Hepatitis C Virus Nucleotide Inhibitors PSI-352938 and PSI-353661 Exhibit a Novel Mechanism of Resistance Requiring Multiple Mutations within Replicon RNA^,

Abstract

PSI-352938, a cyclic phosphate nucleotide, and PSI-353661, a phosphoramidate nucleotide, are prodrugs of β-d-2′-deoxy-2′-α-fluoro-2′-β-C-methylguanosine-5′-monophosphate. Both compounds are metabolized to the same active 5′-triphosphate, PSI-352666, which serves as an alternative substrate inhibitor of the NS5B RNA-dependent RNA polymerase during HCV replication. PSI-352938 and PSI-353661 retained full activity against replicons containing the S282T substitution, which confers resistance to certain 2′-substituted nucleoside/nucleotide analogs. PSI-352666 was also similarly active against both wild-type and S282T NS5B polymerases. In order to identify mutations that confer resistance to these compounds, in vitro selection studies were performed using HCV replicon cells. While no resistant genotype 1a or 1b replicons could be selected, cells containing genotype 2a JFH-1 replicons cultured in the presence of PSI-352938 or PSI-353661 developed resistance to both compounds. Sequencing of the NS5B region identified a number of amino acid changes, including S15G, R222Q, C223Y/H, L320I, and V321I. Phenotypic evaluation of these mutations indicated that single amino acid changes were not sufficient to significantly reduce the activity of PSI-352938 and PSI-353661. Instead, a combination of three amino acid changes, S15G/C223H/V321I, was required to confer a high level of resistance. No cross-resistance exists between the 2′-F-2′-C-methylguanosine prodrugs and other classes of HCV inhibitors, including 2′-modified nucleoside/-tide analogs such as PSI-6130, PSI-7977, INX-08189, and IDX-184. Finally, we determined that in genotype 1b replicons, the C223Y/H mutation failed to support replication, and although the A15G/C223H/V321I triple mutation did confer resistance to PSI-352938 and PSI-353661, this mutant replicated at only about 10% efficiency compared to the wild type.

INTRODUCTION

According to the World Health Organization, hepatitis C virus (HCV) currently infects more than 170 million people worldwide. The majority of these patients develop chronic liver disease, with a 20% rate of liver cirrhosis and fibrosis, and up to 5% could progress to hepatocellular carcinoma. There is no vaccine available, and the current standard of care (SOC), which combines pegylated alpha interferon (peg-IFN-α) and ribavirin, has limited efficacy and may be associated with a number of side effects (3, 5). Efforts to develop direct-acting antiviral agents (DAA) have focused on compounds that target viral proteins critical for HCV replication. Recently, two DAA targeting the NS3/4A protease, boceprevir (Victrelis) and telaprevir (Incivek), in combination with the SOC have been approved as treatment for hepatitis C. Other antiviral compounds currently in clinical development include NS5A inhibitors, nucleoside/nucleotide analogs, and nonnucleoside inhibitors that target the NS5B polymerase. While DAA strategies have shown promising results in reducing viral load, viral breakthrough related to the emergence of resistance has been observed and has since become a major concern in the development of DAA as potential therapy for HCV infection (8, 28).

The high genetic diversity of HCV and error-prone nature of the NS5B polymerase generate a pool of HCV variants with naturally occurring resistant mutations that can be selected for in the presence of antiviral compounds (10). Mutations that confer resistance to the NS3 protease, NS5A, and NS5B nonnucleoside inhibitors have been readily selected both in vitro and in vivo (7, 16, 22, 24). In contrast, a higher barrier of resistance exists for NS5B nucleoside analogs (17). To date, only two mutations within HCV NS5B have been found to be associated with decreased susceptibility to nucleoside/-tide analogs: S96T, which confers resistance to 4′-azidocytidine (R1479) (13), and S282T, which confers resistance to 2′-C-methyl-modified nucleoside/-tides (e.g., IDX184 and INX-08189) (18, 29; J. F. McCarville et al., presented at the 5th International Workshop on Hepatitis C—Resistance and New Compounds, Boston, MA, 24 to 25 June 2010) and 2′-F-2′-C-methyl pyrimidine nucleoside/-tide analogs (e.g., RG7128 and PSI-7977) (1, 26). Since these are highly conserved residues, replicons containing the S96T or S282T amino acid change are severely impaired for replication (1, 18). Consequently, emergence of resistance associated with nucleoside/-tide inhibitors in clinical studies has been uncommon: to date, no resistance-related breakthrough has been observed in patients treated with either RG7128 or PSI-7977 alone or in combination with the SOC (14, 20).

Recently we described the in vitro anti-HCV activities and cytotoxicity profiles of PSI-352938 and PSI-353661, both prodrugs of 2′-F-2′-C-methylguanosine-monophosphate (4, 11). PSI-352938 is a cyclic monophosphate prodrug and is the first of its class to be evaluated in clinical trials, while PSI-353661 is a phosphoramidate prodrug designed to improve the delivery of the compound to the liver. PSI-352938 and PSI-353661 inhibit HCV genotype (GT) 1b replicon replication with 50% effective concentrations (EC₅₀s) of 0.13 ± 0.076 μM and 3.0 ± 1.4 nM, respectively, and are similarly active against GT 1a and 2a replicons and infectious viruses (4, 11). Metabolism of PSI-352938 and PSI-353661 generates the same 5′-triphosphate metabolite, PSI-352666 (2′-F-2′-C-methylguanosine-triphosphate), which is similarly active against NS5B polymerases from GT 1 to 4 (11). Our previous cross-resistance studies also showed that PSI-352938 and PSI-353661 remained fully active against both the S96T and S282T replicons (4, 11).

