Consistent with prior reports, our analyses suggest that PD-1 expression is driven in large part, but not solely, by viral antigen expression. Moreover, PD-1 expression alone does not reflect the functional capacity of CD8+ T cells. Factors other than PD-1 appear to participate in: 1) modulating EBV-specific CD8+ T cell proliferative capacity during both AIM and convalescence, 2) determining states of differentiation and activation, 3) and the degree of contraction of EBV-specific CD8+ T cells in convalescence. These observations serve to refine and extend our understanding of the complex network of factors that govern immune responses and tolerance. To our knowledge, this is the first demonstration that PD-1 expression on virus-specific CD8+ T cells is influenced significantly by factors intrinsic to the responding cell that segregate with epitope specificity and V-beta usage. While these factors have yet to be defined, these observations would be consistent with variable TCR binding and signaling characteristics that may play a significant role in shaping effective immune responses. Price et al have shown that immunodominant clonotypes of EBV and CMV specific CD8+ T cells tend to have greater avidity for their cognate antigen 
. While we have not shown a relationship between PD-1 and the frequency of V-beta usage (i.e. immunodominance), it remains possible that avidity relates to PD-1 expression.
Previous studies have emphasized a relationship between high levels of antigenic stimulation and high PD-1 expression levels on virus-specific CD8+ T cells in chronic viral infections. In a murine model of LCMV infection, graded levels of PD-1 expression on virus-specific CD8+ T cells correlated with viral clearance rates at day 30 after infection 
. The highest levels of PD-1 expression were found on exhausted virus-specific CD8+ T cells in animals with chronically elevated LCMV titers. Whereas PD-1 expression on virus-specific CD8+ T cells in untreated HIV-1 (or SIV in animal models of AIDS) infection is high and is directly related to viral replication 
, PD-1 expression on CMV-specific CD8+ T cells is reportedly lower; presumably due to intermittent or lower levels of viral antigen expression. Relatively high levels of PD-1 expression on EBV-specific CD8+ T cells from chronically infected individuals have been reported by a number of groups 
. Most studies that have compared PD-1 expression on EBV vs. HIV or CMV specific CD8+ T cells have focused on A2-BMLF-1 CD8+ T cells 
. Our results are in agreement with these previous reports, however, we have also demonstrated consistent and significant differences in PD-1 expression among different epitope specific CD8+ T cell subsets.
The concept that circulating virus-specific CD8+ T cells with different epitope specificities may display consistent differences in PD-1 expression has precedent in the literature. Using a broad panel of pMHCI tetramers to characterize HIV-specific CD8+ T cell responses in infected individuals, Day and colleagues observed lower levels of PD-1 expression on certain epitope-specific CD8+ T cells among a majority that expressed PD-1 at high levels 
. A dramatic difference in PD-1 expression on virus-specific CD8+ T cells with different epitope specificities has also been reported in an individual with resolved HCV infection 
. Our data demonstrated consistent differences in PD-1 expression on A2-BMLF-1 and A2-BRLF-1 CD8+ T cells over the course of infection from AIM to convalescence, and between different individuals. This suggests a mechanism that relates to differences in TCR signaling that are shared by HLA-A*0201 positive individuals responding to these epitopes. Possible mechanisms include: 1) differences in the stability of the peptide/MHC-1 association, or 2) a common limitation on TCR signaling (perhaps defined by the avidity of the TCR for the pMHCI complex). Our data show that exposure of epitope-specific CD8+ T cells to high (presumably saturating) peptide concentrations result in different degrees of PD-1 upregulation, consistent with PD-1 expression ex vivo. This is unlikely to be a function of limiting concentrations of pMHCI complexes on the antigen presenting cells. MHC-peptide binding algorithms actually predict that the A2-BRLF-1 epitope has a longer half-life of dissociation from HLA-A*0201 than the A2-BMLF-1 epitope. If antigen presentation were the most critical factor, the less stable pMHCI complex might be expected to be associated with lower PD-1 expression. Therefore, it is more likely that there is a determinant of PD-1 upregulation specific to the cells responding to these particular epitopes. The observation that PD-1 expression may vary significantly in different V-beta subsets that recognize the same epitope suggests that V-beta usage itself may influence PD-1 expression. Further studies are necessary to better define determinants of PD-1 expression on virus-specific CD8+ T cells and their functional consequences.
