Functional Correlation between Subcellular Localizations of Japanese Encephalitis Virus Capsid Protein and Virus Production

This study is the first to report the comprehensive mutagenesis of a flavivirus capsid protein. We assessed the requirement of each molecular surface for infectious viral particle formation as well as for LD and nucleolar localization and found functional relationships between the subcellular localization of the virus capsid protein and infectious virus particle formation. We developed a system to independently assess the packaging of viral RNA and that of the capsid protein and found a molecular surface of the capsid protein that is crucial for packaging of viral RNA but not for packaging of the capsid protein itself. We also characterized the biochemical properties of capsid protein mutants and found that the capsid protein localizes at the nucleolus in a different manner than for its localization to the LD. Our comprehensive alanine-scanning mutagenesis study will aid in the development of antiflavivirus small molecules that can target the flavivirus capsid protein.

ABSTRACT

The flavivirus capsid protein is considered to be essential for the formation of nucleocapsid complexes with viral genomic RNA at the viral replication organelle that appears on the endoplasmic reticulum (ER), as well as for incorporation into virus particles. However, this protein is also detected at the lipid droplet (LD) and nucleolus, and physiological roles of these off-site localizations are still unclear. In this study, we made a series of alanine substitution mutants of Japanese encephalitis virus (JEV) capsid protein that cover all polar and hydrophobic amino acid residues to identify the molecular surfaces required for virus particle formation and for localization at the LD and nucleolus. Five mutants exhibited a defect in the formation of infectious particles, and two of these mutants failed to be incorporated into the subviral particles (SVP). Three mutants lost the ability to localize to the nucleolus, and only a single mutant, with mutations at α2, was unable to localize to the LD. Unlike the cytoplasmic capsid protein, the nucleolar capsid protein was resistant to detergent treatment, and the α2 mutant was hypersensitive to detergent treatment. To scrutinize the relationship between these localizations and viral particle formation, we made eight additional alanine substitution mutants and found that all the mutants that did not localize at the LD or nucleolus failed to form normal viral particles. These results support the functional correlation between LD or nucleolus localization of the flaviviral capsid protein and the formation of infectious viral particles.

IMPORTANCE This study is the first to report the comprehensive mutagenesis of a flavivirus capsid protein. We assessed the requirement of each molecular surface for infectious viral particle formation as well as for LD and nucleolar localization and found functional relationships between the subcellular localization of the virus capsid protein and infectious virus particle formation. We developed a system to independently assess the packaging of viral RNA and that of the capsid protein and found a molecular surface of the capsid protein that is crucial for packaging of viral RNA but not for packaging of the capsid protein itself. We also characterized the biochemical properties of capsid protein mutants and found that the capsid protein localizes at the nucleolus in a different manner than for its localization to the LD. Our comprehensive alanine-scanning mutagenesis study will aid in the development of antiflavivirus small molecules that can target the flavivirus capsid protein.

INTRODUCTION

Viruses of the Flaviviridae family, such as dengue virus (DENV), West Nile virus (WNV), tick-borne encephalitis virus, Zika virus (ZIKV), yellow fever virus, and Japanese encephalitis viruses (JEV), comprise a large group of emerging and reemerging pathogens causing morbidity and mortality in humans (1). However, vaccines have not yet been developed for some flaviviruses, like DENV and ZIKV, and there is a compelling need for the development of therapeutic agents to prevent clinical exacerbation (1).

The flavivirus belongs to the single-stranded RNA viruses, which have a single genome containing a long open reading frame (ORF) that encodes a single polypeptide (2). This polypeptide is posttranslationally cleaved into three structural proteins—capsid protein (C), precursor membrane (prM) and envelope (E)—and seven nonstructural (NS) proteins (2). Among them, the capsid protein forms a protective shell around the RNA genome, thus playing a central role in the assembly of virus particles (3). Furthermore, the structure of the capsid protein is well conserved in Flaviviridae, and this protein has been considered to be a good target for the development of antiviral reagents (4). The mature capsid protein is a highly basic 12-kDa protein that forms homodimers in solution (5–7). The crystal and nuclear magnetic resonance (NMR)-based structures indicate that the core region of the capsid protein dimer has two major surfaces: one surface formed by positively charged side chains on α4 and the other surface formed by hydrophobic side chains on α2 and α1 (5–7). These structural studies also suggested that the hydrophobic surface on α2 and that on α1 interact with the membranes (5, 7, 8). The positively charged α4 surface and N-terminal structurally unsolved region are proposed to be responsible for interaction with nucleic acid (3, 4, 9), although there is no strong experimental evidence. In flavivirus-infected cells, capsid protein was detected not only at the viral replication site (known as replication organelle) that appears on the endoplasmic reticulum (ER) membrane but also in the nucleolus (10–12) and lipid droplet (LD) (7, 13–15). Amino acids required for LD localization have been identified in the hydrophobic surface on α2 (13) and the N-terminal region (15, 16), based on studies on the DENV capsid protein. It has been proposed that the capsid protein localizes at the LD to utilize it for temporary storage or to modulate lipid metabolism for facilitating efficient viral growth (13), although direct evidence remains missing. For the DENV capsid protein, multiple basic amino acid clusters at the N-terminal region and C terminus of α4 have been proposed to be important for nuclear localization (17). More specifically, capsid proteins localized in the nucleus have been known to accumulate at the nucleolus. The amino acid residues required for such nucleolar localization have been mapped to G42 and P43 in the JEV capsid protein (10); another study showed that capsid protein interacts with a host nucleolar protein, B23, via these amino acids (18). However, the physiological role of nucleolar localization of the capsid protein is still unknown.

In this study, we performed exhaustive alanine-scanning mutagenesis covering all polar and hydrophobic amino acids located on the surface of JEV capsid protein dimers and determined which molecular surface is required for viral particle formation. Furthermore, we also investigated the molecular surface necessary for nucleolus and LD localization to systematically reveal the functional relationships between these localizations and viral particle formation.

RESULTS

Alanine-scanning mutagenesis for the JEV capsid protein.

To identify the molecular surface required for capsid protein function, we performed alanine-scanning mutagenesis for the JEV capsid protein. Initially, we made 23 alanine substitution mutants in a manner such that one to four polar or hydrophobic amino acids were changed to alanine in each mutant. All polar and hydrophobic residues on the surface of the dimer model (19) and in the N-terminal structurally unsolved region are covered by this scanning (Fig. 1A).

FIG 1.

Capsid protein residues required for oligomerization. (A) Structure of the JEV capsid protein dimer model deduced from dengue virus capsid protein structure (PDB accession code 1R6R) (19) and the position of the alanine substitution mutations. One monomer is colored in gray and the other is in white. The three sections depict the 90°-rotated views along the y axis. (B) Dimer formation of the capsid protein visualized with yeast two-hybrid assay. WT or mutant capsid proteins fused with AD and with DBD were coexpressed in yeast cells, and WT-mutant heterodimer or mutant-mutant homodimer formation validated with yeast two-hybrid assay. (C and D) Mutant-mutant homodimer (C) or WT-mutant heterodimer (D) formation of the capsid protein validated by pulldown assay. 293T cells were cotransfected with two plasmids encoding capsid protein with C-terminal FOS tag and Myc tag. Cell lysates were precipitated with Strep-Tactin beads and capsid protein was visualized with Western blotting using anti-FLAG or anti-Myc antibodies. (E) Oligomeric state of the capsid protein validated by gel filtration chromatography. Capsid protein purified from E. coli cells expressing N-terminally His10-tagged capsid protein (recombinant capsid), medium supernatant (virus), and crude cell lysate (infected cell lysate) of 293A cells infected with JEV and cell lysates of 293T cells expressing WT or mutant capsid proteins were fractionated via gel filtration and detected by Western blotting. (F) Oligomeric state of the purified recombinant capsid protein. Sedimentation velocity of purified N-terminally His10-tagged recombinant capsid protein (left) was quantitated through analytic ultracentrifugation (right). CBB, Coomassie brilliant blue.

