Molecular and Structural Basis for the Roles of Hepatitis C Virus Polymerase NS5B Amino Acids 15, 223, and 321 in Viral Replication and Drug Resistance

Abstract

Resistance to the 2′-F-2′-C-methylguanosine monophosphate nucleotide hepatitis C virus (HCV) inhibitors PSI-352938 and PSI-353661 was associated with a combination of amino acid changes (changes of S to G at position 15 [S15G], C223H, and V321I) within the genotype 2a nonstructural protein 5B (NS5B), an RNA-dependent RNA polymerase. To understand the role of these residues in viral replication, we examined the effects of single and multiple point mutations on replication fitness and inhibition by a series of nucleotide analog inhibitors. An acidic residue at position 15 reduced replicon fitness, consistent with its proximity to the RNA template. A change of the residue at position 223 to an acidic or large residue reduced replicon fitness, consistent with its proposed proximity to the incoming nucleoside triphosphate (NTP). A change of the residue at position 321 to a charged residue was not tolerated, consistent with its position within a hydrophobic cavity. This triple resistance mutation was specific to both genotype 2a virus and 2′-F-2′-C-methylguanosine inhibitors. A crystal structure of the NS5B S15G/C223H/V321I mutant of the JFH-1 isolate exhibited rearrangement to a conformation potentially consistent with short primer-template RNA binding, which could suggest a mechanism of resistance accomplished through a change in the NS5B conformation, which was better tolerated by genotype 2a virus than by viruses of other genotypes.

INTRODUCTION

Hepatitis C virus (HCV) is a single-stranded positive-sense RNA virus and the cause of hepatitis C. Nonstructural protein 5B (NS5B) is the RNA-dependent RNA polymerase (RdRp) responsible for replication of the viral genome (1) and adopts a right-hand structure with palm, finger, and thumb domains common to nucleotide polymerases (2–4). In addition, HCV NS5B contains an extension of the finger region called the fingertips, as well as a β-hairpin loop or β-flap insertion into the thumb domain that protrudes deep into the active site and appears to be a hallmark of Flaviviridae RdRps (5). It has been suggested that a significant structural rearrangement of the thumb domain, including relocation of the β-flap and a C-terminal-membrane-anchoring linker from the active-site cavity, must take place to accommodate a growing primer-template RNA chain (3, 5, 6). Indeed, such a rearrangement was observed in an elongation-phase binary complex of an NS5B β-flap deletion mutant with primer-template RNA (7). A crystal structure of an HCV NS5B ternary assembly is not yet available, and thus, insights into nucleoside triphosphate (NTP) binding and incorporation have been derived from biochemical studies and homology modeling.

NS5B is critical for viral replication, exhibits important differences from cellular polymerases, and has become a major target for antiviral drug development. Chain-terminating nucleotide analog inhibitors are promising drugs against HCV because they target the conserved active site and exhibit pan-genotype activity and a high barrier of resistance (8). One such compound, sofosbuvir (GS-7977) (Fig. 1), was recently approved by the FDA. Because clinical resistance is a major concern when treating viral infections, understanding what mutations might arise and confer viral resistance is important in the development of any antiviral agent. We recently reported that the 2′-F-2′-C-methylguanosine (2′-F-2′-C-MeGTP) analogs PSI-352938 and PSI-353661 selected for a number of amino acid changes within the finger and palm domains (changes of S to G at position 15 [S15G], C223H, and V321I) that, in combination, conferred resistance to this series of inhibitors (9). Here, we examine the importance of these residues in viral replication using mutational analysis coupled with assays of replicon fitness and inhibition by nucleotide analog inhibitors, as well as a crystal structure of the NS5B S15G/C223H/V321I mutant of the JFH-1 isolate that we obtained. Our combined cell-based and structural results provide a better understanding of the roles of residues 15, 223, and 321 in HCV replication.

FIG 1.

Chemical structures of HCV NS5B nucleotide analog inhibitors used in this study.

MATERIALS AND METHODS

Compounds.

All nucleoside/-tide analog inhibitors (Fig. 1) were synthesized at Pharmasset and shown to be more than >95 to 99% pure by proton nuclear magnetic resonance, mass spectrometry, and high-performance liquid chromatography analyses.

Replicons and cells.

