Domain III of NS5A contributes to both RNA replication and assembly of hepatitis C virus particles

Abstract

The hepatitis C virus (HCV) NS5A protein plays a critical role in viral RNA replication and has recently been shown to play a role in particle production in the infectious genotype 2a HCV clone (JFH-1). Here, we show that alanine substitutions of serines 2428/2430 within the C-terminal domain III of NS5A do not affect subgenomic replicon RNA replication but do reduce particle production. In contrast, substitution of serines 2390/2391 had no effect on either RNA replication or particle production. Relative to genotype 1, all genotype 2 HCV isolates contain a 19 residue insertion near the C terminus of domain III which, when deleted (Δ2408–2426), resulted in a delay to both RNA replication and particle production. None of these mutations affected the ratio of basal to hyperphosphorylated NS5A, suggesting that serines between residues 2390 and 2430 are not phosphorylated. We propose that although domain III is dispensable for RNA replication, it nevertheless influences this process.

Hepatitis C virus (HCV) is estimated to infect some 123 million individuals (Shepard et al., 2005). In the majority of cases, the virus establishes a chronic infection which can ultimately result in liver fibrosis, cirrhosis or hepatocellular carcinoma. Combination therapy comprising pegylated alpha interferon (IFN-α) and ribavirin is only successful in approximately 50% of patients. HCV, a member of the family Flaviviridae, is an enveloped virus with a positive sense RNA genome of 9.6 kb. A single open reading frame is flanked by 5′ and 3′ untranslated regions (UTRs) containing cis-acting elements required for RNA replication; the 5′ UTR also contains an internal ribosome entry site (IRES) which mediates cap-independent translation of the 3000 residue polyprotein. This is cleaved co- and post-translationally by host cell and viral proteases to release the structural (core, E1, E2 and p7) and non-structural (NS2, NS3, NS4A, NS4B, NS5A and NS5B) proteins.

NS5A has been shown to have many functions; foremost, as a component of the RNA replication complex, it is absolutely required for viral RNA replication. Structural analysis has revealed that NS5A comprises three domains separated by short low complexity regions (Tellinghuisen et al., 2004) (Fig. 1a). The structure of domain I has been determined; it coordinates a zinc ion and is postulated to dimerize forming a groove through which RNA is predicted to pass (Tellinghuisen et al., 2005). Domains II and III are less structured and more flexible; domain III appears to be dispensable for RNA replication (Tellinghuisen et al., 2008b) and can accommodate a green fluorescent protein (GFP) insert at the C terminus with no adverse effects (Appel et al., 2005; McCormick et al., 2006b; Moradpour et al., 2004). NS5A is a phosphoprotein existing in both a basally phosphorylated (p56) and a hyperphosphorylated (p58) state. Proline-directed kinases such as casein kinase (CK)II have been implicated in basal phosphorylation (Reed et al., 1997) and CKIα has been implicated in hyperphosphorylation (Quintavalle et al., 2007); however, there is little consensus as to the sites of phosphorylation. Early deletion analysis predicted that basal phosphorylation sites were present in both domains II and III (Tanji et al., 1995) and hyperphosphorylation sites have also been mapped to domain II (Katze et al., 2000; Tanji et al., 1995). Interestingly, inhibition of hyperphosphorylation, either pharmacologically or by mutation, enabled replication of a non-culture-adapted genotype 1b subgenomic replicon (Appel et al., 2005; Neddermann et al., 2004).

Fig. 1. Characterization of RNA replication of domain III mutants in the context of the mSGR-luc-JFH-1.

(a) Schematic of the three mutants characterized in this study. S2390/2391A and S2428/2430A mutations are indicated by bold, underlined type. (b) Huh7 cells were electroporated with replicon RNA and harvested into passive lysis buffer at the indicated time points. Luciferase activity was measured as previously described (Macdonald et al., 2003). Data from three independent experiments are shown; bars indicate SEM. (c) At 72 h p.t., samples were analysed by Western blotting with antibodies to NS3 (Aoubala et al., 2001), NS5A (Macdonald et al., 2003) or GAPDH. GND represents a corresponding polymerase-defective construct.

