Potential conflict of interest: Nothing to report.
Supported by the University of Leeds Biomedical Health Research Centre; a Medical Research Council New Investigator Research Grant (G0700124) and Yorkshire Cancer Research Grant (PP025) (to S. G.); a Wellcome Trust Ph.D. studentship (to T. L. F.); a Cooperative Awards in Science and Engineering Ph.D. studentship from the Biotechnology and Biological Sciences Research Council (BBSRC) and Pfizer (to E. A.); and a BBSRC studentship (to M. V.).
The hepatitis C virus (HCV) p7 ion channel plays a critical role during infectious virus production and represents an important new therapeutic target. Its activity is blocked by structurally distinct classes of small molecules, with sensitivity varying between isolate p7 sequences. Although this is indicative of specific protein–drug interactions, a lack of high-resolution structural information has precluded the identification of inhibitor binding sites, and their modes of action remain undefined. Furthermore, a lack of clinical efficacy for existing p7 inhibitors has cast doubt over their specific antiviral effects. We identified specific resistance mutations that define the mode of action for two classes of p7 inhibitor: adamantanes and alkylated imino sugars (IS). Adamantane resistance was mediated by an L20F mutation, which has been documented in clinical trials. Molecular modeling revealed that L20 resided within a membrane-exposed binding pocket, where drug binding prevented low pH-mediated channel opening. The peripheral binding pocket was further validated by a panel of adamantane derivatives as well as a bespoke molecule designed to bind the region with high affinity. By contrast, an F25A polymorphism found in genotype 3a HCV conferred IS resistance and confirmed that these compounds intercalate between p7 protomers, preventing channel oligomerization. Neither resistance mutation significantly reduced viral fitness in culture, consistent with a low genetic barrier to resistance occurring in vivo. Furthermore, no cross-resistance was observed for the mutant phenotypes, and the two inhibitor classes showed additive effects against wild-type HCV. Conclusion: These observations support the notion that p7 inhibitor combinations could be a useful addition to future HCV-specific therapies. (HEPATOLOGY 2011;)
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Hepatitis C virus (HCV) infects over 3% of the population, causing severe liver disease. Current therapy comprising pegylated interferon (IFN) and ribavirin (Rib) is inadequate, which, combined with high cost and poor patient compliance, has driven the demand for new virus-specific drugs.1 Future standard of care will replace IFN/Rib with combinations of specific inhibitors, such as seen for human immunodeficiency virus (HIV) therapy. However, extensive HCV variability raises concerns over the ability of relatively few compounds to suppress resistance. Thus, great effort focuses on expanding the repertoire of HCV drug targets, expedited by the availability of the Japanese fulminant hepatitis clone 1 (JFH-1) infectious isolate.2
HCV is the prototype member of the Hepacivirus genus within the Flaviviridae.3 It is enveloped and possesses a positive-sense single-stranded RNA genome of ∼9.6 kb. An internal ribosome entry site in the 5′ untranslated region drives translation of a polyprotein that is cleaved into 10 mature products. The core and envelope glycoproteins with the RNA genome comprise the virion, whereas nonstructural (NS) proteins modulate host metabolism and replication of the viral RNA. JFH-1 has permitted the study of particle production, and it has become clear that, in addition to canonical virion components, other viral proteins are required.4-13
HCV p7 forms a cation channel in vitro,14-16 and both deletions and point mutations markedly reduce the production of infectious virions in culture.4, 5 It is comprised of two trans-membrane domains separated by a cytosolic loop and forms both hexameric and heptameric complexes.14, 17, 18 We have recently shown that p7 acts as a proton channel within infected cells, which is directly required for the production of infectious virions.19 p7 is required for HCV to replicate in chimpanzees20 and small molecules block both channel function in vitro and virion production in culture, rendering it an attractive antiviral target.21, 22
Skepticism concerning p7 inhibitors heralds from trials where p7 inhibitor monotherapy, or combinations with IFN/Rib failed to significantly improve responses.23 However, evidence from meta-analyses24, 25 and patient virus loads at early time points26, 27 supports a specific antiviral effect, and selection of specific nonsynonymous mutations occurs within patient isolate p7 sequences.28, 29 Because HCV displays genotype (GT)-dependent p7 inhibitor sensitivity,21 changes in amino acid sequence could interfere with the binding of drug molecules, making it likely that the emergence of resistant quasispecies accounts for trial outcomes.