In order to identify mutations selected for by PSI-352938 and PSI-353661, we initiated selection studies first using GT 1b replicon cells. Our effort was then expanded to include cells containing either GT 1a or 2a replicons. The results showed that PSI-352938 and PSI-353661 selected a number of novel amino acid changes within the GT 2a NS5B region. Phenotypic analysis was performed to identify which of these amino acid changes were responsible for mediating the observed resistance. The effect of these changes on HCV replicon replication was also determined. The susceptibility of these resistant replicons to other compounds was examined using a panel of nucleoside/-tide analogs and other classes of HCV inhibitors currently in clinical development.

MATERIALS AND METHODS

Compounds.

PSI-352938, PSI-353661, PSI-7977, PSI-6130, IDX-184, INX-08189, 2′-C-methyladenosine, RG7227, BMS-790052, and HCV-796 were synthesized at Pharmasset. Telaprevir was synthesized by ACME Biosciences (Palo Alto, CA). Interferon and ribavirin were purchased from Sigma (St. Louis, MO).

Cells and viruses.

The genotype (GT) 1b clone A replicon cell line (Apath LLC, Brooklyn, NY) was maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 0.1 mM nonessential amino acids, 2 mM l-glutamine, 1 mM sodium pyruvate, 50 U/ml penicillin, 50 μg/ml streptomycin, and 1 mg/ml G418 (Invitrogen, Carlsbad, CA). The GT 1b-derived ET replicon (Con1 strain; GenBank accession no. AJ238799.1) with the adaptive mutations E1202G, T1280I, and K1846T (15) and the Lunet cell line (9) were kindly provided by R. Bartenschlager (University of Heidelberg, Heidelberg, Germany). Plasmid DNA containing the GT 1a replicon (H77 strain; NCBI reference no. NC_004102.1) with the adaptive mutations P1496L and S2204I (2) and the GT 2a J6/JFH-1 replicon were licensed from Apath. The J6/JFH-1 replicon contains a partial core (first 19 amino acids) and 3′ nontranslated region (NTR) from strain J6 (GenBank accession no. AF177036) and the 5′ NTR and NS3-to-NS5B region from strain JFH-1 (GenBank accession no. AB047639). The ET and J6/JFH-1 replicons also contain the firefly and Renilla luciferase reporter genes, respectively. The GT 1a, GT 1b, and GT 2a replicon cell lines were each generated by electroporating the corresponding replicon RNA (10 μg) into the Lunet cells as described previously, followed by G418 selection (12).

In vitro selection of HCV replicon cells.

GT 1a, 1b, and 2a replicon cells were cultured in the presence of G418 (0.75 mg/ml for GT 1a, 0.25 mg/ml for GT 1b and 2a) and increasing concentrations of PSI-352938 or PSI-353661 starting at their respective EC₅₀ or EC₉₀. As a no-compound control, replicon cells were maintained in parallel in the equivalent percent volume (0.2%) of dimethyl sulfoxide (DMSO) (Sigma). Cells were passaged whenever they reached ∼80% confluence and replenished with G418 medium containing fresh compound. At various passages, cells were tested for sensitivity to PSI-352938 and PSI-353661. For each assay, 3-fold dilutions of test compound were added to cells in duplicate and incubated at 37°C in a humidified 5% CO₂ atmosphere for 4 days. Inhibition of HCV replicon RNA replication was determined by real-time PCR (RT-PCR) using primers that anneal to the 5′ untranslated region (27) or by measuring the levels of luminescence expressed via the firefly or Renilla luciferase reporter gene using the Bright-Glo or Renilla-Glo reagent, respectively (Promega, Madison, WI). EC₅₀ and EC₉₀, the concentrations at which 50% and 90% inhibition were achieved, were determined using GraphPad Prism software (San Diego, CA). Aliquots of cells were also saved for RNA isolation, cDNA synthesis, and PCR amplification for sequencing analysis.

Total RNA transfection experiment.

Total RNA was isolated from about 1 × 10⁶ DMSO control or compound-selected replicon cells using an RNeasy minikit (Qiagen, Valencia, CA), and 10 to 15 μg of this RNA was electroporated into Lunet cells. Stable cell lines containing the transfected replicon RNA were established by culturing in the presence of G418, which were tested for sensitivity to PSI-352938 and PSI-353661 as described above. Other nucleoside/-tide analogs, including PSI-7977, PSI-6130, IDX-184, INX-08189, and 2′-C-methyladenosine, were similarly tested. The change (fold) in activity was determined by dividing the EC₅₀s for cells transfected with total RNA from either the PSI-352938 selection or the PSI-353661 selection by those for cells transfected with total RNA from the DMSO control cells.

Genotypic analysis.

RNA extracted from the selected cells was reverse transcribed into cDNA, and the NS5B region was amplified with a Titan One-Tube RT-PCR kit (Roche Applied Science) and the appropriate primer sets (IDT DNA Technologies). Primers for GT 1b NS5B were 1bNS5B S1 (5′ CGT AAG CGA GGA GGC TAG T) and 1bNS5B A1 (5′ GTG TTT AGC TCC CCG TTC ATC). GT 1a NS5B primers were 1aNS5B S1 (5′ GTT GAG TCC TAT TCT TC), 1aNS5B S2 (5′ GGC CGA CAC GGA AGA TGT C), 1aNS5B A1 (5′ GAG TGT TTA CCC CAA CCT TCA TCG G), and 1aNS5B A2 (5′ GGC CTA AGA GGC CGG AGT G). Primers for GT 2a NS5B were 2aNS5B S1 (5′ CGA GGA GGA CGA TAC CAC CGT GTG CTG CTC C), 2aNS5B S2 (5′ GGT AGC TCC CGG TTC GGG C), and 2aNS5B A1 (5′ GTG TAC CTA GTG TGT GCC GCT CTA CCG AGC GG). For the GT 2a NS3-NS5B replicon sequencing, cDNA was first reversed transcribed from the selected cells using a Transcriptor first-strand cDNA synthesis kit (Roche). The NS3-to-NS5B region was amplified using the primers 2aNS3 S1 (5′ CCG TCT TTT GGC AAT GTG) and S2 (5′ GAT AAT ACC ATG GCT CCC), 2a3UTR A1 (5′ AAA GGG ACA GTT AGC TAT GGA GTG), and 2aNS5B A2 (5′ GTG CCG CTC TAC CGA GCG) and the Expand Long Range dNTPack (Roche). The amplified products were gel purified and cloned into the TA TOPO pCR4 sequencing vector (Invitrogen) for subsequent clonal sequencing (Genewiz, South Plainfield, NJ). Sequence alignments were performed using Lasergene (Madison, WI) DNAStar software.