First, we checked the dimer formation ability of each mutant by a series of yeast two-hybrid assays. Gal4 activation domain (AD)-fused wild-type (WT) or mutant capsid proteins and DNA binding domain (DBD)-fused WT or mutant capsid proteins were expressed in yeast cells simultaneously, and the interaction between AD- and DBD-fused capsid proteins was evaluated in terms of expression levels of genes controlled by the GAL promoter. In this study, we used ADE2 and HIS3 as reporter genes. Therefore, we were able to monitor the protein-protein interaction by measuring yeast growth in a medium depleted of histidine and adenine. We tested WT-mutant heterodimer as well as mutant-mutant homodimer formations. As a result, we could confirm normal dimer-forming function for most of our mutants except K31A/R32A, G42A/P43A/R45A, F46A/L50A/F53A/F54A, and E87A/I92A mutants (Fig. 1B). We examined the dimer-forming function of these mutants, which failed to exhibit positive signals in the yeast two-hybrid assay, directly by using a pulldown assay in mammalian cells (Fig. 1C and D). 293T cells coexpressing (i) WT or mutant capsid proteins C-terminally tagged with a FLAG-One-STrEP (FOS) epitope and (ii) WT or mutant capsid protein C-terminally tagged with a Myc epitope were lysed and purified with Strep-Tactin Sepharose beads. In all combinations tested, FOS-tagged capsid protein was copurified with Myc-tagged capsid protein, suggesting that none of the mutations affect capsid-capsid interactions (Fig. 1C and D).

Next, we performed gel filtration mobility assays using lysates of cells expressing capsid proteins (Fig. 1E). In JEV-infected cells, the majority of capsid protein was detected in fractions corresponding to about 140 kDa (Fig. 1E). This mobility is similar to that of the recombinant pure JEV capsid protein purified from bacterial cells (Fig. 1E, recombinant capsid) and the capsid protein from virus particles released into cell-culture supernatants (Fig. 1E, virus). In a manner similar to that of WT capsid protein, K31A/R32A, F46A/L50A/F53A/F54A, K85A/R86A, and E87A/I92A mutant capsid proteins were also detected in ∼140-kDa fractions, indicating that these mutants form dimers. An exception is the G42A/P43A/R45A mutant, for which capsid proteins were detected in both ∼140-kDa fractions and in the void fraction, indicating that some portion of the capsid proteins were aggregated in cells; nevertheless, 28% of capsid proteins were still detected in the dimer (∼140-kDa) fraction.

In order to characterize the oligomeric state of capsid proteins in these fractions, we performed an analytical ultracentrifugation (AUC) assay. The results revealed that the purified recombinant JEV capsid protein exists as a homodimer in solution (Fig. 1F), indicating that capsid proteins detected in the ∼140-kDa fractions of the gel filtration assay are dimers. Notably, the recombinant capsid protein was detected in not only the ∼140-kDa fractions but also in heavier fractions. Although we cannot provide a straightforward explanation for this varied mobility of the recombinant capsid protein in gel filtration, the result suggests that the mammalian cell-borne capsid proteins present in the ∼140-kDa fractions form monomeric dimers rather than higher-order oligomers.

Taken together, these results indicate that almost all mutants that we constructed form monomeric dimers in cells, as does the WT capsid protein. This was expected, considering that we chose the surface residues of the capsid protein dimer for mutations.

Identification of the molecular surface required for formation of infectious viral particles.

To determine the capsid protein region required for the formation of infectious particles, we employed single-round infectious particle (SRIP) production assays (20), using a series of alanine substitution capsid protein mutants. The capsid protein, prM-E, and a subgenomic replicon RNA were coexpressed separately from three plasmids (Fig. 2A) in 293T cells (producer cells). The subgenomic replicon RNA contains an ORF for NanoLuc reporter gene, so efficiency of SRIP formation can be quantified by inoculating Huh7 cells (target cells) with culture supernatant of the producer cells, which contains SRIPs, and measuring NanoLuc luciferase activity in the target cells after cultivation.

FIG 2.

Capsid protein residues required for formation of infectious viral particles. (A) Design of SRIP-producing plasmids used in this study. The subgenomic replicon plasmid pCMV-JErep-nluc contains an ORF encoding a chimeric polyprotein in which NanoLuc is fused to nonstructural proteins via an F2A “self-cleaving” peptide. A cytomegalovirus (CMV) promoter, a hepatitis D virus (HDV) self-cleaving ribozyme, and a simian virus 40 (SV40) polyadenylation signal are positioned such that the transcript from CMV promoter forms a subgenomic replicon in which the ORF encoding the chimeric polyprotein is flanked by 5′ and 3′ untranslated region (UTR) sequences of the viral genome, after self-cleavage of the HDV ribozyme. pCAG-JEprME encodes the prM and E structural proteins, and pCAG-JEC encodes the WT capsid protein. (B) Effect of alanine substitutions on production of the SRIPs. 293T cells were cotransfected with pCMV-JErep-nluc (or its nonreplicable variant pCMV-JErep-nluc-fs [fs]), pCAG-JEprME (or an empty vector pCAG-FOS [ΔprM]), and pCAG-JEC (or its variant with alanine substitutions or empty vector pCAG-FOS [ΔC]) and cultivated for 72 h (producer cells). Huh7 cells were inoculated with the culture supernatant of producer cells and cultivated for 72 h (target cells). Inner cellular NanoLuc luciferase activities of target cells were quantified to estimate the amount of SRIPs in the culture supernatant of the producer cells. NanoLuc luciferase activities in producer cells were also quantified to estimate efficiencies of transfection and replication of the subgenomic RNA. Means and SDs from three independent experiments are presented. Expression of the capsid protein and its variants in producer cells were checked using Western blotting. IB, immunoblotting.

The cells coexpressing WT capsid protein, prM-E, and the NanoLuc-subgenomic RNA produced a substantial amount of SRIPs in the culture supernatants (Fig. 2B, WT, bottom graph), while the cells not expressing the capsid protein or prM-E produced little or no SRIPs (Fig. 2B, ΔC and ΔprME, both graphs), indicating that this SRIP system works as expected. Furthermore, the cells transfected with plasmids that express a replication-defective NanoLuc-subgenomic RNA with a frameshift in the NS5 replicase coding region produced few or no SRIPs (Fig. 2B, fs, both graphs). This indicates that in our system, SRIP production requires not only expression of three functional structural proteins (C, prM, and E) but also replication of subgenomic RNA.

Using this system, we tested the capability of each capsid protein mutant for assisting the formation of infectious particles. As shown in Fig. 2B, we found five capsid protein mutations (K31A/R32A, G42A/P43A/R45A, M78A/K79A, K85A/R86A, and R98A/K100A/K101A) that impair SRIP production. These mutations did not significantly reduce NanoLuc activities in the producer cells, excluding the possibility that the impairment of SRIP production is caused by inefficient viral genome replication. Our results suggest that the residues altered by these mutations are responsible for the functional interaction between the capsid protein and viral RNA during packaging of RNA or other viral processes such as uncoating after internalization. The F46A/L50A/F53A/F54A mutation, which abolishes the hydrophobic residue cluster on the α2 helix, also impaired SRIP production but to a lesser extent. The K63A/L66A mutation slightly diminished SRIP production, but this effect was not significant. Residue L66 is mutated from leucine to serine in the live vaccine strain SA14-14-2. It has been reported that the L66S mutation contributes to the attenuation of this vaccine strain (21), which is consistent with our result. It is noteworthy that the majority of other capsid protein mutants that were tested retained the ability to form infectious particles, indicating that regions other than those that we discussed above do not have critical roles in infectious particle formation. We also identified mutations (K3A/K4A/P5A, P8A, R10A/R12A, P22A/R23A/V24A, and V34A/M35A) that enhanced SRIP production without enhancing replication of subgenomic RNA in producer cells (Fig. 2B). Residues in these mutations may have repressive functions in the formation of infectious particles.

