|Home | About | Journals | Submit | Contact Us | Français|
BMS-790052, targeting nonstructural protein 5A (NS5A), is the most potent hepatitis C virus (HCV) inhibitor described to date. It is highly effective against genotype 1 replicons and also displays robust genotype 1 anti-HCV activity in the clinic (M. Gao et al., Nature 465:96-100, 2010). BMS-790052 inhibits genotype 2a JFH1 replicon cells and cell culture infectious virus with 50% effective concentrations (EC50s) of 46.8 and 16.1 pM, respectively. Resistance selection studies with the JFH1 replicon and virus systems identified drug-induced mutations within the N-terminal region of NS5A. F28S, L31M, C92R, and Y93H were the major resistance mutations identified; the impact of these mutations on inhibitor sensitivity between the replicon and virus was very similar. The C92R and Y93H mutations negatively impacted fitness of the JFH1 virus. Second-site replacements at NS5A residue 30 (K30E/Q) restored efficient replication of the C92R viral variant, thus demonstrating a genetic interaction between NS5A residues 30 and 92. By using a trans-complementation assay with JFH1 replicons encoding inhibitor-sensitive and inhibitor-resistant NS5A proteins, we provide genetic evidence that NS5A performs the following two distinct functions in HCV RNA replication: a cis-acting function that likely occurs as part of the HCV replication complex and a trans-acting function that may occur outside the replication complex. The cis-acting function is likely performed by basally phosphorylated NS5A, while the trans-acting function likely requires hyperphosphorylation. Our data indicate that BMS-790052 blocks the cis-acting function of NS5A. Since BMS-790052 also impairs JFH1 NS5A hyperphosphorylation, it likely also blocks the trans-acting function.
Hepatitis C virus (HCV), a positive-strand RNA virus of the Flaviviridae family, is a leading cause of liver disease and hepatocellular carcinoma (35, 59). The HCV genome encodes a single polyprotein that is cleaved by cellular and viral proteases into at least 10 individual proteins (40, 48). Nonstructural protein 3 (NS3), NS4A, NS4B, NS5A, and NS5B are sufficient for HCV RNA replication in cell culture (6, 30). NS3-4A is the primary viral protease, and NS5B is an RNA-dependent RNA polymerase (4, 5, 14). NS4B, a hydrophobic protein with multiple trans-membrane domains, induces an endoplasmic reticulum-derived membranous web that harbors the HCV replication complex (10, 32). NS5A is also essential for HCV RNA replication, but its precise roles in this process are poorly understood (21, 33).
NS5A is an RNA binding protein that interacts with other HCV nonstructural proteins and is capable of altering NS5B polymerase activity in vitro (9, 20, 41, 46, 47). NS5A also interacts with a variety of cellular proteins and is potentially involved in modulating multiple aspects of the cellular environment (33, 43). In addition to its role(s) in RNA replication, NS5A is also required for viral assembly (3, 50). NS5A is a modular protein with an N-terminal amphipathic α-helix membrane anchor (amino acids [aa] 5 to 25) and three distinct structural domains (11, 39, 52). Domain I (aa 28 to 213) is essential for RNA replication and has been crystallized as a dimer in two different configurations, suggesting the potential for distinct functional conformations and/or higher-order multimers (31, 53). A basic groove proposed to function in RNA binding is present in one structure (53) but absent in the other (31). Domains II (aa 250 to 342) and III (aa 356 to 447) are natively unfolded, and roles for these domains in RNA replication and viral assembly are still emerging (19, 29, 51). NS5A is highly phosphorylated and is expressed in cell culture as basally phosphorylated (p56) and hyperphosphorylated (p58) forms (24). Basal phosphorylation is believed to occur at residues in the central and C-terminal parts of the protein, while several highly conserved serine residues in the central part of the protein (aa 214 to 249) are required for NS5A hyperphosphorylation (2, 49). Phosphorylation has been implicated as a regulatory switch, modulating multiple NS5A functions (2, 13, 21, 50).