Phenotypic analysis.

Amino acid substitutions, as either single changes or combinations of changes, were introduced into the wild-type ET or J6/JFH-1 replicon plasmid using a QuikChange II site-directed mutagenesis kit (Stratagene) and primers with nucleotide substitutions that corresponded to changes at sites 15, 222, 223, 320, and 321. Replicon plasmids containing the appropriate mutation(s) were confirmed by sequencing (Genewiz). Plasmids were each linearized with ScaI for ET replicons, HpaI for H77 replicons, or XbaI for J6/JFH-1 replicons. The linearized plasmid was purified as previously described (12), and 10 to 20 μg was used for in vitro transcription to generate replicon RNA using a RiboMAX large-scale RNA T7 kit (Promega). Replicon RNA (10 μg) was electroporated into Lunet cells to test for replication fitness and sensitivity to PSI-352938, PSI-353661, or control compounds using a 4-day transient-transfection assay. Replication fitness was determined by first normalizing the luciferase expression at 96 h to expression at 4 h and then dividing the normalized level of luciferase expression of the replicon mutant by that of the wild type. Stable cell lines containing mutated replicons were generated by selection in the presence of G418. Stable cells were also tested for sensitivity to PSI-352938 and PSI-353661 as described above.

RESULTS

Selection studies using GT 1a and 1b replicon cells.

The chemical structures of PSI-352938 and PSI-353661 are shown in Fig. 1. Previously we reported that replicons containing the NS5B amino acid change S96T or S282T, which confers resistance to certain nucleoside/-tide analogs, or amino acid changes (C316Y, M414T, M423T, or P495L) that confer resistance to various classes of nonnucleoside inhibitors remained fully susceptible to both PSI-352938 and PSI-353661 (4, 11). In order to identify the mutation(s) that confers resistance to these compounds, selection studies were performed using replicon cells and increasing concentrations of PSI-352938 or PSI-353661.

Fig. 1.

Chemical structures of nucleoside/-tide inhibitors PSI-352938, PSI-353661, PSI-7977, PSI-6130, 2′-C-methyladenosine, IDX-184, and INX-08189.

In the case of PSI-352938, multiple selection studies were performed using both GT 1b (clone A or ET) and 1a (H77) replicon cells (representative experiments are summarized in Table S1 in the supplemental material). We observed that cells did not survive selection if the starting concentration of PSI-352938 was 1 μM (∼10-fold over its EC₅₀) and the concentration was increased by 1 μM each week (GT 1b pool 1). Crisis began on day 17, and complete cell death occurred by day 49, indicating that the replicons were cleared under these conditions. Subsequently we performed selections by starting at concentrations near the EC₅₀ (0.15 μM) and increasing the concentration more gradually. While cells did survive in most cases (GT 1b pools 2 and 3, GT 1a clone 1 and 3), genotypic analysis of the NS5B region showed mainly random changes across multiple cDNA clones from different cell passages. Two mutations, C445F and S556G, did appear at high frequency in pool 3 of the PSI-352938 selected GT 1b cells. These mutations were introduced into the ET replicon and tested for sensitivity to PSI-352938. Results showed that PSI-352938 remained fully active against replicons containing C445F or S556G, either alone or in combination, indicating that these amino acid changes did not confer resistance to PSI-352938. To further evaluate whether mutation(s) outside the NS5B region occurred during the selection process, total RNA from GT 1b pool 3 and GT 1a clone 3 replicon cells were extracted and transfected into the highly permissive Lunet cell line. We examined the total RNA from GT 1b pool 3 and GT 1a clone 3 because a significant increase in EC₅₀s for PSI-352938 was observed with the selected replicon cells. Results showed that cells transfected with PSI-352938-selected GT 1b pool 3 or GT 1a clone 3 total RNA were as sensitive to PSI-352938 as cells transfected with total RNA from the no-drug control cells (EC₅₀ fold change < 2), indicating that no replicon-based resistant variants were selected. Selection and total RNA transfection studies performed using PSI-353661 also indicated that no resistant GT 1a or 1b replicons were selected with PSI-353661.

Selection studies using JFH-1-derived GT 2a replicon cells.

Since the JFH-1-derived replicon has been shown to replicate more efficiently than GT 1a and 1b replicons (6), we performed selection studies with PSI-352938 and PSI-353661 using a JFH-1-derived GT 2a replicon cell line. Compared to GT 1 replicons, GT 2a replicons were more tolerant of PSI-352938 and PSI-353661: overall, the compound-treated cells propagated similarly to the DMSO control cells, with no major crisis occurring during the course of the selection studies. The changes in sensitivity of GT 2a replicon cells to PSI-352938 or PSI-353661 at various passages are summarized in Tables 1 and 2. A >10-fold reduction in activity for PSI-352938 was observed starting on day 85, when cells were treated with PSI-352938 at concentrations ∼25-fold over its EC₅₀ (Table 1). Similarly, cells treated with PSI-353661 at concentrations ∼30-fold over its EC₅₀ showed a >10-fold reduction in activity compared to the no-drug control cells (Table 2).

Table 1.

