Mutagenesis of the DI/DIII Linker in Dengue Virus Envelope Protein Impairs Viral Particle Assembly

Abstract

The dengue virus (DV) envelope (E) protein is important in mediating viral entry and assembly of progeny virus during cellular infection. Domains I and III (DI and DIII, respectively) of the DV E protein are connected by a highly conserved but poorly ordered region, the DI/DIII linker. Although the flexibility of the DI/DIII linker is thought to be important for accommodating the structural rearrangements undergone by the E protein during viral entry, the function of the linker in the DV infectious cycle is not well understood. In this study, we performed site-directed mutagenesis on conserved residues in the DI/DIII linker of the DV2 E protein and showed that the resulting mutations had little or no effect on the entry process but greatly affected virus assembly. Biochemical fractionation and immunofluorescence microscopy experiments performed on infectious virus as well as in a virus-like particle (VLP) system indicate that the DI/DIII linker mutants express the DV structural proteins at the sites of particle assembly near the ER but fail to form infectious particles. This defect is not due to disruption of E's interaction with prM and pr in immature and mature virions, respectively. Serial passaging of the DV2 mutant E-Y299F led to the identification of a mutation in the membrane-proximal stem region of E that fully compensates for the assembly defect of this DI/DIII linker mutant. Together, our results suggest a critical and previously unidentified role for the E protein DI/DIII linker region during the DV2 assembly process.

INTRODUCTION

Dengue virus (DV) is a significant human pathogen and the cause of dengue fever and dengue hemorrhagic fever. The four dengue virus serotypes (DV1, DV2, DV3, and DV4) are members of the Flaviviridae family and have a positive-sense RNA genome encoding a single polyprotein. This polyprotein is processed by host- and DV-encoded proteases into 10 proteins: three structural proteins—core (C), premembrane (prM), and envelope (E)—and seven nonstructural proteins—NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5. Following RNA replication and translation, the viral RNA is encapsidated by C to form the nucleocapsid that buds at the endoplasmic reticulum (ER) membrane to associate with the prM and E proteins and form an immature DV virion (25). This immature virion then transits through the secretory pathway, where the virion is matured through cleavage of prM into the membrane (M) protein by furin in the trans-Golgi (21, 22). The mature virion then exits the cell and enters new cells via endocytosis. The acidic environment in the host endosome serves as the physiological trigger for a major conformational change in the E protein that leads to the insertion of its fusion loops in the host endosomal membrane (2, 16). This results in fusion of the viral membrane with the host membrane and delivery of the viral genome to the cytoplasm.

The DV E protein belongs to the class II fusion proteins, which share a tertiary structure (11). The DV E protein is anchored in the membrane by two transmembrane domains (28). The remainder of the E protein is composed of three globular domains that are linked to the anchor by the stem region, a membrane proximal helix-loop-helix (28). Domain I (DI) is a β-barrel at the core of the E monomer that connects at one end to DII, a “finger-like” domain that bears the fusion loop at its tip. At its other end, DI connects to DIII via a short polypeptide, the DI/DIII linker (Fig. 1A). The 11 residues that comprise this linker are moderately conserved but exhibit poorly ordered structure in high-resolution crystal structures of the DV2 soluble E (sE) prefusion dimer (15) and postfusion trimer (16). From these structures, it appears that the DI/DIII linker accommodates a large displacement during the fusion process when domain III packs against domain I (16). Although a role for the analogous DI/DIII linker in the trimerization of the Semliki Forest virus (SFV) class II fusion protein E1 has been recently demonstrated (30), the absence of the DI/DIII linker did not prevent trimerization of a recombinant DV2 E protein comprised of domains I and II (DI/DII) in biochemical assays (12). This suggests variability in the function of the DI/DIII linker across different class II fusion proteins and raises the question of what other aspects of viral replication might be affected by the conserved residues of this linker.

Fig 1.