BMS-790052, a first-in-class inhibitor targeting NS5A, is extremely potent against genotype 1 replicons in vitro and has exhibited impressive anti-HCV activity in early clinical trials (16). The precise mode of action of BMS-790052 is unknown, but related inhibitors have been shown to bind directly to NS5A and to disrupt NS5A hyperphosphorylation (16, 28). Resistance studies with genotype 1 replicons identified NS5A residues 28, 30, 31, and 93 as primary sites for drug-induced changes (15, 16). Substitutions at these positions were also observed in HCV-infected individuals treated with BMS-790052, suggesting a correlation between in vitro and in vivo resistance (16). BMS-790052 is active against multiple HCV genotypes (16), including the JFH1 genotype 2a strain, which is capable of producing infectious virus both in vitro and in vivo (55).
Here we report the results from BMS-790052 resistance studies performed on JFH1 replicons and cell culture infectious virus. Similar to genotype 1, resistance mutations were found in the N-terminal region of the NS5A protein, with F28, L31, C92, and Y93 identified as primary sites of resistance. In addition, mutations at residue 30 acted as compensatory mutations, enhancing viral replication and modulating inhibitor sensitivity. The resistance analysis established a correlation between BMS-790052 resistance profiles in the replicon and virus cell culture systems and provided a predictor for the emergence of resistance in the clinical setting. We have also used BMS-790052 as a tool to explore NS5A functions. Our results suggest that NS5A performs at least two distinct functions in RNA replication: a cis-acting function that is hyperphosphorylation independent and likely occurs in the context of the HCV replication complex and a trans-acting function that requires hyperphosphorylation and may occur outside the replication complex. Both functions are likely blocked by BMS-790052.
Human hepatoma cells (Huh-7.5; Apath, Brooklyn, NY) were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), l-glutamine (2 mM), penicillin (100 units/ml), and streptomycin (100 μg/ml). Huh-7 cells harboring the JFH1 replicon were established as described previously (27) and were maintained in complete DMEM supplemented with 0.5 mg/ml G418 (Geneticin; Invitrogen Corp., Carlsbad, CA). The NS5A inhibitor BMS-790052 has been described (16).
A bicistronic JFH1 viral clone with a Renilla luciferase reporter and adaptive mutations in the core (K74T), E2 (G451R), NS3 (M1051T), and NS5A (C2219R) (60) was constructed as described previously (55). A subgenomic JFH1 replicon clone with a neomycin resistance gene (neo) was constructed as described previously (25). The con1 replicon clone has been described (15, 16). Reporter replicons for transient assays were made by replacing the neo gene with genes encoding Renilla luciferase or β-lactamase (15, 18). NS5A mutations in replicon and virus clones were introduced with recombinant PCR and standard cloning techniques and were verified by sequence analysis.
A fluorescence resonance energy transfer (FRET) assay based on NS3 protease activity was used to assess inhibitor sensitivities of replicon cell lines as described previously (15, 38). Transient-transfection assays with replicon and viral clones were performed as described previously (15). RNA transcripts were prepared with a RiboMax T7 express system (Promega Corp., Madison, WI) from replicon and viral clones that were linearized with XbaI. In cotransfection and trans-complementation experiments, equal amounts of RNA (~10 μg) were transfected into Huh-7.5 cells. Duplicate 96-well cell culture plates were prepared for Renilla luciferase and β-lactamase assays. Renilla luciferase assays were performed as described previously (15, 38). β-Lactamase activity was assayed with a LiveBlazer CCF2-AM kit by following the enhanced loading protocol recommended by the manufacturer (Invitrogen, Corp., Carlsbad, CA). Inhibitor 50% effective concentrations (EC50s) were calculated as described previously (15, 38). Replication levels of NS5A variants (see Tables Tables2,2, ,4,4, and and6)6) were determined from the replication window (uninhibited luciferase signal/background) at 72 h after transfection relative to the parental control (15). Background was derived from wells transfected with the parental clone and treated with 1 μM BMS-790052. For virus infection experiments (see Fig. 2), media were collected from JFH1 virus RNA-transfected cultures and used to infect Huh-7.5 cells plated in 100-mm dishes. For the experiments shown in the figure (see Fig. 2B), the amount of cleared media used for infection was normalized according to the luciferase signal obtained from cell lysates and was adjusted to a total of 10 ml with complete media. After 4 h, media were replaced with complete media, and subconfluent cultures were maintained. At each passage, equivalent proportions of the cultures (approximately one-eighth) were harvested for luciferase activity as a means of monitoring viral growth.
