Failure of interferon plus ribavirin therapy for HCV can occur in two different patterns: nonresponse and relapse. HCV is highly diverse genetically, and this diversity could affect how the virus responds to therapy. Furthermore, HCV can evolve rapidly, and hence therapy could drive evolution of the virus by selecting relatively resistant variants. We previously found that high genetic diversity in genotype 1a pre-therapy NS3 and NS5A sequences correlated robustly with response to therapy 
. The previous study addressed variability related to both response to therapy and the race of the patient. Here, we employed the pre- and post-therapy sequences from genotype 1a Virahep-C treatment failures to evaluate both the extent to which variations in viral protein sequences may affect the pattern of failed therapy and to determine the effects of failed therapy on the viral sequences. Issues of variability associated with the race of the patient were not addressed in this study because the low number of samples available does not provide enough power to make useful comparisons.
To evaluate how the viral protein sequences changed during treatment, we examined the number and pattern of amino acid mutations that occurred in each sequence during therapy. We expected to find a relatively high mutation frequency due to HCV's high genetic plasticity, but contrary to our expectations, the 0.9% mutation frequency observed in the samples was relatively small in relation to the variability among independent HCV isolates (10–12%), and was in the range of quasispecies variability typically found in a given individual (~1–4%). Furthermore, about one third of the mutations we observed were in E2, especially in HVR1, so the mutation rate of the HCV ORF outside E2 was just 0.59%. Thus, a primary conclusion of this study is that the HCV consensus sequence is relatively stable during failed antiviral therapy. Most previous studies of HCV evolution during therapy have focused on small regions of the genome, primarily in NS5A and E2, and especially on the highly variable regions in these genes 
. These regions were chosen because their high variation makes them ideal for evolutionary analyses. However, focusing on theses regions had the unintended effect of helping to form the perception that HCV sequences are similarly mutable throughout the genome.
The second major conclusion from this study was that genetic differences in NS2 correlated with the pattern of failed response to interferon-based therapy. These correlations were evident in the number of pre-therapy unique variations (), the pre-therapy protein distance (), and the nature of mutations that occurred during therapy (). Relapsers had higher variability in NS2 than nonresponders in pre-therapy samples, and mutations that occurred in NS2 over the course of therapy were more likely to affect the function of the protein in relapsers than in nonresponders. Therefore, variability in NS2 may help determine whether a patient will be a nonresponder or relapser. However, this possibility is difficult to interpret at a functional level because the roles of NS2 in viral replication and pathology are poorly understood. NS2 has been shown to inhibit the interferon response when expressed in cells 
, and the observed differences in variability could lead to differences in the effectiveness of this inhibition. Alternatively, since NS2 is involved in protein processing and is required for virion formation 
, variability could also affect the effectiveness of viral protein processing and virus production. Further study of NS2, including identifying possible cellular targets, may clarify how NS2 affects response to interferon-based therapy.
Despite the overall stability of the HCV sequences, a substantial number of mutations did occur during failed therapy, but no significant difference was observed in the number of mutations between the nonresponders and relapsers (). We had hypothesized that relapsers would evolve more than nonresponders as the virus adapted to the pressures of therapy. This was not observed at the level of the number of mutations, possibly because both groups passed through at least a weak bottleneck. However, despite equivalent numbers of mutations in the relapser and nonresponder groups, the overall intra-group genetic distance in the relapsers increased while it did not change in nonresponders (). Therefore, the mutations in relapsers created new sequences, whereas those in nonresponders largely alternated between sequences already present within the group.
To determine which sequence motifs may have been under positive selection during therapy, we examined the dN/dS ratio in a sliding window of ten codons. Many regions of strong positive selection were observed (), but the number and distribution of regions of positive selection were similar in the nonresponder and relapser sequences. This implies that both nonresponders and relapsers evolved to similar degrees under the pressures induced by therapy and that the targets of the selective pressure were broadly distributed throughout the polyprotein. The exception to this pattern was in E2, where there were more regions of positive selection in nonresponders than in relapsers. As E2 is a primary target of humoral immune responses, this difference may be due to the difference in neutralizing antibody titers throughout therapy. Brown et al. showed that E2 and not E1 evolve in chronically infected patients in solvent exposed regions 
, and our data show that there were differences in evolution of the nonresponder and relapser sequences. Since the nonresponders had relatively high viral titers throughout therapy, the humoral immune response may have been be constantly stimulated by a relatively high antigen load, leading to an evolving humoral pressure. In relapsers, viral titers declined below the detection limit, and the humoral immune response may have declined during therapy due to the drop in antigen load. These analyses compared pre- and post-therapy sequences, and the patterns of evolution observed in samples during therapy may be different than those that were prevalent in post-therapy samples. Therefore, studies of samples from sustained viral responders, nonresponders, and relapsers at early time-points during therapy, such as 2 or 4 weeks, could be useful in further understanding the evolution of these groups in response to interferon-based therapy. However, the relatively small number of changes observed in the viral consensus sequence between the pre- and post-therapy time points implies that a detailed quasispecies analysis over the early phases of therapy would be needed to substantially advance this understanding.
