting 3 functions (CD107a⁺IFN-γ⁺TNF⁺; orange; P=0.01)and a decay in mono- or bi-functional cells (yellow and green; Figure 3B-D). Of note, after antigen rebound due to cessation of HAART, the frequency of HIV-specific CD8 T cells exhibiting 3 or 4 functions (orange, red) decreased in subjects 1 to 4. No significant changes in functionality were apparent for the corresponding CMV-specific CD8T cell populations (Supplementary Figure 3B). Indeed,CMV-specific CD8 T cells were consistently poly-functional, exhibiting up to 4 (CD107a⁺IFN-γ⁺TNF⁺IL-2⁺), 3 (CD107a⁺IFN-γ⁺TNF⁺) or 2 (IFN-γ⁺TNF⁺ and CD107a⁺IFN-γ⁺) functions.Thus, HIV-specificCD8 T cells display enhanced poly-functional profiles coincident with the decay of antigen, whether due to effective HAART or the emergence of escape mutations, consistent with previous reports (22, 23).

Changes in the HIV-specific CD8 T cell repertoire after antigen decay

To determine if the changes in phenotype and functionality that followed the decay of antigen levels could be attributed to alterations in the TCR repertoire of HIV-specific CD8 T cells, we performed a comprehensive clonotypic analysis of HIV-specific and CMV-specific CD8 T cell populations using a template-switch anchored RT-PCR to amplify all expressed TRB gene products without bias as described previously(34, 41). All TCR sequences from this longitudinal analysis are shown in Supplementary Table 1. Figure 4A summarizes the dynamic evolution of the major clonotypes, color-coded to match the sequences shown in Supplementary Table 1. Dynamic longitudinal changes in clonotypic composition were observed within HIV-specific CD8 T cell populations, concomitant with changes in antigen load. Several clonotypes persisted throughout the duration of the study,while others could no longer be detected. Furthermore, we observed the emergence of new clonotypes during periods of HAART administration and after the emergence of escape mutations. Persistent clonotypes showed quantitative changes as a consequence ofantigen load decay. In subject 1, for example, the TRBV6-2/CASSYVGGDGYT/TRBJ1-2 clonotype, which was the second most frequent clonotype in the pre-HAART repertoire (27.7%), increased in frequency during therapy and became the prevalent clonotype after cessation of HAART and viral load rebound (87.6%). In subjects 6 and 8, an even more dramatic shift was observed, with a monoclonal repertoire emerging after immune escape; the detected TRBV13/CASSPGLDGEQY/TRBJ2-7 clonotype for subject 6 and the TRBV9/CASSTKAGGLADTQY/TRBJ2-3 clonotype for subject 8 were not present at the pre-escape time point. In subject 5, we observed a major bias in the TCR repertoire specific for the HLA-B*0702-FPQGEAREL epitope, with preferential usage of TRBV18 and TRBJ2-5 combined with the presence of a discernable RGR motif in the CDR3 loop. More over, clonotype TRBV18/CASSPRGREETQY/TRBJ2-5, which was a subdominant clonotype at 7 and 21 months (13.4% and 16.1% respectively), became dominant at 31 and 43 months (100% and 98%, respectively). Of note, the dominant clonotype at 7 and 21 months, TRBV18/CASSPRGRDETQY/TRBJ2-5, could no longer be detected within the tetramer⁺ pool at 31 and 43 months. Hence, in the context of the E8K and L9F mutations, we observed the selection of a single clonotype from a heavily biased repertoire that persisted over time. In subject 7, the TCR repertoires specific for the wild type (RLRPGGKKK) and variant (RLRPGGRKR) epitopes were analyzed. Responses to both epitopes were detected and decayed over time (Figure 1C). The TCR repertoire specific for the wild type epitope remained stable. The dominant clonotype specific for the variant epitope that emerged under low levels of antigen at 16 months (TRBV10/CASSDTLNTEAF/TRBJ1-1) was not detected in either repertoire prior to immune escape; however, it was detected as a subdominant clonotype in the wild type epitope-specific repertoire post-escape. Other cross-reactive clonotypes were also detected at 2 and 16 months (TRBV28/CASRDSSYEQY/TRBJ2-7 and TRBV24-1/CATSDDGTPNNEQF/TRBJ2-1), and both repertoires contained mutually exclusive clonotypes. Thus, immune escape through mutation can impact the clonotypic repertoire of wild type epitope-specific CD8 T cell populations.

