Peptidomimetic Escape Mechanisms Arise via Genetic Diversity in the Ligand-Binding Site of the Hepatitis C Virus NS3/4A Serine Protease

Abstract

Background & Aims

It is a challenge to develop direct-acting antiviral agents (DAAs) that target the NS3/4A protease of hepatitis C virus (HCV) because resistant variants develop. Ketoamide compounds, designed to mimic the natural protease substrate, have been developed as inhibitors. However, clinical trials have revealed rapid selection of resistant mutants, most of which are considered to be pre-existing variants.

Methods

We identified residues near the ketoamide-binding site in X-ray structures of the genotype 1a protease, co-crystallized with boceprevir or a telaprevir-like ligand, and then identified variants at these positions in 219 genotype 1 sequences from a public database. We used side-chain modeling to assess the potential effects of these variants on the interaction between ketoamide and the protease, and compared these results with the phenotypic effects on ketoamide resistance, RNA replication capacity, and infectious virus yields in a cell culture model of infection.

Results

Thirteen natural binding-site variants with potential for ketoamide resistance were identified at 10 residues in the protease, near the ketoamide binding site. Rotamer analysis of amino acid side-chain conformations indicated that 2 variants (R155K and D168G) could affect binding of telaprevir more than boceprevir. Measurements of antiviral susceptibility in cell culture studies were consistent with this observation. Four variants (Q41H, I132V, R155K, and D168G) caused low-to-moderate levels of ketoamide resistance; 3 of these were highly fit (Q41H, I132V, and R155K).

Conclusions

Using a comprehensive sequence and structure-based analysis, we showed how natural variation in the HCV protease NS3/4A sequences might affect susceptibility to first-generation DAAs. These findings increase our understanding of the molecular basis of ketoamide resistance among naturally existing viral variants.

Until recently, the standard of care (SOC) for patients with chronic hepatitis C virus (HCV) infection has consisted of a combination of pegylated interferon-α plus ribavirin (Peg-IFN/RBV), administered for 24- to 48-weeks depending on the HCV genotype^1–3. The sustained virologic response (SVR) rate for this SOC has been only about 50% in patients infected with genotype 1 HCV, the most prevalent genotype in Europe and North America. The addition of a direct-acting antiviral agent (DAA) targeting the NS3/4A serine protease of HCV significantly improves the SVR rate, and two such drugs have recently been approved for clinical use in the United States. The ketoamide compounds boceprevir (SCH503034) and telaprevir (VX-950) were both designed to mimic the natural substrate of the protease^4–6. Clinical trials in treatment-naïve genotype 1-infected patients have revealed significant improvements in the kinetic of the virologic response with the addition of a DAA to the prior SOC, leading to improved SVR rates of up to 74%^7–9.

Unlike Peg-IFN/RBV, the selection of drug-resistant virus variants during treatment with protease inhibitors (PIs) is a major concern. According to recent calculations by Rong et al.¹⁰, most, if not all, potential drug-resistant viral variants pre-exist at low frequencies within the viral quasispecies population in untreated patients. The highly replicative nature of HCV infection, with approximately 10¹² new virions produced each day in the typical infected individual¹¹, coupled with the lack of proofreading activity in the RNA-dependent RNA polymerase, NS5B, results in the generation of every possible viral variant every day. Thus, each patient is infected with a viral quasispecies “cloud” comprised of genetically distinct but closely related viral genomes. In the absence of concomitant Peg-IFN/RBV therapy, drug-resistant viral variants are rapidly selected and may emerge at frequencies as high as 5–20% in the quasispecies of patients as early as the second day of treatment. Unless suppressed by concomitant PegIFN/RBV, these pre-existing resistant variants are likely to be selected with subsequent treatment failure¹⁰.

