Chronic hepatitis C viral infection subverts vaccine‐induced T‐cell immunity in humans

Adenoviral vectors encoding hepatitis C virus (HCV) nonstructural (NS) proteins induce multispecific, high‐magnitude, durable CD4+ and CD8+ T‐cell responses in healthy volunteers. We assessed the capacity of these vaccines to induce functional HCV‐specific immune responses and determine T‐cell cross‐reactivity to endogenous virus in patients with chronic HCV infection. HCV genotype 1‐infected patients were vaccinated using heterologous adenoviral vectors (ChAd3‐NSmut and Ad6‐NSmut) encoding HCV NS proteins in a dose escalation, prime‐boost regimen, with and without concomitant pegylated interferon‐α/ribavirin therapy. Analysis of immune responses ex vivo used human leukocyte antigen class I pentamers, intracellular cytokine staining, and fine mapping in interferon‐γ enzyme‐linked immunospot assays. Cross‐reactivity of T cells with population and endogenous viral variants was determined following viral sequence analysis. Compared to healthy volunteers, the magnitude of HCV‐specific T‐cell responses following vaccination was markedly reduced. CD8+ HCV‐specific T‐cell responses were detected in 15/24 patients at the highest dose, whereas CD4+ T‐cell responses were rarely detectable. Analysis of the host circulating viral sequence showed that T‐cell responses were rarely elicited when there was sequence homology between vaccine immunogen and endogenous virus. In contrast, T cells were induced in the context of genetic mismatch between vaccine immunogen and endogenous virus; however, these commonly failed to recognize circulating epitope variants and had a distinct partially functional phenotype. Vaccination was well tolerated but had no significant effect on HCV viral load. Conclusion: Vaccination with potent HCV adenoviral vectored vaccines fails to restore T‐cell immunity except where there is genetic mismatch between vaccine immunogen and endogenous virus; this highlights the major challenge of overcoming T‐cell exhaustion in the context of persistent antigen exposure with implications for cancer and other persistent infections. (Hepatology 2016;63:1455‐1470)

H epatitis C viral (HCV) infection is a global epidemic and a leading cause of death and morbidity from liver disease. Recent estimates show a seroprevalence of 2.8%, with 185 million people infected. (1) While the incidence rate of HCV infection is decreasing in the developed world, HCVrelated deaths from advanced liver disease are predicted to increase over the next two decades. (2) We recently developed an HCV T-cell vaccine based on a chimpanzee adenovirus (ChAd3-NSmut) and an adenovirus derived from a rare human serotype (Ad6-NSmut), both encoding the nonstructural (NS) proteins of HCV genotype 1b, assessed in a heterologous prime/ boost vaccination strategy in healthy volunteers. (3,4) The vaccine was safe and well tolerated, and the magnitude and breadth of T cells induced after a single priming vaccination were the most potent described in human HCV studies to date. We now assess the capacity of the Abbreviations: CMV, cytomegalovirus; DAA, direct-acting antiviral; DMSO, dimethyl sulfoxide; ELISpot, enzyme-linked immunospot; HCV, hepatitis C virus; HLA, human leukocyte antigen; IFN, interferon; IL, interleukin; nAb, neutralizing antibody; NS, nonstructural; PBMC, peripheral blood mononuclear cell; PCR, polymerase chain reaction; PEG, pegylated; RIB, ribavirin; SFC, spot-forming cell; SI, stimulation index; TNF-a, tumor necrosis factor-a; vp, viral particle. same vaccine strategy to induce T cells in patients chronically infected with HCV genotype 1.
HCV may be particularly susceptible to a T-cell vaccine, as evidenced by human leukocyte antigen (HLA) genetic association studies, (5) chimpanzee T cell-blocking experiments, (6,7) and the temporal association of the magnitude and breadth of the T-cell response with viral eradication. (8) In general, broad, high-magnitude T-cell responses are seen in primary HCV infection (9) and are maintained in people who spontaneously resolve infection. (10) However, once persistent disease is established, T-cell responses are generally weak and narrowly focused (10) ; and although they may be detected ex vivo and expanded in vitro, (11) they are generally dysfunctional. (12) To date, the capacity of an immunogenic T-cell vaccine to recover T-cell immunity through vaccination in the periphery, remote from the tolerogenic liver environment, in people with persistent HCV infection is unexplored.
Previous immunotherapeutic vaccine approaches have included HCV peptide vaccines, (13) recombinant yeast expressing an HCV NS3-core fusion protein, (14) autologous dendritic cells loaded ex vivo with lipopeptides, (15) and DNA vaccines encoding HCV proteins. (16) In each case transient, very low-level effects were seen on T-cell induction or HCV viral load. More recently repeated vaccination in HCV-infected patients with modified vaccinia Ankara encoding HCV NS proteins in combination with pegylated interferon-a (PEG-IFN-a)/ribavirin (RIB) was associated with the induction of HCV-specific T cells at low level with a nonsignificant increase in sustained virological response in the vaccinated group. (17,18) However, previous studies of HCV immunotherapy have not evaluated the effect of vaccination in the context of circulating viral variants.
In this study we determine the capacity of a T-cell vaccine to induce HCV-specific T cells in patients with chronic HCV infection. We assess vaccination in the setting of both high and low viral loads following treatment with PEG-IFN-a/RIB because mouse studies of lymphocytic choriomeningitis viral infection have suggested that T-cell responses may be best recovered after viral suppression. (19) We also assess in detail the relationship between T-cell induction and endogenous circulating viral variants before vaccination. Our findings have important implications not only for HCV vaccine strategies but also for immunotherapy against other persistent pathogens and cancer.

