|Home | About | Journals | Submit | Contact Us | Français|
Filibuvir (PF-00868554) is an investigational nonnucleoside inhibitor of the hepatitis C virus (HCV) nonstructural 5B (NS5B) RNA-dependent RNA polymerase currently in development for treating chronic HCV infection. The aim of this study was to characterize the selection of filibuvir-resistant variants in HCV-infected individuals receiving filibuvir as short (3- to 10-day) monotherapy. We identified amino acid M423 as the primary site of mutation arising upon filibuvir dosing. Through bulk cloning of clinical NS5B sequences into a transient-replicon system, and supported by site-directed mutagenesis of the Con1 replicon, we confirmed that mutations M423I/T/V mediate phenotypic resistance. Selection in patients of an NS5B mutation at M423 was associated with a reduced replicative capacity in vitro relative to the pretherapy sequence; consistent with this, reversion to wild-type M423 was observed in the majority of patients following therapy cessation. Mutations at NS5B residues R422 and M426 were detected in a small number of patients at baseline or the end of therapy and also mediate reductions in filibuvir susceptibility, suggesting these are rare but clinically relevant alternative resistance pathways. Amino acid variants at position M423 in HCV NS5B polymerase are the preferred pathway for selection of viral resistance to filibuvir in vivo.
Hepatitis C virus (HCV) is a leading cause of chronic liver disease, liver cancer, and cirrhosis, with an infection prevalence of 2 to 3% worldwide (9). Approximately 70% of people who become infected with HCV are unable to clear the virus naturally and develop a chronic infection. The current treatment regimen requires 6 to 12 months of combination therapy with pegylated interferon (pegIFN) (weekly injections) and ribavirin (RBV) (daily oral dosing). This standard of care results in cure rates of ~50% in patients with the most common genotype virus (genotype 1) and is associated with multiple side effects (2, 6, 13). Therefore, new, improved treatments are urgently needed.
Multiple candidate small-molecule therapeutics that specifically target the virus (direct-acting antivirals [DAAs]) are currently in development, including inhibitors of the HCV NS3 protease, NS5A protein, and NS5B RNA-dependent RNA polymerase (10, 22). The most advanced compounds, telaprevir and boceprevir, are NS3 protease inhibitors and have been approved for use in combination with pegIFN and RBV. In phase 3 clinical studies, these compounds significantly increased rates of sustained viral response (SVR) compared to the standard of care alone in both treatment-naïve and treatment-experienced patient populations (16). However, both compounds have significant side effects, including severe rash and anemia, which contribute to the overall adverse-event burden associated with treatment. Patients failing therapy also typically develop drug class resistance that may prevent effective retreatment with other HCV protease inhibitors (19). Therefore, more tolerable and efficacious DAAs with unique mechanisms of action are needed. Ultimately, it is hoped that combinations of multiple DAAs targeting different parts of the viral life cycle will provide an oral regimen and a new standard of care for hepatitis C virus infection.
Filibuvir (PF-00868554) is a novel, selective nonnucleoside inhibitor (NNI) of the HCV NS5B RNA-dependent RNA polymerase that binds noncovalently in the “thumb 2” pocket (12, 21). Filibuvir potently inhibits HCV genotype 1 polymerases, with a mean in vitro 50% inhibitory concentration (IC50) of 19 nM (21). Filibuvir also exhibits potent antiviral activity against subgenomic HCV replicons in cell culture assays; it is active against a majority (95.8%) of genotype 1 replicons (mean 50% effective concentration [EC50], 59 nM) with similar potencies against replicons containing patient-derived subtype 1a and subtype 1b NS5B sequences (21). In two phase 1b clinical studies in HCV genotype 1-infected patients, filibuvir potently inhibited viral replication in a dose-dependent manner (23), and doses of ≥450 mg twice a day (BID) resulted in >2.0-log-unit reductions in HCV RNA. Filibuvir was well tolerated, with no discontinuations, serious adverse events, or deaths reported. Its safety and efficacy are currently being assessed in combination with pegIFN and RBV in a phase 2b study.
