HCV Drug Pre-Existing Resistance in Protease/Polymerase

[Guest guest]

_______________________________________________

Naturally occurring dominant resistance mutations to hepatitis C virus protease and polymerase inhibitors in treatment-naïve patients

Hepatology Dec 2008

"In summary, dominant mutations described to confer resistance to STAT-C drugs in vivo or in vitro can be observed as sporadic, unrelated cases in up to 8.6% of treatment-naïve patients infected with HCV genotype 1 from North America and Europe. High viral loads in patients carrying drug resistance mutations suggest unimpaired viral replication, thus questioning previous speculations about maintained drug efficiency in the setting of pre-existing drug resistance mutations. These data provide a rationale to further characterize possible correlations between dominant STAT-C resistance mutations at baseline and treatment outcome, to evaluate a possible need for drug resistance testing and individual tailoring of drug combinations. Sequences published with this article may be used as a baseline reference for cost-effectiveness calculations for current and future anti-HCV drugs."

Kuntzen 1 *, Joerg Timm 2, Berical 1, Niall Lennon 3, M. Berlin 3, K. Young 3, Bongshin Lee 4, Heckerman 4, Carlson 4, L. Reyor 1, nna Kleyman 1, Cory M. McMahon 1, Birch 1, n Schulze zur Wiesch 5, Ledlie 2, Koehrsen 3, Chinnappa Kodira 3, D. 3, Georg M. Lauer 1, Hugo R. Rosen 6, Florian Bihl 7¶, s Cerny 8¶, Ulrich Spengler 9, Zhimin Liu 10, Arthur Y. Kim 1, Yanming Xing 10, Arne Schneidewind 1, Margaret A. Madey 10, Jaquelyn F. Fleckenstein 10, Vicki M. Park 10, E. Galagan 3, Chad Nusbaum 3, Bruce D. 1 16, Gerond V. Lake-Bakaar 11, S. Daar 12, Ira M. son 11, D. Gomperts 13, R. Edlin 11, Sharyne M. Donfield 14, T. Chung 15, H. Talal 11, Tony n 10, Bruce W. Birren 3, R. Henn 3||, Todd M. 1||

1Partners AIDS Research Center, Massachusetts General Hospital, Harvard Medical School, Boston, MA

2Department of Virology, Essen University Hospital, Germany

3Broad Institute of Massachusetts Institute of Technology and Harvard University, Cambridge, MA

4Microsoft Research, Redmond, WA

51. Medizinische Klinik und Poliklinik, Universitätsklinikum Hamburg-Eppendorf, Germany

6Division of Gastroenterology and Hepatology, University of Colorado Health Science Center, Denver, CO

7Department of Gastroenterology and Hepatology, University Hospital Geneva, Switzerland

8Clinica Moncucco, Lugano, Switzerland

9Department of Internal Medicine I, Bonn University Hospital, Bonn, Germany

10University of Tennessee Health Science Center, Hepatitis C ative, Memphis, TN

11Center for the Study of Hepatitis C, Weill Cornell Medical College, New York, NY

12Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center, Geffen School of Medicine at UCLA, Los Angeles, CA

13The Saban Research Institute, Childrens Hospital Los Angeles, Los Angeles, CA

14Department of Biostatistics, Rho Inc., Chapel Hill, NC

15Gastrointestinal Unit, Massachusetts General Hospital, Harvard Medical School, Boston, MA

16 Medical Institute, Chevy Chase, MD

Resistance mutations to hepatitis C virus (HCV) nonstructural protein 3 (NS3) protease inhibitors in <1% of the viral quasispecies may still allow >1000-fold viral load reductions upon treatment, consistent with their reported reduced replicative fitness in vitro. Recently, however, an R155K protease mutation was reported as the dominant quasispecies in a treatment-naïve individual, raising concerns about possible full drug resistance.

To investigate the prevalence of dominant resistance mutations against specifically targeted antiviral therapy for HCV (STAT-C) in the population, we analyzed HCV genome sequences from 507 treatment-naïve patients infected with HCV genotype 1 from the United States, Germany, and Switzerland. Phylogenetic sequence analysis and viral load data were used to identify the possible spread of replication-competent, drug-resistant viral strains in the population and to infer the consequences of these mutations upon viral replication in vivo.

