Infrequent Development of Resistance in Genotype 1–6 Hepatitis C Virus–Infected Subjects Treated With Sofosbuvir in Phase 2 and 3 Clinical Trials

Abstract

Data demonstrate a uniform susceptibility of subject-derived hepatitis C virus (HCV) to sofosbuvir, and also show that selection of sofosbuvir-resistant HCV is exceedingly rare and is associated with a significant reduction in viral fitness.

Background. Sofosbuvir is a chain-terminating nucleotide analogue inhibitor of the hepatitis C virus (HCV) NS5B RNA polymerase that is efficacious in subjects with HCV genotype 1–6 infection. Sofosbuvir resistance is primarily conferred by the S282T substitution in NS5B.

Methods. NS5B sequencing and susceptibility testing of HCV from subjects infected with genotypes 1–6 who participated in phase 2 and 3 sofosbuvir clinical trials was performed.

Results. No NS5B variants present at baseline among 1645 sofosbuvir-treated subjects were associated with treatment failure; sofosbuvir susceptibility was within 2-fold of reference. Among 282 subjects who did not achieve sustained virologic response, no novel sofosbuvir resistance–associated variants were identified, and the NS5B changes observed did not confer significant reductions in sofosbuvir susceptibility. In 1 subject with S282T observed at relapse 4 weeks after sofosbuvir monotherapy, the resistant variant (13.5-fold reduced sofosbuvir susceptibility, replication capacity <2% of control) became undetectable by deep sequencing 12 weeks after treatment. L159F and V321A were identified as treatment-emergent variants but did not confer resistance to sofosbuvir in the replicon system.

Conclusions. These data demonstrate a uniform susceptibility of subject-derived HCV to sofosbuvir, and also show that selection of sofosbuvir-resistant HCV is exceedingly rare and is associated with a significant reduction in viral fitness.

(See the Major Article by Howe et al on pages 1657–65, and Editorial Commentary by Tong and Kwong on pages 1675–7.)

It is estimated that >170 million individuals are infected with hepatitis C virus (HCV) globally [1]. At least 6 genetically distinct genotypes of HCV have been described, with varying prevalence in specific areas of the world [2, 3]. HCV genotype is an important component of the information needed to guide treatment strategies [4, 5].

Four direct-acting antivirals (DAAs) have been approved for treatment of HCV infection: 3 NS3/4A protease inhibitors (PIs)—telaprevir [6, 7], boceprevir [8, 9], and simeprevir [10, 11]—and 1 NS5B polymerase inhibitor (sofosbuvir) [12–14]. Many other DAAs are in various stages of clinical development [15, 16]. Despite the promise of new DAA-based regimens, the potential for selection of drug-resistant HCV remains a significant concern. For example, virologic breakthrough following treatment with a PI plus pegylated interferon and ribavirin is associated with selection of PI-resistant virus in the majority of treated subjects failing therapy [17]. The propensity for HCV to develop antiviral resistance is influenced by many factors including drug susceptibility, pharmacokinetics, the genetic barrier, and the fitness of the resistant variant(s) [18, 19]. However, the clinical relevance of resistant variants of HCV remains uncertain until additional experience with approved DAAs is gathered [20, 21].

Sofosbuvir has been extensively studied in clinical trials [13, 14, 22, 23] and was recently approved by the US Food and Drug Administration. Resistance to sofosbuvir is conferred by the S282T substitution in NS5B [24]; S282T was first described as the major resistance-associated variant for other nucleoside inhibitors (NIs) [25–28]. In addition, the combination of S96T and N142T has been observed following in vitro selection with the NI R1479 [29], M289I/L/V were selected in vitro by various NIs [24, 30], and a combination of L159F and L320F was observed in 1 subject who had a partial response during treatment with mericitabine [31].

In this report, we present comprehensive resistance analyses performed as part of the sofosbuvir phase 2 and 3 clinical development program, involving sequence-based and phenotypic methods in subjects infected with HCV genotypes 1–6.