Selection of JFH1 replicon cells with reduced sensitivity to BMS-790052 was performed as described previously (15). To generate resistance in the full-length infectious virus, Huh-7.5 cells were transfected with transcripts prepared from the Renilla luciferase reporter JFH1 viral clone. After 72 h, cells were transferred to media containing inhibitor or DMSO (control) and maintained as subconfluent cultures. Fresh Huh-7.5 cells were added to the cultures as needed to offset the cytopathic effects of the JFH1 virus (45). Virus replication was monitored by measuring Renilla luciferase activity at each passage until levels approached that in the DMSO control. NS5A cDNAs were recovered from cells by reverse transcriptase PCR (RT-PCR) as described previously (15). The oligonucleotide primers used to amplify JFH1 NS5A cDNA were 5′-CTTACTATAACCAGCCTACTCAGAAGACTCCAC (forward) and CTCAAAGGGTTGATTGGCAACTTTTCCTCTTC (reverse). NS5A amplicons were used for direct population sequencing and to generate cDNA clones with a TOPO-TA cloning kit (Invitrogen Corp., Carlsbad, CA).
HCV replicon clones were expressed in BHK-21 cells with a modified vaccinia virus-T7 (MVA-T7) expression system (58). Briefly, cells were seeded (~200,000 cells/well) in 24-well tissue culture dishes. After ~16 h, cells were rinsed once with 2 ml infection media (DMEM with 2% FBS) and infected with ~3 to 5 PFU of MVA-T7 per cell in 150 μl (56). After 2 h, media were aspirated, and the cells were transfected (0.2 μg replicon DNA in 200 μl) by using Effectene transfection reagent (Qiagen Corporation, Valencia, CA). Approximately 2 to 3 h after transfection, an additional 200 μl of complete media supplemented with the inhibitor or DMSO was added. After 12 to 16 additional hours, cells were washed with phosphate-buffered saline (PBS; 0.5 ml), incubated for 5 min at room temperature in 200 μl enzyme-free cell dissociation buffer (Invitrogen Corp., Carlsbad, CA), transferred to microcentrifuge tubes, and centrifuged for 1 min (14,000 × g). Pellets were resuspended in 100 μl ice-cold lysis buffer [50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 0.5% NP-40, 1× complete protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN), 135 μM sodium orthovanadate (Sigma-Aldrich Corp., St. Louis, MO)] and incubated on ice for 15 min. Lysates were cleared by centrifugation (14,000 × g, 15 min, 4°C). Supernatants (20 μl) were electrophoresed on 7.5% polyacrylamide gels (Criterion system, Bio-Rad Laboratories, Hercules, CA) and transferred to 0.45-μm nitrocellulose membranes (Bio-Rad Laboratories, Hercules, CA). Membranes were blocked in 0.1% Tween 20 in PBS containing 5% milk and incubated with a 1:5,000 dilution of rabbit anti-NS5A antibody (27), followed by incubation with a 1:10,000 dilution of goat anti-rabbit horseradish peroxidase-conjugated antibody (Bio-Rad Laboratories, Hercules, CA). Visualization was performed with a SuperSignal West Pico detection kit (Pierce Biotechnology Inc., Rockford, IL).
BMS-790052 is a potent inhibitor of genotype 1a and 1b replicons and also inhibits JFH1 subgenomic replicon cells and full-length infectious virus with EC50s of 46.8 ± 18.5 pM and 16.1 ± 12.4 pM, respectively (Table 1) (15, 16). A previously described NS5A inhibitor was shown to suppress production of the hyperphosphorylated form of NS5A expressed from a genotype 1b replicon (28). To determine if BMS-790052 has a similar activity, genotype 1b (con1) and 2a (JFH1) replicon clones were expressed in cell culture by using a vaccinia virus-T7 expression system (Fig. 1). Treatment of the cells with BMS-790052 resulted in a substantial decrease in the expression of hyperphosphorylated NS5A in both genetic backgrounds (Fig. 1), indicating that this activity is conserved among inhibitors of this class and is not restricted to genotype 1b replicons.
