The Presence of a Serine Protease in HCV NS3 Protein

In 1989, two groups presented seminal comparative sequence studies suggesting the presence of a trypsin/chymotrypsin-like serine protease in the N-terminal one-third of the NS3 protein of flaviviruses and pestiviruses (Bazan and Fletterick, 1989; Gorbalenya et al., 1989a). Although the identity of the HCV proteins had not been determined yet at that time, these groups showed that HCV might encode a homologous trypsin/chymotrypsin-like serine protease as well. A catalytic triad of His¹⁰⁸³, Asp¹¹⁰⁷, and Ser¹¹⁶⁵ (based on the polyprotein number of HCV-H strain) (Fig. 2A) was identified by sequence alignment of the HCV NS3 protein with many known viral and cellular serine proteases, which belong to the trypsin/chymotrypsin superfamily of serine proteases (Bazan and Fletterick, 1989; Bazan and Fletterick, 1990; Gorbalenya et al., 1989a; Miller and Purcell, 1990). In this chapter, a numerical system starting with the N terminus of the HCV NS3 protein itself will be used, in which the number of this catalytic triad would be His⁵⁷, Asp⁸¹, and Ser¹³⁹. Mutagenesis studies of the catalytic triad by several laboratories showed that the HCV NS3 serine protease is necessary for cleavage at four junctions in the HCV polyprotein, including NS3/NS4A, NS4A/NS4B, NS4B/NS5A, and NS5A/NS5B sites (Bartenschlager et al., 1993; Eckart et al., 1993; Grakoui et al., 1993a; Hijikata et al., 1993a; Manabe et al., 1994; Tomei et al., 1993). Substitution of any of the catalytic triad residues, His⁵⁷, Asp⁸¹, or Ser¹³⁹, abolished the cleavage at these four junctions, but have no effect on cleavage at other sites of the HCV polyprotein (C/E1, E1/E2, E2/p7, p7/NS2, and NS2/NS3), all of which are located upstream of the four NS3 serine protease-dependent cleavage sites (Fig. 1). Cleavage at the NS2/NS3 junction requires the presence of both NS2 and the N-terminal 180 residues of NS3, i.e., the serine protease domain, but not the catalytic triad of the serine protease per se. The NS2-3 protease will be the topic of another chapter in this book (reviewed in chapter 5).

Fig. 2

Sequence alignment of the HCV NS3-4A serine protease and its substrates. The genotypes (1a, 1b, 2a, 2b, and 3a) are indicated at the left, and the strain names are shown in a bracket. (A) HCV NS3 (more...)

The Minimal Domain of Serine Protease: N-Terminal 180 Amino Acids

The 631-residue HCV NS3 protein is a dual-function protein, containing the trypsin/chymotrypsin-like serine protease in the N-terminal region and a helicase in the C-terminal region (Gorbalenya and Koonin, 1993; Gorbalenya et al., 1989b). Co-transfection studies using constructs in which either the catalytic triad of the serine protease was mutated or the entire protease domain was deleted demonstrated that the NS3 serine protease domain, in the absence of its C-terminal helicase counterpart, is capable of mediating cleavage of polyprotein substrates (Bartenschlager et al., 1994; Lin et al., 1994; Tanji et al., 1994b). The minimal sequences required for a functional serine protease activity were determined by these groups to be the N-terminal 180 amino acids of the NS3 protein. Deletion of up to 14 residues from the N terminus of the NS3 protein is tolerated, although a further deletion of the N-terminal 22 amino acids resulted in significantly poorer processing of HCV polyprotein. On the other hand, deletions from C terminus of this minimal serine protease domain completely abolished proteolytic activity (Bartenschlager et al., 1994; Failla et al., 1995; Tanji et al., 1994b).

The HCV NS3 Serine Protease-Mediated Cleavage of HCV Polyprotein

The N-terminal amino acid sequences of the HCV NS3 serine protease-dependent cleavage products, including NS4A, NS4B, NS5A, and NS5B were determined by radioactive labeling of the HCV polyprotein with specific amino acids followed by N-terminal sequencing (Grakoui et al., 1993a). The nomenclature of Schechter and Berger (Schechter and Berger, 1967), which has been widely accepted for description of the proteases and the corresponding substrates, will be used for the HCV NS3-4A serine protease, its substrates and inhibitors in this review. A decapeptide substrate for the HCV NS3-4A protease, with 6 residues on the N-terminal side and 4 residues on the C-terminal side (Fig. 2C and 4A), would be described as NH₂-P6-P5-P4-P3-P2-P1-P1’-P2’-P3’-P4’-OH, with the scissile bond locating between the P1 and P1’ residues. The cleavage products of this substrate would be the P-side product (NH₂-P6-P5-P4-P3-P2-P1-OH) and the P’-side product (NH₂-P1’-P2’-P3’-P4’-OH). The pockets on the HCV NS3-4A protease in which the P1 side chain binds would be called the S1 pocket, and so forth for S2, S3, S4, S5, or S6 pockets and S1’, S2’, S3’, or S4’ pockets. Sequence alignment of these cleavage sites among various HCV isolates indicates that a consensus sequence would include the following elements: an acidic residue (Asp or Glu) at the P6 position, a thiol-terminating residue (Ser for the NS3/NS4A and Thr for the other three sites) at the P1 position, and a small side chain residue (Ala or Ser) at the P1’ position (Fig. 2C).

Fig. 4

Chemical diagrams of the HCV NS3-4A protease inhibitors. (A) Chemical structures of the natural HCV NS5A/NS5B deca-peptide substrates (from P6 to P4’), and two NS3-4A protease inhibitors, (more...)

Site-directed mutagenesis studies showed that blockage of any one of these four junctions has little or no impact on cleavage at other sites (Kolykhalov et al., 1994; Leinbach et al., 1994), which is consistent with kinetic analyses that have demonstrated a preferential, but clearly not obligatory, order of cleavage of the NS polyprotein by the NS3-4A serine protease (Bartenschlager et al., 1994; Lin et al., 1994; Tanji et al., 1994a). However, proteolysis of the NS3/NS4A junction is believed to be a co-translational, cis-cleavage event since an NS3-NS4A precursor were not detected and this cleavage was insensitive to dilution (Bartenschlager et al., 1994; Lin et al., 1994; Tanji et al., 1994a). The other three junctions can be cleaved by the NS3-4A serine protease in trans (Bartenschlager et al., 1994; Failla et al., 1994; Lin et al., 1994; Tanji et al., 1994a; Tanji et al., 1994b). However, it is possible that all these proteolysis events occur in a localized, i.e. cis-cleavage environment, since the NS3-4A protease is expressed as part of the same polyprotein molecule as all of its substrates. In trans-cleavage reactions, processing at the NS5A/NS5B site occurred more rapidly than those at the NS4A/NS4B and NS4B/NS5A sites since rather stable NS4A-NS4B-NS5A processing intermediates were detected (Bartenschlager et al., 1994; Failla et al., 1995; Lin et al., 1994). Two additional cleavage sites in NS4B and NS5A have been identified (Kolykhalov et al., 1994; Markland et al., 1997) although their significance in viral replication is unclear. The first one is located near the N terminus of NS4B protein and its cleavage was observed only when the NS4A/NS4B cleavage was blocked (Kolykhalov et al., 1994). The second one, in the middle of NS5A protein, was seen in cell-free proteolysis experiments (Markland et al., 1997).