Recently, studies utilizing the genotype 2a infectious clone of HCV (JFH-1) (Wakita et al., 2005) have demonstrated a role for domain III of NS5A in the process of virus assembly (Appel et al., 2008; Tellinghuisen et al., 2008a). In particular, a serine residue at position 2433 was predicted to be phosphorylated by CKII and, when mutated to alanine, reduced virus titres 10 000-fold (Tellinghuisen et al., 2008a). However, neither of these studies considered other conserved serines in domain III so we proceeded to investigate their role in both RNA replication and virus production. NS5A in genotype 2 HCV isolates contains a 19 aa insertion relative to other isolates, located between residues 2408 and 2426 of the JFH-1 NS5A protein. To investigate the potential role of this sequence, we precisely excised it and analysed the RNA replication and virus assembly phenotypes of the resulting mutant.

To facilitate rapid mutagenesis of NS5A in the context of both the luciferase-based subgenomic JFH1 replicon (SGR-luc-JFH-1) and the full-length infectious clone of JFH-1, we engineered unique silent restriction sites flanking the NS5A coding region. Full materials and methods for all experiments are available in JGV Online. The resulting constructs were designated mSGR-luc-JFH-1 and mJFH-1. The RNA replication capacity of the modified subgenomic replicon in a transient luciferase assay, and the ability of the full-length clone to produce infectious virus, were indistinguishable from the parental unmodified versions (Supplementary Fig. S1, available in JGV Online). The modified clones were therefore used as the basis for mutational analysis of NS5A.

To inform a mutagenic analysis of domain III in HCV replication, we first identified highly conserved residues. Two clusters of serines close to the C terminus of the protein, residues 2390/2391 and 2428/2430 (polyprotein numbering), were 100% conserved in all NS5A isolates from all genotypes within the database (Kuiken et al., 2005), suggesting that they played a key role in viral replication. Furthermore, between these two clusters of serines, genotype 2 isolates contained a 19 residue insert (residues 2408–2426 in JFH-1) not present in any other genotypes. We therefore generated three mutants: serines 2390/2391 and 2428/2430 were mutated to alanines (S2390/2391A and S2428/2430A) and the 19 residue insert was precisely excised (Δ2408–2426) (Fig. 1a). These mutations were cloned into both the subgenomic replicon (mSGR-luc-JFH-1) and full-length infectious virus (mJFH-1).

RNAs from these mutant subgenomic replicons were electroporated into Huh7 cells and replication was assessed by measuring luciferase activity over a 72 h time-course. All of these mutations supported robust RNA replication that was indistinguishable from wild-type by 48 h post-transfection (p.t.) (Fig. 1b); however, it is noteworthy that Δ2408–2426 exhibited a delay in RNA replication compared with the wild-type mSGR-luc-JFH-1; this was statistically significant at 24 h p.t., but not at later time points. Lysates were analysed for expression of NS3 and NS5A at 72 h p.t. (Fig. 1c). Interestingly, viral protein expression was higher for the S2390/2391A and S2428/2430A mutants compared with wild-type. As expected, the Δ2408–2426 NS5A migrated more rapidly than wild-type protein, but it also exhibited a lower abundance. We do not believe this is due to loss of an epitope, as the polyclonal serum used for Western blotting was raised to bacterially expressed genotype 1b NS5A (lacking the 19 residue insert) (Macdonald et al., 2003). Of note, none of the mutants exhibited an alteration in the ratio of basal to hyperphosphorylated NS5A, suggesting that neither the specific serine residues that we mutated, nor the two serines within the 19 residue insertion, are sites of phosphorylation.