Mutations described to confer resistance to the protease inhibitors Telaprevir, BILN2061, ITMN-191, SCH6 and Boceprevir; the NS5B polymerase inhibitor AG-021541; and to the NS4A antagonist ACH-806 were observed mostly as sporadic, unrelated cases, at frequencies between 0.3% and 2.8% in the population, including two patients with possible multidrug resistance. Collectively, however, 8.6% of the patients infected with genotype 1a and 1.4% of those infected with genotype 1b carried at least one dominant resistance mutation. Viral loads were high in the majority of these patients, suggesting that drug-resistant viral strains might achieve replication levels comparable to nonresistant viruses in vivo.

Conclusion: Naturally occurring dominant STAT-C resistance mutations are common in treatment-naïve patients infected with HCV genotype 1. Their influence on treatment outcome should further be characterized to evaluate possible benefits of drug resistance testing for individual tailoring of drug combinations when treatment options are limited due to previous nonresponse to peginterferon and ribavirin.

Hpatitis C virus (HCV) is a global health problem infecting more than 170 million persons worldwide.[1] The current standard treatment with pegylated-interferon and ribavirin is complicated by frequent adverse reactions, and a sustained virological response can only be achieved in 50% of patients infected with the most prevalent genotype 1.[2] Mutation F415Y in nonstructural protein 5B (NS5B) has been detected in patients on ribavirin monotherapy,[3][4] and similar to mutations G404S and E442G in NS5A,[5] it has been found to confer ribavirin resistance in vitro.[4] However, the clinical significance is unclear because patients exhibiting these mutations still achieved sustained virological responses on combination therapy.[3][4] No single site in the HCV genome has consistently been found to be associated with resistance to interferon, most likely because this drug acts more indirectly through several antiviral pathways.[6]

With the development of the HCV replicon system, a new tool has become available for in vitro testing of drugs that specifically inhibit HCV proteins essential for viral replication. For this new class of specifically targeted antiviral therapies for hepatitis C (STAT-C), several new compounds have been investigated in recent years, among these the NS3/4A protease inhibitors (PIs) Telaprevir (VX-950), Boceprevir (SCH503034), SCH6, BILN2061, and ITMN-191 and the NS4A antagonist ACH-806,[7] along with two types of NS5B RNA-dependent RNA polymerase (RdRp) inhibitors, the nucleoside analogs Valopicitabine (NM283), JTK-109, and R1479, and the non-nucleosides HCV-796, A-837093, A-782759, and AG-021541, some of which have recently attracted attention with promising results in primary clinical trials (reviewed in Cholongitas and Papatheodoridis[8]). However, the rapid selection of viral variants displaying drug-resistant phenotypes has been observed in patients experiencing viral rebound during treatment as well as in replicon experiments. Variants resistant to STAT-C reported to date are summarized in Table 1.

Consensus amino acids are numbered according to their position in NS3, NS4A or NS5B (column 2) or the full HCV genome (column 3) relative to the H77 HCV genotype 1a reference sequence. Amino acid substitutions selected during drug treatment in patients are listed separately from substitutions observed to confer drug resistance in replicon experiments or enzymatic assays. Drugs inducing resistance mutations and relevant references are given. Resistance mutations detected as the dominant viral sequence in treatment naïve patients in this study are listed on the right, and their proportion among genotype 1a or 1b infected patients is shown. (Other observed variants not described to confer resistance included A39T, Q41H and S489A/F [NS3], E442D [NS5A] and M423A [NS5B] in genotype 1a, and S489V [NS3], G404C/D/N [NS5A], C316N/S and M414I [NS5B] in genotype 1b.)

* Includes one patient who carries two mutations (V36M and V23A).