The NS3/4A protease plays an essential role in the HCV replication cycle by proteolytically processing non-structural proteins from the viral polyprotein downstream of the NS2–3 junction¹². The protease domain of NS3, comprising the amino-terminal third of the protein contains a catalytic triad, H57, D81 and S139, and an “oxyanion hole” at G137. It acts in concert with its cofactor, NS4A, which intercalates into its structure and is required for full enzymatic activity and proper subcellular localization. The carboxy-terminal two-thirds of NS3 comprises a DExD-box RNA helicase domain that is essential for productive viral infection¹³. NS3 thus appears to be a critical component of the macromolecular viral RNA replicase that directs the synthesis of new viral RNAs. Genetic evidence indicates that NS3 has an additional distinct function in assembly of virus particles^{14, 15}. Since viral RNA replication capacity and virus assembly are crucial determinants of viral fitness, mutations in NS3 that contribute to PI resistance can also profoundly influence virus fitness¹⁶. The probability of a resistant variant emerging from the quasispecies population during treatment with a DAA is determined not only by its degree of resistance, but also by its fitness. Many mutations associated with PI resistance negatively impact the replication of genotype 1a HCV RNA in cell culture, while some have additional effects on the production of infectious virus¹⁴. Compensatory second-site mutations may enhance the fitness of resistant viruses¹⁶, but current understanding of these is rudimentary.

Given the genetic diversity that exists among different HCV strains, it is possible that naturally occurring polymorphisms in the NS3/4A sequence could provide a priori resistance to DAAs, and thus negatively impact the success of future treatment regimens. Here, we have studied the variation in amino acid residues that neighbor ketoamide compounds in the ligand-binding site of the protease. We identified natural amino acid substitutions at these positions in NS3 among genotype 1a sequences deposited in a public database, and modeled their side-chain conformations to assess their potential impact on ketoamide binding. To corroborate these in silico predictions, we then introduced these amino acid substitutions into a cell culture-infectious genotype 1a virus (H77S.3)¹⁴ and determined their impact on both susceptibility to ketoamide PIs and replication fitness in a cell culture system.

MATERIALS AND METHODS

Details of the materials and methods can be found in the Supplementary Material.

In silico analysis

We used X-ray structures of the genotype 1a HCV NS3/4A protease from the Protein Databank RCSB PDB¹⁷ co-crystallized with boceprevir (PDB 2OC8) or a telaprevir-like ligand (TLL, PDB 2P59) to deduce sets of ketoamide-neighboring residues. We designated the P₄ to P₁ and P₁’ groups for ligands and their corresponding specificity pockets within the ligand-binding site, S₄ to S₁ and S₁’, according to the numbering scheme of Schechter and Berger¹⁸. We then analyzed 219 genotype 1a HCV NS3/4A sequences deposited in the European HCV database¹⁹, which contains sequences collected from around the world, to identify potential natural binding site variants (BSVs) at residues that neighbor the ketoamides within the structure of the protease. The side-chain conformations of these BSVs were modeled using IRECS²⁰ (details in Supplementary Material).

Cell culture and reagents

Details of the cells and reagents used in this study are provided in Supplementary Material.

Plasmids

pH77S.3 and pH77S.3/GLuc2A are molecular clones of the genotype 1a H77 strain of HCV. Synthetic RNA transcribed from these plasmids replicates in transfected Huh7 cells and produces infectious virus¹⁴. pH77S.3/GLuc2A RNA also produces secreted Gaussia luciferase (GLuc) reporter protein. Amino acid substitutions in BSVs expected to impact ketoamide binding were created in these plasmids by site-directed mutagenesis¹⁴.

Virus fitness and antiviral resistance

Genome-length RNA was transcribed from the mutated pH77S.3 and pH77S.3/GLuc2A plasmids in vitro, and studies to assess antiviral resistance and viral fitness carried out as described previously¹⁴.

