Requirement of cellular DDX3 for hepatitis C virus replication is unrelated to its interaction with the viral core protein

The cellular DEAD-box protein DDX3 was recently shown to be essential for hepatitis C virus (HCV) replication. Prior to that, we had reported that HCV core binds to DDX3 in yeast-two hybrid and transient transfection assays. Here, we confirm by co-immunoprecipitation that this interaction occurs in cells replicating the JFH1 virus. Consistent with this result, immunofluorescence staining of infected cells revealed a dramatic redistribution of cytoplasmic DDX3 by core protein to the virus assembly sites around lipid droplets. Given this close association of DDX3 with core and lipid droplets, and its involvement in virus replication, we investigated the importance of this host factor in the virus life cycle. Mutagenesis studies located a single amino acid in the N-terminal domain of JFH1 core that when changed to alanine significantly abrogated this interaction. Surprisingly, this mutation did not alter infectious virus production and RNA replication, indicating that the core–DDX3 interaction is dispensable in the HCV life cycle. Consistent with previous studies, siRNA-led knockdown of DDX3 lowered virus production and RNA replication levels of both WT JFH1 and the mutant virus unable to bind DDX3. Thus, our study shows for the first time that the requirement of DDX3 for HCV replication is unrelated to its interaction with the viral core protein.

INTRODUCTION Persistent hepatitis C virus (HCV) infection is a major cause of chronic hepatitis, cirrhosis and hepatocellular carcinoma (HCC). Current treatments for chronic infection are ineffective in approximately 50 % of patients (Chen & Morgan, 2006). The virus, which belongs to the family Flaviviridae, has a positive-sense RNA genome encoding a polyprotein that is cleaved by cellular and viral proteases to yield mature structural and non-structural proteins. The structural proteins are core and the envelope glycoproteins E1 and E2, while the non-structural proteins are p7, NS2, NS3, NS4A, NS4B, NS5A and NS5B (reviewed by Moradpour et al., 2007). HCV exhibits a high degree of genetic variability with six distinct viral genotypes, each further divided into subtypes, and within a single In addition, D1 also interacts with several cellular factors (McLauchlan, 2000). The hydrophobic D2 domain, which spans amino acid residues 118 to approximately 169, is required for correct folding of D1. D2 consists of two amphipathic a-helices connected by a hydrophobic loop and mediates core association with lipid droplets (LDs) and endoplasmic reticulum membranes (Barba et al., 1997;Boulant et al., 2006;Hope & McLauchlan, 2000). Recent evidence strongly suggests that the HCV core-LD association is important for the production of infectious virus particles (Boulant et al., 2007;Miyanari et al., 2007;Shavinskaya et al., 2007).
Previously, we proposed that the HCV core-DDX3 interaction might function in virus replication and/or pathogenesis (Owsianka & Patel, 1999). Two separate studies recently provided evidence for the involvement of DDX3 in HCV replication (Ariumi et al., 2007;Randall et al., 2007). The importance of DDX3 in the life cycle of other viruses has also become apparent, including its requirement for human immunodeficiency virus-1 (HIV-1) replication (Yedavalli et al., 2004). Furthermore, the DDX3 homologue Ded1 is required for brome mosaic virus replication (Noueiry et al., 2000). In contrast, DDX3 inhibits hepatitis B virus transcription by its incorporation into nucleocapsids (Wang et al., 2009). Thus, there is increasing evidence that DDX3 is important in the life cycle of several diverse viruses.