Resistance to filibuvir has been selected in vitro. Serial passaging of Huh7 cells containing subgenomic HCV replicons in suboptimal compound concentrations selected for replicons with amino acid changes in the NS5B thumb 2 region (21). The M423 codon was the predominant site of mutation (90% of sequences analyzed), and M423T was the most prevalent amino acid change, accounting for 76% of the mutants analyzed. Replicons encoding these M423 mutations demonstrated a 706- to >2,202-fold increase in the filibuvir EC50, confirming their relevance to viral escape from filibuvir.
The aim of this study was to characterize filibuvir-resistant NS5B variants selected in HCV genotype 1-infected individuals enrolled across two filibuvir monotherapy studies (23). Samples from patients at baseline, end of therapy, and follow-up were analyzed genotypically for the NS5B sequence and phenotypically for filibuvir susceptibility. Site-directed mutagenesis was also used to confirm the impacts of specific mutations on filibuvir susceptibility and replicative capacity (RC) relative to the wild type (WT).
The details of the two phase 1b clinical studies in which HCV-infected patients were dosed with filibuvir (A8121002 and A8121006) are described elsewhere (23). Treatment groups covered ascending doses of the compound and varied between 3 and 10 days in these studies (8 days in most subjects). All patients provided written informed consent before study participation. The protocol and informed-consent forms were approved by independent ethics committees at all study centers in accordance with national procedures (23). The studies were conducted according to the ethical principles in the Declaration of Helsinki and in compliance with all good clinical practice guidelines and local laws and regulations (24).
Plasma samples were collected from all patients (before, during, and after therapy) and stored at −80°C until analysis. Extraction of viral RNA from plasma was performed using the QIAamp Ultrasens Virus Kit (Qiagen, Germantown, MD). HCV cDNA was prepared by reverse transcription, and the NS5B gene was amplified using nested PCR. Population sequencing was performed across the entire NS5B coding sequence, using multiple primers targeting both cDNA strands; these data were used to generate a single consensus sequence according to published protocols (15). An internally written Perl program script was used to translate the nucleotide sequences to determine every possible encoded amino acid at each residue position. For each individual patient, all translated sequences were aligned to each other and to genotype 1a (H77; NC_004102) or genotype 1b (Con1; AJ238799) reference sequences, as determined by the genotype of the patient's HCV. Listings of amino acid changes from the reference sequence and from baseline were generated for all sequences for each patient.
Genotyped full-length genotype 1a and 1b NS5B amplicons from 20 patients were cloned into an HCV genotype 1b (strain Con1) replicon system for phenotypic analysis using a transient-transfection protocol, together with an additional nine baseline samples (see Table S1 in the supplemental material). Standard population-based cloning procedures were employed to introduce the genotyped NS5B amplicons into the BB7M4hRLuc.ribo-Sbf.Pac HCV subgenomic replicon vector (21). Additionally, site-directed mutations (SDMs) were created in the NS5B gene carried in the parent replicon vector using the Quikchange II Mutagenesis Kit (Agilent Technologies, Santa Clara, CA) according to the manufacturer's instructions. Mutations introduced into the WT (Con1) NS5B gene were those identified during filibuvir phase 1b studies (R422K, M423A/I/T/V, M426A/L/T/V, and V494I), together with those previously shown to affect susceptibility to other thumb 2 binding compounds (L419I/M/S, I482L/S/T, and V494A/T). Nucleotide sequencing confirmed all desired constructs.
To determine the in vitro RC of RNA derived from the engineered replicon plasmids and to assess the susceptibility of the replicating subgenomic RNA to HCV DAAs in tissue culture cells, RNA was first generated from the BB7M4hRLuc.ribo-Sbf.Pac-based vectors using the T7 Megascript Transcription Kit (Ambion, Austin, TX). The resulting RNA concentration and quality were assessed by spectrophotometry and gel electrophoresis. Replicon RNA (10 μg) was electroporated into 5 × 106 Huh7.5 cells (21) using an Amaxa Nucleofector II (Lonza AG, Allendale, NJ) according to the manufacturer's instructions. Following electroporation, 9.6 × 103 cells were seeded into each well of a 96-well white culture plate. Twenty-four hours later, compounds were added to the cells to give appropriate final assay concentrations in 1% dimethyl sulfoxide (DMSO). Plates seeded for EC50 determination were harvested 72 h later. For determination of RC, plates were set up in the absence of compound and harvested at 4 h (input) and 96 h. In all cases, luciferase activity was assayed using a Renilla luciferase assay kit (Promega, Madison, WI) as directed by the manufacturer's instructions. The RC was determined as the 96-h to 4-h relative light unit (RLU) ratio in comparison to the luciferase activity of the WT Con1 replicon.