To identify amino acid substitutions that confer resistance to BMS-790052, Huh-7 cells harboring the JFH1 subgenomic replicon or full-length virus were treated with inhibitor until resistant populations emerged (~4 to 5 weeks). Selections were performed with BMS-790052 at concentrations ranging from 1 to 10 nM (Table 1). Control cells were maintained with equivalent concentrations of DMSO. Mutations within NS5A associated with reduced susceptibility to BMS-790052 were identified by sequence analysis of NS5A cDNA isolated by RT-PCR from control and inhibitor-treated cells. Analysis focused on the N-terminal region of NS5A since prior resistance studies with genotype 1 replicons exclusively identified BMS-790052 resistance mutations in this region (15, 16). The occurrence and frequency of NS5A amino acid substitutions identified in JFH1 NS5A cDNA clones are summarized in Table 1. In cDNA from replicon cells, single amino acid substitutions were found at residues 28 (F28I/S/V), 31 (L31M/V), 92 (C92R), and 93 (Y93H). The C92R mutation was also found linked to changes at residue 30 (K30E or K30Q). In clones from virus-infected cells, F28S and L31M were the only single amino acid substitutions observed. F28S was also found linked to K30E and L31M. Mutations at NS5A residues 28, 30, 31, 92, and 93 were not identified in 49 cDNA clones from DMSO-treated replicon cells or in 31 clones from DMSO-treated virus-infected cells.
To evaluate the contributions of specific amino acid substitutions to resistance, the F28S, L31M, C92R, and Y93H mutations were introduced into the JFH1 replicon and viral clones. The sensitivity of these variants to BMS-790052 was assessed in transient replication assays. Each of these mutations resulted in substantial decreases in BMS-790052 potency, with the L31M and C92R mutations having the least effect (~100-fold decrease) and the F28S mutation having the greatest effect (>7,500-fold decrease). The relative levels of resistance conferred by the mutations (fold resistance) were very similar between the replicon and virus assays, even though the BMS-790052 EC50s were generally 2- to 4-fold lower in the virus assay (Table 2). A control HCV NS3 protease inhibitor was also more potent in the virus assay than in the replicon assay, but the potency of this inhibitor was not affected by the NS5A mutations (data not shown).
In addition to determining the effect on inhibitor sensitivity, the relative replication ability (fitness) of the variants was assessed (Table 2). The F28S and L31M mutations had relatively little impact on fitness in the context of either the virus or replicon, with replication levels ranging from 54 to 125% compared to that of the parental wild-type (WT) controls (Table 2). In contrast, the C92R variant replicated poorly (≤10% of the WT level). The Y93H mutation had relatively little impact on replication in the context of the replicon (81% compared to that of the WT), but a much greater impact was observed with the virus (4% of the WT level).
In an attempt to identify compensatory mutations that enhance replication of the C92R or Y93H variants, cell cultures were infected with mutant viruses and maintained for ~3 weeks or until a substantial increase in viral replication (monitored by luciferase activity) was observed (Fig. 2A and B). NS5A cDNA was then recovered from the cultures by RT-PCR and sequenced. In two experiments with the C92R variant, luciferase levels increased substantially within 20 days after infection. In each case, the C92R mutation was found to be linked to additional mutations at residue 30. In cDNA from the first experiment (Fig. 2A), K30E and K30Q were the primary mutations observed. In the second experiment (Fig. 2B), a more complicated pattern of changes at codon 30 made it difficult to discern the precise identity of the amino acid replacements (data not shown). Sequence analysis of individual NS5A cDNA clones confirmed the occurrence of mutations at residue 30 linked to C92R (Table 3). In clones obtained from both experiments, the most common combination was K30E plus C92R, present in 13 out of 29 clones (Table 3). The combination of K30Q plus C92R was present in 6 clones, while K30M plus C92R was found in 3 clones. Other changes included reversion to C92 ± K30Q and C92L (Table 3).