Here, we identify p7 resistance mutations specific to adamantane and IS drugs, indicative of a genuine antiviral effect that supports their inclusion in future combination therapies.
Ama, amantadine; DHPC, 1,2-diheptanoyl-sn-glycero-3-phosphocholine; GT, genotype; HCV, hepatitis C virus; HIV, human immunodeficiency virus; IFN, interferon; IS, imino sugar; JFH-1, Japanese fulminant hepatitis clone 1; LMPG, lyso-myristoylphosphatidylglycerol; NN-DGJ, N-nonyl deoxygalactonojirimycin; NN-DNJ, N-nonyl deoxynojirimycin; NS, nonstructural; PAGE, polyacrylamide gel electrophoresis; PCR, polymerase chain reaction; Rib, ribavirin; Rim, rimantadine.
Materials and Methods
pJFH-1, pCON-1/JFH-1c3, p452/JFH-1c6, pJ4(CVL6S), and pGEX-p7(J4/JFH-1/452) have been described.2, 21, 30-32 pGEX-p7 mutants were generated by fusion polymerase chain reaction (PCR). pJFH-1 mutagenesis: a unique BsiWI–KpnI fragment was ligated into pLitmus28i (NEB): pLitJFH-B/K and a silent AvrII site introduced 5′ of p7: pLitJFH-B/K(A). The BsiWI–KpnI fragment containing the AvrII site was reintroduced into pJFH-1: JFH(A), which replicated and produced particles as wild-type (data not shown). Mutations were generated in pLitJFH-B/K(A) by fusion PCR. pCON-1/JFH-1c3 mutagenesis: a unique BglII–AflII fragment was ligated into pLitmus28i (NEB): pLitCON-1-B/A. Fusion PCR was used to generate an L20F amplimer; this was digested with NotI and ligated into pLitCON-1-B/A. The BglII–AflII fragment was then reintroduced into the full-length chimeric sequence. Constructs were confirmed by double-stranded DNA sequencing; primers and details are available on request.
In Silico Structure Modeling and Drug Binding.
p7 channel models were generated as described31 using Maestro (Schrödinger Inc.). Point mutations were introduced into wild-type structures with subsequent reminimization. The Maestro draw function was used to design molecules that would fit within the density associated with L20. Molecules were subjected to free-energy minimization and stable, bound conformations used as templates for rapid overlay of chemical structures, generating a small panel of molecules including CD. These and adamantane analogues were available from commercial libraries (Maybridge). Pdb files were analyzed and images were captured using PyMol version 0.9 (Delano Scientific). Drug-binding studies against full-channel complexes employed Autodock 4 (Scripps Research Inst., San Diego, CA), Glide (Schrödinger Inc.) and E-Hits (Symbiosys Inc.). Details are available on request.
Bacterial p7 Expression and In Vitro Channel-Forming Assay.
Wild-type and mutant flu antigen–tagged p7 was expressed as a glutathione S-transferase fusion in Escherichia coli, then cleaved and purified as described.17 Real-time measurements of channel activity were performed as described.33
Virus Culture, Drug Inhibition and Live Cell pH Assays.
Huh7 cells were maintained, transfected, and treated with inhibitors as described.21 Intracellular virions were liberated by freezing/thawing,11 and HCV titres were determined by focus-forming assay.21 For live cell imaging, infected cells seeded onto poly-D-lysine–coated cover slips were grown overnight, prior to labeling with Lysosensor Yellow/Blue DND-160 and quantitation of cytoplasmic vesicle pH as described.19
Viral protein western blots of Huh7 lysates at 72 hours posttransfection were performed as described21 using rabbit anti-core (308), mouse anti-E2 (AP33), rabbit anti-p7 2715 (GT2a) and 1055 (GT1b),21, 34 rabbit anti-NS2, sheep anti-NS5A, and mouse anti–glyceraldehyde 3-phosphate dehydrogenase (6CS, Invitrogen), with appropriate horseradish peroxidase–conjugated secondary antibodies (Sigma).
Channel Oligomerization Native Polyacrylamide Gel Electrophoresis.