HCV NS4A Protein Is a Cofactor Essential for Activity of the NS3 Serine Protease

It was demonstrated during these trans-cleavage and processing kinetics studies that an additional HCV-encoded protein, NS4A, is required as an activating cofactor for the optimal activity of the NS3 serine protease (Bartenschlager et al., 1994; Bartenschlager et al., 1995b; Failla et al., 1994; Lin et al., 1994; Tanji et al., 1995). In the absence of NS4A protein, only the NS5A/NS5B site, but not the other three cleavage sites (NS3/NS4A, NS4A/NS4B, and NS4B/NS5A), was partially processed by the NS3 serine protease alone. The presence of NS4A not only enables efficient processing at these three junctions and but also resulted in enhancement of the NS5A/NS5B cleavage. Deletion analysis showed that the central region (residues 21 to 34) of the NS4A (Fig. 2B), a 54-residue protein, is essential and sufficient for the cofactor function of the NS3 serine protease (Bartenschlager et al., 1995b; Failla et al., 1995; Lin et al., 1995; Satoh et al., 1995; Tanji et al., 1995). In addition, NS4A forms a non-covalent complex with the NS3 serine protease, which was stable in the presence of non-ionic detergent (Bartenschlager et al., 1995b; Failla et al., 1995; Hijikata et al., 1993a; Hijikata et al., 1993b; Lin et al., 1995; Satoh et al., 1995). Deletion studies indicate that the N-terminal 22 residues of NS3 and the above-mentioned central region of NS4A is involved in interaction between two proteins (Bartenschlager et al., 1995b; Failla et al., 1995; Koch et al., 1996; Lin et al., 1995; Satoh et al., 1995; Tanji et al., 1995). Substitutions that disrupted the interaction between NS3 and NS4A also resulted in reduction or loss of protease activity, suggesting that formation of an NS3-NS4A complex could be a pre-requisite for a functional serine protease (Butkiewicz et al., 1996; Koch et al., 1996; Lin et al., 1995; Shimizu et al., 1996; Steinkühler et al., 1996a; Tomei et al., 1996).

NS3 Serine Proteases and Their Cofactors in the Flaviviridae Family

The activation of a virus-encoded protease by peptide(s) derived from another viral protein is not uncommon for viruses. This phenomenon seems to be a common feature of members of Flaviviridae family. As mentioned earlier, comparative sequence analysis suggests the presence of a chymotrypsin-like serine protease in the N-terminal one-third of NS3 protein of flaviviruses, pestiviruses, and HCV, which accounts for all three known genera of the Flaviviridae family at that time (Bazan and Fletterick, 1989; Bazan and Fletterick, 1990; Gorbalenya et al., 1989a; Miller and Purcell, 1990). A chymotrypsin-like serine protease was later identified in the newly discovered 4^th genus of this family, hepatitis G virus (HGV) or GB virus (Leary et al., 1996; Scarselli et al., 1997). In each member of the Flaviviridae family that has been studied so far, a virus-encoded cofactor is needed to activate the corresponding NS3 serine protease. It is the NS4A protein, located immediately downstream of the NS3 protein, in the case of HCV, bovine viral diarrhea virus (BVDV) (Wiskerchen and Collett, 1991; Xu et al., 1997) of pestiviruses, and HGV or GB virus (Butkiewicz et al., 2000; Sbardellati et al., 2000). For flaviviruses, it is the NS2B protein, located immediately upstream of the NS3 protein, that fulfills the role of activator for the corresponding NS3 serine protease of yellow fever virus (Chambers et al., 1991) or dengue virus (Cahour et al., 1992; Falgout et al., 1993; Falgout et al., 1991). Again, a short peptide corresponding the central region of the NS4A cofactor of BVDV (Tautz et al., 2000) or GB virus (Butkiewicz et al., 2000), or a central 40-residue fragment of the dengue NS2B protein (Falgout et al., 1993) was sufficient for activation of the corresponding NS3 serine protease. It should be noted that there is little sequence homology between different genera with regard to these activators of the NS3 serine proteases. In fact, the NS3 serine proteases of different genera of the Flaviviridae family have quite distinct specificity for substrates, especially on the P1 residues: Cys or Thr for HCV (Grakoui et al., 1993a) and HGV or GB virus (Scarselli et al., 1997), Leu for pestiviruses (Xu et al., 1997), and Lys or Arg for flaviviruses (Chambers et al., 1991).

Adenovirus Protease Has Three Activating Cofactors

The examples of virus-encoded proteases and their activating cofactors are not limited to Flaviviridae family. In fact, the very first example was discovered in human adenovirus. A cysteine protease encoded by human adenovirus (AVP) (for a review see Mangel et al., 2003) was found in virions and activated by an 11-residue peptide cofactor from the C-terminus of pVI protein (pVIc) (Mangel et al., 1993; Webster et al., 1993). In contrast to the HCV serine protease, the pVIc cofactor is covalently linked to the AVP by a disulfide bond (Ding et al., 1996). This protease requires binding of not one, but two co-factors, pVIc and adenovirus DNA for the optimal activity (Baniecki et al., 2001; Mangel et al., 1993; Webster et al., 1993). One possible explanation for the DNA cofactor is that AVP moves along the viral DNA looking for precursor protein cleavage sites much like RNA polymerase moves along DNA looking for promoter (McGrath et al., 2001). In addition, actin, a cellular cytoskeletal protein, was found to activate AVP as well (see below).

Virus-Host Interaction of HCV NS3-4A Serine Protease

Several recent studies suggest that HCV NS3-4A protease could be one of the weapons that HCV uses to breakdown the host antiviral response (reviewed in chapter 13). Innate immune response is trigged upon recognition by a family of toll-like receptor (TLR) of certain pathogen-associated molecular patterns. TLR-3 binds one of these patterns, dsRNA, a replication intermediate of many viruses, and initiates a massive type-I IFN-mediated antiviral response through activation of IFN regulatory factor 3 (IRF-3), a transcription activator. Recently, it was shown that binding of dsRNA to two DExD/H box RNA helicases, retinoic acid inducible gene I (RIG-I) and Helicard (Mda5) in cytoplasm induces a TLR-3-independent IFN response pathway, through activation of IRF-3 and NF-κB. The fact that the Huh7 hepatoma cell line, which is deficient in TLR-3-dependent pathway, became highly permissive for replication of HCV replicon RNA when an inactivating mutation in the RIG-I gene was selected in a subclone, Huh7.5 (Sumpter et al., 2005) suggests that both pathways play a role in anti-HCV immune responses. It was reported that expression of active HCV NS3-4A protease blocked activation of both TLR-3-dependent (Foy et al., 2003) and TLR-3-independent (Foy et al., 2005) signal transduction cascades, and therefore prevented the IFN-induced antiviral response against HCV RNA replication. In the case of TLR-3-dependent pathway, the substrate of HCV NS3-4A protease was reported to be Toll-IL-1 receptor domain-containing adaptor inducing IFN-β (TRIF or TICAM-1) (Li et al., 2005). TRIF, a critical player in the TLR-3-dependent pathway, recruits two kinases, TBK1 and IKKɛ, to phosphorylate and activate IRF-3. In the case of TLR-3-independent cascade, the HCV NS3-4A protease was reported to cleave Cardif, a new CARD-domain containing adaptor protein that interacts with RIG-I and recruits IKKα, IKKβ, and IKKɛ kinases, resulting in activation of IRF-3 and NF-κB (Meylan et al., 2005).

Interference of host function by virus-encoded proteases is not limited to HCV. Actin, a cellular cytoskeletal protein, was found to interact and activate AVP (Brown et al., 2002). More specifically, it is an 11-residue peptide corresponding to the C terminus of actin, which is highly homologous to pVIc, was shown to bind and stimulate AVP activity (Brown and Mangel, 2004). The activation of AVP by actin has been proposed to be a mechanism for adenovirus to facilitate the cleavage of cytoskeletal proteins, preparing the infected cells for lysis and release of nascent virions (Brown and Mangel, 2004). In the presence of actin, the cellular skeletal protein, AVP is activated and then is able to mediate proteolysis of cytokeratin 18, another cytoskeletal protein and, not surprisingly, actin itself (Brown et al., 2002).