Figure 4B displays the similarity indices for TCR sequences between time points with high and low antigen load for HIV-specific and CMV-specific CD8 T cell populations calculated using the Morisita-Horn coefficient for all subjects (36). All values were generated in comparison to the first time point, when antigen levels were high. HIV-specific CD8 T cell repertoires exhibited drastic changes as the Morisita-Horn coefficients were lower than 1, except for subject 3. For all 4 subjects in whom CMV-specific responses were detected, the Morisita-Horn coefficient was equal or close to 1, indicating a conserved repertoire throughout the study period.The TCR repertoire similarity between the high and low antigen load time points was significantly lower than the reference similarities for HIV-specific CD8 T cells (P=0.001), suggestingthat the HIV-specific CD8 T cell repertoire population was highly dynamic upon decrease of antigen load. For subjects 1 to 4, the Morisita-Horn coefficients ranged from 1 to 0.56 for the HAART-induced low antigen load time points studied (Figure 4C). After antigen rebound, these coefficients were significantly lower compared to those observed between high and low antigen load states (P=0.04), indicating a further turnover of the HIV-specific TCR repertoire upon antigen rebound (Figure 4C). Collectively, these results indicate that the HIV-specific TCR repertoire is highly dynamic and that fluctuations in antigen levels (decay and rebound) lead to changes in clonal dominance by selection of particular persistent clonotypes and the emergence of new HIV-specific CD8 T cell clonotypes.

Functionally superior HIV-specific CD8 T cells persist under conditions of limited antigenic stimulation

In 3 subjects, we were able todissect the functional characteristics of individual persistent clonotypesbased on the availability of specific TCRVβmAbs. The TRBV25-1/CASSVLRAAF/TRBJ1-1 clonotype dominated the HLA-B*0801-restricted FL8-specific CD8 T cell population in subject 3 at the 1 month (high antigen load, 97.7%) and 8 month(low antigen load on HAART, 100%)time points. This clonotype displayed a change in phenotype between the pre-HAART and HAART time points, predominantly due to increased expression of CD45RA (dark blue) and CD127 (purple), and a decrease in PD-1 expression (medium pink)(Figure 5A). Gain of function between these two time points was also apparent, as evidenced by a substantial increasein the frequencies of CD107a⁺IFN-γ⁺ TNF⁺ (3+) cells, from 1.6% to 33.6% of the responding population (Figure 5A). However, this gain in functionality under conditions of limited antigenic stimulation was not sufficient to maintain the TRBV25-1/CASSVLRAAF/TRBJ1-1 clonotypeafter antigen rebound, as it was replaced by a new clonotype post-HAART (Figure 5A and Supplementary Table 1). Thus, a decrease in antigen load can lead to phenotypic changes and increased functionality at the clonal level.

Figure 5. Gain of function and persistence of individual HIV-specific CD8 T cell clonotypes under conditions of limited antigenic stimulation.