RESULTS

To identify amino acid residues in NS3 that are in close proximity to ketomides in the ligand-binding site of the protease, we analyzed PDB structure 2OC8 in which boceprevir is co-crystallized with the NS3/4A protease²¹. Since no similar co-crystallized structure is publically available for telaprevir, we used PDB structure 2P59 in which the protease is co-crystallized with a telaprevir-like ligand (TLL) that has two small differences from telaprevir²². Its P₂ group and the P₄ capping group are slightly modified, with the P₄ providing a pyrrole NH for H-bond interactions with the protease (Supplementary Fig. S1)²³. We identified 20 residues interacting with or neighboring boceprevir and TLL in these X-ray structures (Table 1 and Supplementary Materials and Methods), and analyzed genotype 1a HCV sequences in the European HCV database to identify differences from the consensus sequence at these residues. The residues were relatively well conserved in 219 genotype 1a sequences from diverse geographic regions. However, we identified 13 different binding site variants (BSVs) involving 10 ketoamide-neighboring residues (Supplementary Table S1 and Fig. S2), 8 of which have not been described previously as PI resistance-associated variants. None of the patients from which these sequences were derived appear to have been treated previously with a DAA (see Supplementary Table S3). Most of these naturally occurring BSVs were single amino acid substitutions without changes in other ketoamide-neighboring residues. However, T42A and K136R were both present within a single sequence (EU677251). Two different BSVs were identified as substitutions at T42 (A/S), V55 (A/I) and D168 (G/E), while single BSVs were identified at the other 7 residues (Supplementary Table S1 and Fig. S2). We examined the first-neighbor residues of these BSVs in an effort to identify potential second-site variants in the BSV strains (see Supplementary Material and Methods, Fig. S3 and Table S2). While there were no additional substitutions at these first-neighbor residues in most BSVs, T54S was found in both V55I strains (EF407443 and EU781818) and I170V was found in a R155K strain (EU781805).

Table 1.

Ketoamide-neighboring residues in NS3¹.

Boceprevir		Telaprevir-like ligand (TLL)
RCSB PDB	UniProtKB	RCSB PDB	UniProtKB
2OC8 (Chain B)	P27958	2P59 (Chain B)	P27958
Q41	Q41	Q1067	Q41
T42	T422	T1068	T42
F43	F43	F1069	F43
V55	V55	-	-
H57	H57	H1083	H57
D81	D81	D1107	D81
R123	R123	R1149	R123
I132	I132	I1158	I132
L135	L135	L1161	L135
K136	K136	K1162	K136
G137	G137	G1163	G137
S138	S138	S1164	S138
S139	S139	S1165	S139
F154	F154	F1180	F154
R155	R155	R1181	R155
A156	A156	A1182	A156
A157	A157	A1183	A157
V158	V158	V1184	V158
C159	C159	C1185	C159
-	-	T1186	T160
D168	D168	D1194	D168

¹Residues are numbered relative to the PDB residue sequence and H77c reference sequences (UniProtKB P27958).

²Residues shown in italics do not directly interact with boceprevir or TLL, but are within 5.0 Å distance of the ligand (see Supplementary Materials and Methods).

Rotamer analysis of BSVs and expected impact on ketoamide binding

We modeled the energetically most favorable side-chain conformations for the genotype 1a consensus and BSV sequences (Fig. 1). Since V55 is buried in the protease domain²⁴, it was not amenable to rotamer analysis. Molecular dynamics simulation of variants at this position will be reported separately (Welsch et al., in revision).

Figure 1. Ligand-binding site of NS3/4A with ketoamides and BSVs.

(A) Surface representation from PDB structure 2OC8 with boceprevir shown in magenta. Side-chains at amino acid residues with BSVs are shown in stick format (light blue). (B) Similar representation of TLL within the PDB structure 2P59 with side-chains at amino acid residues with BSVs (dark blue). (C) Superposition of boceprevir and TLL.

Expected impact of Q41H and K136R at the ketoamide P₁’ position

Q41 is located adjacent to S₁’. The H41 variant represents a non-conservative change from the consensus sequence involving substitution of an uncharged polar side chain with a charged aromatic residue. Telaprevir possesses a cyclopropyl group at P₁’ that is oriented away from the Q41 side chain without noncovalent interactions. The conformation of Q41 suggests a potential H-bond interaction of its carboxamide chain, including the H-bond donor NH₂ and acceptor C=O, with the backbone of boceprevir and TLL (OE1—HN distance for Q41-ketoamide: 2.7 Å). However, the H41 BSV has no side-chain group allowing an H-bond interaction with the ligand (Fig. 2A). This suggests H41 will be associated with a minor decrease in binding affinity for boceprevir and telaprevir.

Figure 2. Rotamer analysis of Q41H and K136R.