METHODS

Sofosbuvir Clinical Trials

We analyzed plasma samples from subjects enrolled in 5 phase 2 and 4 phase 3 clinical trials of sofosbuvir-containing regimens. Details salient to the resistance analysis are summarized in Table 1. Information regarding each clinical trial can be found in the references listed in Table 1 or at www.clinicaltrials.gov. Before enrollment and before any study procedures were undertaken, written informed consent was obtained from all subjects. The study was conducted in accordance with the principles of the Declaration of Helsinki and Good Clinical Practice.

Table 1.

Sofosbuvir Clinical Trials

Trial Number (Name)	Phase	Subject Population	HCV Genotype	Regimen	Duration	Reference	ClinicalTrials.gov Identifier
P2938–0721 (QUANTUM)	2	Naive	1, 2, 3	SOF/RBV	12 or 24 wk	[32]	NCT01435044
P7977-0221	2	Naive	1	SOF/RBV/PEG	4 wk	[33]	NCT01054729
P7977-0422 (PROTON)	2	Naive	1, 2, 3	SOF/RBV/PEG	12 wk	[22]	NCT01188772
P7977-0523 (ELECTRON, treatment arms 1–9)	2	Naive	2, 3	SOF/RBV	12 wk	[23]	NCT01260350
				SOF/RBV/PEG	12, 4/8 or 8/4 wk
				SOF	12 wk
				SOF/RBV/PEG	8 wk
		Experienced	2, 3	SOF/RBV	12 wk
		null responders	1	SOF/RBV	12 wk
		Naive	1	SOF/RBV	12 wk
P7977-0724 (ATOMIC)	2	Naive	1, 4, 6	SOF/RBV/PEG	12 or 24 wk	[12]	NCT0132997
GS-US-334-0107 (POSITRON)	3	IFN intolerant	2, 3	SOF/RBV	12 wk	[13]	NCT01542788
GS-US-334-0108 (FUSION)	3	IFN failure	2, 3	SOF/RBV	12 or 16 wk	[13]	NCT01604850
GS-US-334-0110 (NEUTRINO)	3	Naive	1, 4, 5, 6	SOF/RBV/PEG	12 wk	[14]	NCT01641640
P7977-1231 (FISSION)	3	Naive	2, 3	SOF/RBV	12 wk	[14]	NCT01497366

Abbreviations: HCV, hepatitis C virus; IFN, interferon; PEG, pegylated interferon; RBV, ribavirin; SOF, sofosbuvir.

Viral load was measured using the Roche TaqMan version 2 assay. The lower limit of quantitation was 25 IU/mL.

Resistance testing was performed on specimens from subjects who met the criteria for inclusion in the resistance analysis population (RAP). Subjects who received at least 1 dose of a sofosbuvir-containing regimen, but did not achieve sustained virologic response (SVR) due to virologic failure or early discontinuation, had HCV RNA ≥1000 IU/mL, and had a plasma sample available for analysis were included in the RAP.

Resistance Analysis

Population Sequencing

Population sequencing of the full-length HCV NS5B coding region was performed by DDL Diagnostic Laboratory (Rijswijk, the Netherlands), or Monogram Biosciences (South San Francisco, California) using reverse transcription polymerase chain reaction (PCR) and standard Sanger sequencing of the bulk PCR product. The sensitivity for detection of resistant variants is approximately 10%–20% [34]. Variants are reported as differences compared with a genotype-specific reference strain: GT1b Con1 (AJ238799); GT1a H77 (GenBank accession number NC_004102); GT2 JFH-1 (AB047639); GT3 S52 (GU814263); GT4 ED43 (GU814265); GT5 SA13 (AF064490); and GT6 EUHK2 (Y1208).

Deep Sequencing

For subjects experiencing virologic failure, NS5B PCR amplicons at baseline and postbaseline timepoints, generated by DDL Diagnostic Laboratory, were subjected to deep sequencing primarily using the Illumina MiSeq deep sequencing platform (Illumina, San Diego, California) at WuXi AppTec (Shanghai, China). Internally developed software was used to process and align sequencing data via a multistep method to identify the substitutions present at levels >1%. In addition, “consensus” sequences were generated with inclusion of mixtures of amino acids when present between 15% and 85%.

