Focus on flaviviruses: current and future drug targets

NS3Pro activity

An important target within NS3 is the trypsin-like serine protease domain (NS3Pro) [12,13]. For NS3Pro to be active it must be in a complex with its cofactor NS2B. This protease (NS2B/NS3Pro) plays an essential role in the cleavage of the viral precursor polyprotein and disruption of this function is lethal to viral replication [14]. There are two primary avenues for antiviral inhibition of NS3Pro: the first is to target the enzymatic activity of an existing NS2B/NS3Pro complex and the second is to block association of NS2B with NS3Pro.

Initial efforts to optimize existing substrate-like inhibitors of protease activity were unsuccessful [15], with compounds such as benzamidine, PMSF, leupeptin and tosyl-L-lysine chloromethyl ketone failing to inhibit the WNV protease at 100 μM concentrations [16]. X-ray crystallography has since provided a wealth of information for identifying and optimizing inhibitors of NS3 enzymatic function (details of available structures are provided in Table 1). Most recently, the crystal structure of the full length NS3Pro domain from DEN-4 was reported, and this structure provides tremendous insight into the structural complexity of NS3 [17]. Structural information is also available for substrate-free and inhibitor-bound WNV NS2B/NS3Pro [18]. Collectively, these data have advanced our knowledge of critical interactions between NS2B and NS3, and provide the foundation for the rational design of inhibitors that mimic substrate association. Computational docking of virtual compound libraries into the substrate-binding cleft and high-throughput screening (HTS) of millions of compounds have yet to produce a significant number of new NS2B/NS3Pro inhibitor candidates [19]. Bioassay systems have been more successful in identifying compounds active against NS3 protease function. Using luciferase-expressing WNV replicons, substrate-based aldehyde analogs have been identified as potential inhibitors with EC₅₀ values of approximately 1.4 μM and selectivity indexes of approximately 100 for WNV [20]. In another study, a cationic tripeptide with a non-peptidic cap at the N-terminus and an aldehyde at the C-terminus was shown to have a K_i of 9 nM against WNV protease activity, to be stable in serum (>90% intact after 3 h at 37°C), cell permeable, and exhibit an EC₅₀ of 1.6 μM with low cytotoxicity (CC₅₀ > 400 μM) [21].

A second avenue for antiviral development targeting NS3 protease is the identification of compounds able to block NS2B/NS3Pro association. Molecular dynamic simulations, principal-component analysis, molecular docking, mutagenesis and bioassays revealed the importance of the NS2B/NS3 complex and of the conformational changes resulting from the association for protease activation [22]. Targeting complex formation has yet to be vigorously tested as a strategy for inhibition, but it has significant potential. The interface between NS2B and NS3Pro has been resolved following crystallography [18,23]. These structural studies and a number of biochemical and virologic experiments have defined the specific interactions between NS2B and NS3 that are important to NS2B/NS3Pro activity [19,24–27], and have led to the preliminary identification of a 5′--amino-1-(phenyl)sulfonyl-pyrazol-3-yl class of compounds suggested by in silico docking to interfere with NS2B/NS3Pro association [28].

An increased understanding of the interactions between NS2B/NS3Pro may be gained from structural investigations of the hepacivirus NS3 and NS4A complex, and similarities may be exploited for flavivirus antiviral discovery. Historically, poor bioavailability and rapid degradation of peptidomimetic inhibitors identified for inhibition of hepatitis C virus (HCV) has compromised their effectiveness as antivirals. However, recent evaluation of the peptidomimetic inhibitor compound 3 and the tripeptide ketoamide compound, SCH 503034, suggest that the use of peptidomimetic inhibitors may indeed be viable [21,29]. Macrocyclic compounds are also performing well in both enzymatic assays and in cell-based replicon systems, and are yielding promising results in clinical trials. BILN 2061 provides confirmation for the application of macrocyclic compounds: after patients infected with HCV were treated with the macro-cyclic BILN 2061, there was an impressive reduction in HCV RNA plasmid levels 2 days after treatment [30]. BILN 2061 has also undergone numerous pharmacological tests to investigate its preclinical safety, and it was found to be well tolerated in various in vitro and in vivo assays with good metabolic properties in a variety of species [30]. The positive results obtained for HCV protease inhibitors are encouraging and reinforces the potential for developing protease inhibitors targeting the flavivirus NS2B/NS3Pro.