Selection of GT 2a JFH-1 replicon cells using PSI-352938^a

Days under selection	PSI-352938 concn (μM)	EC50 (μM)		EC50 fold change
		No-drug control	PSI-352938
16	1.0	0.22	0.73	3.3
31	1.4	0.13	0.42	3.2
43	1.6	0.21	0.85	4.0
68	3.0	0.19	0.80	4.2
85	3.8	0.11	1.94	17.6
105	4.2	0.20	2.32	11.6
137	6.5	0.23	3.58	15.6
154	8.0	0.23	3.36	14.6
158	8.0	0.16	3.07	19.2

^aGT 2a JFH-1 replicon cells were cultured in the presence of G418 and increasing concentrations of PSI-352938 or 0.2% DMSO (no-drug control) for 158 days. At various passages, the sensitivities of the no-drug control and PSI-352938-selected cells to PSI-352938 were tested. The EC₅₀ fold shift was determined by normalizing the values from the PSI-352938-selected cells to those from the no-drug control cells.

Table 2.

Selection of GT 2a JFH-1 replicon cells using PSI-353661^a

Days under selection	PSI-353661 concn (μM)	EC50 (μM)		EC50 fold change
		No-drug control	PSI-353661
7	0.015	0.010	0.011	1.1
17	0.020	0.012	0.019	1.6
20	0.020	0.014	0.032	2.3
45	0.040	0.023	0.091	4.0
63	0.060	0.013	0.073	5.6
83	0.070	0.013	0.067	5.2
115	0.30	0.0049	0.055	11.2
132	0.30	0.0083	0.22	26.5

^aGT 2a JFH-1 replicon cells were cultured in the presence of G418 and increasing concentrations of PSI-353661 or 0.2% DMSO (no-drug control) for 132 days. At various passages the sensitivity of the no-drug control and PSI-353661-selected cells were tested against PSI-353661. The EC₅₀ fold shift was determined by normalizing the values from the PSI-353661 selected cells to those from the no-drug control cells.

In order to verify that the increase in EC₅₀s for PSI-352938 and PSI-353661 was mediated by changes within the replicon, total RNA was extracted from the PSI-352938-resistant (day 148) or PSI-353661-resistant (day 127) cells and transfected into Lunet cells. Lunet cells were also transfected with total RNA isolated from the no-drug control cells from the corresponding days of selection. Our results showed that the reduction in sensitivity to PSI-352938 and PSI-353661 was indeed mediated by changes within the GT 2a replicons: a >10-fold reduction in activity was observed for both PSI-352938 and PSI-353661 in cells transfected with RNA isolated from the corresponding compound-selected replicon cells (Fig. 2). In addition, data showed that cross-resistance exists between PSI-352938 and PSI-353661. Cells transfected with total RNA from PSI-352938-selected replicon cells were 12.3-fold less sensitive to PSI-353661, while cells transfected with total RNA isolated from PSI-353661-selected replicon cells were 16.2-fold less sensitive to PSI-352938. In contrast, we did not observe any cross-resistance between PSI-352938/PSI-353661 and other classes of nucleoside/-tide analogs (chemical structures shown in Fig. 1). Both PSI-352938- and PSI-353661-resistant replicon cells remained fully susceptible to PSI-6130 (a 2′-F-2′-C-methylcytidine nucleoside analog of RG7128), PSI-7977 (a prodrug of 2′-F-2′-C-methyluridine-monophosphate), 2′-C-methyladenosine, IDX-184, and INX-08189 (both prodrugs of 2′-C-methylguanosine-monophosphate) (Fig. 2).

Fig. 2.

Total-RNA transfection studies. Lunet cells stably expressing RNA transfected from either the PSI-352938 (white bars) or PSI-353661 (gray bars) selection studies were tested for their sensitivity to nucleoside/-tide inhibitors. Values are averages plus standard deviations from at least three independent experiments performed in duplicate.

Genotypic analysis of resistant replicons.

Since the mechanism of action for nucleoside/-tide inhibitors involves the inhibition of the HCV NS5B polymerase, we sequenced the NS5B region from both PSI-352938- and PSI-353661-selected replicons. Results showed that PSI-352938 and PSI-353661 selected very similar sets of amino acid changes (Table 3). Among these mutations, residue 223 was changed from its wild-type Cys amino acid (codon TGC) to either a Tyr (TAC in most analyzed cDNA clones) or a His (CAC) in all 18 different cDNA clones sequenced. The next most commonly found amino acid change was S15G, followed by V321I and L320I. An R222Q change was observed only in PSI-352938-selected replicons and not in PSI-353661-selected replicons. None of these mutations were found in replicons isolated from the DMSO control cells. Alignment of HCV GT 1 to 6 NS5B sequences (∼700 different isolates) from the Los Alamos Database indicates a high level of conservation for amino acids 222, 223, 320, and 321 (>99% conserved). Residue 15 is relatively more polymorphic, but it appears to be associated with some genotype specificity: it is primarily an Ala in GT 1, 4, and 6, primarily a Ser in GT 3 and 5, and a Ser or Gly (∼1:1 ratio) in GT 2.

Table 3.

Genotypic analysis of the GT 2a NS5B gene from PSI-352938- and PSI-353661-selected replicons^a

Mutation	Codon		Frequency in replicon
	WT	Mutant	PSI-352938 selected	PSI-353661 selected
S15G	AGC	GGC	7/10	6/8
R222Q	CGA	CAA	2/10	0/8
C223Y	TGC	TAC	3/10	6/8
C223H	TGC	CAC	7/10	2/8
L320I	CTA	ATA	3/10	1/8
V321I	GTA	ATA	4/10	4/8

^aTotal RNA was extracted from PSI-352938- or PSI-353661-selected cells and no-drug control cells on day 127. JFH-1 NS5B region was reversed transcribed into cDNA and amplified for clonal sequence analysis. Nucleotide changes are underlined.