The direct relevance of the core-DDX3 interaction for the HCV life cycle has not yet been elucidated. In this study, we show that this interaction does occur in cultured cells infected with HCV strain JFH1 (Wakita et al., 2005), and that DDX3 co-localizes directly with core on LDs. Using alanine substitution mutagenesis, we identified a key amino acid residue within the D1 domain of core that is (a) Schematic representation of DDX3 protein structure showing the nine conserved motifs, including motif II (or the DEAD motif). The arrows mark the epitopes recognized by the mouse mAbs. (b) Characterization of anti-DDX3 antibodies. HuH-7 cell lysates were immunoprecipitated with the immune (I) and pre-immune (PI) antiserum R648, or mAb AO196 or an isotype control mAb. The immune complexes were analysed by Western blotting using biotinylated R648. Positions of molecular mass markers are shown (in kDa). (c) Western blotting of HuH-7 cell lysates with R648-I, -PI, mAb AO196 and isotype control mAb. (d) Western blotting of lysates of HEK cells transfected with a plasmid expressing EGFP alone or an EGFP-DDX3 fusion protein using anti-GFP mAb, AO196 and R648. critical for this interaction. Furthermore, we examine the impact of abrogation of the core-DDX3 interaction on virus RNA replication and infectious progeny yields in cultured cells.

Generation of DDX3 antibodies
We previously showed that DDX3 co-localizes with HCV genotype 1a core in cytoplasmic punctate spots (Owsianka & Patel, 1999). To better understand the nature of this interaction, we generated several mouse mAbs (prefixed with 'AO', see Fig. 1a) and two rabbit polyclonal antisera (R647 and R648) to human DDX3. Both rabbit antisera and the majority of mAbs were found to interact with DDX3 by ELISA, Western blotting and immunoprecipitation assays (data not shown). The epitopes of mAbs reactive to DDX3 in Western blots were broadly mapped using truncated forms of DDX3 expressed in bacteria as GST-fusion proteins. Most of the mAbs recognized aa 409-473, a region located between motifs III and V of DDX3, whereas mAb AO190 recognized motifs V and VI. mAbs AO166 and AO196 bound to the N-terminal region, whereas mAbs AO2 and AO35 recognized the region downstream preceding the ATPase domain of DDX3 (Fig. 1a).
The specificities of antibodies AO196 and R648 for DDX3 in immunoprecipitation and Western blot assays are shown in Fig. 1(b, c, d). The polyclonal serum R648 (but not the pre-immune control serum) specifically immunoprecipitated DDX3 from a cytoplasmic extract of HuH-7 cells, as did mAb AO196 (Fig. 1b). R648 and AO196 also recognized DDX3 in cytoplasmic extracts of HuH-7 cells by Western blotting (Fig. 1c). The specificity of these antibodies was further demonstrated by their recognition of both endogenous DDX3 and an enhanced green fluorescent protein (EGFP)-DDX3 fusion protein expressed in HEK cells (Fig. 1d). mAb AO196 and R648 recognized only the cytoplasmic form of DDX3 by immunofluorescence ( Fig. 2a and data not shown), even though they were able to detect DDX3 in the nuclear extracts of cells in both immunoprecipitation and Western blotting assays (data not shown). In contrast, another mAb (a) DDX3 co-localizes with core in JFH1infected cells. HuH-7 cells were either mockinfected (lower panel) or infected with JFH1 (top panel). At 3 days post-infection, cells were fixed and probed with HCV core and DDX3 using antibodies R308 and A0196, respectively. Cell nuclei were detected by counterstaining with DAPI. (b) Co-immunoprecipitation of core by anti-DDX3 antiserum. HuH-7 cells electroporated with viral RNA were lysed at 72 h post-incubation, and the lysate immunoprecipitated with anti-DDX3 R648 immune (I) or pre-immune (PI) serum. The resulting precipitates were examined by immunoblotting using anti-core mAb c7-50 (top panel). One twentieth of the cell lysate used in the coimmunoprecipitation assay was immunoblotted for core with mAb c7-50 as the input control (bottom panel). (c) Localization of DDX3 and core on LDs. HuH-7 cells electroporated with the JFH1 WT RNA were fixed at 72 h posttransfection and probed using antibodies to core (R308), DDX3 (AO196) and ADRP. Zstack analysis of all three proteins was performed by recording a series of approximately 20 images. 3D reconstructions of the boxed areas are shown (i) and a selected area is shown in greater detail (ii) where DDX3 is depicted as a wire frame to reveal the core-ADRP association. Bars, 10 mm for confocal images and 2 mm for 3D image.