For each replicon population, the dose-response curve for BILN-2061 was assessed initially. A single dose-response curve was fitted (GenStat v.11; VSN International) to the percent inhibition responses using nonlinear regression with ordinary least squares. When the BILN-2061 dose-response curves passed assay validation criteria (coefficient of variation, <75%; sigma, <25; and R2, >70), the dose-response curve for filibuvir was assessed. The filibuvir EC50 was calculated from the fitted dose-response curve relative to the mean activity of the plasma HCV RNA-derived replicon without compound present. For the replicon mixture experiments, a mixture of two dose-response curves was fitted (“biphasic fit”) using nonlinear regression with ordinary least squares, in addition to the single dose-response curve fit. This mixture model estimated two separate slopes, two EC50 parameters, and the ranges of the two individual dose-response curves, with the lower asymptote constrained to zero. Given that the sum of the two dose-response ranges represents the total replicon present, the percentage of each replicon present in the mixture was calculated. When the biphasic-mixture model fit successfully (convergence of curve-fitting routines within GenStat), the fits of the two models were compared using information criteria (Aikake and Bayesian), and the model with the best fit was selected.
Statistical analyses were performed on the logarithmic scale using a linear mixed model allowing for within and between patient/replicon variation. Pairwise comparisons were performed and summarized using geometric means, 95% confidence intervals (not reported), and P values.
The GenBank accession numbers for the baseline NS5B consensus sequences are JF424723 to JF424774.
The samples analyzed in this study were obtained from 52 patients (8 receiving placebo, 6 receiving filibuvir at 100 mg BID, and 38 receiving filibuvir at ≥300 mg BID) enrolled across the two filibuvir monotherapy studies (23). Of the 38 patients receiving filibuvir at doses of ≥300 mg BID, 37 achieved a maximum change from baseline in plasma HCV RNA of >1.0 log10 IU/ml. This suggests that the majority of patients were infected with a virus that was susceptible to filibuvir. The remaining patient (patient 13; filibuvir, 450 mg BID) did not respond to therapy, as defined by a maximum change from baseline in plasma HCV RNA of <0.5 log10 IU/ml over the 8 days of dosing.
Full-length population sequencing of the HCV NS5B coding region (codons 1 to 581) was successfully performed for 166/176 (95%) plasma samples collected for genotypic analysis across all 52 patients at ≥3 time points (pretherapy, end of therapy, and up to 25 days after therapy cessation). The remaining 10 plasma samples were successfully assessed using a shortened sequencing protocol covering NS5B codons 380 to 505; this region of NS5B encompasses all amino acid positions at which mutations that are known to affect susceptibility to thumb 2 NNIs occur (1, 20, 21).
Variants in NS5B at positions L419, M423, M426, I482, and V494 have previously been shown to affect susceptibility to thumb 2 NNIs (1, 20, 21). Variants at residues M423, L419, and I482 were not detected in virus from any of the 52 enrolled patients at baseline. NS5B variants M426A, M426L, and V494I were observed as the dominant species in baseline samples from 3 patients (see Table 2). Virus from patient 13, who failed to show a viral response when treated with filibuvir at 450 mg BID, encoded the NS5B variant R422K at baseline (see Table 2). These NS5B variants were retained in the 4 patients at all time points analyzed (pretherapy, on therapy, and posttherapy).
M423 variants were the most common NS5B mutations selected after filibuvir monotherapy. At clinically meaningful doses of filibuvir (≥300 mg BID), 29/38 (76%) patients displayed M423 variants on therapy (Table 1). A similar frequency was observed in patients infected with viral subtype 1a (19/25; 76%) or subtype 1b (10/13; 77%) (Fisher's exact test; P = 1.00). Mutations of the methionine codon to encode isoleucine, threonine, or valine were identified either as a single detectable species or in a mixture with methionine (20/29; 69%) (Table 1). In addition, the HCV sequence obtained from the plasma HCV RNAs of 9 patients encoded mixtures at multiple nucleotides in the 423 codon from which, in some instances, the presence of alanine was also inferred (Table 1).
