Mutations Identified in the Hepatitis C Virus (HCV) Polymerase of Patients with Chronic HCV Treated with Ribavirin Cause Resistance and Affect Viral Replication Fidelity

Ribavirin has been used for 25 years to treat patients with chronic hepatitis C virus (HCV) infection; however, its antiviral mechanism of action remains unclear. Here, we studied virus evolution in a subset of samples from a randomized 24-week trial of ribavirin monotherapy versus placebo in chronic HCV patients, as well as the viral resistance mechanisms of the observed ribavirin-associated mutations in cell culture. Thus, we performed next-generation sequencing of the full-length coding sequences of HCV recovered from patients at weeks 0, 12, 20, 32 and 40 and analyzed novel single nucleotide polymorphisms (SNPs), diversity, and mutation-linkage.

ABSTRACT

Ribavirin has been used for 25 years to treat patients with chronic hepatitis C virus (HCV) infection; however, its antiviral mechanism of action remains unclear. Here, we studied virus evolution in a subset of samples from a randomized 24-week trial of ribavirin monotherapy versus placebo in chronic HCV patients, as well as the viral resistance mechanisms of the observed ribavirin-associated mutations in cell culture. Thus, we performed next-generation sequencing of the full-length coding sequences of HCV recovered from patients at weeks 0, 12, 20, 32 and 40 and analyzed novel single nucleotide polymorphisms (SNPs), diversity, and mutation-linkage. At week 20, increased genetic diversity was observed in 5 ribavirin-treated compared to 4 placebo-treated HCV patients due to new synonymous SNPs, particularly G-to-A and C-to-U ribavirin-associated transitions. Moreover, emergence of 14 nonsynonymous SNPs in HCV nonstructural 5B (NS5B) occurred in treated patients, but not in placebo controls. Most substitutions located close to the NS5B polymerase nucleotide entry site. Linkage analysis showed that putative resistance mutations were found in the majority of genomes in ribavirin-treated patients. Identified NS5B mutations from genotype 3a patients were further introduced into the genotype 3a cell-culture-adapted DBN strain for studies in Huh7.5 cells. Specific NS5B substitutions, including DBN-D148N+I363V, DBN-A150V+I363V, and DBN-T227S+S183P, conferred resistance to ribavirin in long-term cell culture treatment, possibly by reducing the HCV polymerase error rate. In conclusion, prolonged exposure of HCV to ribavirin in chronic hepatitis C patients induces NS5B resistance mutations leading to increased polymerase fidelity, which could be one mechanism for ribavirin resistance.

INTRODUCTION

Over the past 25 years the broad-spectrum antiviral nucleoside analogue ribavirin (1, 2) has been a key component for the treatment of patients with chronic hepatitis C virus (HCV) infection (3). Despite the recent advent of highly efficacious direct-acting antivirals (DAAs) targeting HCV, its global burden remains high. At least 70 million people worldwide are estimated to be chronically infected with HCV, resulting in about 400,000 deaths annually (4, 5). The high genetic heterogeneity of HCV represents a challenge for antiviral therapy, since different susceptibilities to treatment have been shown among genotypes, with genotype 3 having a more resistant phenotype (6, 7).

Although a limited effect of ribavirin monotherapy on plasma HCV RNA levels was observed, ribavirin improves treatment outcome in combination with other antiviral drugs, including DAAs (3). A recent randomized controlled trial found that patients with advanced chronic genotype 3 HCV infection and baseline resistance-associated substitutions (RASs) in nonstructural protein 5A (NS5A) were more likely to achieve sustained virological response (SVR) if ribavirin was added to the pangenotypic sofosbuvir and velpatasvir DAA regimen (8). Another recent study also suggested that inclusion of ribavirin improves the chances of SVR in treatment-experienced chronic HCV patients (9). Furthermore, ribavirin is recommended for the treatment of chronic hepatitis E virus (HEV) infection and was recently shown to be highly efficacious in treating HEV infection after organ transplantation (10).

HCV is an enveloped, positive-sense single-stranded RNA virus with a single open reading frame (ORF) that encodes three structural and seven nonstructural (NS) proteins (5). Amplification of the full-length HCV genome by reverse transcription and PCR is possible (11, 12) but requires optimization for analysis of patient samples. Several studies analyzing either NS5A/NS5B or the entire ORF of genotype 1 HCV-infected patients identified a single putative resistance mutation to ribavirin in NS5B encoding the F415Y substitution (13–18). This change, however, did not lead to treatment failure in patients treated with interferon and ribavirin or to ribavirin resistance in cell culture assays (17, 19). A study of patients receiving interferon and ribavirin combination therapy found a positive correlation between NS5B mutations and outcome but did not identify specific resistance mutations (20). Recently, two putative ribavirin resistance mutations were identified in patients infected by HCV genotype 3a who had failed regimens that included sofosbuvir and ribavirin. However, both mutations had limited effect on ribavirin susceptibility when tested in a cell culture assay (21).

Ribavirin’s antiviral mode of action is not fully understood. Different mechanisms, such as depletion of nucleotide pools, stimulation of Th1 immune cells, and increased mutagenesis, have been suggested (22, 23). A mutagenic effect of ribavirin monotherapy on HCV genotype 1 was detected in most studies (13–18, 24), but it has not been investigated in patients infected with other HCV genotypes.

In recent years several HCV cell-culture-adapted strains, representing the full-length sequence of different genotypes and subtypes, have been developed (5, 25). These systems make it possible to study HCV drug susceptibility and resistance in the context of the entire viral life cycle.