HCV core interacts with cellular DDX3 in cells replicating WT JFH1
The core-DDX3 interaction was previously demonstrated in cells ectopically expressing HCV genotype 1a core, and in the absence of virus RNA replication and productive infection (Owsianka & Patel, 1999). To test whether DDX3 interacts with JFH1 core, HuH-7 cells transfected with virus RNA were first analysed by indirect immunofluorescence. DDX3 was found to be sequestered by core in JFH1replicating cells as detected using mAbs AO196 and AO322 [ Fig. 2a and Supplementary Fig. S1a (available in JGV Online), respectively]. It should be noted that the corebound DDX3 emitted a stronger fluorescent signal compared with free DDX3 occupying the cell cytoplasm. Thus, adjustment of the core-bound DDX3 signal to prevent overexposure resulted in the apparent loss of free DDX3 from the image. A representative image of core and DDX3 in virus-infected and surrounding non-infected cells taken at two different intensities is shown in Supplementary Fig. S1(b) (available in JGV Online). The anti-DDX3 antiserum R648 specifically co-immunoprecipitated core from JFH1-replicating cells, further confirming a direct interaction between these two proteins ( Fig. 2b).

DDX3 interacts with HCV core on LDs
The association of HCV core with LDs is essential for infectious virus production (Boulant et al., 2007;Miyanari et al., 2007;Shavinskaya et al., 2007). We sought to gain further insight into the core-DDX3 complex in the context of the JFH1 system by analysing the intracellular distribution of core and DDX3. HuH-7 cells electroporated with JFH1 RNA were analysed by indirect immunofluorescence for HCV core, DDX3 and adipocyte differentiation-related protein (ADRP), a protein abundant on the surface of LDs (Fujimoto et al., 2004). As shown in Fig. 2(c) left, core associated with ADRP and the localization of DDX3 was precisely coincident with both proteins, indicating its redistribution to LDs. To examine the subcellular position of DDX3 and core in relation to ADRP in greater detail, a series of Z-stacks was obtained and used after blind deconvolution to create a three-dimensional model of LDs coated by core, ADRP and DDX3. As reported before (Boulant et al., 2007), core fully coated LDs. We found that DDX3 by virtue of its interaction with core was also fully associated with LDs [ Fig. 2c (i) and (ii)].

Identification of core residues required for DDX3 interaction
We previously reported that aa 1-59 of core protein are involved in its interaction with DDX3 (Owsianka & Patel, 1999). To identify critical residues within this region, a library of core 1-59 mutants containing single or multiple amino acid substitutions was generated by error-prone PCR (EP-PCR). The mutated sequences were fused inframe with GFP and expressed in bacteria. The mutant fusion proteins were screened by incubation with GST-DDX3 fusion protein immobilized in ELISA wells, and the bound mutants detected using a rabbit polyclonal anti-GFP antiserum. Of 130 clones screened, only nine were found to be defective in binding to GST-DDX3 (data not shown). In order to confirm these results, all nine mutants were individually subcloned into a sequence encoding the HCV genotype 1a strain H77c core, E1 and E2 (pCE1E2) in a mammalian expression vector pcDNA3.1/Zeo+ (Invitrogen), and their expression was analysed by indirect immunofluorescence. The results were in accordance with the data from the initial ELISA screen in that all nine mutated core proteins failed to redistribute DDX3 (data not shown). Nucleotide sequence analysis showed that each of the mutants carried between one and four amino acid substitutions (Fig. 3). Of interest, mutant 90 had only one substitution (I30N), indicating that this residue must be required for the interaction of core with DDX3. All nine mutant proteins had at least one amino acid substitution