(A) Structure of E protein in the dimeric conformation on the virion surface at neutral pH (Protein Data Bank [PDB] accession number 1OAN). The DI/DIII linker is colored in green, and an arrow indicates its localization. (B) The sequences for dengue virus serotypes 1, 2, 3, and 4 (DV1, DV2, DV3, and DV4), Japanese encephalitis virus (JEV), tick-borne encephalitis virus (TBEV), West Nile virus (WNV), and yellow fever virus (YFV) envelope proteins were aligned using ClustalW2. Conserved (*) and semiconserved (: and .) amino acids are indicated below the alignment. Domain I (DI) and domain III (DIII) are underlined, and the DI/DIII linker is highlighted in green. Amino acid residues 296, 298, and 299, mutagenized in this study, are highlighted in black.

To probe the function of the DI/DIII linker of the DV2 E protein in viral replication, we chose residues that are conserved across all four DV serotypes as well as in other members of the flavivirus genus and introduced conservative amino acid changes that were unlikely to disturb the flexible structure of the linker. Here, we show that these mutations have little effect on the entry step of DV2 infection but strongly impair the production of both DV2 virions and virus-like particles (VLPs). Further characterization of the phenotype of the DI/DIII linker mutants revealed an unexpected role of the unstructured DI/DIII linker in promoting the formation of viral particles on the ER membranes. In addition, we identified a mutation in the DV2 E stem region that completely restored the defect in viral particle formation caused by mutations in the DI/DIII linker. Overall, our results demonstrate for the first time the critical role of the E protein DI/DIII linker in mediating DV assembly and highlight the presence of a network of interactions among E proteins that facilitate this process.

MATERIALS AND METHODS

Cell lines.

Huh7 and HEK293T cells were grown in Dulbecco's minimal essential medium (DMEM) supplemented with 10% fetal bovine serum (FBS). BHK21 cells were grown in MEM-α supplemented with 5% FBS. LoVo cells were grown in Ham's F-12 medium supplemented with 20% FBS.

Antibodies.

Mouse hybridomas producing the monoclonal antibody 6F3.1 anti-DV2 C were generously provided by John Aaskov (7). Anti-DV2 C antibody was produced, purified, and conjugated with biotin by Genscript. Mouse monoclonal antibodies 2E11 (anti DV2 prM-E heterodimer [19]) and prM6.1 (anti DV2 prM [10]) were kindly provided by Chunya Puttikhunt, Watchara Kasinrerk, and Prida Malasit. Mouse hybridomas producing the monoclonal antibody 4G2 anti-DV E were purchased from ATCC (HB-112). Mouse monoclonal antibody 3H5-1 that recognizes DV2 E was purchased from Millipore (MAB8702). Mouse monoclonal antibody E1D8 anti-DV2 NS3 was a kind gift from Eva Harris (5). Antibodies against GAPDH (glyceraldehyde-3-phosphate dehydrogenase), blasticidin-S deaminase (BSD), and calnexin were respectively purchased from GeneTex (GTX100118) and Abcam (ab38307 and ab22595). Horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG and anti-rabbit IgG antibodies were obtained from Bio-Rad Laboratories (170-6516 and 170-6515, respectively). Fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse IgG antibody, FITC-conjugated goat anti-mouse IgM antibody, FITC-conjugated streptavidin, and Alexa Fluor 647-conjugated goat anti-mouse IgG antibody were obtained from Jackson ImmunoResearch Laboratories (115-095-003), Abcam (ab97229), BioLegend (405201), and Life Technologies (A-21235), respectively.

Production of recombinant DV2.