Results from the Y93H variant-infected cultures were less informative. In one experiment (Fig. 2A), luciferase activity remained relatively low, and sequence analysis of NS5A cDNA recovered at day 23 revealed that the Y93H mutation was still present. No additional mutations within NS5A were observed. In a second experiment, higher luciferase levels were achieved by day 24 (Fig. 2B). NS5A cDNA recovered from these cells encoded a tyrosine at NS5A residue 93, indicating reversion to the WT residue (data not shown). To more directly compare growth of the Y93H variant with the WT virus, luciferase signals from cell cultures infected with very similar titers of mutant or WT virus were monitored over a period of several weeks. As shown in Fig. 2C, even when very similar luciferase signals were obtained at 4 h after transfection, the wild-type virus rapidly outpaced the Y93H variant, confirming the negative impact of this mutation on cell culture growth of the JFH1 virus.
The identification of mutations at residue 30 in NS5A cDNA recovered from cultures infected with the C92R variant suggested that these mutations might act as compensatory mutations to increase replication of the C92R variant. To test this possibility, clones with single or linked mutations (K30E/Q ± C92R) were evaluated in transient replication assays (Table 4). As expected, replication of the C92R variant was substantially enhanced by either the K30E or K30Q mutation (Table 4). Interestingly, a virus with only K30E replicated as poorly as the C92R variant; however, when both mutations were present, the virus replicated at WT levels (Table 4). As individual mutations, neither K30E nor K30Q conferred resistance to BMS-790052. In combination with C92R, the K30E mutation moderately enhanced the BMS-790052 resistance, but the K30Q-plus-C92R variant had WT sensitivity (Table 4). In genotype 1a, amino acid replacements at NS5A residue 30 yielded similar effects on BMS-790052 sensitivity. In this case, glutamine (Q) was the WT residue, and Q30E and Q30K replacements were associated with decreases in BMS-790052 potency (15).
NS5A has been proposed to function as a dimer or higher-order multimer (31, 53) and also to be capable of functioning in trans (1, 23, 54). Given these possibilities, we decided to investigate the potential effect that coexpression of WT and resistant replicons would have on NS5A inhibitor potency. For example, would either WT or resistant NS5A proteins possess a trans-dominant inhibitor-sensitivity phenotype? To this end, replicons with different sensitivities to BMS-790052 were coexpressed in Huh-7.5 cells. Distinct reporter genes which could be used to independently monitor inhibitor susceptibilities (Renilla luciferase or β-lactamase) were used to differentiate expression of sensitive (JFH1-WT) and resistant (JFH1-L31M) replicons. Results from the coexpression experiments are summarized in Table 5. BMS-790052 EC50s of 6.4 and 0.07 nM were obtained with the JFH1-L31M (Renilla luciferase) and JFH1-WT (β-lactamase) replicons, respectively, when these replicons were individually expressed in transient assays (Table 5). When equal amounts of RNA from these replicons were cotransfected, EC50s from the luciferase and β-lactamase assays were 7.1 and 0.04 nM, respectively (Table 5), indicating that the replicons each maintained their original sensitivities to the NS5A inhibitor.
To exclude the possibility that results from the cotransfection experiments (Table 5) were disproportionately influenced by cells expressing only one of the transfected replicons, luciferase reporter replicons were transiently expressed in cells containing replicons that were maintained by antibiotic selection. The replicons in the host cells expressed the neo gene and were either sensitive (JFH1-WT-neo) or resistant (JFH1-052R-neo) to BMS-790052. Since the resident replicons did not contain reporter genes, luciferase signals were used to measure the BMS-790052 sensitivity of the transfected replicon. Dose response curves from these experiments are shown in Fig. 3. When the sensitive replicon (JFH1-WT-luc) was transiently expressed in replicon cells that were highly resistant to BMS-790052 (JFH1-052R-neo; EC50 = 182 nM), the mean EC50 derived from luciferase assays was 20 pM, indicating that the replicon retained an inhibitor-sensitive phenotype (Fig. 3). Similarly, the JFH1-L31M-luc replicon did not adopt an inhibitor-sensitive phenotype when it was expressed in the JFH1-WT-neo replicon cells. In this case, an increase in the luciferase signal from the introduced replicon was observed with increasing concentrations of BMS-790052 until between 1 and 10 nM, at which point a sharp inhibition of the luciferase signal was observed (Fig. 3). Although the shape of the dose response curve precluded accurate calculation of the inhibitor EC50, it is clear that the resistant luciferase replicon retained its BMS-790052-resistant phenotype. The increase in luciferase signal observed in these experiments occurred concomitantly with inhibition of the host cell replicon (Fig. 3), suggesting an increase in replication of the transfected replicon due to a release of competition with the sensitive replicon for limiting host cell factors (12). Collectively, the results obtained from coexpressing replicons with different sensitivities to BMS-790052 suggest that very limited mixing of NS5A occurs among replication complexes encoded by different replicons.