(A)Phenotypic and multi-functional assessment of HLA-B*0801-restricted FL8-specific CD8 T cells from subject 3 before (1 month) and during (8 months) HAART. The dominant TRBV25-1/CASSVLRAAF/TRBJ1-1 clonotype showed changes in phenotype and gained functionality after antigendecay. (B)Simultaneousphenotypic and multi-functional assessment of bulk and TCRVβ6-2⁺ HLA-A*0201-restricted FK10-specific CD8 T cells before and during HAART in subject 1.Tetramer⁺ cells comprising all clonotypes did not display an altered functional profile after antigendecay, whereas epitope-specific TCRVβ6-2⁺ CD8 T cells gained functionality within the same time frame.The experiment shown in this figure was performed at a different timethan the experiment presented in Figure 3 without co-stimulationto compare total tetramer⁺ cells and tetramer⁺TCRVβ6-2⁺ cells. No differences in phenotype were observed between tetramer⁺TCRVβ6-2⁺and tetramer⁺TCRVβ6-2⁻ CD8 T cells at 17 months under HAART.(C)Clonotypic composition of functional subsets within the HLA-B*0701-restricted FL9-specific CD8 T cell response at the 21 month pre-escape time point for subject 5, as determined by DNA-based sequence analysis. Pie colors represent the different functional combinations that were sorted: 3⁺ (orange), 2⁺ (yellow) and 1⁺ (green). The boxes show the CDR3 amino acid sequence, TRBV and TRBJ usage of the different functional cells.At 21 months, the tetramer⁺ CD8 T cell population was composed of one dominant clonotype (blue) and one subdominant clonotype(orange), both expressing TRBV18. All functional cells correspond to the subdominant clonotype at 21 months, which became dominant after antigen decay. No phenotypic differences were observed between functional and non-functional epitope-specific CD8 T cells at 21 months.

The clonotypic and functional analysis from subject 1supported this observation and suggested another mechanism for the gain of function after antigen decay (Figure 5B). ClonotypeTRBV6-2/CASSYVGGDGYT/TRBJ1-2 persisted throughout the study period, although its relative frequency progressively increased such that it became the dominant clonotype during and after HAART (58.7% at 17 months and 87.6% at 20 months, respectively). Functionality was assessed both within the total HLA-A*0201-restricted FK10-specificCD8 T cell population encompassing several clonotypes and within the TCRVβ6-2⁺subset, identified by mAb staining; this analysis was performed at time points that preceded and followed antigen decay (1 month and 17 months, respectively), when the TRBV6-2/CASSYVGGDGYT/TRBJ1-2 clonotype (in blue) was subdominant and dominant respectively (Figure 5B). At 1 month, tetramer⁺CD8 T cells were predominantly IFN-γ⁺. Interestingly, the functional profile of TCRVβ6-2⁺CD8 T cells was similar to that of the total tetramer⁺CD8 T cell population at this time point. After antigen decay on HAART at 17 months, the functional profile of the tetramer⁺CD8 T cell population remained predominantly mono-functional. In contrast, TCRVβ6-2⁺CD8 T cells acquired novel functions, demonstrated by an increase in the proportion of CD107a⁺IFN-γ⁺TNF⁺ (3+) cells. At 17 months, no differences in phenotype were observed between the TCRVβ6-2⁺cells and the total FK10-specific CD8 T cell population (Figure 5B). Thus, fluctuations in antigen load can lead to clonotype-specific functional changes. Furthermore, clonotypes that display enhanced functionality can become numerically dominant under conditions of limited antigenic stimulation.

To extend these observations, experiments were performed to determine if clonotypes with superior functional profiles were endowed with a greater capacity to persist. The clonotypic and functional data from subject 5 allowed us to address this question (Figure 5C). Functional improvements in the HLA-B*0702-restricted FL9-specific CD8 T cell population after antigen decay at 31 and 43 months were described above (Figure 3). The TCR repertoire of the HLA-B*0702-restricted FL9-specific CD8 T cell population was comprised of a single clonotype,TRBV18/CASSPRGREETQY/TRBJ2-5, at these time points. To determine the functional compartment within which this specific clonotype resided at the 21 month time point, before antigen decay, we sorted 4 different functional populations from the HLA-B*0702-restricted FL9-specific CD8 T cell population and performed a DNA-based clonotypic analysis as described previously(35). The following 4 functional CD8 T cell populations were sorted: 3+ (IFN-γ⁺TNF⁺IL-2⁺), 2+ (IFN-γ⁺TNF⁺ and IFN-γ⁺IL2⁺) and 1+ (IFN-γ⁺). The TRBV18/CASSPRGREETQY/TRBJ2-5 clonotype was present in all 4 functional populations sorted; in contrast, the dominant TRBV18/CASSPRGRDETQY/TRBJ2-5 clonotype prior to escape was not detected(Figure 5C). We also performed a phenotypic analysis of FL9-specific CD8 T cells at the 21 month time point with or without stimulation, and compared the tetramer⁺ cells on the basis of IFN-γ production. No significant phenotypic differences were detected between the responding and non-responding cells that could differentiate the TRBV18/CASSPRGREETQY/TRBJ2-5 clonotype (Figure 5C). Thus, functionally superior clonotypes can persist preferentially under conditions of limited antigenic stimulation.