Phenotypic Analysis

Phenotypic analyses of select samples were performed by Gilead Sciences, Inc [35], Janssen Diagnostic (Beerse, Belgium), or Monogram Biosciences. These assays all use chimeric Con1 replicon vectors with luciferase readout into which the subject specimen-derived HCV NS5B coding sequence is inserted. RNA is transcribed from the vector in vitro and transfected by electroporation into “cured” Huh-7 cell lines. Luciferase activity measured 3 days posttransfection is used to derive inhibitory drug concentrations associated with 50%, 90%, or 95% inhibition (EC₅₀, EC₉₀, or EC₉₅, respectively). In general, intra- and interassay precision is approximately 2- to 3-fold [36]. Luciferase activity in the absence of inhibitor, expressed as a percentage of the reference replicon, is a measurement of replication capacity.

For all trials, phenotypic analysis was attempted if a previously identified NI resistance-associated variants (RAVs) was detected by sequence analysis. In addition, for all phase 2 studies and the FISSION (P7977-1231) and NEUTRINO (GS-US-334-0110) phase 3 trials, phenotypic analysis was attempted for all subjects not achieving SVR. The panel of subject samples subjected to phenotypic analysis included at least 1 sample containing each amino acid substitution observed in >2 subjects with virologic failure by population/consensus sequencing.

Data Analysis and Statistics

Known NI RAVs were defined as S96T, N142T, L159F, S282T, M289I/L/V, and L320F [24, 26, 27, 29, 30, 37]. The potential impact of specific NI RAVs as well as any other change compared with reference detected at baseline on treatment outcome (SVR₁₂) was evaluated for each individual study and for all trials together, for each genotype/subtype. For each position in NS5B, each subject was categorized as matching or differing from the reference sequence. Mixtures of mutant and wild-type at the population level were treated as mutant. The association of variants with treatment outcome was tested using Fisher exact test. The Hochberg procedure [38, 39] was used to control for false discovery due to multiple comparisons.

Sofosbuvir susceptibility data were analyzed using Prism 6.0 (GraphPad, La Jolla, California). The distributions of baseline or posttreatment EC₉₀ vs EC₉₅ results from different assays, when expressed as the fold change (FC) vs reference, were not significantly different from each other (unpaired t test of log-transformed FC values P > .5; data not shown), allowing for them to be combined to increase the power of comparison analyses. This parameter is referred to throughout as FCEC_90/95.

RESULTS

Baseline Sequence and Association With Treatment Outcome

NS5B population sequencing was attempted at baseline for all 1662 subjects (834 subjects with genotype 1, 291 with genotype 2, 486 with genotype 3, 39 with genotype 4, 1 with genotype 5, and 11 with genotype 6) who participated in the phase 2 and 3 clinical trials of sofosbuvir, and was successful for 1645 of them. No subjects were identified with the N96T, S282T, or L320F RAVs; 2 were found with N142T, 11 with L159F, and 25 with M289I or L (Table 2). No subject had >1 of these substitutions. It was noted that a variant (C316N) at a position where a known nonnucleoside inhibitor RAV occurs [40–43] was frequently associated with the L159F variant: 9 of 11 subjects with L159F also had C316N. Of the 38 subjects with an NI RAV, 35 (92%) achieved SVR, which is comparable to the overall SVR rate across all studies and the SVR rate among subjects without known RAVs. Two subjects with L159F (1 associated with C316N), and 1 M289I at baseline, did not achieve SVR; however, statistical analysis showed no significant association between the presence of any RAVs with treatment outcome (SVR₁₂).

Table 2.

