1 Introduction

The discovery of the hepatitis C virus (HCV) nonstructural 5A (NS5A) replication complex inhibitor daclatasvir (1) began with the development of a genotype 1b (GT-1b) replicon that was implemented as a phenotypic screen using a design that conferred a stringent triaging of hit molecules [1,2,3,4,5,6,7,8,9,10,11]. This screening campaign was part of a broader strategy that pursued the identification of mechanistically orthogonal inhibitors of HCV that could be used in combination therapy, an approach that anticipated clinical use of drug cocktails to minimize the selection of resistant viruses. More specifically, the replicon screen comprised of simultaneously assessing the antiviral activity of test molecules toward a sub-genomic HCV GT-1b construct and a bovine viral diarrhea virus (BVDV) replicon, both replicating in the same human liver Huh-7 cell line background and seeded in the same well of a 96-well plate [6]. HCV inhibition was determined indirectly using a fluorescence resonance energy transfer (FRET)-based assay that assessed NS3 protease activity toward a synthetic substrate incorporating both a fluorescence donor [(5-((2-aminoethyl)amino)naphthalene-1-sulfonic acid (EDANS)] in the amino terminus and a fluorescence acceptor [4-((4-(dimethylamino)phenyl)azo)benzoic acid, succinimidyl ester (DABCYL)] at the carboxyl terminus. The BVDV replicon incorporated a firefly luciferase reporter gene that provided an orthogonal readout of replication activity based on the emission of light, reflecting the amount of enzyme present after adding a substrate. The third piece of information harvested from the assay, which was also the first experimental data obtained, was AlamarBlue cell viability staining which assessed mitochondrial function, providing an indication of the cytotoxicity of test compounds. This assay configuration was used to interrogate a representative selection of the Bristol-Myers Squibb proprietary compound collection and identified the iminothiazolidinone 2 as hit that met the criteria of exhibiting activity toward the HCV replicon at concentrations at least tenfold lower than either inhibitory activity toward the BVDV replicon or cytotoxicity [12]. The specific details of the antiviral profile of 2 in the screening assays and toward a panel of viruses are summarized in Table 1. Compound 2 had an interesting background since it had its origin in a prospective library that had been prepared as part of a campaign to embellish the Bristol-Myers Squibb proprietary compound collection. A notable structural feature of 2 is that it had been designed to include a phenyl substituent at C-5, unusual since C-5 benzylidene derivatives are far more prevalent in the literature, a function of convenient access by way of a Knoevenagel condensation reaction between a C-5 unsubstituted iminothiazolidinone and an aldehyde [13, 14].

Table 1 Antiviral activity of 2 in the GT-1b HCV replicon and toward a panel of viruses
figure a
figure b

The data accumulated for 2 that are compiled in Table 1 revealed a profile of significant and selective inhibitory potency toward the GT-1b HCV replicon with an EC50 value of 570 nM, a finding that encouraged further study of the chemotype [15]. Adding to the intrigue with 2 as a lead inhibitor was the generation and validation of resistant mutants arising in the replicon in response to selective pressure that mapped to the amino terminus of the NS5A protein, specifically a Tyr98His and a Leu31Val/Gln54Leu combination [12]. At the time of this discovery, little was known about the specific functions of the NS5A protein in viral replication although, perhaps not surprisingly for such a small virus, it was known to be an essential protein [16,17,18,19,20].

The Z-configuration of the 2-arylimino moiety of 2 was assigned based on prior studies which indicated that this topology minimized steric effects. However, attempts to separate the individual stereoisomers at C-5 were thwarted by facile racemization of the benzylic center after chiral chromatographic separation [15, 21]. Seminal insights into structure-activity relationships (SARs) were gleaned from a survey that explored variation of the three major peripheral elements of the molecule – the furanylmethyl substituent, the arylimino moiety, and the amino acid amide. Changes to both the furanylmethyl and arylimino moieties in the context of the Cbz-alanine amide were found to influence potency, with compounds incorporating polar heterocycles being the more potent, while more lipophilic substituents generally exhibited poorer replicon inhibitory activity. Overall, a good dynamic range of potency was observed with the structural variations sampled, with EC50 values ranging from 10 to 2.5 μM. However, more profound and precise effects on HCV GT-1b inhibitory potency were associated with changes to the Cbz-alanine amide element, with the key findings summarized in Fig. 1. The unnatural alanine analogue 3 was 170-fold less potent, while the glycine homologue 4 was eightfold weaker than 2. The strong dependence of antiviral activity on the absolute configuration of the amino acid element was reproduced in the proline derivatives 5 and 6, both of which were fivefold more potent than their alanine congeners [15]. Replacing the Cbz element with a dihydrocinnamate moiety resulted in an erosion of potency in both series, as exemplified by 7 and 9. However, a significant increase in potency was observed with the phenylacetamide moiety found in 8 and 10, with both compounds expressing single digit nanomolar EC50 values in the replicon and confirming the inherent advantage offered by proline. Only weak antiviral effects were observed with the many other amino acids and additional structural variations that were explored at this site of the pharmacophore, with some representative examples of those sampled compiled in Fig. 2 [15].