All on-therapy NS5B sequences were additionally assessed for amino acid mutations selected at positions L419, R422, M426, I482, and V494. Only mutations at R422 (n = 1) and M426 (n = 3) were observed. In patient 21, selection of the mutation R422K/R (Table 2) was not associated with a rebound of plasma HCV RNA (data not shown). For the other 3 patients in whom NS5B M426 mutations were selected, mutations at M423 were also selected on therapy (Fig. 1 and Table 2).
To determine whether additional residue mutations were potentially associated with filibuvir administration, end-of-therapy amino acid sequences were compared to the corresponding baseline amino acid sequence for each patient. Mutations were identified at a total of 84 residue positions (84/581; 14%) across the 44 filibuvir-dosed patients (mean number of changes per patient, 3.3). There were 3 positions in addition to M423 where a mutation from baseline was observed in 4 or more patients (S/T5, G/S543, and A/V581). At all 3 positions, both polymorphism variants individually appeared on therapy, indicating that, for each position, neither polymorphism was selected preferentially during filibuvir administration.
Plasma HCV RNA in samples taken after therapy cessation were assessed for further genotypic changes in NS5B. In all instances, assessments were performed at day 28 (18 to 25 days after therapy cessation); patients who received 700 mg BID for 3 days were also assessed at day 5. A genotypic change to the pretherapy NS5B sequence was observed in posttherapy population sequences from 24 of 29 (83%) patients in whom M423 variants were selected on therapy (see Table S2 in the supplemental material). For 8 patients who received filibuvir for 3 days, reversion of the population NS5B sequence was rapid (i.e., within 2 days); complete reversion of the sequence to M423 was observed for 7 (88%), and partial reversion was observed for the remaining patient. For 21 patients with longer durations of therapy (8 to 10 days), the frequency of complete reversion of the population sequence by day 28 was reduced (7/21; 33%), with partial reversion to WT more common (9/21; 43%). Mutations that arose on therapy at NS5B positions 422 and 426 were not detected at the day 28 follow-up, indicating reversion of the population sequence at these positions.
In summary, mutations at NS5B position 423 were selected in 29 of 38 (76%) patients receiving filibuvir at ≥300 mg BID and showed reversion to WT following therapy cessation. Mutations at positions 422 and 426 also arose in a small minority of patients. There was evidence of the reestablishment of WT virus in plasma after withdrawal of filibuvir therapy.
In order to confirm that mutations selected in the HCV-infected individuals during filibuvir therapy were associated with resistance, drug susceptibility assays were performed using transient replicons constructed from NS5B amplicons derived from baseline, on-therapy, and follow-up plasma samples. Specifically, samples from 16 patients were selected because of the presence on therapy of M423 mutations (n = 13) or M426A (n = 1) or R422K (n = 2; one with R422K at all time points) variants. Baseline samples were selected from another 9 filibuvir-treated patients to provide additional information on the filibuvir susceptibilities of clinically derived subtype 1a and subtype 1b viruses. A flow diagram detailing the selection of samples for phenotypic assessment of filibuvir susceptibility is shown in Fig. 1.
Filibuvir inhibited both subtype 1a and subtype 1b replicons with comparable potencies for the 23/26 patients with NS5B sequences that were WT at residues 422, 423, 426, and 494 for whom valid dose-response inhibition curves were obtained at baseline (Table 2). The geometric mean EC50s were 29.2 nM (range, 7.1 to 77.0 nM) and 21.9 nM (range, 5.8 to 59.8 nM) for subtype 1a (n = 10) and subtype 1b (n = 13), respectively, which is consistent with preclinical data obtained previously (21). Consistent with the phenotypic susceptibilities of their NS5B sequences in the transient-replicon assay, patients infected with subtype 1a or 1b viruses showed a time- and dose-dependent response to filibuvir treatment, as determined by maximum drops in the viral load from baseline (Table 2). Four patients had NS5B sequences encoding a relevant NS5B resistance variant at baseline. The phenotypic susceptibilities of replicons derived from patient 18 (subtype 1b; NS5B mutation M426A; EC50, 170 nM), patient 23 (subtype 1a; M426L; EC50, 69 nM), and patient 28 (subtype 1a; V494I; EC50, 89 nM) were approximately 2- to 8-fold higher than the mean EC50 for the WT sequences (Table 2). All 3 patients responded to treatment, with maximal reductions in the viral load that were comparable to those of others within their dose group, suggesting that the presence of these NS5B mutations did not impair the patient's ability to respond to treatment during the short period of therapy in the two phase 1b studies. The replicon derived from patient 13 (subtype 1a; R422K) had very low RC, and therefore, no valid susceptibility data were obtained.