in the region spanning residues 24-36, indicating that this region may harbour residues that are critical for the core-DDX3 interaction. To test this hypothesis, site-directed mutagenesis was carried out to revert any mutations outside of this 13 aa region back to the WT residue. These new mutant core 1-59 fragments were then subcloned into pCE1E2, and their expression analysed following transient transfection into HuH-7 cells as above. Again, none of these new core mutants redistributed DDX3 (data not shown), indicating that the 13 aa region between residues 24 and 36 of core is indeed involved in the interaction between core and DDX3. To determine which residues in this 13 aa region were essential, we carried out alaninescanning mutagenesis across aa 24-36, individually substituting each residue in this region with alanine. As before, these alanine mutant sequences were subcloned into pCE1E2 and transiently transfected into HuH-7 cells. Immunofluorescence analysis revealed seven mutants (P25A, G26A, G28A, Q29A, V31A, G32A and L36A) that showed distinct co-localization between core and DDX3 (similar to that seen with WT HCV core), whilst the other mutants (F24A, G27A, I30A, G33A, V34A and Y35A) displayed no interaction with DDX3 at all ( Supplementary  Fig. S2, available in JGV Online). Thus, these results indicate that core residues F24, G27, I30, G33, V34 and Y35 are critical for its interaction with DDX3.
Disrupting the core-DDX3 interaction Sequence comparison of core protein from strains H77c (genotype 1a) and JFH1 (genotype 2a) revealed 96.6 % identity within the first 59 residues of core, with the key residues (F24, G27, I30, G33, V34 and Y35) for interaction with DDX3 being fully conserved (data not shown). Thus, recombinant JFH1 genomes each containing one of the six mutations were constructed. All mutants were replication competent as seen by the expression of NS5A protein at 3 days post-transfection (Fig. 4). The viral core protein in these cells was detected using two different anti-core antibodies, which bound to the mutant proteins with varying affinity. mAb c7-50 bound minimally or not at all to I30A and G33A mutants, but had increased affinity for the V34A protein. The anti-core rabbit serum R308 on the other hand recognized all core forms equally except for G33A (Fig. 4). Notably, all six core mutations are located within the epitope (aa 21-40) recognized by c7-50 (Moradpour et al., 1996), which may account for the altered affinity of this mAb to some mutants. The antiserum R308 was raised against a peptide corresponding to core aa 5-25 (Hope & McLauchlan, 2000). Therefore, the smaller quantities of G33A core detected using R308 suggest either this mutation lowers the stability of the protein or simply reduces the binding efficiency of the antibody. Importantly, there was no detectable change in DDX3 protein levels in any of the transfected cell cultures tested (Fig. 4).
We next tested by co-immunoprecipitation if these mutations had effectively disrupted the core-DDX3 interaction in virus-replicating cells. We excluded mutants I30A, G33A and V34A from this assay due to their altered affinity to mAb c7-50 (Fig. 4), which is the antibody we found to be most compatible in our co-immunoprecipitation assay. As shown in Fig. 5, the F24A and G27A core mutants were co-immunoprecipitated by the anti-DDX3 antiserum R648 in much lower amounts than the WT protein. However, the Y35A core was not co-immunoprecipated at all, indicating that its interaction with DDX3 was abrogated. Consistent with these findings, no co-localization of core and DDX3 was seen in cells replicating JFH1 Y35A by immunofluoresence analysis (Fig. 5b). Importantly, this mutant core remained associated with LDs, which is not surprising since motifs responsible for targeting core to LDs are located in the D2 domain (Hope & McLauchlan, 2000).