Identification of the molecular surface required for incorporation of capsid protein into infectious and uninfectious particles.

We evaluated the effect of mutations in the capsid protein on the formation of infectious particles with the conventional SRIP system described above, using a NanoLuc gene on the subgenomic RNA as a reporter. Next, we evaluated the packaging efficiency of the capsid proteins into viral particles. To assess the packaging efficiency of the capsid protein, we developed a modified SRIP system designated HiBiT-SRIP, which employs a plasmid encoding HiBiT-tagged capsid protein in place of the capsid protein plasmid (Fig. 3A). HiBiT-tag, a C-terminal 11-amino-acid fragment of NanoLuc luciferase, spontaneously assembles with LgBiT, the N-terminal 15.7-kDa fragment of NanoLuc, to reconstitute an active luciferase. Using the HiBiT-SRIP system, the packaging efficiency of the HiBiT-tagged capsid protein can be estimated by measuring HiBiT-dependent luciferase activity in the culture supernatant of producer cells. In this analysis, we fractionated culture supernatant 72 h posttransfection using sucrose density gradient centrifugation to separate infectious and uninfectious particles released from cells.

FIG 3.

Capsid protein residues required for assembly of viral particles. (A) Design of HiBiT-tagged capsid containing particle-producing plasmids used in this study. (B and E to H) Sedimentation analysis of SRIPs containing HiBiT-tagged capsid protein. Culture supernatant containing HiBiT-SRIPs was loaded on a linear sucrose density gradient. After centrifugation, HiBiT luciferase activities in each fraction were quantified. (C) Huh7 cells were inoculated with culture supernatant fractions from sucrose density gradient centrifugation. Inner cellular NanoLuc luciferase activities 72 h postinoculation were quantified. (D) Sedimentation analysis of JEV live virus-infected cell culture supernatant. After centrifugation, viral genomic RNA and infectivity of each fraction were estimated. (I) Amounts of subgenomic RNA in fractions 16, 19 (SVP), and 25 (SRIP) of the experiment summarized in panels B, E, and H, respectively. (J) JEV-infected Vero cells were immunostained with anti-capsid protein and anti-envelope antibodies. (K) Degree of colocalization of capsid protein and envelope quantified from images of three immunostained HeLa cell cultures.

The culture supernatant of positive-control cells expressing WT capsid-HiBiT, prM-E, and the NanoLuc-subgenomic RNA formed three major capsid-HiBiT peaks after centrifugation (Fig. 3B, black circles). The first, second, and third peaks appeared in fractions 25 and 26, 19 to 21, and 1 to 7, respectively. The first peak was not detected in culture supernatant from cells not expressing the subgenomic RNA, suggesting that this peak corresponds to SRIPs (Fig. 3B, open triangles). To confirm that the first peak corresponds to SRIPs, we checked the infectivity of each of fractions corresponding to the three peaks. As expected, only the first peak showed infectivity as demonstrated by NanoLuc luciferase activity in infected (target) cells (Fig. 3C, closed circles). Appearance of the second peak was dependent on expression of prM-E (Fig. 3B, open squares) and independent of expression of the subgenomic RNA (Fig. 3B, open triangles), suggesting that this peak corresponds to a type of subviral particle (SVP) consisting of the capsid protein, M, and E. We also observed a minor peak in fractions 11 to 14, whose appearance was dependent on expression of the capsid protein and independent of expression of prM-E or the subgenomic RNA (Fig. 3B); this might correspond to some type of uninfectious nucleocapsid particle consisting of the capsid protein and subgenomic RNA or nonspecific host nucleic acid.

To characterize particles in the capsid protein-positive peaks, we measured the amount of JEV live virus-borne genomic RNA in fractions after sucrose density gradient centrifugation. The distribution of RNA separated into two major peaks, infectious and uninfectious (Fig. 3D). The live viral infectious peak formed in fractions 24 to 27, which overlapped with fractions of the HiBiT-SRIP peaks (fractions 25 and 26 in Fig. 3B and C), suggesting similarity in size between HiBiT-SRIP and live viral particles. The live viral uninfectious peak formed in fractions 2 to 5 and overlapped with the HiBiT-SRIP system-borne, capsid protein-positive uninfectious peak (fractions 1 to 7 in Fig. 3B).

Next, we applied the HiBiT-SRIP system to scrutinize a subset of capsid protein mutations which severely impaired infectious particle formation in a conventional SRIP assay whose results are presented in Fig. 2. As a result, mutations K31A/R32A and K85A/R86A impaired formation of HiBiT-SRIPs without ruining that of SVP, indicating that two basic surfaces consisting of K31/R32 and of K85/R86 are important for packaging of the viral RNA, but not the capsid protein itself, into the viral particles (Fig. 3E, H, and I). On the other hand, mutations G42A/P43A/R45A and F46A/L50A/F53A/F54A impaired formation of both SRIP and SVP (Fig. 3F and G). This indicates that two hydrophobic molecular surfaces consisting of G42/P43/R45 and of F46/L50/F53/F54 are important for incorporation of the capsid protein itself into prM-E particles, which is a prerequisite for incorporation of the viral RNA into the particles. While the F46A/L50A/F53A/F54A mutant with a HiBiT tag formed a significantly smaller peak in SRIP fractions (Fig. 3G), the mutant without the HiBiT tag produced only a moderately smaller amount of SRIPs (Fig. 2B). Therefore, we checked the infectivity of fractions after sucrose density gradient centrifugation of HiBiT-SRIPs along with this mutant. The mutation significantly decreased HiBiT-SRIP infectivity compared to that of the WT (Fig. 3C, open circles), confirming the discrepancy between SRIP and HiBiT-SRIP assays. The insertion of the HiBiT tag might have had a synergistic effect on the F46A/L50A/F53A/F54A mutant, resulting in the significant phenotype in the HiBiT-SRIP assay. We also found a small shift in the SRIP peak of this mutant compared to that of the WT (Fig. 3C), indicating that the mutation affects infectious particle formation quantitatively as well as qualitatively.

To characterize the HiBiT-positive peaks that appeared in the SVP and SRIP fractions, we used quantitative reverse transcription-PCR (qRT-PCR) to directly measure the amount of the reporter RNA in these peaks, formed in experiments whose results are summarized in Fig. 3C (WT), Fig. 3E (K31A/R32A), and Fig. 3H (K85A/R86A) (Fig. 3I). Only the SRIP fraction (fraction 25) from the WT experiment contained a substantial amount of the reporter RNA. The SVP fraction (fraction 19) and control fraction with little HiBiT activity (fraction 16) contained little RNA, confirming that the uninfectious particles in the SVP fraction did not contain subgenomic RNA. These findings also indicated that K31A/R32A and K85A/R86A are defective in the packaging of viral RNA rather than other viral processes.