All plasmid sequences were confirmed by sequencing. Plasmid DNA containing the genotype 1a (GT 1a) replicon of isolate H77 (NCBI reference NC_004102.1) with adaptive amino acid changes P1496L and S2204I and the J6/JFH-1 isolate-derived GT 2a (JFH-1 GT 2a) were licensed from Apath. The JFH-1 GT 2a replicon contains the first 19 amino acids and 3′-nontranslated region (NTR) from the J6 strain (GenBank accession no. AF177036) and the 5′ NTR and NS3-to-NS5B region from the JFH-1 strain (GenBank accession no. AB047639). The JFH-1 replicon open reading frame also contains a Renilla luciferase reporter gene upstream from the neomycin phosphotransferase gene (Neo). We generated a GT 1a replicon with a Renilla luciferase reporter gene in between the 5′ NTR and Neo by cloning and introduced an adaptive change (NS3 K583E) by site-directed mutagenesis to enhance its replication fitness. The Lunet cell line and the Con1 isolate-derived GT 1b ET replicon (GenBank accession no. AJ238799.1) with the firefly luciferase reporter gene and adaptive changes E1202G, T1280I, and K1846T were provided by R. Bartenschlager (University of Heidelberg). The plasmid containing the J6 GT 2a NS5B sequence was synthesized at IDT. GT 3a NS5B was cloned from a human serum sample (EMBL ENA accession no. LN555582). J6 and GT 3a NS5B were each cloned into a Con1 GT 1b replicon that encodes a Renilla luciferase reporter gene to generate the chimeric replicons GT 1b/J6 GT 2a NS5B and GT 1b/GT 3a NS5B.

Mutagenesis and HCV replicon transient-transfection assays.

Mutations were introduced into the appropriate replicon plasmids using QuikChange II site-directed mutagenesis (Agilent) and primers (IDT). Replicon RNA was generated by in vitro transcription from linearized plasmids (digested with ScaI for Con1 GT 1b and GT 1b/J6 NS5B, HpaI for H77 GT 1a, and XbaI for JFH-1 GT 2a) using the RiboMAX large-scale RNA T7 kit (Promega). Replicon RNA (10 μg) was electroporated into Lunet cells as described previously (10) to evaluate replication efficiency and susceptibility to inhibition by nucleoside/-tide analogs. Transient-transfection assays were performed by transfecting cells and seeding them at a density of 8,000 cells/well. Cells were incubated with serially diluted compounds for 4 days prior to the luciferase assay. Replication fitness was determined by normalizing the luciferase expression at 96 h with that at 4 h and then dividing the normalized level of replicon mutant luciferase expression by that of the wild type. HCV replicon inhibition was determined by measuring the levels of luminescence expressed via the firefly or Renilla luciferase reporter gene using Bright-Glo or Renilla-Glo, respectively (Promega). The concentrations at which 50% inhibition was achieved (50% effective concentration [EC₅₀]) were determined using GraphPad Prism.

Expression and purification of JFH-1 NS5B.

JFH-1 NS5B Δ21 was cloned from the plasmid DNA described above by amplification using Platinum PCR supermix high fidelity (HF) (Invitrogen), with forward primer 5′-GACGCTCATGAGCTCCATGTCATACTCCTGGACCGGGGCTC and reverse primer 5′-GTGCTCGAGGCGGGGTCGGGCGCGC, into the pET28a(+) expression vector (EMD) containing a C-terminal His₆ tag. Site-directed mutagenesis was performed using QuikChange II (Agilent) to obtain the plasmid encoding the S15G/C223H/V321I mutations. The plasmid was transformed into Rosetta DE3 E. coli competent cells, and the cells were grown in LB medium at 37°C to an optical density at 600 nm (OD₆₀₀) of 0.6, followed by induction with 0.4 mM isopropyl-β-d-thiogalactopyranoside (IPTG) at 25°C for 4 h. The cells were harvested by centrifugation at 4,000 rpm for 12 min, resuspended in buffer A (50 mM Na₂HPO₄, pH 7.5, 400 mM NaCl, 0.25% Triton X-100, 5 mM β-mercaptoethanol, 20 mM imidazole, and 10% glycerol) containing Complete protease inhibitor cocktail (Roche), lysed by sonication, and clarified by centrifugation at 15,000 rpm for 30 min. NS5B was purified using a Ni-nitrilotriacetic acid (NTA) superflow cartridge (Qiagen) and a linear gradient of 20 mM to 500 mM imidazole. Fractions containing the enzyme were pooled, concentrated, and dialyzed in 50 mM HEPES, pH 7.5, 200 mM NaCl, 0.25% Triton X-100, 0.1 mM dithiothreitol (DTT), 0.1 mM EDTA, and 30% glycerol using a 12- to 14-kDa-cutoff D-tube dialyzer (EMD). The purified protein showed >95% purity based on SDS-PAGE and was stored at −80°C.

RdRp assays.