Fig. 1
figure 1

Structure-activity relationships associated with variation of the amino acid moiety of 2

Fig. 2
figure 2

Structure-activity relationships associated with the amino acid element of 2

While these collective SARs appeared to be coherent and were readily interpretable, as highlighted by the sensitivity of potency to changes to the amino acid element, as exploration of 2 and its analogues continued, it became apparent that the iminothiazolidinone chemotype was unstable under some conditions [15, 22]. The first indication of a problem was the observation that a sample of 2 degraded upon standing as a solution in DMSO for several days, with the thiohydantoin 15 (R3 = NHCO2CH2Ph) and thiourea 17 (R3 = NHCO2CH2Ph) determined to be two of the degradation products after a preparative experiment (Scheme 1). In the replicon, 15 exhibited a modest inhibitory activity, EC50 = 13 μM, while 17 was essentially inactive. Compounding the intrigue, incubating the more potent 8 in replicon media until degradation was complete and then assessing the antiviral activity of the preparation in the GT-1b replicon demonstrated that the HCV inhibitory activity and potency were fully preserved [15, 22]. A more detailed analysis of the degradation products from 8 guided by high-performance liquid chromatography (HPLC) biogram analysis, in which fractions taken from the chromatography column are concentrated and assessed for replicon inhibition, led to the identification of two potent constituents present in just trace amounts in the cell media [22, 23]. A preparative experiment allowed isolation of a sufficient quantity of each component to allow for a more detailed characterization. 1H-NMR and mass spectrum analyses indicated that the two compounds were dimers with an isomeric relationship and assigned as 18 based on the absence of the C-5 hydrogen atom of the iminothiazolidinone ring in the 1H-NMR spectrum and connectivity that was confirmed after the analysis of the 13C-NMR spectrum. Although neither the exact stereo-composition of the benzylic centers of 18 nor the precise relationship between these two compounds was determined, it is of note that the isomer that was less mobile on a reversed phase column converted to the more mobile isomer upon heating in CD3CN at 50°C, an observation made while conducting an NMR experiment as part of the structure determination process [22]. With the acquisition of these data, the mechanistically holistic picture of the degradation process that is presented in Scheme 1 was proposed. Abstraction of the C-5 hydrogen atom of 8 by oxygen, which is a diradical in the ground state, would lead to the C-5 radical 11 that is stabilized in a classic captodative fashion by the adjacent C=O moiety, the sulfur atom, and the phenyl substituent [24, 25]. Combination of 11 with O2 and hydrogen atom acquisition would lead to the hydroperoxide species 12 which in DMSO would be reduced to alcohol 13, an unstable ring system that would be expected to undergo ring opening to give 14. Reclosure of 14 would then afford the thiohydantoin 15, which had been isolated in the original DMSO degradation experiment, while hydrolytic decomposition of 14 would afford acid 16 and the thiourea 17, the latter of which had also been isolated. However, in cell culture media, the stability of radical 11 is presumably such that dimerization can occur, a process that may well be facilitated by aggregative association of 8 and/or 11 in the aqueous medium.

Scheme 1
scheme 1

Chemical degradation path elucidated for 2

figure l

The replicon inhibitory potency of the dimers 18a and 18b was striking, with the less mobile isomer eluting from the reversed phase chromatography column exhibiting an EC50 value of 600 pM in the GT-1b replicon, while the thermodynamically more stable and chromatographically more mobile isomer was 70-fold weaker, EC50 = 43 nM [15, 22]. With the elucidation of the structures of 18a and 18b, consideration was given to the concept that the NS5A-inhibiting pharmacophore represented by these dimers may be the embedded symmetrical bibenzyl element. This notion was based on an appreciation of the precise SARs surrounding the amino acid moiety in contrast to the relatively more nebulous effects associated with structural variation to the iminothiazolidinone substituents. Under this concept, the iminothiazolidinone ring system of 2 was suggested to act as a scaffolding element by which a bivalent NS5A inhibitor pharmacophore was convened through a radical-mediated dimerization process [15, 22]. This hypothesis was readily tested in the context of the proline-based chemotype, with 19 demonstrating an EC50 value of 30 nM and confirming the pharmacophore proposal. More interestingly, the unsaturated synthetic precursor 20 was almost 350-fold more potent than 19, EC50 = 0.086 nM, providing further insight into the topography of the NS5A-inhibiting pharmacophore.

