High-Programmed Death-1 Levels on Hepatitis C Virus-Specific T Cells during Acute Infection Are Associated with Viral Persistence and Require Preservation of Cognate Antigen during Chronic Infection

Study subjects

Blood samples were obtained from consenting HCV-infected adults participating in a prospective study of young intravenous drug users in Baltimore, MD, as previously described (34). Table I⇓ shows the demographics of subjects who were studied. At each visit, participants were provided counseling to reduce the risks of drug use. Blood was drawn for the isolation of serum, plasma, and PBMC in a protocol designed for monthly follow-up. Serum and plasma were stored at −80°C, PBMC were isolated by Ficoll-Hypaque (GE Healthcare Bio-Sciences) gradient centrifugation and cryopreserved in liquid nitrogen.

HCV RNA levels

HCV RNA levels were determined using the quantitative RT-PCR assay (COBAS AMPLICOR HCV Monitor version 2.0 or COBAS TaqMan HCV Test (Roche Molecular Systems) according to the manufacturer’s instructions. The COBAS AMPLICOR HCV and COBAS TaqMan assays have a lower limit of quantitation of 2.8 log₁₀ and 1.7 log₁₀ IU/ml, respectively. HCV clearance was defined as the presence of anti-HCV Ab with HCV RNA undetectable by the COBAS TaqMan HCV Test in serum or plasma specimens from ≥2 consecutive visits obtained at least 300 days after initial detection of viremia. Persistence was defined by detection of HCV RNA by either COBAS HCV assay in serum or plasma specimens obtained at least 300 days after initial viremia.

Multimers

Allophycocyanin-labeled multimers were obtained through ProImmuneand Beckman Coulter Immunomics. They comprised a panel of nine HCV multimers: A2 (KLVALGINAV_{NS3:1406–1415}, ALYDVVTKL_{NS5:2594–2602}, RLWHYPCTV_{E23:614–622}, CINGVCWTV_{NS3:1073–1081}, VLSDFKTWL_{NS4:1987–1995}); A01 (ATDALMTGY_{NS3:1435–1443}); B7 (GPRLGVRAT_Core:41–49); B35 (HPALVFDIT_{NS2:881–889}, HPNIEEVAL_{NS3:1359–1357}). Eight CMV, EBV, and influenza (Flu) multimers were used as controls and were included in the evaluation. These consisted of A01-CMV (VTEHDTLLY_{pp50:245–253}), A2-CMV (NLVPMVATV_{pp65:495–504}), B7-CMV (TPRVTGGGAM_{pp65:417–426}), A1-Flu (CTELKLSDY_{NP 44–52}), A2-Flu (GILGFVFTL_{MP 58–66}), A2-EBV (GLCTLVAML_{BMLF-1 259–267}, CLGGLLTMV_{LMP-2 426–434}), and B8-EBV (RAKFKQLL_{BZLZ-1:190–197}).

Phenotypic analysis

PBMC were thawed and rested at 37°C, 5% CO₂ for at least 5 h; 1–2 × 10⁶ cells were resuspended in 50 μl/well of 10% human serum (Gemini Bio-Products), 10% FCS (HyClone). 50 μg of human IgG (Jackson ImmunoResearch Laboratories) was added; the cells were incubated at room temperature for 10 min, 5 μl of Ag-specific multimer were added, and the incubation was continued for a further 20 min. Either biotinylated PD-1 (anti-human PD-1.5) or IgG4 isotype control (Medarex) was added at a final concentration of 5 μg/ml, and the cells were incubated for an additional 20 min at room temperature (35). PBMC were washed twice with FACS buffer (1% BSA, 0.1% NaN₃, PBS) and resuspended in 50 μl of 10% human serum (Gemini Bio-Products), 10% FCS. To this, streptavidin-PE (10 μg/ml; eBioscience) and CD8-PerCP Cy5.5 (BD Pharmingen) were added, and PBMCs were incubated at room temperature for 20 min. PBMCs were washed twice with FACS buffer and resuspended in 200 μl of 2% paraformaldehyde (J. T. Baker) and analyzed on a FACSCanto flow cytometer within 24 h. At least 50,000–100,000 cells were collected in the lymphocyte gate, from which CD8⁺ cells were isolated. Subsequently, CD8⁺multimer⁺ cells were selected, and PD-1 expression was measured in the gated total CD8⁺multimer⁺ T cells.