Polymerase reactions were performed in 20-μl mixtures containing various concentrations of the test compound, 5 μM all four natural NTPs, appropriate amounts of [α-³²P]UTP, 20 ng/μl of a 341-nucleotide internal ribosome entry site (IRES) minus-strand fragment as the RNA template (11), 1 unit/μl of SUPERase•In (Ambion), 40 ng/μl of the recombinant NS5B described above, 1 mM MgCl₂, 0.75 mM MnCl₂, and 2 mM DTT in 50 mM HEPES buffer (pH 7.5). After incubation at 27°C for 120 min, the reaction was quenched with 80 μl of stop solution (12.5 mM EDTA, 2.25 M NaCl, and 225 mM sodium citrate). The radioactive RNA products were separated from unincorporated NTP substrates by passing the quenched reaction mixture through a Hybond N+ membrane (GE Healthcare) using a dot blot apparatus. The membrane was washed four times with a solution containing 0.6 M NaCl and 60 mM sodium citrate, followed by rinsing with water and then with ethanol. Product formation was quantified using a phosphorimager (PerkinElmer). All assays were performed in duplicate, and IC₅₀s (the concentrations at which 50% of the enzyme activity was inhibited) were calculated using GraphFit (Erithacus Software).

Structure determination.

JFH-1 NS5B S15G/C223H/V321I was expressed as described above with the following modifications. To increase yields, NS5B was expressed with the surface solubilization amino acid changes E86Q and E87Q in autoinduction medium overnight at 20°C. After nickel affinity chromatography, the protein was >95% pure as determined by SDS-PAGE and Bio-Rad Experion analysis. Sitting-drop vapor diffusion experiments were initiated at 16°C with 0.4 μl of protein (5 mg/ml in 5 mM Tris, pH 7.5, 200 mM ammonium acetate, 1 mM EDTA, 1 mM DTT, as reported previously [12]) and 0.4 μl of precipitant equilibrated against a reservoir of 80 μl of precipitant. Crystals grown from 0.1 M MMT {malic acid, 2-(N-morpholino)ethanesulfonic acid [MES], Tris} buffer, pH 8.0, and 25% polyethylene glycol (PEG) 1500 were harvested, cryoprotected in 0.1 M MMT buffer, pH 8, 30% PEG 1500, and 20% glycerol, and then vitrified in liquid nitrogen. A data set was collected, and the data were reduced with XDS/XSCALE (13). The structure was solved by molecular replacement in Phaser (14), using the structure deposited in PDB under PDB ID 1YUY (15) as a search model. The final model was obtained after interspersed rounds of refinement in REFMAC5 (16) and manual building in Coot (17).

Protein structure accession number.

The structure of JFH-1 HCV NS5B S15G/C223H/V321I has been deposited in the Protein Data Bank under PDB ID 4OBC.

RESULTS

Effects of amino acid changes at positions 223 and 321 on HCV replicon replication.

Alignment of NS5B isolates from the Los Alamos Database indicates that residues 223 and 321 in the palm domain are both highly conserved. C223 is positioned in a region implicated as the NTP entry channel (3), and the backbone amide of the equivalent residue in the Norwalk virus RdRp recognizes the β-phosphate of the incoming NTP, while its side chain is pointed away from the NTP (18). We hypothesized that large or charged side chains could partially block and interfere with NTP entry. Thus, C223 was changed to Ser, Gln, Trp, Asp, or Arg (Fig. 2A). The isosteric change C223S was the most tolerable, whereas the bulky substitution C223W was not as well tolerated. Replacement with charged amino acids (C223R or C223E) yielded the least fit replicons. Interestingly, the addition of S15G and V321I improved replication for C223Q, -W, and -R (C223Q/-W/-R) replicons.

FIG 2.

Effects of size and polarity at positions 223 and 321 on JFH-1 GT 2a replicon replication. (A) White bars indicate results for mutants with single amino acid substitutions at position 223. Gray bars indicate results for mutants with triple substitutions of C223 variants along with S15G and V321I. (B) White bars indicate results for mutants with single amino acid substitutions at position 321. Gray bars indicate results for mutants with triple substitutions of V321 variants along with S15G and C223H. Means and standard deviations were determined from at least three independent experiments performed in duplicates. Data from previous fitness analyses of C223H and V321I are included for comparison. *, value was below 0.05%.

Residue V321 is buried within the palm domain, surrounded by other hydrophobic residues. To evaluate the effects of size and charge, we replaced V321 with Ala, Leu, Phe, Glu, or Arg (Fig. 2B). Residue 321 favors larger hydrophobic side chains (V321L/-F) over smaller ones (V321A). The addition of S15G and C223H improved replication for V321A/-L/-F replicons. Charged side chains were not tolerated at position 321, consistent with its hydrophobic environment.

Effects of amino acid changes at position 15 on HCV replicon replication.

Residue 15 is located near the incoming RNA template (7) and may be important in HCV replication, since our previous fitness studies (9) showed that a JFH-1 GT 2a replicon containing S15G replicated with lower fitness (Fig. 3A). Both neutral and basic residues were tolerated at this position. In contrast, the acidic substitutions S15D and S15E significantly reduced replicon fitness. This was not surprising, given that this residue resides near the entrance for the polyanionic RNA template and S15D and S15E would be negatively charged at neutral pH, thereby repelling the RNA template from entering the active site. Again, the addition of C223H and V321I to the single amino acid variants S15G/-K/-R/-N improved replication.