Of the 13 patients with NS5B M423 mutations identified on therapy who were selected for phenotypic analysis, valid dose-response inhibition curves were obtained from on-therapy HCV replicons generated from 8; the remaining 5 failed due to inadequate replication (Fig. 2). Six of the 8 (75%) replicon populations with a valid titration result showed >10-fold increases in resistance relative to the replicon population generated from their baseline sample (P < 0.0001) (Fig. 2a and Table 2). Of these 6, a <3-fold reduction in filibuvir susceptibility relative to baseline was observed for the day 28 follow-up replicons for 4. This reestablishment of phenotypic susceptibility coincided with genotypic reversion to the WT M423 residue, either complete reversion or reversion to a mixture of residues that included methionine (Table 2; see Table S2 in the supplemental material). The posttherapy replicon population from patient 24 retained reduced susceptibility to filibuvir, consistent with the persistence of NS5B M423T in this patient (Table 2), while a valid titration result could not be generated for the posttherapy replicon from patient 8. The functional replicons from all patients remained sensitive to the NS3 protease inhibitor BILN-2061 at all 3 time points (data not shown).
A modest (3.8-fold) reduction in filibuvir susceptibility relative to baseline was observed in the replicon derived from the day 5 plasma sample for patient 24 treated with filibuvir for 10 days at 450 mg BID (Table 2). This shift in susceptibility coincided with selection of an NS5B M426T mutation. At day 10, a genotypic mixture of resistance-associated variants at residue 423 was detected in the absence of M426T; this genotypic change was associated with high-level (>727-fold) phenotypic resistance (Table 2). For the patient with M426A present at all time points (patient 18) (Table 2), a reduction of filibuvir susceptibility was not observed relative to baseline either on therapy or at day 28 follow-up. Similarly, the on-therapy sample-derived replicon that encoded an R422K/R mixture (patient 21) did not show reduced susceptibility to filibuvir relative to baseline. Replicons derived from the nonresponder (patient 13), whose NS5B gene encoded R422K at all time points, could not be assessed for filibuvir susceptibility due to extremely poor replication (fitness).
The drug susceptibility assays were performed in parallel with control analyses using replicons derived from the plasma of 4 patients treated with placebo (Fig. 1). Reductions in filibuvir susceptibility were not observed with these placebo plasma-derived on-therapy and follow-up replicons relative to their corresponding baselines (Fig. 2a).
The RCs of the replicons containing clinically derived NS5B sequences can also be quantified in the transient-assay format. The RCs of the on-therapy- and follow-up-derived replicons were therefore determined for each patient and compared to the corresponding baseline sequence. As shown in Fig. 2b, statistically significant reductions in RC were observed with the 13 on-therapy replicons that encoded NS5B M423 mutations (P = 0.0003), but not with the 4 replicons obtained from the on-therapy plasma of placebo-treated patients (P = 0.74). Indeed, for 5 of the 13 M423-encoding on-therapy replicons, the RC was sufficiently affected to prevent the accurate determination of filibuvir susceptibility (Fig. 2b, black squares). Furthermore, in patient 13, assessment of the filibuvir susceptibilities of replicons with the R422K variant was prevented by the low RCs of the replicons derived from the plasma from baseline, on-therapy, and day 28 follow-up samples.
In summary, mutations at NS5B residue 423 were associated with a statistically significant reduction in filibuvir susceptibility and RC. The presence of M426 mutations correlated with small reductions in filibuvir susceptibility, but no effect of the R422K mutation was observed, possibly due to the presence of a mixed genotype (i.e., WT and mutant residues were detected) and low replicative capacity in vitro.
To further characterize the genotypic changes seen in the patients, specific NS5B mutations were incorporated into the Con1 replicon by site-directed mutagenesis, and these mutants were then assessed for filibuvir susceptibility (Table 3). Isoleucine, threonine, and valine mutations at residue 423 mediated very high-level resistance to filibuvir (>560-fold relative to Con1), with the M423T mutation being the most fit and M423V the least. In contrast, mutation of M423 to alanine resulted in a modest 4.3-fold change.