Viral sequencing

From 140 to 280 μl of serum or plasma, the 5.2-kb region from the 5′-untranslated region to the NS3-NS4A junction was amplified as previously described (36). For each specimen, 40 clones were isolated and stored, and the first 20 were selected for preliminary clonal analysis. Preliminary clonal analysis was performed by sequencing and phylogenetic analysis of a 450-nt region spanning the E1-E2 junction including hypervariable region 1. Sequence contigs were assembled and aligned using Aligner version 2 (CodonCode) and trimmed to equal length using BioEdit (37). The general time-reversible (GTR) + I + G analytical model (parameters available on request from the authors) was selected using the Akaiki criterion as implemented in ModelTest version 3.7 (38) and PAUP* version 4b10 (Sinauer Associates). Phylogenetic trees were estimated using the neighbor-joining algorithm implemented in PAUP*. For each specimen, the sequence closest to the phylogenetic center of the tree was identified as previously described (39) and sequenced. These 5.2-kb contigs were assembled, aligned, and edited as described in this paragraph. Because longitudinal sequencing provided validation of conserved clonal sequences, only novel epitope substitutions required validation by sequencing additional stored cDNA clones.

Assessment of potential escape mutations

Peptides originating from the HCV H77 genome (genotype 1a) were used to assess T cell function by the ELISPOT assay. Amino acid replacements in epitopes were tested as potential immune escape substitutions by synthesizing each of the variant peptides and testing them in serial dilution as Ags for bulk PBMC and T cell lines raised to original and mutated versions of the epitope in an IFN-γ ELISPOT assay. These methods have been described in detail previously (11). The observed amino acid replacements were classified as escape substitutions if they reduced autologous T cell recognition of the epitope in which it was located, as measured by at least a 2-fold reduction in SFC at two or more peptide dilutions.

Statistical analysis

Linear regression was used to examine the relationship between the covariates and the level of PD-1. Generalized estimating equations methods were used to account for the multiple observations per participant (40). Analysis was performed using SAS version 9. Differences were considered significant if p values were <0.05.

Characteristics of study populations

The demographic characteristics of the 20 subjects studied are listed in Table I⇑. Subjects were selected on the basis of expression of HLA alleles for which multimers were available as well as on the basis of having previously demonstrated peptide-specific responses by IFN-γ ELISPOT. Subjects with HIV infection were excluded. Using peptides representing T cell epitopes identified previously, HLA-matched MHC multimers with peptides recognized by the subject were used to assess PD-1 expression on HCV-specific CD8 T cells obtained from the subjects at time points from initial detection of CD8 T cell responses through the first 4 years after infection. As controls, we performed multimer-Ab costaining for EBV-, CMV-, and Flu-specific T cells. Statistical analysis of the relationship between expression of PD-1 with gender, age, HCV RNA levels, status of Ag (intact or mutated), and virological outcome of infection was then performed. PD-1 expression on HCV-specific T cells was essentially uniformly positive and generally unimodal (Fig. 1⇓). Therefore, we used mean levels of PD-1 expression (mean fluorescence intensity; MFI) on the entire multimer⁺ population of T cells rather than percentage positivity, as an unbiased quantitative estimate of PD-1 expression. Other studies of PD-1 have also demonstrated that the relative expression level of PD-1 is probably the more useful marker than is the presence or absence of PD-1 as measured by percentage (41).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

PD-1 staining patterns for two representative subjects. PD-1 expression on HCV-specific T cells was essentially uniformly positive and generally unimodal. Therefore, the mean levels of PD-1 expression (MFI) on the entire multimer⁺ population of T cells rather than percentage positivity was used as an unbiased estimate of PD-1 expression. The numbers in the top boxes indicate the percentage of isotype control (top row) or PD-1⁺ (bottom row) CD8⁺ HCV-specific T cells. HIgG4, Human IgG4. PE-A, PE area; APC-A, APC area.