A Double β-Barrel Fold of the HCV NS3-4A Serine Protease

In 1996, two groups presented seminal studies showing the X-ray structures of the HCV NS3 serine protease domain (Kim et al., 1996; Love et al., 1996). Kim et al. solved an X-ray structure of NS3 serine protease domain (residues 1 to 181) of the HCV H strain of genotype 1a in a non-covalent complex with a NS4A cofactor peptide (residues 21 to 39) (Kim et al., 1996). The HCV serine protease forms a double-barrel fold (Fig. 3A), which is similar to that of serine proteases from the chymotrypsin/trypsin super-family. The catalytic triad is located in a cleft between two sub-domains (or barrels), with His⁵⁷ and Asp⁸¹ in the N-terminal sub-domain and Ser¹³⁹ in the C-terminal one. The C-terminal sub-domain (residues 96–180) contains the conventional six-stranded β barrel, common to most members of the chymotrypsin family, followed by a structurally conserved α helix. The N-terminal sub-domain (residues 1–93) consists of eight β-strands, including seven from the NS3 protein (Fig. 3A, colored in green) and one from the NS4A peptide (Fig. 3A, colored in magenta), and the latter one is sandwiched between two β-strands of the N-terminal sub-domain of NS3. In addition, the C-terminal sub-domain contains a tetrahedrally coordinated metal ion, presumably a zinc atom, located at one end of the β barrel, opposite to the catalytic triad.

Fig. 3

X-ray structures of the HCV NS3-4A serine protease. (A) Ribbon diagram of the NS3 protease domain in a complex with an NS4A cofactor peptide. The N-terminal sub-domain of the NS3 protease (in green), (more...)

Effects of NS4A Binding

The extensive interaction between NS3 and NS4A results in a tightly packed β-barrel and buries an additional 2400 Ų of surface area of the NS3 protease (Fig. 3C). All but two of the main-chain carbonyl and amide groups of the NS4A residues 23–31 (Fig. 2B) form hydrogen bonds with NS3. Hydrophobic side chains of several conserved NS4A residues (Val²³, Ile²⁵, Ile²⁹, and Leu³¹) (Fig. 2B) are buried in hydrophobic core of the N-terminal sub-domain of NS3 (Kim et al., 1996) (Fig. 3C). These observations in the X-ray structure confirm the previous deletion and mutagenesis studies, in which the central region of NS4A (residues 21–33), and in particular, the several conserved hydrophobic residues (Ile²⁵, Ile/Val²⁹), were shown to be critical for optimal binding of NS4A to the NS3 serine protease and activation of the protease activity (Butkiewicz et al., 1996; Lin et al., 1995; Shimizu et al., 1996; Tanji et al., 1995). The mechanism of activation by the NS4A cofactor on the NS3 serine protease activity were best illustrated by a direct comparison of three X-ray structures of HCV NS3 protease domain. In an X-ray structure of uncomplexed NS3 serine protease domain (residues 1 to 189) of the HCV BK strain of genotype 1b, solved in the absence of an NS4A cofactor, the N-terminal sub-domain of NS3 protease has only six β-strands and the N-terminal 28 residues of NS3 are extending away from the core of the protein as a flexible loop (Love et al., 1996). In two X-ray structures of the NS3 protease-NS4A cofactor complex of HCV H strain (Kim et al., 1996) or BK strain (Yan et al., 1998), respectively, the first 28 NS3 residues fold into an α-helix and a β-strand. This additional NS3 β-strand is spatially located next to the β-strand of NS4A, and both contribute to the formation of an eight-strand β-barrel (Fig. 3A). These observations in X-ray structures are consistent with earlier results that deletion of the N-terminal 22 residues of NS3 resulted to loss of the NS4A binding (Bartenschlager et al., 1995b; Failla et al., 1995; Koch et al., 1996; Satoh et al., 1995). The more striking effect of NS4A binding is on the orientation of the catalytic triad, which is highly conserved in the chymotrypsin serine protease family so that the Asp residue stabilizes the His residue after the imidazole ring of His deprotonate the OH group of the Ser nucleophile. In the absence of the NS4A cofactor, the conformation of the triad is significantly distorted so that the protease is not expected to have an appropriate activity, as observed in biochemistry studies. In the X-ray of NS3 serine protease domain alone, the imidazole ring of His⁵⁷ is located too far away to effectively deprotonate the nucleophilic OH of Ser¹³⁹ or to be stabilized by the Asp⁸¹ (Love et al., 1996). In the presence of NS4A cofactor peptide, the catalytic triad is in the characteristic position expected for a chymotrypsin-like serine protease (Kim et al., 1996; Yan et al., 1998).

Although both HCV serine protease and adenovirus cysteine protease require the binding of the corresponding activator(s) to be fully functional, the mechanisms of cofactor binding and activation are quite different. In the X-ray structure of human adenovirus-2 protease in a complex with its pVIc cofactor, the catalytic triad of AVP (Cys-His-Glu) adapted an arrangement similar to that of papain, and the pVIc cofactor extends a β-sheet, which is distant from the active site (Ding et al., 1996). Another difference is that pVIc binds to AVP by covalent disulfide bond and forms a 6^th strand on the β-sheet (McGrath et al., 2003).

Zinc-Binding Site

Alignment of amino acid sequences showed that three Cys and one His residues are well conserved in the NS3 protease domain of various HCV strains (Fig. 2A) and closely related GB viruses or hepatitis G viruses. Computational modeling analysis suggests that these four residues could form a zinc-binding site, which is located opposite to the catalytic triad in the NS3 serine protease model (Failla et al., 1996). This prediction was quickly confirmed by both X-ray structural and biochemical studies. In the X-ray structure of the NS3 serine protease domain, a zinc ion is tetrahedrally coordinated by four residues, Cys⁹⁷, Cys⁹⁹, Cys¹⁴⁵, and through a water molecule, His¹⁴⁹ (Fig. 3B) (Kim et al., 1996; Love et al., 1996). Substitution of any of the three Cys residues involved in zinc-binding with Ala led to significantly reduced protease activity, whereas mutation at His¹⁴⁹ had much less effect (Hijikata et al., 1993a). This zinc-binding pocket is located in a cleft between two β-barrels and at the opposite end of the catalytic triad (Fig. 3A). The zinc atom is at least 20 Å away from the catalytic Ser¹³⁹, suggesting the coordination of a zinc atom plays more a structural role rather than a catalytic function, as in case of several other viral proteases, including poliovirus and rhinovirus 2A cysteine proteases (Sommergruber et al., 1994; Voss et al., 1995; Yu and Loyd, 1992). A similar arrangement of Cys and His residues, as in Cys-X-Cys……Cys-X-His, where X is any residue, is also found in the cysteine protease 2A of the Picornavirae family. In these picornavirus 2A cysteine proteases, a tightly bound zinc atom is found to be critical for integrity and stability of the properly folded protein structure, rather than proteolysis. In many other serine proteases, such as chymotrypsin, a disulfide bond is found in a similar location, suggesting the zinc-binding pocket in the HCV NS3 serine protease plays a similar role as the disulfide bond in stabilizing the relative position of two β-barrels. Additional confirmation came from biochemical experiments, which showed that the purified, active HCV NS3 protease domain contained an equimolar amount of zinc (De Francesco et al., 1996; Stempniak et al., 1997). NS3 protein expressed in the absence of zinc in E. coli was not folded properly and therefore deficient of the protease activity (De Francesco et al., 1996; Stempniak et al., 1997).