PATIENT ENROLLMENT
Patients aged 18-65 with HCV genotype 1 were eligible for inclusion. Patients with human immunodeficiency virus, hepatitis B virus, immunosuppressive illness, Ad6 or ChAd3 neutralizing antibody (nAb) titer >200, or evidence of cirrhosis (clinical, biochemical, or histological) were excluded (for patient demographics, see Supporting Table S1). Patients were recruited at Oxford University NHS Trust and Queen Elizabeth Hospital Birmingham, Birmingham, UK. The study (EudraCT: 2008-006127-32) was approved by the Medicines and Healthcare products Regulatory Agency, registered in the ClinicalTrial.gov database (ID: NCT01094873), and ethically approved (GTAC reference: GTAC162). All volunteers gave written informed consent, and the studies were conducted in accordance with good clinical practice.
Two volunteers withdrew from arm A2 prior to vaccination after starting IFN-a, and two patients received prime vaccine only (021 rash postprime, 031 difficulties in venesection). Overall, 23 patients in arm A received both vaccines. All volunteers in arm B received both vaccines. Five patients stopped PEG-IFN-a/RIB early: three with IFN-a-related side effects (16-30 weeks), one after virological breakthrough (24 weeks), and one for personal reasons unrelated to the study (12 weeks). Volunteers documented all symptoms and recorded temperature daily. Solicited and unsolicited adverse events were recorded. All volunteers who received any vaccine were included in the analysis.

ADENOVIRAL CONSTRUCTS
The Ad6 and ChAd3 vectors encoding NS3-5B (1985 amino acids) of genotype 1b BK strain HCV (accession number M58335) have been described (20) and were manufactured at the Clinical BioManufacturing Facility, Oxford University, Oxford, UK.
To determine a positive cutoff, T-cell responses to DMSO were analyzed in 58 individuals with chronic HCV genotype 1 (Supporting Fig. S1). An HCVspecific response was defined as an IFN-c-ELISpot response 39 spot-forming cells (SFCs)/10 6 PBMCs/ pool (mean 6 3 standard deviations) and three or more times the background DMSO level. Positive pools were summed and background-subtracted. Epitope specificities were determined using frozen PBMCs rested overnight, by dividing positive peptide pools into minipools of approximately 10 peptides, then mapping to 15-amino acid-long peptides and optimal-length peptides. CD8 1 T-cell depletion IFNc-ELISpots were performed using individual peptides and negative depletion beads (Dynal magnets). Positivity in response to vaccination was defined as a 30% increase in preexisting responses or the induction of T cells targeting new antigens.