Due to the high replication rate and the error-prone nature of the viral RdRp, HCV circulates within a given host as a mixture of viral strains each differing slightly from one another, referred to as the viral quasispecies. Two recent reports suggested that resistant variants may already be present at frequencies <1% in the quasispecies population in treatment naïve patients,[9][10] consistent with their dominant emergence only days after treatment initiation.[11][12] However, drug treatment in the setting of resistance mutations may still be beneficial, because a decreased in vitro replicative capacity has been demonstrated for many viral strains resistant to protease or RdRp inhibitors.[13-19] Patients may therefore benefit from elimination of the dominant, drug-susceptible viral quasispecies with more than 1000-fold reductions in their viral load, until only drug-resistant but replication-deficient quasispecies constitute the residual viral population.[11][12] Although in some cases replication levels may later be restored by compensatory mutations, it seems possible that this effect could suffice to achieve treatment success if several drugs were combined to suppress viral replication before compensatory mutations or additional resistance mutations could evolve. Despite the presumed impaired viral replication, however, the PI-resistant mutation R155K was recently detected as the dominant quasispecies in a treatment-naïve patient,[20] raising concerns that drug resistance may not necessarily be associated with reduced viral fitness in the individual in vivo situation. Consequently, preliminary data indicate reduced early treatment responses in patients with 100% R155K PI-resistant quasispecies at baseline, indicating possible primary drug resistance.[21]

Because this finding may have important implications for the response rates to new drugs in the population with a possible need for drug resistance screening, we have investigated the prevalence of dominant mutations described to confer protease or polymerase inhibitor resistance in an international multicenter cohort of treatment-naïve patients infected with HCV. Phylogenetic sequence analysis and viral load data were used to identify the possible spread of replication-competent, drug-resistant viral strains in the population and to infer the possible consequence of these mutations upon replicative capacity in vivo.

Discussion

In this study, we analyzed the prevalence and distribution patterns of dominant drug resistance mutations in 507 treatment-naïve patients infected with HCV genotype 1. Dominant mutations were mostly observed as sporadic, unrelated cases at frequencies between 0.3% and 2.8% in the population. Taken together, however, 8.6% of patients infected with genotype 1a and 1.4% of patients infected with genotype 1b exhibited at least one drug resistance mutation, including two cases with possible multidrug resistance. Viral loads were high in the majority of patients carrying these mutations, suggesting that resistant viruses might achieve replicative capacities comparable to nonresistant strains in vivo.

STAT-C drugs offer new opportunities to overcome limited response rates to interferon and ribavirin. Treatment success as measured by decreasing viral loads represents a balance between the achievable drug concentration in the plasma, the frequency and degree of drug resistance among the viral quasispecies,[21] and the replication efficiency of drug-resistant viral strains. Numerous studies have investigated the impact of amino acid substitutions on viral replication and drug susceptibility to investigational NS5B polymerase or NS3/4A protease inhibitors in the replicon model. Here, a pronounced reduction in the replicative capacity was described for the highly resistant A156 variants in NS3 and M423 variants in NS5B, and also to a lesser extent for the low-level PI-resistant R155, T54, and V36 mutations.[12][13][19][27] It has been speculated that these lower replication levels could facilitate eradication through combination treatment,[11] because STAT-C resistant viral strains appear to remain sensitive to interferon and ribavirin.[19] However, the high viral loads observed in patients with dominant drug resistance mutations in our study demonstrate that viral fitness may be unimpaired in the individual in vivo situation, suggesting possible limitations of in vitro studies for the assessment of viral fitness. Possibly, as yet unknown secondary mutations may also be present which might compensate any replication defects. Unfortunately, the low number of individual resistance mutations in our study prevented the identification of candidate residues for compensatory mutations.