To assess the capsid protein incorporation into viral particles, we surveyed cells by superresolution microscopy and quantified colocalization of the capsid protein and E proteins. In JEV-infected cells, E-positive dots with a size of 50 to 100 nm, which is similar to the size of the JEV particle, as well as capsid protein-positive smaller dots were observed in the cytoplasm (Fig. 3J). A substantial fraction of E-positive and capsid protein-positive dots overlapped with each other, suggesting that these overlapping dots correspond to the viral particles incorporating the capsid protein (Fig. 3J). Similar dots were also observed in cells in which capsid protein and prM-E were coexpressed from plasmids, indicating that incorporation of the capsid protein into viral particles consisting of prM-E does not require other viral factors. We examined the effect of mutations in the capsid protein on colocalization efficiency with E in cells, using the plasmid-based expression system. As a result, the G42A/P43A/R45A and F46A/L50A/F53A/F54A mutations impaired colocalization with E, while the K31A/R32A and K85A/R86A mutations did not (Fig. 3K). This result is well consistent with our results using HiBiT-SRIP, again supporting the notion that hydrophobic molecular surfaces consisting of G42/P43/R45 or F46/L50/F53/F54, but not those consisting of K31/R32 or K85/R86, are required for incorporation of the capsid protein into prM-E particles.

Identification of the molecular surface required for nucleolar localization.

As several other groups reported previously, flavivirus capsid proteins are detectable in the nucleolus of infected cells (10–12, 22), though the physiological role of this localization is still unclear. To identify the molecular surface required for nucleolar localization, we performed immunocytochemistry analyses to test the subcellular localization of each alanine substitution capsid protein mutant in HeLa cells. The WT capsid protein expressed from a plasmid was detected in both the cytoplasm and nucleolus in a manner quite similar to that of capsid proteins in JEV-infected cells (Fig. 4A and B), suggesting that nucleolar localization of the capsid protein does not require other viral factors. In this screening, we found three mutations (L38A/G40A/R41A, G42A/P43A/R45A, and F46A/L50A/F53A/F54A) which decrease nucleolar localization (Fig. 4B and C). The G42A/P43A/R45A mutant exhibited the most pronounced phenotype, which localized almost exclusively at the cytoplasm, which is consistent with a report proposing that residues G42 and P43 of the JEV capsid proteins are responsible for nucleolar localization (10). We also found that L38A/G40A/R41A, a mutation in a site adjacent to these residues, has a similar effect on nucleolar localization, albeit to a lesser extent (Fig. 4C). The F46A/L50A/F53A/F54A mutant exhibited an aberrant localization different from that of the other two mutants; this mutant formed small dots at the nucleus without overlapping with the RPL11 nucleolar marker (Fig. 4B and C). These results support the idea that the molecular surface consisting of L38, G40, R41, G42, P43, R45, F46, L50, F53, and F54 has an important role in localization at the nucleolus. Despite an exhaustive search, the other mutants that we examined were found to retain normal nucleolar and cytoplasmic localization, suggesting that this molecular surface is exclusively responsible for nucleolar localization of the capsid protein. To characterize biochemical properties of capsid protein localization at the nucleolus, we investigated the persistence of capsid proteins in their original sites, under various conditions. JEV-infected cells were sequentially treated with buffers containing detergent, RNase A, high salt, and DNase I, in this order (Fig. 5A), and then the capsid proteins in the soluble fractions were detected with Western blotting (Fig. 5B); 4% and 40% of capsid proteins were retained after detergent and high-salt treatments, respectively, as evidenced from their band intensities. The capsid proteins that remained in cells after each steps of treatments were also detected by immunostaining (Fig. 5C). These behaviors were similar to that of two typical nucleolar proteins, nucleolin and B23 (Fig. 5B). While a significant fraction of nucleolar capsid proteins remained after treatment with detergent, cytoplasmic capsid protein almost completely disappeared (Fig. 5C), suggesting that capsid protein localization at the cytoplasm, but not the nucleolus, is dependent on membrane association. Nucleolar capsid protein disappeared after treatment with RNase or high salt (Fig. 5C), suggesting that capsid protein localization at the nucleolus is dependent on RNA association. These behaviors of the capsid protein in JEV-infected cells were also observed when WT capsid protein alone was expressed from a plasmid (Fig. 5D and F), indicating that other viral factors are not required for such behaviors.

FIG 4.

Capsid protein residues required for nucleolar localization. (A and B) HeLa cells were inoculated with JEV at an MOI of 1 and immediately transfected with pQC-RPL11-mCherry (A) or cotransfected with pCAG-JEC or its variants and pQC-RPL11-mCherry (B) and cultured for 24 h. Microscopic images of fixed cells stained with anti-capsid protein antibody and Hoechst 33342 are presented. (C) Percentage of cells in which nucleolar localization of the capsid proteins was observed. Means and SDs from three independent experiments are shown.

FIG 5.

Characterization of nucleolar localization of the capsid protein. (A) Schematic representation of the sequential treatments to assess retention of the capsid protein in the nucleolus. (B) JEV-infected HeLa cells were cotransfected with plasmids pCAG-OSF-B23 and pCAG-OSF-Nucleolin and treated as for panel A, and the capsid proteins in fractions after each treatment was examined by Western blotting. (C) Cells after each treatment step, immunostained with anti-capsid protein antibody. (D) HeLa cells were cotransfected with pCAG-JEC or its variants, treated as for panel A, and the capsid proteins in fractions after each treatment was examined. (E) The amount of capsid protein after each treatment was quantified by Western blotting, Means and SDs from three independent experiments are shown. (F) HeLa cells after each treatment step, immunostained with anti-capsid protein antibody.

To investigate the biochemical properties of the aberrant dots that the F46A/L50A/F53A/F54A mutant formed in the nucleus (Fig. 4B), we examined retention of this mutant in the nucleus after each treatment step, using cells expressing this mutant from a plasmid. A majority of the nuclear dots formed by the F46A/L50A/F53A/F54A mutant disappeared after treatment with detergent (Fig. 5F), suggesting that this mutant resides in the nucleus in a manner biochemically different from that of WT capsid protein. We also tested the K85A/R86A mutant and found that localization of this mutant at the nucleolus was sensitive to high-salt treatment but remained resistant to detergent treatment, suggesting that mutation at this site attenuates the association to RNA (Fig. 5F).

Identification of the molecular surface required for lipid droplet localization.

It has been reported that in DENV-infected cells, the capsid protein is detectable not only at sites of replication but also at sites around the LDs in the cytoplasm (13). We examined whether JEV capsid protein also localizes at the LD, using BODIPY, a fluorescent dye which stains neutral lipids. As a result, the capsid protein was detected around the BODIPY-positive compartment, indicating that JEV capsid protein also accumulates around the LD in cells infected with JEV (Fig. 6A), in a way similar to that of DENV capsid proteins in DENV-infected cells. This LD localization was also recapitulated in cells expressing the capsid protein from a plasmid, suggesting that no other viral components are required (Fig. 6B).

FIG 6.

Capsid protein residues required for localization to the lipid droplet. (A) JEV-infected Huh7 cells immunostained with anti-capsid protein antibody. (B) Huh7 cells transfected with pCAG-JEC or its variant and immunostained with anti-capsid protein antibody. (C) Percentage of cells in which capsid protein was localized to the lipid droplet, quantified from capsid protein-immunostained cell images. Means and SDs from three independent experiments are shown.

Next, we tried to identify the molecular surface of JEV capsid proteins responsible for LD localization, using the whole set of alanine substitution mutants. As a result, the F46A/L50A/F53A/F54A mutant alone failed to localize at the LDs, suggesting that a hydrophobic cluster consisting of F46/L50/F53/F54 on the molecular surface of the α2 helix is exclusively responsible for LD localization of the JEV capsid protein (Fig. 6B and C). This is consistent with a past study showing that a similar hydrophobic cluster in the α2 helix of DENV is important for LD localization (13).

Dissection of the G42A/P43A/R45A and F46A/L50A/F53A/F54A mutants.