The infectious cDNA clone of DV2-NGC used in this study, pRS-D2, was kindly provided by Barry Falgout (18). The strategy to create mutant pRS-D2 plasmids by exploiting yeast recombination was adapted from a previously described protocol (8). We introduced E-G296A, E-S298A, and E-Y299F mutations in a PCR fragment encompassing DV2 nucleotides 121 to 2246 using primers listed in Table S1 in the supplemental material. The double mutations E-Y299F/Q400H were introduced directly through amplification of the compensatory virus cDNA using the primers D2-121 FW and D2-2246 RV. The single mutation E-Q400H was produced by removal of mutation E-Y299F from the plasmid pRS-D2(E-Y299F/Q400H). The mutagenized PCR fragments were then cotransformed in yeast (INVSc1; C810-00; Life Technologies) with the pRS-D2 plasmid linearized with SalI and BamHI (located at nucleotide positions 166 and 2203). Yeast colonies containing the recombined pRS-D2 plasmids were selected on tryptophan dropout medium (Y1876; Sigma). Yeast DNA was extracted as previously described (4, 8) and used to transform Stbl2 cells (10268-019; Life Technologies) for multiplication. In vitro transcripts were synthesized from SacI-linearized pRS-D2 using an SP6-Scribe standard RNA IVT kit (C-AS3106; CellScript) and m7G(5′)ppp(5′)A RNA cap structure analog (S1405L; New England BioLabs) following the manufacturers' instructions. Huh7 cells were washed twice in phosphate-buffered saline (PBS), and 1 × 10⁶ cells were electroporated with DV2 in vitro transcripts using an ECM 830 electroporator (BTX Harvard Apparatus) at the following settings: 5 pulses at 820 V, 100 μs per pulse with 1.1-s intervals. Following electroporation, the cells were plated in DMEM supplemented with 2% FBS in a 24-well plate, 6-well plate, or 10-cm dish format when performing, respectively, immunofluorescence analysis, Western blotting, or subcellular fractionation. Although viral protein expression of E mutants typically peaked at early times postelectroporation (days 1 and 2), we observed some variability from one electroporation to another. This was manifested in differing numbers of cells with detectable viral protein expression on day 1 as well as by a difference in the kinetics of steady-state expression of E protein for the mutants, which in some cases disappeared as early as day 2 but in other cases disappeared on day 3. We attribute this variability to differences in the efficiencies of transfection, fitness of the cells, and/or quality of the in vitro transcripts from one experiment to another.

RVP production and entry assay.

The plasmids used for recombinant viral particle (RVP) production were kindly provided by Ted Pierson (4). The pCDNA6.2-D2.CprME plasmid was mutagenized using the primers listed in Table S1 in the supplemental material and following the instructions of the QuikChange Lightning site-directed mutagenesis kit (210518; Stratagene). The resulting plasmids were amplified in Stbl2 cells (10268-019; Life Technologies). To produce RVPs, HEK293T cells were plated in a 10-cm dish and then cotransfected with pCDNA6.2-D2.CprME (4) and pWIIrep-REN-IB (17) using Lipofectamine 2000 (11668-019; Life Technologies) according to the manufacturer's instructions. The supernatants were collected at 2 days posttransfection and precipitated for 3 h at 4°C by addition of 0.075% of polyethylene glycol 8000 (PEG 8000). The RVPs were pelleted by centrifugation at 10,000 × g for 15 min at 4°C. The pellet was suspended in 200 μl of 1× TNE (10 mM Tris-HCl [pH 7.5], 2.5 mM EDTA, 50 mM NaCl), loaded over a sucrose cushion (800 μl of 12.5% sucrose in 1× TNE), and centrifuged at 100,000 × g for 2.5 h at 4°C. The supernatants were discarded, and the purified RVPs were suspended in 50 μl of 1× TNE. Ten microliters of purified RVPs was diluted with 90 μl of Earle's balanced salt solution (EBSS) to infect Huh7 cells seeded in 48-well plates. At 24 h postinfection, the cells were collected and the samples were processed following the instructions in the Renilla luciferase assay system (E2810; Promega). The Renilla luciferase signal was read using a Perkin Elmer EnVision plate reader.

Production and purification of VLPs.

The plasmid used for production of virus-like particles (VLPs) was kindly provided by Stephen Harrison. Briefly, the codon-optimized sequence of DV2-FGA/02 prM-E described by Wang and colleagues (24) was introduced into the pcDNA3.1 vector. The pcDNA3.1-D2.VLP plasmid was mutagenized using the primers listed in Table S1 in the supplemental material and following the instructions of the QuikChange Lightning site-directed mutagenesis kit (210518; Stratagene). To produce VLPs, Huh7 cells were plated in a 10-cm dish and transfected with pcDNA3.1-D2.VLP using Lipofectamine 2000 (11668-019; Life Technologies) according to the manufacturer's instructions. The supernatants were collected at 2 days posttransfection and purified as described above for the RVPs.

Immunofluorescence analysis (IFA).