NS5B Resistance-Associated Variants Detected at Baseline

Trial	Treatment	No. Enrolled (SOF-Treated)a	No. Sequenced	N142T (GT)b	L159F (GT)	M289I/L (GT)	Any NI RAVc	No. With Any RAV and SVR
P2938-0721 (QUANTUM)	SOF/RBV	50	48	…	1 (1b)	…	1	0
P7977-0221	SOF/RBV/PEG	49	49	…	1 (1b)	…	1	1
P7977-0422 (PROTON)	SOF/RBV/PEG	120	120	…	…	…	…	…
P7977-0523 (ELECTRON)	SOF	10	10	…	…	…	…	…
P7977-0523 (ELECTRON)	SOF/RBV	70	70	…	…	1 (1b)	1	1
P7977-0523 (ELECTRON)	SOF/RBV/PEG	40	40	1 (3a)	…	1 (1b)	2	2
P7977-0724 (ATOMIC)	SOF/RBV/PEG	332	326	…	6 (1b)	…	6	6c
GS-US-334-0107 (POSITRON)	SOF/RBV	207	207	…	…	9 (2-2a, 7-2b)	9	9
GS-US-334-0108 (FUSION)	SOF/RBV	201	201	…	…	9 (2b)	9	8
GS-US-334-0110 (NEUTRINO)	SOF/RBV/PEG	327	324	…	3 (1b)	…	3	2c
P7977-1231 (FISSION)	SOF/RBV	256	251	1	…	5 (1–2a, 4-2b)	6	6
Total		1662	1645	2	11	25	38	35

Abbreviations: GT, genotype; NI, nucleoside inhibitor; PEG, pegylated interferon; RAV, resistance- associated variants; RBV, ribavirin; SOF, sofosbuvir; SVR, sustained virologic response.

^a Among the 1662 subjects: 834 subjects with genotype 1, 291 subjects with genotype 2, 486 subjects with genotype 3, 39 subjects with genotype 4, 1 subject with genotype 5, and 11 subjects with genotype 6.

^b The genotype of the virus in subjects with the indicated RAVs is shown in parentheses.

^c No S282T was detected at baseline by population or deep sequencing. All RAVs and SVR outcome were assessed in subjects who completed treatment.

Of the 591 codons in NS5B, 138, 150, 179, and 141 codons were found to be polymorphic (different amino acid compared to subtype-specific reference) at baseline in at least 1% of genotype 1a, 1b, 2a, 2b, or 3a sequences, respectively. The most polymorphic positions (variation observed in >20% of subjects) for each genotype are shown in Supplementary Data along with the SVR rates for subjects with or without these variants. Statistical analysis showed no significant association between the presence of polymorphic substitutions at any of these positions, including those seen in <20% of subjects, with treatment outcome (SVR₁₂).

Phenotypic Analysis of Baseline Samples

Baseline sofosbuvir susceptibility data were generated for 183 subjects. The distribution in FC in susceptibility vs reference using either the EC₅₀ (FCEC₅₀) or EC_90/95 (FCEC_90/95) measurement within a population of untreated subjects provides an estimation of the extent of natural variability in susceptibility and is shown in Figure 1. The mean was 0.92-fold (range, 0.21- to 1.8-fold) for FCEC₅₀ and 0.92-fold (range, 0.32- to 2.0-fold) for FCEC_90/95, and the 99th percentile of the distribution was 1.7-fold for FCEC₅₀, and 1.9-fold for FCEC_90/95. The 99th percentile can be considered as a preliminary biological cutoff to define “normal” sofosbuvir susceptibility. Thus, despite large genotypic variability across HCV genotypes and subtypes, there was a relatively narrow distribution of sofosbuvir susceptibility values across this population of HCV-infected subjects.

Figure 1.

Sofosbuvir susceptibility at baseline (BL) and posttreatment (PT). Susceptibility is presented as fold change (FC) in half maximal effective concentration (FCEC₅₀) vs reference (Ref), fold-change in 90% or 95% effective concentration (FCEC_90/95) vs reference (Ref), fold-change in EC₅₀ PT vs BL, and FCEC_90/95 PT vs BL. Horizontal bars represent the mean for each group. The data for the subject with the S282T-containing virus is not included.

Sequencing Results in the Resistance Analysis Population

Among all sofosbuvir-treated subjects in the phase 2 and 3 studies, a total of 302 of 1662 subjects qualified to be part of the RAP due to failure to achieve SVR. All but 1 of these involved virologic relapse following cessation of therapy, and the 1 subject with virologic breakthrough had documented nonadherence based on drug level assessment. Posttreatment population/consensus NS5B sequences were obtained for 300 of 302 RAP subjects. Deep sequencing (mutant detection sensitivity 1%) was available for 294 of 300 subjects, with >1000 times coverage at position 282 in 272 of 294 subjects.