To further investigate whether NS5A can be efficiently exchanged among replication complexes, a NS5A trans-complementation assay similar to that described by Jones et al. (23) was employed. In the context of the JFH1 replicon, a serine-to-isoleucine replacement at amino acid 232 of NS5A (S2204I) produces a severe replication defect (23). However, replication of a S2204I replicon can be rescued (trans-complemented) by coexpression of a helper replicon encoding a WT NS5A protein (23). By incorporating replicons with different sensitivities to BMS-790052 into a trans-complementation assay, we were able to investigate the possible mechanisms of NS5A trans-complementation. For this assay, defective luciferase reporter replicons (S2204I) were transiently coexpressed with helper replicons that carried β-lactamase reporters and encoded either inhibitor-sensitive (WT) or inhibitor-resistant (L31M) NS5A proteins. Complementation was noted as an increase in luciferase activity relative to that of the background, and inhibitor susceptibilities of the rescued and helper replicons were determined from Renilla luciferase and β-lactamase assays, respectively. Results from three independent assays are summarized in Table 6. As expected, no replication of the S2204I replicon was detected when this replicon was expressed without a helper replicon (luciferase signals were similar to those of a negative-control replicon) (Table 6). However, coexpression of the S2204I replicon with either the WT or resistant helper replicon resulted in replication of the defective replicon to a level of ~8% of that of the WT control (Table 6). Assaying with β-lactamase revealed that the helper replicons maintained their original inhibitor sensitivities when they were coexpressed with the defective replicon (WT, EC50 = 0.06 nM; L31M, EC50 = 6.8 nM) (Table 6). Most notably, the rescued S2204I replicon retained an inhibitor-sensitive phenotype regardless of whether it was rescued by the sensitive (WT) or resistant (L31M) replicon (EC50s = 0.04 and 0.02 nM, respectively) (Table 6). These results suggest that trans-complementation did not occur simply as a result of NS5A expressed from the helper replicon functionally replacing NS5A in the defective replication complex, since in this case the inhibitor sensitivity should correspond to that of the helper NS5A. Instead, the results are more consistent with a scenario in which the defective NS5A protein still performed a cis-acting replication function which was blocked by BMS-790052.
While the S2204I mutation in the JFH1 replicon yields a severe replication defect, the equivalent mutation in genotype 1b replicons is associated with an enhanced replication phenotype and a decrease in NS5A hyperphosphorylation (6, 13). To investigate the nature of the defect caused by different S2204I mutation on the JFH1 replicon, the impact of the amino acid 2204 substitutions on NS5A hyperphosphorylation and replicon replication was examined. As shown in Fig. 4, the S2204I mutation severely reduced hyperphosphorylation of NS5A expressed from the JFH1 replicon. In contrast, a replicon with a S2204G mutation still expressed substantial levels of hyperphosphorylated NS5A and also replicated at near WT levels (Fig. 4). These results suggest that at least some NS5A hyperphosphorylation, or a conformation associated with hyperphosphorylation, is required for efficient JFH1 replicon replication. When taken together with the findings of the trans-complementation assays, these results further suggest that a hyperphosphorylated form of NS5A performs a trans-acting function that is essential for RNA replication.
Resistance selection studies with the HCV NS5A inhibitor BMS-790052 identified F28S, L31M, C92R, and Y93H as primary drug-induced mutations in NS5A in the JFH1 subgenomic replicon and/or full-length virus systems. The relative impact of these mutations on BMS-790052 potency in the replicon and infectious virus assays was very similar. We previously noted a correlation between resistance mutations selected in genotype 1 replicons and those enriched upon dosing of BMS-790052 in the clinic (15, 16). Collectively, these findings support the use of in vitro replicon and cell culture virus systems to assess the emergence of BMS-790052-resistant variants in a clinical setting.