Detail of PDB structure 2OC8 showing boceprevir (magenta) and the superposed TLL (yellow). The panels on the left show the side-chain conformation of the consensus residue, while those on the right show the predicted BSV conformation. (A) (left) The Q41 side-chain OH-group with predicted H-bond interaction (dotted line) with ketoamide backbone Cα, (right) this H-bond interaction is absent in H41. (B) Predicted R136 side-chain conformation and the TLL P₁’ cyclopropyl showing closer van der Waals contacts in (right) R136 than in (left) the consensus K136.

In contrast, K136R is a conservative change between polar residues, both with long, flexible and charged side chains at S₁’. The K136 side chain is not defined in the electron density map of PDB structure 2P59. As modeled, it has close contacts with the TLL P₁’ cyclopropyl group (Fig. 2B). Boceprevir possesses no comparable P₁’ group (Supplementary Fig. S1). We predict the R136 side-chain conformation is similar to the wild-type K136, but that it will wrap around the TLL cyclopropyl group providing an increasing number of van der Waals contacts (Fig. 2B). This suggests that the R136 variant may have stronger effects on telaprevir than on boceprevir binding. We expect hydrophilicity to play a role in ketoamide binding at S₁’, and the R136 side chain possesses more polar contacts and potentially binds more water molecules than the wild-type K136. Thus, the release of water molecules bound to the R136 side chain could increase systemic entropy and potentially add to the free binding energy and affinity for telaprevir, although decreased enthalpy may compensate for the effects on entropy. Since the R136 side chain does not have a tight H-bond or salt bridge interaction at its end, electrostatic attraction between the nonpolar P₁’ cyclopropyl group of telaprevir and the polar R136 side chain is unlikely. Overall, however, the impact of K136R on binding of telaprevir is not readily predicted by rotamer analysis alone.

Expected impact of I132V and F154Y on ketoamide P₁ and P₃ interactions

I132 is located adjacent to S₁ and S₃, and its substitution with valine in V132 represents a conservative change between hydrophobic residues. The consensus I132 side-chain conformation forms a hydrophobic S_1/3 interface with the ketoamide P₁ and P₃ groups. The I132 side-chain C_δ carbon points towards the ketoamide P₃ group (Fig. 3A). The variant V132 shows a similar side-chain conformation but without a C_δ carbon. This suggests a loss of van der Waals contacts for V132, resulting in a minor reduction in binding affinity for boceprevir and telaprevir. F154Y is also a relatively conservative change, in this case from nonpolar towards polar among aromatic residues. This residue is close to the ketoamide P₁ and P₃ groups at the bottom of the S₁ pocket. The wild-type F154 and the predicted variant Y154 side chains are oriented towards the ketoamide P₁ group, potentially influencing boceprevir and telaprevir binding (Fig. 3B). The Y154 side-chain has an OH group that is not present in the wild-type F154. Ketoamides do not offer an opportunity for H-bond interactions at P₁, but this OH group may provide for alternative binding of a water molecule. The polar, hydrophilic nature of the Y154 side chain reduces the hydrophobic properties of the S₁ pocket, and this variant is expected to cause a significant reduction in binding affinity for boceprevir and telaprevir.

Figure 3. Rotamer analysis of I132V and F154T.

(A) (left) The I132 side chain provides a hydrophobic S_1/3 interface for the ketoamide P₁ and P₃ moieties. The I132 C_δ carbon is oriented towards the ketoamide P₃, while (right) V132 lacks a C_δ carbon. (B) (left) F154 neighbors S₁ underneath the ketoamide, while (right) Y154 is predicted to assume a similar side-chain conformation but with an additional OH-group.

