Hepatitis C Virus Variants with Decreased Sensitivity to Direct-Acting Antivirals (DAAs) Were Rarely Observed in DAA-Naive Patients prior to Treatment

Abstract

The prevalence of naturally occurring hepatitis C virus (HCV) variants that are less sensitive to direct-acting antiviral (DAA) inhibitors has not been fully characterized. We used population sequence analysis to assess the frequency of such variants in plasma samples from 3,447 DAA-naive patients with genotype 1 HCV. In general, HCV variants with lower-level resistance (3- to 25-fold increased 50% inhibitor concentration [IC₅₀]) to telaprevir were observed as the dominant species in 0 to 3% of patients, depending on the specific variant, whereas higher-level resistant variants (>25-fold-increased IC₅₀) were not observed. Specific variants resistant to NS5A inhibitors were predominant in up to 6% of patients. Most variants resistant to nucleo(s/t)ide active-site NS5B polymerase inhibitors were not observed, whereas variants resistant to non-nucleoside allosteric inhibitors were observed in up to 18% of patients. The presence of DAA-resistant variants in NS5A, NS5B, or NS3 (including telaprevir-resistant variants), in baseline samples of treatment-naive patients receiving a telaprevir-based regimen in phase 3 studies did not affect the sustained viral response (SVR). Treatment-naive patients with viral populations containing the telaprevir-resistant variants NS3 V36M, T54S, or R155K at baseline achieved a 74% SVR rate, whereas patients with no resistant variants detected prior to treatment achieved a 76% SVR rate. The effect of specific resistant variant frequency on response to various DAA treatments in different patient populations, including interferon nonresponders, should be further studied.

INTRODUCTION

The hepatitis C virus (HCV) NS3-4A protease, NS5A protein, and NS5B polymerase are essential for viral replication. Inhibitors of their function comprise the vanguard of a new class of HCV inhibitors, namely, direct-acting antivirals (DAAs) for HCV (1, 2). HCV has high sequence diversity both between and within the various genotypes and subtypes (3, 4). The combination of the low fidelity of the HCV polymerase, high replication rate, and strong selective pressures on the virus lead to a unique viral quasispecies in each patient. This viral quasispecies in patients exists as a mixed population of viruses, with the most fit viruses being the predominant viral populations observed by population sequence analysis. In addition, new populations with every potential substitution are likely created and lost each day, some of which convey various degrees of resistance to DAAs (4–6). Thus, it is likely that all patients have DAA-resistant variants that require one or two mutations present at some level prior to treatment, albeit usually transient and below the current detection limits, since they are typically less fit than drug-sensitive (wild-type [WT]) virus. Indeed, drug-resistant substitutions have been shown to emerge in vitro and in vivo with all classes of DAAs after inhibition of the drug-sensitive virus population or after treatment failure (7–10).

The prevalence of a resistant variant in a patient's viral quasispecies is generally determined by its replicative fitness and selective advantage compared to the rest of the viral population (5). In general, substitutions in critical residues near the highly conserved active site of an enzyme [e.g., the protease catalytic site or the nucleo(s/t)ide incorporation site], are likely to impair enzyme function, resulting in diminished replicative capacity and decreased viral fitness. In contrast, substitutions at less critical residues which participate in the binding of the DAA have the potential to be better tolerated and result in enzymes and viral variants with a relatively lower fitness consequence (e.g., substitution at an allosteric site) (11).

Because of the intrinsic fitness cost of resistant mutations, patients possessing DAA-resistant variants as the predominant viral quasispecies have rarely been observed in spite of the expected presence of these variants in most patients at a low level (6). There are few large studies that report the prevalence of DAA-naive patients with viral populations predominantly resistant to DAAs (12, 13). DAA inhibitors are now approved in North America, Europe, Asia, and South America (protease inhibitors telaprevir [VX-950] and boceprevir) (14–17). In addition, there are a large number of investigational DAA inhibitors that continue to advance in development, some which target the NS5B polymerase or the NS5A protein. It is therefore important to assess the natural prevalence of patients with HCV viral populations containing variants resistant to DAAs and understand the clinical consequences of these variants on different treatment regimens. Here, we report the prevalence of NS3-4A, NS5A, or NS5B polymerase (active site and allosteric) inhibitor-resistant variants as the majority viral population in 3,447 genotype 1 patients who are naive to DAA treatment and their subsequent antiviral response to a telaprevir-based treatment regimen.

Materials and Methods

Patient population.

This study included 3,447 DAA-naive patients who had chronic, genotype 1 (subtype 1a or 1b) HCV infection and who were enrolled in phase 2 and 3 clinical studies of telaprevir (18–24). Studies were conducted in full compliance with the guidelines of Good Clinical Practice and of the World Medical Assembly Declaration of Helsinki. Prior to study initiation, protocols and informed consent forms were reviewed and approved by institutional review boards at each site. All patients provided written informed consent before participating in any study-related activity.

HCV RNA sequencing.

Prior to treatment, population sequence analysis (sensitivity ∼20%) of the NS3-4A region was performed in 2,111 subtype 1a and 1,336 subtype 1b samples from DAA treatment-naive patients enrolled in the phase 2 or phase 3 trials of telaprevir. All variants observed were considered, including minority species, which appeared as mixtures in the population sequencing. Population sequence analysis of the NS5A and NS5B regions was performed in 538 subtype 1a and 239 subtype 1b samples from patients participating in the phase 2 trials of telaprevir, and samples were also collected prior to treatment.