As far as comparative data on drug susceptibility are available, mutations observed in our study at residues M423 in NS5B,[27] and in some cases R155 in NS3,[13][17-19][28] were described to confer high levels of drug resistance compared to the wild-type virus in the replicon model, whereas amino acid substitutions at residues V36, T54, and D168 resulted in relatively moderate resistance.[13][19][28] In a recent clinical trial, the in vivo relevance of these findings was supported by the observation that weakly Telaprevir-resistant variants arose predominantly in patients who only achieved lower plasma drug levels, whereas increasing drug concentrations selected for mutants that were also highly resistant in vitro.[12] Although this suggests that viruses carrying low-level resistance mutations can be eradicated at high drug concentrations, the clinical impact of dominant amino acid substitutions observed in our study is unclear, because in vivo data published so far have originated from patients with resistance only in subdominant quasispecies. However, preliminary results indicate that dominant resistance mutations can potentially reduce the early treatment response to STAT-C drugs,[21] and it must be assumed that low-level resistant strains will sustain viral replication in patients who do not achieve optimal drug levels with standard dosing, in cases with poor adherence, or when dose reductions are inevitable due to adverse events.

Importantly, continued viral replication in the presence of the selecting drug would put the patient at risk to develop additional resistance mutations. Here, observations from HIV infection suggest that baseline resistance even against only one drug in a multidrug antiviral regimen may jeopardize treatment success,[12] and further indicate that apart from single mutations conferring high-level resistance, stepwise accumulation of subtle but synergistically acting resistance mutations may also eventually lead to treatment failure.[30] Data from a recent study using Telaprevir has indicated a similar pathway in HCV infections, where combination of the low-level resistance mutations V36M and R155K resulted in a highly drug-resistant phenotype, the appearance of which coincided with viral breakthrough.[11][12] In our treatment-naïve patients, we did not detect any combinations of resistance mutations against the same drug, neither in the bulk sequences nor by clonal analysis. However, although the odds for simultaneous occurrence of two specific resistance mutations in the same viral strain has been calculated at 10-8 to 10-10, the chance to add a single mutation to a pre-existing, dominant quasispecies is at least 1000-fold higher.[19] In this context, our finding that up to 8.6% of patients infected with genotype 1a carry dominant low-level resistance mutations (including 3.0% with V36M/L or R155K) may be relevant, because these patients would be expected to have a significantly lower genetic barrier to developing full resistance during drug treatment.[30] Therefore, a combination of drugs might be necessary to rapidly suppress viral replication before additional resistance mutations can evolve. Unfortunately, no combination therapy is currently available for peginterferon nonresponders, the patient group that needs new treatment options most urgently. Novel STAT-C drugs as alternative treatment combinations appear to be prone to viral resistance mutations, likely due to their direct interaction with HCV proteins. Because many mutations in NS3 confer cross-resistance to several PIs, the options to combine drugs from this currently more advanced STAT-C class are limited. This problem might be overcome by the addition of polymerase inhibitors or NS4A antagonists. However, with M423I/V and A39V, we have already identified dominant resistance mutations to these classes of drugs in treatment-naïve patients. In addition, combinations of mutations V36M/V23A or M423I/F415Y that confer resistance to the PIs Telaprevir and ITMN-191 or to the polymerase inhibitor AG-021541 and ribavirin indicate possible multidrug resistance.

In HIV infection, resistance testing was calculated to be cost-effective when the prevalence of drug resistance becomes >1% at baseline[31] and is currently recommended in areas with more than 5% prevalence of resistant strains, and in all cases of treatment failure.[32] For novel STAT-C drugs, important factors such as their cost and treatment response rates are not yet available to derive similar calculations, as is detailed information about the effectiveness of these therapies in the face of pre-existing resistance mutations. However, with a baseline frequency of drug resistance mutations between 5.0% and 8.6% in patients infected with genotype 1a in our study, and possibly increasing rates due to transmission of resistant variants once new drugs are widely used, it seems possible that pretreatment screening for baseline resistance mutations might be warranted to enable the individual tailoring of specific drug combinations for those patients whose treatment options are limited because of previous nonresponse to peginterferon and ribavirin.