Statistical analysis for correlations with PD-1 level

Given that we had different numbers of subjects in each outcome group and they were sampled at varying numbers of time points after infection, we chose the generalized estimating equations as our statistical method. Using this method, which accounts for different numbers of subjects sampled at different numbers of time points, we determined that PD-1 levels, as measured by the MFI on HCV-specific T cells, correlate positively with HCV RNA levels but not with subject gender or age (Table II⇓). Race could not be considered given that all study subjects were Caucasian. When the levels of PD-1 were compared among subjects based on outcome of infection, they were significantly higher on HCV-specific T cells in subjects with persistent infection than on T cells from subjects who cleared infection, both in the first 180 days of infection and using all time points assessed. In a multivariate model, both log₁₀ HCV RNA and clearance were significantly associated with the PD1 level. When the model was changed so that clearance was outcome, PD-1 levels were again associated with clearance (p = 0.0007). Due to the limited variability of HCV RNA in the clearance group a model which included HCV RNA could not be tested. Earlier studies on PD-1 expression in chronic viral infections have not assessed whether the difference in T cell PD-1 expression between subjects who control HIV or HCV and those with progressive HIV or chronic HCV infection is based on the level of virus even though our study and others have shown that levels of virus (HIV or HCV) and PD-1 expression on virus-specific T cells are positively correlated. Given that HCV RNA levels are higher in those with persistent infection than in those who have cleared infection, we analyzed this association controlling for HCV RNA levels and found that PD-1 levels are independently associated with outcome of infection. In fact, multivariate analysis revealed that in the first 180 days of infection, PD-1 levels were correlated only with outcome of infection, not with HCV RNA levels. These data are most compatible with the notion that PD-1 expression or a factor that regulates it has causal effects on outcome of infection rather than that PD-1 expression simply represents a marker for presence and levels of Ag.

PD-1 levels on HCV-specific T cells vs the total CD8 population T cell population

PD-1 is significantly up-regulated on HCV-specific T cells vs the total CD8 T cell population at all stages of infection regardless of the outcome of HCV infection (Fig. 2⇓). Given that the level of PD-1 remains high on the cell surface of T cells specific for recognized HCV Ags relative to the level on the general CD8 T cell population long after control of HCV infection, continued recognition of Ag does not seem to be required to maintain PD-1 levels above that of the general CD8 population (Fig. 2⇓b).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

PD-1 expression on HCV-specific T cells is higher than on Flu-specific T cells and the general CD8 T cell population regardless of outcome of infection or the phase of infection. The MFI of PD-1 on the T cell surface was compared among HCV-specific T cells from subjects with either outcome of infection, control-specific T cells, and the general CD8 population in the first 180 days of HCV infection (HCV⁺ acute to chronic or cleared (a) and at time points after 180 days of HCV infection (b, HCV⁺ chronic or cleared). The level of PD-1 expression is also higher in acute infection on HCV-specific T cells from subjects who remain persistently infected than those who clear HCV infection. Each data point represents the MFI of the PD-1 level on a specific tetramer⁺ population within an individual.

PD-1 levels on HCV-specific T cells in subjects who progress to chronic infection vs those who clear virus

PD-1 is significantly up-regulated on HCV-specific T cells vs T cells specific for influenza from the acute phase through chronic infection in individuals who progress to chronic HCV infection (Fig. 2⇑). PD-1 levels are also significantly higher on HCV-specific T cells than on T cells specific for EBV and CMV in the first 6 mo of HCV infection but not in later stages of HCV infection in subjects who progress to chronic infection (Fig. 2⇑). This suggests a change in PD-1 modulation on HCV-specific T cells in the later stages of infection relative to EBV- or CMV-specific T cells and that regulation of PD-1 might differ in the acute and chronic phases of infection.

There were not enough data points in the first 6 mo of infection for control (EBV, CMV or Flu) Ag-specific T cells from subjects who cleared HCV infection to compare them with HCV-specific responses in the early phase of infection. However, the level of PD-1 on HCV-specific T cells for subjects who cleared infection was significantly higher than on Flu-specific but not other control Ag-specific T cells after the first 6 mo of infection (Fig. 2⇑b). HCV infection is characterized by a longer duration of viremia than is Flu, and PD-1 may be regulated in part by duration of viremia or another difference between Flu and cleared HCV infection. Significantly, the level of PD-1 on HCV-specific T cells in the first 6 months of infection for subjects who cleared infection was lower than on HCV-specific T cells from subjects who failed to clear HCV infection (Fig. 2⇑a; p < 0.0001). Smaller differences were observed between these two groups when PD-1 levels were assessed beyond 6 mo after initial infection (Fig. 2⇑b; p = 0.06). The range of PD-1 expression on HCV-specific T cells in the chronic phase of HCV infection is broad with some levels <100 and others exceeding 2000. However, the bulk of the data points for the HCV⁺ chronic population in Fig. 2⇑b falls in the lower half of PD-1 MFIs measured, resulting in an insignificant difference between the two groups >6 mo after infection. This further suggests that regulation of PD-1 may differ between the acute and chronic phases of infection. We therefore further analyzed the association between PD-1 levels, infection outcome, and HCV RNA levels in early and later infection.

