Programmed death-1 expression on HIV-1-specific CD8+ T cells is shaped by epitope specificity, T-cell receptor clonotype usage and antigen load

Objectives: Although CD8+ T cells play a critical role in the control of HIV-1 infection, their antiviral efficacy can be limited by antigenic variation and immune exhaustion. The latter phenomenon is characterized by the upregulation of multiple inhibitory receptors, such as programmed death-1 (PD-1), CD244 and lymphocyte activation gene-3 (LAG-3), which modulate the functional capabilities of CD8+ T cells. Design and methods: Here, we used an array of different human leukocyte antigen (HLA)-B∗15 : 03 and HLA-B∗42 : 01 tetramers to characterize inhibitory receptor expression as a function of differentiation on HIV-1-specific CD8+ T-cell populations (n = 128) spanning 11 different epitope targets. Results: Expression levels of PD-1, but not CD244 or LAG-3, varied substantially across epitope specificities both within and between individuals. Differential expression of PD-1 on T-cell receptor (TCR) clonotypes within individual HIV-1-specific CD8+ T-cell populations was also apparent, independent of clonal dominance hierarchies. Positive correlations were detected between PD-1 expression and plasma viral load, which were reinforced by stratification for epitope sequence stability and dictated by effector memory CD8+ T cells. Conclusion: Collectively, these data suggest that PD-1 expression on HIV-1-specific CD8+ T cells tracks antigen load at the level of epitope specificity and TCR clonotype usage. These findings are important because they provide evidence that PD-1 expression levels are influenced by peptide/HLA class I antigen exposure.


Introduction
Persistent viral infections necessitate lifelong host immunity. In this setting, efficacy is mediated predominantly by CD8 þ T cells, which recognize virus-derived peptide epitopes presented on the surface of infected cells by human leukocyte antigen (HLA) class I molecules. Polymorphisms within the HLA class I locus contribute significantly to disease outcome in many such infections, including HIV-1, most likely via effects that intertwine with the quality of the cognate CD8 þ T-cell response [1][2][3].
In addition to the requirement for durable protection and surveillance, there is a concomitant need to limit the pathology associated with continuous immune activation in the face of viral persistence [4]. The systematic upregulation of multiple inhibitory receptors on the surface of antigen-specific CD8 þ T cells provides one such mechanism, which operates via the delivery of negative signals at various stages of the cellular differentiation programme [5,6]. This process is intimately linked with the phenomenon of 'exhaustion,' whereby effector T-cell functions are progressively lost according to a predictable hierarchy [7]. A finely tuned balance between the differentiation-linked acquisition and inhibitory receptor-mediated modulation of functional competence must therefore be achieved, either within a specialized phenotype [8] or within the heterogeneous pool of memory CD8 þ T cells, to ensure an optimal outcome for the host.
Programmed death-1 (PD-1), a member of the CD28/ CTLA-4 family, represents the prototype inhibitory receptor. The importance of PD-1 with respect to CD8 þ T-cell exhaustion was first realized in the lymphocytic choriomeningitis virus (LCMV) model with the demonstration that antibody-mediated blockade reversed effector dysfunction and enhanced viral control [9]. Subsequent studies confirmed the human disease relevance of these findings, describing profound PD-1 upregulation on CD8 þ T-cell populations specific for hepatitis C virus (HCV) [10,11], hepatitis B virus (HBV) [12,13] and HIV-1 [14][15][16][17]. In the latter case, PD-1 expression levels correlated with disease progression [14] and were shown to be reduced by either viral escape or treatment-induced suppression of antigen exposure. These observations strongly suggest that antigen-specific stimulation via the T-cell receptor (TCR) plays an important role in the regulation of PD-1 expression [14,16,18]. Moreover, treatment of simian immunodeficiency virus (SIV)-infected macaques with PD-1blocking antibodies enhanced CD8 þ T-cell immunity and prolonged survival [19].