Functionally sensitive HIV-specific CD8 T cell clonotypes persist after antigen decay

HIV-specific CD8 T cells from subjects 5, 7 and 8 (Table 1) were stimulated with the corresponding cognate peptides to determine if increased functional sensitivity contributed to the selective persistence of HIV-specific TCR clonotypes under conditions of limited antigenic stimulation. Intracellular cytokine staining was performed to assess functionality at different peptide concentrations, with the underlying hypothesis being that clonotypes with the highest levels of functional sensitivity would be enriched in the persisting CD8 T cell population. In all cases, the EC50 values for IFN-γ production bytetramer⁺ CD8 T cell populations decreased over time, concomitant with the emergence of escape mutations within the corresponding cognate epitopes. Thus, the functional sensitivity of persistent antigen-specific CD8 T cells increased in the presence of low antigen levels in these 3 subjects, suggesting that this parameter confers individual clonotypes with a selection advantage in vivo in the presence of limited antigen load.

Table 1. Antigen sensitivity in the context of antigen decay.

EC50 values (μg/ml peptide) for IFN-γ production by antigen-specific CD8 T cells are shown at time points before and after antigen decay for subjects 5, 7 and 8.

	Time point	EC50	Time point	EC50
Subject 5	7 months	0.24	31 months	0.12
Subject 7	2 months	0.59	16 months	0.08
Subject 8	1 month	0.6	7 months	0.09

Discussion

In this study, we investigated the impact of declining antigen load on the persistence, phenotypic status andfunctionality of HIV-specific CD8 T cell populations at the clonal level. A marked degree of clonotypic turnover was observed within HIV-specific CD8 T cell populations as a consequence of antigen decay after the initiation of HAART or upon the emergence of viral epitope mutations. Furthermore, new cognate clonotypes emerged under conditions of limited antigen load. In contrast, the contemporaneous CMV-specific CD8 T cell repertoires remained stable. Antigen decay led to changes in the phenotype of HIV-specific CD8 T cells and to the acquisition of novel functions; CMV-specific CD8 T cells remained polyfunctional throughout and maintained their phenotypic profiles. HIV-specific CD8 T cell clonotypes that persisted over time exhibited functional and phenotypic changes that paralleled alterations in antigen load. Moreover, particular clonotypes that became dominant after antigen decay were selected for their higher functional capacities. Overall, these data provide clear evidence of a functional reconstitution within the HIV-specific CD8 T cell compartment upon antigen decay, which can be attributed to two non-exclusive mechanisms: (i) gain of function by persistent clonotypes; and, (ii) selection of clonotypes with high levels of functional sensitivity.

After antigen decay, we observed increases in CD127 and CD45RA expression and decreases in PD-1 expression within HIV-specific CD8 T cell populations, consistent with previous reports (22, 23, 42-45).However, these changes were not drastic; indeed, HIV-specific CD8 T cells remained predominantly CD27⁺. The functional improvement of HIV-specific CD8 T cell populations observed after antigen decay was primarily due to the acquisition of TNF production. Of note, the loss of TNF production has been described as an early functional marker of exhaustion (10). Furthermore, CMV-specific CD8 T cell populations in the present study exhibited similar functional and phenotypic properties throughout all time points studied. Thus, the gain of TNF production is limited to the HIV-specific CD8 T cell compartment and can be attributed to antigen decay.It should be noted that we cannot comment directly on changes in cytolytic activity because CD107 mobilization is an indicator of degranulationand may not reflect the expression of key cytolytic molecules, which show substantial heterogeneity within CD8 T cell populations (46, 47). Nonetheless, it is intriguing to postulate that the improved functionality of wild type epitope-specific CD8 T cells observed after viral escape might not only be the consequence of antigen decay but could also contribute to the generation and maintenance of HIV escape mutants. In addition, althoughthe functional improvementof HIV-specific CD8 T cells was partially sustained after the discontinuation of HAART, this was not associated with improved control of viral replication in the 4 subjects studied in this context.Given that a previous report has suggested such an association in a small number of individuals(48), further work is warranted to determine the conditions under whichthese parameters can be linked.