As the three mutants discussed above did not abrogate RNA replication, we proceeded to investigate their effects on virus assembly and release in the context of the mJFH-1 infectious virus. Following transfection of in vitro-transcribed RNA, both extra- and intracellular virus titres were determined over a 72 h time-course. At 24 h p.t., mJFH-1(Δ2408–2426) exhibited significantly reduced levels of both extra- and intracellular virus; however, at later time points, this difference became less apparent (Fig. 2). At all time points, mJFH-1(S2390/2391A) was indistinguishable from wild-type; however, mJFH-1(S2428/2430A) exhibited a reduction of between 10- and 100-fold in both extra- and intracellular virus titres at 48 and 72 h p.t. Interestingly, at 24 h p.t., no extracellular virus was detectable for the S2428/2430A mutant [compared with a wild-type titre of 10⁴ focus-forming units (f.f.u.) ml⁻¹], whereas the defect in intracellular virus titre was less apparent at this time point (approximately 1-log reduction). This observation suggested that the S2428/2430A mutation not only caused a reduction in overall levels of virus assembly but also resulted in a delay to release.

Fig. 2. Characterization of virus assembly and release of domain III mutants.

(a) Huh7 cells were electroporated with the indicated virus RNAs and virus release into the culture supernatant was measured by a focus-forming assay. (b) Intracellular virus titres were measured by a focus-forming assay following cell disruption by repetitive freeze–thaw at 72 h p.t. (c, d) Extracellular (c) and intracellular (d) genomic RNA levels were measured at 120 h p.t. by quantitative RT-PCR. Data from three independent experiments are shown; bars indicate SEM. GND represents a corresponding polymerase-defective construct.

We used quantitative RT-PCR to determine the abundance of both extra- and intracellular viral genomic RNA. As previously noted (Wakita et al., 2005), the number of genomes released from wild-type-infected cells was approximately 1000-fold higher than the infectivity (as measured by focus formation) (Fig. 2c). The genome release from cells infected with mJFH-1(Δ2408–2426) or mJFH-1(S2390/2391A) was comparable to wild-type, whereas mJFH-1(S2428/2430A) showed a 100-fold reduction. To an extent, these results were mirrored by the abundance of intracellular genomes, although interestingly, mJFH-1(S2428/2430A), despite having little effect on subgenomic replicon RNA replication (Fig. 1b), exhibited a level of intracellular RNA that was 10-fold lower than wild-type. This apparent discrepancy may point to additional controls of RNA replication in the context of the full-length genome that are not apparent in the context of the subgenomic replicon. The fact that mJFH-1(S2428/2430A) showed only a 10-fold reduction in intracellular RNA, yet all other parameters (extracellular RNA and both intra- and extracellular infectivity) were reduced 100-fold, is consistent with defects in both virus assembly and release.

To characterize the phenotypes of the three mutants further we examined the intracellular accumulation and localization of viral structural and non-structural proteins. Interestingly, at 48 h p.t., levels of all viral proteins analysed were reduced for mJFH-1(Δ2408–2426), whereas expression levels were comparable to the wild-type for the other two mutants (Fig. 3a). The same overall pattern of expression was observed at 24 and 72 h p.t. (data not shown). Again, in common with the subgenomic replicons, there was no significant effect of the mutations on the ratio of basal to hyperphosphorylated NS5A, further strengthening the notion that these serines are not sites of phosphorylation. Analysis of the subcellular distribution of NS5A and Core in infected cells revealed that in cells infected with wild-type or the three mutants, the distribution of Core was similar, showing a cytoplasmic ring-like pattern consistent with the localization of Core to lipid droplets, as previously described (Barba et al., 1997). In wild-type-, mJFH-1(Δ2408–2426)- and mJFH-1(S2390/2391A)-infected cells, NS5A was closely associated with the Core staining; however, in the case of S2428/2430A, the juxtaposition of the NS5A and Core staining was less apparent (Fig. 3b) and the distribution of the two proteins was largely distinct.

Fig. 3. Viral protein expression in infected cells.

(a) Huh7 cells were electroporated with the indicated virus RNAs and harvested at 48 h p.t. by lysis in Glasgow lysis buffer (Harris & Coates, 1993). Protein (10 μg) was analysed by Western blotting with antisera to Core, NS2, NS5A and GAPDH as indicated. (b) Immunofluorescence analysis of NS5A (green) and Core (red) in infected cells at 72 h p.t. Representative images are shown.