HCV replication in the absence of the core-DDX3 interaction
To determine the importance of this interaction in the virus life cycle, we examined the phenotype of JFH1 Y35A . Fig. 3. Identification of core residues critical for its interaction with DDX3. Amino acid substitutions in residues 1-59 of core in nine mutants unable to interact with DDX3. Residues in the region 24-36 (shaded box) were targeted for alanine-scanning mutagenesis, which identified residues at positions 24, 27, 30, 33, 34 and 35 (underlined in the sequence at the top) that were critical for DDX3 interaction. Fig. 4. Analysis of JFH1 core mutant viruses. HuH-7 cells electroporated with viral RNAs as shown were subjected to Western immunoblotting at 72 h post-incubation using anti-NS5a mAb 9E10, anti-core antibodies mAb c7-50 or R308, anti-DDX3 mAb AO196 and an anti-tubulin antibody.
HuH-7 cells were electroporated with JFH1 WT , JFH1 Y35A or the replication-deficient JFH1 GND RNAs and incubated for various time points before determining the progeny virus yields released into the medium and intracellular viral RNA levels. As shown in Fig. 6(a), the number of infectious particles released by JFH1 Y35A -replicating cells was lower in comparison with JFH1 WT -transfected cells at 24 and 48 h post-transfection, although parity was achieved at 72 h. The intracellular RNA replication levels of JFH1 Y35A were similar to JFH1 WT throughout the time-course (Fig. 6b). As expected, no virus release or intracellular viral RNA replication was detected in cells harbouring JFH1 GND RNA (Fig. 6a, b). We next examined the infectivity and replication of the core mutant virus released from electroporated cells. To do this, we infected naïve cells at an equal m.o.i. and quantified the released infectious virus progeny and the intracellular viral RNA at various times post-infection. As shown in Fig. 6(c, d), respectively, both the virus yields and RNA replication levels of JFH1 Y35A were very similar to JFH1 WT . We found no direct reversion or second-site mutation in JFH1 Y35A core sequence. Thus, our data indicate that the core-DDX3 interaction does not play a role in HCV morphogenesis.

Replication of HCV following siRNA-mediated knockdown of DDX3
Recent reports have shown that siRNA-mediated knockdown of DDX3 reduces JFH1 WT replication in infected cells (Ariumi et al., 2007;Randall et al., 2007). To assess the influence of DDX3 abundance on virus infection in the absence of the core-DDX3 interaction, we measured the replication of JFH1 WT and JFH1 Y35A following infection of  cells that had undergone efficient siRNA knockdown of DDX3 (Fig. 7a). As previously described, DDX3 knockdown cells were substantially less permissive for virus replication than normal cells following infection with JFH1 WT . Similar reductions in JFH1 Y35A replication levels were also observed (Fig. 7b). Collectively, the data presented in Figs 6 and 7 indicate that DDX3 promotes efficient HCV infection by processes that are independent of its interaction with the viral core protein.

DISCUSSION
The recent advent of the HCV cell culture system has enabled us to further characterize the core-DDX3 interaction in the context of the complete HCV life cycle. Our overall goal is to deduce the role of DDX3 in normal cells with a view to understanding the significance of its interaction with HCV core in the virus life cycle and pathogenesis. Towards this end, we generated a large panel of mAbs recognizing different regions of DDX3. Using a subset of these, we confirmed that this interaction is genuine in JFH1-infected cells.