We mapped the locations of these amino acids using the GT 2a JFH-1 NS5B crystal structure (25). As shown in Fig. 3, residues R222, C223, L320, and V321 are located within the palm domain close to the active site of the polymerase. In particular, R222 and C223 are located on the same strand as the highly conserved aspartic acid D220, while L320 and V321 are located on the opposite strand just adjacent to the GDD motif (G317, D318, and D319) that is highly conserved among RNA-dependent RNA polymerases. Residue S15 is located on the surface of the finger domain far from the catalytic site: the distance between S15 and the Cα of D318 is >30 Å.

Fig. 3.

Structure of JFH-1 NS5B (PDB: 3I5K). Shown are the positions of key amino acid changes selected by PSI-352938 (orange), the conserved GDD motif and D220 residue (yellow), and amino acid changes selected by other 2′-modified nucleoside/-tide analogs (S282T) or 4′-azidocytidine (S96T) (purple). Red is the palm domain, green is the thumb domain, and blue is the finger domain of the NS5B polymerase.

Additionally, we sequenced other nonstructural regions of the replicons to determine if PSI-352938 or PSI-353661 selected for amino acid changes outside the NS5B region. Analysis of the various cDNA clones showed that a number of changes occurred throughout NS3, NS4A/B, and NS5A: a total of 28 and 18 amino acid changes were detected at various frequencies in PSI-352938- and PSI-353661-selected replicons, respectively. With the exception of one mutation (NS3 Q552L), which was observed in both PSI-352938- and PSI-353661-selected replicons as a low-frequency change (two of eight cDNA clones in both cases), all the other amino acid changes within the PSI-352938- and PSI-353661-selected replicons were different. It is possible that some of these mutations were selected in order to improve the replication efficiency of the resistant replicons. For the present study, we chose to focus on the amino acid changes within the NS5B polymerase, as it is the target of the active 5′-triphosphate metabolite generated from both PSI-352938 and PSI-353661.

Phenotypic analysis of amino acid changes.

In order to examine which of the selected mutations within NS5B confer resistance to PSI-352938 and PSI-353661, GT 2a JFH-1 replicons containing the corresponding amino acid changes were evaluated using a 4-day transient-transfection assay (data summarized in Table 4). Overall, the EC₅₀ fold changes for PSI-352938 and PSI-353661 in the panel of replicon variants tested were comparable. Certain replicons appeared to have smaller EC₅₀ fold changes for PSI-353661 than for PSI-352938; however, the difference was mostly modest (∼2-fold in general) and could be due to variations within the transient-transfection assays.

Table 4.

Sensitivity of GT 2a replicon variants to PSI-352938 and PSI-353661^a

Mutation	EC50 fold change
	PSI-352938	PSI-353661
Single
S15G	1.3 ± 0.5	0.7 ± 0.2
R222Q	1.0 ± 0.3	0.5 ± 0.1
C223Y	2.2 ± 0.5	1.5 ± 0.2
C223H	3.7 ± 1.4	2.2 ± 0.3
L320I	1.5 ± 0.5	2.0 ± 0.9
V321I	2.1 ± 1.0	1.3 ± 0.4
Double
C223H/S15G	6.4 ± 1.8	3.7 ± 1.5
C223H/R222Q	3.3 ± 1.3	2.3 ± 0.4
C223H/L320I	9.2 ± 3.0	4.1 ± 1.3
C223H/V321I	4.6 ± 1.8	4.2 ± 1.3
C223Y/S15G	3.1 ± 0.9	2.5 ± 0.9
C223Y/L320I	2.6 ± 1.5	1.4 ± 0.7
S15G/L320I	1.7 ± 0.5	1.7 ± 0.5
S15G/V321I	1.8 ± 0.3	1.7 ± 0.4
Multiple
S15G/C223H/L320I	6.2 ± 2.7	5.3 ± 0.9
S15G/C223H/V321I	16.5 ± 4.6	12.3 ± 2.5
S15G/R222Q/C223H/V321I	19.8 ± 5.3	9.9 ± 1.9
S15G/C223Y/L320I	5.6 ± 1.5	2.5 ± 0.8
S15G/C223Y/V321I	7.8 ± 2.5	3.5 ± 0.7
S15G/R222Q/C223Yb	5.0 ± 1.3	5.2 ± 0.7
S15G/C223Y/L320I/V321I	3.2 ± 1.8	1.3 ± 0.7

^aThe effect of amino acid changes at positions 15, 222, 223, 320, and 321 in JFH-1 replicons was examined using a 4-day transient-transfection assay. The sensitivities of mutated replicons to PSI-352938 and PSI-353661 were examined and compared to that of the wild type. Values are averages ± standard deviations from at least three independent experiments performed in duplicate.

^bEC₅₀ for S15G/R222Q/C223Y replicon was determined using stable cells due to poor fitness of the mutated replicon.

Studies examining the single amino acid substitutions S15G, R222Q, C223Y, L320I, or V321I showed that these changes had no significant impact on the activity of PSI-352938 or PSI-353661 (Table 4). Replacing the Cys at position 223 with a His caused a modest increase in EC₅₀s, 3.7-fold for PSI-352938 and 2.2-fold for PSI-353661. As these EC₅₀ changes were significantly lower than those observed in the selection and total-RNA transfection studies, replicons with multiple amino acid changes were evaluated. Among variants with two substitutions, those that contained C223H and either S15G, L320I, or V321I had higher EC₅₀ changes, up to 9.2-fold for PSI-352938 and 4.2-fold for PSI-353661. Those with C223Y, however, showed relatively lower EC₅₀ fold shifts compared to the corresponding replicons that contained C223H. Addition of R222Q, the amino acid substitution selected by PSI-352938 but not PSI-353661, to C223H did not result in a further increase in EC₅₀s. No significant effect on the activity of PSI-352938 and PSI-353661 was observed in replicons with double mutations, such as S15G/L320I and S15G/V321I, without a change at position 223.