PD-1 levels on HCV-specific T cells stratified by HCV RNA levels and outcome of infection

As shown in Table II⇑ and Fig. 2⇑, the level of PD-1 is significantly higher in early infection on HCV-specific T cells from those who fail to control infection vs T cells from those who clear HCV infection. Despite high HCV RNA levels being generally associated with high PD-1 levels, the levels of PD-1 on HCV-specific T cells ranged widely at all HCV RNA levels both early and late in infection (Fig. 3⇓). We observed that some HCV-specific T cells from subjects with chronic infection expressed low PD-1 levels in the setting of high circulating HCV RNA levels (Fig. 3⇓). Low levels of PD-1 have also been observed on some HIV-specific T cells with high HIV RNA levels (28). We hypothesized that T cells with low levels of PD-1 expression in the setting of high viral RNA levels recognize epitopes that have mutated to escape the immune response. Viral amino acid substitutions in an epitope that abrogate the generation of peptide-MHC complexes on the cell surface result in functional Ag loss even in the setting of high HCV RNA levels. If intact Ag is required for a T cell to receive the signals that induce PD-1 up-regulation, T cells specific for HCV epitopes that have escaped via impaired presentation on APCs would be expected to have low levels of PD-1 expression.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

PD-1 expression on HCV-specific T cells stratified by HCV RNA level during early and late infection. HCV RNA levels at the time the T cells were acquired are indicated by colored symbols: red, >100,000 copies/ml; yellow, 600–100,000 copies/ml; green, <600 copies/ml. Acute infection was defined as time points <180 days from initial viremia. PD-1 levels are highly variable in chronic infection. Some HCV-specific T cells from subjects with chronic infection expressed low PD-1 levels in the setting of high circulating HCV RNA levels (red dots at low PD-1 MFI values).

Correlations between viral sequence evolution and PD-1 levels on HCV-specific T cells

To determine whether the low PD-1 levels seen on T cells in patients with high circulating HCV RNA levels were due to escape mutations, the viral sequence was assessed at initial infection and surrounding the time at which the PD-1 levels were found to be low relative to initial PD-1 levels. We compared the levels of PD-1 on T cells specific for epitopes that underwent substitutions that impaired T cell recognition to that on T cells specific for epitopes that remained constant over the same time period in the same subjects. For two subjects with some epitopes that underwent substitution while other epitopes were maintained constant (subjects 17 and 30), the decline in PD-1 levels was observed only on T cells specific for the epitopes that underwent substitution (Fig. 4⇓). In contrast, PD-1 levels remained stable or increased for epitopes in the same host that did not undergo substitution. Subject 17 recognized three epitopes, two of which underwent substitution in the first 6 mo of infection (880TL10 and B35-135) and the other of which remained intact (140G), as shown in Fig. 4⇓a). T cells specific for 880TL10 and B35-135 showed 3- and 10-fold decreases in PD-1 levels following substitution, respectively. In contrast, subject 17’s T cells recognizing 140G, which remained constant over that time period, demonstrated a 3-fold increase in PD-1 expression. Subject 17’s 880TL10 and B35-135 HCV epitopes are both HLA B*3501 restricted. The substitutions resulted in 365- and 1690-fold reductions in HLA B*3501 binding, respectively. We have previously demonstrated that these mutations significantly impair T cell recognition in vitro but do not completely abrogate recognition (11). With these substitutions, T cell recognition of those epitopes is likely significantly impaired in vivo despite high levels of circulating virus. Similarly, in subject 13, who had one HCV-specific T cell response that could be assessed with multimers, a substitution in that T cell epitope that resulted in impaired but not complete loss of recognition (Fig. 5⇓) was associated with nearly a 3-fold drop in PD-1 expression (Fig. 4⇓b). Two additional findings further highlight the role of Ag in modulating PD-1 levels on HCV-specific T cells. In subject 175, HCV levels transiently fell to below the limit of quantitation at day 206 postinfection followed thereafter by re-emergence of high viral titers. PD-1 levels drop precipitously on HCV-specific T cells at the time when virus becomes nearly undetectable and are reinduced after viral reemergence (Fig. 4⇓c). The kinetics and magnitude of PD-1 re-expression are quite different for the two distinct epitopes, suggesting that additional factors may contribute to regulation of PD-1 expression besides simply Ag dosage. In subject 30, sequence escape at epitope C63B is observed at day 1144 after infection and is associated with a 5-fold drop in PD-1 levels (Figs. 4⇓d and 5⇓b). On Day 1521 after initial infection, the epitope reverted back to the original sequence. This reversion is associated with a 6-fold increase in PD-1 expression on the cognate T cells. The A2-61 epitope recognized by subject 30 does not undergo substitution, and the level of PD-1 expression over the same time frame varies <2-fold. In contrast to levels on T cells specific for epitopes that underwent substitution, PD-1 levels remained stable (declined <2-fold) or increased over time for epitopes in four additional subjects with no viral escape (Fig. 4⇓, e–h). Seven of the eight subjects studied had detectable T cell responses to at least one control (EBV, CMV, or Flu) Ag, and PD-1 levels differed by <2-fold over multiple time points on control epitope-specific T cells in the same time periods. (data not shown) In summary, epitope escape (or as in subject 175, decrease in viral titer to nearly undetectable levels) consistently resulted in decreases in PD-1 levels of 3- to 10-fold, whereas decreases in PD-1 levels were never >2-fold and usually increased over time when the cognate epitope sequence was maintained.