The R422K mutation identified in the nonresponder (patient 13), whose NS5B sequence could not be phenotyped, was found to be very unfit in vitro (2.6% relative to Con1) but highly resistant to filibuvir (>340-fold change). The M426A, M426T, and V494A mutations mediated low-level filibuvir resistance (9.6-, 3.4-, and 6.9-fold changes, respectively), while the M426V, M426L, V494T, and V494I mutations resulted in <2-fold change in the Con1 background. Assessment of mutations I482S, I482T, L419I, and L419M showed 82-, 12-, 2.1-, and 2.3-fold changes in filibuvir susceptibility, respectively. These NS5B mutations have been associated with resistance to thumb 2 binding investigational NNIs (1, 20, 21). However, none were selected by filibuvir during the brief treatment in the patients described in this study.
Given the high-level resistance mediated by NS5B M423 mutations when cloned into Con1 by mutagenesis (Table 3), it was surprising that phenotypic resistance to filibuvir was not observed for 2/8 patients in whom mutations at M423 were selected for on therapy (patients 6 and 19) (Table 2). Therefore, in order to mimic the clinically derived replicons, simplified two-component systems containing different percentages of resistant replicon mixed with Con1 WT were assessed for filibuvir susceptibility. Using standard sigmoid analyses, the presence of the M423T replicon did not substantially increase the filibuvir EC50 until the resistant replicon was present at ≥75% in the mixture, while M423V and I482S had little impact on filibuvir susceptibility until they were present at 90% (data not shown). The reason for this is apparent when the data are fitted using biphasic dose-response curves (Fig. 3A to C). As these replicon mixtures contained two components, the dose-response curves obtained with the mixtures of WT and mutant replicons are essentially composites of the two individual curves. In each case, the lower EC50 corresponds to the drug-sensitive WT NS5B sequence, while the higher EC50 corresponds to the resistant replicon EC50 (Fig. 3A to C). Using biphasic fitting, a higher EC50 could be determined (or inferred) when the input mixture contained ≥10% M423T, ≥25% I482S, or ≥50% M423V (data not shown).
In order to further understand the relationship between the relative fold resistance and the RC of WT (Con1) and mutant sequences on phenotypic assay readout for mixed populations, the percent resistant replicon remaining at the assay endpoint was calculated for mixtures to which biphasic dose responses were successfully fitted. As shown in Table 4, for mixtures of M423T and Con1, the percent resistant-replicon sequence remaining at the assay endpoint was in good correlation with the percent resistant replicon incorporated into the starting RNA mixture. This is not surprising, given that M423T is a relatively fit mutation (54% RC relative to Con1) (Table 3). In contrast, for the M423V and I482S replicons, the percent resistant replicon present at the end of the assay was substantially lower than that present in the RNA input (Table 4 and Fig. 3D), consistent with the low RC of the M423V and I482S mutants relative to Con1 (11% and 10%, respectively) (Table 3).
Collectively, these analyses explain why for some patients (e.g., patients 19 and 21), samples encoding mixtures of WT and highly resistant yet relatively unfit mutant NS5B sequences may appear phenotypically sensitive (Table 2). It remains unclear why the on-therapy sample from patient 6 (encoding an M423V mutation) appeared sensitive to filibuvir, although it is possible that a low-level (and therefore undetected) M423 WT variant outgrew the resistant M423V sequence in the transient-phenotype assay.
Filibuvir has demonstrated potential clinical utility in two phase 1b studies conducted in patients chronically infected with HCV subtype 1a or 1b with time- and dose-dependent reductions in plasma HCV RNA (23). In this analysis, the genotypic sequence of the HCV NS5B gene was determined at pretherapy, on-therapy, and posttherapy time points for all patients to enable characterization of filibuvir resistance. NS5B-encoding regions were used to prepare transient replicons to enable phenotypic assessment of filibuvir susceptibility and replicative capacity in vitro. In addition, specific mutations were incorporated into the Con1 replicon for further characterization.