Five micrograms of MeOH-solubilized flu antigen–p7 protein was dried by evaporation, then resolubilized overnight at room temperature in 20 mM sodium phosphate buffer (pH 7.0) containing 100 mM lyso-myristoylphosphatidylglycerol (LMPG) (monomeric) or 100 mM 1,2-diheptanoyl-sn-glycero-3-phosphocholine (DHPC) (oligomeric),31 incorporating 4 mM rimantadine-HCl (Sigma) or 4 mM N-nonyl deoxynojirimycin (NN-DNJ) (Toronto Biochemicals); 2× native polyacrylamide gel electrophoresis (PAGE) loading dye (150 mM Tris-Cl (pH 7.0), 30% glycerol, 0.05% bromophenol blue) was added and samples were separated on a 4-20% TGX gel (Biorad) prior to staining with Coomassie Brilliant Blue.
Molecular Modeling of p7 Inhibitor Interactions.
We have modeled the heptameric GT1b J4 isolate p7 complex31 with lumenal His17.35 We extended these studies to include a low-pH, open form wherein His17 protonation caused p7 protomers to rotate, inducing channel opening (Fig. 1A). This is consistent with p7 opening being stimulated at low pH,33 as well as cellular proton conductance.19 We also generated a GT2a JFH-1 model (Fig. 1B) with similar structural characteristics to the J4 channel, despite significant sequence diversity.
Autodock 4.0 was used to model binding sites (residue interactions <4 Å) on J4 and JFH-1 channels for amantadine (Ama), rimantadine (Rim), and NN-DNJ. Adamantanes bound to a peripheral, membrane-exposed region of the channel complex (Fig. 1B, left panel), preventing channel opening. The location of this pocket agreed with NMR studies of p7-amantadine interactions36 and overlapped with J4 L(50-55)A, a mutation shown to alter amantadine sensitivity in vitro.31NN-DNJ did not interact with channel complexes, instead docking to p7 monomers at the protomer interface (Fig. 1B, right panel), thus potentially disrupting oligomerization. Accordingly, active nonyl-IS derivatives were predicted to bind this site with >10-fold higher affinity than inactive butyl-derivatives15 (data not shown). Although relatively well conserved in other genotypes (Fig. 1C), variation at these positions may alter compound binding, providing a basis for genotype-dependent sensitivity.21
Adamantane Resistance Is Conferred by a Mutation Observed in Clinical Trials.
J4 and JFH-1 adamantane binding sites contained L20, which mutated to F20 in GT1b patients unresponsive to IFN/Rib/Ama.29 Comparison of predicted binding affinities (Autodock) revealed that Rim bound to wild-type channels with higher affinity compared with Ama, explaining its increased potency.19, 21 Ama-resistant JFH-1 p7 provided a threshold value for effective drug binding (Kd>7.41 μM). L20F increased predicted Kd values for both Ama and Rim above 7.41 μM (Fig. 2A), with one exception. We therefore tested JFH-1 L20F p7 Rim sensitivity in vitro, assessing steady state activity values to measure open channel complexes in drug-bound equilibrium (Fig. 2B). As predicted, L20F was Rim resistant, whereas the number of open wild-type complexes reduced with increasing drug concentration.
JFH-1 and CON-1/JFH-1c3 viruses are Ama resistant, yet susceptible to Rim and NN-DNJ.21 At 72 hours posttransfection and through earlier time points (data not shown), L20F caused no significant defect in the production of intracellular or extracellular infectious virions, and did not disrupt viral protein expression or processing (Fig. 2C). JFH-1 L20F p7 showed slight stabilization compared with wild-type (Fig. 2C), though this was less apparent in the CON-1/JFH-1 background. Addition of p7 inhibitors at ∼IC80 concentrations had no effect on the levels of intracellular infectious HCV, consistent with ion channel activity acting late during infectious virion production. Wild-type secreted infectivity was reduced by Rim and NN-DNJ, but not Ama, whereas Rim had no effect on secreted L20F infectivity (Fig. 2D). L20F NN-DNJ sensitivity was retained, however, and combining Rim with NN-DNJ had an additive effect on wild-type virus but not L20F, supporting separate modes of action (Fig. 3A). Secreted infectivity could not be reduced by more than ∼2 log10 at higher drug concentrations (data not shown), indicative of a low level of ion channel-independent virion production. p7 channel activity therefore enhances, rather than permits, production of infectious HCV.