Evaluation of replicons with multiple amino acid changes showed that replicons containing S15G/C223H/V321I had 16.5- and 12.3-fold increases in EC₅₀s for PSI-352938 and PSI-353661, respectively (Table 4). The corresponding triple mutant containing L320I instead of V321I was less resistant to PSI-352938 and PSI-353661. As was observed in replicon variants with two mutations, replicons with three or more amino acid substitutions that included C223Y had lower EC₅₀ fold shifts than the C223H replicon variants.

Replication fitness of the mutated replicon.

The relative fitness of each of the replicon mutants can be inferred from the expression of luciferase, which is driven by the HCV internal ribosome entry site (IRES) during mRNA translation. GT 2a JFH-1 replicons with a single amino acid change containing the C223Y/H, L320I, or V321I substitution replicated with an efficiency similar to that of the wild type, while replicons with R222Q and S15G replicated at 53% and 20% of the efficiency of the wild-type replicon, respectively (Fig. 4A). The replication capacity of the S15G replicon was improved to at least 80% of that of the wild-type replicon when it was combined with C223Y/H, L320I, or V321I (Fig. 4B). Adding C223H to R222Q also increased the replication fitness of the R222Q replicon from 53% to about 80% (Fig. 4B).

Fig. 4.

Replication fitness analysis of JFH-1 replicons containing single (A), double (B), and multiple (C) amino acid changes. Values are averages plus standard deviations from at least three independent experiments performed in duplicate.

Certain combinations of three or more amino acid changes could mediate a significant reduction in replicon replication fitness (Fig. 4C). Among JFH-1 replicons with multiple mutations, the least fit mutant was that with the mutations S15G/R222Q/C223Y (0.4%), followed by the S15G/C223Y/L320I/V321I (11%) and S15G/C223H/L320I (18%) mutants. However, the triple and quadruple mutants with the highest EC₅₀ fold shifts for PSI-352938 and PSI-353661, the S15G/C223H/V321I and S15G/R222Q/C223H/V321I replicons, replicated at about 80% of the efficiency of the GT 2a wild-type replicon. Replicons with the triple mutations S15G/C223Y/L320I and S15G/C223Y/V321I replicated at 43% and 75% of the efficiency of the wild-type replicon, respectively.

Progression of resistant mutations.

Data from the phenotypic studies indicated that resistance to PSI-352938 and PSI-353661 required multiple amino acid changes. In order to track the progression of these mutations, total RNA was isolated from earlier passages of the PSI-352938 selection study, from which clonal sequencing of the NS5B region was performed (data summarized in Table 5). We examined the NS5B sequence as early as day 31, in which no amino acid changes at positions 15, 222, 223, 320, and 321 were detected (0 of 10 cDNA clones). On day 43, a minority of the cDNA clones contained a Val-to-Ile substitution at position 321 (2 of 15 cDNA clones). Amino acid changes at position 223 were first detected on day 57 (7 of 9 cDNA clones). Among these seven clones, four had a single-nucleotide substitution resulting in the C223Y change (in one clone, this mutation was combined with S15G), while the other three had a two-nucleotide substitution resulting in C223H (in two of these, the mutation was combined with S15G or S15G/V321I). Analysis of replicons from day 61 onward showed that all cDNA clones sequenced contained at least one of the amino acid changes that occurred at positions 15, 222, 223, 320, and 321. Up to day 68, the majority of the cDNA clones still contained single amino acid changes, with C223H being the predominant alteration, but by day 95 all cDNA clones contained either double or triple mutations. By day 127, 60% of the cDNA clones sequenced contained triple mutations, and the rest contained double mutations.

Table 5.

Progression of amino acid changes within GT 2a NS5B

Day	No. of cDNA clones sequenced	Mutation(s) (no.b)
		Single	Double	Triple
31	10
43	15	V321I (2)
57	9	C223Y (3), C223H (1)	S15G/C223Y (1), S15G/C223H (1)	S15G/C223H/V321I (1)
61	14	C223Y (1), C223H (11), V321I (1)	S15G/V321I (1)
68	11	C223Y (2), C223H (8)	S15G/C223Y (1)
95	11		S15G/C223Y (1), S15G/C223H (5), C223H/V321I (1)	S15G/C223Y/V321I (2), S15G/C223H/V321I (2),
127	10		S15G/C223Y (1), C223H/L320I (2), C223H/V321I (1)	S15G/C223H/L320I (1), S15G/R222Q/C223Y (2), S15G/C223H/V321I (3)

Total RNA was extracted from PSI-352938 selected cells from various passages, starting on day 31. JFH-1 NS5B region was reversed transcribed into cDNA and amplified for clonal sequence analysis.

^bNumber of cDNA clones containing the indicated amino acid substitution(s).

These data indicate that GT 2a JFH-1 replicons exposed to PSI-352938 accumulated multiple amino acid changes in a gradual and progressive manner. The predominant amino acid change appeared to be associated with position 223. Throughout the selection period, 90% of cDNA clones with single amino acid changes contained either a C223Y or C223H mutation, 93% of cDNA clones with double amino acid changes contained C223Y/H, and 100% of the cDNA clones sequenced with triple mutations contained C223Y/H.

The S15G/C223H/V321I mutant lacks cross-resistance with other HCV inhibitors.

As we showed in the studies with cells transfected with total RNA (see above), no cross-resistance exists between the 2′-F-2′-C-methylguanosine prodrugs and other classes of nucleoside/-tide analogs (Fig. 2). As future antiviral therapy for HCV would likely involve combinations of compounds with different mechanisms of action, we tested the activity of various classes of HCV inhibitors using cells that stably express the GT 2a S15G/C223H/V321I replicon. These include compounds that inhibit the NS3 protease (telaprevir and RG7227), NS5A (BMS-790052), NS5B (nonnucleoside NS5B inhibitor HCV-796), interferon, and ribavirin. Our results showed that the S15G/C223H/V321I replicon was fully susceptible to inhibition by compounds from other classes of HCV inhibitors (Fig. 5). PSI-352938, which was included as a control in these studies, showed an 11.8-fold reduction in activity against the GT 2a S15G/C223H/V321I replicon.