We investigated viral evolution during treatment on a unique set of serial samples obtained from a randomized trial of ribavirin monotherapy versus placebo for patients with chronic HCV infection (26). We studied the mutagenic effects of ribavirin and HCV adaptation to ribavirin treatment using next-generation sequencing (NGS) on the entire HCV ORF of isolates recovered from patients. Finally, we investigated the effect of identified mutations on ribavirin treatment using cell-culture-based assays.

RESULTS

HCV adaptation to ribavirin monotherapy.

Serial samples were retrieved from a subset of 9 chronic HCV patients, 5 of whom received ribavirin (2 at 1,000 mg per day and 3 at 1,200 mg per day) and 4 of whom received placebo for 24 weeks. The alanine aminotransferase (ALT) levels at baseline were comparable in ribavirin- and placebo-treated patients (Table 1). Plasma HCV RNA titers were determined at baseline (week 0), during treatment (weeks 12 and 20), and after treatment (weeks 32 and 40) (Fig. 1a and b). Comparison of the HCV RNA levels in the treated and placebo groups showed no statistical differences between the two (P = 0.74; unpaired t test) (Fig. 1c).

TABLE 1.

Novel SNPs of HCV acquired during and after ribavirin or placebo treatment in patients with chronic hepatitis C

	Genotype	Baseline ALT (U/ml)	Ribavirin dose (mg/day)	No. of novel SNPsa
				Wk 12	Wk 20	Wk 32	Wk 40
Ribavirin patients
1	2b	108	1,200	721	898	844	774
2	3a	114	1,000	638	723	455	242
3	3a	95	1,000	585	617	507	419
4	3a	110	1,200	959	1087	968	812
5	3a	112	1,200	551	642	540	363
Mean (SEM)				690 (72)	793 (88)	663 (102)	522 (114)
Placebo patients
6	2b	169		15	12	12	14
7	1b	94		55	78	78	74
8	3a	308		303	249	163	214
9	1a	52		49	38	125	145
Mean (SEM)				105 (66)	94 (53)	95 (33)	112 (43)
P				<0.001	<0.001	0.002	0.019

^aCompared to baseline sequences (week 0).

FIG 1.

Characterization of HCV in hepatitis C patients treated with ribavirin or placebo. Five patients treated with ribavirin (orange-red) and four patients that received placebo (blue) for 24 weeks had plasma HCV RNA titrated and were NGS sequenced at weeks 0, 12, 20, 32, and 40. (a and b) Viral RNA titers of patient samples, determined by real-time quantitative PCR. Values are expressed in IU/ml on a logarithmic scale. (c) Average RNA titer values for ribavirin- and placebo-treated groups at five time points, expressed as means and standard deviations. (d) Phylogenetic tree of 45 full-length HCV ORF sequences, based on consensus sequences aligned to the indicated genotype reference sequences. Bayesian estimates of posterior probability (%) are indicated for relevant branches. Genetic distance as measured by nucleotide changes per site is indicated by a horizontal scale bar. pt, patient.

To investigate how ribavirin affected HCV evolution, we extracted viral genomic RNA from patient samples and analyzed it using NGS. First, we characterized the phylogeny of recovered ORF sequences (Fig. 1d). All sequences obtained from the same patient clustered together, excluding the possibility of cross-contamination during sample processing (6). Five patients were infected with HCV genotype 3a, 2 patients were infected with genotype 2b, and 2 patients were infected with genotypes 1a and 1b, respectively.

We next evaluated the virus genetic variation during treatment by calculating the increase in sequence diversity in the last on-treatment samples (week 20) compared to baseline (week 0) (Fig. 2a). Synonymous diversity was higher throughout the complete HCV ORF in patients treated with ribavirin compared to patients receiving placebo. Nonsynonymous diversity of HCV was also increased in samples from ribavirin-treated patients, particularly in the NS5B coding region. During amplification of HCV RNA from genotype 2b patients, we consistently detected shorter genomes (replicons) containing in-frame deletions spanning either E1-p7 or E1-NS2 (see Fig. S1 in the supplemental material). Nested PCR of the cDNA confirmed the presence of shorter replicon species, indicating that the genomes with deletions represented more than 50% of the viral population (see the Fig. S1 legend for details).

FIG 2.

Next-generation sequencing of the complete HCV ORF from viruses obtained from ribavirin- and placebo-treated patients. (a) Diversity analyses of synonymous (green) and nonsynonymous (red) novel SNPs at week 20 in ribavirin (patients 1 to 5)- and placebo (patients 6 to 9)-treated patients, calculated using a 50-amino-acid sliding window. The separation between structural and nonstructural proteins is indicated by a blue dotted line; the beginning of the NS5B region is indicated by a black dotted line. (b and c) Distribution of specific SNPs in HCV of ribavirin (b)- and placebo (c)-treated patients. HCV RNA in plasma at illustrated time points was sequenced by NGS, and new SNPs compared to the baseline sequence were calculated and presented with the 95% confidence interval (CI). A, adenine; C, cytosine; G, guanine; U, uracil.

Taken together, these results indicate a higher genetic variation in treated patients compared to controls and show positive selection in the NS5B region of HCV treated with ribavirin, suggesting treatment-driven virus adaptation.

Increased mutation rates in HCV patients treated with ribavirin.