Expected impact of R155K, D168E and D168G on ketoamide P₂ and P₄ interactions

R155K is a conservative change between positively-charged residues within the S₂ pocket The R155 side chain is predicted to participate in a pattern of noncovalent interactions involving its neighboring residues, R123 and D168, at S₄ (the O-H distance for D168-R155 is 2.6 Å; for R123-D168, it is 2.5 Å) (Fig. 4). The combination of H-bond and electrostatic interactions is predicted to result in a particularly strong noncovalent salt ‘bridge’ interaction. Polar interactions between the D168, R123 and R155 side chains contribute to the strength of this noncovalent interaction network. The D168 side chain is predicted to be tightly fixed, allowing no other H-bond interaction and contributing to a nonpolar S_2/4 interaction interface for the P₂ and P₄ groups in boceprevir and telaprevir. The variant K155 disrupts this nonpolar S_2/4 interface by leaving the negatively-charged D168 unbound, and is expected to reduce binding affinity for boceprevir and telaprevir. A stronger effect is expected for telaprevir than for boceprevir, since the telaprevir P₂ cyclopental-proline is larger than the respective boceprevir isopropyl-proline (Supplementary Fig. S1). Compared with R155, the K155 side-chain is also predicted to be shifted slightly away from the ketoamide P₂ group. This shift is likely to cause a loss in van der Waals contacts with the ketoamide P₂ group, and may reduce binding affinity for telaprevir. The effects on polarity and loss of van der Waals contacts suggest that there will be a significant decrease in binding affinity for both ketoamides, but a larger impact on telaprevir.

Figure 4. Rotamer analysis of R155K.

(left) The H-bond pattern (dotted lines) in the R155 structure with D168 serving as a nonpolar S₄ contact interface for the ketoamide P₄. (right) K155 disables this H-bond pattern, leaving D168 unbound with a polar OH-group at S₄. The K155 side-chain conformation is also predicted to be shifted slightly away from the ketoamide P₂.

Two variants were observed at the D168 position that contributes to the S₄ pocket in the protease. D168G is a non-conservative change from aliphatic and polar to a smaller, nonpolar side chain. The wild-type D168 side-chain conformation is oriented towards the ketoamide P₄ group, which is smaller in boceprevir than telaprevir. The extended P₄ group in telaprevir (Fig. 1) points away from D168. Thus, the D168 variant is expected to have only minor effects on ketoamide binding. G168 is expected to have effects on the nonpolar S_2/4 interface similar to K155, since it offers no polar side-chain interaction partner for R155. It is predicted to change the polarity of the S_2/4 interface and expected to cause a minor decrease in binding affinity for both ketoamides. D168E is a conservative change as both side-chains are negatively charged and aliphatic. The side-chain conformation is predicted to be similar to the wild-type D168, and to preserve S_2/4 interface polarity. D168E is not expected to impact ketoamide binding.

Additional BSVs without expected impact on ketoamide binding

T42A is located at the periphery of the S₁’ pocket, and is a non-conservative change from aliphatic, polar and negative towards aliphatic and hydrophobic. A side-chain OH group is lost in the A42 variant, but preserved in the S42 variant at this position, allowing for a potential H-bond interaction. The wild-type T42 and variant A42 and S42 side-chain conformations are predicted to be similar. No ketoamide H-bond donor or acceptor group is found in close proximity to T42 or S42. Due to its relative distance from the ketoamide, the polarity change in T42A is predicted not to have a significant effect on ketoamide binding. Thus, the T42A/S variants are expected to have no impact on ketoamide binding. F43S is a non-conservative change from aromatic and nonpolar towards aliphatic and polar, and is found at the bottom of S₁’. No change is predicted in the side-chain conformation. There is an additional OH group in the variant, but since there is no ketoamide H-bond donor or acceptor group in close proximity, this by itself is unlikely to have any impact on ketoamide binding. Nevertheless, F43S may cause ketoamide resistance since the F43 aromatic ring directly participates in the formation of the S₁’ pocket, which impacts binding of the ketoamide P₁’ group²⁵. Substitutions at F43 have been shown previously to cause resistance to ketoamide compounds²⁵. T160A is a non-conservative change from polar towards nonpolar and hydrophobic, located at a distance from S₄ near the boundary of an extended S₅ pocket. T160 interacts directly with the extended P₄ group in TLL in the PDB structure 2P59, but has no noncovalent interactions with boceprevir in PDB structure 2OC8. The variant A160 side-chain is not expected to influence ketoamide binding.

Replication capacity and infectious virus yield from RNAs containing BSVs

Those BSVs for which the rotamer analysis suggested a possible impact on ketoamide binding (Q41H, I132V, F154Y, R155K and D168G) were selected for phenotypic characterization. We similarly tested K136R, for which the rotamer analysis provided no clear predictions. The amino acid substitutions were created within the background of the genotype 1a H77S.3 genome, and their impact on replication of the viral RNA and production of infectious virus determined in RNA-transfected cells.