NS3 helicase activity

Helicase activity is required for the separation of dsRNA formed during viral replication and for the removal of proteins bound to the viral RNA during replication. For either activity, ATP is required and the two C-terminal domains of the NS3 protein comprise the ATPase activity [20]. Mutational analysis has shown that alteration of helicase, RTPase or ATPase activities results in the reduction or elimination of viral replication [10,31,32].

Recently, several high-resolution structures of the flaviviral NS3 have been solved (details provided in Table 1) which illustrate the expected three protein domains. However, the protease and helicase domains build an unexpected elongated structure with an RNA-binding tunnel that transverses the length of the protein. The structures are important because they represent a divergence from the hitherto accepted (HCV) model for helicase structure/conformation in mosquito-borne flaviviruses [33,34].

The NS3 helicase is a challenging target for drug design because its enzymology is not clearly understood. The most likely target of the NS3 helicase is the ATPase-binding site, however, this target is potentially problematic because of selectivity issues [20,35]. Currently, halogenated benztrioles have been shown to inhibit WNV helicase activity by targeting ATPase activity and ring-expanded nucleosides have also been reported to inhibit JEV and WNV helicases [35,36]. Sampath et al. demonstrated that an alanine substitution of Lys 396 drastically reduces helicase activity [10]. Interestingly, the amino acid substitution decreased activity to a greater degree then current small-molecule inhibitors [10,20]. The location of Lys 396 on the surface of domain II suggests that rather than targeting the ATP-binding sites for inhibition, perhaps the nucleoside-binding pockets surrounding the RNA-binding tunnel should be considered (see ‘Future perspective’ section).

RNA-dependent RNA polymerase

The RNA-dependent RNA polymerase (RdRp) is a strong candidate target for antiflaviviral drug development. The RdRp is located within residues 273–900 of NS5, and is responsible for synthesis of both the negative-strand RNA and positive-strand genomic RNA. Because the RdRp is essential for viral replication and a similar enzyme is not found within host cells, it is considered of great value as a target for the development of potent and specific antivirals.

Flavivirus negative- and positive-strand RNA synthesis is a complex and dynamic process, and an excellent review of the replication process and RdRp function was recently published [37]. Briefly, viral RNA released into the cytoplasm from infecting virions is first used for translation of the viral polyprotein, and the newly translated NS5 RdRp enzyme initiates synthesis of a negative-sense copy of the genomic viral RNA. Once bound to the 3′-untranslated region, the RdRp (in conjunction with the helicase activity of NS3) will polymerize a full-length negative-strand copy of the genomic RNA. The function of the negative-strand RNA is to serve as a template for the synthesis of a large number of positive-strand genomic RNAs. The RdRp recognizes and binds to the double-stranded, positive-negative RNA duplex and, in combination with NS3 helicase activity, synthesizes a nascent positive-strand RNA, displacing the original positive strand from the RNA duplex in the process. In a coupled process, the newly synthesized positive-strand RNA is capped by the combination of the NS3 RTPase, NS5 MTase and an as yet unidentified GTase.

The X-ray crystallographic structures for WNV and DEN-3 RdRp domains were recently solved [38,39] and both structures have the same basic structural components that characterize all known RdRp enzymes (i.e., palm, finger and thumb domains) [40]. When comparing the WNV and DEN-3 RdRp structures, most of the variability is observed within the finger domains, with the palm domain being the most conserved. DEN-3 RdRp assumed an open conformation and the WNV structure showed the finger subdomain rotated towards the thumb by an angle of approximately 8° [39]. The individual domains are similar to those of a DNA polymerase, however, the flavivirus RdRp does appear to have some distinctive characteristics. For example, the DEN-3 RdRp domain has fingertips that connect the thumb and the finger domains, allowing for a more compact, spherical shape [39].

