CHAPTER 1: Discovery and Clinical Validation of HCV Inhibitors Targeting the NS5A Protein

Significant effort has been invested in elucidating the exact role and function of the NS5A protein in the hepatitis C virus (HCV) replication cycle. Although, unlike the NS3 and NS5B proteins, no enzymatic function has been identified thus far for NS5A, it has become apparent that this protein plays a diverse and critical set of roles both in the replication of the virus and in the mediation of host–virus interactions. Despite its multifunctional role, the lack of a well‐characterized function coupled with the limited availability of structural information, compared with the NS3 protease and NS5B polymerase, initially made the NS5A protein a less compelling target for therapeutic intervention. That changed, however, with the validation of NS5A as a clinically relevant target by daclatasvir (1), where single doses effected pronounced and rapid declines in viral RNA in HCV‐infected subjects (Figure 1.1).¹  Highlights of the biochemical pharmacology of the NS5A protein, along with the discovery, the mode of action and the clinical characterization of a potent class of NS5A inhibitors, are discussed in this chapter.

The HCV RNA genome encodes a ∼3000 amino acid polypeptide that is processed by both viral and cellular proteases into structural proteins (Core, E1 and E2), an ion channel (p7) and non‐structural proteins (NS2, NS3, NS4A, NS4B, NS5A and NS5B).²  The non‐structural proteins are responsible for replication of the viral genome and for the assembly of the viral particle from the structural proteins, with the assistance of host factors. HCV NS5A is a 447 residue peptide that is comprised of three domains, which are interlinked with short fragments designated as low‐complexity sequences (LCSs) (Figure 1.2).³  Various studies have demonstrated that NS5A is an RNA‐binding protein, although the specific elements of the protein that are establishing the biologically relevant interactions with ribonucleic acid still need to be identified.^4–6  For example, one study has indicated that all three domains of NS5A exhibit RNA‐binding properties, albeit with differential affinities, whereas a different study showed that the Domain I/LCS I peptide fragment exhibited RNA‐binding affinity that is comparable to that of the full‐length NS5A protein, supported by the observation that the RNA‐binding property of NS5A is abolished if Domain I is deleted.^4,5  Whatever their specific RNA‐binding properties may be, all three domains contribute to genome replication, while Domain III plays a key role in viral particle assembly.^7,8  In addition, all three domains play a diverse set of regulatory roles in modulating host–virus interactions so as to facilitate the establishment of an environment conducive to successful viral replication.

Domain I of NS5A is a Zn²⁺‐binding moiety with an amphipathic α‐helix at its N‐terminal that is believed to anchor the protein to cellular membranes. X‐ray structural studies by two independent groups on similar amino acid constructs of Domain I, both lacking the amphipathic α‐helix motif, revealed that the protein crystallizes as a homodimer (Figure 1.3).^9,10  Interestingly, although the monomeric units in the two X‐ray studies were highly structurally conserved, their modes of dimerization were different and involved non‐overlapping contact surfaces. A positively charged groove created by the dimeric interface of the X‐ray structure reported by Rice’s group, which had the appropriate dimensions to support the hypothesis that it could be an RNA‐binding site, is fully exposed in Love’s X‐ray structure. The reason for the differing modes of dimerization and how well either one may reflect a biochemically relevant structure of the NS5A protein, especially since about half of the protein is missing from the structural analyses, is not apparent at this stage. Some have postulated that the two dimeric modes may represent snapshots of an oligomeric state, the functional significance of which has yet to be revealed. In general agreement with the X‐ray structural findings, a glutathione‐S‐transferase (GST)‐tagged NS5A Domain I was able to pull down a His‐tagged NS5A protein, presumably through a dimeric interaction, whereas this was not possible with GST alone. This interaction does not appear to be mediated by the presence of nucleic acids and yet, interestingly, there is a similarity between the minimal NS5A fragment required to effect this pull‐down and the minimal fragment required to maintain the RNA‐binding affinity of the full‐length NS5A protein. It is also noteworthy that the minimal peptide fragment required to effect the dimerization in the pull‐down study was longer than the peptide constructs used in the X‐ray studies (amino acids 1–240 versus 25–215 for the Rice dimer and 33–202 for the Love dimer), and although the reason for this disparity is not apparent, it could be a result of the distinct physical states that the two studies are dealing with and of differences in experimental parameters, such as protein concentration.¹¹ 

In another study, a glutaraldehyde cross‐linking experiment demonstrated that either Domain I or the full‐length version of NS5A (but not Domain II–III) dimerize in solution, that the dimer is in equilibrium with the monomer and that the presence of uracil‐rich RNA, which is known to bind to NS5A, shifts the equilibrium in favor of the dimer.⁵  Interestingly, in the same study, NS5A–RNA cross‐linking followed by the mapping of the amino acids involved in the cross‐linking on to either of the two X‐ray structures indicated that, for the Rice dimer, the amino acids decorate the positively charged groove of the protein and not the similarly charged back side of the dimer, whereas for the Love dimer, a ribbon pattern surrounding the structure is observed. It was claimed that this cross‐linking result is more consistent with Rice’s dimer. Moreover, others have hypothesized that the groove in the Rice dimer may serve as an ‘RNA railway system’ that connects different functional states that the RNA has to traverse, along with providing a role of protecting the viral RNA from cellular factors that degrade exogenous RNA.¹²  Whatever the case may be, the fact that highly potent NS5A inhibitors with resistance mutations that map to Domain I constitute a dimeric pharmacophore that complements the symmetrical features revealed by the X‐ray studies and supported by biochemical studies, is unlikely to be coincidental (see below).