Given the putative function of DDX3 in RNA metabolism, initial identification of the core-DDX3 interaction led us and others to postulate its possible role in virus replication and assembly (Mamiya & Worman, 1999;Owsianka & Patel, 1999;You et al., 1999). Recent data using the JFH1 system suggest that HCV RNA replication and virus assembly occur in LD-associated membranes (Boulant et al., 2007;Miyanari et al., 2007). Core protein, which associates with LDs (McLauchlan, 2000;Roingeard & Hourioux, 2008), recruits the viral non-structural proteins, replication complexes and envelope glycoproteins to these sites, allowing virus assembly to proceed in this local environment (Miyanari et al., 2007). We show here that core also recruits the cellular DDX3 to LDs, suggesting that it may have a function in HCV replication. In keeping with this, two recent studies have shown reductions in JFH1 replication when DDX3 is removed from the cell by siRNA (Ariumi et al., 2007;Randall et al., 2007). We hypothesized that disrupting the association between HCV core and DDX3 might be detrimental to HCV replication, and provide insights into the purpose of this interaction. Our mutagenesis analysis revealed that the Y35A substitution in the JFH1 core molecule abrogated this interaction without affecting core-LD association. No alteration to virus RNA replication, translation or infectious particle production was observed following transfection of this mutant viral RNA into the cells. Similarly, JFH1 Y35A virus yields from cells infected with the mutant progeny were equivalent to those of the WT virus. Interestingly, the two other mutants identified in this study (JFH1 F24A and JFH1 G27A ) showed greater reductions in virus replication, particularly postinfection ( Supplementary Fig. S3, available in JGV Online). However, given their interaction with DDX3 is only slightly reduced (Fig. 5), this impairment is unlikely to be related to the disruption of the core-DDX3 interaction. A more plausible explanation for their unusual phenotypes could be an alteration in core function and/or RNA structure resulting from an alanine substitution at these positions, and as such these mutants require further investigations. Nevertheless, our results collectively suggest that the core-DDX3 interaction plays no role in virus morphogenesis.
Subgenomic replicons that do not possess core replicate as well as, if not better than, replicons encoding the entire polyprotein (Blight et al., 2000;Ikeda et al., 2002;Lohmann et al., 1999), which indicates that the core-DDX3 interaction per se is not essential for HCV RNA replication. Nevertheless, Ariumi et al. (2007) reported that subgenomic replicons showed a twofold decrease in RNA replication in DDX3 knockdown cells. This finding raises an intriguing prospect that DDX3 may play a direct role in HCV RNA replication independent of and in addition to its interaction with core. Nonetheless, these workers observed a much greater reduction in JFH1 RNA replication in cells supporting replication of full genomelength RNAs, indicating greater importance of DDX3 in this setting and supporting the functional relevance of the interaction of DDX3 with core. We reproduced the deleterious effects of DDX3 knockdown on the JFH1 WT replication but also found a similar phenotype for JFH1 Y35A (Fig. 7). This finding further supports the notion of the core-DDX3 interaction having no function in HCV replication. Therefore, our data make a clear distinction between the effects of siRNA knockdown of DDX3 from the cells and disruption of the core-DDX3 interaction on HCV replication.
The possible involvement of the core-DDX3 interaction in pathogenesis cannot be discounted. Indeed, core protein has been widely implicated in modulating cellular functions, mainly due to its interaction with numerous host factors (McLauchlan, 2000;Ray & Ray, 2001;Watashi & Shimotohno, 2003). Differential regulation of DDX3 has been reported in a number of tumours, including HCC, suggesting that it may be involved in HCV-associated pathogenesis (Botlagunta et al., 2008;Chang et al., 2006;Chao et al., 2006;Huang et al., 2004). DDX3 is also one of several DEAD-box proteins that are upregulated during HIV-1 replication (Krishnan & Zeichner, 2004). Interestingly, it is also upregulated during immune response to lipopolysaccharide-induced inflammation and during interferon treatment (de Veer et al., 2001;Saban et al., 2006). Thus, it is possible that during active viral replication HCV recruits DDX3 protein to evade host immune recognition, thereby adversely affecting one or more of its roles in normal cells (Rosner & Rinkevich, 2007) and in turn contributing to viral pathogenesis. Although we observed no differences in DDX3 expression levels between JFH1-infected cultured cells and uninfected cells, its relevance in vivo from a long-term disease perspective warrants further investigation.

METHODS
Cell culture and antibodies. Human hepatoma HuH-7 (Nakabayashi et al., 1982) cells were propagated as described previously (Witteveldt et al., 2009). The anti-NS5A mouse monoclonal antibody (mAb) 9E10 was a kind gift from Charles M. Rice (Center for the Study of Hepatitis C, The Rockefeller University, New York, USA) (Lindenbach et al., 2005). The anti-HCV core rabbit serum R308, and the anti-ADRP sheep serum have been described previously (Hope & McLauchlan, 2000;Targett-Adams et al., 2003). The anti-HCV core mAb c7-50 and the anti-a-tubulin and anti-GFP mAbs were purchased from Bioreagents and Sigma, respectively.