Recently, it has been shown that PD-1 inhibits CD8 þ T-cell function via upregulation of the basic leucine zipper transcription factor (BATF) [20]. It is also clear that other inhibitory receptors contribute to the functional impairment of CD8 þ T cells in the setting of persistent viral infections, including CD160, CD244, lymphocyte activation gene-3 (LAG-3) and T-cell immunoglobulin and mucin domain-3 (Tim-3) [21][22][23][24][25][26]. Although these molecules offer novel opportunities for therapeutic intervention, a detailed understanding of the corresponding immunobiology will likely be required to inform rational progress in the clinic. In this light, it is established that inhibitory receptor expression profiles are tightly regulated at the transcriptional level [5,6]. Beyond a requirement for antigen exposure per se, however, the environmental cues associated with these transcriptional programmes remain less well defined.
In this study, we set out to determine the factors that govern inhibitory receptor expression on HIV-1-specific CD8 þ T cells during the chronic phase of infection. Our analysis was restricted to two HLA class I molecules, HLA-B Ã 15 : 03 and HLA-B Ã 42 : 01, both of which occur at high frequencies in sub-Saharan Africa and present multiple different HIV-1-derived epitopes. This experimental design enabled controlled comparisons across epitope specificities within and between individuals. The importance of our approach is highlighted by observations of epitope-linked differences in HIV-1-specific CD8 þ T-cell efficacy [1,[27][28][29][30], which can further segregate at the level of individual TCR clonotypes [31,32].

Subjects
Individuals expressing either HLA-B Ã 15 : 03 (n ¼ 15) or HLA-B Ã 42 : 01 (n ¼ 17) were selected from a total cohort of 237 antiretroviral treatment-naïve participants with chronic HIV-1 infection [33,34]. The only additional criterion for selection was the availability of cryopreserved peripheral blood samples. All 32 participants were infected with HIV-1 clade C and harbored virus-specific CD8 þ T-cell responses characterized previously by comprehensive interferon (IFN)g ELISpot screening [28]. For the purposes of this study, 11 different CD8 þ T-cell specificities were considered for detailed evaluation (Table S1, http://links.lww.com/QAD/A536). Informed consent was obtained from all participating individuals, and institutional review boards at the University of Oxford approved the study (E028/99). The use of material from human participants was conducted in accordance with the guidelines of the World Medical Association's Declaration of Helsinki (59th General Assembly).
Human leukocyte antigen class I genotyping and viral load determination Four-digit genotyping of HLA-A, HLA-B and HLA-C alleles was performed using Dynal REALTIME reverse sequence-specific oligonucleotide kits as described previously [28]. Viral loads were determined using the Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Quantification of functional sensitivity (EC 50 )
The peptide concentration required to elicit 50% of the maximum response magnitude [EC 50 (mg/ml)] was determined by IFNg ELISpot analysis [28]. Optimal peptides were used as stimulants and titrated across a concentration gradient of eight logs in 10-fold serial dilutions.
Autologous proviral DNA sequencing Genomic DNA was extracted from PBMCs and amplified by nested PCR using previously published primers [41,42]. The resultant PCR products were purified as described previously [43]. Sequencing was performed using the Big Dye Terminator v3.1 Cycle Sequencing Kit (Life Technologies) [44,45].

Statistical analysis
The Mann-Whitney U test was used to compare median values with respect to the expression of phenotypic markers on bulk and tetramer-positive CD8 þ T cells, both in terms of cell percentages and fluorescence intensities. The Holm-Sidak analysis of variance test was used for multiple comparisons across responses with respect to both parent gate percentage and MFI values. The Wilcoxon signed-rank test was used to compare median values with respect to differences between CD8 þ T-cell memory populations. The Spearman rank test was used to determine correlations between cell percentages with respect to the parent gate and MFI values. Analyses were conducted using GraphPad Prism version 6.0 (GraphPad Software, La Jolla, California, USA). The Student t test was used to calculate differences between CD8 þ T-cell populations specific for FL9-Vpr and other HIV-1-derived epitopes as determined by Boolean gating (SPICE version 4.3).