Sequencing methods have been presented elsewhere (12, 25). Briefly, a blood sample was collected from patients by venipuncture of a forearm vein into tubes containing EDTA (K₂) anticoagulant. Plasma was separated by centrifugation, divided into aliquots, and stored at −80°C. Sequence analysis of HCV utilized nested reverse transcriptase PCR (RT-PCR) amplification of an ∼9-kb HCV RNA fragment spanning the HCV polyprotein coding region. The DNA from this PCR was purified using a QIAquick 96 PCR purification kit (Qiagen, Valencia, CA) and was analyzed on an agarose gel. Purified DNA was sent to Beckman-Coulter (Danvers, MA) for sequencing of the NS3-4A, NS5A, or NS5B regions. Sequences were aligned and analyzed using the software Mutation Surveyor (SoftGenetics, State College, PA).

Analysis of drug-resistant variants.

For the approved protease inhibitors, telaprevir and boceprevir, variants reported in the U.S. primary indications were analyzed, including rarely observed (<10% for boceprevir and <2% for telaprevir) variants (14, 17). For the investigational protease, NS5A, and polymerase inhibitor compounds, variants reported from clinical trials were analyzed when available, and the use of variants not observed clinically is noted in the text. These resistant variants and their associated increase in resistance were recently summarized by the HCV Drug Resistance Advisory Group (DRAG) group (26).

Analysis for association of variants.

In the subset of patients with NS3, NS5A, and NS5B, the sequence associations of resistant variants within and between regions was investigated. To determine whether any resistant variants whose presence may be associated with another resistant variant, a chi-square test was performed comparing all possible resistant variant positions with each other, with the likelihood-ratio test used to calculate χ². A two-tailed test was utilized, such that both significantly more or less frequent occurrences could be detected to identify both pairs of variants that may result in fitness improvements and pairs that may result in reductions in relative fitness.

GenBank accession numbers.

The GenBank accession numbers for the baseline sequences of the NS3-4A and NS5AB regions for the patients in the present study are KC123434 to KC127656.

RESULTS

Baseline prevalence of NS3-4A protease inhibitor variants.

HCV population sequence analysis of the complete NS3 and 4A regions was obtained from 2,111 subtype 1a and 1,336 subtype 1b DAA-naive patients. Approved and investigational protease inhibitors in phase 2 or 3 trials have identified a consistent clinical resistance pattern. For the approved, covalent, linear inhibitors (telaprevir and boceprevir) the primary treatment emergent variants have been observed at NS3 positions 36, 54, 155, 156, and 168. For the investigational, noncovalent, linear (asunaprevir [BMS-650032], BI 201335 and ABT-450), and macrocyclic (simeprevir [TMC435] and danoprevir [RG7227]) inhibitors, resistant variants have been observed predominantly at NS3 positions 155 and 168 (26).

The prevalence of telaprevir- and boceprevir-resistant variants was low in both subtype 1a and 1b DAA-naive patients (Table 1). In subtype 1a patients, the variants most commonly observed during treatment are V36M and R155K alone or in combination. Both of these variants were observed by population sequence analysis in <1% of patients (V36M, 0.5%; R155K, 0.9%) prior to treatment, and the V36M+R155K combination variant was not observed in this data set. In subtype 1b patients, the variants commonly observed during treatment include A156S/T, V36A, and T54A. These variants were not observed prior to treatment. Other variants—V36M (0.07%), V36L (0.90%), and T54S (2.0%)—were observed prior to treatment in subtype 1b patients. Importantly, all variants seen in DAA-naive patients were lower-level resistant variants, and none of the higher-level resistant (>25-fold increased 50% inhibitor concentration [IC₅₀]) variants were observed.

Table 1.

Prevalence of naturally occurring dominant protease inhibitor-resistant variants in patients naive to DAA treatment

Inhibitora	NS3 majority amino acid		Majority amino acid prevalence		Observed variant(s) (%)b		Variant(s) not observed
	1a	1b	1a	1b	1a (n = 2,111)	1b (n = 1,336)
1,2	V36	V36	97.6	98.8	I (0.09), L (1.89), M (0.47)	I (0.30), L (0.90), M (0.07)	A, G
3	Q41	Q41	99.3	99.9			R
8	F43	F43	99.9	100	Y (0.05)
1,2	T54	T54	97.4	98.2	A (0.05), S (3.08)	S (1.95)	C, G
2	V55	V55	95.4	99.5	A (2.75), I (2.13)	A (0.37)
8	Y56	Y56	99.9	74.2			H
3	Q80	Q80	60.9	96.3	K (37.57), R (0.76)	K (0.82), R (0.60)
2	V107	V107	99.9	99.4	I (0.24)	I (0.60)
1	I132	V132c	99.6	73.0	V (0.52)
1–7	R155	R155	99.2	100	K (0.85)		I, G, M, T, Q, S
1,2	A156	A156	100	100			F, N, S, T, V
2	V158	V158	100	99.9		I (0.07)
1–7d	D168	D168	99.8	99.6	E (0.24)	E (0.60)	G, H, V, T, Y
2	I170	V170	96.4	73.5	T (0.14)	A (0.07), T (0.15)	F
2	L175e	M175	99.9	99.0		L (1.05)

^aKey: 1, telaprevir; 2, boceprevir; 3, simeprevir; 4, asunaprevir; 5, BI-201335; 6, danoprevir; 7, ABT-450; 8, MK-5172.

^bThis includes all variants observed, including variants observed as a minority viral species (∼20%), allowing for multiple variants in a single sample.