The Substrate Binding Groove

The HCV NS3-4A serine protease retains some highly conserved features of the chymotrypsin family, such as spatial location of the catalytic triad of His⁵⁷, Asp⁸¹, and Ser¹³⁹, as well as the positions of backbone amides of Gly¹³⁷ and Ser¹³⁹, which forms the oxyanion hole (Kim et al., 1996). A twisted strand (residues Arg¹⁵⁵-Ala¹⁵⁶-Ala¹⁵⁷-Val¹⁵⁸) of the HCV protease superimposes well with the corresponding strand of residues in other serine proteases in the chymotrypsin family, which makes hydrogen bonds with the P3 carbonyl and the P1 and P3 amides of peptidomimetic inhibitors of serine proteases (Edwards and Bernstein, 1994). However, the HCV serine protease does display some significant difference from other serine proteases, such as chymotrypsin. For examples, several loops that interact with P4, P3, and P2 moieties of inhibitors and form a well-defined substrate-binding pocket in many other serine proteases are either shortened significantly or deleted in the HCV serine protease. Lack of side chain interaction is compensated by an extensive network of hydrogen bonds between the main chain atoms of the protease and substrate. However, the absence of these long loops results in a shallow, solvent-exposed substrate-binding groove and renders the design of small-molecule, peptidomimetic inhibitors against the HCV NS3-4A serine protease an extremely challenging task.

The S1 Pocket

Sequence alignment of four cleavage sites in the HCV polyprotein indicates that there are three conserved positions in the HCV NS3-4A protease substrates: an Asp or Glu at P6, a Cys or Thr at P1, a Ser or Ala at P1’ (Fig. 2C). Three of the cleavage sites, NS4A/NS4B, NS4B/NS5A, and NS5A/NS5B, have a Cys at the P1 position, while the NS3/NS4A site has a P1 Thr residue. In the X-ray structures, the S1 pocket is primarily determined by three hydrophobic residues, Leu¹³⁵, Phe¹⁵⁴, and Ala¹⁵⁷ (Kim et al., 1996; Love et al., 1996). Phe¹⁵⁴ is located at the bottom of the S1 pocket and clearly in position to make favorable van der Waals interactions with the P1 side chain. The small and hydrophobic nature of this S1 pocket is complementary to a relatively small and lipophilic side chain of a Cys. In addition, it is known that the sulfhydryl group of a Cys residue forms a favorable electrostatic interaction with the aromatic ring of Phe (Burley and Petsko, 1988). Proteolysis of substrates with larger P1 residues was allowed when Phe¹⁵⁴ was substituted with a residue that has a smaller side chain, such as Thr, along with replacement of Ala¹⁵⁷ with a Gly, which altered the specificity of the mutated NS3 protease (Failla et al., 1996; Koch and Bartenschlager, 1997).

The S6 Pocket

All HCV NS3-4A cleavage sites contain an acid residue (usually Asp) at the P6 position (Fig. 2C). The P6 acid residue, sometimes along with an acidic residue at the P5 position, is believed to form electrostatic interactions with a cluster of positively charged residues of the NS3 protease, Arg¹²³, Arg¹⁶¹ and Lys¹⁶⁵ (Koch et al., 2001; Steinkühler et al., 2001).

Evolution of In vitro Biochemical Assays

While cell-based expression and proteolytic processing experiments were useful for providing the first glimpses of the activity of the HCV NS3-4A serine protease, their limitations were quickly reached given the complexity of cellular environment. Further understanding of the functions of this protease and substrate, which would enable meaningful drug discovery efforts, required establishment of efficient biochemical assays. In the early-generation cell-free trans-cleavage assays, the protease, the substrate or both were derived from cell lysates or in vitro translation, which may or may not have been coupled with in vitro transcription (Bouffard et al., 1995; Hahm et al., 1995; Koch et al., 1996; Lin and Rice, 1995; Suzuki et al., 1995; Tomei et al., 1996). Later, in vitro translated polyprotein substrates were used for monitoring activity of purified NS3 protease, which was generated as recombinant proteins in E. coli, baculovirus, or yeast expression systems (Butkiewicz et al., 1996; D’Souza et al., 1995; Markland et al., 1997; Steinkühler et al., 1996a). However, these assays using in vitro translated polyproteins are much less quantitative than those using synthetic peptides as substrates, which are cleaved by the HCV serine protease and the products are analyzed on high performance liquid chromatography (HPLC) (Bianchi et al., 1996; Inoue et al., 1998; Kakiuchi et al., 1995; Landro et al., 1997; Shimizu et al., 1996; Steinkühler et al., 1996a; Steinkühler et al., 1996b; Sudo et al., 1996; Urbani et al., 1997; Zhang et al., 1997). A colorimetric substrate, EDVVα AbuC-p-nitroanilide (5A-pNA), in which the P2 Cys was substituted with Abu and the whole P’-side replaced with a colorimetric leaving group, pNA, was used to increase assay convenience (Landro et al., 1997). It was observed when the scissile amide bond was replaced with an ester linkage, which allows ready trans-esterification of the scissile bond to the acyl-enzyme intermediate, these substrates displayed much improved catalytic efficiency (k_cat/K_m), allowing detection of activity with sub nM of NS3 protease (Bianchi et al., 1996). An internally quenched fluorogenic donor/acceptor couple, based on resonance energy transfer, was then incorporated into a depsipepetide substrate, which was shown to be suitable for high-throughput screening (Taliani et al., 1996).

Role of NS4A Cofactor

It was shown in these studies that a synthetic peptide corresponding to the central region (residues 21–34) of NS4A could be used to substitute for the full-length NS4A protein in activation of the HCV NS3 serine protease domain (Butkiewicz et al., 1996; Lin et al., 1995; Shimizu et al., 1996; Steinkühler et al., 1996a; Tomei et al., 1996). Substitution study of NS4A peptide indicates that the following residues of NS4A, Val²³, Gly²⁷, Arg²⁸, and in particular, Ile²⁵ or Ile²⁹ are critical for its cofactor function (Butkiewicz et al., 1996; Koch et al., 1996; Lin et al., 1995; Shimizu et al., 1996; Tomei et al., 1996). The activator function of NS4A peptide was largely due to an increase in the turnover rate (k_cat) with a small or little change in substrate affinity (K_m) for substrates corresponding to all three cleavage sites that can be cleaved in trans, namely, NS4A/NS4B, NS4B/NS5A, and NS5A/NS5B (Landro et al., 1997; Shimizu et al., 1996; Steinkühler et al., 1996a; Steinkühler et al., 1996b). The increase is more dramatic (>100-fold) for the less efficient substrate, the NS4B/NS5A substrate (Steinkühler et al., 1996b). An 1:1 stoichiometric amount of NS3 protease domain protein and NS4A peptide is sufficient for the maximal activation, suggesting the active form of HCV serine protease is a heterodimer (Steinkühler et al., 1996b). The dissociation constant (k_d) was determined to be in low μM range, but the association rate (k_on) was very low, suggesting conformational transitions to be the rate limiting event for the formation of NS3-4A protease complex (Bianchi et al., 1997).