The S282T substitution was detected in 1 subject infected with genotype 2b HCV who received sofosbuvir monotherapy in a phase 2 study (P7977-0523 ELECTRON; see below) [23]; no other S282T substitutions were detected in posttreatment sequences from the remaining RAP subjects by deep sequencing. Substitutions at other positions were observed, but were not associated with reduced susceptibility to sofosbuvir as described below. Comprehensive analysis of all substitutions identified 2 substitutions, L159F and V321A, that emerged to levels >10% of viral variants at relapse. L159F and V321A emerged in 6 (0.4% of all subjects, or 3.2% of the 189 genotype 3–infected subjects) and 5 (0.3% of all, or 2.6% of genotype 3) subjects, respectively, who were treated with sofosbuvir/ribavirin; neither substitution was observed in patients infected with genotypes other than 3 or treated with sofosbuvir/ribavirin/pegylated interferon. L159 and V321 were 100% conserved in baseline sequences from subjects with genotype 3a infection. However, neither L159F nor V321A were associated with significant reduced susceptibility to sofosbuvir, as described below.

Phenotypic Analysis of RAP Samples

To determine whether any substitutions that developed in the RAP subjects at virologic failure conferred reduced susceptibility to sofosbuvir, the NS5B coding region derived from postbaseline samples was analyzed in a transient replicon phenotyping assay. Postbaseline phenotypic analysis was attempted for 205 subjects and was successful for 176 subjects. Three hundred and forty-five different amino acid substitutions at 202 positions were observed in postbaseline sequences for these subjects, only 1 of which is a recognized RAV (S282T in subject from the ELECTRON study). Ninety-four of the 345 substitutions were observed in >2 subjects either as full substitutions or in mixtures with the wild-type amino acid (Table 3). Excluding the subject with S282T, the distribution of FCEC₅₀ and FCEC_90/95 vs reference or each subject's baseline sample is shown in Figure 1. The mean was 0.92-fold (range, 0.13- to 2.3-fold) for FCEC₅₀ and 0.94-fold (range, 0.13-fold to 2.7-fold) for FCEC_90/95. No statistically significant differences between posttreatment and baseline susceptibility distributions were found (unpaired t test, P > .7). Four samples had FCEC₅₀ greater than the 1.7-fold biological cutoff determined from the baseline analysis (ranging from 1.8- to 2.3-fold), and 2 had FCEC_90/95 >1.9-fold (2.0-fold and 2.7-fold). The susceptibility and sequence data for these 6 outliers are shown in Supplementary Data. In addition, 23 of the 94 substitutions (along with the L159F + C316N double mutant) were analyzed by site-directed mutagenesis and also demonstrated no change in sofosbuvir susceptibility (FCEC₅₀ <1.5; Table 4). These data indicate that, apart from the 1 subject with S282T (described below), no significant reduction in sofosbuvir susceptibility was observed following virologic failure in subjects taking sofosbuvir-containing regimens, with mean, median, and distribution of FCEC₅₀ or FCEC_90/95 vs reference or baseline being similar for all samples regardless of genotype (Figure 1).

Table 3.

Sofosbuvir Susceptibility Among Subject-Derived NS5B Bearing Substitutions Occurring in >2 Subjects