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

PD-1 up-regulation requires maintenance of intact Ag and restoration of intact Ag after escape is associated with an increase in PD-1 levels. Viral sequence was assessed at initial infection and at multiple subsequent time points at which T cells specific for known Ags were detectable, including those surrounding the time at which the PD-1 levels were found to be low relative to initial PD-1 levels. Levels of PD-1 on T cells specific for epitopes that either underwent substitutions that impaired T cell recognition or remained constant are shown. ○, Time points at which the substituted epitopes were present; •, points at which the initial viral sequence was present. a, Subject 17 recognized three epitopes: 880TL10 and B35–135 underwent substitution in the first 6 mo of infection; and 140G remained intact. PD-1 levels decline 3- and 10-fold following substitution of 880TL10 and B35-135, respectively. T cells recognizing 140G demonstrated a 3-fold increase in PD-1 expression. b, Subject 13 had one HCV-specific T cell response that underwent substitution with an associated drop of nearly 3-fold in PD-1 expression. c, In subject 175, HCV levels transiently fell to below the limit of quantitation with a decline in PD-1 levels followed by re-emergence of high viral titers and an increase in PD-1 levels on T cells specific for the epitope. d, Subject 30. Sequence escape at epitope C63B is associated with a 5-fold decline in PD-1 levels. C63B reverted back to the original sequence in association with a 6-fold increase in PD-1 expression on the cognate T cells. A2-61 does not undergo substitution, and the level of PD-1 expression varies <2-fold. PD-1 levels are maintained constant or increase over time in the absence of viral sequence variation. e–h, Levels of PD-1 on T cells specific for epitopes that remained constant are shown for four different HCV-infected subjects who recognized at least one T cell epitope for which multimers were available. For all four subjects, PD-1 expression on HCV-specific T cells decreases <2-fold or increases over time.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

Amino acid substitutions in T cell epitopes reduce recognition by cognate T cells, validating them as escape mutations. PBMC were cultivated in the presence of the peptide present at initial infection (•) or at the point where PD-1 levels decreased (○) in subject 17 (a) or subject 30 (b). After 20 days of stimulation with the peptide present in initial infection (left) or the variant peptide (right), cells were tested for recognition of both peptides in serial dilutions in an IFN-γ ELISPOT assay. The number of spot-forming colonies (SFC) per 30,000 cells is shown at each peptide dilution. In both cases, the variant peptide was less well recognized than the initial peptide by T cells expanded with the initial peptide (left). Stimulation with either of the variant peptides also resulted in reduced recognition of either the initial or variant peptides (right) and in reduced expansion of T cells as measured by T cell numbers (data not shown) compared with stimulation with the initial peptide. These substitutions were thus deemed escape mutations.