Substrate Specificity

In enzymatic assays, a decapeptide substrate, with 6 residues on the P-side and 4 more on the P’-side, was found to be optimal for efficient proteolysis by the HCV NS3-4A protease. Further truncation from either the P-side or the P’-side resulted in a significantly drop in catalytic efficiency (Landro et al., 1997; Steinkühler et al., 1996a; Zhang et al., 1997). Sequence alignment of natural decapeptide substrates for the HCV NS3-4A serine protease revealed a conserved acidic residue (Asp or Glu) at the P6, a Cys or Thr at the P1, and a Ser or Ala at the P1’, resulting in the consensus substrate sequence of (Asp/Glu)–X–X–X–X–(Cys/Thr)↓(Ser/Ala)–X–X–X, whereas X indicates an variable residue (Grakoui et al., 1993a). In cell culture transfection experiments, it was found that the P6 acidic residue was dispensable and substitutions at the P1’ were reasonably tolerated. However, most mutations introduced at the P1 inevitably resulted in a significant loss of proteolytic processing, indicating that the P1 Cys or, to a less extent, Thr, is the major determining factor for substrate recognition (Bartenschlager et al., 1995a; Kolykhalov et al., 1994; Komoda et al., 1994; Leinbach et al., 1994; Tanji et al., 1994a). The preference for the peptide substrate sequence in enzyme assay is also reminiscent of the consensus sequences of the HCV natural polyprotein substrates. The optimal peptide substrate in enzyme assay has an acidic residue (Glu or Asp) at P6, a Cys at P1, and a Ser or Ala at P1’ (Landro et al., 1997; Urbani et al., 1997; Zhang et al., 1997). The other residues besides these three key positions (P6, P1, and P1’) may also play roles in recognition by the NS3-4A serine protease, as evidenced by the drastically different catalytic efficiency in the following order: NS5A/NS5B > NS4A/NS4B >> NS4B/NS5A (Landro et al., 1997; Steinkühler et al., 1996b). These differences in catalytic efficiency are consistent with the processing order and polyprotein intermediates observed in cell culture.

Cross Talk between the Protease and Helicase Domains of Ns3?

In Flaviviridae family, the NS3 protein is invariably an at least bi-functional protein, with the serine protease in the N terminal one-third and the helicase in the C-terminal two-thirds. It is not unreasonable to suggest that there may be cross communication between these two enzyme functions residing in the same protein. Indeed, polynucleotides, especially poly(U), which are stimulants for the ATPase activity of the NS3 helicase, also enhanced the serine protease activity in a full-length NS3-4A protein complex purified from over-expressing COS cells (Morgenstern et al., 1997). Because the poly(U) did not stimulate the protease activity of the purified NS3 serine protease domain, these results suggest that poly(U) enhances the protease activity of the full-length NS3-4A complex through its interaction with the helicase domain, and the latter then interacts and stimulates the protease domain. The presence of the serine protease domain seems to be required for optimal binding of poly (U) to the helicase domain because the dissociation constant, K_d, of poly(U) was 10-fold lower against full-length NS3 protein than that against the helicase domain alone (Kanai et al., 1995). However, another study showed that poly (U) inhibited the protease activity of both protease domain alone and full-length NS3, in the presence of a synthetic NS4A cofactor peptide (Gallinari et al., 1998).

Biochemistry of Adenovirus Protease, AVP

The pVIc peptide also forms a 1:1 complex with the adenovirus protease, AVP, and enhances its catalytic efficiency by hundreds fold, while the addition of adenovirus DNA increased the k_cat/K_m even further. Both cofactors enhance AVP activity by increasing k_cat, not by decreasing K_m (Baniecki et al., 2001; Mangel et al., 1996; McGrath et al., 2001). These two cofactors bind to the AVP with a dissociation constant (Kd) in the μM range, 4.4 μM for pVIc and 0.09 μM for a 12-mer ssDNA (Baniecki et al., 2001). On the other hand, actin binds to the adenovirus protease AVP with a much lower equilibrium dissociation constant, 4.2 nM, than those exhibited by two viral, nuclear cofactors for AVP, the 11-amino acid peptide pVIc and the viral DNA (Brown and Mangel, 2004). The catalytic efficiency, k_cat/K_m, for substrate hydrolysis by AVP increased 150,000-fold in the presence of actin (Brown and Mangel, 2004).

Non-Covalent, Product-Based Inhibitors

In 1998, two groups presented critical findings that, after an oligopeptide corresponding to the NS4A/NS4B or NS5A/NS5B substrate was cleaved by the HCV NS3 protease in the presence of the NS4A cofactor, the C-terminal cleavage product quickly dissociated from the enzyme but the N-terminal product was released rather slowly. This resulted in feedback inhibition by the N-terminal cleavage product, i.e., the hexa-peptide from the C terminus of the NS4A or NS5A, respectively (Llinas-Brunet et al., 1998b; Steinkühler et al., 1998). In addition, the free carboxylic group of the P1 residue, which is liberated by the cleavage of the substrate peptide bond, was recognized as an essential feature imparting selectivity with respect to other serine proteases (Llinas-Brunet et al., 1998a). The X-ray structure of a full-length NS3-4A protein provided an excellent explanation for the observed product-based inhibition (Yao et al., 1999). In the X-ray structure, the last residue of the NS3 protease, Thr⁶³¹, which is the P1 residue of the NS3/NS4A cis-cleavage site, was bound in the active site of the NS3 serine protease domain. Apparently, after the NS3 serine protease cleaves the NS3/NS4A peptide bond, the N-terminal cleavage product, with the Thr⁶³¹ as the C-terminal residue, is not released from the NS3 protease in the X-ray structure. With this seminal finding, two groups presented extensive structure-and-activity relationship (SAR) results using the hexa-peptide from the C terminus of the NS4A or NS5A, respectively. These SAR studies demonstrated that a combination of a thiol-containing residue at the P1 position and acidic residues at the P5–P6 positions is required for optimal binding and resulted in potent hexa-peptide inhibitors (Ingallinella et al., 1998; Llinas-Brunet et al., 2000). The preference displayed by NS3 protease for a thiol-containing residue, such as Cys, as the P1 anchor results from the shape of the S1 pocket. This small, hydrophilic pocket, lined by the hydrophobic residues of Val¹³², Leu¹³⁵, and Phe¹⁵⁴, is complementary to the small and hydrophilic side chain of Cys (Kim et al., 1996). In addition, the sulfhydryl (SH) group of the P1 Cys can interact in a unique way with the aromatic ring of Phe¹⁵⁴. The second anchor of the hexa-peptide inhibitors is the pair of P5–P6 acidic residues, which is thought to form electrostatic interactions with a cluster of basic amino acids of the NS3 protease, including Arg¹²³, Arg¹⁶¹, and Lys¹⁶⁵ (Di Marco et al., 2000; Koch et al., 2001).

The presence of a thiol-containing side chain in the P1 position presents a major hurdle for the chemical synthesis and stability of potential inhibitors, and therefore, limits its use in clinical development (Emery et al., 1992). Substitution of P1 Cys with amino acids containing small hydrophobic side chains, such as Ala or α-aminobutyric acid, resulted in a decline in inhibitory potency due to the sub-optimal filling of the P1 subsite. Replacement with amino acids containing larger hydrophobic side chains led to a significant loss of activity presumably due to steric hindrance (reviewed in Steinkühler et al., 2001). It was shown that a difluoromethyl group is an effective mimetic of the thiol, likely due to the similarity of their steric and electrostatic properties (Narjes et al., 2002a). In addition, 1-aminocyclopropylcarboxylic acid was shown to be an effective surrogate as well, with little impact on potency when it replaced the natural P1 Cys in a hexa-peptide inhibitors (Llinas-Brunet et al., 2000).