To further investigate the functional relationships between particle formation and subcellular localization, we delved into two mutants with the most pronounced phenotypes in terms of nucleolus and LD localization, the G42A/P43A/R45A and F46A/L50A/F53A/F54A mutants, respectively. We generated single and double amino acid mutants to elucidate which amino acids are actually responsible for localization at the nucleolus and LDs (Fig. 7).

FIG 7.

Surfaces on the capsid protein responsible for infectious particle formation and subcellular localizations. (A) Percentage of capsid protein-mutant expressing cells which exhibited nucleolar localization of the capsid protein, quantified from capsid protein-immunostained cell images. (B) Percentage of capsid protein-mutant expressing cells which exhibited LD localization of the capsid protein, quantified from capsid protein-immunostained cell images. (C) Effect of alanine substitutions on production of SRIPs. 293T cells were transfected with pCMV-JErep-nluc (or its nonreplicable variant pCMV-JErep-nluc-fs [fs]), pCAG-JEprME, and pCAG-JEC (or its variant with alanine substitutions or an empty vector pCAG-FOS [ΔC]) and cultivated for 72 h (producer cells). Huh7 cells were inoculated with the culture supernatant of producer cells and cultivated for 72 h. Inner cellular NanoLuc luciferase activities of producer and target cells were quantified. Means and SDs from three independent experiments are presented. Expression of the capsid protein and its variants in producer cells was checked with Western blotting. (D) G42, P43, R45, F46, L50, F53, and F54 residues mapped on a structural model of the JEV capsid protein (19).

G42A/P43A/R45A, a mutation in the loop between α1 and α2 helices, abolished localization of the capsid protein at the nucleolus (Fig. 4B and C). Double mutations G42A/P43A and G42A/R45A, but not P43A/R45A, impaired localization of the capsid protein at the nucleolus, while single-residue substitutions G42A, P43A, and R45A did not, suggesting that G42, P43, and R45 contribute redundantly to localization at the nucleolus (Fig. 7A). The combination of mutations which impaired localization suggests that the G42 residue has a larger contribution among the three residues for nucleolus localization. On the other hand, all mutations that abolished nucleolus localization (G42A/P43A/R45A, G42A/P43A, and G42A/R45A) concomitantly impaired SRIP production (Fig. 7C). Thus, we could not identify any mutations that impair nucleolus localization without disrupting infectious particle formation.

F46A/L50A/F53A/F54A, a mutation which abolishes the hydrophobic surface on the α2 helix, decreased packaging efficiency of the capsid protein (Fig. 3G) and formation of infectious particles (Fig. 2B), although this mutation severely impaired localization at the LD (Fig. 6B and C). To scrutinize the contribution of residues altered in this mutation for localization at the LD, we constructed double mutants F46A/L50A and F53A/F54A. These double mutants exhibited clear phonotypes in terms of infectious particle formation (Fig. 7C) and localization efficiency at the LD (Fig. 7B), indicating that there is a functional correlation between LD localization and viral particle formation capability.

DISCUSSION

In flavivirus-infected cells, the capsid protein localizes not only at viral replication organelles but also at the LD and nucleolus, which are not related to viral particle formation, although the physiological roles of these localizations are still unclear (3). In this study, we conducted functional mapping of the molecular surface of JEV capsid proteins and alanine scanning to reveal the relationship between these localizations and viral particle formation. All of the alanine mutants analyzed in this study formed soluble homodimers in the cell (Fig. 1), indicating that the majority of phenotypes seen in this study are not a result of misfolding or conformational abnormality that could occur due to the introduction of mutations.

A molecular surface consisting of two amino acids, G42 and P43, on the loop between the α1 and α2 helices has been considered to be necessary for the viral life cycle (10). In this study, we revealed that in addition to these two amino acids, R45 is essential for incorporation of the capsid proteins into the viral particles. Mutations in two of these three important amino acids were required to impair SRIP formation (Fig. 7C), suggesting that these amino acids cooperate to form the functional surface required for forming viral particles. Several reports have revealed that mutations in this site in flavivirus capsid protein increases SVP production (10, 23, 24), which might be explained by impairment of capsid protein packaging described in this study. It is worth stressing that some portion of the G42A/P43A/R45A mutant was detected in heavier fractions in size exclusion chromatography analyses (Fig. 1E). Taking into account a report showing that flavivirus capsid-dimer forms higher-order oligomers in solution (6, 7), these residues might be responsible for making such higher-order complexes.

Another mutant, the F46A/L50A/F53A/F54A mutant, also showed a defect in capsid protein incorporation into the viral particle (Fig. 3G). Substituted amino acids in this mutant are located in the α2 helix; this helix is composed of a hydrophobic surface, which is important for association of the capsid protein with the ER membrane and with the LD in DENV (5, 8, 13, 25). The F46A/L50A/F53A/F54A mutant seemingly failed to localize at the LD (Fig. 6), suggesting that the loss of membrane binding function of the α2 helix causes impairment of incorporation of this mutant into the viral particle (Fig. 3G). A recent report based on the structure of immature Zika virus particles proposed that direct interaction with the transmembrane domains of M and E is important for capsid protein incorporation into the viral particles (26). However, other reports based on structures of DENV particles proposed that membrane binding, but not direct interaction with M or E, is crucial for packaging of the capsid protein (27, 28). Our results using the F46A/L50A/F53A/F54A mutant revealed the importance of association with membrane, rather than with M or E, for the capsid protein to be incorporated into viral particles, thus supporting the last reports.

The K31A/R32A mutant exhibited defects in the formation of infectious particles (Fig. 2B), retaining normal levels of incorporation into uninfectious particles (Fig. 3E). The inner cellular level of the K31A/R32A mutant was low compared to that of the WT (Fig. 2B), suggesting that this mutation impairs the stability of the capsid protein and reduces its amount to a level sufficient for incorporation into SVPs but not for formation of infectious particles. Furthermore, it has been shown that a DENV capsid protein with a similar mutation, K31A/R32A, is unstable in cells (25). On the other hand, a structural study of Zika virus capsid protein demonstrated the functional importance of pre-α1 helix, where K31 and R32 reside, for the LD/membrane targeting of the capsid protein (7). However, the JEV capsid protein K31A/R32A mutant did not show any abnormalities in localization at the LD in our study (Fig. 6C), which is consistent with previous studies showing that a similar mutation in pre-α1 helix of DENV capsid protein does not affect its localization at the LD (25). Capsid protein structure around the K31 and R32 residues might have diversity within the flavivirus family (5–7); regarding these residues, JEV and DENV might share similar structures, while Zika virus does not.

Several residues make up the basic surface on the α4 helix of the flavivirus capsid protein and are considered to be responsible for direct interaction with viral genomic RNA (29–31). M78A/K79A, K85A/R86A, and R98A/K100A/K101A mutations in the α4 helix surface impaired formation of infectious particles (Fig. 2). In the case of K85A/R86A, incorporation into uninfectious particles was not disrupted (Fig. 3H). This result, together with the similar result obtained for the K31A/R32A mutant, indicates that the ability of genome association can be separated from the ability of viral-particle formation of capsid protein in JEV, which seemingly contradicts a previous study proposing that packaging and replication of viral RNA are functionally coupled in WNV (32).

During analysis of HiBiT-SRIP system-borne particles, we found uninfectious, capsid protein-containing SVPs (Fig. 3B). These particles did not contain subgenomic RNA (Fig. 3I), indicating that these particles could be composed of capsid, membrane, and envelope proteins and lipid membrane. Alternatively, it is also possible that these particles contain nonspecific host nucleic acid instead of subgenomic RNA.