Given our current understanding of EBV replication, it is likely that different EBV lytic antigens are presented to CD8+ T cells in a similar microenvironment, and in a similar time frame. The differences in PD-1 expression on A2-BMLF-1 and A2-BRLF-1 CD8+ T cells, despite highly similar differentiation phenotypes, suggests a dissociation between signaling pathways that determine differentiation states and PD-1 expression. We have examined the relationship between PD-1 expression and CD127 in detail because a recent report indicates that CD127 expression is tightly linked with proliferative capacity, a function that is lost early as PD-1 is upregulated 
. Furthermore, upregulation of both PD-1 and CD127 has been reported to precede disappearance of an EBV-specific CD8+ T cell from the circulation 
. We found that few CD8+ T cells expressing high levels of PD-1 also expressed CD127 in AIM, but this relationship was not exclusive; higher frequencies of CD127+, PD-1+ cells were seen in convalescence. As reported by Salisch et al, PD-1 expression on virus-specific CD8+ T cells is a sensitive indicator of lentiviral replication 
. Kasprowicz et al have also suggested that PD-1 staining is compatible with it being an activation marker 
. Our analyses are consistent with these interpretations. While PD-1 levels correlated with the contraction of CD8+ T cells, we did not find a simple association between higher PD-1 levels on EBV-specific CD8+ T cells and the degree of contraction from AIM to convalescence. Within a particular epitope specific subset (e.g. A2-BRLF-1 CD8+ T cells), there can be a significant correlation, however between subsets, similar levels of contraction occur with significantly different PD-1 expression (, right panel). This suggests that PD-1 may be necessary, but not sufficient in this process.
Sauce et al 
have described transient increases in PD-1 expression on epitope-specific CD8+ T cells that are associated with an “accelerated” expression of CD127 and subsequent disappearance of that epitope-specific response from the circulation. Consistent differences in PD-1 staining intensity were not reported in their analysis of A2-BMLF-1 (labeled “GLC”) than A2-BRLF-1 (labeled “YVL”) CD8+ T cells in acutely infected individuals at all time points to convalescence. However, careful inspection of the data presented actually demonstrates higher PD-1 expression on A2-BMLF-1 than A2-BRLF-1 CD8+ T cells in three of four individuals at multiple time points, and in the fourth subject during acute infection. Differences in PD-1 expression may have been less evident due to the binding characteristics of the anti-PD-1 antibody used. Indeed, we had collected additional earlier data using a different anti-PD-1 antibody that failed to clearly distinguish positive and negative populations, yet still demonstrated these differences in PD-1 expression at all time points (data not shown). Using another antibody (Medarex), we observed PD-1 expression profiles with clearly defined bimodal distributions (data reported here in , , , ). The ability to distinguish differences in intensity of PD-1 expression, particularly PD-1high
expression, has been previously shown to be critical for describing meaningful biological differences.
The functional impact of PD-1 upregulation on EBV-specific CD8+ T cells during AIM is unclear. We found virtually no proliferation to peptide stimulation during AIM, and no enhancement of proliferation in the presence of antibody blocking the PD-1/PD-L1 interaction. While PD-1 upregulation may contribute to loss of responsiveness, it certainly cannot explain it in entirety. Our findings are consistent with other reports that PD-1 expression alone does not explain loss of proliferative capacity or sensitivity to apoptosis 
. In convalescence, the inhibitory activity associated with PD-1 is more evident; enhanced proliferative responses are observed in the presence of anti-PD-L1. Our observations thus highlight the complexities of PD-1 expression and its functional correlates; in some settings (e.g. chronic LCMV infection in mice), its expression is consistent with cellular exhaustion 
, while in others (e.g. acute hepatitis C infection), it appears primarily to be a marker of activation 
. Blackburn et al have shown that there may be multiple inhibitory molecules that contribute to an “exhausted” phenotype 
, and that PD-1/PD-L1 blockade primarily rescues proliferation in less terminally differentiated cells 
. We have now shown that these complexities must be viewed through a lens that accounts not only for antigen expression, but also epitope specificity and V-beta usage.
In aggregate, studies indicate that PD-1 expression and signaling play an important role in limiting pathogenic consequences of chronic virus infection. Efforts are underway to determine if disrupting the interaction of PD-1 with its ligands will have a beneficial effect in chronic viral infections in which cellular immunity fails to provide control. We have shown that PD-1 expression is induced by antigenic stimulation, but is dependent in part on intrinsic properties of the responding cell. Consistent differences, and significant variability in PD-1 expression observed in two lytic epitope specific CD8+ T cell subsets, and subpopulations defined by V-beta usage point to properties that are likely linked to TCR signaling, TCR avidity and recruitment of molecules involved in signal transduction. Properties of TCR-pMHCI interactions that influence expression of PD-1 are likely to inform the future designs of vaccines and immunotherapeutics.