2 Optimization of Dimeric NS5A Replication Complex Inhibitors to Daclatasvir

The discovery of the structurally simpler symmetrical pharmacophore represented by 19 and 20 turned out to be prescient since X-ray crystallographic structure data for NS5A constructs that were published several years after this discovery revealed a dimeric, C2-symmetric protein complex [26,27,28]. Of note, this structure was determined using a protein construct lacking the membrane-associating amino terminus but, nevertheless, containing some of the key elements of the anticipated binding site for NS5A inhibitors based on the location of resistance mutations. The availability of a structure of the amino terminus of NS5A, acquired by NMR methodology, facilitated the construction of models of NS5A that, when combined with the resistance mutation data, suggested that these compounds bound across the NS5A dimer interface at a site residing between the membrane and the core of the protein. Dimerization of the NS5A protein in cells was subsequently confirmed as was association of NS5A with RNA, a prediction from the X-ray data based on the preponderance of basic amino acids lining the inner surface of the U-shaped dimeric form of NS5A in the solid state [29,30,31,32].

While the antiviral activity of 20 was attractive, there were concerns around elements embedded within the structure, with the olefin observed to be configurationally unstable in some analogues, while the potential for release of one or both aniline moieties after the action of an amidase or protease in vivo raised the specter of toxicity. However, a considerably more challenging problem arose when 20 was evaluated in a newly developed GT-1a replicon where the compound was found to be poorly active, with an EC50 value that was in excess of 10 μM. The GT-1a strain of HCV is clinically relevant, with prevalence that varies across the globe; consequently, securing activity toward this genotype was considered to be a critical objective. Enhancing GT-1a inhibitory activity became the immediate focus of further study, and the approach adopted, while primarily directed toward the peripheral elements, was broad-based in prosecution. Modifications to all facets of the molecule were explored, some of which were also directed toward simultaneously addressing the structural liabilities highlighted above. However, as described below, introducing and retaining GT-1a coverage while optimizing ADME properties would end up posing a considerable challenge. While deeper insight into SARs for GT-1b inhibition emerged from the initial phases of this effort, only a small number of compounds were identified that exhibited modest but reproducible GT-1a inhibition. Among these were the oxazole 21 and the substituted proline derivative 22, both of which were inhibitory in the GT-1a replicon with EC50 values of less than 1 μM [33]. However, attempts to optimize these molecules were unproductive, in each case leading to SAR cul-de-sacs. As the studies progressed, the 2-ethyl-substituted benzamide derivative 23 and its unsaturated homologue 24 were discovered to exhibit weak inhibition of GT-1a and GT-1b replicons, with modeling studies suggesting that the effect of substitution was not a function of modulating the conformation between the phenyl ring and proline carbonyl [34]. The pyridine derivative 25 further advanced the SAR associated with this cap element but, more importantly, formed the basis of the discovery of the isoquinoline derivative 26, in which the ethyl substituent was incorporated into a fused ring, as the first compound to exhibit potent and balanced antiviral activity toward GT-1a and GT-1b replicons [34]. The promise of this compound was further underscored when it was screened in replicons or hybrid replicons representing genotypes 2a, 3a, and 5a where the EC50 values ranged from 2.2 to 14 nM.

figure c

Further examination of the SARs revealed that GT-1a inhibitory activity was much more sensitive to the nature and the substitution pattern of the isoquinoline ring than GT-1b [34]. For example, 27, the parental analogue of 26, exhibited GT-1a and GT-1b EC50 values that were 20- and 3-fold weaker than that of 26, respectively. In addition, the methoxy-substituted derivative 28 and its chloro-substituted analogue 29 retained potent GT-1b inhibition, but their GT-1a EC50 values were >10 μM. However, a more fruitful avenue of investigation was found when the effects of deannulating the isoquinoline ring were probed [35]. The α-keto amide 30 preserved the GT-1a inhibition exhibited by 26, while the derived secondary alcohols 31 and 32 demonstrated that planarity at this site was not a specific requirement. The tertiary alcohol homologues 33 and 34 added further to the SARs, with the (S,S )-analogue 34 the more potent isomer, particularly toward the GT-1b replicon where the EC50 value was 6 pM. Another encouraging observation was made with 35 which, although poorly active in the GT-1a replicon, demonstrated 64% bioavailability following oral dosing to rats, indicating that securing systemic exposure after oral delivery of these symmetrical stilbene derivatives was feasible.

figure d

Replacing the stilbene element with an alkyne, which resolved a cis-trans isomerization issue observed with some analogues, added further to the understanding of the topography of the pharmacophore. Additional probing of the amino acid terminal region using this alkyne-based scaffold identified potent arylglycine-based analogues for which the GT-1a EC50 values for some compounds, including 3638, were single digit nM [35]. Notably, the change in stereochemical preference in evolving the chemotype from the mandelamide analogues 33 and 34 to the phenylglycine analogues 36 and 37 further highlighted the relatively intricacy and sensitivity of the GT-1a inhibition SARs that were being uncovered during this phase of the project. Equally intriguing was the accumulating evidence indicating that the GT-1b virus was highly tolerant of many structural changes, an observation that could not readily be explained based on the differences in amino acid composition at the putative binding site of the compounds.