(1) RNA replication capacity

The replication capacity of H77S.3/GLuc2A RNA mutants with BSV substitutions in NS3 was assessed by measuring GLuc activities in supernatant media collected at 24h intervals following transfection of synthetic RNA, as described previously¹⁴. Results were normalized to the activity present at 6h after transfection, as this represents GLuc expressed directly by the transfected input RNA. The RNA replication capacity of the Q41H variant was similar to the parental H77S.3/GLuc2A RNA, while I132V and K136R were minimally impaired (Fig. 5A). In contrast, as shown previously¹⁴, the replication of R155K was moderately impaired while D168G was severely handicapped for RNA replication. The maximum RNA replication capacity observed was with Q41H (89% of parental RNA) while the lowest was with D168G (5.5%). The F154Y substitution had a lethal effect on RNA replication in the H77S.3 background, suggesting that viruses containing this sequence variant (and possibly also G168) possess one or more compensating changes at other amino acid positions (see Discussion). None of the BSVs demonstrated enhanced RNA replication capacity compared to the parental H77S.3/GLuc2A RNA (Fig. 5B).

Figure 5. Replication capacity and infectious virus yield of H77S.3 RNAs with BSV substitutions in NS3.

(A) RNA replication capacity reflected of H77S.3/GLuc2A BSV into Huh7.5 cells. Results are normalized to the 6h GLuc activity, and represent the mean of triplicate samples. (B) Comparison of RNA replication capacity (lightly shaded bars) and infectious virus yields (dark shaded bars). Both are normalized to those obtained with the relevant parental RNA (100%). Data represent the mean ± SD from at least 3 independent experiments

(2) Infectious virus yield

We also assessed the impact of these BSVs when placed within the background of H77S.3 RNA, which lacks the GLuc2A-coding sequence and produces infectious virus as described previously¹⁴. Cell culture supernatant fluids were collected 72h after transfection of the RNA, and subsequently inoculated onto naïve cells, with foci of infected cells detected by immunofluorescence 72h later. Each of the BSVs tested produced infectious virus yields in the range of that expected from their RNA replication capacity, although the low replication capacity of the D168G variant precluded a careful analysis on production of infectious virus (Fig. 5). Thus, none of these amino acid substitutions were documented to directly impair infectious virus assembly or release, as described previously for some resistance-associated NS3 variants¹⁴. Reductions in the fitness of these particular BSVs are confined primarily to defects in viral RNA replication.

Ketoamide resistance of BSVs

We measured the antiviral activities (EC₅₀) of boceprevir and telaprevir against each of the H77S.3/GLuc2A mutants by determining the concentration of each compound required to cause a 50% reduction in RNA replication (GLuc expression by RNA-transfected cells). Resistance testing could not be performed for the F154Y variant because it was not competent for replication.

Boceprevir demonstrated antiviral activity against each of the BSVs (Table 2). The EC₅₀ value against the parental H77S.3/GLuc2A was 870 ± 48 nM. The maximum fold-change in the EC₅₀ for any of the BSVs was 2.1 for R155K¹⁴. Only R155K showed significant, albeit low-level resistance against boceprevir. Telaprevir showed greater molar activity than boceprevir, with an EC₅₀ against the H77S.3/GLuc2A of 120 ± 10 nM. As with boceprevir, telaprevir was active against each of the BSVs, but low- to medium-level resistance was evident with Q41H, I132V, R155K and D168G (Table 2). The maximum fold-change in the EC₅₀ was 8.8 for R155K. As expected from the in silico analysis, the range of fold-changes in EC₅₀ was broader for telaprevir than boceprevir. In general, these changes were in good agreement with the impact of these BSVs on ketoamide binding predicted from the rotomer analysis, except for K136R which was difficult to predict and showed greater antiviral activity than anticipated against both ketoamide compounds (Table 2).

Table 2.

Predicted and measured impact of BSVs on antiviral activity of ketoamides¹.