No. of Subjects in Whom Substitution Is Present Posttreatment (No. of Variants)	Amino Acid Variants Observed	FCEC50 vs Ref, Mean (Range)	FCEC50 vs BL, Mean (Range)
>10 (n = 7)	11I, 116I, 116V, 150V, 150T, 517K, 543S	0.96 (0.13–1.8)	1.0 (0.44–1.7)
5–9 (n = 31)	62N, 67V, 77K, 79R, 110S, 114R, 114K, 130N, 147V, 148N, 150A, 159F, 189P, 212R, 244D, 270R, 300Q, 304R, 304K, 335T, 335N, 376D, 517R, 519R, 520I, 520T, 535K, 543G, 571H, 571Y, 580A	0.96 (0.6–1.4)	1.1 (0.4–2.0)
3–4 (n = 57)	11V, 16V, 63Y, 65R, 66T, 66I, 66A, 71I, 81K, 84S, 106R, 113S, 117S, 124K, 130S, 151K, 156P, 178V, 184T, 189S, 206K, 206E, 212K, 213T, 244N, 247L, 250R, 250K, 254R, 310D, 311S, 321A, 327G, 327E, 333K, 333E, 334A, 357I, 374Q, 374L, 377A, 379K, 379R, 390I, 424I, 431T, 438I, 473M, 480I, 480V, 523R, 531K, 531R, 544R, 553I, 564V	0.95 (0.33–1.7)	1.0 (0.44–1.8)

Abbreviations: BL, baseline; FCEC₅₀, fold change in half maximal effective concentration; Ref, reference.

Table 4.

Sofosbuvir Susceptibility of NS5B Site-Directed Mutants^a

Genotype	Mutation	Replication Capacity, % Relative to WT, Mean ± SDb	SOF FCEC50 vs WT	SOF FCEC90 vs WT
1a	V11I	215 ± 15	0.59	0.91
	C110S	292 ± 26	1.1	1.2
	A117S	257 ± 10	1.1	1.5
	L159F	8.9 ± 1.0	1.2	1.1
	Q206K	106 ± 14	1.1	0.94
	K254R	88 ± 1.1	1.1	1.1
	R300Q	146 ± 2.2	1.1	0.91
	S335N	186 ± 16	1.2	0.94
	T390I	192 ± 11	0.89	0.85
	I520T	309 ± 190	1.3	1.1
	R531K	121 ± 3.3	1.1	0.91
	G543S	159 ± 6.2	0.94	0.93
1b	N110S	166 ± 2.9	0.94	0.89
	T130N	71 ± 2.7	0.99	1.0
	L159F	24 ± 6.3	1.3	1.4
	S189P	113 ± 7.0	1.2	1.1
	K212R	71 ± 4.0	1.0	1.3
	S335N	66 ± 8.0	0.75	0.89
	T520I	76 ± 2.0	0.85	1.0
	L564V	<0.1	NA	NA
	L159F + C316N	65 ± 3.5	1.6	0.9
2a	H254R	90 ± 5.4	0.90	1.3
3a	T130N	138 ± 5.9	0.99	0.96
	A150T	90 ± 0.88	1.1	1.1
	A150V	133 ± 9.7	0.99	1.0
	L159F	10 ± 2.7	1.3	1.2
	K206E	115 ± 3.7	0.94	0.84
	K212R	<0.1	NA	NA
	V321A	15 ± 2.2	1.3	1.8
	A335T	90 ± 1.4	0.97	1.0
	G543S	73 ± 2.0	0.95	0.95
	V553I	88 ± 2.8	0.99	0.95

Abbreviations: FCEC₅₀, fold change in half maximal effective concentration; FCEC₉₀, fold change in 90% effective concentration; NA, not available; SD, standard deviation; SOF, sofosbuvir; WT, wild type.

^a NS5B substitutions detected by population sequencing in >2 subjects from the resistance analysis population across all genotypes. Number of subjects shown in the table is for each genotype.

^b Average of at least 2 measurements.

The impact of treatment-emergent variants L159F and V321A on susceptibility to sofosbuvir and ribavirin was assessed by phenotypic analysis of site-directed mutants containing these substitutions. L159F imparted a 1.2- to 1.3-fold shift in EC₅₀ to sofosbuvir, whereas V321A demonstrated a 1.3-fold shift in EC₅₀ to sofosbuvir (Table 4). No reduced susceptibility to ribavirin was observed for either L159F or V321A. Both L159F and V321A demonstrated reduced replication capacity in replicon system in the absence of the sofosbuvir.

Analysis of the Subject Sample With S282T

The S282T variant was detected in 1 subject following sofosbuvir monotherapy at 4 weeks posttreatment (Figure 2). As determined by deep sequencing, the S282T variant in this sample was present in >99% of the viral population. The S282T variant decreased to 27% at 8 weeks posttreatment. Ultimately, S282T became undetectable (below the 1% assay cutoff) at weeks 12 and 24 posttreatment, suggesting impaired fitness of S282T in vivo.