A concurrent effort examined scaffold modifications directed toward the identification of a less problematic replacement for the anilide moiety that would decrease or eliminate the potential for aniline release in vivo which led to the emergence of two noteworthy SAR findings. Firstly, the replacement of the anilide moieties of 36 and 38 with a benzimidazole, a design intended to preserve both the H-bond donor and acceptor properties of the anilide, resulted in a 4- to 30-fold reduction in potency toward the GT-1a replicon, as exemplified by 39 and 40 [36]. Secondly, a mix and match SAR exercise accentuated the sensitivity of GT-1a inhibitory potency to topological parameters, exemplified by the 70-fold difference in GT-1a inhibitory potency between regioisomers 41 and 42.

These SAR findings were attributed to the altered topology of the peripheral pharmacophoric elements with respect to the core, a shortcoming that was ultimately addressed by the biphenyl derivatives 43 and 44 which recapitulate the linearity associated with the core alkynes in 3640. In 43 and 44, a motif that was arrived at only after considerable experimentation, deannelation of the benzimidazole ring provided a structural arrangement that compensated for the reduced length of the core of the pharmacophore relative to 39 and 40. The success of this design strategy is readily apparent since both 43 and 44 are exquisitely potent HCV antiviral agents with balanced GT-1a and GT-1b inhibition, with EC50 values ranging from 7 to 42 pM [4]. However, the oral bioavailability and systemic exposure of both 43 and 44 in the rat were poor, a result attributed, in part, to the high molecular weight (747 and 807 Da, respectively) and structural composition. This notion was reinforced by PK studies with the smaller (MW = 713) and unsymmetrical tetrahydrofuran 45 which divests of a H-bond donor. Although the oral bioavailability of 45 was low in rodents (F in mouse = 17%, F in rat = 6.8%), its exposure in the dog was much improved, with bioavailability measured at 45%. In an effort to reduce the molecular weight of the carbamate 44, the two aromatic rings of the phenyl glycine moiety were curtailed to isopropyl substituents affording the D-valine derivative 46. However, this structural modification resulted in a significant reduction in potency toward both HCV genotypes, with the GT-1a inhibition particularly sensitive, eroding by 44,000-fold. This SAR observation took some time to understand and was resolved only after further study of the chemotype, which revealed that the preferred absolute configuration of the alkyl-glycine caps was the opposite of that of the aryl-glycine caps. The initial observation in this direction was made when the tetrahydrofuran ring of 45 was replaced with L-alanine to provide 47, which restored potency to sub-nanomolar levels. In an observation that proved to be pivotal, the symmetrical alanine derivative 48 performed similarly, and further optimization of the amino acid appendage led to the discovery of the bis-L-valine derivative 1, an exercise that also included assessing the potential of unsymmetrical derivatives. The decision to advance 1 into IND-enabling toxicology evaluation was taken after a careful comparison with 49, an analogue with two changes to the periphery that demonstrated similar antiviral properties to 1 (Table 2) but a different PK profile (Table 3). The decision to select 1 for development rather 49 was based on the observation of a twofold accumulation of the latter compound in the plasma, livers, and hearts of mice after 4 days of daily drug administration which occurred at all of the dose levels (15, 50, and 100 mg/kg) examined.

figure e
Table 2 Inhibition of replicons and hybrid replicons by 1 and 49
Table 3 Pharmacokinetic profile of 1 and 49 in preclinical species

The antiviral profiles of 1 and 49 toward wild-type and hybrid replicons representing all of the genotypes and subtypes that were available at the time are summarized in Table 2 [1,2,3,4, 37,38,39,40]. Adding to confidence in the potential of 1 as it negotiated the development path toward clinical evaluation was the potent inhibition observed in a newly established GT-2a JFH-replicating virus assay. The EC50 value of 28 pM in this assay exhibited a good correlation with the inhibitory potency toward the GT-2a JFH replicon [1,2,3,4].

The pharmacokinetic parameters of 1 and 49 in rat, dog, and cynomolgus monkey are compiled in Table 3 and demonstrate good systemic exposure following oral administration, with the drug concentration measured at 24 h post-dose well in excess of the EC50 values for the GT-1a and GT-1b replicons and the protein-binding-adjusted EC90 value of 383 pM determined for the GT-1a replicon. More importantly, target organ exposure was also demonstrated, with liver levels of 103 nM measured in the rat 24 h following a 5 mg/kg dose of 1, a concentration that was fivefold higher than that in plasma. The favorable absorption properties of 1 have been attributed, in part, to the formation of an intramolecular H-bond between the carbamate C=O and imidazole N-H moieties that enhances lipophilicity and reduces the apparent PSA of the molecule based on a chromatographic analysis and which is supported by modeling studies [41].