To further investigate the genetic variation of HCV during treatment, we determined the number and type of novel single nucleotide polymorphisms (SNPs) in on-treatment (weeks 12 and 20) and posttreatment samples (weeks 32 and 40) compared to their baseline sequence (week 0) (Table 1; Fig. 2b and c; see also Fig. S2 to S10 in the supplemental material). The number of novel SNPs in ribavirin-treated samples was significantly higher than in placebo controls both during and after treatment (Table 1). Although a few new nonsynonymous and synonymous mutations increased in frequency after end of treatment (see Fig. S2 to S6), the overall number of new mutations decreased. Since both weeks 12 and 20 had similar accumulation of mutations (Fig. 2b and c), we conclude that most mutations were acquired between week 0 and 12. The increase was driven by a rise in ribavirin-associated G-to-A and C-to-U transitions, indicating a direct mutagenic effect of ribavirin on HCV replication (Fig. 2b and c).

The NS5B region accumulates putative resistance mutations.

Nonsynonymous mutations were particularly concentrated in the NS5B region (Fig. 2a; see Fig. S2 to S6). Overall, we identified 14 high-frequency novel SNPs (≥15% of NGS reads in at least one sample) in the HCV NS5B polymerase of ribavirin-treated patients (Table 2), whereas none were detected in the polymerase of placebo controls. Thirteen substitutions were identified during treatment (weeks 12 and 20), and 10 of these persisted after ribavirin discontinuation. The remaining three NS5B substitutions—S15N, D148N, and T227A—became less prevalent in posttreatment samples. Finally, one NS5B substitution, N535T, was detected only after treatment.

TABLE 2.

Frequency of nonsynonymous mutations of the HCV NS5B polymerase in chronic hepatitis C patients receiving ribavirin

Patient (genotype)	Frequency (%) of nonsynonymous mutation(s)a
	Wk	S15N	V147M, I147V*	D148N	A150V, A150T*	Y162F	S183P	T227A	T227S	T340A	T357I, E357K*	I363V	T390N, T390I*†	S393T, A393V/T*	N535T, K535R*
1 (2b)	12	–	0.3*†	4.4*	2.0*†	2.7	–	–	–	–	0.2*	0.4*	2.7*	3.6	16.8*†
	20	–	0.2*†	2.6*	2.2*†	6.2	–	–	–	–	–	0.2	4.3*	44.6	10.9*†
	32	–	0.3*†	0.8*	1.5*†	13.2	–	–	–	–	–	–	2.5*	68.3	3.0*†
	40	–	0.4*†	–	1.4*†	24.2	–	–	–	–	–	–	0.8*	60.2	1.3*†
2 (3a)	12	0.2	6.4	4.2	1.1	–	0.2	0.8	–	0.4	0.6	1.7	2.3	–	–
	20	0.2	13.6	12.6	2.7	1.4*†	20.7	25.6	1.9	0.2	1.6	5.7	6.9	0.2*	–
	32	–	1.5	2.4	0.2	–	80.1	10.7	82.6	–	–	0.3	0.7	–	–
	40	–	2.6	0.2	1.8	–	14.8	0.6	15.0	–	–	–	0.5	–	–
3 (3a)	12	6.6	23.5	7.4	4.8†	0.2*	–	16.7	1.0	–	11.2	21.9	6.1	1.1*	–
	20	3.8	78.9	2.2	1.9†	0.1*	–	6.7	0.4	–	12.2	62.2	8.7	2.0*	–
	32	0.9	95.8	0.3	1.0†	–	–	0.2	–	–	40.6	49.0	9.8	2.5*	–
	40	0.8	98.6	–	0.6†	–	–	–	–	–	71.4	14.5	5.7	1.3*	–
4 (3a)	12	4.4	–	39.6	3.1	0.2*	0.3	20.3	–	0.9	1.6	12.0	1.5	–	–
	20	3.1	–	37.8	4.3	0.3*	0.3	43.3	1.2	1.0	12.1	12.7	1.9	0.9*	–
	32	0.7	–	59.7	14.1	1.1*	–	1.5	–	39.8	23.9	15.0	17.7	0.5*	10.8
	40	–	–	8.0	85.8	–	0.2	–	–	4.6	2.6	76.0	2.3	0.2*	88.1
5 (3a)	12	3.4	4.2	2.5	–	–	–	20.0	1.0	0.3	4.0	34.2	5.9†	–	–
	20	55.9	9.8	0.4	–	–	–	27.2	0.8	–	1.5	10.8	59.2†	–	–
	32	38.8	11.7	–	–	–	–	6.9	–	–	0.8	52.8	40.7†	–	–
	40	4.7	1.2	–	–	–	–	0.3	–	–	–	91.6	12.2†	–	–

^a*, mutations identified in the aligned positions in another genotype strain; †, mutations present at baseline; –, no mutations detected.

HCV NS5B substitutions T227A and I363V were detected in all four ribavirin-treated genotype 3a patients, whereas V147M, D148N, and T390N/T390I were observed in three of these patients. None of the NS5B mutations identified in ribavirin-treated patients emerged in placebo controls, even at lower cutoff values (≥5% of reads). These results suggest adaptation of the HCV NS5B polymerase during ribavirin treatment, possibly in the form of resistance development.

Linkage analysis of NS5B putative resistance mutations.

Next, we assessed the linkage between mutations observed in the NS5B region of samples from week 20 of ribavirin-treated patients by further analysis of NGS reads and by molecular cloning (Fig. 3a). Several substitution associations were observed across patient samples: T227A/S was detected in combination with S183P, V147M and D148N were found in combination with either T357I or I363V, and S15N was always observed in combination with T390I. Overall, the proportion of putative resistant haplotypes accounted for 60 to 100% of the total population in the four genotype 3a patients and almost 50% in the 2b patient (Fig. 3a).