Different rates of resistance mutations between HCV genotype 1a and 1b have previously been observed in patients on Telaprevir monotherapy.[11][12] Here, mutations V36M or R155K require only one nucleotide change from the genotype 1a consensus sequence, but two substitutions in genotype 1b, resulting in a relative genetic barrier to resistance.[11] Different constraints on viral evolution due to different protein structures between viral subtypes may provide alternative explanations. Although clonal sequence analysis in a subset of patients demonstrated a high concordance with the bulk sequencing results, it is also likely that our approach underestimates the number of patients with relevant proportions of resistance mutations in their quasispecies, which may have randomly skewed mutational frequencies toward genotype 1a. Finally, mutations T54S, R155K, and M423I/V fall within the described CD8+ T cell epitopes CINGVCWTV, HAVGLFRAA and ARMILMTHF, and it would be intriguing to speculate that viral evolution driven by escape from CD8+ T cell pressures might differentially influence the prevalence of drug resistance mutations.

In summary, dominant mutations described to confer resistance to STAT-C drugs in vivo or in vitro can be observed as sporadic, unrelated cases in up to 8.6% of treatment-naïve patients infected with HCV genotype 1 from North America and Europe. High viral loads in patients carrying drug resistance mutations suggest unimpaired viral replication, thus questioning previous speculations about maintained drug efficiency in the setting of pre-existing drug resistance mutations. These data provide a rationale to further characterize possible correlations between dominant STAT-C resistance mutations at baseline and treatment outcome, to evaluate a possible need for drug resistance testing and individual tailoring of drug combinations. Sequences published with this article may be used as a baseline reference for cost-effectiveness calculations for current and future anti-HCV drugs.

Results

Within a larger cohort study, we sequenced the entire HCV proteome from 507 patients infected with HCV genotype 1 recruited from multiple clinical centers in the United States, Germany, and Switzerland. Demographic characteristics are given in Table 2. To investigate the prevalence of described drug resistance mutations to established or investigational anti-HCV drugs in the population, we analyzed dominant amino acid substitutions separately for the 362 patients infected with genotype 1a and the 145 infected with genotype 1b.

Prevalence of Dominant Mutations Described to Confer Resistance to Ribavirin, NS5B Polymerase, and NS3 Protease Inhibitors in a Treatment-Naïve Patient Population.

In searching for dominant drug-resistant variants, we first screened sequences for substitutions in positions G404 and E442 in NS5A and position F415 in NS5B, which have been associated with exposure to ribavirin.[4][5] Here, in the study population infected with genotype 1a, only one (0.3%) patient displayed the dominant F415Y substitution and one (0.3%) patient expressed substitution E442G. Similarly, only two (1.4%) patients with G404S substitutions and two (1.4%) patients exhibiting the E442G mutation were observed among the patients infected with genotype 1b. Thus, dominant ribavirin-associated mutations appeared uncommon and therefore unlikely to explain frequent nonresponse to combination therapy.

Further examination of viral sequences in NS5B revealed that variation in positions previously associated with resistance to the RdRp inhibitors A-837093, A-782759, R1479, 2-C-methyl-ribonucleosides, and JTK-109 was similarly rare. With the exception of substitutions C316N/S and M414I in genotype 1b that have not been described as resistance variants, the residues H95, S96, S282, S365, N411, Y448, P495, G554, and D559 were 100% conserved in our patient population. In contrast, however, mutations M423V and M423I, which conferred a more than 44-fold decrease in susceptibility to the non-nucleoside polymerase inhibitor AG-021541,[27] were present in 10 (2.8%) of the patients with genotype 1a infection (Table 1), including one individual additionally carrying an F415Y mutation associated with ribavirin resistance.

In NS3, among the 362 patients infected with genotype 1a HCV,we observed three (0.8%) individuals carrying the R155K mutation that exhibits moderate to high levels of drug resistance to the protease inhibitors Telaprevir, BILN2061, ITMN-191, and Boceprevir.[11-13][19][28] Mutations V36L and V36M conferring low-level resistance were found in six (1.7%) and two (0.6%) patients, respectively, with mutation T54S observed in an additional seven (1.9%) subjects.[12][13] Of note, none of our study subjects showed substitutions at position A156 that is known to confer the highest level of resistance to Telaprevir, BILN2061, Boceprevir, or SCH6,[12][17][18] consistent with the reported high impact of these mutations on the replicative capacity in vitro.[11-13][18][19][28] Mutations D168E and V23A that confer reduced susceptibility to ITMN-191[29] were found in one (0.3%) and two (0.6%) patients, respectively. Remarkably, in one of these individuals, the V23A substitution was combined with mutation V36M that confers resistance to Telaprevir and Boceprevir.[11][12][19][28] Last, mutation A39V, which is associated with resistance to the NS4A antagonist ACH-806, was observed in one (0.3%) patient. Beyond that, no variants associated with drug resistance in vitro were observed in positions Q41, F43, R109, S138, V170, and S489. In the 145 patients infected with genotype 1b, only the T54S substitution was detected in two (1.4%) individuals (Table 1).