Unlike Domain I, Domains II and III are disordered proteins that lack secondary structural elements.^13,14  It is hypothesized that disordered proteins have an extended surface area that promotes simultaneous interactions with multiple proteins and/or an interaction with RNA, which could be a reflection of the multifunctional nature of these domains, the details of which still need to be delineated.¹⁵ 

HCV replicons, cell‐based assay systems that support the autonomous replication of the subgenomic and genomic HCV, have played a central role in the HCV drug discovery field since the introduction of the first genotype 1b (G‐1b) system in 1999, most notably in creating opportunities to exploit the potential of viral targets devoid of enzymatic functions.^16,17  As part of a campaign directed at identifying inhibitors of HCV that act by novel mechanisms to disrupt replication, scientists at Bristol‐Myers Squibb (BMS) devised a unique, dual‐replicon assay system that was used to conduct a phenotype‐based, high‐throughput screening (HTS) campaign.¹⁸  Specifically, this assay system utilized a mixture of a G‐1b HCV replicon and a replicon of a closely related virus, bovine viral diarrhea virus (BVDV), in the same well. The two replicons had the same Huh‐7 cellular background but orthogonal activity reporters – a FRET assay based on NS3 protease activity for HCV and a luciferase expression assay for BDVD. In addition, cell toxicity was assessed in the same well using a standard Alamar Blue assay. It is noteworthy that since a luciferase enzyme assay is more sensitive than a FRET assay, this reporter combination placed a stringent criterion for the identification of HCV‐specific inhibitors. The BMS compound collection was screened with this dual replicon assay system and initial hits that had either cytotoxic properties or poor HCV specificity, as reflected by a <10‐fold potency spread between HCV and BVDV inhibitory activities, were discarded. Counter‐screening of the resultant hit set with NS3 protease, NS3 helicase and NS5B polymerase enzymatic assays afforded a thiazolidinone chemotype, exemplified by carbamate 2, as a novel class of HCV inhibitor (Figure 1.4). It is noteworthy that over one million compounds were assayed in the HTS effort and only this single chemotype class met the stringent screening criteria to qualify as a suitable lead structure. Carbamate 2 exhibited moderate HCV potency (G‐1b EC₅₀=0.58 µM), very good HCV specificity (BVDV EC₅₀ >50 µM) and a CC₅₀ of >100 µM.^19,20 

In order to identify the HCV protein that carbamate 2 might be targeting, passage of a G‐1b replicon system through increasing concentrations of the compound resulted in a resistant phenotype that was >10‐fold less sensitive to the inhibitor. After confirming that the mutation that caused the resistance phenotype was associated with viral RNA and not cellular RNA, sequence analysis of viral RNA from resistant cell lines was conducted and two dominant mutations were identified in the Domain I region of NS5A (Y93H and Y93C). Either mutation, when introduced individually into a G‐1b replicon, was sufficient to confer the observed resistant phenotype and no cross‐resistance was observed with inhibitors targeting alternative HCV mechanisms. This resistance analysis was the first indication that the thiazolidinone chemotype might be engaging the NS5A protein.

Preliminary structure–activity relationship (SAR) studies directed at establishing the fidelity of the lead revealed that there was a preference for the S stereochemistry at the amino acid moiety and that changing the benzyl carbamate to phenylacetamide, as in amide 3, effected a ∼100‐fold potency enhancement, an SAR observation that was recapitulated in the analogous proline series (see amides 4 and 5). The SAR survey of the iminothiazolidinone region of the lead molecule revealed that variation of the substituent pattern also modulated potency; a >10‐fold dynamic potency range that was dependent on structure was noted. However, the patterns of SAR associated with this region were less discrete than that of the amino acid moiety. Resistance selection with amide 3 yielded additional mutations in NS5A Domain I (L31V and Q54L) that resulted in a 9–60‐fold potency loss and were cross‐resistant to amide 2, suggesting commonality of the inhibitory mechanism between these two molecules, despite the difference in their resistance mutations.