FIG 3.

Frequently changed amino acids in HCV NS5B in ribavirin-treated patients. (a) Proportion (%) of high-frequency NS5B haplotypes (≥15% of NGS reads) identified in ribavirin-treated patients #1 to #5 at week 20. (b) Structure of the aligned HCV genotype 1a polymerase (PDB 2XI2) (27), highlighting the location of high-frequency NS5B mutations with labels and red color. The polymerase domains are shown in different colors for reference: blue for fingers, yellow for palm, green for thumb, and gray for the beta-hairpin region. The nucleotide entrance site is indicated by an arrow. Mutations were identified by NGS, and the frequencies and developments of specific mutations are shown in Table 2.

To clarify the role of these substitutions in NS5B adaptation to ribavirin treatment, we mapped their location onto the crystal structure of the HCV polymerase (27, 28). Most mutations were located at the proximity of the nucleotide entry site distributed in the three structural domains of NS5B: thumb (green in Fig. 3b), fingers (blue in Fig. 3b), and palm (yellow in Fig. 3b). Taken together, these data suggest that different mutations within the NS5B region emerge during ribavirin treatment and could be associated with a drug-resistant phenotype.

Viability in cell culture of genotype 3a variants carrying putative ribavirin resistance mutations.

To characterize the effect of the observed putative resistance substitutions from patients treated with ribavirin, we cloned several mutations alone and in combination into DBN3acc (referred to also as DBN), an HCV cell-culture-adapted full-length construct of genotype 3a based on the DBN HCV isolate obtained from a patient with chronic HCV (29, 30), for a total of 20 variants (see Table S1 in the supplemental material). The viability and genetic stability of the mutant viruses were assessed by monitoring viral spread and the production of infectious particles during transfection and infection in Huh7.5 cells and by analyzing recovered viral HCV genomes with NGS (see Fig. S11 and S12 and Table S1).

Overall, most mutants were viable in cell culture, but some variants acquired additional mutations (Fig. 4a). We selected the following subset of mutant viruses, which had maintained the introduced mutations at >95% at passage 2, for further analysis: DBN-V147M, DBN-A150V, DBN-S183P, DBN-I363V, DBN-D148N+I363V, DBN-A150V+I363V, and DBN-T227S+S183P. All introduced mutations and combinations thereof had been observed among the four genotype 3a patients treated with ribavirin.

FIG 4.

Differential suppression of DBN3acc mutants with ribavirin 50 μM. (a) Graphical representation of consensus mutations (>50%) identified in viral constructs. The top diagram represents the HCV genomic structure and indicates the locations of both engineered and newly detected mutations. Colored lines represent different mutant genomes, solid triangles indicate inserted mutations, and empty triangles represent acquired mutations. (b) Infection spread and infectivity titers of each DBN mutant with or without ribavirin 50 μM (the data represent the results from one of two independent experiments). The number of infected Huh7.5 cells was scored by intracellular flow cytometry using antibodies against the HCV NS5A protein. Infectivity represent the log of FFU/ml, with error bars indicating the standard deviations. The break in the right y axis represents the assay sensitivity threshold. Colors represent different DBN mutants. Solid lines and bars indicate untreated samples; dotted lines and hatched bars represent ribavirin-treated samples. (c) Viral suppression achieved during treatment, calculated as the percentage of infected cells in treated cultures relative to untreated cultures, at each time point. Asterisks indicate statistical difference from the DBN3acc control (two-way ANOVA; *, P < 0.01; **, P < 0.0001).

NS5B substitutions in DBN3acc have limited effect on ribavirin and sofosbuvir IC₅₀.

The susceptibility to ribavirin, measured as inhibitory concentration 50% (IC₅₀), was estimated for each mutant virus in a short-term dose-response assay in Huh7.5 cells (Table 3). All introduced mutations had limited effect on short-term HCV ribavirin susceptibility, resulting in 0.8- to 1.6-fold changes of IC₅₀ compared to the parental DBN3acc virus. The mutants showing the largest change were DBN-V147M and the combinations DBN-A150V+I363V and DBN-T227S+S183P.

TABLE 3.

Treatment susceptibility in Huh7.5 cells of DBN3acc mutants carrying putative NS5B resistance mutations

Construct	Average IC50 (CI) or fold difference (CI)a
	Ribavirin		Sofosbuvir
	IC50 (μM)	Fold difference	IC50 (nM)	Fold difference
DBN	37.1 (±2.7)	1.0 (±0.0)	767 (±59)	1.0 (±0.0)
DBN-V147M	58.2 (±4.1)	1.6 (±0.2)	875 (±81)	1.1 (±0.1)
DBN-A150V	47.4 (±3.6)	1.3 (±0.1)	710 (±64)	0.9 (±0.1)
DBN-S183P	29.4 (±3.5)	0.8 (±0.1)	1,679 (±322)	2.2 (±0.5)
DBN-I363V	53.0 (±4.2)	1.4 (±0.2)	1,144 (±102)	1.5 (±0.2)
DBN-D148N+I363V	48.2 (±6.4)	1.3 (±0.2)	524 (±63)	0.7 (±0.1)
DBN-A150V+I363V	59.9 (±5.9)	1.6 (±0.2)	419 (±45)	0.5 (±0.1)
DBN-T227S+S183P	60.3 (±6.6)	1.6 (±0.2)	1,071 (±153)	1.4 (±0.2)

^aThe weighted average of two to three independent experiments is presented. The fold difference values were calculated relative to the parental DBN3acc. CI, 95% confidence interval.