Mutations at the NS5B residue M423 were the predominant change associated with filibuvir therapy. They were observed in 29 of 38 patients (76%) receiving filibuvir at ≥300 mg BID. The frequencies of mutations at this residue were similar in subtype 1a and subtype 1b viruses, consistent with filibuvir having similar potencies (and selection pressures) against each subtype. Mutations to isoleucine, threonine, and valine, requiring a single-nucleotide mutation in the methionine codon, were all observed both individually and in genotypic mixtures with the WT. These observations were entirely consistent with the in vitro observations; M423 mutations arose in 90% of resistant replicons sequenced (21). Some of the clinical-population sequences also potentially encoded alanine at residue 423 when complex mixtures were observed. M423A was found to mediate low-level resistance to filibuvir when it was generated by site-directed mutagenesis on a WT (Con1) background. Two nucleotide mutations are needed to encode an M423A mutation, either of which on its own results in the encoding of threonine or valine, both of which are associated with high-level resistance to filibuvir. Therefore, it is considered less likely that viruses harboring an M423A mutation will be selected during therapy with filibuvir.
The high-level resistance of the M423 isoleucine, threonine, and valine mutations (>560-fold) was confirmed using site-directed mutagenesis. In addition, the site-directed mutants highlighted the variation in the in vitro replicative capacity between these mutations, with threonine being the most fit and valine the least, thereby explaining the high prevalence of M423T observed in the in vitro resistance study (21). In the phase 1b studies, isoleucine mutants were observed in genotype 1a but not genotype 1b viruses. The clinical relevance of this is unknown and will be monitored in future studies.
M423 variants were not detected using population sequencing in any patient at baseline, which is consistent with their low clinical prevalence (<2%) in treatment-naïve HCV genotype 1 cohorts (3, 7, 11) and their reduced RCs compared to viruses with a WT M423 sequence. However, mathematical modeling of the HCV mutation and replication rates has predicted that every possible single and double HCV mutant is generated many times each day (4, 17). Therefore, it is likely that treatment with filibuvir at doses of ≥300 mg BID potently inhibits the WT viruses and confers a selective replicative advantage on preexisting minority M423 mutants, which becomes detectable within 48 h of the first dose and before any viral rebound is observed. Conversely, when treatment is stopped, the WT M423 viruses regain their competitive advantage and become the dominant circulating sequence once again. While we have not directly shown this in our study (it requires clonal analyses or ultradeep sequencing of baseline samples), this is the mechanism by which resistance to the protease inhibitors is rapidly selected (17, 18). Similarly, treatment of chronic hepatitis B with lamivudine as monotherapy results in selection of lamivudine-resistant variants in most patients (8). The resistant variants have compromised replication compared to WT virus; hence, their high level of drug resistance confers a selective growth advantage only in the presence of lamivudine. These variants are deselected in favor of WT virus once treatment is stopped yet rapidly reemerge in patients following the resumption of therapy.
Though variants at NS5B residue M423 were not detected by population sequencing at baseline, the NS5B R422K, M426A, M426L, and V494I variants were observed, albeit at low frequency. There was no evidence of any effect of the M426A, M426L, or V494I variants on virological response to filibuvir, although they each conferred low-level phenotypic resistance in vitro (approximately 2- to 8-fold change relative to the mean of the WT baseline sequences). Conversely, the R422K variant was identified in a patient who did not respond to therapy (450 mg BID). The R422K mutation was identified on treatment in another patient, but its appearance did not correlate with loss of virologic response. When introduced by site-directed mutagenesis into a Con1 background, the R422K mutation conferred high-level resistance and very poor RC. Taken together, these data suggest that R422K is a clinically relevant resistance mutation, but under some circumstances, it confers poor RC on the virus.
X-ray crystallography studies of NS5B support the relevance of residues R422, M423, and M426 in the selection of filibuvir resistance, as seen in the phase 1b studies. The most intimate noncovalent interaction with NS5B is mediated by filibuvir's cyclopentyl ring, which fits into a small complementary hydrophobic pocket (12). Residues M423 and R422 help form the walls of this pocket, with M426 in close proximity (Fig. 4). Thus, it is likely that the mutations observed in the monotherapy studies at these residues alter the structure of the hydrophobic pocket, thereby reducing the affinity for filibuvir binding. The compound VX-759 also binds in the thumb 2 pocket and shows an overlapping resistance pattern (1). Therefore, although we were not able to measure this directly, we would expect that the filibuvir-resistant NS5B mutants selected in the clinical studies reported here would also be cross-resistant to VX-759 in the transient-replicon assay. Consequently, a DAA combination regimen of filibuvir and VX-759 or sequential use of these compounds in HCV patients is not recommended. However, sequences resistant to filibuvir are fully sensitive to the NS3 protease inhibitor BILN-2061 and would be expected to retain full sensitivity to other protease inhibitors, as well as interferon and ribavirin, NS5A inhibitors, and the NS5B nucleoside inhibitors and NS5B NNIs that bind other pockets (e.g., “thumb 1” and “palm 1”). In vitro combinations of filibuvir with compounds in other DAA classes greatly reduce the emergence of resistance in replicon cultures (5) and support the further development of filibuvir as part of a multi-DAA regimen.