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Results
Increased programmed death-1 and CD244 expression on HIV-1-specific CD8 R T cells To investigate the expression of exhaustion markers on HIV-1-specific CD8 þ T cells across multiple epitope targets with identical restriction elements, we used four HLA-B Ã 15 : 03 and seven HLA-B Ã 42 : 01 tetramers (Table S1, http://links.lww.com/QAD/A536) to stain PBMC samples directly ex vivo from individuals with chronic untreated HIV-1 clade C infection (n ¼ 15 and n ¼ 17, respectively). Surface expression of the differentiation marker CD57 and three inhibitory markers (PD-1, CD244 and LAG-3), previously shown to be upregulated during chronic viral infections [21,24,25], was determined by polychromatic flow cytometry. The vast majority of HIV-1-specific CD8 þ T cells were found to reside in the PD-1 high , CD244 high and CD57 low compartments ( Fig. 1). Analysis of all HLA-B Ã 15 : 03-restricted (n ¼ 52) and HLA-B Ã 42 : 01-restricted (n ¼ 76) HIV-1-specific CD8 þ T-cell populations revealed that both PD-1 and CD244 were upregulated compared with tetramer-negative bulk CD8 þ T cells (P < 0.0001), whereas CD57 expression was decreased (P ¼ 0.0001) ( Fig. 1a and b). No differences in LAG-3 expression were detected between HIV-1-specific and bulk CD8 þ T cells (Fig. 1b). as cytomegalovirus and Epstein-Barr virus (EBV) [25,38,46]. However, such comparisons ignore potential differences related to the targeted viral proteins or epitopes, even though fine specificity is linked to disparate CD8 þ T-cell-mediated outcomes in HIV-1 infection [2]. To seek evidence of differential epitopelinked exhaustion, we first examined the expression of PD-1, CD57 and CD127 on CD8 þ T-cell populations specific for distinct HIV-1-derived epitopes (n ¼ 4) restricted by HLA-B Ã 15 : 03 (Table S1, http:// links.lww.com/QAD/A536). Substantial differences were apparent across epitope specificities, most notably with respect to PD-1 expression ( Fig. 2a and b). In particular, CD8 þ T-cell populations specific for VF9-p24, previously associated with effective immune control of HIV-1 replication [28], expressed high levels of PD-1 and low levels of CD127. The converse applied to CD8 þ T-cell populations specific for FY10-Tat, which are not associated with protection [28]. These epitope-specific phenotypic differences were also apparent in terms of percentage expression frequencies ( Fig. S1a and c, http:// links.lww.com/QAD/A536). The expression of CD57 was more consistent across epitope specificities ( Fig. 2a  and b, Fig S1b,    Adjusted P values (P < 0.05) for multiple comparisons in (b) and (d) were calculated using the Holm-Sidak analysis of variance test. Significant differences with respect to bulk CD8 þ T cells are not shown.
phenotypic differences related to CD244 and LAG-3 suggests that distinct mechanisms may be driving the differential expression of these exhaustion markers.
A recent study found a positive correlation between TCR avidity and PD-1 expression [47], thereby providing a potential explanation for differential epitope-linked phenotypes. To test this possibility, we examined the functional sensitivity of HIV-1-specific CD8 þ T-cell responses in IFNg ELISpot assays conducted directly ex vivo using sample-matched PBMCs (Fig S3a-

Programmed death-1 expression on HIV-1specific CD8 R T cells is a measure of antigen load
A previous study demonstrated that different epitopespecific CD8 þ T-cell populations in the same individual expressed different levels of PD-1 [14]. However, the basis for such disparities was not fully elucidated. To pursue this line of investigation, we analyzed PD-1 expression at a given time point in an individual with CD8 þ T-cell responses directed against five different epitopes derived from four different HIV-1 proteins restricted by two different HLA-B molecules (Fig. 3a). The PD-1 high population varied from 86% (FL9-Vpr) to 37% (TL9-p24) of tetramer-positive CD8 þ T cells. In contrast, CD244 expression exceeded 96% for all five CD8 þ T-cell populations. Furthermore, we found distinct patterns of PD-1 expression across different HIV-1-derived epitope-specific CD8 þ T-cell populations in participants with different levels of viremia (Fig. 3b). These epitope-linked differences within and between samples applied to each of 33 participants analyzed in a similar manner (data not shown).