HCV RNA QUANTIFICATION
HCV RNA levels (lower limit of quantification 15 IU/mL) and subtypes were determined at the Virology Laboratory, John Radcliffe Hospital Oxford (COBAS AMPLICOR HCV test v2).

HCV VIRAL SEQUENCING
Autologous virus was sequenced at baseline and viral relapse. Thawed plasma (500 lL) was pelleted by centrifugation (23,600g at 48C for 60 minutes) and resuspended in 140 lL of plasma. Viral RNA was extracted (Qiagen). RNA was reverse-transcribed, and firstround polymerase chain reaction (PCR) was performed using the Superscript III One-Step RT-PCR System (Invitrogen) with specific primers and PCR cycling conditions (Supporting Table S2). Second-round PCR used High Fidelity Taq DNA polymerase (Roche). PCR products were gel-purified or PCR-purified (Qiagen). Products were sequenced bidirectionally using second-round internal primers and Prism Big Dye (Applied Biosystems) on an ABI 3100 automated sequencer. Cycling conditions were 968C for 1 minute, followed by 30 cycles of 968C for 15 seconds, 508C for 10 seconds, and 608C for 4 minutes. Sequences were analyzed and aligned using Sequencher (v. 4.10.1) and Se-AI (v.2.0a11 http://tree.bio.ed.ac.uk/software/).

SEQUENCE VARIABILITY AT T-CELL EPITOPES AT THE POPULATION LEVEL
HCV sequences were downloaded from the Los Alamos database (http://hcv.lanl.gov/content/index). Single variants with a prevalence of >5% were identified at each immunodominant epitope. Incomplete sequences and nongenotype 1a or 1b sequences were excluded.

nAbs TO ADENOVIRAL VECTORS
nAb assays were performed as described using recombinant adenoviruses expressing the reporter gene SEAP. (21) IL-28B GENOTYPING DNA was tested for polymorphism rs8099917 using the ABI TaqMan allelic discrimination kit (Applied Biosystems) where G is the protective allele and T is the risk allele, as described. (22)

STATISTICAL ANALYSIS
Data were assumed to have a nongaussian distribution. Nonparametric tests were used throughout: paired tests within an individual (Wilcoxon matched pairs signed rank test) or unpaired tests between individuals (Mann-Whitney). For correlations, Spearman's r test was used. Two-tailed P values were used throughout. P < 0.05 was considered significant. Prism v. 5.0d for Mac was used for analysis.

VACCINATION WITH Ad6-NSmut AND ChAd3-NSmut IS WELL TOLERATED
The primary endpoints of the study were safety and immunogenicity. Predominantly mild local and systemic side effects were reported. Overall vaccination was well tolerated, with no vaccine-related serious adverse events (Supporting Fig. S2.1). A transient increase in liver transaminases was observed in two patients, peaking at 103 IU/L and 339 IU/L, respectively (Supporting Fig. S2.2). A liver biopsy was performed in patient 038, which showed mild nonspecific inflammation only, in keeping with HCV infection.