Genetic bottlenecks can cause a constriction of the genetic variability within a population. In HCV, this is reflected in the breadth of the quasispecies within an individual. We expected the difference in the strength of the bottlenecks experienced by nonresponder and relapser groups to cause a greater decline in the quasispecies breadth in the relapsers. We found that the intra-patient quasispecies breadth declined in both nonresponders and in relapsers, and that the decline in relapsers was greater than in nonresponders, but this difference did not reach statistical significance (). Other groups have shown that the breadth of the quasispecies is correlated with response to interferon-based therapy 
. Our study indicates that changes in the breadth of the quasispecies also correlated with the difference in response between nonresponders and relapsers.
This study was designed to assess that role of HCV genetic variation at the protein level on outcome of therapy. Variability in the RNA itself can also predicted to affect the response to therapy by altering the RNA structure or the interactions of the RNA with host proteins and/or the viral replication machinery. RNA elements associated with protein binding could occur anywhere in the viral RNA, but they are most likely to occur in the 3′ and 5′ UTRs since these areas are known to contain the promoters for viral replication and the viral internal ribosome entry site. The sequences obtained for this study include part of the UTRs, but these sequences are of varying length, and in some cases are absent. This precludes meaningful analysis of these samples outside the ORF.
This study is the largest examination to date of genetic changes in the full-length viral open-reading frame during interferon-based therapy, and it is the first study comparing nonresponders to relapsers in genotype 1a. Previous studies have examined the changes in patient samples over the course of therapy. Enomoto identified the ISDR 
by examining pre- and post-therapy sequences in full-length sequences from three nonresponding genotype 1b infected patients. The three patients also had many mutations scattered elsewhere throughout the structural and nonstructural genes. We saw similar patterns of mutations in the 1a sequences. Other studies of evolution during therapy focused on smaller portions of the genome. Vuillermoz examined the changes in genotype 1b responders, nonresponders, and breakthrough patients in E2, NS5A, and NS5B 
. They showed higher mutation rates in responders in the V3 region of NS5A (amino acids 2356-2379) and conservation of the PePHD region in E2 in all samples. We also found no mutations in the PePDH region, but unlike Vuillermoz, we did not find a difference between nonresponders and relapsers in the V3 region. Differences in the observed numbers of mutations between our study and the Vuillermoz study could be due to the different genotypes studied or the different definitions of the response types analyzed. Evolution of HCV during therapy has also been noted in NS5B 
, NS5A 
, and in the structural proteins, especially HVR1 
. While correlations between diversity and evolution between relapsers and nonresponders were noted in some of these studies, others showed no difference between the groups. We did not observe significant differences in NS5A or NS5B between the nonresponders and relapsers but we did find many regions of positive selection in E2 as well as HVR1, similar to earlier studies.
Genetic variability between HCV genotypes can lead to difference in response to therapy (e.g. genotype 1 vs. 2), and we have shown that variability differences are also associated with early response to therapy within a given viral subgenotype 
. Here, we divided the nonresponders into two phenotypes, nonresponders and relapsers, and found a spectrum of diversity associated with failed response to antiviral therapy, where relapsers fell between responders and nonresponders. We interpret this pattern to indicate that viral variability forms a continuum from sequences that are close to an “optimal” sequence to sequences that are more divergent. The optimal sequence would be most resistant to the effects of therapy, and the degree of resistance would decline with genetic distance from the optimum. Those samples that were furthest from the optimum sequence would be unable to withstand the super-physiological interferon response induced by therapy, and hence would be cleared. Therefore, although variability of the virus is clearly not the only factor affecting response to therapy, it appears to play an important role in determining the pattern of response of HCV to interferon-based therapy. Further sequencing of isolates that are nonresponders to therapy and characterization of these sequences in in vitro studies could reveal this optimum sequence, and in vitro studies of this sequence could reveal how HCV inhibits the type 1 interferon response. Understanding how the HCV proteins are involved in resistance to interferon and ribavirin could identify new drug targets that improve or replace the current therapy.