Currently, HTS against DEN-2 RdRp using chemical libraries from both academic and commercial sources is being performed to identify both nucleoside inhibitors (NI) and non-NIs(NNIs) in order to find a suitable antiviral drug [37]. NIs are usually converted to nucleotide analogs by kinases present in the host cells and target the active site of the polymerase [41]. If the NI binds to the RdRp domain, the elongation processes that the polymerase performs is terminated. NNIs tend to bind to allosteric surface cavities that are located throughout the target polymerase, do not directly compete with NTPs and generally inhibit the polymerase at a stage prior to the elongation reaction [41]. NNIs have been shown to inhibit the replication of bovine virus diarrhea virus with high efficiency, and these allosteric inhibitors may also have an effect on WNV, DEN and YF. Targeting NNI-binding sites in HCV is a current focus for drug development [37], and covalent inhibitors with submicromolar biochemical affinity have been observed [42]. To date, two cavities predicted to be potential target locations for NNIs have been identified within the DEN-3 RdRp domain and five cavities identified within the WNV RdRp domain [37].

NS5 N-terminal methyltransferase adomet binding

The N-terminal, approximately 260 amino acids, of the NS5 protein contains the methyltransferase enzyme necessary for the formation of the mature RNA cap structure present at the 5′-- end of the viral genomic RNA. S-adenosyl methionine (AdoMet) is used by methyltransferases as a methyl group donor for substrate modification. Mutation of residues involved in NS5 AdoMet binding disrupt viral replication [43,44]; however, identifying compounds that specifically inhibit flavivirus MTase AdoMet binding, but not cellular methyltransferase activity, may be problematic due to the conserved nature of the AdoMet binding motif. Indeed, the conserved KDKE tetrad motif was initially used to classify the N-terminus of NS5 as a methyltransferase [45]. Additionally, the postreaction methyltransfer product AdoHyc is copurified and co crystallized in the AdoMet-binding site of all flavivirus MTase proteins crystallized to date [44,46,47], indicating that the site is likely to be occupied at high frequency in vivo. The K_d of AdoMet binding to the flavivirus MTase has not been empirically determined, but the presence of AdoHyc in the MTase crystal structures indicates that at least AdoHyc is relatively tightly bound (Figure 3) [44,47–50]. Furthermore, published methyltransferase assays detect post-reaction products rather than directly detecting ligand displacement, making in vitro HTS problematic [51]. However, the tight association of AdoMet to the MTase and the well-resolved crystal structures have allowed for virtual screening of compounds that could displace AdoMet. Luzhkov et al. identified one compound from in silico screening (compound 7) that could displace AdoMet from the MTase protein in an in vitro assay [52]. No toxicity or selectivity data has been reported for compound 7, but as a novel scaffold, it may be a valuable starting point for subsequent medicinal chemistry efforts to identify potent and selective AdoMet analogs that possess suitable toxicity and bioavailability profiles for further drug development.

Open in a separate window

Figure 3

(A) Surface representation of the yellow fever NS5 N-terminal methyltransferase domain indicating locations of GTP cap-binding domain, AdoMet-binding domain and RNA/phosphate-binding site. (B) Relative location of the three binding sites. (C) Yellow fever NS5 N-terminal methyltransferase residues associated with GTP binding to the GTP cap-binding site. (D) Yellow fever NS5 N-terminal methyltransferase residues associated with AdoMet binding to the AdoMet-binding site [47].