HETEROLOGOUS Ch3Ad-NSmut PRIME/Ad6-NSmut BOOST INDUCES HCV-SPECIFIC T-CELL RESPONSES
We first assessed the immunogenicity of ChAd3-NSmut prime/Ad6-NSmut boost vaccination in a doseescalation regime in patients receiving concurrent PEG-IFN-a/RIB (arm A). Vaccines were administered 10 weeks apart following a 2-week or 14-week lead-in with PEG-IFN-a/RIB and as a single prime or double prime (ChAd3NS) with a heterologous boost (Ad6NS). No responses were seen at low and medium vaccine doses (data not shown). At the highest dose (2.5 3 10 10 vp, arms A3 and A4; Table 1) HCV-specific T cells were detected by IFN-c ELISpot in eight of 12 individuals receiving a single prime/boost (median 107 SFCs/10 6 PBMCs, range 10-2953), peaking 2-4 weeks following boost vaccination (Fig. 1A). Double-prime vaccination (4 weeks apart, arms A5 and A6) induced a response in four of eight vaccinees (median 36.5 SFCs/10 6 PBMCs, range 10-317) (Fig. 1B) and did not significantly enhance the response (Fig. 1C). Paradoxically, the highest responders were observed in the singleprime arms. Overall, there was a significant increase in peak HCV T-cell responses after boost but not after priming vaccination (Fig. 1D). A transient depletion of T cells specific for CMV, influenza, and Epstein-Barr virus after PEG-IFN-a/RIB therapy has been described (23) ; however, a clear effect of treatment on antiviral T-cell responses was not observed in this study (Supporting Fig. S3).

VACCINATION AT HIGH VERSUS LOW HCV VIRAL LOAD
Mouse models of lymphocytic choriomeningitis virus have shown that therapeutic vaccination may be more efficacious after viral suppression. (13) We compared vaccination 14 weeks after PEG-IFN-a/RIB therapy, when HCV RNA was undetectable in nine of 10 volunteers (arms A3, A5), with vaccination 2 weeks after therapy (arms A4, A6) when HCV viral loads were significantly higher (median 0 versus 11, 441 IU/mL, P 5 0.0002) and in untreated patients in arm B, with high viral loads (median 2,300,000 IU/mL). There was no difference in vaccine responses between the three groups ( Fig. 1F) and overall no correlation between the viral load at time of priming vaccination and the peak T-cell response thereafter (Supporting Fig. S4).

HCV-SPECIFIC T-CELL RESPONSES IN CHRONIC HCV INFECTION COMPARED TO HEALTHY VOLUNTEERS
The magnitude of the response was significantly lower in HCV-infected patients compared with healthy volunteers vaccinated with an identical regimen as part of a previous study (3) (mean 6 standard deviation HCV versus healthy volunteers: postprime 49.25 6 54.15 versus 1476 6 978.8, P < 0.0001, postboost 317.3 6 704.7 versus 21836 4633, P 5 0.08, respectively) ( Fig. 2A).
The kinetics of the response also differed between these groups; the high-magnitude response that was observed after priming vaccination in healthy volunteers was markedly attenuated in HCV-infected patients (Fig. 2B).

ANTIVECTOR IMMUNITY IN RESPONSE TO VACCINATION
Following priming vaccination, ChAd3 nAbs that were cross-reactive with Ad6 were readily detectable and maintained to the time of heterologous boost in both single-prime and double-prime study groups (Fig. 3A,B). We found a negative correlation between anti-ChAd3/Ad6 nAbs at the time of boost vaccination, with peak IFN-c-ELISpot thereafter (r 5 -0.48, P 5 0.03, and r 5 -0.44, P 5 0.05) (Supporting Fig.  S5). Robust antivector T-cell responses were observed following both priming and heterologous boosting vaccination, showing that T-cell responses are not generally defective in patients with HCV infection or during PEG-IFN-a/RIB therapy (Fig. 3C,D).