3 Mode of Action Studies with Daclatasvir

Despite its high potency and broad genotype inhibitory activity, the precise mode of inhibition of HCV replication by 1 remains as enigmatic as does the biochemical function of the NS5A protein [16, 18, 42,43,44,45,46,47,48]. HCV NS5A has no known enzymatic activity but is a critical element in the assembly and function of the replication complex on intracellular membranes and also in virion assembly [16, 42,43,44,45,46,47,48]. HCV NS5A is a 447-residue phosphoprotein that is comprised of three functional domains and an amphipathic helix at the amino terminus that associates with but does not traverse to biological membranes. Domain 1 contains a Zn2+ binding motif and several serine residues that are sites of basal phosphorylation and hyperphosphorylation. The phosphorylation state of NS5A may modulate its function in virus replication and assembly with the hyperphosphorylated form, which can be produced by the action of the host cell lipid kinase, phosphatidylinositol 4-kinase, involved in the assembly of virions. Domain 2 has been shown to bind to the NS5B RNA-dependent RNA polymerase and has been associated with the sensitivity of the virus to interferon therapy although that function is controversial. While domain 3 appears to play a role in virus replication, it has most prominently been associated with virion assembly. The mapping of resistance mutations arising in response to selective pressure exerted by 1 and related analogues to domain 1 of NS5A is consistent with the effect on virus replication, but studies with infectious virus have demonstrated that 1 also interferes with the assembly of virions [42,43,44,45,46,47,48,49,50]. The latter effect has been postulated to explain the rapid decline in viremia observed in HCV-infected patients administered clinically effective doses of 1 (vide infra) [3, 51]. In addition to associating with all of the viral nonstructural proteins, HCV NS5A has also been shown to bind to an extensive repertoire of host cell proteins that exceeds 130 entities [52,53,54,55,56,57]. As a consequence, the NS5A protein is typically viewed as a master regulator of virus replication and virion production, orchestrating both viral proteins and the host cell environment to ensure the successful production and release of progeny virus [42,43,44,45,46,47,48].

An association of NS5A inhibitors with the viral protein was originally demonstrated by studies with the biotin-labeled derivative 50 which is an effective inhibitor of GT-1b replication [3]. The antiviral activity of 50 is highly sensitive to the absolute configuration of the proline moiety since the (R,R)-isomer 51 is inactive, while inhibition is substantially reduced by the Tyr93His resistance mutation that arises in response to virus passaging in the presence of 1. This SAR profile is strictly analogous to that established for the stilbene chemotype, and 50 was thus viewed as a useful tool molecule with which to probe drug-target binding interactions. In an initial experiment, GT-1b replicon lysate was incubated with 50, and the mixture passed over streptavidin immobilized on beads; however, this experimental protocol failed to pull down any viral protein. In contrast, incubating replicon cells with 50 for 18 h before lysing and passing the lysate over streptavidin beads identified only NS5A as a binding partner, while a control experiment with inactive diastereomer 50 failed to isolate any virial proteins. These results indicate that 50 binds to HCV NS5A and that binding is dependent on the absolute configuration of the proline element, an observation concordant with the SARs developed in the stilbene-based series.

figure h

While the experiments conducted with 50 indicate that the binding of inhibitors to the NS5A is choreographically somewhat complex, the binding of inhibitors of NS5A to both domain 1 and the full-length viral protein was subsequently demonstrated in a series of independent biochemical experiments [58, 59]. These studies have suggested that the binding of inhibitors to NS5A interferes with the association of viral RNA with the protein, with the binding of compounds competed out by other NS5A inhibitors and demonstrating diminished affinity for the Tyr93His mutant protein [58, 59]. However, profiling of inhibitors in cell-based assays has indicated that disruption of RNA binding to NS5A does not appear to occur and that the introduction of key resistant mutations leads to only a modest reduction in the binding of inhibitors [60, 61]. Studies with 50 in resistant GT-1b replicons indicated that while the Tyr93His-resistant mutation reduced inhibitory potency by 220-fold, an estimate of the amount of NS5A protein pulled down by the chemical probe, as determined by an analysis of Western blots, suggested similar levels of protein-drug association for the resistant and wild-type strains [60, 61]. Adding further to the complexity of the biochemistry was the observation that in a protein pull-down experiment, an attempt to outcompete the biotinylated tool compound 50 with a non-biotinylated analogue in a GT-1b Tyr93His-containing mutant (EC50 = 290 nM for stated analogue vs >7 μM for 50) not only failed but, at low concentration, appeared to have incrementally enhanced the amount of NS5A pulled down. From these data, it was inferred that the development of resistance to 1 does not lead to exclusion of binding to the NS5A protein, as is often observed with other classes of antiviral agents. Rather, these observations suggested a scenario in which HCV NS5A develops resistance by accommodating rather than expelling inhibitors, with the mutations presumably allowing restoration of protein function in the presence of the bound inhibitor. Consistent with this suggestion, several of the resistant mutations incorporate smaller or more flexible amino acid side chains, exemplified by Tyr93His, Tyr93Cys, Leu31Met, and Leu31Val, which may restore conformational flexibility compromised by the binding of inhibitors. These observations stimulated an experiment designed to evaluate the effect of combining 1 with structurally related compounds on the function of HCV NS5A incorporating resistance mutations. Two possible outcomes were contemplated: in the first scenario, a molecule would simply compete with bound 1 and the observed effect would be one of silence. However, the alternative scenario speculated on the potential of a second molecule to act in conjunction with 1 to restore inhibition by binding to an adjacent site on the NS5A molecule. A screen of compounds selected from the library of HCV NS5A inhibitors assessed in the presence of 1 using the Tyr93Asn GT-1a-resistant replicon, followed by SAR optimization, identified Syn-395 (52) as a molecule capable of restoring the sensitivity of resistant virus to the inhibitory effects of 1. For example, in a typical experiment, 1 exhibited EC50 values of 33 pM and 339 nM toward the wild type and Tyr93Asn mutant replicons, respectively, whereas 52 was poorly active in both replicons, with EC50 values of 214 and 215 nM, respectively. However, the EC50 of 1 toward the Tyr93Asn mutant replicon improved to 0.13 nM when titrated in the presence of a suboptimal concentration (40 nM) of 52. This result represented a 2,600-fold increase in the sensitivity of the Tyr93Asn replicon to 1 in the presence of 52. The synergistic relationship between 1 and 52 was confirmed in a reciprocal experiment where 52 was titrated in the presence of suboptimal amount of 1 affording a similar outcome [60, 61].