^cI132V did not shift the IC₅₀ for telaprevir in a subtype 1a replicon, and V132 is present in the subtype 1b replicon and the majority of subtype 1b sequences.

^dOnly D168N was reported as a rare treatment emergent variant for telaprevir and boceprevir, and D168N or the more commonly observed treatment emergent D168 variants did not confer resistance to telaprevir in the replicon system.

^eThe majority amino acid in subtype 1a (L175) is the resistant variant (M175L) in subtype 1b.

The prevalence of resistant variants for the investigational protease inhibitors simeprevir, danoprevir, asunaprevir, BI 201335, and ABT-450 were also low in the DAA-naive population (Table 1). Additional variants resistant to these drugs are observed at NS3 position 168; the most frequent of these was D168E, which was observed in 0.24% of subtype 1a patients and 0.60% of subtype 1b patients. Variants at NS3 positions 41 and 80 have been reported for simeprevir (27). The NS3 Q80K variant was observed here in 38% of subtype 1a patients, similar to the recently reported 47% prevalence (28). Although NS3 Q80K was enriched in patients receiving simeprevir in the PILLAR study, the Q41R variant was not. NS3 Q41R was also not observed in the present study. The prevalences of NS3 variants enriched for by MK-5172 were also very low: F43S (0.05%) and Y56H (0%).

Baseline prevalence of NS5A inhibitor variants.

HCV population sequence analysis of the complete NS5A region was obtained from 538 subtype 1a and 239 subtype 1b DAA-naive patients. For the NS5A inhibitor daclatasvir (BMS-790052), the primary resistant variants selected in vitro were NS5A M28T, Q30E/H/R, L31M/V, P32L, and Y93C/H/N for genotype 1a and L31F/V, P32L, and Y93H/N for genotype 1b, with L23F, R30Q, and P58S selected as secondary resistance substitutions (7). Although the subtype 1b variants conferred lower-level resistance (a 5- to 28-fold shift in EC₅₀), the variants in subtype 1a conferred higher levels of resistance (233- to 5,367-fold shift in EC₅₀) (7). The reported clinical resistance pattern appears similar to that seen in vitro; patients with viral breakthrough on a dual combination of protease inhibitor asunaprevir and NS5A inhibitor daclatasvir had NS5A Q30E/R, L31M/V, and Y93C/N variants (29). The prevalence of resistant variants in the context of the NS5A inhibitors is highly dependent on subtype due to several of the positions having different baseline amino acids in each subtype (see individual subtype majority amino acid in Table 2). In subtype 1a DAA-naive patients, NS5A M28T (0.37%), Q30H (1.3%), Q30R (0.74%), L31M (0.93%), P58S (0.19%), Y93C (0.37%), and Y93N (0.37%) were observed, whereas no variants were observed at NS5A positions 23 or 32. In subtype 1b patients, NS5A Q30H (0.42%), L31M (6.28%), P58S (3.35%), and Y93H (3.77%) variants were observed.

Table 2.

Prevalence of naturally occurring dominant NS5A inhibitor-resistant variants in patients naive to DAA treatment

NS5A majority amino acid		Majority amino acid prevalence		Observed variant(s) (%)		Variant(s) not observed
1a	1b	1a	1b	1a (n = 538)	1b (n = 239)
L23	L23	99.8	99.6			F
M28	L28	96.5	99.2	T (0.37)
Q30	R30a	98.7	92.9	H (1.3), R (0.74)	H (0.42)	E
L31	L31	99.3	94.6	M (0.93)	M (6.28)	F, V
P32	P32	100	100			L
Q58	P58	100	92.9	S (0.19)	S (3.35)
Y93	Y93	98.7	97.5	C (0.37), N (0.37)	H (3.77)

^aThe majority amino acid in subtype 1b (R30) is the resistant variant (Q30R) in subtype 1a.

Baseline prevalence of NS5B inhibitor variants.

HCV population sequence analysis of the NS5B region was completed from 538 subtype 1a and 239 subtype 1b DAA-naive patients. Compounds that inhibit the NS5B polymerase through different mechanisms of action have been characterized (30, 31). These include the active-site nucleo(s/t)ide inhibitors, as well as non-nucleoside inhibitors that bind at least three distinct allosteric sites designated as the palm, thumb, and finger-loop sites.

Baseline prevalence of NS5B polymerase nucleo(s/t)ide inhibitor variants.

The nucleo(s/t)ide class of NS5B polymerase inhibitors bind the catalytic site and inhibit replication (32, 33). Currently, a number of inhibitors in this class have advanced in the clinic (34). Compared to the other classes of HCV DAA inhibitors, the number of resistant variants selected in the clinic is significantly lower, most likely due to the impaired replicative fitness conferred by amino acid changes in the polymerase active site (35, 36). The variant NS5B S282T conveys decreased susceptibility to the 2′-methyl-cytidine inhibitor mericitabine (RG7128) and sofosbuvir (GS-7977) in the replicon system. NS5B S282T has been observed in a very small number of patients either receiving NM283 monotherapy (37) or after failing treatment with mericitabine in combination with danoprevir (38). In addition, variants NS5B S96T and N142T appear to be compensatory mutations for S282T in the replicon (39). The passaging of the 2′-methyl-guanosine inhibitor PSI-352938 did not readily select for variants in a genotype 1 replicon system; however, variants NS5B C223H/Y and V321I were selected after multiple passages in a genotype 2a replicon (J6/JFH-1), conveyed a 2.1- to 3.7-fold-decreased susceptibility, and are included in this analysis, but resistance >2-fold in genotype 1b was only conveyed by the combination of variants which had a significant impact on viral fitness (40). The NS5B S96T, N142T, C223H/Y, and S282T variants were not observed as the dominant viral population in any patients in the present study, and the NS5B V321I variant was observed in 0.19 and 2.51% of subtype 1a and 1b patients, respectively (Table 3).