Recent data suggested that specific substitutions within NS5B can simultaneously affect the response of HCV to ribavirin and sofosbuvir treatment (21, 31, 32). We thus performed dose-response assays to sofosbuvir on all viruses tested against ribavirin (Table 3). Similarly, the weighed estimates of the sofosbuvir IC₅₀ fold change showed very small differences among the tested viruses, with the most resistant mutant (DBN-S183P) displaying only a 2.2-fold-higher IC₅₀ compared to DBN3acc.

Overall, these results indicate that the tested putative resistance mutations have limited effect on viral susceptibility to ribavirin and sofosbuvir in short-term treatment assays.

NS5B substitutions reduce the effect of ribavirin long-term treatment in cell culture.

To better understand the effect of the selected mutations on HCV viral spread and genetic evolution over time, we performed long-term ribavirin treatment of DBN3acc and the mutant viruses in Huh7.5 cells (Fig. 4b). The three mutants DBN-D148N+I363V, DBN-A150V+I363V, and DBN-T227S+S183P showed statistically significant decreased viral suppression under ribavirin treatment, compared to DBN3acc (P < 0.0001; two-way analysis of variance [ANOVA]) (Fig. 4b and c). The latter mutant displayed particularly low susceptibility to ribavirin treatment, reaching 60% of infected cells and with a limited decrease in infectivity titers. The remaining mutants showed a reduced but detectable effect on ribavirin susceptibility. NGS analyses performed on virus supernatant from day 10 showed small differences between treated and untreated viruses, and no changes in the sequence consensus (see Table S2). These results indicate that mutations observed in patients caused reduced susceptibility to ribavirin during long-term treatment compared to the original virus.

Resistant HCV variants accumulate fewer ribavirin-associated mutations during treatment.

We and others have previously shown that ribavirin treatment in cell culture increases ribavirin-associated mutations (C to U and G to A) in newly produced viral genomes (33–36). We analyzed the effect of ribavirin on our ribavirin putative resistant mutants by performing NGS on viral supernatants obtained at day 10 of the long-term treatments (Fig. 4b). All treated samples showed greater accumulation of ribavirin-associated mutations compared to untreated controls; however, the total number of new mutations varied among different mutants (Fig. 5a and b). Mutants carrying substitutions V147M, A150V, and S183P showed a <2-fold difference in the number of new mutations compared to DBN3acc. Conversely, mutants DBN-I363V, DBN-D148N+I363V, DBN-A150V+I363V, and DBN-T227S+S183P displayed a >3-fold reduction in the number of accumulated substitutions, with the latter showing a 10-fold difference, relative to the original DBN3acc (Fig. 5a and b). These data suggest that reduced susceptibility to ribavirin in these mutants is linked to a reduction in the accumulation of ribavirin-associated mutations.

FIG 5.

Accumulation of mutations in DBN3acc mutants. New mutations were identified by comparing the initial viral inoculum to NGS of day 10 samples of untreated control cultures (a) and cultures treated with ribavirin 50 μM (cutoff value of 0.5%) (b). (c) Supernatants of untreated mutant viruses were collected and deep sequenced before and after the experiment shown in Fig. 4b. The sequences were compared to the original plasmid constructs, and the number of new mutations was plotted against total time in cell culture. (d) Data points from panel c were used to calculate the number of expected mutations at day 34 by linear regression analysis. The mutation numbers were normalized to the DBN3acc parental control and are shown as the fold change in total accumulated substitutions. (e) Reduction of infection spread upon treatment with sofosbuvir 2 μM cell cultures fully infected with DBN mutant viruses. Histograms represent the percentages of infected cells scored at different time points after treatment initiation. HCV positivity was measured by intracellular flow cytometry using antibodies against HCV NS5A protein. The colors represent different DBN mutants.

Ribavirin-resistant HCV mutants exhibit higher intrinsic polymerase fidelity.

To better elucidate the possible mechanism of resistance to ribavirin observed in our mutant variants, we analyzed the rate of accumulation of mutations in the nontreated control viruses during long-term experiments. Mutant viruses displayed comparable infection spread during passage 2 (stock preparation) and passage 3 (long-term experiment) (Fig. 4b; see Fig. S11 in the supplemental material). We thus analyzed NGS sequences from passages 2 and 3 and compared them to their original plasmid reference sequences, relative to cumulative time in cell culture (Fig. 5c). The data provide a rough estimate of the intrinsic replication fidelity of each variant and show that some mutants display markedly slower accumulation of mutations compared to DBN3acc. To better compare the different constructs, we then estimated the number of accumulated mutations at day 34 posttransfection for all viruses, normalized to the DBN3a parental virus (Fig. 5d). Mutants DBN-I363V, DBN-D148N+I363V, DBN-A150V+I363V, and DBN-T227S+S183P had accumulated fewer mutations, suggesting that their polymerases could exhibit higher intrinsic fidelity. Interestingly, these mutants also showed reduced susceptibility to ribavirin, hinting at polymerase fidelity as one possible mechanism for ribavirin resistance in these viruses.

Mutations T227S+S183P conferred increased susceptibility to sofosbuvir.