While the P5–P6 acidic residue pair contributes significantly to the binding of hexa-peptide inhibitors to the NS3-4A protease, it does bring two major disadvantages that render the hexa-peptide inhibitors not suitable for clinical development as oral drugs: negative charges and a rather large molecular weight (over 1,000 Daltons). The negative charges of the P5–P6 acidic pair would most likely prevent the hexa-peptides from penetrating into cells, which is reflected in the lack of cellular potency of these inhibitors in HCV replicon cells. Removal of the P5 and P6 acidic residues resulted in, as expected, a significant loss in potency of tetra-peptide inhibitors, which has to be compensated by improvement in other subsites of the inhibitors. Significant enhancement in potency was achieved with the addition of large, hydrophobic aromatic rings to the P2 Pro group, resulting in potent tetra-peptide inhibitors (Goudreau et al., 2004b). In addition, a macrocyclic ring was designed to link the side chain of the P1 and P3 residues to reduce the peptidic nature and provide rigidity to pre-order the binding conformation (Tsantrizos et al., 2003). The rigidity imparted by the ring structure constricts the molecule into exclusively adopting the correct rotamer for binding to the backbone of the NS3 protease. A 15-membered ring macrocycle was found to be optimal (Goudreau et al., 2004a). All these efforts, coupled with the previously described aminocyclopropane carboxylic acid at P1 (Rancourt et al., 2004), resulted in identification of a clinical candidate, BILN 2061 (ciluprevir) (Fig. 4A) (Llinas-Brunet et al., 2004; Tsantrizos, 2004). This compound has an excellent potency against the HCV NS3-4A serine protease, with estimated K_i values of 0.66 nM and 0.30 nM against the genotype 1a and 1b HCV protease, respectively (Lamarre et al., 2003). Treatment of the genotype 1a and 1b HCV replicon cells with BILN 2061 for 3 days resulted in a dose-dependent decrease of HCV RNA with a mean IC₅₀ of 4 nM and 3 nM, respectively (Lamarre et al., 2003).

Reversible, Covalent Inhibitors

The prospect of the hexapeptide inhibitors of the HCV NS3-4A protease being developed into potential oral therapies is rather low because of their large molecular weight and the presence of multiple carboxylic acids, both of which are major obstacles for achieving high oral bioavailability. Another approach to compensate for the loss of P5–P6 acidic pair is to design a “warhead” that forms covalent bonds to the catalytic Ser nucleophile, or so-called serine-trap. Ideally, these serine-trap inhibitors will form a covalent bond to the catalytic Ser¹³⁹, but this covalent bond will be cleaved so that the inhibition is not irreversible against the proteases. In the early stages of research on covalent HCV protease inhibitors, most of the standard serine-trap warheads, such as α-haloketones or heterocyclic ketones, displayed poor inhibition of this enzyme (Perni et al., 2004b). Although aldehydes were useful tools for SAR (Perni et al., 2003a), the inherent instability of aldehydes, in particular, aliphatic aldehydes, which are oxidized, rendered this warhead unsuitable for further development. The sulfonamido group is also an attractive and effective warhead for peptidomimetic scaffolds (Johansson et al., 2003). One of boronate esters demonstrated very strong, but presumably reversible, binding to the enzyme (K_i = 8 nM) (Priestley et al., 2002).

Another potent warhead is the α-ketoacid, which is capable of both covalent attachment to the catalytic Ser¹³⁹ and electrostatic attraction of the carbonyl terminus (Colarusso et al., 2002). The most potent inhibitors in a dipeptide series with the α-ketoacid warhead inhibited the HCV serine protease with an IC₅₀ of 3 μM in enzyme assay (Nizi et al., 2002). Diketones and α-ketoamides are also covalent but reversible functionalities that could serve as stable, reversible and effective binding groups (Han et al., 2003; Perni et al., 2004b), although in general, diketones are less potent than the corresponding α-ketoamides (Han et al., 2000). The covalent attachment of α-ketoamide group with Ser¹³⁹, as well as an unusual interaction between the dicarbonyl motif and the oxyanion of the protease (Fig. 4B), provide exceptional potency against the HCV NS3-4A protease. This class of inhibitors has a slow-binding mechanism of action due to the requirement of an unusual re-arrangement in the active site of the NS3 similar to that observed for ketoacids (Liu et al., 2004; Perni et al., 2004b). Extensive optimization of the P3 and P4 hydrophobic groups, removal of the acidic charge at P1’ and, finally, the modification of the P2 Pro substituent into a bicyclic Pro motif (Perni et al., 2003a; Perni et al., 2004a; Perni et al., 2004b; Sun et al., 2004; Victor et al., 2004; Yip et al., 2004a; Yip et al., 2004b) resulted in identification of a clinical candidate, VX-950 (Fig. 4A) (Perni et al., 2003b). While the optimal length for recognition by the HCV NS3-4A protease is 10 amino acids in natural HCV substrates, the backbone of these inhibitors was truncated to a tetrapeptide scaffold while maintaining significant binding affinity for the NS3-4A serine protease (Perni et al., 2003b). This compound had excellent potency against the HCV NS3-4A serine protease, with a K_i* value of 7 nM against the genotype 1a HCV protease (Perni et al., 2003b). Despite of the large difference in the IC₅₀ values in a 2-day replicon cell assay, these two inhibitors had a comparable ability to induce a 4-log₁₀ reduction of HCV RNA levels after an extended incubation with replicon cells (Lin et al., 2003; Lin et al., 2006).

Non-Covalent Substrate Mimetics and P' Inhibitors

To date, most peptidomimetic, active site inhibitors of the HCV serine protease were designed against those binding pockets on the P-side, and much less effort has been spent on the P’-side. It was found that the replacement of the P1’ residue of a decapeptide substrate based on the NS5A/NS5B site led to poor turnover of the substrate by the HCV protease (Steinkühler et al., 1996b; Urbani et al., 1997), indicating that the S1’ pocket can accommodate residues much larger than the natural Ser residue. In addition, a favorable interaction was found between the P4’ residue (Tyr) of these inhibitors and the NS4A cofactor (Landro et al., 1997). Incorporation of the optimized P’-side sequence and an N-terminal carboxylic acid, which is well-positioned in the active site to engage in interactions similar to those previously described for the C-terminal carboxylic acid of non-covalent product-based inhibitors, resulted in a novel series of HCV NS3-4A protease inhibitors that bind exclusively to the P’-side, without any contact with the P-side of the enzyme (Ingallinella et al., 2002). Competitive, capped tripeptide inhibitors of the NS3-4A protease, with low μM potency, were identified with the addition of a proper linkage between these two elements from a small combinational library. Another series of reversible, competitive inhibitors binding to the substrate-binding cleft across the active site has been described (Colarusso et al., 2003). The presence of a C-terminal phenethylamide group in these inhibitors allows an interaction with the S’-side of the enzyme, which might present a potential alternative to the C-terminal free carboxylic group present in the non-covalent, product analogue inhibitors.

Irreversible Inhibitors

As described above, the vast majority of HCV NS3-4A protease inhibitors that have been described to date are either non-covalent or covalent but reversible inhibitors. Recently, a series based on a pyrrolidine-5,5-trans-lactam core was reported to inhibit the HCV protease with an IC₅₀ up to 0.30 μM in replicon cell assay. These inhibitors bind irreversibly to the enzyme through opening of the lactam ring, with a biochemical potency (k_obs/I) of 7,760 M⁻¹s⁻¹ for the best compound in this series (Andrews et al., 2003a; Andrews et al., 2002; Andrews et al., 2003b; Andrews et al., 2003c; Slater et al., 2002).