Our data point to an important role for domain III of NS5A in the process of virus assembly and/or release, and in this regard are in broad agreement with the findings of previous studies (Appel et al., 2008; Tellinghuisen et al., 2008a). During the course of this study, it was reported (Masaki et al., 2008) that a triple serine–alanine substitution mutation (S2428/2430/2433A) resulted in a 1000-fold reduction in intracellular virus and a 10 000-fold reduction in extracellular virus, accompanied by a 10-fold reduction in Core release. In this study, a 10-fold reduction in Core release for the S2428/2430A mutant was observed, although a corresponding reduction in infectivity was not presented. Thus, although postulated (Masaki et al., 2008), it has not been formally demonstrated that serines 2428/2430 are required for assembly and release of infectious virus particles, as even a ΔE1/E2 construct (lacking any measurable infectivity) released only ~20-fold less Core than wild-type JFH-1 (Pietschmann et al., 2006). Thus, our data confirm that S2428/2430A exhibits reductions in both intra- and extracellular virus, consistent with a defect in virus assembly, although interestingly this mutant also exhibits a delay in virus release. In contrast, the S2390/2391A mutant was indistinguishable from wild-type, consistent with the lack of effect seen for a multiple alanine substitution mutation targeting these residues together with serines 2384 and 2388 (Masaki et al., 2008). All of these studies point to a role of the 15 C-terminal residues of NS5A in virus assembly.

An interaction between NS5A and Core in JFH-1-infected cells has been described previously (Masaki et al., 2008) and the two proteins were reported to co-localize on lipid droplets (Miyanari et al., 2007). Furthermore, alanine substitution of serines 2428/2430 was shown to disrupt both the NS5A–Core interaction and colocalization (Masaki et al., 2008). Although we have not undertaken a coimmunoprecipitation analysis, we did observe a lack of NS5A colocalization with Core in cells infected with the S2428/2430A mutant, compared with wild-type and the S2390/2391A or Δ2408–2426 mutants. These data are consistent with a model in which the 15 C-terminal residues of NS5A interact with Core and promote virus assembly. However, we should stress that this is not an absolute requirement, as the S2428/2430A mutant is still able to produce infectious virus particles, albeit at only ~1% of wild-type efficiency.

Our study also investigated the potential role of a 19 residue insertion in NS5A, located just N-terminal to serines 2428/2430 and found exclusively in genotype 2 isolates. When precisely excised (Δ2408–2426) in the context of the subgenomic replicon, this mutant displayed delayed kinetics of RNA replication that were reminiscent of the genotype 1b subgenomic replicons and suggest that this sequence may play some role in the efficient establishment of an RNA replication complex at early time points. Once genome replication is ongoing, however, there is clearly no defect in the rate of RNA synthesis. Deletion of the 19 residue insertion appeared to have a detrimental effect on viral protein expression levels in both the subgenomic replicon and the full-length infectious system. As replicated viral RNA must be utilized for both virus assembly and protein translation, it is possible that this sequence influences the fate of the genomic RNA such that, when it is deleted, less RNA is directed to ribosomes for protein translation. In this regard, we have previously proposed a link between NS5A and control of the replication/translation switch that must occur in all positive-strand RNA viruses (McCormick et al., 2006a). Alternatively, it may be that the deletion induces instability or degradation of the viral replication complex. The high level of sequence conservation within this insertion (Kuiken et al., 2005) suggests an important functional role, although both 1b and JFH-1 will tolerate the insertion of foreign sequences at a site just N-terminal to the genotype 2 insertion. For example, replication of the JFH-1 subgenomic replicon is unaffected by insertion of GFP at residue 2398 (Jones et al., 2007), although in the full-length JFH-1, this insertion results in a 10-fold reduction in Core release (Masaki et al., 2008). These data also point to a role of the 19 residue genotype 2 insertion in virus assembly or release, consistent with our results (Fig. 2). Further analysis will be required to dissect the complex interplay between virus RNA replication and particle assembly in HCV and to elucidate the multiple roles of NS5A.