At this juncture, it became apparent that this thiazolidinone chemotype was exhibiting chemical instability in certain organic solvents and in the replicon medium.²¹  Careful analysis of degradation products revealed that when 3 is stored in dimethyl sulfoxide (DMSO) under ambient conditions, it undergoes an oxidative rearrangement to afford the thiohydantoin 8, which was inactive in the replicon assay, EC₅₀ >20 µM (Scheme 1.1). Incubation of 3 in the replicon assay medium initially afforded thiohydantoin 8, which degraded to thiourea 9, which also lacked replicon inhibitory activity. A critical and enlightening experiment in which 3 was pre‐incubated in the assay medium until complete degradation had occurred, followed by assessment in the replicon, revealed that the HCV inhibitory activity was maintained, clearly indicating that some chemical entity other than the parental analog was likely responsible for the observed effect. A careful HPLC biofractionation study conducted on 3 after incubation in assay medium coupled with detailed spectroscopic analyses of the degradation products revealed the presence of two dimeric derivatives (see 14 in Scheme 1.2), both of which demonstrated inhibitory activity in the G‐1b replicon with EC₅₀s of 0.6 and 43 nM. Although the precise stereochemical relationship between these two dimers has not been established, the more potent dimer converts to the weaker dimer when heated at 55 °C in CD₃CN, which is suggestive of either a rotameric or a stereoisomeric relationship. It is hypothesized that the formation of the dimeric species from 3 arises from its susceptibility to form a captodative radical (11) in the presence of a radical initiator such as molecular oxygen, which has a diradical ground state. Since 11 is a stabilized radical, it persists such that it can undergo either dimerization to afford 14 or combine with molecular oxygen to afford the peroxy intermediate 12, which would be susceptible to reduction to alcohol 6 followed by rearrangement to afford thiohydantoin 8 via ketoamide 7.²²  Transfer of the ketoacyl moiety of intermediate 7 to a nucleophilic species in the replicon medium would afford thiourea 9. Although a simple hydrolysis of ketoamide 7 in assay medium is also possible, the byproduct of such a hydrolytic process, keto acid 10, was not identified. It is noteworthy that acetate 13, which can be prepared from 3 in 78% yield by oxidation with Mn(OAc)₃–Cu(OAc)₂–AcOH, afforded thiohydantoin 8 when treated with MeOH–K₂CO₃, providing supportive evidence for a key step of the proposed mechanism.²³ 

Although the identification of the dimeric derivatives represented marked progress for the medicinal chemistry effort, optimizing these architecturally complex leads to a drug candidate appeared to be a challenging task, given that their physical properties fall far outside conventional drug space.^24–26  However, based on insights gleaned from the preliminary SAR investigation, a significant simplification of the dimeric species was achieved when the key pharmacophoric elements were successfully captured in the bibenzyl 15, which exhibited a G‐1b EC₅₀ of 30 nM, potency that was improved further with the structurally more rigid stilbene analog 16, which displayed an EC₅₀ of 0.086 nM in the G‐1b replicon assay.²⁰  This new lead molecule was relatively stable when incubated in replicon medium for the length of the assay period and exhibited a resistance profile that mirrored that of 3, supportive of a similar mode of inhibitory effect and confirming that this molecule contains the key pharmacophore within dimers 14. With its impressive potency and simplified structure compared with 14, stilbene 16 served as the starting point for the next phase of the medicinal chemistry campaign. This enterprise focused on expanding genotype coverage, since the EC₅₀ of 16 in a G‐1a replicon was >10 µM, and optimizing ADME properties. The effort involved significant chemotype evolution based on the application of bioisostere concepts and ultimately culminated in the discovery of the highly potent, first‐in‐class NS5A replication complex inhibitor daclatasvir (1).¹  Daclatasvir inhibits G‐1b and G‐1a replicons with EC₅₀s of 0.009 and 0.05 nM, respectively. In addition, it inhibits G‐2a to G‐5a replicons with EC₅₀s ranging from 0.033 to 0.146 nM. This unprecedented in vitro potency spectrum established a new benchmark for the HCV field.

A similar cell‐based screening of compound libraries conducted by scientists at Arrow Pharmaceuticals (subsequently acquired by AstraZeneca) led to the identification of two distinct hits (17 and 19) with HCV inhibitory activity that also appeared to target the NS5A protein and which were optimized to the two clinical candidates AZD‐2836 (18) and AZD‐7295 (20) (Figure 1.5).²⁷  Interestingly, although resistance associated with the quinazoline series mapped to the NS5A protein, along with some accompanying mutations in the NS4B and NS5B regions, reverse genetic engineering of the mutations into a G‐1b replicon, either alone or in combination, failed to recapitulate the resistant phenotype. On the other hand, the biphenyl carboxamide series afforded mutations in Domain I, the Y93H/C change being the hallmark and in the C‐terminal region of the NS5A protein, for which additional details were not provided.

The high in vitro inhibitory potency associated with 1 and its clinical validation of NS5A as a target for therapeutic intervention in HCV infection have generated considerable interest in this class of HCV inhibitor. Over the past 4 years, more than 100 patent applications have been published claiming various NS5A inhibitors, the majority of which are based on structural variation of the dimeric pharmacophore element pioneered by 1.²⁸  Comprehensive and insightful overviews of the NS5A patent literature that provide distinct perspectives have been published, and in the next section highlights of more recent developments in the field are provided.^27,29,30 

A dimeric pharmacophore that does not necessarily embrace chemical symmetry is a common theme throughout the majority of the published patent applications. Each pharmacophore unit typically contains a spacer element that projects hydrogen bond‐donating and ‐accepting properties attached to a pyrrolidine‐like fragment that is derivatized with an amino acid moiety (Figure 1.6). Most of the molecules disclosed maintain some variation of either a methyl carbamate of an alkyl‐ or an arylglycine or N,N‐dialkyl derivatives of an arylglycine for the amino acid fragment, and have primarily focused on modifications to the central core and the pyrrolidine regions.