Genotype 1a HCV	IRECS		EC50 [nM]		Fold-change in EC50
	Boceprevir	TLL	Boceprevir	Telaprevir	Boceprevir	Telaprevir
H77S.3/GLuc2A	-	-	870 (±48)	120 (±10)	1.0	1.0
Q41H	MiN2	MiN	1040 (±60)	410 (±10)	1.2	3.5
T42A	NoI	NoI	-	-	-	-
T42S	NoI	NoI	-	-	-	-
F43S	NoI	NoI	-	-	-	-
I132V	MiN	MiN	980 (±60)	280 (±40)	1.1	2.4
K136R	NoI	NP	440 (±19)	100 (±10)	0.5	0.9
F154Y	SN	SN	ND3	ND	ND	ND
R155K	MeN	SN	1830 (±240)	1010 (±240)	2.1	8.8
T160A	NoI	NoI	-	-	-	-
D168G	MiN	MeN	490 (±50)	260 (±30)	0.6	2.2
D168E	NoI	NoI	-	-	-	-

¹The table compares the in silico predictions of the impact of BSVs on ketoamide binding from the IRECS²⁰ analysis, and corresponding EC₅₀ values determined from H77S.3/GLuc2A-transfected cell cultures. Results shown represent the mean ± SD and fold changes (FC) compared to wild-type.

²Predicted impact: NoI, no impact; MiN, minor negative impact, change only in a single ketoamide-BSV interaction (H-bond, van der Waals) or in polarity; MeN, moderate negative impact, combined changes in ketoamide-BSV interaction and polarity; SN, strong negative impact, same as for MeN but at particularly close ketoamide-BSV interaction sites; NP, not predictable;

³ND, not determined.

DISCUSSION

Mathematical arguments suggest that every possible drug-resistant viral variant is likely to pre-exist at a low frequency in the replicating viral quasispecies population of the typical HCV-infected patient¹⁰. Whether this is actually the case, and at what frequency such variants actually exist, may never be formally demonstrated due to technical difficulties. In this study, we analyzed the natural variation present among ketoamide-neighboring residues in 219 genotype 1a HCV sequences collected from geographically diverse sites and deposited in a public database. We cannot exclude the possibility that some of the BSVs we identified in this set of sequences may represent variants that were present at low frequency in their source patient, or even unrecognized sequencing errors. However, it is likely that they represent true variants present within the dominant quasispecies of the patients from which these sequences were derived, since multiple BSVs were identified at some residues (T42, V55, and D168) (Supplementary Fig. S2), while others (H41, A42, A55, I44, and K155) were present in more than one sequence. We identified BSVs at one or more ketoamide-neighboring residues in 17 of 219 (7.8%) genotype 1a sequences. Importantly, 8 of these variants (Q41H, T42A, T42S, V55I, I132V, K136R, F154Y, and T160A) have not been identified, to our knowledge, in previous in vivo or in vitro studies of ketoamide resistance.

Although they are both linear ketoamide compounds, boceprevir and telaprevir have distinct structural features (Supplementary Fig. S1). Telaprevir possesses an extended P₄ capping group and a P₁’ cyclopropyl group. The P₂ group is different from the isopropyl-proline in boceprevir, which is smaller than the telaprevir P₂ cyclopental-proline²³. These structural differences likely contribute to the lower EC₅₀ of telaprevir against genotype 1a HCV in cell culture (Table 2), but could also pose a higher risk for telaprevir resistance among BSVs. Rotamer analysis predicted that two BSVs (R155K and D168G) would exert a greater negative effect on the binding of telaprevir than boceprevir. The impact of the K136R substitution proved difficult to predict on the basis of rotamer analysis alone, however, and subsequent tests in cell culture demonstrated that it imposes no resistance against either ketoamide (Table 2).

On the other hand, the R155K and D168G substitutions led to a 2- to 4-fold greater increase in the EC₅₀ of teleprevir compared with boceprevir, and almost a 9-fold increase in the teleprevir EC₅₀. This was consistent with predictions from the rotamer analysis, which also agreed with previous crystallographic studies²⁶. Whether such changes in the EC₅₀ result in clinical resistance would be dependent upon the potency of a drug and the drug exposure achieved in a typical patient. A 2-fold change might not be clinically relevant, but a 9-fold increase such as that found with R155K is likely to be significant. Importantly, R155K has been associated previously with resistance to both ketoamide and macrocyclic compounds^{23, 26}.