Interaction between NS3 and NS4A

Because the interaction with the NS4A co-factor is critical for maintenance of a stable, active conformation of the NS3 protease, it is thought that agents competing against the NS4A binding could be used to inhibit the HCV protease activity. A 14-mer NS4A peptide (residues 21–34) with a substitution of Arg²⁸ with Glu yielded an IC₅₀ of 20 μM against activation of the NS3 protease by the wild-type NS4A peptide (Shimizu et al., 1996). 13-mer NS4A analogs (residues 22–34) assembled from D-amino acids (instead of the normal L-amino acids) in a standard order, or from L-amino acids in a reverse order, inhibited the NS3 protease activity with an IC₅₀ of 0.2 μM (Butkiewicz et al., 1996; Walker, 1999). In addition, several bivalent inhibitors in which the above-mentioned NS4A analogs were fused in frame to a NS5A/NS5B substrate have been described with IC₅₀ values ranging from 0.2 μM to 3 μM. In these cases, a hexapeptide (Glu-Asp-Val-Val-Cys-Cys), corresponding to a P6–P1 portion of the NS5A/NS5B substrate, was fused via a short linker of 1–3 amino acids to the NS4A analogs (Walker, 1999). However, it is unclear how these NS4A analogs inhibit the NS3-4A protease activity. Furthermore, since formation of the NS3-NS4A non-covalent complex may occur co-translationally during synthesis of the HCV polyprotein, it remains to be seen whether these NS4A analogs will be able to effectively compete against an already formed, tightly bound NS3-NS4A complex.

Aptamers

Another strategy to inhibit the HCV NS3-4A protease is to select aptamers (Biroccio et al., 2002; Fukuda et al., 1997; Fukuda et al., 2000; Sekiya et al., 2003; Urvil et al., 1997), which are single-stranded nucleic acids binding into a specific pocket of the target protein with high affinity and interfering with function(s) of the protein. Aptamers can be identified via multiple rounds of selection and amplification from a pool of random nucleic acids against any protein or small-molecule target. Two of these aptamers inhibit the HCV NS3 serine protease with an excellent potency (K_i = 3 μM for the better aptamer inhibitor) (Kumar et al., 1997). In addition, the same two aptamers were shown to block the helicase activity of NS3 (Kumar et al., 1997). Inhibition of the NS3 protease activity by a different set of RNA aptamers (Fukuda et al., 2000; Sekiya et al., 2003) was also demonstrated in transfected cells (Nishikawa et al., 2003).

Other Non-Peptidic Small-Molecule Inhibitors

Several groups have undertaken high-throughput screening of large libraries of chemical or natural products to identify novel inhibitors of the HCV NS3-4A serine protease that are not peptidomimetic. Many of the confirmed hits are non-competitive against the substrates (for a detailed review see Beaulieu and Llinas-Brunet, 2002). One of the hit series is 2,4,6-trihydroxyl-3-nitrobenzamides (THNBs) with an IC₅₀ of 3.0 or 5.8 μM in the absence or presence of NS4A, respectively (Sudo et al., 1997b). However, the major challenge for THNBs as potential therapeutic agents against HCV is the lack of selectivity against human serine proteases, such as chymotrypsin and elastase. Another series with a thiazolidine core was identified by the same group of scientists, with an IC₅₀ of 2.3 μg/mL (Sudo et al., 1997a). Again, improvement in selectivity is the major challenge for this series as the most selective derivative in this series showed a slight decline in potency. In a recent structure-based NMR screening of a customized fragment library, 16 small-molecule hits were discovered to bind weakly, with a K_d in the range of 100 μM to 10 mM, to substrate binding sites of the HCV NS3-4A protease (Wyss et al., 2004). NMR chemical shift perturbation data were then used to identify the binding location and the orientation of the active site directed scaffolds. Two of these compounds, which bind at the proximal S1–S3 and S2’ substrate binding pockets, were linked together to generate competitive inhibitors with relatively high molecular weight and IC₅₀ values in the μM range (Wyss et al., 2004). Novel, Zn²⁺-dependent benzimidazole-based inhibitors of the HCV serine protease have also been reported to form a ternary complex with a Zn²⁺ ion and the catalytic residues, His⁵⁷ and Ser¹⁹⁵, as determined in X-ray crystallography. However, the SAR for this series of compounds was established using a Zn²⁺-independent system, and did not appear to be consistent in the presence of Zn²⁺ (Sperandio et al., 2002) The highly charged nature and the dependence on a high concentration of Zn²⁺ ion of this class of inhibitors present significant challenges in terms of cellular penetration and oral bioavailability. Finally, a β-sheet mimetic has been combined with the boronate ester warhead to create a potent series of inhibitors (Glunz et al., 2003; Zhang et al., 2003).

In vitro Resistance Mutations

Because of the error-prone nature of the viral reverse transcriptase of retroviruses or RNA-dependent RNA polymerase of RNA viruses, drug resistance frequently emerges in patients treated with antiviral drugs and therefore limits the efficacy of these therapies. For new HCV NS3-4A serine protease inhibitors, resistance could become a major issue in the treatment of patients. The HCV subgenomic replicon system (Blight et al., 2000; Lohmann et al., 1999) was used for identification of in vitro resistance mutations against two HCV protease inhibitor clinical candidates, BILN 2061 or VX-950. All of the in vitro resistance mutations selected against either inhibitor were substitutions of a single amino acid in the NS3 serine protease domain and resulted in significant reduction in susceptibility to the respective inhibitor (Lin et al., 2005a; Lin et al., 2004a; Lu et al., 2004). Two of the primary resistance mutants against BILN 2061, Asp¹⁶⁸-to-Val (D168V) and Asp¹⁶⁸-to-Ala (D168A), were highly resistant to BILN 2061 as reflected in at least a 63-fold increase in K_i values in the FRET substrate-based enzyme assay and a more than several hundred-fold jump in the IC₅₀ values in the subgenomic replicon cell assay (Lin et al., 2004a). However, both mutants at Asp¹⁶⁸ remained fully susceptible to VX-950 because there was only a slight decrease in both K_i and IC₅₀ values in enzyme and replicon cell assays, respectively (Lin et al., 2004a). Asp¹⁶⁸ is in salt-bridge interactions with the side-chains of Arg¹²³ and Arg¹⁵⁵, and is also part of the S4 binding pocket. Computational modeling analysis suggests that substitution of Asp¹⁶⁸ with a non-acidic residue, such as Val or Ala, results in the loss of salt-bridge interaction with the Arg¹⁵⁵ side-chain on the neighboring β-strand (Fig. 5, color coded in light green), which in turn makes multiple contacts with the large P2 group of BILN 2061 in the model. Therefore, the conformation of the Arg¹⁵⁵ in the BILN 2061-wild-type NS3 protease complex is no longer energetically favored in the D168V or D168A mutant. Instead, Arg¹⁵⁵ in these two mutants appears to clash with the P2 quinoline group of BILN 2061 and destabilizes its binding. On the other hand, the conformation of Arg¹⁵⁵ in the two published crystal structures of the NS3 protease-inhibitor complex is similar to that in the VX-950-protease complex (Fig. 5, color coded in orange). In addition, this conformation of Arg¹⁵⁵ confers stabilization of VX-950 binding as it allows the maximal number of van der Waals contacts between the Arg¹⁵⁵ side-chain and the inhibitor. Therefore, VX-950 is not expected to be affected by the substitutions at Asp¹⁶⁸ compared with BILN 2061 (Lin et al., 2004a). It should be noted that substitutions at Asp¹⁶⁸ have been identified in a previous study as the resistance mutations against a less potent HCV protease inhibitor, which had an IC₅₀ of about 1 μM in the replicon cell assay (Trozzi et al., 2003). Another BILN 2061-resistant mutation, substitution of Arg¹⁵⁵ with Gln (R155Q), was identified in a separate in vitro study. The R155Q mutant was moderately resistant to BILN 2061 (a 24-fold increase in replicon cell IC₅₀) (Lu et al., 2004), although it is not clear whether this mutation confers resistance to VX-950 or not.

Fig. 5

Computational models of the HCV NS3-4A protease in complex with its inhibitors. The protein is shown as a schematic based on its secondary structure in light gray. The inhibitors (VX-950 in yellow (more...)