It has been reported that substantial amounts of capsid proteins are detected in the nucleolus in flavivirus-infected cells, although the physiological meaning of this nucleolar localization is still unclear (3). In our study, JEV capsid protein remained at the nucleolus in the presence of detergent but not in the presence of high salt or RNase (Fig. 5C), suggesting that nucleolar localization of the capsid is driven by strong interaction with RNA. We identified three mutations, L38A/G40A/R41A, G42A/P43A/R45A, and F46A/L50A/F53A/F54A, that impair nucleolar localization (Fig. 4C). These mutations were located around the loop between the α1 and α2 helices. Among them, the G42A/P43A/R45A mutation induced the most pronounced phenotype in terms of nucleolar localization, supporting a previous report indicating that G42 and P43 are required for nucleolar localization of JEV capsid proteins (10). From our detailed analyses, we identified an additional residue, R45, which also contributes to nucleolar localization (Fig. 7A). Considering that G42 and P43 have been shown as residues responsible for interaction with host factor B23 (18), R45 may also have a role in the interaction with some nucleolar factors. The F46A/L50A/F53A/F54A mutant impaired localization at the nucleolus in a manner different from that of the G42A/P43A/R45A mutant (Fig. 4B and C). The capsid protein with this mutation was detected in the nucleus, forming aberrant clumps that did not overlap nucleolar makers (Fig. 4B). This aggregate formation may hide the molecular surface required for capsid protein localization at the nucleolus. All alanine substitution mutants that lost nucleolar localization concomitantly lost the function to form reporter-containing SRIPs (Fig. 4 and 7A and C), suggesting that there is a functional correlation between nucleolar localization of the capsid protein and efficiency of infectious particle formation. Functional modulation of the nucleolus by association with nucleolar host factors such as DDX56 (11, 33), Jab1 (34), hnRNP A2 (35), and B23 (18), which have been reported to interact with the capsid protein, might be crucial for efficient packaging of the genomic RNA. However, we could not exclude the possibility that the capsid protein passes the nuclear pore by passive diffusion and passively accumulates at the nucleolus, an RNA-rich compartment, via its genomic RNA binding site.

Although apparent localization of the K85A/R86A mutant was normal, it was slightly more sensitive to RNase in comparison to the WT (Fig. 5D and E). It has been proposed that α4, the helix on which K85 and R86 reside, directly makes contact with RNA, cooperatively with the N-terminal unstructured region (29–31); this is in agreement with our result. The K85A/R86A mutant failed to packaging of subgenomic RNA (Fig. 3H and I) and to form infectious particles (Fig. 2B and 3H), while it retained the ability of incorporation into SVPs (Fig. 3H), indicating that K85 and R86 are crucial for packaging of RNA, but not the protein itself, into the viral particle.

In this study, we found that the WT capsid protein in the cytoplasm was washed out by treatment with detergent (Fig. 5C), indicating that cytoplasmic capsid proteins are essentially membrane bound. The cytoplasmic capsid protein is known to accumulate around LDs in cells, and this LD localization has been considered to depend on membrane binding abilities of the capsid protein (13). In our study, despite an exhaustive search, we could find only a single mutation (F46A/L50A/F53A/F54A) that impairs LD localization (Fig. 6), suggesting that hydrophobic residues at the α2 helix are exclusively responsible for membrane binding of the JEV capsid protein. This is consistent with previous findings that the α2 helix is important for LD localization of the capsid protein in DENV (13, 25); additionally, in the case of DENV capsid protein, N-terminal region residues are also crucial (15, 16). Despite impairment of LD localization, the F46A/L50A/F53A/F54A mutation in the α2 helix of the JEV capsid protein had only a moderate effect on formation of the infectious particles (Fig. 2B and 7C). All mutations in the hydrophobic cluster of the α2 helix have been reported to impair LD localization of the capsid protein in other flaviviruses and simultaneously impair viral propagation (13, 23, 25, 36). This might be attributed to functional differences between the α2 helix of the capsid protein of JEV and those of other flaviviruses. Alternatively, it could be related to technical differences between our study using SRIP systems to focus on single-round particle formation, which scores small effects as small, and other studies using full-genome viruses to evaluate entire virus proliferation process, in which small effects could be amplified after several rounds of propagation. Further research is required to estimate the functional importance of membrane binding, via hydrophobic residues in α2 of the capsid protein, in JEV particle formation.

In this study, we constructed a set of mutants in which each polar and hydrophobic residue on the surface of the JEV capsid protein is changed to an alanine residue and examined the mutants using biochemical and cell biological approaches. We could map the essential and nonessential residues for packaging of the capsid protein or the viral RNA into a viral particle or uninfectious SVP, although there is room for further study to fine-tune this mapping by checking the effect of replacement of residues tested in our investigation with amino acids other than alanine. We also mapped the surface residues important for all flaviviruses, for JEV only, and for other flaviviruses apart from JEV. This functional mapping could be useful especially for the development of antiflavivirus drugs, whereby with the help of other techniques, such as in silico docking simulations, small-molecule compounds targeting the flavivirus capsid protein could be identified.

MATERIALS AND METHODS

Cells, viruses, and transfection.

293A, 293T, HeLa, Huh7, and Vero cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing 100 U/ml of penicillin, 100 μg/ml of streptomycin, and 10% (vol/vol) fetal bovine serum (FBS) in humidified air containing 5% CO₂ at 37°C. JEV strain AT31 (kindly provided by Eiji Konishi, Osaka University) was grown in 293A cells. 293T, HeLa, and Huh7 cells were transfected with plasmids using polyethylenimine (PEI; 25,000 Da; Polysciences, Warrington, PA) or Lipofectamine 3000 (Thermo Fisher Scientific, Waltham, MA) following the manufacturer’s protocol.

Expression plasmids.

A cloning vector, pCAG-MCS2-FLAG-HiBiT, was constructed by inserting a sequence encoding FLAG and HiBiT tag peptides to the cloning site of pCAG-MCS2 (37). The gene for the mature JEV capsid protein (amino acid [aa] residues 2 to 105) was amplified from a JEV Mie/41/2002 strain cDNA clone kindly gifted from Shigeru Tajima (38) and cloned into mammalian expression vectors pCAG-MCS2-FOS, pCAG-MCS2-Myc (39), pCAG-MCS2-FLAG-HiBiT, pGADT7, and pGBKT7 (Clontech, Mountain View, CA). The resulting plasmids were designated pCAG-JEC-FOS, pCAG-JEC-Myc, pCAG-JEC-FLAG-HiBiT, pGADT7-JEC, and pGBKT7-JEC, respectively. The plasmid encoding JEV capsid protein without any tag, designated pCAG-JEC, was prepared by inserting a TAA stop codon into pCAG-JEC-FOS. Variants of these plasmids encoding alanine substitution mutant forms of JEV capsid proteins in place of the wild-type (WT) capsid protein were constructed using overlap extension PCR. pCMV-JErep-nluc is a plasmid that encodes a JEV subgenomic replicon with a NanoLuc reporter gene, and pCMV-JErep-nluc-fs is its variant that encodes a defective replicon with a loss-of-function mutation in NS5, as described previously (20, 40). pCAG-JEprME is a previously described plasmid encoding JEV prME (40). pQC-RPL11-mCherry, a plasmid encoding a human ribosomal protein L11 fused with an mCherry fluorescent protein, was constructed by subcloning an RPL11 gene amplified from a 293T cell-derived cDNA into a mammalian expression vector, pQCXmCherryIN (39). Human nucleolin and B23 genes were amplified from cDNA prepared from 293T cells and cloned into pCAG-OSF (41) to construct pCAG-OSF-Nucleolin and pCAG-OSF-B23, respectively.

Preparation and titration of SRIPs.