Because p7 inhibitors specifically block HCV-mediated alkalinization of intracellular vesicles required for virion production,19 we assessed whether L20F prevented Rim inhibition of p7 activity in infected cells using Lysosensor yellow/blue (Fig. 3B). In accordance with infectivity data, vesicular pH in JFH-1 L20F–infected cells was unaffected by increasing Rim concentration, whereas JFH-1–infected cells experienced a Rim-dependent reacidification. L20F adamantane resistance therefore unequivocally links the antiviral effect of p7 inhibitors to the prevention of vesicle alkalinization.
Validation of Molecular Models and the Predicted Adamantane Binding Site Using Novel Inhibitors.
The L20F phenotype provided compelling evidence for the validity of drug binding predictions, yet the possibility remained that resistance occurred by an alternate mechanism. We therefore validated predicted p7–adamantane interactions using drugs as probes for specificity. First, we selected a group of amantadine analogues in rank order of JFH-1 p7 binding from three docking programmes (see Materials and Methods) (Fig. 4A). With one exception, these molecules behaved as expected; those predicted to bind equally or better than Rim inhibited JFH-1 p7 in vitro (Fig. 4B) and achieved equivalent or improved results in culture when added at Rim IC50 (Fig. 4C). Those predicted to bind less well than Rim had no effect. The exception (compound D) indicated that although our models provide a reliable guide to compound binding, they (and molecular docking programmes) are not 100% accurate. Interestingly, effective analogues were not affected by the L20F mutation, despite adamantyl moieties interacting identically with the Ama/Rim binding pocket. However, extended analogue side chains formed additional interactions with A41 and G46, which presumably overcame disruption caused by L20F.
We next designed nonadamantane molecules using the “Draw” function in Maestro with a high predicted affinity for the J4 and JFH-1 binding sites. These were screened in a subgenomic replicon for effects on HCV RNA replication and cell viability (data not shown).21 Compound CD (Fig. 5A) both inhibited GT1b p7 activity in vitro and showed an equivalent antiviral effect to Rim, to which L20F virus was resistant (Fig. 5B,C). To our knowledge, CD is the first molecule designed entirely against a de novo molecular model to display an antiviral effect in culture.
IS Resistance of GT3a HCV Supports an Antiviral Effect Targeting Channel Oligomerization.
GT3a 452 isolate p7 displays resistance to NN-DNJ in vitro and in culture.21 This provided an excellent basis to investigate whether IS targeted oligomerization and to identify resistance polymorphisms. DHPC induces oligomerization of IS-sensitive J4 p7 in vitro, inducing heptameric complexes equivalent to liposomes.31 We therefore assessed whether IS or Rim blocked oligomerization of J4 and 452 p7. NN-DNJ abrogated J4 p7 oligomerization and channel activity, yet 452 p7 activity was insensitive to this drug and oligomerization was not affected (Fig. 6A). Rim did not affect oligomerization, but it inhibited channel activity in both cases, confirming separate modes of action for these inhibitor classes.
Comparing NN-DNJ binding sites revealed variation between J4 and 452 (Fig. 1C), however alignment with other p7 sequences revealed an F25A polymorphism to be covariant with IS resistance. F25 is located on a predicted bulge in the p7 N-terminal helix, which may link with adjacent protomers, but is also predicted to interact with IS head groups (Fig. 1B). We previously showed that J4 F(22, 25, 26)/A p7 formed hyperactive channels in vitro that retained Ama sensitivity.31 We therefore tested whether this mutant or F25A in isolation could rescue p7 oligomerization from NN-DNJ. Both J4 mutant proteins and JFH-1 F25A p7 were insensitive to NN-DNJ in vitro and displayed hyperactive channel phenotypes, consistent with a more open-form channel structure (Fig. 6B). Native PAGE again correlated IS resistance with the formation of drug-resistant oligomeric complexes (Fig. 6C). Interestingly, the major species formed by JFH-1 F25A p7 oligomer migrated more rapidly than other proteins, yet was stable in the presence of NN-DNJ; some heptameric JFH-1 F25A protein was also apparent. All mutant proteins remained sensitive to Rim in vitro (data not shown). We next tested F25A in cell culture and, despite a modest decrease in particle production, the mutant was resistant to both NN-DNJ and N-nonyl deoxygalactonojirimycin (NN-DGJ), but not Rim (Fig. 6D).