Transient-replicon systems (based on bulk cloning populations of quasispecies) have been used successfully for determining the phenotypic susceptibilities of clinically derived HCV gene sequences to investigational DAAs (11, 14). It has been argued that these systems are relevant, as they mimic the intrinsic HCV genetic heterogeneity present in infected patients. For example, they avoid the need to test a large number of individual clones or the selection and testing of individual clones from the population that could potentially have genetic and phenotypic characteristics significantly different from those of the isolate's overall population (11). We used a transient system to phenotype a relatively large number of plasma-derived NS5B sequences, including serial samples prior to, during, and after filibuvir therapy. In the majority of cases, the phenotypic data were consistent with the genotypic data obtained for the samples and confirmed that M423 mutations are the dominant resistance pathway selected for filibuvir in HCV patients. For patients whose virus appeared phenotypically sensitive despite the presence of mutant residues at key positions in NS5B, controlled-mixture experiments provide a plausible explanation for the apparent discrepancy. The approach we followed is consistent with mixture experiments performed using an NS5B L419M mutant that confers resistance to another thumb 2 NNI (11). For mutants such as M423T, which are relatively fit compared to the drug-sensitive WT sequence, the sensitivity of the phenotypic assay to deliver a resistant EC50 readout is comparable to the population-sequencing results. For mutants with poor RCs, e.g., M423V, it is likely that even low levels of a WT sequence in the input to the assay can result in a phenotypically sensitive readout. The sensitivity of phenotypic assays to detect mutant sequences in a mixed population can be enhanced by culturing the transfected cells in the presence of inhibitor, thus suppressing the WT sequence and giving a selective growth advantage to less fit drug-resistant variants (14). We did not follow this method in our analysis because we wanted to correlate the population-sequencing results with the phenotypic filibuvir susceptibility for the same NS5B amplicons.
Although the data we have obtained from the phenotypic analyses are broadly consistent with the genotypic changes we observed, there are limitations to this approach. For example, replicative capacity was estimated using only the patient-derived NS5B gene, whereas the real replicative capacity depends on the NS5B sequence in the context of the entire viral genome in the same patient. This is most notable for patient 13, for whom the R422K mutant (a highly “unfit” NS5B polymerase mutation in the context of Con1) was the dominant species (and therefore “fit”) prior to any filibuvir treatment. Second, the transient-replicon system is an imperfect surrogate system for a replicating virus in a patient in whom host factors (such as immune pressure) also play a significant part in the selection of a specific viral sequence.
In conclusion, mutations at the HCV NS5B residue M423 form the predominant pathway to filibuvir resistance in the clinic. Isoleucine, threonine, and valine mutations mediate high-level resistance, albeit with reduced replicative capacity relative to the WT, resulting in the reversion of residue 423 to methionine in many patients following cessation of short-term therapy.
We gratefully acknowledge all the patients who participated in the study; all the investigators, nursing staff, and research support staff involved in the study; and the research team at Pfizer Global Research and Development. We also acknowledge Stanley M. Lemon, University of North Carolina, Chapel Hill, NC, and Paul Targett-Adams for critical reading of the manuscript; Mark Gardner for generation of the figure of the NS5B-filibuvir model; and Steve Shaw for technical assistance in the phenotypic-characterization work. The extraction of RNA from plasma samples and amplification of NS5B were performed under contract by Lab 21 Healthcare, Cambridge, United Kingdom. The transient-replicon assays were performed under contract by HD Biosciences, Shanghai, China.
This study was sponsored by Pfizer, Inc. Editorial assistance provided by Susanne Gilbert of ACUMED (New York, NY) was funded by Pfizer.
Published ahead of print 27 December 2011
Supplemental material for this article may be found at http://aac.asm.org/.