figure i

These observations, taken together with the absence of structural data that holistically captures the HCV NS5A drug-binding sites in the context of the membrane-bound replication complex, confer considerable complexity on the nature of drug-target interactions, the mode of inhibition, and the function of the NS5A protein. This has presented a significant challenge to developing a more coherent and detailed understanding of the mechanism of action of HCV NS5A inhibitors and the modeling of drug-target interactions of these potent antiviral agents and the allosteric modulators [60,61,62,63,64,65,66,67]. While the bivalent nature of NS5A inhibitors, including the allosteric modulators represented by 52, complements the dimeric form of the protein observed in solid-state structures of elements of domain I, the binding interfaces between the proteins vary [26,27,28, 60, 61]. One interpretation of this observation anticipates an oligomeric form of HCV NS5A in cells that can bind to the viral RNA and protect it from chemical and enzymatic degradation while providing a platform for RNA presentation to the polymerase and translocation to the developing virion [68,69,70]. However, the biochemical pharmacological effects of NS5A inhibitors are multifaceted and complex and include altering the subcellular distribution of NS5A, modulating the phosphorylation state of the protein, interfering with the formation of the membranous factories where virus replication occurs, and blocking the transfer of the viral genome to assembly proteins, leading to a clustering of HCV proteins at endoplasmic reticulum membranes [71,72,73,74,75,76].

4 Clinical Trials with Daclatasvir

The phase I clinical trial with 1 comprised of a randomized, double-blind, placebo-controlled, single ascending dose study involving administration of 1, 10, 25, 50, 100, and 200 mg of the drug to normal healthy volunteers (NHVs) [3]. A dose-proportional increase in plasma exposure was observed over the dosing range, and the concentration of 1 in plasma 24 h after dosing exceeded the protein-binding-adjusted EC90 values of 49 pM (0.04 ng/mL) and 383 pM (0.28 ng/mL) recorded for the GT-1b and GT-1a replicons, respectively [3]. Compound 1 was quickly absorbed and exposure extended beyond 24 h, with plasma drug concentration maintained above the less sensitive GT-1a protein-binding-adjusted EC90 value of 383 pM at 72 h for all but the 1 mg dose, predicting the potential for once-daily dosing [51]. The absolute oral bioavailability of 1 in humans is 67%, and the compound is ~99% bound to plasma proteins [77,78,79,80]. In this trial, 1 was safe and well tolerated at all of the administered doses with no clinically significant adverse effects observed, a profile that set the stage for a proof-of-concept study in HCV GT-1-infected subjects. Doses of 1, 10, and 100 mg were administered in a randomized, double-blind, placebo-controlled, single ascending dose format similar to that used for the NHV study, and plasma HCV RNA levels were monitored until 172 h post-dose. The results of this study are compiled in Table 4 with mean plasma HCV RNA declining by 1.8 log10 IU/mL 24 h following the 1 mg dose, while the 10 and 100 mg doses provided increased efficacy, with viral load declines of 3.2 and 3.3 log10 IU/mL, respectively, measured at 24 h. The mean maximal viral load reduction in the 100 mg dose cohort was 3.6 log10 IU/mL with HCV RNA measured at 35 IU/mL in one of the GT-1b-infected subjects, while plasma RNA in another was below the lower limit of quantification (25 IU/mL) at 144 h post-dose. The decline in plasma viral RNA concentration following administration of the 10 and 100 mg doses of 1 was both rapid and profound in nature, with a mean reduction of 1.95 log10 IU/mL measured at 6 h post-dose for nine of the patients [49, 51]. The steep decline in viral load was subsequently explained after the development of a multiscale model of viral kinetics that took into account the effects of 1 on both viral replication and virus assembly and secretion, with the latter being the source of an immediate effect on virion production. The mean effectiveness of 1 on virus RNA production and virion assembly/secretion was estimated to be 99 and 99.8%, respectively, and yielded an estimate of plasma HCV half-life of 45 min, significantly shorter than the 2.7 h half-life estimated from an analysis of viral kinetics during treatment with older, interferon-based therapies [49].