Taken together, various described drug-resistance mutations were detectable by bulk sequencing at individual frequencies between 0.3% and 2.8% in treatment-naïve patients infected with HCV. Two subjects each carried two amino acid substitutions conferring resistance to different drugs. Collectively, NS3 protease or NS5B polymerase inhibitor resistance mutations were prevalent at 8.6% (31 of 362 patients) in those infected with genotype 1a and 1.4% (2 of 145 patients) in those infected with genotype 1b. When the analysis was more conservatively restricted to resistance variants described to be selected during drug treatment in vivo, 5.0% of the genotype 1a isolates and 1.4% of the genotype 1b isolates harbored a resistance variant (Table 1). However, mutations in NS3 position A156 conferring >100-fold resistance to most PIs were undetectable by bulk population sequencing, consistent with their reported high impact on the replicative capacity in vitro.

Unimpaired In Vivo Fitness of Viral Strains Carrying Resistance Mutations.

Because most protease and polymerase inhibitor resistance mutations described to date have been associated with reduced replicative capacity of varying degrees,[11-19] it seemed remarkable to find these as dominant viral sequences in treatment-naïve patients. Moreover, the emergence of double mutants in positions V36 and R155 has been reported in patients experiencing viral rebound during Telaprevir monotherapy,[11][12] with double mutant viruses displaying enhanced replication[12] while also exhibiting increased drug resistance compared to single mutant viral strains.[12][28]

To confirm that amino acid substitutions detected in our bulk sequences indeed represented the dominant quasispecies in the patients, and to evaluate whether these were present in combinations with other drug resistance mutations at subdominant levels, we analyzed 153 clones from 11 patients generated from new, independent RT-PCR products spanning the NS3 protease catalytic domain. Here, for all sites of drug resistance, the bulk sequence data was correctly representative of 100% of the viral quasispecies in 9 of 11 patients and of 13 of 14 and 14 of 15 clones in the remaining two patients, thus confirming the detected mutations as the predominant viral strain in each patient. Within the limited clonal depth applied here we did not, however, detect any additional combinations of drug resistance mutations on a subdominant level.

In examining viral replication levels in patients carrying dominant drug resistance mutations, 19 of 33 (58%) of these displayed viral loads in the range of >500,000 to 17 × 106 IU/mL (Fig. 1), including two patients with R155K substitutions and one patient with the V36M/V23A double mutant, indicating that drug-resistant strains were not necessarily impaired in their ability to replicate in vivo.

Sporadic Nature of Naturally Occurring Drug Resistance.

We next asked whether the presence of naturally occurring protease or polymerase inhibitor resistance mutations was influenced by transmission of resistant viral strains, high-risk behavior, human immunodeficiency virus (HIV) coinfection, race, sex, or geographic location. In examining relationships between amino acid sequences from patients with identical drug resistance mutations, only two viral strains carrying the M423I mutation resistant to the polymerase inhibitor AG-021541 clustered closely together in a phylogenetic analysis (Fig. 2). Upon closer examination, these differed in 50 of 3012 (1.7%) amino acid residues, arguing against a recent transmission event, but compatible with transmission or infection from a common source further in the past.[24] Both samples came from injection drug users from the same small town in Switzerland and were drawn 3 years apart from each other, suggesting possible transmission and long-term stability of drug resistance mutations. More distant relationships were observed in two other clusters in the tree between sequences from two patients from Boston sharing the M423I mutation, and between sequences from three patients from different cities in Switzerland carrying the V36L mutation (Fig. 2). These differed by 105-130 amino acids and were thus not suggestive of direct transmission. Beyond that, no other viral strains sharing resistant variants displayed similar close relationships, with resistant mutants spread across the entire tree (Fig. 2). Among patients carrying drug resistance mutations, the proportions of transmission risk factors, HIV/HCV coinfection, male sex, and ethnic groups were similar to those observed in the entire cohort, and drug resistance mutations were present in patients from all involved international HCV cohorts from the United States, Germany, and Switzerland. Thus, whereas transmission of a resistance mutation seemed possible between two patients in our cohort, the majority of resistant variants represented sporadic cases rather than locally spread, closely related viral strains or mutants specifically selected within a certain subgroup of patients.