Finally, we evaluated the susceptibility of the mutants to sofosbuvir in long-term treatments by measuring the clearance rate of fully infected cell cultures (Fig. 5e). Viruses that are more susceptible to sofosbuvir should be cleared more rapidly than more resistant ones. The data showed very similar susceptibility to sofosbuvir across all mutants. Interestingly, however, mutant DBN-T227S+S183P was cleared significantly faster than the others (P = 0.02; two-way ANOVA), showing higher susceptibility to sofosbuvir. The effect of the combined mutations was thus opposite of what had been observed under ribavirin treatment. These results support the notion that different molecular mechanisms modulate HCV susceptibility to ribavirin and sofosbuvir in cell culture.

DISCUSSION

This study analyzed HCV in blood samples from a subset of patients participating in a clinical trial of 24 weeks of ribavirin monotherapy versus placebo for chronic HCV infection (26). Our study showed that ribavirin treatment caused a significant viral mutagenic effect in genotype 2b and 3a infected patients, particularly an increase in G-to-A and C-to-U mutations, that persisted after treatment cessation (Fig. 2b and c). In addition, we detected accumulation of synonymous SNP across the entire ORF in ribavirin-treated patients, strongly supporting mutagenesis as a main mechanism of action of ribavirin on HCV in patients (Fig. 2a). A mutagenic effect of ribavirin on HCV genotype 1 in patients had been previously described (13–18). The same effect was also identified previously in culture by analyzing viruses recovered from a genotype 2a recombinant (33, 36–38) and recently in recombinants of genotypes 1a and 3a (34).

The study has the obvious limitation of analyzing a limited number of historical samples. In addition, four of five ribavirin-treated patients had genotype 3a HCV compared to one of four placebo-treated patients. There may be intergenotypic differences in fidelity of HCV polymerases. However, a decline of total new SNP from week 20 (on treatment) to week 32 (post treatment) in the ribavirin-treated patients, but not in the placebo-treated patients, was detected (Table 1; Fig. 2b and c). The observed increase of SNPs in ribavirin- compared to placebo-treated patients in this study is therefore most likely not determined by genotype differences in the ribavirin-treated and placebo-treated patients. In addition, the in-depth analysis allowed by NGS technology, as well as confirmation of the sequencing findings by in vitro experiments, lend confidence to our results. In particular, the mutations identified in the highly represented genotype 3a samples showed an effect in a genotype 3a cell culture system (Fig. 4b and c).

Despite repeated efforts to identify ribavirin resistance mutations, only a few candidates have been described. By analyzing novel SNPs in samples from patients treated with ribavirin alone, we were able to identify mutations in the HCV NS5B polymerase that were most frequently observed at treatment week 20 or later, suggesting that a long exposure time to ribavirin is necessary for viral adaptation, and possibly resistance, to occur (Fig. 3a). Many nonsynonymous substitutions emerged within the NS5B region and were located at the periphery of the nucleotide entry site in the predicted NS5B structure (Fig. 3b). These findings suggest that such mutations could modulate ribavirin access to the polymerase. Particularly, mutations at position 227 were detected in all genotype 3a ribavirin-treated patients and promptly reverted to wild-type sequences after treatment discontinuation, strongly suggesting a role of these substitutions in ribavirin adaptation. Interestingly, position 227 is located near the border between palm and fingers domains of the HCV NS5B polymerase and relatively close to Asp225, which was found to be important for nucleotide recognition (28).

Substitutions at positions 15, 147, 150, and 162 in NS5B of HCV were previously detected in patients treated with a nucleoside polymerase inhibitor (RG7128) combined with interferon and ribavirin (39), but their effect on ribavirin susceptibility was not evaluated. A recent study identified substitution A150V in genotype 3a patients failing sofosbuvir-based regimens and confirmed that it conferred resistance to both sofosbuvir and ribavirin in cell culture. Interestingly, substitutions at NS5B positions 147, 148, and 150 were genetically linked to those at positions 357 or 363, suggesting a cooperative or compensatory effect between them. Analogously, positions 183 and 227 are located in the same general area within the structure of NS5B and were observed in combination in treated genotype 3a patients (Fig. 3). The patient sequence data, coupled with the in vitro results of DBN mutants showing ribavirin resistance and high genetic stability of 227S when combined with 183P, strongly suggest a cooperative and/or compensatory effect between these mutations in the context of a genotype 3a polymerase. Although most of the analyzed mutations are located far apart within the polymerase structure, their effect on the overall conformation of NS5B could be far reaching and result in functional cooperation of different substitutions. None of the remaining mutations in NS5B have previously been identified in patients treated with ribavirin or DAA (13–18, 40, 41).

Mutations conferring ribavirin resistance were shown to affect the overall replication fidelity of the virus, thus possibly preventing incorporation of incorrect nucleotides and reducing the effect of mutagenic nucleotide analogues such as ribavirin (Fig. 5c and d). Although these results are based on a single long-term experiment, to our knowledge this is the first in vitro estimation of the fidelity of HCV using full-length cell culture systems of genotype 3a. Our findings will need confirmation using more thorough analyses and complementary approaches, but they lend support to the involvement of polymerase fidelity as a mechanism of ribavirin resistance. Nonetheless, DBN-A150V showed little effect on ribavirin susceptibility, despite displaying fidelity comparable to some of the resistant mutants, whereas DBN-V147M showed susceptibility similar to that of A150V but no increase in fidelity, suggesting that fidelity is only one of several molecular mechanisms involved in ribavirin resistance. In addition, despite the weak mutagenic effect of ribavirin on DBN-T227S+S183P, this virus was suppressed by ribavirin, pointing at other mechanisms of action of ribavirin besides mutagenesis (36, 42–44).