The survey of central core elements has largely been directed towards uncovering alternate scaffolds that maintain the topological disposition of the key peripheral pharmacophoric moieties of 1 (Figure 1.7). Conceptually, the least disruptive strategy involves utilizing the biphenyl core and examining the effect of substitution at every position of the biphenyl moiety, including the installation of bridging elements, as exemplified by compounds 21–23.^31–33  In a case where the bridging element is a single bond, as in 23, the point of attachment of one of the imidazole moieties is changed to a meta position, presumably to reestablish a more linear disposition of the peripheral entities.³⁴ 

Replacement of one of the phenyl groups with a bicylooctane group, as in compound 24 and central scaffold elongations (see 25, 26 and 29 in Figure 1.7) resulted in compounds claimed to exhibit sub‐nanomolar inhibitory potencies in a G‐1b replicon.^35–38  Numerous combinations of bridging and elongation strategies that have resulted in tri‐ and tetracyclic scaffolds claiming to possess potent G‐1b inhibitory activity have also been reported (see 28).³⁹  Scaffolds with increased flexibility (see 27) suffered a loss in inhibitory activity.⁴⁰ 

Another strategy explored the utility of benzimidazole and its aza variants as bioisosteric replacements for the aryl imidazole moiety (Figure 1.8). Here again, analogs with a wide range of core lengths have exhibited G‐1b EC₅₀s of <1 nM.^33,41–43  Although detailed data are not available to make a comparative assessment regarding inhibitory activity towards other genotypes, it is noteworthy that 33b was reported to exhibit G‐1a and G‐1b inhibitory potencies of 50 and 3 pM, respectively, along with 23% oral bioavailability in monkeys when administered as a dihydrochloride salt.⁴⁴ 

Numerous hybrid scaffolds have been explored that lie outside the bis‐imidazole or bis‐benzimidazole scaffolds (see Figure 1.9). The most common structural elements include combinations of imidazole and benzimidazole moieties, as in 34, or cases where one of these two moieties is hybridized with other potential bioisosteres, including a primary amide (35), a thienoimidazole (36) or a quinazolinone (37).^45–48 

Although the pyrrolidine moiety has been replaced by acyclic amines or by homologous heterocycles, the majority of the reported analogs incorporate functionalized pyrrolidines or five‐membered heterocyclic variants in this region (Figure 1.10). Examples include the spirocyclic analog 38 and the bridged analog 39, where the latter molecule also incorporates a number of the modifications described earlier and has featured prominently in a recent patent application disclosing combinations of advanced HCV therapeutic agents that encompass a range of alternate mechanisms.^49,50  In another example, a dimethylsilane moiety was incorporated into the 3‐position of the pyrrolidine ring to afford 40, claimed to be a potent G‐1a/1b inhibitor.⁵¹  Compound 41, which is a peptidomimetic variant of 1, exhibited potent G‐1b inhibitory activity but is significantly weaker towards a G‐1a replicon.⁵² 

Finally, additional distinct chemotypes with moderate levels of inhibitory activity in replicon systems and resistance mutations that map to HCV NS5A have been reported (see 42–44 in Figure 1.11).^53–56  Although some of the resistance mutations overlap with those observed for 1, it is not apparent at this stage if they share a similar mode of inhibitory mechanism(s). It is noteworthy that a hybrid chemotype containing pharmacophore elements derived from 1 and 43 has been claimed to exhibit potent inhibitor activity towards both G‐1a and G‐1b replicons (see 45).⁵⁷ 

Clinical validation that inhibitors of NS5A represented an effective approach to the control of HCV replication was obtained in Phase 1 studies with 1.¹  A single ascending dose (SAD) study conducted in normal healthy volunteers with doses of 1 ranging from 1 to 200 mg established dose‐proportional exposure of the drug, with plasma concentrations at 24 h post‐dose in all subjects significantly exceeding that required to express antiviral activity in replicons. Administration of an oral solution of 1 to subjects chronically infected with G‐1a and G‐1b HCV at doses of 1, 10 and 100 mg produced a rapid and dose‐related reduction of virus RNA levels. Viral RNA was reduced by an average of 1.8 log₁₀ measured 24 h following a single 1 mg dose of 1, whereas the 10 and 100 mg doses exerted more profound effects on viral load, with reductions of 3.2 log₁₀ and 3.3 log₁₀, respectively, at 24 h post‐dose. The single 100 mg dose was associated with a maximum 3.6 log₁₀ decline in viral RNA levels that was maintained for 144 h following drug administration, with one G‐1b virus‐infected subject in this cohort achieving RNA levels below the 25 IU mL^–1 level of quantification at the 144 h time point, and the viral load for a second subject was measured at 35 IU mL^–1.