FIG 3.

Effects of size and polarity at position 15 on JFH-1 GT 2a and Con1 GT 1b replicon replication. (A) White bars indicate results for JFH-1 mutants with single amino acid substitutions. Gray bars indicate results for JFH-1 mutants with triple substitutions of S15 variants along with C223H and V321I. Data from previous fitness analyses of S15G are included for comparison. (B) White bars indicate results for Con1 mutants with single amino acid substitutions. Gray bars indicate results for mutants with triple substitutions of A15 variants along with C223H and V321I. Means and standard deviations were determined from at least three independent experiments performed in duplicates. *, value was below 1%.

To further verify that acidic side chains were not tolerated at this position, we introduced substitutions into Con1 GT 1b replicons (Fig. 3B). Instead of a serine residue, Con1 GT 1b replicons contained an alanine at position 15. The results showed that Con1 replicons containing A15D/-E were severely impaired for replication, while A15K yielded wild-type replication levels. In addition, Con1 NS5B with A15G exhibited patterns of reduction as a single point variant and when rescued with C223H/V321I that were similar to the patterns observed for JFH-1, albeit with lower fitness.

Effects of mutagenesis at positions 15, 223, and 321 on cross-resistance.

The effects of these substitutions at positions 15, 223, and 321 on inhibition by nucleoside/-tide analogs PSI-352938, GS-7977, PSI-6130, IDX-184, and INX-189 (Fig. 1) were examined (Table 1). Among the C223 JFH-1 variants, none of the single amino acid substitutions conferred resistance to PSI-352938. Only the S15G/C223S/V321I and S15G/C223Q/V321I substitutions shifted the EC₅₀ of PSI-352938. As expected, no cross-resistance was observed for the other nucleoside/-tide analogs. Interestingly, JFH-1 replicons containing C223W/-R alone or in combination with S15G and V321I became significantly more sensitive to the other nucleoside/-tide analogs. In general, viruses with resistance variants against one nucleotide analog inhibitor that exhibit loss of fitness may appear more susceptible to different nucleotide analog inhibitors than wild-type strains.

TABLE 1.

Effects of C223 mutagenesis in JFH-1 replicons on nucleoside/-tide inhibition

JFH-1 variantb	Avg EC50 ± SD (fold change) ofa:
	PSI-352938	GS-7977	PSI-6130	IDX-184	INX-189
C223S	2.6 ± 1.3	0.96 ± 0.36	0.57 ± 0.14	0.95 ± 0.19	0.96 ± 0.099
S15G/C223S/V321I	5.7 ± 3.1	0.99 ± 0.44	0.69 ± 0.13	0.49 ± 0.14	0.60 ± 0.21
C223Q	2.2 ± 1.1	0.86 ± 0.40	0.46 ± 0.090	0.49 ± 0.15	0.66 ± 0.18
S15G/C223Q/V321I	6.0 ± 3.2	0.76 ± 0.44	0.48 ± 0.091	0.29 ± 0.084	0.34 ± 0.054
C223W	1.3 ± 0.29	0.27 ± 0.12	0.082 ± 0.025	0.082 ± 0.027	0.064 ± 0.033
S15G/C223W/V321I	2.0 ± 1.2	0.14 ± 0.091	0.085 ± 0.020	0.053 ± 0.028	0.036 ± 0.020
C223R	0.19 ± 0.082	0.062 ± 0.012	0.041 ± 0.012	0.054 ± 0.015	0.042 ± 0.025
S15G/C223R/V321I	0.69 ± 0.29	0.13 ± 0.058	0.097 ± 0.035	0.065 ± 0.026	0.051 ± 0.018

^aFold changes in EC₅₀s were determined by normalizing the EC₅₀s from the replicon variants with that of the wild type. All values are the average ± SD of the results of at least three independent experiments performed in duplicates.

^bThe C223 substitution is in boldface.

Among the V321 variants capable of replication (the variant with a mutation to Asp failed to replicate), we observed that reduction in PSI-352938 activity was dependent both on the combination with S15G and C223H and on the size of the side chain (Table 2). Essentially no change in EC₅₀s was observed for the variants with single point changes, and resistance was not observed for the other nucleoside/-tide analogs.

TABLE 2.