Table 4 Dose, HCV genotype distribution, and plasma HCV RNA levels following administration of 1 to HCV-infected subjects

The profile of 1 was further explored in a double-blind, placebo-controlled multiple ascending dose study in which the drug was administered for 14 consecutive days to GT-1-infected subjects at doses of 1, 10, 30, 60, and 100 mg once daily and 30 mg twice daily [51]. Each panel comprised of five patients randomized such that four received drug and one was administered a placebo control, with drug PK parameters determined on days 1 and 14. Median peak plasma concentrations of 1 occurred 1–2 h after dosing, and the PK profile was supportive of once-daily dosing with a mean terminal half-life of 12–15 h and steady state achieved after 3–4 days of drug administration. The mean maximal reduction in HCV RNA levels in plasma are compiled in Fig. 3 with 30 and 60 mg QD cohorts comprised solely of GT-1a-infected subjects. In the other dosing cohorts, those infected with GT-1b virus exhibited a greater response compared to those infected with GT-1a virus. However, most patients experienced viral rebound on or before day 7 of therapy with viral RNA levels below the mean maximal decline except in the 30 mg BID cohort (Fig. 4). Rebound was typically more rapid in the GT-1a-infected subjects which can be explained by a lower genetic barrier to resistance in this subtype [81,82,83]. In HCV GT-1a, only a single base pair change in the viral RNA is typically required to code for an alternative amino acid, while GT-1b frequently requires two base pair changes for coherent coding [81,82,83]. Population sequencing indicated the appearance of mutations at Met28, Gln30, Leu31, and Tyr93 all of which had been identified as resistance mutation hotspots in response to selective pressure by 1 in replicon studies in vitro [84, 85].

Fig. 3
figure 3

Mean maximum decrease in plasma viral RNA in HCV GT-1-infected subjects following dosing of 1 for 14 days

Fig. 4
figure 4

Mean viral load reductions on days 2 and 7 compared with the maximal viral RNA decline following dosing of 1

While the results of this trial further confirmed the potential of HCV NS5A as a therapeutic target, the rapid emergence of resistance to 1 anticipated that optimal clinical application would be as part of combination therapy [79, 81,82,83,84,85,86,87,88,89,90,91]. The combination of 1 as add-on therapy to the extant standard of care, pegylated IFNα and ribavirin (53), was explored clinically in patients infected with HCV genotypes 1–4 and a subset of patients who were co-infected with HIV-1. The results indicated that sustained virologic responses could be achieved with shorter 24-week durations of therapy with a burden of side effects comparable to that of pegylated IFNα and 53 [92,93,94,95,96]. However, it was the opportunity availed by the contemporaneous development of the HCV NS3 protease inhibitor asunaprevir (54) that allowed the pursuit of a parallel clinical program that had a significant impact on the course of the clinical development of HCV therapeutic agents [97,98,99,100,101,102]. In a small open-label clinical trial conducted in HCV GT-1-infected subjects who had no evidence of cirrhosis and who had previously failed to respond to peg-IFN/53 therapy (referred to as null responders), a combination of 1 (60 mg QD PO) and 54 (600 mg BID PO) with peg-interferon α2a (180 μg per week subcutaneously) and 53 (1,000 or 1,200 mg QD PO, depending on body weight) administered for 24 weeks was compared with a regimen comprised of only the two direct-acting antiviral agents (DAAs) [99, 102]. All of the ten patients receiving the quadruple drug regimen had undetectable levels of HCV RNA in plasma measured at 12 weeks following the last dose (SVR12), while nine also achieved SVR24. One patient in this group had detectable HCV RNA in plasma at week 24, but this was not quantifiable, and viral RNA was not detected in plasma 5 weeks later [99, 102]. Of the 11 patients receiving only the dual DAA combination, five had undetectable levels of HCV RNA in plasma at the end of therapy, and four maintained this status at weeks 4, 12, and 24 after the last drug dose. This cohort was comprised of nine subjects infected with GT-1a and two infected with GT-1b, with both GT-1b-infected patients achieving SVR24, while six patients infected with GT-1a virus experienced virological breakthrough while on therapy and the remaining patient infected with GT-1a virus relapsed after completing drug therapy. This study, which was conducted in a very challenging patient population, provided the first indication that a chronic HCV infection could be cured solely by treatment with DAAs in the absence of the immune stimulation provided by peg-interferon α2a and 53 [100].