Virus sequencing, conducted prior to dosing and at 24 and 144 h post‐dosing, revealed that when reduction in viremia was significant, HCV variants were detectable, with M28T, Q30H/R and L31M observed in G‐1a‐infected subjects and Y93H in the G‐1b patients. These observations presumably reflect potent and effective restriction of wild‐type virus replication by 1 that reveals virus‐possessing resistant mutations as the thriving species. This scenario is anticipated based on the replication rate of the virus, estimated to be 10¹² virions per day, combined with the low fidelity of the polymerase, the error rate of which is estimated to range from 0.1 to 1 nucleotide per RNA synthesized.^58–63 

These statistics contribute to the significant population of viruses harboring single (87×10⁹ virions per day), double (4.2×10⁹ virions per day) and triple (0.13×10⁹ virions per day) mutations produced in an infected individual every day.^58–63  Indeed, in a multiple ascending dose study with 1 administered at doses of 1, 10, 30, 60 and 100 mg qd and 30 mg bid for 14 days to chronically infected G‐1‐infected subjects, the viral load was rapidly reduced by 2.8–4.1 log₁₀ IU mL^–1 across the cohorts, but most patients experienced rebound on or before day 7 of the treatment period.⁶⁴  Viral rebound was associated with the emergence of resistant variants with major substitutions identified at residues M28T/A/V, Q30H/R/K/E, L31M/V and Y93H/C/N for G‐1a‐infected subjects and L31M/V and Y93H/C for G‐1b.⁶⁵  These mutant viruses were also observed in vitro in G‐1a and G‐1b replicons placed under selective pressure from 1. One patient in the 60 mg qd cohort, all of whom were infected with G‐1a, had a Q30R mutation detectable at baseline and experienced initial viral suppression, but viral breakthrough occurred by day 14 with a Q30H and Y93H linkage detected that in vitro exhibited high resistance to 1.⁶⁵  For two additional patients, Q30E and Y93N variants were detected at day 14, a double mutant associated with high resistance to the drug. The final patient in this cohort experienced failure of therapy with a Q30R virus that emerged within 12 h of drug dosing, despite the fact that the plasma exposure of 1 at day 14 (117 nM) substantially exceeded the in vitro replicon EC₅₀ of 7 nM.⁶⁵  A closer analysis revealed a baseline E62D polymorphism that by itself did not confer resistance to 1 but, when linked to Q30R, conferred high levels of resistance in vitro.⁶⁶ 

Analysis of a cohort of 78 HIV‐HCV co‐infected subjects and 635 NS5A sequences deposited in the Los Alamos database for the occurrence of baseline resistant mutations to 1 revealed an absence in G‐1a and G‐3 whereas all G‐4 sequences had L31M; the double mutant L31M+Y93H occurred in 7% of G‐1b and 13% of G‐4 sequences.⁶⁷  In a cohort of Japanese subjects infected with G‐1b, overwhelmingly the most prevalent in that population, resistance‐conferring amino acid substitutions were detected in 11.2% of 294 patients, with Y93H (8.2%) predominating over L31M (2.7%).^68–70 

Taken together, these observations emphasize the anticipation based on virus replication kinetics that combination therapy will be required, either by adding a direct‐acting antiviral agent (DAA) to interferon‐α/ribavirin therapy or by combinations of DAAs with orthogonal patterns of resistance, to suppress the virus effectively and durably.^{61–63,71–75} 

In a Phase 2a clinical trial of 1 in conjunction with PEG interferon‐α/ribavirin (PEG‐IFN/RBV), doses of 3, 10 and 60 mg of the drug were compared with a placebo control arm over a 48 week time span.⁷⁶  The primary efficacy endpoint focused on an extended virological response (eRVR), which is defined as undetectable levels of viral RNA at both weeks 4 and 12 after initiation of therapy. Secondary endpoints were rapid virological response (RVR; HCV RNA undetectable at 4 weeks), complete early virological response (cEVR; HCV RNA undetectable at 12 weeks) and sustained virological response at 12 and 24 weeks after the end of the dosing period (SVR₁₂ and SVR₂₄). The results of this trial are compiled in Table 1.1 and indicate that the two higher doses of 1 are associated with greater efficacy, with 5 of 12 (42%) patients in the 3 mg group achieving eRVR compared with 10 of 12 (83%) and 9 of 12 (75%) of those receiving 10 and 60 mg of drug, respectively, whereas only 1 of the 12 (8%) administered with just PEG‐IFN/RBV achieved this endpoint.⁷⁶  Based on results from subsequent Phase 2b trials, the 60 mg dose of 1 was selected for Phase 3 studies.

The first step towards a PEG‐IFN/ribavirin‐free therapy was a short clinical trial in which 74 treatment‐naive and null‐responding HCV‐infected patients – the latter are the most difficult to treat patient subgroup – were administered a combination of the HCV NS3 protease inhibitor danoprevir (46) at doses of 100 or 200 mg tid or 600 or 900 mg bid and the nucleoside prodrug mericitabine (47) at doses of 500 or 1000 mg bid, for up to 13 days against a placebo control arm (n=14) (Figure 1.12).⁷⁷  In the treated subjects who completed the 13 days of therapy, the viral load declined by a median 3.7–5.2 log₁₀ IU mL^–1, which compared with an increase of 0.1 log₁₀ IU mL^–1 observed in the placebo arm.