The major in vitro resistance mutant against VX-950, Ala¹⁵⁶-to-Ser (A156S), was moderately resistant to VX-950 with a ~12-fold and ~29-fold increase in enzyme K_i and replicon cellular IC₅₀ values, respectively (Lin et al., 2004a). The HCV replicon cells containing the A156S substitution remained as sensitive to BILN 2061 as the wild-type replicon cells (Lin et al., 2004a). The Ala¹⁵⁶ side-chain is in van der Waals contact with the P2 group of these two inhibitors (Fig. 5, color coded in green). In a computational model of the A156S mutant, the terminal oxygen of Ser¹⁵⁶ is too close to the P4 cyclohexyl group of VX-950, and it is also close to the terminal cyclopentyl cap of BILN 2061. Because the cyclopentyl cap of BILN 2061 is at the flexible end of the inhibitor, it can be moved away from this unfavorable contact without significantly affecting BILN 2061 binding. A similar movement of the P4 cyclohexyl group of VX-950 causes destabilization of the interactions between the inhibitor and S4 and S5 sub-sites of the protease. Therefore, a larger loss in binding affinity is expected for VX-950 than for BILN 2061 with the A156S mutant protease.

The lack of overlap between the dominant in vitro resistance mutations of BILN 2061 and VX-950 raised an interesting question – whether it is possible for a combination of these two protease inhibitors to suppress the emergence of these resistance mutations. While the combination did suppress the occurrence of A156S, D168V or D168A mutants, two other single-residue substitutions, Ala¹⁵⁶-to-Thr (A156T) and Ala¹⁵⁶-to-Val (A156V), were selected and found to confer cross-resistance to both VX-950 and BILN 2061 (Lin et al., 2005a). In a computational model, two out of the three possible conformations of the A156S side chain have unfavorable contacts with both the inhibitors either at the P2 side chain or P3 carbonyl group. In the A156T or A156V mutation, the additional hydroxyl or methyl group, respectively, at the C_β atom of the residue 156 side-chain is forced to occupy one of these two unfavorable positions, which leads to a repulsive interaction with the inhibitor and/or enzyme backbone atoms. Therefore, A156T and A156V mutants are expected to be resistant to both inhibitors (Lin et al., 2005a). It also remains to be seen which of these resistance mutations identified in cell culture, if any, will be observed in patients treated with HCV protease inhibitors.

In Vitro Combinations

The current standard care for hepatitis C is a combination of weekly injections of pegylated IFN-α and daily oral doses of ribavirin, which results in a SVR in roughly half of treated patients. The SVR is higher (~80%) in patients infected with genotype 2 or 3 HCV, but much lower (40–50%) in genotype 1 HCV-infected patients, who account for the majority of hepatitis C population in developed countries. It remains to be seen whether a single direct antiviral agent, such as a protease inhibitor, is sufficient to induce a more favorable SVR in chronically infected hepatitis C patients. One possible strategy to increase efficacy and help suppress the emergence of resistance mutations in HCV protease inhibitor-based therapy is to combine it with other antiviral agents, such as IFN-α or a polymerase inhibitor. It has been shown that a combination of an HCV NS3-4A protease inhibitor with IFN-α resulted in a synergistic reduction of HCV RNA in the subgenomic replicon cells after a 2-day incubation (Lin et al., 2004b). Furthermore, the benefit of the combination was sustained over time such that a nearly 4-log₁₀ or 10,000-fold reduction of HCV RNA was achieved following a 9-day treatment of the HCV replicon cells. The viral RNA dropped by more than 4-log₁₀ to below the detection limit after a-14 day combination treatment.

In the presence of G418 (neomycin), which allows selective growth of replication-competent HCV replicon cells over “cured” Huh7 cells, no replicon cells were recovered three weeks after withdrawal of the inhibitors, suggesting that the HCV RNA has been cleared from the cells by the 14-day combination treatment (Lin et al., 2004b). In each case, the antiviral effects obtained with higher concentrations of either the protease inhibitor alone or IFN-α alone can be achieved by a combination of both agents at lower concentrations, which potentially may reduce the risk of possible adverse effects associated with high doses of either agent (Lin et al., 2004b). Given the observation that HCV protease may interfere with the IFN signal transduction pathway, one of the major components of the host anti-viral response, these data suggest that HCV protease inhibitors may have a dual anti-HCV function, blocking the HCV RNA replication and restoring the host antiviral response.

Clinical Development

BILN 2061 (Fig. 4A, ciluprevir) was the first NS3-4A protease inhibitor to enter clinical development. Despite its peptidic nature, BILN 2061 showed a low-to-moderate oral bioavailability after a single dose in multiple species of animals, including rats and dogs (Lamarre et al., 2003; Narjes et al., 2002b). In a single-dose escalation phase 1a trial in healthy adults, this compound showed dose-proportionality up to the 1,200 mg dose, and was well tolerated up to the 2,000 mg dose. In several two-day, twice-daily dosing studies, BILN 2061 has been shown to possess proof-of-concept antiviral activity in chronic genotype 1 HCV-infected patients with minimal or advanced fibrosis (Benhamou et al., 2002; Hinrichsen et al., 2002; Lamarre et al., 2003). At 200 mg/administration, a 2-3 log₁₀ or greater reduction in viral load was observed after the 2-day treatment. In some patients, the viral load dropped below the limit of detection using a relatively insensitive assay (<1500 copies/mL), although it remained positive in a more-sensitive assay (>50 copies/mL). However, BILN 2061 was much less effective against genotypes 2 and 3 HCV, as evidenced by the K_i values (80–90 nM) against these proteases in vitro (Thibeault et al., 2004), and an uneven, less pronounced viral load reduction in the genotypes 2 and 3 HCV infected patients (Reiser et al., 2003). Unfortunately, further development of BILN 2061 was put on hold due to safety concerns in animals, which has not yet been fully disclosed.

As described before, VX-950 (Fig. 4A) has a different mechanism of inhibition than BILN 2061. VX-950 was the first covalent, reversible inhibitor of the HCV NS3-4A protease to enter clinical development for hepatitis C. VX-950 showed a moderate oral bioavailability and a much higher exposure in the livers than in the plasma after a single dose in both rats and dogs (Perni et al., 2003b). In a single-dose escalation (range 25–1,250 mg) phase 1a trial in healthy adults, VX- 950 was well tolerated up to the 1,250 mg dose level and exhibited good systemic exposure, which was greater than proportional to dose (Chu et al., 2004). In a recently completed 14-day study, VX-950 demonstrated excellent antiviral activity in chronic genotype 1 HCV-infected patients (Reesink et al., 2005) (Fig. 6). Again, VX-950 was well tolerated in all three groups, 450 mg thrice daily, 750 mg thrice daily, and 1,250 mg twice daily. In all three groups treated with different VX-950 regimens, a 2–3 log₁₀ or greater reduction in viral load was observed after the first 2 days of treatment, as in the case of BILN 2061. In some patients, the viral load dropped by more than 4 log₁₀ to below the limit of detection of a very sensitive assay (<10 IU/mL) after 14 days of dosing. In addition, VX-950 was equally potent against the HCV NS3-4A protease of genotype 2, but not genotype 3, in enzyme assays (Taylor et al., 2004). A close homolog of VX-950 inhibited the replication of a genotype 2a full-length HCV replicon that is capable of generating infectious virus particle (Lindenbach et al., 2005), suggesting that it may also be an effective agent for genotype 2 HCV-infected patients.

Fig. 6

Antiviral antivity of VX-950 in chronic HCV-infected patients. Chronic genotype 1 HCV-infected patients were treated with VX-950 at the following doses for 14 days: 450 mg thrice daily (open square, (more...)