Effects of V321 mutagenesis in JFH-1 replicons on nucleoside/-tide inhibition

JFH-1 variantb	Avg EC50 ± SD (fold change) ofa:
	PSI-352938	GS-7977	PSI-6130	IDX-184	INX-189
V321A	0.68 ± 0.10	1.1 ± 0.29	1.2 ± 0.47	1.1 ± 0.14	0.93 ± 0.17
S15G/C223H/V321A	1.2 ± 0.25	0.56 ± 0.14	0.63 ± 0.15	0.40 ± 0.072	0.23 ± 0.046
V321L	1.2 ± 0.34	1.0 ± 0.32	1.0 ± 0.34	0.69 ± 0.25	0.67 ± 0.087
S15G/C223H/V321L	3.0 ± 0.70	1.0 ± 0.41	0.91 ± 0.30	0.44 ± 0.055	0.28 ± 0.12
V321F	0.70 ± 0.17	0.37 ± 0.089	0.44 ± 0.21	0.20 ± 0.043	0.12 ± 0.045
S15G/C223H/V321F	5.7 ± 1.9	0.34 ± 0.072	0.48 ± 0.11	0.23 ± 0.050	0.11 ± 0.029

^aFold changes in EC₅₀s were determined by normalizing the EC₅₀s from the replicon variants with that of the wild type. All values are the average ± SD of the results of at least three independent experiments performed in duplicates.

^bThe V321 substitution is in boldface.

The effects of modifying residue 15 were evaluated using both JFH-1 (Table 3) and Con1 (Table 4) replicons. JFH-1 S15 variants with single amino acid substitutions remained susceptible to nucleoside/-tide analogs, and reduction of PSI-352938 activity was only observed in the presence of C223H and V321I together with a nonacidic amino acid at position 15 (Table 3). The only residue 15 substitution that produced replication-competent replicons in Con1 was A15K as a single mutation or part of a triple mutation, and a more modest increase in the EC₅₀ was observed for Con1 (2.6-fold) (Table 4) than for JFH-1 (5.9-fold) (Table 3) with the A15K/C223H/V321I substitutions against PSI-352938 only.

TABLE 3.

Effects of S15 mutagenesis in JFH-1 replicons on nucleoside/-tide inhibition

JFH-1 variantb	Avg EC50 ± SD (fold change) ofa:
	PSI-352938	GS-7977	PSI-6130	IDX-184	INX-189
S15E	0.097 ± 0.039	0.11 ± 0.022	0.11 ± 0.026	0.13 ± 0.082	0.10 ± 0.085
S15E/C223H/V321I	0.41 ± 0.11	0.13 ± 0.057	0.11 ± 0.046	0.068 ± 0.028	0.040 ± 0.020
S15K	0.77 ± 0.17	1.1 ± 0.16	1.1 ± 0.11	0.86 ± 0.084	0.94 ± 0.19
S15K/C223H/V321I	5.9 ± 0.94	0.79 ± 0.085	0.84 ± 0.26	0.44 ± 0.079	0.24 ± 0.068
S15R	0.86 ± 0.24	1.0 ± 0.20	1.0 ± 0.17	0.94 ± 0.24	0.94 ± 0.086
S15R/C223H/V321I	5.0 ± 0.56	0.58 ± 0.037	0.70 ± 0.28	0.37 ± 0.074	0.29 ± 0.034
S15N	0.77 ± 0.086	0.91 ± 0.045	0.98 ± 0.17	1.1 ± 0.16	1.0 ± 0.025
S15N/C223H/V321I	4.1 ± 0.59	0.52 ± 0.10	0.63 ± 0.28	0.38 ± 0.073	0.29 ± 0.061

^aFold changes in EC₅₀s were determined by normalizing the EC₅₀s from the replicon variants with that of the wild type. All values are the average ± SD of the results of at least three independent experiments performed in duplicates.

^bThe S15 substitution is in boldface.

TABLE 4.

Effects of A15 mutagenesis in Con1 replicons on nucleoside/-tide inhibition

Con1 variantb	Avg EC50 ± SD (fold change) ofa:
	PSI-352938	GS-7977	PSI-6130	IDX-184	INX-189
A15K	1.1 ± 0.33	1.3 ± 0.39	1.3 ± 0.21	1.1 ± 0.21	1.1 ± 0.46
A15K/C223H/V321I	2.6 ± 0.71	1.1 ± 0.32	0.84 ± 0.27	0.62 ± 0.21	0.27 ± 0.093

^aFold changes in EC₅₀s were determined by normalizing the EC₅₀s from the replicon variants with that of the wild type. All values are the average ± SD of the results of at least three independent experiments performed in duplicates.

^bThe A15 substitution is in boldface.

Effects of the S15G/C223H/V321I mutations in GT 1a and in 1b/2a and 1b/3a chimeras.