Three BSVs (Q41H, I132V and F154Y) were predicted to interact with ketoamide structural features common to both telaprevir and boceprevir, and thus to influence the binding of these compounds equally. Although Q41H was expected to cause only a minor decrease in affinity for both ketoamides, in vitro assays revealed a 3.5-fold increase in the EC₅₀ for teleprevir versus 1.2 for boceprevir. Its greater impact on telaprevir potency is likely related to its P₁’ cyclopropyl group and induced structural changes in the corresponding S_1’ pocket. Such changes were not detectable using rotamer analysis, but nonetheless probably influenced the binding of telaprevir. A similar difference in the magnitude of the change in EC₅₀ was observed for the I132V variant (1.1 fold-change for boceprevir versus 2.4 for teleprevir). This could result from a loss of van der Waals contacts with the ketoamide P₃ group, which differs slightly in its orientation in the co-crystallized boceprevir and TLLstructures (see Fig. 1). The F154Y substitution was lethal for RNA replication when placed in the background of the H77S.3 virus, and thus we could not measure antiviral activity against it.

Viral fitness coupled with the degree of resistance conferred by a BSV are likely to be the major determinants driving selection of a variant from within the viral quasispecies during therapy. RNA replication capacity is one measure of the fitness of the virus, and this is dependent on proper processing of the polyprotein by NS3/4A. Ketoamides mimic the natural substrate of the protease at the site of NS3-NS4A scission, and it is likely that the negative influence of BSVs on RNA replication (Fig. 5A) reflects altered recognition of the polyprotein substrate related to structural changes similar to those leading to drug resistance. Despite this, there is no strong correlation between the degree of PI resistance and the impact on RNA replication¹⁴. This is reflected in the marked deficit in replication demonstrated by the D168G substitution (Fig. 1A), which confers only a minimal increase in the teleprevir EC₅₀ (Table 2). There is no obvious structural or molecular explanation for this difference, and it is not possible to exclude the possibility that G168 might provide for more robust RNA replication when placed in the context of a different virus sequence.

In contrast F154Y was lethal for replication in the context of the H77S.3 virus. This substitution occurs at a central position within the ligand-binding site at the bottom of the S₁ pocket. A recent crystallographic study found the F154 aromatic ring to directly contact the substrate P₁ side-chain²⁷. Thus, the complete loss of RNA replication observed with F154Y could be due to either intrusion of the Y154 side-chain into the space normally occupied by the polyprotein substrate, or a polarity shift within S₁ due to the Y154 OH group. Nonetheless, the presence of Y154 in the database suggests that it is capable of functioning in an alternative sequence context. This highlights a limitation of the technical approach we have taken here to study the phenotypic effect of BSVs, as second-site substitutions in the same strain might compensate for defects fitness. In fact, the sequence of the F154Y BSV (EU677253) contains two additional substitutions in NS3 that differ from the genotype 1a consensus: R11G and H110R (Supplementary Fig. S2). H110R is in close structural proximity to F154Y and the natural substrate of the protease (Supplementary Fig. S3). While it might be preferable in investigating BSVs to engineer swaps of the complete NS3 sequence into the background of a replication-competent clone, it is not clear how often this approach would fail due to sequence incompatibilities between the “donor” and “recipient” viruses. NS3 appears to interact with several other viral proteins²⁸.

Our findings provide a molecular basis for ketoamide resistance among BSVs that exist naturally as dominant quasispecies in some patients prior to treatment with DAAs (Supplementary Table S3). Such natural variants may be of limited clinical significance at present as they are likely to be suppressed by Peg-IFN/RBV in current SOC regimens, but they can be expected to be of substantial importance to the outcome of future interferon-sparing, all-DAA combination therapies. These variants might also affect future generations of inhibitors depending upon their chemical structures. Knowledge on the natural variability in structures targeted by antivirals, as presented in this study, may help guide the development of future generation PIs.

Abbreviations

BSV

binding-site variant

DAA

direct-acting antiviral agent

EC

effective concentration

FFU

single infectious focus-forming unit

GLuc

Gaussia Luciferase

HCV

hepatitis C virus

NS

non structural protein

PDB

Protein Databank

PI

protease inhibitor

RNA

ribonucleic acid

SOC

standard-of-care

SVR

sustained virological response

TLL

telaprevir-like ligand