NS5 N-terminal methyltransferase GTP binding

The GTP-binding domain of the NS5 N-terminal MTase domain is involved in the 2′-O methylation reaction [43,53,54]. Mutation of amino acids in the DEN MTase RNA cap-binding domain abolishes viral replication [43,55,56], indicating that GTP and RNA cap binding are essential to viral replication and may represent a novel drug target site. Comparison of the GTP-binding site structure from the various guanine-binding proteins in the PDB database with the flavivirus MTase GTP-binding site suggests that the flavivirus MTase proteins utilize a novel architecture for binding guanosine [49]. Most GTP-binding proteins with known structures interact with the guanine base via the guanine C-6 carbonyl group and two aromatic or charged residues that sandwich the guanine ring [57–59]. By contrast, the flavivirus MTases bind the cap on only one face of the guanine base, leaving the other highly exposed (Figure 3). The MTase Leu 16 and Leu 19 peptide backbone carbonyl groups interact with the guanine N-1 atom (via a conserved water bridge) and C-2 amino group. There is also a strong, directed stabilizing effect observed between the guanine ring and Phe24 [47,49,50]. These data suggest that the N-1 and C-2 positions represent the major determinants for specificity. This novel guanosine-binding configuration, including the lack of interaction with the C-6 carbonyl and the presence of the N-1 water bridge coupled with the open guanine-binding architecture, represents a significant difference as compared with known endogenous GTP-binding proteins, and may allow for virus-specific compounds to be identified. In addition, the GTP-binding site is highly amenable to in vitro HTS due to the availability of stable fluorescent GTP analogs [47]. Finally, all flavivirus NS5 MTase proteins appear to adopt the same mechanism to bind GTP, indicating that broadly active antivirals can be developed toward the GTP-binding site. Collectively, these factors advocate the flavivirus MTase RNA-cap-binding site as an ideal target for antiviral drug development. Our group has performed in vitro HTS for compounds that displace GTP from the MTase cap-binding pocket, and has identified a number of molecules with K_i values of less than 10 μM (Unpublished Data). Additional ex vivo and structure-based drug-design testing is underway to further optimize potential inhibitors. While little has been published to date on antiviral development targeting this domain, the conservation of the GTP-binding site amongst flaviviruses, ease of HTS assay development, and critical nature of GTP binding for viral replication makes the MTase an exciting new target for novel drug development [60].

Inhibitors of virion assembly

Flavivirus virions are relatively simple structures composed of a large number of repeating protein units. The flavivirus virion is composed of an inner shell formed by interacting subunits of highly basic capsid (C) protein that interact with genomic RNA, an intermediate shell composed of membrane derived from the host ER, and an outer shell containing the viral glycoproteins M and E proteins. The nucleocapsid (RNA + C) is assembled in direct association with the RNA replication complex on the cytoplasmic face of the ER, and the viral glycoproteins (E and prM) drive virion budding [81–83]. The RNA-containing nucleocapsid associates with the budding glycoproteins, and the immature virion is released into the lumen of the ER. Immature virions are trafficked through the Golgi apparatus, where both the prM and E proteins are N-glycosylated and prM is cleaved to M by the furin protease. Maintenance of the immature virion until release from the cell is dependent on the interaction of E with prM, as prM inhibits the rearrangement of the E protein to a fusogenic conformation until the mature virion is released thus preventing reinfection of the cell. It is apparent that virion assembly is a highly orchestrated process, and chemotherapeutic disruption of any step of the process will severely affect viral viability.