Table 3.

Prevalence of NS5B inhibitor-resistant variants naturally occurring in patients naive to DAA treatment

NS5B majority amino acid		Majority amino acid prevalence		Observed variant(s) (%)		Variant(s) not observed
1a	1b	1a	1b	1a (n = 538)	1b (n = 239)
Nucleo(s/t)ide
S96	S96	100	100			T
N142	N142	100	98.7			T
C223	C223	100	100			H, Y
S282	S282	100	100			T
V321	V321	99.8	98.3	I (0.19)	I (2.51)
Palm site
C316	C316	99.8	88.7	Y (0.19)	N (10.88)	F, S
M414	M414	99.6	98.3	L (0.19), V (0.19)	L (1.26), T (0.42)	I
Y448	Y448	100	98.7	H (0.19)	H (1.26)	C
C445	C445	100	99.6		F (0.42)
Y452	Y452	100	99.6		H (0.42)
G554	G554	100	99.2			D
S556	S556	99.2	99.2	G (0.38)	G (0.42)	D, N
D559	D559	100	98.4	N (0.57)		G
Thumb site
L419	L419	100	99.2			M, S
R422	R422	99.6	100	K (0.56)
M423	M423	98.1	99.6	A (0.19), I (1.49), T (0.19), V (0. 19)	A (0.42)
M426	M426	90.5	99.2			T, V
I482	I482	100	100			L, T
A486	A486	100	100			V
V494	V494	99.6	98.8		A (0.8)
Finger-loop site
A421	A421	83.3	94.1	V (17.84)	V (6.28)
P495	P495	100	99.2			A, L, S, T
P496	P496	100	99.2		A (0.84)
A499a	V499	96.5	81.2		A (14.23)

^aThe majority amino acid in subtype 1a (A499) is a resistant variant in subtype 1b (V499A).

Baseline prevalence of NS5B allosteric palm site inhibitor variants.

The resistance profile of inhibitors that bind allosterically in the palm domain of NS5B, near the elongation pocket of the polymerase, has been thoroughly characterized using the replicon system (41, 42) and the clinical resistance pattern from early studies for some of the compounds in development has been reported (2). In early clinical trials with ABT-333, NS5B variants C316Y, M414T, Y448H, S556G, and D559G/N were observed (43). A partially overlapping resistance profile of NS5B C316N, C445F, Y448H, and Y452H was reported for tegobuvir (GS-9190) in patients that failed in the combination of tegobuvir, and the protease inhibitor GS-9256 and ribavirin (44, 45). The primary clinical resistance profile for ANA598 includes M414T and G554D (46). Of the clinically observed resistant variant positions reported for the palm site inhibitors, only variants at NS5B 554 were not observed in DAA-naive patients (Table 3). The highest prevalence (11%) was observed for C316N, which was originally identified as conveying a modest 3-fold decrease in susceptibility to HCV-796 (47), and a 5.2-fold decrease in susceptibility to tegobuvir (48). In patients that failed treatment with tegobuvir and the protease inhibitor GS-9256, the presence of C316N, C445F, or Y452H prior to treatment in subtype 1b patients was associated with the emergence of NS3 D168 variants without selection of additional NS5B mutations such as was those observed in subtype 1a patients (44, 45). The prevalence of the remaining resistance variants in subtype 1a were all below 1%: C316Y, M414L, M414V, Y448H, S556G, and D559N. In subtype 1b the prevalence of most variants also remained below 1% (M414T, C445F, Y452H, and S556G), with a slightly higher prevalence for M414L and Y448H (1.26%).

Baseline prevalence of NS5B allosteric thumb site inhibitor variants.

Clinical candidates VCH-759, VX-222, and filibuvir (PF-00868554), bind allosterically in the thumb region of NS5B, and the resistance profiles, including early clinical resistance data, have been reported (10, 49, 50). Variants selected clinically during a 3-day dosing study of VX-222 were at NS5B 419, 422, 423, 482, 486, and 494 (51); however, in the ZENITH study in patients that had virologic breakthrough during treatment with VX-222 and telaprevir the NS5B variants observed were L419S and R422K (52). A partially overlapping resistance profile was reported for filibuvir with variants at NS5B 423, 426, 482, and 494 identified in vitro (53). Variants at NS5B 423 were the predominant treatment emergent variant in patients that failed a filibuvir-containing regimen (54). Of the clinically observed resistant variant positions reported for the thumb site inhibitors, most were not observed in the present study, including variants at NS5B 419, 426, 482, and 486. NS5B V494A was observed in 0.80% of subtype 1b patients and R422K in 0.56% of subtype 1a patients (Table 3). Only variants at NS5B 423 had a >1% prevalence in subtype 1a patients, with M423I at 1.49% and M423A/T/V at 0.19%.

Baseline prevalence of NS5B allosteric finger-loop site inhibitor variants.