Finally, the result obtained by long-term treatment with sofosbuvir showed that the mutant virus DBN-T227S+S183P was slightly more sensitive to this nucleotide analog despite being less prone to mutation accumulation compared to the parental DBN3acc, suggesting a different molecular mode of action for sofosbuvir compared to ribavirin (Fig. 5e). It has been recently shown that a genotype 3a virus with the sofosbuvir resistance mutation S282T exhibited hypersusceptibility to ribavirin (31, 32). These data support the idea that multiple complementary molecular mechanisms regulate HCV susceptibility to nucleotide analogues.

In conclusion, this study found a mutagenic effect of ribavirin in the complete ORF of HCV genotype 2b and 3a in vivo. We further showed HCV NS5B adaptation to ribavirin treatment and documented highly prevalent NS5B substitutions causing resistance to ribavirin. Finally, we confirmed the ribavirin resistance potential of mutations identified in these patients in an infectious cell culture system. We found that ribavirin-associated mutations increase the overall fidelity of the viral polymerase as a putative mechanism for ribavirin resistance. If confirmed, the selection of viral genomes with increased polymerase fidelity during ribavirin treatment could increase the barrier to resistance of other antivirals, and thus ribavirin could represent a relevant addition to current regimens to hinder the development of resistance.

MATERIALS AND METHODS

Patient sample collection and analysis.

A subset of nine samples from a randomized double-blind and placebo-controlled study of the efficacy of ribavirin monotherapy were retrieved based on samples and clinical data availability (26). Nine participants with chronic HCV infection (five ribavirin treated and four placebo treated) were included at Huddinge Hospital, Stockholm, Sweden, in 1992 and 1993. Patients had been randomized to receive either weight-based ribavirin or placebo for 24 weeks. Plasma samples obtained at baseline (week 0) and at weeks 12, 20, 32, and 40 were stored at −70°C. HCV RNA titers of patient samples were determined as previously described (45). ALT was measured in μkat/liter and converted to U/liter with Conversion-Factor 59.9 (26).

HCV RNA isolation, amplification, sequencing, and cloning from patient samples.

Patient samples were processed as previously described (46). Specifically, 100 μl of plasma was suspended in 300 μl of TRIzol LS (Thermo Fisher Scientific), and RNA was extracted using an RNeasy MinElute kit (Qiagen). Reverse transcription of full-length ORF HCV RNA was performed with Maxima H Minus reverse transcriptase (Thermo Scientific) with RNasin Plus RNase inhibitor (Promega) added at 50°C for 120 min, followed by 5 min at 85°C using genotype-specific primers described below. cDNA was treated with RNase H (20 min at 37°C) and amplified (35 cycles at 98°C for 10 s, 65°C for 10 s, and 72°C for 8 m) using Hot Start High-Fidelity Q5 DNA polymerase (New England Biolabs). The applied primers were as follows: for genotype 1a, RT-PCR (TAAGAGGCCGGAGTGTTTAC), PCR forward (TGCCTGATAGGGTGCTTGCG), and PCR reverse (AGGCCGGAGTGTTTACCCCA); for genotype 1b, RT-PCR (CCTATTGGCCTGGAGTGTTTAG), PCR forward (TGCCTGATAGGGTGCTTGCG), and PCR reverse (TGGCCTGGAGTGTTTAGCTCC); for genotype 2b, RT-PCR (CTATGGAGTGTAGCTAGGGTTTGC), PCR forward (TGCCCCGGGAGGTCTCGTAGACC), PCR reverse (GTAGCTAGGGTTTGCCGCTC), and PCR reverse junction (CCAGGTCCCGCAAACCCT); and for genotype 3a, RT-PCR (AAAAGAATGGAGTGTTATC), PCR forward (GATAGGGTGCTTGCGAGTGCC), and PCR reverse (AGAATGGAGTGTTATCCTACCAGCTCA). The purity and lengths of PCR products were confirmed on GelRed (Biotium)-stained 1% agarose gels.

Four separate PCR products of each sample were pooled, cleaned (DNA Clean & Concentrator; Zymo Research), run on a 1% agarose gel stained with SYBR Safe (Invitrogen) for 60 min, cut by blue light imaging, and recovered (Zymoclean large fragment DNA recovery kit; Zymo Research). The concentrations of DNA were quantified (Qubit Fluorometer; Thermo Fisher Scientific), and 50 ng of gel-extracted PCR product was fragmentized (Fragmentase kit M0348A; New England Biolabs) with 1 μl of MgCl₂ added. Libraries were prepared with the NEBNext Ultra II DNA kit, size selected (500 bases) using Ampure XP beads (Beckman Coulter), and multiplexed with index primers (New England Biolabs). The sizes and concentrations of the library preparations were quantified (2100 Bioanalyzer; Agilent Technologies), and quantitative PCR (Library Quant kit for Illumina; New England Biolabs, Roche LightCycler 96 system) and diluted library preparations (4 nM) were pooled for NGS using the Illumina MiSeq platform applying the v3 600 cycle kit (Macrogen).

To verify mutations linkage in the NS5B region of HCV, 4 μl of ORF amplicons was cloned with a TOPO XL-2 kit (Thermo Fisher Scientific); 18 to 20 clones of each sample were sequenced and analyzed using Sequencher 5.1.

SNP detection and computation of HCV recovered from patients.