Different T-cell receptor clonotypes within individual HIV-1-specific CD8 R T-cell populations express different levels of programmed death-1 To dissect inhibitory receptor expression patterns in more detail, we examined the phenotypic profiles of individual clonotypes within HIV-1-specific CD8 þ T-cell populations. Initially, we determined the clonotypic composition of HLA-B Ã 42 : 01 TL9-p24 tetramer-positive CD8 þ T cells in a single participant (Fig. 4a). The corresponding aTCRVb mAbs were then used to identify clonotypic subsets within the epitope-specific CD8 þ T-cell population, enabling their phenotypic characterization by flow cytometry (Fig. 4b). Clear interclonotypic differences were apparent with respect to PD-1 expression; in contrast, no such differences were detected for CD244 (Fig. 4c). Similar patterns were observed for a further five HIV-1-specific CD8 þ T-cell populations targeting four distinct epitopes across five different participants (Fig. 4d-f). However, there was no correlation between clonotypic dominance and PD-1 expression (Fig. 4g) [47].
Programmed death-1 expression on HIV-1specific CD8 R T cells is driven by the effector memory population It has been shown previously that PD-1 expression is enriched within the effector memory population of HIV-1-specific CD8 þ T cells [46] and differentially expressed with CD57 [48]. To investigate differentiation-linked expression of PD-1 in our cohort, we pregated on PD-1 high and CD57 high populations, then compared tetramer-positive HIV-1-specific CD8 þ T cells and tetramernegative bulk CD8 þ T cells with respect to memory phenotype (Fig. 5a). Both PD-1 high and CD57 high populations were enriched in the T EM compartment (Fig. 5b and c), and differentially expressed on T EM (CCR7 À /CD45RA À ) and terminally differentiated T EMRA (CCR7 À /CD45RA þ ) cells (Fig. 5d and e). Furthermore, PD-1 high CD8 þ T cells across all HIV-1-derived epitope specificities resided predominantly in the effector memory (CCR7 low /CD45RA low ) rather than the T EMRA (CCR7 low /CD45RA high ) pool ( Fig. 5f and g), thereby supporting the hypothesis that PD-1 expression is driven by repetitive antigen exposure [49].

Discussion
In this study, we conducted a detailed analysis of inhibitory receptor expression across a large number of different HIV-1-derived epitope-specific CD8 þ T-cell populations restricted by either HLA-B Ã 15 : 03 or HLA-B Ã 42 : 01 to inform our understanding of the processes that regulate and potentially compromise antiviral immune responses. Major differences were observed for PD-1 expression levels across epitope specificities both within and between individuals. Differential interclonotypic expression of PD-1 was also apparent within individual HIV-1-specific CD8 þ T-cell populations. Positive correlations were detected between PD-1 expression and plasma viral load, which were reinforced by stratification for epitope sequence stability and dictated by effector memory CD8 þ T cells. Thus, PD-1 expression on HIV-1-specific CD8 þ T cells is shaped by epitope specificity as a function of differentiation and driven by antigen load.
Our finding that PD-1 expression is increased on HIV-1specific CD8 þ T cells confirms previous studies [14,16] and concurs with similar observations in other persistent viral infections, including LCMV [9] and SIV [50]. However, it is established that multiple inhibitory receptors beyond PD-1 are involved in the negative regulation of CD8 þ T-cell immunity [21]. Accordingly, we examined CD244 and LAG-3 expression in parallel. Consistent with a recent study [25], CD244 expression levels were increased on HIV-1-specific CD8 þ T cells. In contrast, we found no evidence for elevated LAG-3 expression.