Patients and Methods

Patients.

Plasma samples were obtained from 507 patients who had been infected with HCV genotype 1a or 1b for at least 1 year and were naïve to anti-HCV treatment with interferon, ribavirin, or any novel or investigational anti-HCV drug. Patients were recruited between 1989 and 2007 from the Massachusetts General Hospital, the Lemuel Shattuck Hospital, and the Fenway Community Health Center (Boston, MA), the University of Tennessee Health Science Center (Memphis, TN), the Center for the Study of Hepatitis C, Weill Cornell Medical College (New York, NY), the Hemophilia Growth and Development Study Cohort (USA),[22] the University of Colorado (Denver, CO), the Swiss HCV Cohort Study (Switzerland),[23] and the Bonn University Hospital (Bonn, Germany). The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the Institutional Review Board at Massachusetts General Hospital. Written informed consent was obtained from each study participant.

Amplification and Sequencing of Full-Length HCV Genomes.

Viral loads were determined using the Cobas HCV Amplicor Monitor test (Roche), the Cobas TaqMan test (Roche), or Versant 2.0 or 3.0 bDNA assays (Bayer). Viral RNA was extracted from plasma samples using the vRNA extraction kit (Qiagen, Hilden, Germany). Complementary DNA was amplified and sequenced using a nested reverse transcription polymerase chain reaction (RT-PCR) approach as published elsewhere,[24] with two additional first-round primer pairs 13F_5-CCTYGCCTACTATTCCATGG-3, 13R_5-TGAGCRCGYACRAAGTACGGC-3 and 14F_5-TCTAYGGCAARGCCATCCCC-3, 14R_5-AGGAGCTTGGACTGGAGCC-3 to achieve better coverage in genotype 1b samples. At Massachusetts General Hospital, 89 samples were sequenced to a final average coverage of three-fold. For the remaining 418 samples, products of the first-round RT-PCR were transferred to the Broad Institute NIAID Microbial Sequencing Center. Here, 96 amplicons 500-800 nucleotides in length were generated for each sample and sequenced bidirectionally using an automated high-throughput viral sequencing platform comprising of ABI3730xl sequencers, resulting in an average sequence coverage of eight-fold for each genome.

Sequence Analysis and Phylogeny.

At Massachusetts General Hospital, the Sequencher (Gene Codes Corp., Ann Arbor, MI) program was used to trim, align, and edit sequences. At the Broad Institute, sequences were trimmed for quality and primer sequence, and contigs were assembled using the Broad Institute's ARACHNE assembler[25] (https://www.broad.mit.edu/wga/). If mixed bases were detected as two different chromatogram peaks at the same residue, only the dominant base was called after evaluation of all overlapping fragments. Sequences with insertions or deletions compared to a reference were validated manually. Genbank accession numbers for all sequences are provided in Supporting Table 1. Full-length nucleotide sequences were translated into amino acids and aligned using Se-Al v2.0a11 software (http://evolve.zoo.ox.ac.uk). Phylogenetic trees were constructed in ClustalX[26] using the neighbor joining method and were visualized in FigTree v1.1.2 (http://tree.bio.ed.ac.uk) and Det (http://research.microsoft.com/vibevis/det..).

Generation of Clonal Sequences.

Procedures are described in the Supporting Methods.

_______________________________________________