A more recent clinical trial examined the potential of a combination of daclatasvir (1) (60 mg qd) and the NS3 protease inhibitor asunaprevir (48) (600 mg bid) to cure a small cohort of G‐1‐infected null responders who were administered the drugs for 24 weeks with and without PEG‐IFN/RBV (49).^78,79  In this trial, all 10 patients administered the quadruple drug regimen experienced SVR₁₂ and 9 of the 10 maintained this status to 24 weeks post‐dosing, while the remaining patient had detectable HCV RNA at week 24 after cessation of drug but had undetectable viral RNA 35 days later. Nine of the 10 patients had an SVR at 48 weeks, with the remaining patient having measurable RNA at less than 25 IU mL^–1 at week 48, but undetectable 13 days later. The cohort receiving only the combination of the two DAAs comprised nine G‐1a and two G‐1b infections. Of these, two G‐1b‐ and two G‐1a‐infected subjects experienced an SVR at 12 and 24 weeks after drug therapy was completed and three of these maintained SVR at week 48, while the fourth one had detectable viral RNA that became undetectable upon retest 43 days later.^78,79  One G‐1a subject had undetectable viral RNA at the end of treatment, but relapsed 4 weeks later. However, six of the G‐1a‐infected subjects experienced viral breakthrough, which occurred as early as week 3 and as late as week 12 of therapy, attributed to the selection of resistant variants which included Q30R, L31M/V and Y93C/N in the NS5A protein and R155K and D168A/E/T/V/Y in the NS3 protease domain. This study established for the first time that a chronic HCV infection could be cured without the use of interferon‐α and/or ribavirin and is recognized as a watershed event in the therapy of the disease.⁷⁹ 

Although the results suggest that more difficult to treat G‐1a patients will require additional DAAs or the use of HCV inhibitors with a higher genetic barrier to resistance, the successful treatment of G‐1b‐infected subjects prompted a similar Phase 2a trial in a Japanese cohort with established null response to PEG‐IFN/RBV.⁸⁰  In this study, daclatasvir (1) (60 mg qd) and asunaprevir (48) (initially 600 mg bid but subsequently reduced to 200 mg bid) were administered to 10 patients for 24 weeks, with nine subjects completing therapy and one patient discontinuing after 2 weeks. HCV RNA was undetectable in those who completed the course of therapy by 8 weeks and all nine achieved SVR₁₂ and SVR₂₄, with no evidence of viral breakthrough during treatment or relapse post‐treatment. Most interestingly, the patient who stopped therapy after 2 weeks and who had detectable HCV RNA (1.8 log₁₀ IU mL^–1) at the time that therapy was discontinued, had undetectable levels of viremia measured on follow‐up visits at weeks 2, 3, 4, 13 and 24 after discontinuation.⁸⁰ 

GS‐5885 is an HCV NS5A inhibitor for which early clinical data have been reported.⁸¹  In a Phase 1 trial conducted in G‐1‐infected subjects, GS‐5885 was administered qd for 3 days at doses of 1, 3, 10, 30 and 90 mg and the effect on viral load was compared with placebo. Plasma exposure of GS‐5885 was close to linear with respect to dose, with C_max occurring at 4–6 h after drug administration. Quantifiable concentrations were detectable at 24 h after all doses, exceeding the protein binding‐adjusted EC₉₀ for the less sensitive G‐1a at doses above 10 mg. Once‐daily dosing of GS‐5885 was supported by a long median apparent plasma half‐life that ranged from 22 to 50 h. All doses above 3 mg produced a >3 log₁₀ IU mL^–1 decline in viral load and suppression was more sustained for G‐1b‐infected patients, although the median maximum reductions were similar for both genotypes 1a and 1b.⁸¹  Of the 72 patients enrolled in this study, four G‐1a subjects and one G‐1b subject harbored detectable viruses with resistance‐associated mutations. Two of the four G‐1a‐infected patients experienced maximum HCV RNA reductions of <1.6 log₁₀ IU mL^–1, of whom one was characterized with a Q30E/Q population and the other harbored L31M virus at baseline. However, one subject dosed with 10 mg of GS‐5885 had a maximum viral response of <1.6 log₁₀ and 454 pyrosequencing was necessary to determine that 12% of the virus harbored a Y93C change in the NS5A gene. The emergence of resistance mutations was assessed by population sequencing and all patients dosed at 3 mg or more had virus with detectable changes associated with resistance to GS‐5885 in vitro. M28T, Q30H, L31M, Q30R and Y93C/H were characterized in G‐1a‐infected subjects, with Q30R being the most frequent, while Y93H was detected in all 10 G‐1b‐infected individuals receiving the 10 mg dose of drug. The less resistant M28T and Q30H mutations were not detected in G‐1a patients at doses ≥30 mg, reflective of plasma concentrations at trough that were above the protein binding‐adjusted EC₉₀ values.