This study revealed the mode of action for adamantane and IS p7 ion channel inhibitors and confirmed that single amino acid changes confer resistance to these drugs. The low fitness cost for these mutations observed in culture implies that a minimal genetic barrier to their selection would exist in vivo, explaining the perceived lack of efficacy for p7 inhibitors in clinical trials.
HCV IFN/Rib resistance is a multifactorial phenomenon, involving virus and host-associated factors. This is distinct to resistance against direct-acting STAT-C antivirals, which are host-independent and mediated through single HCV point mutations. According to quasispecies theory, all possible single variants exist within an HCV-infected individual, with selection dependent on fitness. Generation of dual, triple, and further variants becomes exponentially less likely and forms the basis for the successful application of combination therapies. Combination of IFN/Rib with single STAT-C molecules targeting replication therefore suppresses HCV replication through distinct mechanisms. As such, IFN/Rib-resistant HCV will rapidly become resistant to a third STAT-C drug, depending on fitness cost and drug potency, because it is essentially a monotherapy.
For virus assembly inhibitors, resistance would be expected to arise all the more rapidly in IFN/Rib-resistant viruses as no suppression of genome replication occurs. Combinations of assembly inhibitors, however, can suppress RNA virus resistance.37 Our demonstration of distinct, specific antiviral effects for two classes of p7 inhibitor therefore supports that combination with STAT-C therapies, rather than IFN/Rib, may enhance patient responses, because the genetic barrier to dual resistance would be significantly raised. Given that prototype p7 inhibitors have been trialed in patients (amantadine, rimantadine, UT-231b [IS] and BIT225 [amiloride]), these could be rapidly deployed alongside other STAT-C compounds.
Our approach was necessarily based on molecular modeling of p7 ion channel complexes. Models comprised a lumenal N-terminal helix with a conserved His17 proton sensor, analogous to M2 His37. Cu2+-mediated inhibition confirms His17 as lumenal,35 and lowered pH activates GT1b p7.33 Accordingly, modeling p7 under acidic conditions where His17 is protonated induced an opening of the structure (Fig. 1A). We recently showed that p7 induces vesicle alkalinization, protecting intracellular virions from reduced pH.19 Because low pH induces the fusogenic action of HCV glycoproteins,38 p7 may act analogously to M2 from certain influenza A virus strains, where it prevents such change in hemagglutinin.39 Interestingly, secreted HCV virions are acid resistant,19, 40 meaning that an as-yet unidentified maturation event occurs at a late stage of virion production where particles are acid-stabilized. Accordingly, p7 inhibitors do not reduce intracellular infectivity (Fig. 2D), supporting a post-assembly role for p7 proton channel function.
Three separate drug docking programs predicted the same adamantane binding site located on the p7 channel periphery. This reconciles mutagenic33 and NMR36 investigations of inhibitor binding, although Ama was also modeled within the H77 p7 lumen,41 following classical models for M2.42 M2 NMR studies, however, revealed a peripheral binding site,43 although this is much debated.44, 45 For GT1b/2a p7 sequences, the residue composition of the Ama/Rim binding site varied, which caused differences in affinity dependent on both protein sequence and the compound in question; Rim bound significantly more avidly than Ama, explaining Ama's poor antiviral effect in several studies.19, 21, 22, 46 Sequence variation therefore provides a structural basis for altered drug susceptibility.21 In GT1b and 2a sequences, the adamantane binding site contains L20, which was mutated to F20 in unresponsive GT1b IFN/Rib/Ama trial patients.29 L20 does not occur in other HCV genotypes, and GT1a patients in the same study showed no discernable resistance changes, perhaps associated with reduced H77 Ama sensitivity.21, 22 Nevertheless, L20F conferred adamantane resistance to GT1b and 2a in vitro and in culture, and protected proton channel function from Rim in live cells. As resistance denotes specific antiviral effects, this confirms L20F as a genuine resistance mutation arising during natural infection in response to Ama-driven selection. Interestingly, a recent study describing an NMR-based model for monomeric GT1b p7 showed no effect of Ama on channel activity,47 yet of the four amino acid positions where this protein varied from the J4 sequence, two (J4: I19, F44, changed to L19, L44) occurred within the predicted adamantane binding site. These and other variations may affect inhibitor binding to this pocket either directly or indirectly through changes to adjacent residues altering pocket density; further Ama patient studies have observed alternative mutations occurring in this region of the protein.28 No sequence analysis exists on patients receiving IFN/Rib/Rim, yet it would be of interest to determine whether a more potent inhibitor in this setting could drive resistance in other HCV genotypes.