Clinical candidates MK-3281, TMC647055, BI 207127, and BMS-791325 bind allosterically in the finger-loop region of NS5B. Decreased sensitivity to the first clinical candidate in this class JTK-109 was conferred by substitutions at the NS5B positions 495, 496, and 499 (55). A similar resistance profile was recently reported for BI 207127 and BMS-791325. The variant P495L was observed in a patient that failed treatment with BI 207127 in combination with the protease inhibitor BI 201335 (56). The NS5B variants P495L/S variants alone or in combination with NS5B A421V were observed in patients that failed a regimen with BMS-791325 and pegylated interferon (peginterferon) and ribavirin (57). The NS5B A421V variant, while conferring low-level resistance to BMS-791325 was shown in combination with P495L to confer ∼300-fold decrease in susceptibility (57). Variants at NS5B position 495 were not observed in the study and P496A was observed in 0.84% of the subtype 1b patients (Table 3). The NS5B A421V and V499A substitutions are more commonly observed, with A421V in 17.8% of subtype 1a and 6.3% of subtype 1b patients. The V499A variant, which confers a 2- to 4-fold decrease in susceptibility to JTK-109 (55), was observed in 14% of subtype 1b patients and is the majority amino acid observed in subtype 1a patients (96%), whereas no evidence for the clinical impact of V499A has been reported.

Naturally occurring combinations of variants.

Combinations of DAA-resistant variants both in a single target protein and across multiple targets have been reported following failure of single and combination DAA regimens (2, 12, 45, 56, 57). However, the natural pretreatment prevalence of resistant variant combinations, especially across multiple targets has not been reported. To investigate associations between DAA-resistant variants in NS3, NS5A, and NS5B prior to treatment, a chi-square test was performed to identify variants whose co-occurrence significantly deviated from expectation (Table 4).

Table 4.

Associations between DAA-resistant variants in NS3, NS5A, and NS5B

Variant 1	Variant 2	Pa	Prevalence (%)b
			Variant 2/variant 1		Variant 1/variant 2
			Observed	Expected	Observed	Expected
NS3 54	NS3 55	<0.0001	33 (5/15)	2.5	26 (5/19)	1.9
NS5B 421	NS5B 423	<0.0001	7.3 (8/109)	1.6	67 (8/12)	14.1
NS3 41	NS3 168	0.0032	33 (1/3)	0.3	50 (1/2)	0.4
NS3 55	NS3 80	0.0041	63 (12/19)	31.5	4.9 (12/244)	2.5
NS5B 316	NS5B 421	0.0046	0 (0/26)	14.1	0 (0/109)	3.4
NS3 54	NS3 155	0.0035	13 (2/15)	0.8	33 (2/6)	1.9
NS3 155	NS5B 556	0.0135	17 (1/6)	0.4	33 (1/3)	0.8
NS3 80	NS5B 316	0.0150	1.2 (3/244)	3.4	12 (3/26)	31.5
NS3 36	NS3 41	0.0249	10 (1/10)	0.4	33 (1/3)	1.3
NS3 41	NS3 55	0.0359	33 (1/3)	2.5	5 (1/19)	0.4
NS3 54	NS5B 556	0.0402	7 (1/15)	0.4	33 (1/3)	1.9
NS5A 30	NS5A 31	0.0436	10 (1/10)	0.6	20 (1/5)	1.3
NS3 54	NS5B 421	0.0531	33 (5/15)	14.1	4.6 (5/109)	1.9
NS5A 30	NS5A 93	0.1402	10 (1/10)	1.3	8 (1/12)	1.3
NS3 36	NS3 54	0.1818	10 (1/10)	1.9	6.7 (1/15)	1.3
NS3 54	NS5A 93	0.2246	7 (1/15)	1.6	8 (1/12)	1.9
NS3 170	NS5B 423	0.3449	4.5 (1/22)	1.6	8 (1/12)	2.8

^aValues in boldface indicate significant observations.

^bThe observed and expected prevalences of each variant in the presence of the other variant is displayed. For example, the first row indicates that of the 15 patients, a variant at NS3 54 (variant 1), 5 possess a variant at NS3 55 (variant 2) as well. This prevalence of 33% is compared to the expected prevalence of NS3 54 variants (2.5%) observed in the substudy of patients with NS3, NS5A, and NS5B sequence. Similarly, 26% of patients with a variant at NS3 54 also have a variant at NS3 55, in contrast to the expectation of 1.9%.

In the present study, combinations of resistant variants across NS3, NS5A, and NS5B were rare, and most combinations were observed in a single patient. The chi-square test did not identify any combinations of resistant variants in NS3+NS5A or NS5A+NS5B as significant. However, four combinations of resistant variants in NS3 and NS5B had small differences in prevalence compared to the expected prevalence of each variant alone. The combinations NS3 155+NS5B 556 and NS3 54+NS5B 556 are explained by a single patient with the NS3 T54S+NS3 R155K+NS5B S556G variants that was previously described by Bartels et al. (see Fig. 3A in reference 12). The remaining two combinations of resistant variants were observed in five or fewer patients. The combination of NS3 54+NS5B 421 had an ∼2-fold increase in prevalence over the expected level, whereas NS3 80+NS5B 316 had to an ∼2-fold decrease.

Although a variety of NS3 variant combinations were identified in the present study, most confer resistance to different protease inhibitors and have not been reported associated with treatment failure. The strongest association was the combination of variants at NS3 V55, with lower-level resistance to boceprevir, and NS3 T54, with lower-level resistance to boceprevir and telaprevir, likely due to the steric interactions between V55I/A variants and other residues in the protease active site, including T54 (58). In the complete NS3 study data set fully 69% (33 of 48) of patients with NS3 V55I also contained T54S. An association was also observed between NS3 positions 54 and 155, with 17% (3 of 18) of the patients with NS3 T54S substitution also containing R155K. The NS3 T54S+R155K combination was treatment emergent in boceprevir and telaprevir trials, but the more commonly observed treatment-emergent combination of V36M+R155K was not observed here.