Fig. 5.

Cross-resistance studies. The sensitivity of a GT 2a S15G/C223H/V321I replicon mutant to other known HCV inhibitors was examined. Compounds include interferon, ribavirin, and those that target the NS3 protease (RG7227 and telaprevir), NS5A (BMS-790052), and the NS5B allosteric region (HCV-796). Values are averages plus standard deviations from at least three independent experiments performed in duplicate.

Effect of resistant mutations in the GT 1b replicon.

As stated above, we were not able to select for GT 1 resistant replicons with PSI-352938 or PSI-353661. Since our phenotypic studies of the GT 2a NS5B mutations showed that the most critical amino acids involved in reducing the activity of PSI-352938 and PSI-353661 were residues 15, 223, and 321, we constructed GT 1b Con1 replicons containing these mutations either alone or in combination to determine if they had an effect on the activity of 2′-F-2′-C-methylguanosine prodrugs. Alignment of the NS5B peptide sequences from GT 1a, 1b, and 2a replicons used in this study showed that with the exception of residue 15, which is an Ala in GT 1a and 1b but a Ser in GT 2a, all three replicons had the same amino acids at positions 223 and 321. Consequently, we generated five GT 1b replicon variants with different mutations: C223Y, C223H, V321I, C223H/V321I, and A15G/C223H/V321I. In addition to the effects of these mutations on the activities of PSI-352938 and PSI-353661, their effect on GT 1b replicon RNA replication was also evaluated (Table 6).

Table 6.

Effects of PSI-352938- and PSI-353661-selected mutations in GT 1b replicons^a

Mutation(s)	EC50 fold change		Replication fitness (% relative to WT)
	PSI-352938	PSI-353661
Single
C223H	ND	ND	0.9 ± 0.49
C223Y	ND	ND	2.2 ± 0.65
V321I	1.3 ± 0.2	1.2 ± 0.2	126.6 ± 8.2
Multiple
C223H/V321I	2.3 ± 0.2	2.6 ± 0.3	21.9 ± 5.5
A15G/C223H/V321I	4.8 ± 1.3	4.8 ± 0.8	10.1 ± 1.1

^aEffect of amino acid changes at positions 15, 223, and 321 was evaluated using the GT 1b Con1 replicon and a 4-day transient assay. Both replication fitness and sensitivity of the mutated replicon for PSI-352938 and PSI-353661 were examined. Values are averages ± standard deviations from at least three independent experiments performed in duplicate. ND, EC₅₀ could not be determined because the levels of firefly luciferase were too low.

Results from the fitness experiments showed that GT 1b Con1 replicons containing the C223Y or C223H amino acid substitution had replication profiles drastically different from those of GT 2a JFH-1-derived replicons. In particular, the C223Y/H substitution within the GT 1b replicon made it incompetent for replication: changing residue 223 to Tyr or His reduced the replication capacity to only 2.2% and 0.9% of the wild-type capacity, respectively. In contrast, a GT 1b replicon containing the V321I substitution replicated with an efficiency similar to that of its corresponding wild type. As was observed with the GT 2a V321I replicon, the V321I change in GT 1b replicons did not significantly affect the activity of PSI-352938 or PSI-353661 (<2-fold change in EC₅₀s). Interestingly, addition of V321I to the C223H replicon resulted in some compensatory effect in that the replication efficiency for this combination was 22% of the wild-type efficiency. This combination slightly increased the EC₅₀s for PSI-352938 and PSI-353661, by 2.3- and 2.6-fold, respectively. Addition of the third mutation to create the GT 1b Con1 replicon triple mutant, A15G/C223H/V321I, further increased the EC₅₀ changes for both PSI-352938 and PSI-353661, to about 5-fold. In contrast to the GT 2a replicon containing the triple mutation, which was quite fit for replication, the corresponding GT 1b triple mutant exhibited a replication efficiency of only about 10% relative to the wild type.

DISCUSSION

Using the JFH-1-derived GT 2a subgenomic replicon system, we demonstrated that PSI-352938 and PSI-353661 selected for a unique set of amino acid changes that have not been associated with resistance to other nucleoside/-tide analogs. Indeed, our cross-resistance data showed that the PSI-352938- and PSI-353661-resistant replicons remained fully susceptible to inhibition by 2′-F-2′-C-methylpyrimidine (PSI-6130 and PSI-7977) and 2′-C-methylpurine analogs (IDX-184, INX-08189, and 2′-C-methyladenosine). In addition, other classes of HCV inhibitors, including interferon, ribavirin, and those that target the NS3 protease, NS5A, and NS5B allosteric sites were similarly active against both wild-type and S15G/C223H/V321I GT 2a replicons.

Compared to previously reported nucleoside resistance studies, which selected for single resistant mutations (1, 13, 18), resistance associated with PSI-352938 and PSI-353661 is mediated by multiple amino acid changes, in particular, the combination of S15G, C223H, and V321I. Among these, the C223H substitution appears to be the most critical one: it was the predominant mutation, and replicon variants without C223H had lower EC₅₀ fold changes. C223H is located in close proximity to the highly conserved D220 residue, V321I is located on the opposite side of C223H just adjacent to the GDD motif, and S15G is located on the surface of the finger domain. It is not unprecedented for an amino acid substitution distant from the active site to affect the activity of a nucleoside analog: the 4′-azidocytidine nucleoside R1479 selects the S96T mutation (13), which is also located on the surface of the finger domain. Based on a crystallographic complex of GT 1b HCV NS5B polymerase and a short single-stranded RNA (21), residues 15 and 96 are near the template binding groove. It is possible that mutating S15 could affect the interaction of the template with the NS5B polymerase. Also, it is reasonable to speculate that changing residues within the active site (C223H and V321I) could affect the binding of the incoming 2′-F-2′-C-methylguanosine 5′-triphosphate. However, amino acids 223 and 321 are directed away from the metal ion binding site, suggesting that they may not be in direct contact with the incoming nucleotide. In light of this, it is possible that these mutations may function to modulate the conformation of the NS5B active site, subsequently impacting the formation of the replicase complex. One interesting observation from our transient-transfection assays was that the EC₅₀ fold changes of certain replicons with multiple mutations were time dependent (∼4-fold higher in the 4-day assay than in the 3-day assay), whereas this was not observed in the stable mutant cell lines. The kinetics of replicase complex formation is likely different between the transient and stable cells. How this observation relates to the molecular mechanism of the roles of these mutations and how the mutations affect the enzymatic activity of the NS5B polymerase and replicase complex formation will need to be investigated.