A second measure of molecular model accuracy and the validity of the predicted Ama/Rim binding site involved correlating analogue predicted binding with antiviral activity. With one exception, analogue activity in vitro and in culture correlated with their predicted binding scores relative to Rim, providing further support for the predicted binding pocket. Interestingly, the L20F mutation did not confer resistance to these molecules, indicating that additional binding to p7 through side groups may overcome L20F-mediated disruption of the Ama/Rim binding site. This may represent a viable strategy for raising the genetic barrier to resistance against novel p7 inhibitors. Binding of a bespoke nonadamantane molecule, CD, designed targeting the J4 and JFH-1 binding site was disrupted by L20F, despite being an entirely different chemotype, because its major stabilizing contacts were present within the original pocket. These experiments demonstrate the subtlety and complexity inherent to p7/inhibitor interactions and explain why variations in protein sequence or inhibitor structure can result in different experimental outcomes. Such studies will, however, inform the future development of more potent compounds through iterative refinement and improvement of rational drug design.
From a therapeutics perspective, alkylated IS p7 inhibitors acted through a mechanism distinct from that of adamantanes, providing scope for the development of parallel yet complimentary p7 inhibitor series. In agreement with previous studies,15, 22 docking programs predicted that nonylated IS bound p7 protomers >10-fold more avidly than those with butyl side chains, occluded more of the p7 protomer interface and so disrupted channel oligomerization. IS compounds disrupted J4 p7 oligomerization and channel activity, but not that of resistant 452 protein.21 F residues have been purported to stabilize p7 oligomerization through hydrophobic interactions48 and F25 is predicted to interact with IS head groups in GT1b p7. In 452 p7, F25 is changed to A, and this polymorphism was shown to mediate IS resistance both in vitro and in culture while remaining Rim sensitive. F25A mutants also formed hyperactive channel complexes in vitro which, in the case of JFH-1, appeared to be less stable and migrated differently by native PAGE. Nevertheless, F25A HCV genomes were viable in culture, again showing a low fitness cost for the development of p7 inhibitor resistance.
Resistance to p7 inhibitors mediated by single amino acid changes with little consequence for virus fitness readily explains their ineffectiveness in clinical trials combined with IFN/Rib. Virus rebound has been noted during amantadine mono-26 and triple therapy.27 In addition, relatively high IC50 values compared with other STAT-C molecules and a maximal reduction in virus production of ∼2log10 even for combinations of p7 inhibitors understandably generates skepticism over their usefulness. However, Rim and IS IC50 values in HCV culture are similar to those in influenza A virus and HIV in vitro/culture systems, where they progressed to clinical and trial-stage use in humans, respectively. Given the relatively high degree (∼30% of patients) of breakthrough in trials combining NS3 inhibitors with IFN/Rib,49 the recent success of STAT-C combinations,50 and lessons from HIV, expanding the STAT-C repertoire should be an immediate and ongoing priority. The availability of prototype p7 inhibitors could rapidly expedite this process, and future p7 inhibitors could complement STAT-C therapies as these are implemented over the next decade as an understanding of the molecular basis of resistance assists in the design of novel, more potent compounds.
We are grateful to Takaji Wakita (National Institute for Infectious Diseases, Tokyo, Japan), Charles Rice (Rockefeller University, New York, NY) & Apath LLC (Brooklyn, NY), Jens Bukh (Hvidovre University Hospital, Hvidovre, Denmark), Ralf Bartenschlager (University of Heidelberg, Heidelberg, Germany) and Thomas Pietschmann (University of Hannover, Hannover, Germany) for the provision of full-length HCV constructs. The rabbit anti-core (308) antibody was a gift from John McLauchlan (Centre for Virus Research, Glasgow, UK) and the mouse anti-E2 (AP33) antibody was a gift from Arvind Patel (Centre for Virus Research, Glasgow, UK) and Genentech Inc. (San Francisco, CA). The rabbit anti-NS2 antibody was a gift from Gholamreza Haqshenas (Monash University, Victoria, Australia). We are also grateful to Andrew Macdonald and David Rowlands (University of Leeds, Leeds, UK) for critical reading of the manuscript and informative discussion.