To date, most NS5A inhibitors that have entered the clinic have reported a similar resistance pattern to daclatasvir. The high potency of these molecules, usually with replicon IC_50s in the low picomolar range, is counteracted in part by the 100- to 1,000-fold resistance reported for some of the combination variants, especially in subtype 1a (7, 59). Recently, a patient with the NS5A combination L28F+R30N+L31M+Y93N prior to treatment was reported to be highly resistant to the NS5A inhibitor PPI-461 (60). In this study, an association between a single patient with NS5A Q30H and L31M was identified, and three of the remaining seven patients with the NS5A Q30H also contained a second NS5A-resistant variant.

Combinations of NS5B variants were rarely observed, and interestingly appeared to be under-represented relative to the overall prevalence. Subtypic differences in prevalence for the variants confound the analysis; however, the combination of substitutions at the palm site inhibitor site NS5B 316 were not observed in combination the NS5B A421V variant, even though the overall prevalence of substitutions at the positions were 3 and 14%. The only positive association observed was between substitutions at the NS5B 421 and 423 positions. However, the absence of reported variants at NS5B 421 in the filibuvir studies where the primary treatment emergent variants were at NS5B 423 suggests caution in discussion of the potential clinical importance of this association.

Genetic barrier to generating DAA-resistant variants.

An important factor in the generation and enrichment of DAA-resistant variants is the genetic barrier to generating those variants. Genetic barrier refers to the number of nucleotide changes needed to acquire a DAA-resistant variant that can replicate in patients. In addition to the total number of nucleotide changes, the probability of a mutation occurring as a transition (G↔A, C↔T) is much higher than for transversions in HCV (61, 62). A summary of the genetic barrier for each DAA class is shown in Tables 5 to 7, with the wild-type state based on the majority amino acid in Tables 1 to 3.

Table 5.

Genetic barrier for generating protease-resistant variants

No. and type(s) of substitution required	Variant(s)
	Subtype 1a	Subtype 1b
One transition	V36A/I/M, Q41R, T54A, V55A/I, Q80R, V107I, I132V, R155G/K, A156T/V, V158I, D168G/N, V170T	V36A/I, Q41R, T54A, V55A/I, Q80R, V107I, A156T/V, V158I, D168G/N, V170A
One transversion	V36G/L, T54S, Q80K, R155I/M/S/T, A156S, D168A/E/H/V/Y, V170F	V36G/L, T54S, Q80K, R155G, A156S, D168A/E/H/V/Y, V170F
Two transitions	V170A	V170T
One transition + one transversion	T54G, R155Q, A156F/N, D168T	V36M, T54G, R155K/Q/T, A156F/N, D168T
≥2 transversion or substitutions	T54C	T54C, R155I/M/T

Table 7.

Genetic barrier for generating NS5B polymerase inhibitor-resistant variants

No. and type(s) of substitution required	Active-site nucleo(s/t)ide inhibitor	Variant(s)
		Palm	Thumb	Finger-loop
One transition	V321I, C223Y	H95Q/R, M414I/T/V, C316F, Y448C/H, G554D, S556G/N, D559G/N	R422K, M423I/T/V, M426T/V, I482T, A486V, V494A	A421V, P495L/S, V499A
One transversion	S96T, N142T, S282T	C316S/Y, M414L	L419M, I482L	P495A/T, P496A
Two transitions	None	S556D	L419S, M423A	None
One transition + one transversion	None	C316N	None	None
≥2 transversions or substitutions	C223T	None	None	None

Although many of the commonly observed NS3 protease inhibitor variants involve a single transition from the consensus codon, other rarely observed variants require transversions or multiple changes (Table 5). Variants that require transversions in subtype 1a and 1b include NS3 V36L, T54S, A156S, and several variants at 155 and 168. Subtypic differences were also observed, including the commonly observed subtype 1a treatment emergent variants NS3 V36M and R155K. These variants are each created by a single transition in subtype 1a but are rare in subtype 1b, where each variant requires both a transversion and a transition. The V36M+R155K combination variant, which would involve two transversions and two transitions in subtype 1b, has never been reported at baseline or as a treatment emergent variant in a subtype 1b patient. Other subtypic differences include NS3 V170A, which requires a single transition in subtype 1b, but two transitions in subtype 1a.

The genetic barrier for the NS5A inhibitor daclatasvir appears similar to the protease inhibitors, with the majority of treatment emergent variants requiring a single transition (Table 6). Exceptions include the NS5A L31V or L31M variants, which require a transversion in both subtypes. Subtypic differences were also observed with a transversion and transition required to create L23F in subtype 1a and M28T in subtype 1b, compared to single transitions in the opposite subtypes.

Table 6.

Genetic barrier for generating NS5A inhibitor-resistant variants

No. and type(s) of substitution required	Variant(s)
	Subtype 1a	Subtype 1b
One transition	M28T, Q30R, P32L, P58S, Y93C/H	L23F, P32L, P58S, Y93C/H
One transversion	Q30E/H, L31M/V, Y93N	L31F/M/V, Y93N
Two transitions	None	None
One transition + one transversion	L23F, L31F	M28T, Q30E/H
≥2 transversions or substitutions	None	None

The genetic barrier for the various NS5B inhibitor classes differs (Table 7). Interestingly, nearly all of the nucleo(s/t)ide-resistant variants not observed here, including S282T, require at least one transversion. The only active-site variant observed in the present study, NS5B V321I, required a single transition. For the NS5B allosteric inhibitors, while subtypic differences in activity are known, the genetic barrier at the positions conferring resistance were nearly identical in subtypes 1a and 1b, and the majority were single transitions. Exceptions for the palm inhibitors were NS5B C316S/Y and M414L, which required a transversion, and S556D and C316N, which required two mutations. For the thumb inhibitors NS5B L419M and I482L required a transversion and L419S required two mutations, and for the finger-loop inhibitors NS5B P495A/T and P496A variants both required a transversion.

