Patients, Materials, and Methods

PatientsThe design of the HALT-C Trial has been described elsewhere [8]. Briefly, criteria for enrollment were chronic HCV infection with a detectable serum HCV RNA level, prior nonresponsiveness to IFN-α therapy (with or without ribavirin), and the presence of bridging fibrosis or cirrhosis in liver-biopsy samples obtained during the 12 months before enrollment. All liver-biopsy and peripheral blood samples were prospectively collected at 4 of 10 HALT-C clinical sites before the initiation of therapy; the 4 sites that participated did so on the basis of investigator interest. The institutional review boards of all involved institutions approved the study protocols, and written, informed consent was obtained from all study participants. The patients in the present analysis had usable results for any of the 4 substudies (the intrahepatic CTL, peripheral blood LP, NA, or QS substudies). Evaluation of serum HCV RNA levels, HCV genotype, and liver biopsies were performed as described elsewhere [9]. Patients were treated for 20 weeks with pegylated IFN-α2a (180 μg weekly; PEGASYS, Roche Pharmaceuticals) and ribavirin (1000–1200 mg daily, depending on body weight; COPEGUS, Roche Pharmaceuticals) in 2 divided doses. At the week 20 (W20) time point, serum HCV RNA levels were tested in duplicate using the qualitative Roche COBAS Amplicor assay (version 2.0; Roche Molecular Systems). Patients with undetectable serum HCV RNA levels at W20 (virologic responders [VRs]) continued the combination therapy for 28 additional weeks (for a total of 48 weeks) and then were followed for 24 weeks; those with undetectable serum HCV RNA levels at the week 72 (W72) time point were considered to be sustained VRs (SVRs; figure 1). Patients with detectable serum HCV RNA levels at W20 entered the randomized phase of the HALT-C Trial and were considered to be W20 non-VRs as well as W72 non-SVRs. Finally, patients with undetectable serum HCV RNA levels at W20 but in whom serum HCV RNA reappeared at any time point after W20 were considered to be breakthrough/relapse patients and were included in the non-SVR group. Patients who had missing HCV RNA data at W20 or W72 were considered to be non-VRs (intention-to-treat analysis)

LP, CTL, NA, and QS assaysLymphoproliferative responses to superoxide dismutase (SOD)–recombinant HCV genotype 1a protein antigens SOD-c22 (HCV aa 2-120), SOD-c100 (HCV aa 1569-1931), SOD-NS5 (HCV aa 2054-2995), Escherichia coli–derived SOD-c33c (HCV aa 1192-1457) and recombinant human SOD (all gifts from M. Houghton, Chiron) were measured in quadruplicate wells using [³H]-thymidine incorporation assays with freshly isolated peripheral blood mononuclear cells (PBMCs), as described elsewhere [10, 11]. T cell lines were derived from fresh liver-biopsy samples (length, ∼0.5 cm) and tested in triplicate wells for cytolysis of autologous ⁵¹Cr-labeled B lymphoblastoid cell lines (BLCLs) infected with recombinant vaccinia viruses expressing regions spanning the entire HCV genotype 1a polyprotein—aa 1–339 (C-E1; vv9A), aa 347–906 (E2-NS2; vv1H), and aa 827-3011 (NS2-NS5; vv827), or β-galactosidase as a negative control, as described elsewhere [10, 12, 13]. HCV chimeric envelope glycoprotein (E1-G or E2-G; genotype 1a)/vesicular stomatitis virus (VSVts045; temperature-sensitive mutant of VSV) pseudotype viruses (∼100 pfu) were incubated with serial dilutions of heat-inactivated test sera, and plaques on baby hamster kidney cell monolayers were counted to assess NA activity, as described elsewhere [10, 14, 15]

HCV QS diversity and complexity were evaluated by clonal frequency analysis of the E2 HVR1, as described elsewhere [10, 16]. For each patient, 20 individual QS clones from frozen baseline serum samples were analyzed

LP assays were scored as positive when the stimulation index (SI) value for any HCV antigen was >4. The SI was defined as the ratio of the mean test antigen cpm/mean control cpm. CTL assays were scored as positive when the percentage of specific lysis of BLCLs infected with HCV-vaccinia recombinant viruses was >10% above the percentage of specific lysis of the control target cells at 1 or more effector-to-target cell ratios. NA assays were scored as positive when ⩾50% neutralization against either E1- or E2-pseudotyped viruses was detected. These cutoffs were selected before data analysis and were based on previous data, from our own laboratories and others, regarding the sensitivity and specificity of these assays, and they were clearly different from those for normal controls [7, 11–15, 17, 18]. Participating laboratories were blinded to treatment outcome