JFH-1 replicons tolerated the S15G/C223H/V321I mutations much better than Con1 replicons containing these mutations (9). For this study, H77 GT 1a or chimeric replicons containing Con1 GT 1b as the backbone and either J6 GT 2a or GT 3a NS5B were constructed as wild type or as the C223H, C223H/V321I or S15G/C223H/V321I variant and assayed for replication fitness (Table 5). H77 GT 1a replicons containing the C223H substitution showed reduced replication levels, which were further lowered by the addition of A15G and V321I mutations. The J6 GT 2a chimeric replicons tolerated the single, double, or triple mutation. Similar to GT 1 replicons, chimeric Con1/GT 3a NS5B replicons with amino acid substitutions at sites 15, 223, and/or 321 were severely impaired for replication. Dose-response studies using GT 1a and GT 1b/J6 NS5B replicon mutants showed that the single C223H amino acid change did not significantly affect PSI-352938 activity, whereas double or triple mutants showed increased resistance (Table 5).

TABLE 5.

Effects of changes in residues 15, 223, and 321 in NS5Bs from other isolates

Subgenomic replicon variant	Avg ± SD of indicated value for varianta:
	C223H		C223H/V321I		A-/S15Gb/C223H/V321I
	Replication fitness (%)	PSI-352938 EC50 (fold change)	Replication fitness (%)	PSI-352938 EC50 (fold change)	Replication fitness (%)	PSI-352938 EC50 (fold change)
GT 1a	28.3 ± 4.8	1.5 ± 0.2	6.0 ± 2.2	3.7 ± 0.5	4.4 ± 0.7	8.7 ± 0.8
GT 1b/J6 GT 2a	74.6 ± 15.3	2.1 ± 0.5	43.6 ± 12.1	3.7 ± 0.9	69.5 ± 17.5	8.0 ± 1.4
GT 1b/GT 3a	8.5 ± 4.4	ND	5.3 ± 2.8	ND	5.5 ± 1.3	ND

^aAll values are the average ± SD of the results of at least three independent experiments performed in duplicates. Replication fitness values are the percentage of the value for the wild type. The EC₅₀ values are the fold change relative to the value for the wild type. ND, not determined, because luciferase expression levels in GT 3a variants were similar to background values.

^bResidue 15 is an Ala in GT 1a and GT 3a and a Ser in J6 GT 2a.

Biochemical studies of JFH-1 NS5B variants.

The effects of the S15G/C223H/V321I triple variant on RNA synthesis and the apparent IC₅₀s of PSI-352666 (active 5′-triphosphate of PSI-352938 and PSI-353661) and 2′-C-methylguanosine triphosphate (2′-C-MeGTP; the active form of IDX-184) were examined. The activity of the triple variant enzyme was approximately 50% higher than that of the wild-type enzyme, and both PSI-352666 and 2′-C-MeGTP showed dose-dependent inhibition (Table 6). The variant enzyme demonstrated approximately 2-fold-reduced sensitivity to PSI-352666, whereas 2′-C-MeGTP was 2.5-fold more active, demonstrating that 2′-F-2′-C-MeGTP was sensitive to these amino acid changes and 2′-OH-2′-C-MeGTP was not.

TABLE 6.

Relative activity and inhibition of genotype 2a wild-type and mutant NS5B

Enzyme	Avg relative activity ± SD	Avg apparent IC50 ± SD (μM)a
		PSI-352666	2′-C-MeGTP
WT	1	17.1 ± 4.9	17.9 ± 5.3
Triple variantb	1.48 ± 0.14	32.8 ± 7.7	7.05 ± 1.75

^aAll values are the average ± SD of the results of at least four independent experiments performed in duplicates.

^bThe triple variant is S15G/C223H/V321I.

JFH-1 S15G/C223H/V321I NS5B crystal structure.

We obtained a crystal structure of JFH-1 S15G/C223H/V321I NS5B at 2.5-Å resolution (Fig. 4A; Table 7). Each amino acid change was clearly defined by the electron density. C223H adopted the third most common rotamer conformation, while V321I adopted the most common rotamer conformation. Surprisingly, JFH-1 NS5B S15G/C223H/V321I exhibited a significantly different conformation (Fig. 4B) than most other GT 2a crystal structures, such as those of JFH-1 (PDB ID 3I5K [12]; C_α root mean square deviation [RMSD], 1.80 Å, and PDB ID 2XXD [19]; C_α RMSD, 1.76 Å), the GT 2a consensus sequence as apo (PDB ID 1YUY [15]; C_α RMSD, 1.58 Å) or in complex with a thumb site nonnucleoside inhibitor (PDB ID 1YVX [15]; C_α RMSD, 1.10 Å), or the J6 isolate (PDB ID 2XWH [19]; C_α RMSD, 1.79 Å). Although the finger and palm domains were quite similar, significant rearrangement occurred in the fingertip and thumb domains, including a partial disordering of the C terminus and a refolding of the loop that spans the primer grip helix and the primer buttress helix (residues W397 to N406) (Fig. 4B). All of these rearrangements were observed previously in our structures of a β-hairpin loop deletion construct of GT 2a JFH-1 (Fig. 4C) either in the apo state (PDB ID 4E76; C_α RMSD, 0.39 Å) or bound to primer-template RNA (PDB IDs 4E78 and 4E7A; C_α RMSDs, 0.75 to 0.77 Å) (7). The large rearrangement most affected the positioning of S15G, which exhibited a peptide backbone flip and a 2.2-Å C_α distance change (Fig. 5A). The C223H variant had only a small change in the C_α positioning (0.4 Å), and the surrounding residues displayed virtually no change (Fig. 5B). The C_α position of V321I also had only a minor change (0.6 Å, Fig. 5C). Given the V321I rotamer conformation, C_δ of V321I would have closely approached C_γ2 of I363 (2.56 Å) in the apo ground state (e.g., PDB ID 3I5K) (Fig. 5D). However, the thumb domain movement observed here generated more space in the vicinity of this residue, with an observed 4.02-Å distance. Thus, V321I may nudge the loop that connects the palm and thumb domains (residues I363 to N369) and consequently reposition the thumb domain to promote the conformational rearrangement observed with the triple variant (Fig. 5D).