HIV-specific CD8 T cell repertoires tend to be oligoclonal and skewed during chronic HIV-1 infection, likely as a consequence of several factors including avidity-based selection(41), the loss of high avidity clonotypes (49) and a relative lack of precursor T cell diversity due to decreased thymic output(50). In the present study, we observed that reduced antigenic stimulation resulted in dynamic HIV-specific CD8 T cell repertoire changes, including the emergence of new clonotypes, modifications of clonal dominance and reductions in the overall numbers of constituent clonotypes. Of note, we observed a degree of cross-reactivity between wild type and variant epitopes in subject 7. In this scenario, TCR repertoire alterations were minimal, consistent with a degree of ongoing antigenic drive mediated by the variant epitope. Thus, cross-reactivity can modify the impact of epitope mutation on the wild type antigen-specific CD8 T cell repertoire.

The emergence of new clonotypes after antigen decay was observed after the institution of HAART and following the emergence of viral epitope mutations. However, the origins of these clonotypes remain to be elucidated. One possibility is that these clonotypes exit secondary lymphoid organs as antigen loaddecreases. De novo priming of new clonotypes is also known to occur during persistent viral infections(51), and may even be enhanced by improved CD4 T cell help during HAART and residual low level antigen persistence (52, 53). With the application of deep sequencing approaches to the evaluation of mutational immune escape, it is becoming clear that wild type antigen frequently persists to some extent within complex mixtures of viral variants and it is probable that such viral quasispecies establish an equilibrium with the residual cognate CD8 T cell population by priming new functional clonotypes(54). Finally, the apparent emergence of new clonotypes may simply represent a relative increase in the frequency of previously primed memory CD8 T cell clonotypes that were harbored below the level of detection at earlier time points(55). The application of more sensitive methodologies for TCR repertoire evaluation could help to resolve these issues(56).

Our experiments also allowed us to determine whether the overall restoration of functional capacity within HIV-specific CD8 T cell populations was the consequence of a functional improvement at the single cell level or the preferential survival of poly-functional CD8 T cells. Overall, the data indicate that both mechanisms contribute to the observed functional restoration within HIV-specific CD8 T cell populations under conditions of limited antigenic stimulation. Thus, antigen decay, due either to effective HAART or the emergence of epitope mutations, led to a significant change in the TCR repertoiredue to the selection of clonotypes with high levels of functional sensitivity. In addition, individual clonotypes were shown to undergo functional improvement under the same conditions. Consistent with these observations, Reeset al. showed that antigen load in peptide-immunized mice shaped the CD4 T cell repertoire, and that high antigen load led to the emergence of clonotypes with low levels of functional sensitivity (57). Moreover, it was reported previously that CD8 T cells with high levels of antigen sensitivity exhibited superior effector functions resulting in more efficient antiviral activity compared to CD8 T cells with poor antigen sensitivity(41, 58-63). In addition, Almeida et al. showed thatepitope-specific CD8 T cell clones endowed with high levels of antigen sensitivity displayed superior functionality, proliferated more efficiently and mediated more potent HIV-1 suppressive activity(6, 64). Our data suggest that clonotypes with higher levels of antigen sensitivity could also have a superior capacity to persist in vivo, thereby highlighting the need to define optimal concentrations of antigen for the generation of long-lived memory CD8 T cell clonotypes with maximal antiviral efficacy. Collectively, these findings have implications for vaccination and lend support to immunotherapeutic strategies that aim to induce polyclonal responses under conditions of HAART (65).