Clinical outcome of patients with NS5A or NS5B inhibitor-resistant variants prior to initiation of treatment with telaprevir plus peginterferon plus ribavirin.

Although there is ample evidence in vitro that DAA-resistant variants do not alter the susceptibility to other classes of DAA inhibitors, peginterferon alfa-2a (peginterferon), and/or ribavirin (5, 34, 41), clinical demonstration of this has not yet been reported. Sequencing the NS5A and NS5B regions in patients from the phase 2 telaprevir clinical trials (18, 20, 21) allowed their antiviral response to a protease inhibitor containing regimen to be evaluated. Of 56 patients with NS5A or NS5B inhibitor-resistant variants that received a telaprevir, peginterferon, and ribavirin regimen, all had an early response, with 95% of patients having HCV RNA levels below 1,000 IU/ml at week 4 and 67% achieving a rapid virologic response (RVR). The clinical outcome was also indistinguishable, with an overall 57% sustained viral response (SVR) rate (32/56) in patients with preexisting NS5A or NS5B inhibitor-resistant variants and a 55% SVR rate (188/341) for patients with wild-type virus at baseline (Table 8).

Table 8.

Treatment outcome of patients with naturally occurring dominant protease inhibitor-resistant variants in a telaprevir combination regimen

Variant(s)	SVR% (n/N) and P valuesa
	Treatment-naive and prior relapser			Prior nonresponder
	Variants	No variants	P	Variants	No variants	P
NS5A or NS5B inhibitor-resistant variantsb	58 (22/38)	65 (142/219)	0.4656	56 (10/18)	38 (46/122)	0.1979
Telaprevir-resistant variantsc	65 (64/98)	75 (1122/1496)	0.0416	36 (5/14)	42 (153/368)	0.7859
NS3 R155K	70 (7/10)			0 (0/3)
NS3 V36M	43d (3/7)
NS3 T54S	74 (37/50)			14% (1/7)

^aP values were calculated using a two-sided Fisher exact test. The sustained virologic response values (SVR%) are expressed as percentages, where n and N (indicated in parentheses), represent the number of patients that had SVR and the total number of subjects receiving the regimen, respectively.

^bThat is, all of the variants listed in Tables 2 and 3, with the exception of variants with subtypic differences where the resistant variant is the majority population in a subtype such as NS5B V499A in subtype 1a.

^cThis group includes patients in a phase 2 or phase 3 clinical trial enrolled to receive at least 12 weeks of telaprevir in combination with pegylated-interferon and ribavirin, variants included in accordance with the telaprevir USPI label.

^dOnly 1 of 4 patients that did not achieve an SVR experienced an on-treatment virologic failure.

Clinical outcome of patients containing NS3 protease inhibitor-resistant variants prior to initiation of treatment with telaprevir plus peginterferon plus+ ribavirin.

Of the 3,447 genotype 1 patients where NS3-4A sequence data were available, 3,323 (96.5%) of these patients did not have detectable levels of telaprevir-resistant variants at baseline by population sequencing. Of the remaining 3.5% of patients with detectable telaprevir-resistant variants, 93 were enrolled in a telaprevir containing regimen of a phase 3 study. In treatment-naive patients randomized to receive 12 weeks of telaprevir in the phase 3 studies (19, 23), clinical outcomes were similar to the patients with no telaprevir-resistant variants. The SVR rate of patients with telaprevir-resistant variants at baseline was 74% (39/53) compared to an SVR rate of 76% (634/837) in patients who did not have a resistant variant at baseline. In the peginterferon and ribavirin treatment-failure population (REALIZE), SVR rates in patients with telaprevir-resistant variants at baseline were 14% (1/7), 100% (2/2), and 80% (12/15) for prior null-responders, prior partial-responders, and prior relapsers, respectively, compared to the SVR rates of 33% (46/140), 57% (53/93), and 87% (231/267) in patients who did not have a telaprevir-resistant variant at baseline. The results of a pooled analysis of all patients enrolled to receive a telaprevir, peginterferon, and ribavirin regimen in all studies (phase 2 and phase 3) are shown in Table 8. Overall, the results are similar, but the SVR rates were lower for the small number of patients with variants enrolled in the phase 2 studies, and an in-depth analysis of these patients was previously described (12).

Although the overall presence of resistant variants did not have a profound effect on the SVR outcomes, an effect of individual variants was possible, especially early in treatment where the predominance of resistant variants may have the largest affect. The mean HCV RNA decline for all treatment-naive patients containing the lower-level telaprevir-resistant variants V36M (n = 6), T54S (n = 50), or R155K (n = 10) prior to treatment with telaprevir, peginterferon, and ribavirin was greater than patients receiving peginterferon and ribavirin (Fig. 1). Patient responses to peginterferon and ribavirin treatment are variable, and the mean HCV RNA response to telaprevir, peginterferon, and ribavirin in patients with variants at baseline was equal or better than the 9.4% of patients that had undetectable HCV RNA at week 4 (RVR) while receiving peginterferon and ribavirin. Thus, it is likely that telaprevir has at least some antiviral activity in treatment-naive patients with baseline telaprevir-resistant variants. However, due to the limited number of patients with baseline resistant variants in prior nonresponders, the impact in this population on the response to telaprevir, peginterferon, and ribavirin regimen remains unclear.