Statistical analysesStandard statistical methods (χ² test, Fisher’s exact test, odds ratios [ORs], Student’s t test, Wilcoxon&rank sum test, Cochran-Mantel-Haenszel test, Spearman’s correlation coefficient, and multivariate logistic regression) were performed using SAS (release 8.2). All analyses were 2-tailed, with α=.05

Results

Characteristics of the study population and baseline immune responsesOf the 367 patients analyzed in the present study who had treatment outcomes available, not all of them had samples tested for each of the substudies. Sample size limitations for the LP (n=338), CTL (n=210), and NA substudies (n=244) have been described elsewhere [10]. QS analysis was limited to HCV genotype 1–infected patients (n=103). Seventy-one patients had results for all 4 substudies. Demographic and other data describing all of the patients in this cohort are shown in table 1. The major difference between the overall cohort and the substudy cohorts was a somewhat lower prevalence of cirrhosis in the patients in the CTL substudy, which probably resulted from the greater difficulty in obtaining fresh liver tissue from those with cirrhosis for analysis [10]. We also previously reported details on the limited specificity, magnitude, and breadth of immune responses observed in the overall cohort to the genotype 1a antigens used [10]. Positive CTL responses were more frequent in patients with non–genotype 1 infection than they were in patients with genotype 1 infection (30% vs. 20%; P=.31, Fisher’s exact test), but this difference was significant only in the case of CTL responses to C-E1 (vv9A) (20% vs. 7%; P=.047, Fisher’s exact test). Genotype did not appear to influence the frequency of positive LP or NA assay results (data not shown). All analyses were also performed including only those patients with HCV genotype 1 infection; the results obtained were not significantly different from those obtained in the analyses of the entire cohort

Figure 1 illustrates the study design. In the overall study cohort, 35% (127/367) achieved a virologic response at W20, as defined by an undetectable serum HCV RNA level at that time point (table 1). Nineteen percent (70/367) became SVRs (undetectable serum HCV RNA level at W72). Treatment-response rates were not significantly different among the substudy populations and were similar to those observed in the HALT-C cohort (n=604) [9]

Similarity of the frequencies of LP, CTL, and NA responses at baseline in W20 VRs and non-VRsTo investigate the association between baseline immune response and on-treatment virologic response, patients were categorized as either W20 VRs or non-VRs, depending on the absence or presence of detectable serum HCV RNA after 20 weeks of pegylated IFN-α2a and ribavirin combination therapy. The percentage of patients with positive LP, CTL, or NA assay results at baseline was not different between W20 VRs and non-VRs, as is shown in figure 2A (P=.59, P=1.00, and P=.20, respectively; Fisher’s exact test). Patients who had missing serum HCV RNA data at W20 (n=25) or W72 (n=10) were classified as non-VRs. Removal of these patients from the analyses did not change any of the results (data not shown)

More frequent detection of LP and NA responses at baseline in non-SVRs than in SVRsTo determine whether baseline immune responses to HCV antigens might predict subsequent treatment-induced viral clearance, patients were categorized according to their virologic status at W72, after 48 weeks of pegylated IFN-α2a and ribavirin combination therapy and 24 weeks of follow-up. For this analysis, all patients with detectable serum HCV RNA levels at W20 as well as W72 were considered to be non-SVRs. The percentage of patients with positive LP, CTL, or NA assay results at baseline among the SVRs and the non-SVRs is shown in figure 2B. Surprisingly, patients with positive NA assay results at baseline were significantly more likely to become non-SVRs (P=.01, Fisher’s exact test; OR, 0.27 [95% confidence interval {CI}, 0.09–0.80], where an OR of <1.0 indicates a negative correlation between a positive assay and becoming an SVR; negative predictive value [NPV], 93%). Although more non-SVRs had positive LP assay results at baseline than did SVRs, only a trend toward statistical significance was found (P=.10, Fisher’s exact test; OR, 0.53 [95% CI, 0.25–1.09]; NPV, 87%). No such association was demonstrated for the CTL assay (P=.83, Fisher’s exact test). It is noteworthy that, when patients had 1 or more positive assay results (LP, NA, or CTL) at baseline, they were also less likely to become SVRs (P=.04, Fisher’s exact test; OR, 0.38 [95% CI, 0.15–0.96]; NPV, 88%). When patients with positive LP and NA assay results at baseline were compared with those with negative LP and NA assay results at baseline, it was found that no SVRs had positive results for both assays (P=.02, Fisher’s exact test; NPV, 100%). These findings indicate that the presence of an HCV-specific immune response at baseline—whether determined by the NA assay alone, 1 or more of the assays, or both the NA and LP assays together—was associated with subsequent virologic nonresponse to antiviral therapy