Figure 2.

Evolution of sofosbuvir resistance in the subject with S282T. The viral load profile for the subject who received sofosbuvir monotherapy for 12 weeks (treatment period depicted as a gray box) and developed S282T is shown. The percentage of the wild-type serine (S) or resistance-associated threonine (T) at position 282 as determined by deep sequencing and the sofosbuvir fold change in the 50% effective concentration (FCEC₅₀) from the baseline and replication capacity (RC) at each time point tested is shown in boxes. The dotted line at 15 IU/mL represents the limit of detection (LOD) of the viral load assay. Abbreviation: HCV, hepatitis C virus.

The week 4 HCV sample with S282T displayed 13.5-fold reduced susceptibility to sofosbuvir compared to the same subject's baseline sample (8.1-fold compared with reference; Table 5). In addition, a significant reduction in replication capacity (replication capacity, 1.7% of wild-type compared with 16% at baseline) was observed in the recombinant replicon containing NS5B from this sample. S282T site-directed mutants also replicated poorly across genotypes 1–6 with replication capacity ranging from 1% to 11% compared with corresponding wild-type replicons (data not shown). These in vitro replication capacity data are consistent with the observed rapid reduction in the S282T variant prevalence over time after treatment in this subject.

Table 5.

Sequencing and Phenotypic Results for the Genotype 2b Subject With S282T

Visit	Viral Load, IU/mL	Population Sequence Change From Baseline: NS5B	Deep Sequencing: S282T	FCEC50 vs Ref	FCEC50 vs Baseline	FCEC90 vs Ref	FCEC90 vs Baseline	RC
Baseline	23 600 000	NAa		0.60	NA	0.92	NA	16 ± 7.2
FU-4 (16 wk)	21 700	S79N, V/I147I, S282T, I/T309T, T/A/I/V312T	>99%	8.1	14	5.7	6.2	1.7 ± 0.8
FU-5 (17 wk)	252 000	S79N, V/I147I, S282T, I/T309T, T/A/I/V312T	97.3%	NA	NA	NA	NA	NA
FU-8 (20 wk)	647 000	S79N, V/I147I, S282S/T, I/T309T, T/A/I/V312T	27.6%	0.91	1.5	1.4	1.5	13 ± 6.3
FU-12 (24 wk)	8 140 000	S79N, V/I147I, I/T309T, T/A/I/V312T	<1%	0.91	1.5	2.1	2.3	12 ± 3.7
FU-24 (36 wk)	11 400 000	S79N, V/I147I, I/T309T, T/A/I/V312T, S543G/S	<1%	0.95	1.6	1.1	1.2	18 ± 7.1
FU-48 (60 wk)	24 300 000	S79N, V/I147I, I/T309T, T/A/I/V312T	<1%	1.1	1.8	1.2	1.3	19 ± 3.5

S282T mutation is indicated in bold.

Abbreviations: FCEC₅₀, fold change in half maximal effective concentration; FCEC₉₀, fold change in 90% effective concentration; FU, follow-up; NA, not applicable; RC, replication capacity; Ref, reference.

^a Changes compared to reference: S15G, L31M, Y33F, C39S, K43R, Q47L, T57V, D65E, K69Q, I71V, L73R, T84S, L85V, Q90A, R98K, K100R, G113R, K120R, K124E, P130Q, Q131H, P135D, V147V/I, T179A, Q184K, A189S, Q199E, E202D, Y203F, A209G, E210S, E247Q, F267L, L285F, I293M, V297I, V309I/T, A310D, T312T/A/I/V, T329N, P353L, G375D, Y383F, L392I, A393S, I405V, L425I, M434L, V435A, D440N, S450A, M474L, H479P, T483S, S487A, A488T, V499A, S513A, R514Q, K517R, V520I, K531R, R543S, L544R, S549G, R566H, S571L, F574L, G575C, F580S, L585I.