Capsid dimerization

Several points along the virion assembly pathway provide intriguing targets for antiviral drug development. Nucleocapsid assembly is one such target. As mentioned earlier, the nucleocapsid is composed of one copy of the genomic viral RNA and an indeterminate number of the 11-kDa C protein. The flavivirus C protein is separated from prM on the viral polyprotein via NS3-mediated cleavage, and mature C associates with the cytoplasmic surface of the ER. The positive charge in the amino acid region 75–97 of C is thought to interact with genomic viral RNA [84]. The solution structure of the DEN-2 C protein indicates that the C protein forms dimers, with helix α2 and α4 forming the dimerization domain, and helix α4 presenting a highly basic surface that likely interacts with the phosphate backbone of the viral genomic RNA. Both the helical association and charged surface interaction present possible avenues of therapeutic intervention. Disruption of the basic charge on helix α4 may inhibit RNA binding to the capsid dimer and disrupt virion formation. However, it is unlikely that a small-molecule inhibitor could shield sufficient basic charges along the α4/α4′ dimer to significantly disrupt the protein–RNA interaction at a clinically available concentration. A perhaps more promising approach is to either disrupt or destabilize C–C dimerization via the interaction of helix α2/α2′ or α4/α4′. The geometry of C–C dimer interaction is very precise based on the solution structure of the DEN C protein, and chemicals that bind between the homologous or heterologous helices may interfere with nucleocapsid formation and virion assembly, or could stabilize the nucleocapsid such that virions cannot efficiently release their genomic RNA into cells. Viral capsid disruption has been utilized against viruses such as hepatitis B virus [85,86], Sindbis virus [87,88] and HIV-1 [89–91]. Several approaches for identifying capsid-disrupting compounds are possible. Since the NMR structures of the DEN capsid dimer are available, rational drug design methodologies can be used to identify both potential sites of disruption and chemical entities that may bind between the dimers. Alternatively, approaches that monitor capsid dimerization may be utilized. These may include in vitro fluorescence energy transfer (FRET)-style screens with purified and chemically labeled protein, or live cell screening using new fluorescence techniques such as bimolecular fluorescence complementation (BiFC) to identify compounds that disrupt C–C dimerization (Figure 4).

Open in a separate window

Figure 4

BiFC and FRET are two cell-based techniques to monitor protein–protein interactions

(A) In BiFC, the targeted proteins are linked to the N-terminal and C-terminal halves of a fluorescent protein (YFP). If the targeted proteins are within 15 nm of each other and in the correct spatial orientation, the N- and C-terminal halves of YFP can refold and form an active fluorophore that produces a detectable emission signal upon excitation. If the targeted proteins are farther apart than 15 nm, the N- and C-terminal halves of YFP are not able to re associate and will not produce an emission signal upon excitation. In FRET, the two proteins whose interactions are being monitored are linked to two different fluorescent proteins whose emission (CFP) and excitation (YFP) wavelengths overlap. If the targeted proteins (and their fluorescent partners) are within 1–10 nm of each other, the emission of CFP is capable of exciting YFP and producing a YFP emission signal. If the proteins are farther apart than 10 nm, the emission from CFP is not able to excite YFP, and no YFP emission is observed. BiFC: Bimolecular fluorescence complementation; FRET: Fluorescence resonance energy transfer.

Heterologous structural protein interactions

Disrupting the interactions between viral structural proteins (C, M and E) is another potential area for antiviral development. However, these interactions may represent difficult drug development targets for several reasons. Cryoelectron microscopic reconstruction of WNV and DEN virions indicate that the N-terminal receptor binding and fusion domains are on the exterior of the virion and the carboxyl termini of the proteins do not extend significantly into the virion interior [92,93]. The lack of a significant cytoplasmic tail for E or prM makes it difficult to determine how C may specifically interact with the proteins, although the presence of a hydrophobic region on the C dimer (helix α1, α2, α1′ and α2′) may provide some interaction with the minimal E endodomain that is not appreciated in the cryo-EM structures. Alternatively, C may not form direct interactions with E or prM, but rather may interact directly with membranes. If this is the case, compounds that bind to the hydrophobic region in the C dimer and extend past the surface of the helices may hinder capsid dimers interaction with the membrane. In either case, targeting the hydrophobic pocket of the capsid dimer may be a viable option for disrupting the C-C, C-E or C-M interaction and viral budding.