FIG 4.

Crystal structure of JFH-1 S15G/C223H/V321I NS5B. The JFH-1 S15G/C223H/V321I structure is colored with the finger subdomain in blue, the palm subdomain in magenta, the thumb subdomain in green, and the thumb subdomain β-flap insertion in yellow (A) and is aligned against residues 62 to 350 (7) of JFH-1 wild type (PDB ID 3I5K) (12) (gray) (B) or of JFH-1 with a β-flap deletion and bound to primer-template RNA (PDB ID 4E78) (7) (gray, with the primer carbon backbone colored in cyan and the template carbon backbone colored in brown) (C). Residues 15, 223, and 321 are shown as spheres in all structures.

TABLE 7.

Crystallographic statistics for JFH-1 HCV NS5B S15G/C223H/V321I^a

Parameter	Overall	Highest shell
Beamline	SSRL 9-2
Space group	P65
Unit cell
a = b, c (Å)	140.21, 92.57
α = β, γ (°)	90, 120
Solvent content (%)	69.7
Vm (Å3/Da)	4.06
Resolution (Å)	50–2.5	2.57–2.50
I/σ	21.9	2.3
Completeness (%)	99.7	97.3
Rmerge	0.085	0.638
CC1/2	99.9	74.5
Multiplicity	7.8	3.8
Reflections	35,841	2,877
Mosaicity	0.2
R	0.190	0.263
Rfree	0.223	0.335
Ramachandran
Favored (%)	97.8
Allowed (%)	100
MolProbity score	99th percentile

^aThe structure was deposited in Protein Data Bank under PDB ID 4OBC.

FIG 5.

Structural analysis of JFH-1 S15G, C223H, and V321I amino acid changes. The JFH-1 S15G/C223H/V321I structure is shown with colors as described in the legend to Fig. 4, whereas the wild-type structure (PDB ID 3I5K) (12) is colored in gray and aligned against residues 62 to 350. (A) The S15G amino acid change induces a peptide bond flip at this position. (B) The C223H amino acid change has virtually no effect in this region or on nearby residues proposed to be involved in metal ion binding (D220, D318, or D319) or recognition of the triphosphate of the incoming nucleotide (R48 or R158). (C and D) The V321I mutation may reposition the palm-thumb hinge region to facilitate the large-scale rearrangement of the thumb domain as viewed from the interior (C) or exterior (D) of the polymerase. Although some nearby hydrophobic residues (L204 and L314) change very little, I363 moves significantly and consequently may induce movement of the palm-thumb hinge, with long-range effects like the movement of the C_α of R377 by 9 Å (indicated with an arrow).

DISCUSSION

We previously speculated that the fitness profile of the S15G/C223H/V321I variant was unique to genotype 2a JFH-1 and demonstrated that Con1 GT 1b replicons with the A15G/C223H/V321I substitutions were highly unfit for replication (9). We attempted to determine whether this was due solely to the NS5B region or to other cis factors from within JFH-1 by swapping JFH-1 NS5B into a Con1 replicon, but both Con1/JFH-1 NS5B and JFH-1/Con1 NS5B chimeric replicons failed to replicate (9). In the current study, we generated three replication-competent replicons containing the triple mutations within an NS5B from a different genotype or strain (H77 GT 1a, Con1 GT 1b/GT 3a NS5B, and Con1 GT 1b/J6 GT 2a NS5B). Data from these replicon variants, together with data from the JFH-1 replicon variants, confirmed that multiple amino acid changes were required to confer resistance to 2′-F-2′-C-methylguanosine nucleotide analog inhibitors. Furthermore, GT 2a NS5B from both J6 and JFH-1 strains tolerated these changes, while replicons with NS5B from GT 1a, 1b, and 3a containing these changes were significantly less fit. Consistent with the previous study on the S15G/C223H/V321I variant, none of the amino acid changes studied in this work conferred resistance to the approved uridine nucleotide prodrug sofosbuvir (GS-7977) or a cytidine analog, PSI-6130, despite having the same substitutions in the sugar moiety as PSI-352938 (9). In addition, the replicon activity of guanosine nucleotide prodrugs with a 2′-OH-2′-C-methyl sugar substitution, IDX-184 and INX-189, was not affected by the amino acid changes described in this and the previous study (9). Although the exact molecular mechanism for the resistance to 2′-F-2′-C-methylguanosine nucleotide analogs is still unknown, a combination of a guanine (or purine) nucleobase and a 2′-F-2′-C-methyl substitution in the sugar moiety is essential for these three genotype 2a variants to confer resistance in vitro.