Because prior ribavirin therapy was a strong predictor of subsequent virologic nonresponse after pegylated IFN-α2a and ribavirin combination therapy in the overall HALT-C cohort [9], the possibilities that prior ribavirin therapy (1) may have affected the LP and NA assay results at baseline and (2) may have selected for different populations of patients were considered. When the subset of patients without prior ribavirin therapy was examined (n=67), the results were similar to those for the overall cohort. Specifically, non-SVRs were more likely to have positive NA assay results at baseline than were SVRs (P=.04, Fisher’s exact test; data not shown). Non-SVRs were also more likely to have positive LP assay results at baseline (31%) than were SVRs (17%), although the association was not statistically significant (P=.20, Fisher’s exact test). Therefore, prior ribavirin therapy did not appear to significantly affect the relationship between baseline LP or NA responses and subsequent virologic response. The median interval from baseline immunologic testing and receipt of the last dose of ribavirin as part of the prior therapy was 25 months (range, 5–62 months)

Effect of the presence of cirrhosis on the association between baseline LP responses and subsequent virologic response to therapyGiven our previous finding in this cohort of a possible association between the presence of cirrhosis and a decreased frequency of HCV-specific LP responses [10], we evaluated whether the presence of cirrhosis could affect the association between baseline LP responses and subsequent virologic response to therapy. When the presence of cirrhosis was controlled for, the association between a positive LP assay result at baseline and becoming an SVR was statistically significant (P=.04, Cochran-Mantel-Haenszel test). Moreover, when patients with cirrhosis (Ishak fibrosis score of 5 or 6) were excluded from the analysis, the association between a positive LP assay result at baseline and becoming a non-SVR was stronger (P=.02, Fisher’s exact test; OR, 0.38 [95% CI, 0.17–0.86]) (table 2). In the 104 patients with cirrhosis, however, the frequency of positive LP assay results at baseline was low (17/104 [16%]), and a positive result did not predict subsequent virologic nonresponse to therapy (P=.67, Fisher’s exact test). Similar analyses did not reveal a significant influence of liver histology on the relationship between NA or CTL assay results and treatment outcome (data not shown). Furthermore, PBMC viability, which could have affected the results, was similar among the groups analyzed; did not vary with the degree of fibrosis, virologic response, or virologic nonresponse; and did not differ between patients with a positive LP assay result at baseline and those with a negative result (data not shown)

Confirmation, by multivariate logistic regression analysis, of the correlation between baseline immune responses and subsequent virologic responseApart from the presence of cirrhosis and prior ribavirin therapy, other important factors known to affect the virologic response to antiviral therapy are serum HCV RNA level, HCV genotype, and the amount of pegylated IFN-α2a and ribavirin received by patients. We therefore performed multivariate logistic regression analysis on data from our cohort, controlling for the presence of cirrhosis, prior ribavirin therapy, genotype 1 infection, log serum HCV RNA level, and receipt of >80% of the prescribed pegylated IFN-α2a and ribavirin during the first 20 weeks of treatment. As is shown in table 3, nonresponse to therapy at W72 was significantly associated with positive LP assay results at baseline (P=.02, χ² test; OR, 0.37 [95% CI, 0.16–0.86]) and with 1 or more positive assays among patients with results for all 3 (LP, CTL, and NA) (P=.02, χ² test; OR, 0.25 [95% CI, 0.08–0.83]). In this model, positive NA assay results at baseline demonstrated only a trend toward an association with virologic response at W72 (P=.07, χ² test; OR, 0.33 [95% CI, 0.10–1.10])