THE RELATIONSHIP BETWEEN VACCINE-INDUCED T CELLS AND ENDOGENOUS VIRUS
T-cell responses were mapped to individual peptides using IFN-c-ELISpot. Fifteen antigenic targets were identified spanning the vaccine immunogen (Table 2; Supporting Fig. S6). Of these, 13 were CD8 1 T-cell  (6) ) and HCV patients (black, n 5 20, groups A3-6). Responses shown are the total positive ex vivo IFN-c ELISpot responses across all NS peptide pools at peak magnitude postprime and peak postboost. Bars at median (A). Kinetics of the total ex vivo IFN-c ELISpot response across all NS peptide pools in HCV-infected patients (A3-4) and healthy volunteers (group 10 as published (6) ) receiving single-prime and boost vaccination (B). Bars at mean and standard error of the mean. **P 0.01, ****P < 0.0001. Abbreviations: S, screening; PP, prepriming. epitopes that have been described in natural infection and two were novel CD8 1 epitopes mapped to optimal length: NS5B 2804-2814 LTRDPTTPLAR and NS4B 1891-1899 ALVVGVVCA (patients 055 and 053, respectively) (Supporting Fig. S7). The most commonly identified epitope was NS3 1406 (KLSGLGI-NAV), a well-described HLA-A2 epitope, (17) targeted in six of 17 HLA-A2 1 patients. Depletion ELISpot and intracellular cytokine staining confirmed that mapped responses were CD8 1 . No mapped responses were made by CD4 1 T cells.
Next we determined the endogenous viral sequence at the 15 CD8 1 T-cell epitopes at baseline and time of viral relapse in all patients (Supporting Table S3). At 14/15 epitopes, viral and vaccine immunogen sequences differed by one to four amino acids in patients targeting those epitopes (Table 3). We then assessed whether vaccine-induced T cells could recognize endogenous viral peptide variants using IFN-c-ELISpot and HLA class I pentamers. Vaccine-induced T cells targeting the vaccine immunogen at NS3 1406 KLSGLGINAV, NS3 1395 HSKKKCDEL, and NS5B 2804 LTRDPTTPLA failed to recognize circulating autologous variants at these epitopes ( Fig. 4A-C). In addition, T cells targeting the dominant NS3 1406 epitope failed to recognize an additional two variant peptides (KLVAL-GINAV and KLVSLGLNAV) in the patient cohort. HLA class I pentamers (epitopes NS3 1406 and NS3 1073 ) matching either vaccine immunogen or endogenous variants confirmed the loss of cross-reactivity to the NS3 1406 KLSGLGINAV endogenous variants (Supporting Fig. S8). Partially cross-reactive T-cell responses between vaccine immunogen and endogenous viral sequences were observed at epitopes NS3 1073 CVNGVCWTV and NS3 1411-1425 GINA-VAYYRGLDVSV (Fig. 4D,E). Sequence analysis at epitopes where HLA restriction was defined showed that genetic mismatch between immunogen and circulating virus was significantly associated with response to vaccination (P 5 0.0063; Fig. 4F). For example, at the HLA-B8 NS3 1395 HSKKKCDEL epitope, circulating virus matched the immunogen in four of six HLA-B8 patients; however, the only patient responding to this epitope (patient 045) had the well-described escape variant NS3 1395 HSKRKCDEL (24) at baseline. Similarly, at epitope LTRDPTTPLAR, 25 viral sequences matched the immunogen but only the patient with a variant epitope (LTRDPITPLAR) responded to vaccination (Supporting Table S3). However, genetic mismatch did not guarantee a response to vaccination, and no specific viral sequence at any epitope was predictive of a response to vaccination. For example, at epitope HLA-A2 NS5 2594 ALYDDVSTL identical endogenous epitope variants (ALYDVVSKL) were found in eight of 17 HLA-A2 patients, only one of whom (027) had a T-cell response to this epitope after vaccination. Similarly, at epitope HLA-A2 NS3 1406 KLSGLGINAV three of four variant epitopes identified in vaccine "responders" were also found in HLA-A2 nonresponders (Supporting Table S3).
Using genotype 1a/1b sequences from the Los Alamos database, we evaluated the variability of the 15 vaccine-induced T-cell epitopes at the population level by Shannon entropy score (median 2149 sequences, range 1131-3896/epitope). T-cell epitopes were generally variable at the population level (Supporting Fig.  S9A). The most dominant, NS3 1406 KLSGLGINAV, was the most variable, with no single variant present in >30% of sequences at the population level (Supporting Fig. S9B). A minority of epitopes were found to be highly conserved (e.g., NS4B 1851 ILAGY-GAGV in 97.3% of genotype 1 sequences and homologous to vaccine immunogen; Shannon entropy score 0.35; Supporting Fig. S9C). However, at one conserved epitope, NS5B 2804 LTRDPTTPLAR (Shannon entropy score 0.28), where 89% of population genotype 1 sequences were identical with the vaccine immunogen (Supporting Fig. S9D), the viral sequence in the only vaccinee responding to this epitope was a non-cross-reactive variant.