Generation and characterization of anti-DDX3 antibodies.
BALB/c mice and rabbits were immunized with a bacterially expressed glutathione S-transferase-full-length DDX3 (GST-DDX3) fusion protein and antibodies generated essentially as previously described (Clayton et al., 2002). A total of two rabbit polyclonal antisera (R647 and R648) and 16 mouse hybridomas secreting mAbs (see Fig. 1) to DDX3 was generated. The mAbs were initially identified and selected on the basis of their ability to interact with GST-DDX3, but not an irrelevant GST fusion protein, in an ELISA assay (data not shown).
Plasmid constructs and mutagenesis. The full-length DDX3 gene was cloned into the BamHI site, in-frame to sequences encoding the EGFP, in the vector pEGFP-C1 (Clontech) to generate EGFP-DDX3 fusion protein.
To identify residues within the N-terminal 59 aa of HCV core that are critical for interaction with DDX3, EP-PCR was performed to introduce random mutations into the HCV genotype 1a strain H77c (Yanagi et al., 1997) sequence encoding this region. The fidelity of Taq polymerase was decreased by altering the relative dNTP concentrations, using a high Mg 2+ concentration and including Mn 2+ in the reaction as described previously (Pritchard et al., 2005). The mutated sequences were cloned in-frame with (and downstream of) green fluorescence protein (GFP) in the bacterial expression vector pKK223-3 (Pharmacia) and the library of core 1-59 mutant-GFP fusion proteins generated was screened in an in vitro DDX3binding assay as described in Results. Nucleotide sequencing subsequently identified the mutations introduced by EP-PCR in the clones of interest.
The plasmid pUC-JFH1 carries the full-length cDNA of the genotype 2a HCV strain JFH1 (Wakita et al., 2005). The plasmids pUC-GND JFH1 and pUC-JFH1 DE1E2 are identical except that they carry the GND mutation in the NS5B-encoding sequence, or an in-frame deletion in the E1 and E2 sequences (Wakita et al., 2005). Sitedirected mutagenesis was carried out using the QuikChange-II kit (Stratagene) to introduce alanine substitutions at the target sites in core. Briefly, various alanine substitutions in the core-coding region were individually introduced into the plasmid pGEM-T (Promega) carrying nt 1-2614 (corresponding to the 59 UTR and core to E2 coding sequence) of HCV strain JFH1 using appropriate primers (the sequences of which are available upon request). The presence of the desired mutation in the resulting clones was confirmed by nucleotide sequencing. Sequences carrying the appropriate mutation were subcloned back into pJFH1-pUC to generate mutant viruses (see below).
Generation of JFH1 virus and its mutated derivatives. Ten micrograms of RNA synthesized in vitro using linearized plasmid templates carrying the wild-type (WT) or mutated JFH1 genomic cDNA was electroporated into HuH-7 cells as described previously (Wakita et al., 2005). The electroporated cells were seeded into appropriate tissue culture dishes and incubated at 37 uC. At the indicated time period, the medium containing the infectious virus progeny was filtered through a 0.45 mm pore-sized membrane and infectivity determined as described below.