Increased baseline HCV QS diversity in non-VRsBaseline HCV QS diversity (degree of nucleotide heterogeneity) and complexity (number of unique variants) were examined in 83 W20 non-VRs and 20 W20 VRs (8 of whom became SVRs) (figure 3A). For consideration of genetic diversity, a value of 1.0 denotes homogeneity; thus, decreasing values correlate with increasing heterogeneity at the nucleotide level [19, 20]. Both QS diversity and complexity were significantly higher at baseline in those who became W20 non-VRs than they were in those who became W20 VRs (P=.04 and P=.03, respectively; Wilcoxon&amp;rank sum test). Baseline QS diversity was also increased in non-SVRs, compared with that in SVRs (P=.01, Wilcoxon&amp;rank sum test) (figure 3B left panel). Although the median QS complexity was increased in non-SVRs, compared with that in SVRs, only a statistical trend for this association was found (P=.09, Wilcoxon&amp;rank sum test) (figure 3B right panel). When patients were categorized as W20 non-VRs, breakthrough/relapse patients, and SVRs, a stepwise decrement in QS diversity and complexity was found (for diversity, P=.02; for complexity, P=.02; Spearman’s correlation coefficient) (figure 3C). No association between serum HCV RNA levels and QS diversity or complexity was found (P=.31 and P=.56, respectively; Spearman’s correlation coefficient). Finally, no direct associations could be established between results for the LP, CTL, or NA assays and the HCV QS analysis, even when results for individual antigens were considered

Discussion

In our cohort of 367 patients tested for HCV-specific immune responses before the initiation of therapy, we found a surprising association between the presence of baseline LP and NA responses and subsequent failure to achieve viral clearance with state-of-the-art antiviral therapy. The unique features of the present study included the large number of samples tested for multiple aspects of the immune response, the specialized cohort of patients with advanced hepatic fibrosis and prior nonresponse to IFN-α therapy, and our focus on the relationship between baseline immune responses and QS variability before the initiation of therapy and subsequent treatment outcome. Multivariate logistic regression analysis that controlled for the presence of cirrhosis, prior ribavirin therapy, genotype 1 infection, log serum HCV RNA level, and receipt of >80% of the prescribed pegylated IFN-α2a and ribavirin revealed that both a positive LP assay result at baseline and 1 or more positive immunologic assay results at baseline predicted subsequent non-SVR status at W72. This is the first report to describe HCV-specific immune responses before the initiation of therapy and relate them to treatment outcome in such a large group of prior non-VRs with advanced hepatic fibrosis

Earlier studies have described an association between increased QS variability and nonresponse to therapy, although most [19, 21, 22], but not all [23], studied significantly fewer patients. We similarly demonstrated increased QS variation at baseline in patients who proceeded to virologic nonresponse at both W20 and W72, although ours is the first report of such a correlation in prior non-VRs with advanced hepatic fibrosis. Moreover, the present study found both increased QS variability and immune responses to be associated with subsequent virologic nonresponse, which is consistent with the notion that immune responses drive QS variation in vivo [24, 25]

However, we were not able to identify direct relationships between LP, CTL, or NA assay results and QS assay results. Our inability to detect these relationships could have resulted from a number of obstacles, such as the relatively small number of overlapping patients analyzed in the QS and immunologic assays, the nature of the cohort studied (all prior non-VRs with advanced hepatic fibrosis), and the fact that heterologous (rather than patient-specific) antigens were used in the immunologic assays. In addition, the LP assay did not specifically assess responses to the HCV E2 HVR1, which was assessed in the QS assay. Nevertheless, the associations of both increased baseline QS variability and increased baseline LP and NA responses with virological nonresponse lead us to hypothesize that T and B cell responses to HCV may have been sufficient to drive QS diversification but may not have been sufficient to predispose patients to virologic clearance. The parameters by which these immune responses are ineffective or counteractive with regard to viral eradication have yet to be defined

The baseline immune responses present in our eventual non-VR population were clearly inadequate—in terms of magnitude, breadth, and/or specificity (as detailed in Rothman et al. [10])—for the purpose of eradicating HCV. Moreover, the immune responses detected were associated with subsequent nonresponse to antiviral therapy. This finding may appear to conflict with earlier data demonstrating a relationship between enhanced HCV-specific CD4⁺ T cell proliferative responses and subsequent virologic clearance [4–7]. However, these other studies were performed in treatment-naive cohorts and focused on the development of T cell responses during and after therapy. By contrast, our cohort had longstanding advanced hepatic fibrosis, a history of prior nonresponse to IFN–α therapy, and was studied before the initiation of therapy. It is notable that African Americans, another treatment-resistant population, have recently been shown to more commonly exhibit HCV-specific CD4⁺ T cell proliferative responses than are Caucasian Americans [26], further supporting the notion that pretreatment immune responses may be predictive of nonresponse to therapy in certain patient cohorts