The quinazoline derivative AZD‐2836 (18) was the first HCV NS5A inhibitor actually to enter clinical trials, in early 2007, but was abandoned, apparently due to inadequate exposure that was not solved by optimization of the formulation.²⁷  The antiviral activity of AZD‐7295 (20) was explored in treatment‐naive and treatment‐experienced patients infected with genotypes 1a, 1b and 3a virus who were administered the drug for 5 days at doses of 90 and 233 mg tid or 350 mg bid.^27,82  The 90 mg tid cohort comprised five each of G‐1a‐ and G‐1b‐infected subjects, with the G‐1b patients receiving active drug experiencing a mean viral load reduction of over 1 log₁₀ IU mL^–1. The higher doses of AZD‐7295 (20) produced a more pronounced reduction in viral load, up to a maximum of 2.4 log₁₀ IU mL^–1, although 4 G‐1b patients across the cohorts showed no response to the drug. In contrast, the viral load in G‐1a‐infected subjects was not statistically significant from placebo, reflecting the poor in vitro potency of the compound towards G‐1a and a C₂₄ drug concentration that did not surpass the EC₉₀ measured in the replicon assay. Similarly, AZD‐7295 (20) at 233 mg tid exerted no significant effect on viral load in the G‐3a‐infected subjects.

The early clinical studies with HCV NS5A inhibitors revealed a class of antivirals that is characterized by high in vitro potency, inhibition of multiple genotypes and excellent efficacy at low doses. The first‐generation NS5A clinical candidates appear to have a low genetic barrier to resistance. The NS5A inhibitors evaluated to date were generally well tolerated and offer considerable promise for use in combination therapy that has the potential to cure chronic HCV infection in the absence of pegylated interferon‐α and/or ribavirin.^83–85 

Detailed in vitro replicon studies have revealed that the mutations that confer resistance to HCV NS5A replication complex inhibitors, such as 1 and 3, map to the N‐terminal region of Domain I, involving amino acids 24–100.^1,65  Resistance analysis of clinical samples obtained from SAD and MAD studies of daclatasvir (1) have corroborated these in vitro findings.⁸⁶  Although there are differences in the exact mutation composition of different resistant genotypes, there is a broad overlap in the specific locale that are hot‐spots for the emergence of resistance mutations, which is suggestive of a common mode and, possibly, region of interactions between NS5A and its putative inhibitors. Although residues 28–31, which is a key resistance mutation region for both G‐1a and G‐1b viruses, was not captured in the X‐ray structural studies of Domain I, it is noteworthy that the mutation Y93H, which is a clinically relevant mutation common for G‐1a and G‐1b, lies in the dimer interface region of both X‐ray structures.

To gain additional insight into the mode of interaction between 1 and the protein, two models for the G‐1b NS5A Domain I were constructed using the Rice dimer (1ZH1.pdb, model 1) or the Love dimer (3FQM.pdb, model 2) and the NMR structure of the amphipathic N‐terminal α‐helix, amino acids 1–31.^9,10,87  The respective missing amino acids were modeled in so as to complete each monomer. Mapping of the primary resistance mutations observed in clinical studies for both G‐1a and/or G‐1b patients on to the model indicates a clustering of mutation sites that is presumed to indicate the putative inhibitor binding site. Daclatasvir (1) was hand‐docked into the putative binding site symmetrically across the dimer interface of each model. Docking into the model constructed based on the Rice dimer produced the best fit based on the physical properties of both ligand and protein and also the location of resistance mutations. The hydrophobic regions of 1 – the biphenyl core, pyrrolidine and valine side chain moieties – are surrounded by hydrophobic amino acids (residues M28 and L31), while the polar imidazole and carbamate moieties deployed on each end of the inhibitor are in a position to form hydrogen bonds with the monomeric units (Figure 1.13). It is believed that these inhibitors may interact with the NS5A dimer and induce a conformational change that is not functionally viable. Most interestingly, a recent report suggests that disruption of dimerization may not be the mode of action for 1 and related analogs.^11,88 

Additional evidence that further corroborated the importance of the N‐terminal region of Domain I of NS5A as a potential site of interaction for inhibitors of interest was obtained from chimeric replicon studies conducted using compounds with differing inhibitory activities.¹⁹  For example, compound 3, which inhibits a G‐1b replicon with an EC₅₀ of 0.018 µM but is devoid of activity in a G‐1a replicon (EC₅₀>10 µM), exhibited an enhanced inhibitory activity (EC₅₀=0.032 µM) when tested in a hybrid G‐1a replicon in which the first 76 amino acids of its NS5A protein were replaced with the corresponding G‐1b sequence. (The difference in the G‐1b EC₅₀ reported for 3 in different sections (i.e., 18 nM versus 6 nM) is a result of assay variations between studies.) Conversely, replacement of the first 76 amino acids of the NS5A region of a G‐1b replicon with the corresponding G‐1a sequence resulted in a decrease in the inhibitory potency of 3 (EC₅₀ of >10 µM). The fact that the inhibitor sensitivity domain of NS5A overlaps with the region that resistance mutations map to is consistent with the direct engagement of NS5A with inhibitors.