The main goal of this study was to better define the roles of amino acids 15, 223, and 321 in HCV RNA replication. In addition to single point changes at these positions, each substitution was combined with the other two amino acid changes that were previously determined to be essential in conferring resistance to 2′-F-2′-C-methylguanine nucleotides. Mutagenesis studies confirmed the importance of each of these residues. Amino acid 15 is located on the surface of the finger domain near the incoming template (7) and is more polymorphic than positions 223 and 321. It is primarily an Ala in GT 1, 4, and 6, a Ser in GT 3 and 5, and a Ser or Gly (∼1:1 ratio) in GT 2 viruses. The reduced fitness of replicons with an acidic side chain could reflect disruption of interactions of the NS5B with the RNA template. Residue 223 may interact with the incoming NTP (3), and it could not tolerate a change to Glu or Arg. Furthermore, the bulky side chain of a tryptophan reduced viral fitness and could block NTP entry or reduce the size of this channel. V321, located within the hydrophobic interior of the palm domain, is also a highly conserved residue, and its replacement with a smaller residue resulted in loss of fitness, while mutation to a larger hydrophobic residue was tolerated.

The crystal structure of JFH-1 NS5B S15G/C223H/V321I unexpectedly showed the same structural rearrangements as those observed previously for an elongation-phase binary complex of a β-hairpin loop deletion polymerase bound to primer-template RNA, in which the thumb domain rotates around the thumb hinge, accompanied by stretching of the strand that contains S15 and flexing of the fingertip domains (7). In comparison with earlier structures (12, 15, 19), the β-hairpin loop retracted ca. 3 to 5 Å to generate more space in the active-site cavity in the vicinity of the catalytic residues, which may allow entry of the 3′ end of the viral RNA template and nucleotides for replication initiation; further structure studies aimed at nucleic acid binding to this more open conformation are in progress. It is tempting to speculate that the S15G/C223H/V321I triple variant, and specifically the V321I mutation (Fig. 5C and D), may facilitate these rearrangements to promote a catalytically competent conformation for replication initiation. However, we cannot rule out the possibility that the crystal form rather than these three amino acid changes is responsible. The current structure appears in the same crystal form as that reported previously (7), which includes the surface solubilization modification E86Q that is involved in crystal lattice contacts. Interestingly, our biochemical de novo RNA synthesis assays (Table 6) showed that JFH-1 NS5B S15G/C223H/V321I was approximately 50% more efficient at RNA synthesis than the wild type. The two rate-limiting steps during de novo RNA synthesis have been identified as the synthesis of the first dinucleotide and the transition from initiation to elongation, which requires rearrangement of NS5B (6). Therefore, it is possible that the increase in our de novo RNA synthesis activity resulted from enhancement of a rate-limiting step by the triple-mutant enzyme with a conformation compatible with RNA binding (7), which is consistent with the observation that JFH-1 is able to synthesize the first dinucleotide and transition to elongation more efficiently than other isolates (20). The molecular mechanism of how this conformation could confer resistance to 2′-F-2′-C-methylguanosine triphosphates remains unclear. It can be speculated that binding of the triphosphate analog, the initiation, the transition, and/or the elongation steps or transitions between them could each be affected as a result of the changes in NS5B conformation. These changes may not be captured in the current static-state crystal structure. In addition, because there was a discrepancy in the fold changes in activity caused by the S15G/C223H/V321I mutation in the replicon assay (10- to 15-fold) and the biochemical assay (2-fold), interactions with other HCV nonstructural proteins and/or host factors in the replication complex may be involved in the mechanism of resistance and such interactions may be affected by the conformational changes. The site-directed mutagenesis studies of each of the three amino acids as single or triple amino acid changes supported the roles of these residues in regulating HCV replication and the possibility that conformational selection within the NS5B protein could also play a role in conferring resistance to 2′-F-2′-C-methylguanosine analogs.