The question remains, however, as to why some patients with baseline immune responses were unable to convert them to ones that lead to viral clearance. Several potential hypotheses could explain our results. The persistent activity of immunoregulatory T cells in HCV-infected patients that serves to down-regulate the immune response during the chronically infected state [27, 28] could prevent the generation of an appropriate immune response. Other potential mechanisms include host factors (such as cytokine polymorphisms) [29, 30], the absence of an adequate T cell repertoire resulting from immune exhaustion [31], and the phenomenon of “original antigenic sin” [32], wherein the immune responses originally elicited by specific epitopes remain dominant, despite the evolution of variant viral epitopes. Future detailed studies will be required to determine which of these mechanisms is operant. Given the treatment-resistant advanced-disease state of our cohort, the HCV-specific immune responses we detected were, at best, ineffective—and, at worst, they served to impede the development of other, effective immune responses after re-treatment

In conclusion, we found positive baseline HCV-specific immune responses and increased QS variability to be associated with a subsequent lack of sustained virologic response in a large cohort of patients with advanced hepatic fibrosis and prior nonresponse to IFN-α therapy, indicating an association between HCV-specific immune reactivity and a pretreatment state that is resistant to successful pegylated IFN-α2a and ribavirin combination therapy. Earlier findings from our group [10] and from others [33] have suggested that active intrahepatic HCV-specific CTL responses may be responsible for ongoing liver damage and may play a role in the progression of liver disease. Together, these data suggest that, in certain situations, the presence of an immune response may not necessarily benefit the host. Clearly, the role played by the immune response in the pathogenesis of hepatitis C disease is complex, and the difference between a beneficial and deleterious immune response will need to be carefully dissected as we consider the use of immune-enhancing or -modifying therapies in the future treatment of chronic hepatitis C

Author Affiliations

Departments of Laboratory Medicine (C.M., S.J.P., D.G.S., and D.R.G.), Pediatrics (C.M.), and Medicine (D.R.G.), University of Washington, Seattle, Washington; Divisions of Infectious Diseases (R.R.) and Gastroenterology and Hepatology (A.M.D.B.), Saint Louis University School of Medicine, St. Louis, Missouri; New England Research Institutes, Watertown, Massachusetts (M.C.D.); Center for Infectious Disease and Vaccine Research and Department of Medicine, University of Massachusetts Medical School, Worcester, Massachusetts (A.L.R.); Beth Israel Deaconess Medical Center, Boston, Massachusetts (M.J.K.); Division of Digestive and Liver Diseases, University of Texas Southwestern Medical Center, Dallas, Texas (P.F.M.); Liver-Biliary-Pancreatic Center and Departments of Medicine and of Molecular and Structural Biology, University of Connecticut Health Center, Farmington, Connecticut (H.L.B.); Division of Gastrointestinal and Liver Diseases, University of Southern California, Los Angeles, California (K.L.L.)

Study Group Members

In addition to the authors of this article, the following individuals either contributed to the laboratory studies or were instrumental in the planning, conduct, and/or care of patients enrolled in this study at each of the following participating institutions:

University of Washington, Seattle, Washington: Megan Allison, Michael Austin, Michelle Gano, Rachel Life, and Christina Yoshihara

Beth Israel Deaconess Medical Center, Boston, Massachusetts: Qi He

University of Massachusetts Medical School, Worcester, Massachusetts: Kim West and Kelly A. White

University of Massachusetts Medical Center, Worcester, Massachusetts: Dawn Bombard, Maureen Cormier, Donna Giansiracusa, and Savant Mehta

University of Connecticut Health Center, Farmington, Connecticut: Michelle Kelley

Saint Louis University School of Medicine, St. Louis, Missouri: Debra King, Keith Meyer, Patricia Osmack, and Judy Thompson

University of Texas Southwestern Medical Center, Dallas, Texas: William M. Lee, Nicole Crowder, Rivka Elbein, Amanda Reeck, and Marisol Serrano

University of Southern California, Los Angeles, California: Carol B. Jones and Susan L. Milstein

New England Research Institutes, Watertown, Massachusetts: Elizabeth C. Wright