Fig 1.

Comparison of early response to T/PR (telaprevir, peginterferon, and ribavirin) in treatment-naive patients with or without predominant telaprevir-resistant variants prior to treatment. The y axis displays the mean HCV RNA decline from baseline through the first 4 weeks of treatment. The results obtained for treatment-naive patients with the lower-level telaprevir-resistant variants V36M (green; n = 6), T54S (blue; n = 50), or R155K (red; n = 10) prior to treatment with telaprevir, peginterferon, and ribavirin (including patients enrolled to receive 8 weeks of telaprevir [19]) are indicated. For comparison, the median HCV RNA decline for patients who received a regimen of 12 weeks of telaprevir with response-guided peginterferon and ribavirin (black) in the phase 3 treatment, i.e., the Naive ADVANCE Study (19), is also shown. Results for all patients in the ADVANCE study receiving peginterferon and ribavirin (gray) and the subpopulation of patients receiving peginterferon and ribavirin that had undetectable HCV RNA at week 4 (RVR) are indicated by a dashed gray line.

DISCUSSION

In this large (n = 3,447), cross-study analysis of genotype 1 HCV, we determined the prevalence of patients with DAA-resistant variants by population sequencing, determined the probable genetic barriers, and assessed the effect of a suite of resistant variants on response to a telaprevir-based regimen.

Overall, the prevalence of patients with variants resistant to protease, NS5A, or polymerase inhibitors was 0 to 18%. In agreement with previous reports, the primary nucleo(s/t)ide inhibitor-resistant variants were not observed.

The prevalence of patients with telaprevir and boceprevir-resistant variants prior to treatment was low (0 to 3%). The variants observed were exclusively lower-level resistant variants for telaprevir and did not appear to affect response to a telaprevir-based regimen (5). Higher-level resistant variants were not observed, likely due to their impaired fitness (63).

The frequency of NS5A inhibitor-resistant variants observed was also low (0 to 6%) in patients; however, 19% (5/26) of these patients had a second NS5A inhibitor-resistant mutation. Variants conferring resistance to non-nucleoside polymerase inhibitors were observed in 0 to 11% (palm site), 0 to 1.5% (thumb site), and 0 to 18% (finger-loop site) of patients. The use of population sequence analysis in the present study allows the presence of a variant in >20% of the viral population to be determined, and variants that are a smaller minority of the viral population are not detected. However, the prevalence of variants above 20% of the viral population is consistent with data obtained using other sequencing methods (i.e., deep sequencing and ultradeep sequencing) (64, 65). Also of note, the vast majority of the DAA-resistant variants observed here were not polymorphic, allowing a survey of potential combinations of variants. Confirmation of resistant-variant linkage identified by population sequencing at nonpolymorphic sites with clonal sequence analysis has been reported (9).

Due to the decreased fitness of DAA-resistant variants compared to the wild type, it is interesting that they were observed as the predominant population in any patient. One possible explanation for this is that other potential compensatory variants may have evolved to increase the fitness of the DAA-resistant viral population; this is consistent with the observation of viral populations with multiple resistant variants in a single NS protein. Combinations of protease with either NS5A or NS5B inhibitor-resistant variants were observed in a small number of patients. While variants in multiple nonstructural domains have been observed in patients that fail multiple DAA containing regimens, the rarity of the combinations here suggests there were no significant naturally occurring compensatory interactions between the NS3 protease, NS5A, and NS5B polymerase inhibitor-resistant variants.

Effect of naturally occurring variants on treatment outcome.

The presence of preexisting NS5A or NS5B inhibitor-resistant variants did not affect the response to a telaprevir-based regimen, which is consistent with in vitro data showing a lack of cross-resistance between DAAs with different mechanisms of action as well as peginterferon and ribavirin.

The presence of preexisting lower-level telaprevir-inhibitor-resistant variants also did not preclude successful treatment with a telaprevir-based regimen in most patients. In treatment-naive patients the clinical outcomes in phase 3 telaprevir trials were similar for patients with or without telaprevir-resistant variants. The SVR rate of patients with telaprevir-resistant variants at baseline was 74% (39/53) compared to an SVR rate of 76% (634/837) in patients who did not have a variant at baseline (66). The overall SVR rates for treatment-naive and prior-peginterferon and ribavirin-relapser patients in phase 2 and 3 telaprevir trials were 65% for patients with telaprevir-resistant variants compared to 75% without. In the limited number patients with a prior nonresponse to peginterferon and ribavirin, the impact of telaprevir-resistant variants was also similar, with a 36% SVR in patients with telaprevir-resistant variants compared to 42% for patients with wild-type virus. The similar response in most patient populations is likely because the telaprevir levels achieved in patients were sufficient to at least partially inhibit the replication of these viral variants, and the combination with peginterferon and ribavirin allows successful clearance of these variants (5, 51). However, for the very limited number of prior nonresponder patients with specific variants such as NS3 R155K (n = 3) the SVR rate was 0%. Therefore, for patients receiving a DAA, interferon and ribavirin regimen, other factors such as the response to interferon and adherence to the regimen likely play a larger role in the clinical outcome. Still, as various investigational regimens with multiple DAA components advance, particularly regimens without interferon, the potential role of DAA-resistant variants prior to treatment should be evaluated since their clinical relevance could be different than in IFN-containing regimens.