In addition to profound upregulation on HIV-1-specific CD8 þ T cells as a whole, PD-1 expression also exhibited substantial differences between epitope specificities. The large number of HIV-1-derived epitopes restricted by only two HLA-B molecules examined in our study provides a distinct advantage over previous reports [16,17,47,51] in that it enables intraindividual comparisons between epitope specificities with an inbuilt control for the restriction element. In this setting, concomitant epitope-linked differences in CD244 and LAG-3 expression were not detected. These findings indicate that PD-1 expression is regulated more stringently in an antigen-dependent manner.
Previous reports have suggested that TCR avidity is linked to PD-1 expression as a function of signal strength [47]. In functional assays across a subset of matched samples, we found no evidence to support this hypothesis [52]. However, this finding is subject to one important caveat. Specifically, the measurement of functional sensitivity in this setting does not necessarily act as a reliable surrogate for TCR avidity because cytokine output cannot be assumed as a constant. Indeed, PD-1 expression preferentially inhibits poorly sensitive functions, such as IFNg, thereby skewing functional assays toward an inverse relationship.
One important aspect of our study is the identification of multiple different HIV-1-derived epitope-specific CD8 þ Differential PD-1 expression on HIV-1-specific CD8 R T-cells Kløverpris et al. 2015 High epitope sequence variability   T-cell populations within individual samples, thereby providing intrinsic controls for interparticipant variables, such as plasma viral load. Accordingly, we could link differences in PD-1 expression to individual epitopespecific CD8 þ T-cell populations in the absence of important confounders. The large variations in PD-1 expression between different epitope specificities were not limited to participants with high levels of viremia, but applied across the spectrum of plasma viral loads detected in the present cohort. This finding suggests that epitopespecific effects on PD-1 expression levels are paramount. Similar data have recently been reported for different EBV-specific CD8 þ T-cell populations restricted by HLA-A Ã 02 : 01 [53].
Strikingly, PD-1 expression levels also differed between TCR clonotypes within HIV-1-specific CD8 þ T-cell populations. However, we found no relationship with clonal dominance, in contrast to a previous report [47]. Nonetheless, this observation is consistent with the lack of correlation between response magnitude and PD-1 expression in our cohort. A recent study demonstrated that PD-1 expression on CD8 þ T cells is maintained by a mechanism of high production and high clearance [54]. Accordingly, if high avidity clonotypes preferentially acquire PD-1 expression, they may fail to dominate despite the operation of avidity-based selection [39]. These considerations add another layer of complexity to the array of forces that govern the clonotypic architecture of antigen-specific CD8 þ T-cell populations. It remains to be determined whether interclonotypic differences in PD-1 expression are linked to differential antiviral activity [31,32,55,56].
The correlation between plasma viral load and PD-1 expression across all HIV-1-specific CD8 þ T-cell Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
populations is consistent with previous studies [14,17,25]. Moreover, this correlation was strengthened by stratification for targeted epitopes that undergo mutation less frequently, whereas no such association was detected for CD8 þ T-cell populations directed against more variable epitopes. These findings suggest that PD-1 expression is regulated in an antigen-dependent manner, although the effect of viral load in this regard may still be indirect.
Longitudinal studies demonstrating reduced PD-1 expression after mutational escape [18] and treatmentinduced suppression of viral antigen load [25,51], as well as increased PD-1 expression after treatment interruption [52], support this interpretation. Nonetheless, multiple host-specific and virus-specific factors undoubtedly affect epitope processing and presentation [57], confounding any simplistic relationships between the parameters measured in this study. The previously reported inverse correlation between plasma viral load and PD-1 expression on HIV-1-specific CD8 þ T cells during acute infection likely represents a separate scenario occurring before the onset of functional exhaustion [58], although the PD-1 promoter remains unmethylated and active throughout the course of infection [59].
Collectively, our data show that PD-1 expression on HIV-1-specific CD8 þ T cells is tightly linked to epitope specificity and TCR clonotype usage regardless of plasma viral load during the chronic phase of infection. The observation that PD-1 levels are most strongly correlated with antigenemia stratified for epitope sequence stability