Likewise, disruption of the interaction between M and E may prove to be difficult. Based on the cryo-EM reconstructions described earlier, there are no conserved contacts visible between the E or M coiled coil transmembrane helices that could account for specific interaction. While this does not eliminate the helix–helix interaction between M and E as a potential drug target, it does limit the usefulness of rational design methodologies in identifying compounds that block this interaction. A more promising target for interfering with M:E interactions is within the ectodomains of both proteins. During virion maturation, prM inhibits the membrane fusion activity of E during egress of the immature virion through the trans-Golgi network [94]. Cleavage of prM by the cellular furin protease [95] enables the E protein to homodimerize and the virions to become infectious [96,97]. The crystal structure of the ectodomain of prM complexed with the E ectodomain at low pH has been solved [98], a development that appears to provide a number of interesting targets. Inhibition of the interaction between the ectodomains of prM and E could be addressed by several mechanisms, including in silico modeling of compounds that specifically bind to the interaction interface or through in vitro HTS campaigns similar to those described earlier to identify candidates that interfere with C-C homodimer formation. Alternatively, compounds could be identified that bind to the furin protease cleavage site on prM, which would not block the release of virion particles but would make the particles unable to enter naive cells and essentially non-infectious.

Disruption of viral replication

Viral genomic replication is a highly orchestrated process that is essential to the production of progeny virus, and as such is a rich environment for the discovery of new antiviral targets and compounds. Currently, two viral nonstructural proteins, NS3 and NS5, are considered prime targets for antiviral development due to their essential roles for viral RNA replication. The currently targeted enzymatic functions of NS3 helicase/protease and NS5 are described earlier. There are several additional antiviral development targets within both the NS3 and NS5 proteins, as well as within the other viral nonstructural proteins, that have yet to be exploited.

NS3 RTPase domain

The known RNA capping machinery for flaviviruses is encoded within two flaviviral enzymes, NS3 (RTPase) and NS5 (MTase). The NS3 RTPase binds to the 5′-triphosphate of the nascent positive-strand viral RNA and removes the γ-phosphate [99], resulting in a 5′-diphos-phorylated RNA, which is the substrate for the guanosine monophosphate moiety transferred from the 5′-RNA guanylyltransferase (GTase). A low-resolution crystal structure of the Kunjin NS3 helicase/NTPase/RTPase domain was recently solved [33], but is not currently of sufficient resolution to be useful in rational drug design efforts. However, with greater resolution, rational design efforts can be utilized to find compounds that bind to and inhibit the ATP or RNA-binding site. In the meantime, an HTS assay to monitor the release of the γ-phosphate based on scintillation proximity assays could be developed, although this would require a greater understanding of the biochemistry of phosphate removal and RNA binding to the NS3 protein to discriminate between inhibitors of helicase and RTPase activity.

NS5 RNA binding

The NS5 MTase domain binds to viral RNA and methylates both the guanine cap and the ribose 2′-OH on the viral RNA. Disrupting the essential step of RNA association to the methyltransferase would impair its function and halt viral replication. The exact mechanism for RNA binding to the methyltransferase domain has not yet been definitively demonstrated, but there is growing evidence that the RNA could interact with a basic channel formed by helix α3 and the random coil region between β-sheets β2 and β3 [49]. The highly basic nature and apparent conservation of charge structure suggest that this region may bind nonspecifically to RNA. Strategies that can monitor displacement of RNA from a protein, such as fluorescence polarization assays or high-throughput scintillation proximity assays, could be used in HTS to identify compounds that interfere with RNA binding. Alternatively, as the structure of the putative RNA-binding domain is known (Figure 3), structure-based drug design could be could be used to develop compounds that fit tightly and selectively into the RNA-binding channel of the NS5 enzyme. An exciting concept is chemically linking compounds that individually block GTP binding and RNA binding such that both sites are occluded simultaneously (Figure 4). The conjugation strategy has been shown effective using pegylated interferon linked to ribavirin [100]. Such an approach would be advantageous for limiting antiviral resistance, as the likelihood of multiple resistance mutations arising within one generation of viral replication is extremely low.

Financial & competing interests disclosure

This work was supported by a grant from the Rocky Mountain Regional Center for Excellence (U54 AI-065357) to Brian J Geiss and Susan M Keenan. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.