Phenotyping assays showed consistent EC₅₀ fold changes for PSI-352938: 16.5-fold in the S15G/C223H/V321I replicons, 12.6-fold in the PSI-352938 total-RNA transfection studies, and 11.6- to 19.2-fold in the selected replicon cells evaluated from days 85 to 158. However, the PSI-353661 EC₅₀ fold changes in the replicon variants with the site-directed mutations S15G/C223H/V321I and S15G/R222Q/C223H/V321I (9.9- and 12.3-fold, respectively) were approximately 2-fold lower than that obtained from the PSI-353661-selected replicon cells on day 132 (26.5-fold) but were more similar to results of the PSI-353661 total-RNA transfection studies (14.7-fold). These results suggest that the difference in EC₅₀ fold change between the site-directed replicon variants and the PSI-353661-selected cells on day 132 could be due to the selection of a population of cells that metabolize PSI-353661 less efficiently. Since this was observed only with PSI-353661 and not PSI-352938, this suggests that the effect was likely one that affected either the uptake or the metabolism of the phosphoramidate prodrug. Differences in sensitivities of nucleoside analogs in selected cells and site-directed replicon variants were observed previously, the most striking example being the nucleosides PSI-6130 and RG7128, in which the EC₅₀ fold change in the selected cells could be as high as >100-fold, but the S282T replicon displayed only an ∼3-fold shift (1). In our GT 1a and 1b selection studies, we did observe that certain populations of replicon cells became resistant to PSI-352938. However, when the total RNA was isolated from these cells and transfected into Lunet cells, no resistance to PSI-352938 was observed, suggesting that some cell-based resistance arises after prolonged culturing of the replicon cells with PSI-352938.

Data from the fitness experiments showed that the replication capacity of GT 1b replicons containing C223H was in striking contrast to that of the corresponding GT 2a replicon variants. In particular, GT 1b replicon variants containing either the C223Y or the C223H substitution were incapable of replication. Assuming that the pathway to resistance would be similar for both GT 1 and GT 2, i.e., starting predominantly with a change at position 223 and progressing to additional changes, it is not surprising that we were unsuccessful in selecting GT 1b replicon cell lines containing either the C223Y or the C223H amino acid change. It has been shown previously that JFH-1 replicons replicate more efficiently than GT 1-derived replicons (6) and that the JFH-1 HCV is the most infectious laboratory strain so far without any cell culture adaptive mutation (30). The high infection rate has been attributed to, at least in part, the NS5B polymerase (19). It is possible that the JFH-1 NS5B can tolerate amino acid substitutions even in highly conserved residues due to extra hydrophobic interactions between the thumb and finger domains, as shown by crystallographic studies (23, 25). On the other hand, other nonstructural proteins and the 3′ untranslated region could also play a role in regulating the fitness of the replicons. We swapped the GT 2a NS5B into the GT 1b replicon and the GT 1b wild-type and A15G/C223H/V321I NS5B into the JFH-1 replicon to see if other replicon determinants could affect replication fitness. However, in our hands all GT 1b/2a and GT 2a/1b chimeric replicons failed to replicate, indicating that RNA synthesis facilitated by NS5B was highly dependent on intergenotypic interaction among structural proteins and/or the nontranslated regions. Work is ongoing to examine the effect of the amino acid changes S15G, C223H, and V321I on the replication capacity of another GT 2a strain as well as other genotypes.

Despite this observation, we proceeded to study the impact of multiple mutations within GT 1b HCV. Addition of V321I or A15G/V321I had a compensatory effect on the fitness of GT 1b replicons with the C223H change alone: C223H/V321I and A15G/C223H/V321I mutants replicated at 22% and 10% of the rate of the wild-type replicon, respectively. Furthermore, susceptibility assays showed that the reduction in sensitivity to PSI-352938 and PSI-353661 was still associated with a combination of mutations, indicating that the multiple amino acid changes required to confer resistance to these compounds were not just specific to GT 2a JFH-1. It is interesting that the EC₅₀ fold shift in the GT 1b triple mutant was lower than that in the corresponding GT 2a replicon variant, which suggests that the levels of resistance exerted in mutated NS5B polymerases from other genotypes could be somewhat different.

In conclusion, we have identified a novel resistance profile that is specific to 2′-F-2′-C-methylguanosine nucleotide prodrugs. The activity of other nucleoside/nucleotide analogs such as 2′-F-2′-C-methylcytidine (PSI-6130), 2′-F-2′-C-methyluridine (PSI-7977), 2′-C-methylguanosine (IDX-184 and INX-08189), and 2′-C-methyladenosine analogs were not affected by these mutations. Recent efforts to develop DAA strategies have focused on combinations of antiviral regimens without the SOC in order to reduce side effects, enhance efficacy, and suppress the emergence of resistant viruses. The unique resistance profile of PSI-352938 and PSI-353661 makes them desirable compounds for potential combination therapy with other nucleoside/-tide analogs as well as other classes of HCV inhibitors.