Of the mutations identified, the L31M mutation is of particular interest because the majority of genotype 2a sequences deposited in the European HCV database contain a methionine at NS5A residue 31 (7). In contrast, variation among the database sequences at residue 28, 92, or 93 was not observed. A JFH1 virus with an L31M mutation was inhibited by BMS-790052, with an EC50 of 1.9 nM (Table 2). Scheel et al. (44) recently reported a similar potency of BMS-790052 (EC50 = 14 nM) on a hybrid genotype 2a virus that also had a methionine at NS5A residue 31. These findings suggest a low nM range of potencies for BMS-790052 on genotype 2a strains with methionine at residue 31. In an early clinical study, BMS-790052 plasma concentrations well in excess of this range were readily achieved and well tolerated (16), indicating the potential utility of BMS-790052 as a component in genotype 2a HCV treatment regimens.
In the genotype 1a replicon, M28T, Q30H/R, L31M/V, and Y93H were commonly identified BMS-790052-induced mutations; L31V and Y93H were the primary resistance mutations identified in genotype 1b (15, 16). The overall similarity in resistance profiles among genotypes 1a, 1b, and 2a suggests a well-conserved BMS-790052 binding pocket. Of the residues identified as sites of resistance in the current study, only residue 92 (C92R) was not previously associated with BMS-790052 resistance. In both genotype 1a (H77c) and 1b (con1) replicons, an alanine is found at residue 92. Alanine-92 is located at or near the dimer interface in NS5A genotype 1b domain I crystal structures (31, 53) and forms a main chain hydrogen bond with glycine-96 in one of the structures (31). We were unable to assess the impact of A92R mutations on BMS-790052 potency, since genotype 1 replicons containing this mutation failed to replicate sufficiently in transient replication assays (R. A. Fridell, unpublished data). An A92V mutation in the genotype 1b replicon was previously reported as a resistance mutation for a class of piperazinyl-N-phenylbenzamide NS5A inhibitors (8). This mutation, however, does not confer resistance to BMS-790052 (Fridell, unpublished), suggesting that the resistance profiles of NS5A inhibitors are dependent on the specific inhibitor chemotype and the precise amino acid substitutions within the inhibitor binding pocket.
A C92R mutation negatively impacted replication fitness in the JFH1 replicon and infectious virus systems (≤10% compared to that of the WT control) (Table 2). Compensatory mutations were selected at residue 30 (K30E and K30Q) that substantially enhanced replication of JFH1 virus bearing the C92R mutation (Table 4). Interestingly, while virus with either K30E or C92R replicated poorly, the combination of these mutations yielded a virus that replicated at close to the WT levels. These findings indicate the existence of a genetic interaction between NS5A residues 30 and 92, which could occur in the context of NS5A monomers or multimers or could potentially involve associations with other proteins. The Y93H mutation also resulted in poor replication of the JFH1 virus (~8% relative to that of the parental control). Efforts to identify compensatory mutations within NS5A that improve replication of the Y93H variant were unsuccessful. Surprisingly, a JFH1 replicon with the Y93H mutation replicated well (~80% of the WT level). The reason why the Y93H mutation impaired fitness more in the virus than in the replicon is currently unclear. One possibility is that this mutation affects processes downstream of RNA replication. Miyanari et al. (34) reported that a mutation in this region of NS5A (aa 99 to 101 changed to alanines) blocked recruitment of NS5A to lipid droplets and decreased viral infectivity. Further studies are needed to determine if the Y93H mutation is associated with similar phenotypes.