VACCINE-INDUCED T-CELL FUNCTION IN PATIENTS WITH CHRONIC HCV
We used HLA class I pentamers to examine the magnitude, phenotype, and function of vaccineinduced T cells postboost and at the end of the study and compared these with published data using an identical vaccine regime in healthy volunteers (3) (Fig. 5;  Supporting Fig. S10). This analysis confirmed that the magnitude of the responses was lower in HCVinfected patients than healthy volunteers at the epitope level ( Fig. 5A; Supporting Fig. S10A). A subset of patients had small but detectable CD8 1 T-cell responses prevaccination, and changes in the  phenotype of these T-cell responses after vaccination were assessed and compared with healthy volunteers (Fig. 5C). The proportion of pentamer-positive HCVspecific T cells was high in patients for effector memory subset (CCR7 -/CD45RA) at baseline and remained significantly enriched after vaccination compared to healthy volunteers. There was no difference in bulk CD8 1 effector memory subsets in patients compared to healthy volunteers (Supporting Fig. S10B). In patients, pentamer-positive T cells were generally low in markers of cytotoxicity and activation (granzyme B, granzyme A, perforin, and CD38) at baseline; and although expression of these markers increased in response to vaccination, it remained significantly lower than that observed in healthy volunteers at peak response postboost. For granzyme B this difference was maintained to the end of the study. PD-1 expression was high at baseline and remained high after vaccination in patients (Fig. 5C), although interpretation of this observation is complicated by the fact that PD-1 expression is a marker of both exhaustion and activation. In healthy volunteers a small subset of vaccineinduced T cells express CD161, a marker related to a more highly functional subset of antigen-specific Tcells. (25) However, vaccine-induced T cells in patients showed minimal CD161 expression. The expression of CD127 was also lower in patients compared to healthy controls at the end of the study (although not statistically significant) (Supporting Fig. S10C), suggesting a reduced capacity to persist by homeostatic proliferation in response to IL-7. A strong positive correlation between the magnitude of response as measured by ELISpot and pentamer staining suggested that a population of pentamer-positive cells that do not make IFN-c is not present in patients or volunteers (Supporting Fig. S10A).
Using intracellular cytokine staining, we showed that CD4 1 T-cell responses were detected at a low level only, while CD8 1 T cells made some IFN-c and IL-2 but minimal TNF-a in patients ( Fig. 5D; example plots in Supporting Fig. S10D).
Breadth of T-cell response is an important correlate of immunological protection against HCV. (10) In contrast to the broad responses detected in healthy volunteers, responses were significantly narrower in patients (median 5 positive pools, range 2-6, versus 1 positive pool, range 0-5; P < 0.0001 (Fig. 5E). Finally, although we detected robust proliferative responses to CMV control antigens in HCV-infected patients (median SI 173.6, range 30.2-1104.0), in stark contrast to data in healthy volunteers proliferative responses to HCV proteins were largely undetectable (median SI 3.5, range 3-14.6) (Fig. 5F).
Overall these data suggest that T cells induced by vaccination in HCV-infected patients remain at least partially functionally attenuated.