Preparation and titration of single-round infectious particles (SRIPs) were performed as described previously (20). In brief, 293T cells were cultured in 6-well plates and cotransfected with pCMV-JErep-nluc or pCMV-JErep-nluc-fs, pCAG-JEC or its variant, and pCAG-JEprME using PEI. For HiBiT-SRIP preparation, we used pCAG-JEC-FLAG-HiBiT or its variant in place of pCAG-JEC or its variant. The culture medium was replaced twice with fresh medium, at 6 and 48 h posttransfection. The medium was harvested 72 h posttransfection and used as SRIP supernatant. To estimate replication efficiency of the plasmid-borne replicon after 72 h posttransfection, cells were lysed with lysis buffer (150 mM NaCl, 20 mM Tris-HCl [pH 7.5], and 1% [vol/vol] Triton X-100). NanoLuc luciferase activity in the lysate was measured using the Nano-Glo luciferase assay system (Promega, Madison, WI) with a Varioskan LUX (Thermo Fisher Scientific). To assess the infectious titer of the SRIPs, Huh7 cells were grown in a 96-well plate and inoculated with SRIP supernatant, and NanoLuc luciferase activity in the cell lysate was measured 72 h later. The NanoLuc luciferase activity was normalized with protein concentration of the lysate, which was measured using the Bio-Rad protein assay (Bio-Rad, Hercules, CA).

Characterization of particles produced from HiBIT-SRIPs and JEV live virus.

The HiBiT-SRIP-containing supernatant was harvested as described above and treated with 100 U of DNase I to remove plasmid DNA. The JEV live viral particle-containing supernatant was harvested 48 h after inoculating JEV on 293A cells at a multiplicity of infection (MOI) of 0.3. The supernatant was loaded onto a 10% to 45% (wt/vol) linear sucrose density gradient in phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 1.5 mM KH₂PO₄, and 8 mM Na₂HPO₄ [pH 7.4]). Centrifugation was carried out for 3 h at 220,000 × g and 4°C. The gradient was fractionated and HiBiT-tagged capsid protein in each fraction was detected using the Nano Glo HiBiT lytic detection system (Promega). When estimation of the infectious titer of the HiBiT-SRIPs was required, the fractions were inoculated on Huh7 cells, and after 72 h of cultivation, NanoLuc luciferase activity in the cell lysate, which positively correlates with the infectious titer of the HiBiT-SRIPs, was measured. The amounts of viral RNA in fractions were determined with qRT-PCR. The infectious titers of live virus in fractions were determined with a focus-forming assay.

Focus-forming assay.

Vero cell monolayers prepared on 96-well plates were inoculated with virus-containing samples serially diluted with DMEM. After 2 h of incubation, medium in each well was replaced with DMEM supplemented with 1% methylcellulose. Following incubation for 34 h, the cells on plates were treated with 4% paraformaldehyde in PBS for fixation and then treated with 0.1% Triton X-100 and 10% FBS in PBS for permeabilization and blocking. After that the cells were incubated with rabbit polyclonal anti-JEV NS3 antibody (raised against recombinant JEV C-terminal region of NS3, aa 1652 to 2093) in PBS for 30 min at room temperature and then incubated with Alexa Fluor 488-conjugated anti-rabbit antibodies (Jackson ImmunoResearch, West Grove, PA) in PBS for 30 min at room temperature. The virus-positive foci were counted using a fluorescence microscope, and the viral infectious titers were calculated as focus-forming units per milliliter.

qRT-PCR.

The amounts of live viral RNA or HiBiT-SRIPs RNA in fractions after sucrose density gradient centrifugation were determined using quantitative reverse transcription-PCR (qRT-PCR). The fractions were diluted 10-fold with PBS and then treated with 50 μg/ml of proteinase K and 0.1% (wt/vol) SDS for 1 h at 37°C. The fractions were subjected to phenol-chloroform extraction followed by ethanol precipitation and treatment with DNase I. They were subjected to another round of phenol-chloroform extraction and ethanol precipitation and used as the templates for reverse transcription reactions. Reverse transcription reactions were carried out using ReverTra Ace (TOYOBO, Osaka, Japan) with a positive-strand-specific primer for the JEV NS5 gene (42). A control reaction without reverse transcriptase was also carried out. The resulting cDNA was analyzed with a quantitative PCR using Brilliant III Ultra-Fast SYBR green QPCR master mix (Agilent Technologies, Santa Clara, CA) and a DNA Engine Opticon 2 system (Bio-Rad) with a pair of primers specific to the JEV NS5 gene (43). Standard curves were prepared using serially diluted pCMV-JErep-nluc plasmid DNA solution with a defined concentration as the template for the quantitative PCR.

Indirect immunostaining and fluorescence microscopy.

HeLa or Huh7 cells cultured on coverslips were fixed with 4% paraformaldehyde in PBS for 15 min at 25°C. To visualize lipid droplets, Huh7 cells were incubated with 1 μM BODIPY (Thermo Fisher Scientific) for 15 min before fixation. The cells were permeabilized and blocked with 0.1% Triton X-100 and 10% FBS in PBS for 10 min at room temperature. The cells were then incubated with rabbit polyclonal anti-JEV capsid protein antibody (raised against recombinant JEV capsid protein, aa 1 to 100) in PBS for 90 min at room temperature. The cells were incubated with Alexa Fluor 350-, 488- or 594-conjugated anti-rabbit secondary antibodies (Life Technologies, Carlsbad, CA) in PBS for 60 min at room temperature. Nuclear DNA was stained with Hoechst 33342. The coverslips were mounted on slides with Fluoromount-G (SouthernBiotech. Birmingham, AL) and examined with a FluoView FV3000 confocal laser scanning microscope (CLSM) (Olympus, Tokyo, Japan) using an UPLAPO 60×/1.35 numerical-aperture (NA) oil immersion objective (Olympus).

Colocalization analysis of capsid protein and E.

HeLa cells transiently coexpressing capsid protein and E were fixed and coimmunostained with rabbit polyclonal anti-JEV capsid protein antibody and mouse monoclonal anti-JEV envelope antibody (Mab8743; Merck Millipore, Burlington, VT). After treatment with fluorescent-dye-conjugated secondary antibodies, samples were examined by CLSM. Three optical sections in Z direction per 0.1 μm were acquired separately for each channel. The raw Z-stacks were deconvolved and Pearson’s correlation coefficient was obtained using CellSens Dimension software (Olympus).

Stimulated emission depletion (STED) microscopy analysis.

Vero cells cultured on coverslips were infected with JEV at an MOI of 1. Forty-eight hours postinfection, cells were fixed, permeabilized, treated with anti-capsid protein and anti-E antibodies as described above, and stained with Alexa Fluor 488-conjugated anti-mouse and Alexa Fluor 555-conjugated anti-rabbit (Life Technologies) secondary antibodies. The coverslips were mounted on slides with ProLong Diamond antifade (Thermo Fisher Scientific). Images were acquired with a Leica TCS SP8 STED 3× inverted microscope (Leica Microsystems, Wetzlar, Germany) using an HCX PL APO 100×/1.4 NA oil immersion objective (Leica Microsystems). Deconvolution of stimulated emission depletion (STED) images was applied using Huygens software (Scientific Volume Imaging, Hilversum, The Netherlands).

Strep-Tactin pulldown assay and Western blotting.