figure j

The successful treatment of HCV GT-1b infections with 1 and 54 redirected the clinical development plan for this dual combination to Japan where GT-1b is the most prevalent, accounting for approximately 70% of the estimated two million infections at the time [103, 104]. The combination of 1 and 54 has been studied extensively in GT-1b-infected Japanese patients, leading to approval of the drug combination by the Japanese Pharmaceutical and Medical Devices Agency (PMDA) on July 4, 2014 [105,106,107,108,109,110,111,112,113,114,115,116,117,118,119]. The marketing authorization of 1 as Daklinza™ and 54 as Sunvepra® in Japan represented the first approval of a combination of DAAs for the treatment of HCV infection although the combination of sofosbuvir (55) and 53 had been approved by the FDA in December of 2013 [107]. The phase III Japanese clinical trials of 1 (60 mg QD) and 54 (100 mg BID) in GT-1b-infected subjects that subtended marketing approval evaluated 24 weeks of therapy and were associated with SVR12 rates of 81% in non-responders and 87% in those intolerant of or ineligible for pegylated interferon therapy. In a multinational study conducted in a broad GT-1b patient population, the SVR12 rates were 90% in those naïve to therapy and 82% in those intolerant of or ineligible for interferon-based therapies. The pre-existence of the Tyr93His polymorphism in the HCV NS5A gene was a predictor of lower clinical efficacy, with the SVR rates declining to 45 from 95% in those patients that harbored this mutation at baseline, which has a 15% prevalence in the Japanese patient population.

Broadening the utility of 1 and 54 to treat HCV GT-1a infections required the addition of a third agent, the allosteric RNA-dependent RNA polymerase inhibitor beclabuvir (56) which was developed as a fixed-dose combination comprising of 30 mg of 1, 200 mg of 54, and 75 mg of 56 administered as a BID regimen for 12 weeks [120,121,122,123,124,125,126]. In the UNITY 1 international study which was conducted in 415 non-cirrhotic patients with HCV GT-1 infection, 91% of the patients achieved SVR12. SVR12 rates of 92% were observed in treatment-naive patients and 89% in treatment-experienced patients, while virologic failure occurred in 8% of the patients. In the UNITY-2 phase III study conducted in the United States, Canada, France, and Australia in 202 GT-1-infected patients with compensated cirrhosis, the SVR12 rates were 93% for patients in the treatment-naive group and 87% for those in the treatment-experienced group. SVR12 rates were improved to 98% for patients in the treatment-naive group and 93% for those in the treatment-experienced group when 53 was included in the regimen. In a phase III trial (UNITY 3) conducted in 217 Japanese patients infected with GT-1 HCV, SVR12 rates of ≥95% were achieved in both treatment-naive (n = 152) and interferon-experienced (n = 65) cohorts after 12 weeks of therapy. The SVR12 rates were similar across the patient subgroups evaluated that included patients with cirrhosis and those aged ≥65 years. These studies contributed to the approval of the fixed-dose combination of 1, 54, and 56 for marketing in Japan on December 20, 2016.

A number of clinical studies have demonstrated that co-dosing of 1 with the nucleoside-based NS5B RNA-dependent RNA polymerase inhibitor 55, with and without 53, achieves a high cure rate across HCV genotypes and patient population groups, including in those with comorbidities such as HIV-1 infection [127,128,129]. In a compassionate use program that reflected a real-world experience, including some subjects with advanced liver disease that would have been excluded from phase III studies, the combination of 1 and 55 (with and without 53) demonstrated a high efficacy [130]. In addition, long-term follow-up studies have demonstrated a 99% durability for the SVR12 associated with various regimens that include 1 [131].

Daclatasvir (1) has been approved in more than 60 countries for use in combination with 54, 56, or other HCV inhibitors, including 55 [132]. A combination of all four of these agents has been evaluated in GT-1-infected patients as a drug intensification approach to shortening the duration of therapy to 4 or 6 weeks [133]. However, while the majority (96%) of patients experienced undetectable levels of HCV RNA at the end of therapy, relapse occurred in 77% of those treated for 4 weeks and 43% of those subject to 6 weeks of treatment with quadruple therapy [133].

figure k

5 Conclusion

The NS5A replication complex inhibitor class of HCV inhibitor has become established as a critical component of all of approved pan-genotypic DAA combination therapies [132]. The discovery of 1, the prototype NS5A inhibitor that is the founding member of the class, was identified only after considerable optimization of a lead discovered by phenotypic screening, a powerful approach to drug discovery that has proven to be well-suited as a means of identifying mechanistically novel antiviral agents [134,135,136,137,138].