Discussion
This is the first study to assess the capacity of a potent HCV T-cell vaccine to restore antiviral T-cell responses in patients with chronic HCV infection and to assess in detail the complex interplay between HCV-specific T-cell induction and circulating viral variants in the context of T-cell immunotherapy. Although HCV-specific T cells are generated in some patients, these do not typically target circulating virus because of an absence of sequence homology between circulating viral variants and vaccine immunogen at Tcell epitopes. Furthermore, when T cells are induced, these appear to be functionally attenuated.
We have shown that an HCV T-cell vaccine based on chimpanzee adenoviruses (ChAd3) and adenoviruses derived from a rare human serotype (Ad6) can induce high-magnitude, polyfunctional CD4 1 and CD8 1 T cells targeting multiple HCV antigens after a single injection with either vaccine. (3,4) We found that when the same vaccines are given to patients with chronic disease, T-cell responses were rarely detected after priming vaccination. Only after heterologous adenoviral boosting vaccination are T cells generated, but these are at a lower magnitude and target fewer antigens than we observed in healthy volunteers. Further analysis showed that while HCV-specific CD8 1 T cells could be generated in some people, HCVspecific CD4 1 T cells were largely absent.
In a stepwise approach, we vaccinated patients at low and intermediate HCV viral loads following PEG-IFN-a/RIB, before vaccinating treatment-naive patients with high viral loads. Overall, vaccination was well tolerated with no clear evidence of liver immunopathology. This was expected as HCV infects only a minority of hepatocytes (<20%) (26) and in the context of acute HCV infection, when T cells are maximally generated, severe hepatic inflammation is rare. While transient elevation in liver transaminases was observed in two patients, this was not clearly temporally related to vaccination and neither patient had a vaccineinduced T-cell response. Although murine models of lymphocytic choriomeningitis viral infection and other human models of persistent viruses support the hypothesis that T cells may be maximally induced in the context of viral suppression, (19,27) we found no correlation between HCV viral load and T-cell induction. One confounder to consider is that both PEG-IFN-a and RIB, which were coadministered with vaccination, have been shown to attenuate cellular immune responses. (28,29) However, antivector, influenza, Epstein-Barr virus, and CMV-specific T-cell responses could be detected in the presence of PEG-IFN-a/RIB therapy, suggesting that therapy did not induce a global impairment of T-cell function.
It is well established that in chronic HCV disease, T cells are found at a low magnitude and are functionally attenuated with reduced proliferative capacity. (10) The reasons for this may include persistent high-level antigen exposure, inhibition by regulatory T cells, and dysfunctional T-cell priming in the tolerogenic liver environment (reviewed by Bowen and Walker (30) ). These defects appear to persist following viral eradication with PEG-IFN-a/RIB. (31) We assessed whether a potent T-cell vaccine could overcome these factors, either by enhancing existing weak responses or by generating new HCV-specific T-cell responses. However, we found that T-cell responses were generated in a minority of individuals.
In order to better understand why some patients responded to vaccination and others did not, we evaluated the sequence of endogenous virus in patients. We showed that in vaccine responders there was sequence divergence between circulating viral sequence and vaccine immunogen at T-cell targets, with a loss of crossreactivity between these variants. This suggests that HCV T-cell vaccines may induce T-cell responses in chronic infection but primarily in the absence of homologous antigen stimulation. The corollary of this is that these T cells will be unable to target endogenous virus or inhibit viral replication. The finding that vaccination had no effect on HCV viral load is in keeping with this. In contrast, T cells were generally not induced when there was sequence homology between the vaccine immunogen and endogenous virus at known T-cell epitopes; in this context vaccination was unable to restore functional immunity. These observations are relevant for vaccine development against other variable pathogens such as human immunodeficiency virus and for cancer immunotherapy, where mismatch between vaccine immunogen and endogenous antigens will commonly arise.
Failure to respond to vaccine in the context of an epitope which is matched between virus and vaccine may be a result of T-cell exhaustion. Detection of T-cell responses where mismatch occurred may therefore either be the result of de novo priming or represent expanded memory responses, the latter potentially demonstrating the phenomenon of "original antigenic sin" (enhanced memory responses to cross-reactive epitopes). (32) Modifying the antigen to broaden the crossreactivity may recruit further T-cell pools, (33) but it will be challenging to overcome the limitations for therapeutic vaccination posed by T-cell exhaustion and escape.