The Strep-Tactin pulldown assay was performed as described previously (4). In brief, 293T cells were cultured in a 10-cm dish and cotransfected with pCAG-JEC-FOS and pCAG-JEC-Myc using PEI. The cells were harvested 48 h posttransfection and lysed with lysis buffer supplemented with complete protease inhibitor cocktail (Roche Ltd., Basel, Switzerland). Whole-cell lysates (WCL) were clarified by centrifugation (20,000 × g, 10 min, 4°C) and incubated with Strep-Tactin Sepharose beads (IBA GmbH, Göttingen, Germany) for 1 h at 4°C. The beads were washed four times with lysis buffer and suspended in sodium dodecyl sulfate (SDS)-PAGE sample buffer. The WCL and bead-bound fractions were subjected to SDS-PAGE using a 13% WIDE RANGE polyacrylamide gel (Nacalai Tesque, Kyoto, Japan). Proteins were electrically transferred to polyvinylidene difluoride (PVDF) membranes (Immobilon-P; Merck Millipore) and visualized with Western blotting using rabbit polyclonal anti-JEV capsid protein antibody, rabbit polyclonal anti-JEV NS3 antibody, mouse monoclonal anti-FLAG M2 antibody (F1804; Sigma-Aldrich, St. Louis, MO), mouse anti-Myc antibody (9E10; DSHB, University of Iowa), and rabbit monoclonal anti-His tag antibody (Roche Ltd.) as primary antibodies and horseradish peroxidase (HRP)-conjugated anti-mouse IgG (Jackson ImmunoResearch) or rabbit IgG (Jackson ImmunoResearch) as secondary antibodies. Immunoreactive bands were detected using EzWestLumi plus system (ATTO Technology, Amherst, MA) and a chemiluminescence detector (Molecular Imager ChemiDoc XRS Plus; Bio-Rad).

Yeast two-hybrid assay.

Yeast two-hybrid assays were performed using the Matchmaker yeast two-hybrid system (Clontech) as described previously (37). Briefly, Saccharomyces cerevisiae strain AH-109 was cotransformed with pGADT7-JEC and pGBKT7-JEC plasmids encoding the wild type (WT) or alanine substitution mutants of capsid protein using lithium acetate/polyethylene glycol (PEG) method and cultured on a control plate (SD-2 broth: synthetic media without leucine and tryptophan) for 3 days at 30°C. The transformants were then inoculated on a selection plate (SD-4 broth: synthetic media without leucine, tryptophan, histidine, or adenine) and further incubated for 3 days at 30°C.

Gel filtration analyses of JEV capsid protein.

Gel filtration analyses for crude cell lysates were performed as described previously (41). 293A cells infected with JEV or 293T cells transfected with pCAG-JEC were lysed in lysis buffer as described above. The WCL fractions were loaded into a Superdex 200 column equilibrated with gel filtration buffer. The capsid protein in each fraction was detected by Western blotting. Purified JEV capsid protein from Escherichia coli and infectious JEV particles were also analyzed in this assay. Infectious JEV particles were prepared from JEV-infected 293A cell culture medium, which was replaced with CD293 serum-free medium (Thermo Fisher Scientific) 24 h before harvesting. The supernatant containing JEV particles was filtered through a 0.22-μm filter, concentrated by ultrafiltration using an Amicon Ultra 100K filter (Merck Millipore), and then applied to the gel filtration column.

Expression and purification of recombinant JEV capsid protein.

The JEV capsid protein (amino acid residues 2 to 100) was subcloned into a plasmid vector derived from pET22-b (Merck Millipore), which has an in-frame 10×His tag and an in-frame tobacco etch virus (TEV) protease site located at a site upstream of the MCS. The JEV capsid protein expressing plasmids were introduced into E. coli BL21-CodonPlus (DE3)-RIPL (Agilent). Cells were grown at 37°C until the optical density at 600 nm (OD₆₀₀) reached 0.6. Subsequently, the temperature was shifted to 18°C, and isopropyl-β-d-1-thiogalactopyranoside (IPTG) was added to a final concentration of 0.2 mM to induce expression of the capsid protein. Cells were suspended in a solubilization buffer (1 M NaCl, 1 M KCl, 50 mM Tris-HCl [pH 8.5], 0.5% Triton X-100, 10 mM imidazole, 1 mM phenylmethylsulfonyl fluoride [PMSF]) and then sonicated. The cell lysate was clarified by centrifugation at 10,000 × g and 4°C for 20 min. The clarified lysate was loaded on a nickel-charged chelating column (Ni-Sepharose 6 fast flow; GE Healthcare) equilibrated with wash buffer (1 M NaCl, 1 M KCl, 50 mM Tris-HCl [pH 8.5], 10 mM imidazole). After loading and washing, the protein was eluted from the column with elution buffer (1 M NaCl, 1 M KCl, 50 mM Tris-HCl [pH 8.5], 0.5% [vol/vol] Brij L23, 1 M imidazole) and collected. The protein was fractionated with a Superdex 200 column equilibrated with gel filtration buffer (250 mM NaCl, 125 mM KCl, 50 mM Tris-HCl [pH 8.5]). The protein that eluted as a peak corresponding to JEV capsid protein was further purified by anion exchange (Capto adhere; GE Healthcare) and then concentrated using a centrifugal concentrator (VIVASPIN20; GE Healthcare). Concentrated capsid protein was analyzed by sedimentation velocity experiments with an Optima XL-I analytical ultracentrifuge (Beckman).

Capsid protein solubility assay.

The capsid protein solubility assay was carried out following the procedure from a previous study (44), with slight modifications. In brief, HeLa cells were infected with JEV (MOI of 3 at 48 h postinfection) or transfected with pCAG-JEC, pCAG-OSF-Nucleolin, and pCAG-OSF-B23. Cells were suspended in CSK buffer (10 mM PIPES-NaOH [pH 6.8], 100 mM NaCl, 300 mM sucrose, 3 mM MgCl₂, 1 mM EDTA, 0.5% Triton X-100, 1 mM PMSF, and cOmplete protease inhibitor cocktail [Roche Ltd.]) for 10 min at 4°C. After centrifugation at 800 × g and 4°C for 5 min (the supernatant was collected as S1), the pellet was resuspended in RNase-containing buffer, which is the same as CSK buffer but with 50 mM NaCl and 200 μg/ml of RNase A, and incubated for 1 h at room temperature. After centrifugation at 8,000 × g and 4°C for 5 min (the supernatant was collected as S2), the pellet was resuspended in high-salt buffer, which is the same as CSK buffer but with 0.25 M (NH₄)₂SO₄ instead of 100 mM NaCl, and incubated for 5 min at 4°C. After centrifugation at 8,000 × g and 4°C for 5 min (the supernatant was collected as S3), the pellet was resuspended in DNase-containing buffer, which is the same as CSK buffer but with 50 mM NaCl and 20 U/ml of DNase I, and incubated for 30 min at 37°C. After incubation, the reaction was stopped by the addition of 2 M (NH₄₎₂SO₄ to a final concentration of 0.25 M. After centrifugation at 8,000 × g and 4°C for 5 min (the supernatant was collected as S4), the pellet was resuspended in 5% (wt/vol) SDS (the pellet was collected as P). Each fraction, S1, S2, S3, S4, and P, was subjected to SDS-PAGE and capsid protein, NS3, and OSF-nucleolin/OSF-B23 were detected by Western blotting using anti-capsid protein, anti-NS3, and anti-FLAG antibodies, respectively, and quantified using ImageJ software (45). Immunofluorescence assays were performed as follows. JEV-infected or capsid protein-encoding-plasmid-transfected HeLa cells grown on coverslips were treated in the following order: CSK buffer for 10 min at room temperature, RNase-containing buffer for 1 h at room temperature, high-salt buffer for 5 min at 4°C, and DNase-containing buffer for 30 min at 37°C. After each treatment step, cells were fixed, immunostained with anti-JEV capsid protein antibody, and visualized with Alexa Fluor 350-conjugated anti-rabbit antibody (Life Technologies), as described above.

Statistical analysis.

Values in graphs are represented as means ± standard deviations. To compare each group with a control group, a one-way analysis of variance (ANOVA) with Dunnett’s test was used and performed using R (http://www.R-project.org) or GraphPad Prism v 7.0 software (GraphPad Software Inc., La Jolla, CA).