A diastereomeric pair of biotin‐tagged analogs with differential inhibitory potencies was used in an NS5A pull‐down experiment to provide evidence for a direct and specific binding interaction between NS5A and its putative inhibitors.¹  Specifically, NS5A was selectively pulled down when a G‐1b replicon was incubated with the biotinylated inhibitor 50a (EC₅₀=33 nM) (Figure 1.14), lysed and passed over streptavidin–agarose beads, but not when the replicon was lysed prior to treatment with the biotinylated compound. This result signifies that a specific conformation of NS5A is needed for a productive interaction with an inhibitor and is accessible only in a cellular context, presumably in the virus replication complex, which is consistent with a report that NS5A inhibitors failed to bind to the isolated protein.⁸⁹  More importantly, in control experiments conducted in parallel, it was observed that little HCV NS5A was pulled down by the inactive stereoisomer 50b (EC₅₀>10 µM). Moreover, only NS5A, and not NS3 or NS5B, was detected in the bound proteins and 50a failed to pull down the NS5A protein of BVDV under similar conditions.

The availability of potent NS5A‐targeting tool compounds has catalyzed additional efforts that are directed at shedding light on the ramifications of inhibition of the NS5A protein which, in turn, provides additional insights into the mode of action of the inhibitors. NS5A is a phosphoprotein that exists in both basally phosphorylated (p56) and hyper‐phosphorylated (p58) forms. It was discovered that NS5A inhibitors such as daclatasvir (1) dose-dependently inhibit the formation of the p58 form, without affecting basal phosphorylation, in a manner that correlates with their RNA replication‐inhibitory activities.^19,90  Moreover, inhibition of hyperphosphorylation did not depend on the presence of Domains II and III of NS5A. Although the exact significance of phosphorylation of NS5A is still unknown, it is hypothesized that phosphorylation is a regulatory mechanism that toggles the NS5A protein between functional states in the viral replication cycle. Interestingly, it was observed that protease inhibitors could similarly inhibit the formation of p58 in a dose‐dependent manner, clearly indicating that the modulation of the phosphorylation state of NS5A is not unique to NS5A inhibitors, albeit this observation may signify the spatial and functional associations of the NS3 and NS5A proteins within the HCV replication complex or polyprotein.¹⁹  Corroborating evidence for a possible interaction of NS5A inhibitors at the polyprotein‐processing stage came from a recent study that demonstrated that treatment with 1 results in the accumulation of the NS4B–NS5A polypeptide in vitro, an effect that was sensitive to the presence of resistance‐conferring mutations.⁹⁰ 

A combination of morphological and biochemical studies have demonstrated that 1 alters the subcellular distribution of NS5A from that of localized foci to diffuse cytoplasmic patterns.⁸⁸  This inhibitor‐induced effect on the subcellular disposition was specific to the NS5A protein and, in line with expectations, was minimized in a replicon harboring the Y93H‐resistant mutant up to certain concentration ranges. A different set of studies revealed that NS5A inhibitors deregulate the normal distribution of NS5A by relocating the protein from the endoplasmic reticulum to lipid droplets, an effect that is minimized in the context of the Y93H‐resistant mutant and which is also specific to the NS5A replication complex inhibitor class.⁸⁹  Finally, a cellular imaging study that utilized a click chemistry approach involving an azide‐containing NS5A inhibitor (51) and an alkyne‐containing fluorophore (52) demonstrated the colocalization of an NS5A replication complex inhibitor with the NS5A protein (Figure 1.15).⁹¹ 

The unusually high in vitro and in vivo inhibitory potency of the NS5A replication complex inhibitor class, coupled with the absence of clearly defined functions of the NS5A protein and of inhibitor binding data for the purified protein, has catalyzed numerous mode of action studies. Although considerable advances have been made, much remains to be discovered in order to illuminate further the complex set of processes that NS5A appears to orchestrate during the HCV replication cycle.

The complex but ill‐defined role of NS5A in the HCV replication cycle continues to drive the curiosity of investigators across academia and industry. Although NS5A was not among the initial targets of choice in the HCV drug discovery campaigns that started over 15 years ago, the pioneering work by Bristol‐Myers Squibb scientists in this field has culminated in the discovery of daclatasvir (1). The key to success included devising a unique dual‐assay screening system that identified a single hit from a collection of over one million compounds; defining aspects of the instability of one of the lead compounds that uncovered significantly active dimeric degradants; simplifying the dimeric pharmacophore into a progressible lead; and successfully optimizing a chemotype with a molecular footprint that is outside of what is considered to be traditional drug‐like space. Daclatasvir (1) established clinical proof‐of‐concept for the HCV NS5A protein as a therapeutic target and set a potency benchmark for the HCV field. In addition, as part of a DAA combination, it demonstrated for the first time that a PEG‐IFN/RBV‐free regimen could cure HCV infection. The clinical validation of NS5A has made it an attractive target for therapeutic intervention, as evidenced by the significant number of patent filings and the growing number of NS5A‐targeting compounds entering clinical trials. It is anticipated that NS5A replication complex inhibitors will become integral components of the more effective HCV combination therapies that are expected to emerge in the near future. The genesis of Bristol‐Myers Squibb’s NS5A drug discovery effort that enabled this successful endeavor is a clear testament to the utility and power of phenotype‐based screening approaches in uncovering valuable targets that otherwise would have remained unexplored because of the lack of biochemical information.