DISCUSSION

This report represents analysis of sofosbuvir resistance from the largest number of sofosbuvir-treated subjects available to date. To maximize the opportunity to detect resistance to sofosbuvir, we employed multiple approaches including deep sequencing and extensive phenotypic analysis at time points most likely to be enriched for resistant variants (earliest time after relapse). Pretreatment sequence data were generated from 1645 subjects who later received sofosbuvir. No S282T variant was identified, and no polymorphisms present at baseline were associated with failure to achieve SVR. Phenotypic analysis of virus in plasma from 183 subjects failed to reveal the presence of virus with reduced sofosbuvir susceptibility before exposure to the drug. Comprehensive analysis of all NS5B sequences identified 2 variants, L159F and V321A, that appear to be sofosbuvir-treatment emergent substitutions. L159F and V321A emerged in 6 and 5 genotype 3–infected subjects, respectively, without significant effects on in vitro susceptibility to sofosbuvir or ribavirin. L159 is conserved in genotype 3a; however L159F is observed in about 10% of genotype 1b subjects at baseline whereas V321A is not observed at baseline across genotypes (V321I is observed in ∼2.5% of genotype 1b subjects). It is possible that these substitutions provide a fitness advantage compared with the HCV variants without them under drug pressure, and thus are enriched in the viral population in a small subset of patients with relapse. The clinical significance of L159F and V321A in subjects treated with sofosbuvir remains to be determined. In subjects not achieving SVR, no other recognized RAVs were detected in posttreatment (relapse) samples, and no other RAVs or viruses with significantly reduced sofosbuvir susceptibility were found, with a single exception in a subject who received sofosbuvir monotherapy and had S282T [23]. To explain relapse in the absence of detectable viral resistance, we hypothesize that there is residual replication in the liver during drug treatment (with undetectable HCV RNA in the blood) followed by relapse after treatment cessation. For the few subjects with emergence of S282T, L159F, or V321A at relapse, it is conceivable that these viral variants were suppressed by sofosbuvir, but enriched relative to wild-type in the liver during this “undetectable” stage and then become detectable in the blood during posttreatment relapse. Few, if any, other examples of a potent antiviral drug that selects so rarely for viral variants with reduced susceptibility have been reported before. Possible reasons for the rarity of sofosbuvir resistance include the well-conserved binding site within the active site of the viral polymerase, reduced fitness of the resistant (S282T) virus, insufficient magnitude in reduction of sofosbuvir susceptibility conferred by resistant variants to overcome the concentration of drug present at viral replication sites, and control of resistant variants by other antivirals in the combination regimen.

The consistent activity of sofosbuvir across a broad range of subject-derived viruses belonging to multiple genotypes and subtypes is noteworthy. This appears to be a unique characteristic of the class of nucleoside NS5B inhibitors, likely related to the genetic conservation of the active site in which they bind. Nonnucleoside NS5B inhibitors, NS3 inhibitors, and NS5A inhibitors currently approved or in clinical development all have significant variations in susceptibility across HCV genotypes. Furthermore, natural variants within a given genotype can give rise to altered susceptibilities and possibly altered treatment responses [44–46].

This study has several limitations. The number of subjects with specific clinical diagnoses or comorbidities is small. In addition, adherence to antiviral combination regimens in real clinical practice may not be as good as in clinical trials. Finally, the number of sofosbuvir-treated subjects infected with genotype 4, 5, and 6 HCV was relatively small. Therefore, future studies should include continued monitoring of sofosbuvir-treated subjects for new variants, and characterization of these variants and of S282T, L159F, and V321A. Furthermore, the lack of sofosbuvir resistance following relapse supports ongoing clinical investigation of retreatment with alternate sofosbuvir-containing regimens. In addition, it would be interesting to investigate whether viral quasispecies complexity and diversity at baseline can predict SVR or relapse treatment outcome.

In summary, throughout 9 sofosbuvir phase 2 and 3 clinical trials, no antiviral resistance to sofosbuvir was detected either at baseline or in cases of virologic relapse except in 1 monotherapy subject. This favorable feature of sofosbuvir makes it an attractive option for future combination strategies against HCV.