HCV RNA replication occurs in the context of a multiprotein complex that is tightly sequestered within specialized intracellular membrane compartments (10, 17, 42). Several of the HCV nonstructural proteins, including NS3 and NS5B, display a strong cis-acting preference in regards to RNA replication, i.e., their replication function is tightly linked to the genome from which they are translated (1, 12, 54). In contrast, a trans-acting replication function has been demonstrated for NS5A (1, 23, 54). Recently, NS4B was also shown to be capable of functioning in trans in HCV RNA replication, albeit much less efficiently and likely by a different mechanism than NS5A (23). A popular hypothesis to explain the efficient trans-complementation observed for NS5A is that NS5A, unlike other HCV replication proteins, can readily move among replication complexes, possibly because it is less tightly tethered to the intracellular membranes where RNA replication occurs (1, 23, 54). Two lines of evidence from the current study argue against this hypothesis. First, when replicons with different sensitivities to BMS-790052 and different reporter genes were coexpressed, the replicons maintained their original sensitivities to the NS5A inhibitor. No evidence of an intermediate or trans-dominant phenotype was observed, which might be expected if NS5As encoded by the different replicons were freely exchanged among replication complexes. Studies examining the mobility of green fluorescent protein (GFP)-tagged NS5A have also indicated that replication complexes are largely static, with little exchange of NS5A occurring among replication foci (22, 57). The second, and perhaps more convincing, line of evidence derives from trans-complementation assays performed with replicons encoding NS5A proteins with different sensitivities to BMS-790052. Specifically, we found that when a defective replicon encoding a sensitive NS5A was trans-complemented with a replicon encoding a resistant NS5A, the “rescued” replicon retained an inhibitor-sensitive phenotype, even while the helper replicon retained an inhibitor-resistant phenotype. These results argue against the wholesale incorporation of NS5A from the helper replicon into the replication complex of the defective replicon as the mechanism for the observed trans-complementation. Instead, it seems more likely that the “defective” NS5A still performed a cis-acting replication function that was blocked by the NS5A inhibitor, while NS5A from the helper replicon provided an additional trans-acting function that was also required for replication. According to this model, NS5A would be similar to other HCV NS proteins by possessing a cis-dominant replication function, likely the result of RNA and protein associations formed during translation and polyprotein processing. NS5A would also, however, possess a distinct function that can be provided in trans.
At this time, we can only speculate on the nature of the putative cis- and trans-acting functions of NS5A. Since the defective replicon was deficient for NS5A hyperphosphorylation (Fig. 4), it is reasonable to propose that the cis-acting replication function of NS5A does not require hyperphosphorylation. This hypothesis is consistent with basally phosphorylated NS5A performing an essential role as a component of the HCV replication complex, as has been previously postulated (2, 6, 13). A corollary of this hypothesis is that the trans-acting replication function of NS5A requires hyperphosphorylation. In support of this hypothesis, Appel et al. (1) showed that the minimal requirement for NS5A trans-complementation in the genotype 1b con1 strain was an NS3-NS5A polyprotein, the same minimal requirement for NS5A hyperphosphorylation (26, 36). Also, although a decrease in hyperphosphorylation has generally been associated with increased replication of the con1 replicon (6, 13), other findings suggest that some hyperphosphorylation may actually be required for replication (37). For example, while amino acid replacements at individual serine residues in the con1 NS5A hyperphosphorylation region generally promote cell culture replication, replacements of multiple serines in this region abolish replication (2). In addition, cellular kinase inhibitors that reduce NS5A hyperphosphorylation and stimulate replication of WT con1 replicons inhibit replication of cell culture-adapted replicons that already display reduced levels of hyperphosphorylation, likely by reducing the level of hyperphosphorylation even further (37). In one scenario, NS5A hyperphosphorylation could mediate interactions with cellular factors that are essential for HCV RNA replication and likely occur outside the replication complex. In this case, the trans-acting function of NS5A would be to create a cellular environment favorable to HCV replication, similar to the “permissive” model for trans-complementation originally proposed by Appel et al. (1).
In conclusion, our findings suggest that NS5A possesses distinct cis- and trans-acting functions in HCV RNA replication and that BMS-790052 inhibits a hyperphosphorylation-independent cis-acting function. Since BMS-790052 also suppresses NS5A hyperphosphorylation (Fig. 1), it is possible that functions of NS5A that are dependent on hyperphosphorylation, such as the trans-replication function proposed above, are also inhibited. This ability to disrupt multiple functions of NS5A could help to explain the extraordinary potency of BMS-790052 and related NS5A inhibitors.
We thank N. Meanwell and M. Belema for critical reading of the manuscript. We thank M. Cockett for support and leadership. We thank our colleagues D. Tenney, D. O'Boyle, J.-H. Sun, P. Nower, J. Lemm, S. Voss, and M. Liu for generously sharing reagents and ideas.
Published ahead of print on 18 May 2011.