We also observed that T cells generated by vaccination were functionally impaired in HCV-infected patients compared to healthy volunteers: low in proliferative capacity, markers of both cytotoxicity and Tcell activation, and TNF-a production, a phenotype that is characteristic of chronic viral exposure. (12) The marked lack of T-cell proliferation may reflect the fact that the proliferation assay uses recombinant HCV proteins that preferentially detect CD4 1 T-cell responses. However, it has also been shown that there is a hierarchical loss of function during T-cell exhaustion, such that proliferative capacity may be lost relatively early after prolonged antigen exposure in comparison to IFN-c production. (34) A further explanation for partial dysfunction of vaccine-induced T cells is that HCV infection per se inhibits the priming of naive T cells to produce functional memory T cells. In support of this, it has been shown that HCV infection is associated with a loss of other virus-specific (CMV) mature effector memory CD8 1 T cells. (35) However, in the absence of liver cirrhosis, there is little evidence that HCV-infected patients respond abnormally to routine clinical vaccines (36) ; and in our study patients generated robust antivector T-cell responses following vaccination. An alternative explanation is that vaccination is not priming truly naive T cells but is stimulating memory responses that were generated early in infection but that were rendered partially dysfunctional following viral exposure. The degree of dysfunction would then depend on the duration of antigen exposure at the epitope level, with more functional responses associated with viral escape early in infection.
In this study vaccination induced predominantly CD8 1 T cells. The generation of robust CD4 1 T-cell responses is thought to be a key determinant of viral clearance as evidenced both experimentally in natural history studies (37) and through large host genome-wide association studies that identify HLA class II alleles as critical genes associated with spontaneous resolution. (38) Therefore, vaccine strategies that aim to induce robust CD4 1 T-cell immunity should be considered. Heterologous boosting with a modified vaccinia Ankara vector following adenoviral vector prime has recently been shown to induce a more balanced CD4 1 and CD8 1 response compared to heterologous adenoviral vaccination. (4) This vaccine regimen is now in efficacy testing in intravenous drug user populations as a prophylactic HCV vaccine (39) (NCT01436357) and is also being currently evaluated in an HCV therapeutic vaccine study (NCT01296451). The use of an alternative boosting vector also overcomes the issue of antiadenoviral antibodies that may limit T-cell responses to a heterologous adenoviral boost.
The utility of an immunotherapeutic approach for HCV can be debated. Historically, treatment of HCV has relied on PEG-IFN-a/RIB-based therapies associated with severe side effects. Recently licensed all-oral combination direct-acting antiviral (DAA) therapies given for 8-24 weeks will transform our capacity to treat patients, with recent phase 2b/3 studies showing viral clearance rates of >90%. (40) However, impacting on the HCV epidemic may require the treatment of intravenous drug users, who frequently live in chaotic social settings (41) ; therefore, vaccines administered as adjuvant immunotherapy that markedly shorten therapy duration would be beneficial. In addition, the development of a prophylactic HCV vaccine remains an important goal, and the assessment of T-cell induction in the presence of circulating viral variants may be readily assessed through therapeutic vaccine strategies and thus inform rational vaccine design. With the advent of all-oral DAA therapies, the capacity of a T-cell vaccine to induce a functional T-cell response and thus protect individuals from reinfection after drug-induced viral clearance will depend of the recovery of T-cell function after therapy. Recent data suggest that both natural killer function and T-cell function may recover after alloral DAA therapy, in contrast to IFN-based treatments. (42,43) Therefore T-cell vaccine studies in this group will be of particular interest.
Our study highlights the formidable challenges in developing an effective immunotherapeutic vaccine for HCV with implications for immunotherapy against variable pathogens in general and in cancer. Additional strategies, such as the use of checkpoint modulation, may be a useful adjunct, (44) although the safety profiles of these are unlikely to be clinically acceptable for the treatment of HCV, where DAAs represent an alternative curative treatment strategy. In conclusion, we have shown that a highly potent T-cell vaccine regimen is not able to restore T-cell function during chronic HCV disease and that without determining circulating viral sequence for variable pathogens, the true value of T-cell induction cannot be accurately assessed.