NGS data were analyzed as previously described (47) except that baseline samples were de novo assembled using IVA software (48) to get a full-ORF consensus. NGS generated between 12,000 and 16,000 coverages/nucleotides for all samples. Mapping of reads of 2b samples were based on a J8 reference sequence (49) because of the presence of subgenomic bands in both samples that prevented a full ORF assembly by IVA. Briefly, reads were mapped to consensus of the baseline sample by BWA MEM and processed by SAMtools. SNPs were called by Lo-Freq and translated by SnpEff. Subsequently, new SNPs with a 0.5% cutoff were detected by VCFTools using complementary and noncomplementary intersections with the input virus leading to new SNP compared to the baseline sample. Population diversity analysis was performed by SNPGenie (50) on novel SNP calls at a cutoff at 0.5% to calculate π (mean number of pairwise differences per site), together with πN and πS in a 50-amino-acid sliding window across the ORF. Consensus sequences were aligned together with reference sequences of represented genotypes using Mafft (51). A phylogenetic tree was computed by maximum likelihood using PhyML (52) and visualized with FigTree v1.4.3. Identified frequent NS5B mutations were introduced into the aligned HCV genotype 1a polymerase structure (PDB 2XI2 [27, 28]) using the PyMOL molecular graphics system, v2.0 (Schrödinger, LLC).

Short-range (<600 bases) linkage analysis of NS5B mutations was performed using LinkGe on NGS alignment files. Long-range linkage (>600 bases) was determined by analysis of cloned sequences.

Construction of genotype mutant viruses.

The genotype 3a full-length cell-culture-adapted DBN3acc recombinant (29) was used as the template for the introduction of putative resistance mutations in the viral polymerase using the QuikChange lightning kit (Agilent), according to the manufacturer’s recommendations. Plasmids were sequence confirmed by Sanger sequencing (Macrogen). For viral production, plasmids were linearized with the restriction enzyme XbaI (New England Biolabs), in vitro transcribed using T7 RNA polymerase (Thermo Fisher), and transfected into naive Huh7.5 cells using Lipofectamine 2000 (Thermo Scientific). Huh7.5 cells were cultured as previously described (45, 53). Every 2 to 3 days, the cells were split, the supernatants were collected, and viral spread was evaluated by immunostaining with a mouse anti-NS5A antibody (9E10) (54). Viral supernatants collected at full infections spread were used to infect naive cells at an MOI of 0.02 (passage 1), and cells were maintained as detailed above until achieving full viral spread. An additional passage (passage 2) was then performed. Supernatants collected at peak spread of infection in both passages were sequenced by NGS (see the section above for details) to confirm genetic stability of the introduced mutations. In addition, the three supernatants collected closer to the infection peak time points were tested for viral infectivity as previously described (34); plates were scanned and focus-forming units (FFU) automatically counted (45). Constructs which retained the engineered mutations at >95% were used for subsequent experiments.

Ribavirin and sofosbuvir dose-response assays against HCV variants with putative resistance mutations.

Both assays were performed as described previously (29, 34). Huh7.5 cells were plated at 6,000 cells/well and infected for 24 h. Drugs were applied 24 h postinoculation in triplicates, at concentrations ranging from 5 to 400 μM for ribavirin and at 0.26 to 47,000 nM for sofosbuvir. After 48 h, cells were fixed and stained using anti-Core C7-50 (Abcam) and anti-NS5A antibody 9E10 (54) as primary antibodies, horseradish peroxidase conjugated with secondary antibody (GE Life Science), and colored using DAB substrate (Dako). Positive cells were automatically counted as previously described (55, 56); the threshold was set using noninfected controls and normalized to nontreated infected controls.

Long-term culture treatment assays of HCV with putative resistance mutations.

Long-term ribavirin-treatment of culture infections were performed as previously described (34). Briefly, 4 × 10⁴ Huh7.5 cells were infected at an MOI of 0.0005 for 4 h, split at 24 h postinfection, and subsequently grown with or without 50 μM ribavirin. Long-term sofosbuvir-treated infections were performed by infecting 4 × 10⁴ Huh7.5 cells at an MOI of 0.05, beginning treatment with 2 μM sofosbuvir after near-full infection spread. Supernatants were collected every 2 to 3 days upon culture split; viral spread was evaluated by intracellular flow cytometry, performed by staining with anti-NS5A 9E10 antibody (54) and Alexa Fluor 488 (Thermo Fisher). Cells were counted on an LSR Fortessa (Becton Dickinson), and data analysis was performed using FlowJo v10. In addition, infectivity titers were assayed on supernatants collected from ribavirin treatment as previously described (34). Plates were scanned, and the FFU were automatically counted (45).

Statistics.

Statistical analyses were performed using GraphPad Prism 8. Viral suppression in long-term treatment experiments was calculated as the 1’s complement of the ratio between the percentages of infected cells in treated and untreated samples:

$$\left(1-\frac{\%{\text{HCV}}_{\text{treat}}^{\text{+}}}{\%{\text{HCV}}_{\text{untreat}}^{\text{+}}}\right)\times 100.$$

The total numbers of novel SNPs were compared using an unpaired t test. Dose-response curves were obtained by nonlinear regression and used to calculate log₁₀ half-maximal inhibitory concentration (IC₅₀). Results from replicate experiments were used to calculate inverse variance weighted mean log₁₀ IC₅₀ values with 95% confidence intervals (CI) and the fold differences, as previously described (19). P values were determined using a Z test.

Data availability.

Consensus HCV CDS sequences obtained from all patient samples are available through GenBank under accession numbers MT995311 to MT995355.