Determination of virus infectivity and RNA replication. Virus titres in the culture supernatants were determined as TCID 50 (Lindenbach et al., 2005) following immunostaining for NS5A. The intracellular and extracellular HCV RNA content was measured by reverse transcription-quantitative real-time-polymerase chain reaction (RT-qPCR) as described previously (Witteveldt et al., 2009). To determine virus replication after electroporation, cells transfected with the respective transcript were seeded into 10 cm culture dishes, incubated at 37 uC for 4 h and then trypsinized. Total RNA was prepared from 1/15 of the trypsinized cells using the RNeasy kit (Qiagen) to quantify viral RNA by qRT-PCR assay. The remaining cell suspension was split 1 : 3 into three T25 flasks. Following incubation at 37 uC for 24, 48 and 72 h, the infectious virus yields in cell culture supernatants were determined by TCID 50 assay and the intracellular viral RNA levels quantified by RT-qPCR. To measure virus replication after infection, 5.3610 3 naïve HuH-7 cells seeded into six-well culture dishes were infected with virus at the indicated m.o.i. Following incubation at 37 uC for 24, 48, 72, 96 and 144 h, infectious virus yields and intracellular viral RNA levels were determined as described above.
SDS-PAGE and Western blotting. Cell lysates were subjected to SDS-PAGE followed by Western blotting using appropriate antibodies as described previously (Clayton et al., 2002). The bound antibodies were detected using enhanced chemiluminescence reagents (Amersham).
Immunoprecipitation of DDX3. HuH-7 cells were washed with PBS, lysed in lysis buffer (20 mM Tris/HCl pH 7.4, 20 mM iodoacetamide, 150 mM NaCl, 1 mM EDTA and 0.5 % Triton X-100) and the lysate was spun briefly to remove nuclei. The clarified cell lysates were incubated with the anti-DDX3 antibodies as described in the text and the immune complexes precipitated using protein G agarose beads (Sigma). Following washes of the Sepharose beads, the immune complexes were analysed by SDS-PAGE followed by Western blotting using biotinylated R648 and anti-streptavidin-horseradish peroxidase (HRP) conjugate.
Co-immunoprecipitation of HCV core protein. Approximately, 5610 6 HuH-7 cells electroporated with viral RNA were seeded into 100 mm tissue culture dishes. At 72 h post-transfection, the cells were washed and lysed in 0.5 ml lysis buffer (20 mM Tris/HCl, pH 7.4, 135 mM NaCl and 0.1 % Triton X-100) supplemented with 50 mM NAF, 5 mM Na 3 VO 4 and 1 mM PMSF. The lysate was spun briefly to remove nuclei. After pre-clearing, the clarified lysate was immunoprecipitated overnight with protein G agarose beads that had been pre-incubated with the anti-DDX3 antiserum R648. The beads carrying the immune complex were spun at 2000 r.p.m. (microcentrifuge) for 2 min, washed three times with the lysis buffer and subjected to non-reducing SDS-PAGE followed by Western blotting using the anti-core mAb c7-50 and anti-mouse IgG-HRP conjugate.
Indirect immunofluorescence. Cells on coverslips were fixed in methanol and probed with the indicated primary antibody for 1 h at room temperature. After washing with PBS, cells were incubated with anti-species antibodies conjugated with fluorescein isothiocyanate, tetramethyl rhodamine isothiocyanate or Cy5 (Invitrogen) for 1 h and then washed with PBS. The coverslips were examined with a Zeiss Laser Scanning LSM510 META inverted confocal microscope (Carl Zeiss) and the images analysed using LSM510 software. Threedimensional (3D) reconstructions were performed from Z-stack images collected using optimum intervals. Image stacks were deconvolved by 3D-blind deconvolution using Autodeblur software (MediaCybernetics), and 3D reconstructions were generated as described previously (Boulant et al., 2007;Targett-Adams et al., 2008).
RNA interference. Two pre-validated siRNA duplexes (s4004 and s4005, synthesized by Applied Biosystems) targeting different regions of the human DDX3 and a negative control siRNA composed of a scrambled sequence were used. Naïve HuH-7 cells plated overnight were transfected with lipofectamine RNAiMax (Invitrogen) and 50 nM siRNAs according to the manufacturer's protocol. The cells were then incubated for a further 2 days prior to virus infection. The efficiency of DDX3 knockdown was determined by immunoblotting using mAb AO196.