Coverslips were fixed and permeabilized for 15 min in fixative [100 mM piperazine-N,N′-bis(2-ethanesulfonic acid) (PIPES) (pH 6.8), 10 mM EGTA, 1 mM MgCl₂, 3.7% formaldehyde, 0.2% Triton X-100) and then washed twice in PBS. After blocking for 30 min in PBS plus 5% bovine serum albumin (BSA), cells were incubated with primary antibodies for 1 h, washed with PBS, incubated with secondary antibodies for 1 h, washed with PBS, and stained with DAPI (4′,6-diamidino-2-phenylindole) stain (D3571; Life Technologies). The coverslips were mounted with ProLong Gold antifade reagent (P36930; Life Technologies) prior to wide-field imaging (nearest neighbors analysis) using a Marianas spinning disk confocal microscope. The staining for the E protein was performed using the mouse monoclonal antibody 3H5-1 (MAB8702; Millipore). For double staining of E and C proteins, the coverslips were first incubated with anti-DV2 E antibody (MAB8702; Millipore) followed by staining with Alexa Fluor 647-conjugated goat anti-mouse IgG antibody. After several washes, the cells were labeled with anti-DV2 C:biotin antibody, followed by staining with FITC-conjugated streptavidin.

Western blotting.

Protein lysates were prepared by cell lysis in radioimmunoprecipitation assay (RIPA) buffer (BP-115X; Boston Bioproducts) containing protease inhibitors (11836170001; Roche). Equal amounts of proteins were loaded on 10% Tris-glycine polyacrylamide gel and transferred to a polyvinylidene difluoride (PVDF) membrane (P2938-1ROL; Sigma). After blocking the membrane for 1 h at room temperature in PBS-Tween (PBS-T) plus 5% milk, the blot was incubated overnight at 4°C with appropriate dilutions of the primary antibodies. For detection of DV2 E protein, we used the monoclonal antibody 4G2 anti-DV E (HB-112; ATCC). The membrane was then washed in PBS-T and then incubated for 1 h at room temperature in the presence of HRP-conjugated secondary antibodies. After washes in PBS-T, the membrane was developed with enhanced chemiluminescence reagents (32209; Pierce) and exposed to film. When necessary the bands were quantified using ImageJ (W. S. Rasband, U.S. National Institutes of Health, Bethesda, Maryland, USA, http://imagej.nih.gov/ij/, 1997 to 2011).

Focus-forming assay (FFA).

BHK-21 cells were seeded in 24-well plates. Aliquots from infections were thawed at 37°C in a water bath. Ten-fold dilutions in EBSS were prepared in duplicate, and 100 μl of each dilution was added to cells. Plates were incubated for 1 h at 37°C and rocked every 15 min. Unadsorbed virus was removed by washes with PBS, after which 1 ml of MEM-α supplemented with carboxymethylcellulose (CMC), HEPES, and 2% FBS was added to each well and incubated at 37°C for 3 days. The CMC overlay was aspirated, and cells were washed with PBS and fixed with methanol for 15 min at −20°C. After fixation, the cells were washed with PBS and incubated for 1 h at room temperature with anti-DV2 C antibody, followed by incubation with HRP-conjugated anti-mouse IgG antibody. The plates were developed with Vector VIP peroxidase substrate kit (SK-4600; Vector Laboratories) following the manufacturer's instructions.

Intracellular titration.

Huh7 cells were trypsinized, pelleted by centrifugation at 400 × g for 5 min at room temperature, and resuspended in 500 μl of DMEM plus 2% FBS. The cells were then subjected to 4 freeze-thaw cycles (3 min at −80°C, 3 min at room temperature). The cell debris were pelleted by centrifugation at 3,000 × g for 5 min at room temperature. The supernatants were collected, and intracellular virus was titrated by FFA.

Subcellular fractionation.

The subcellular fractionation protocol was adapted from the work of Boson et al. (6). Huh7 cell pellets were trypsinized and washed in PBS. The pellets were suspended in 900 μl of 10 mM HEPES-NaOH, pH 7.8 (hypo-osmotic buffer). The cells were allowed to swell on ice for 10 min and then centrifuged at 800 × g for 2 min at room temperature. The medium was returned to iso-osmocity by the removal of 600 μl of the supernatants and the addition of 300 μl of hyperosmotic buffer (0.6 M sucrose, 10 mM HEPES-NaOH, pH 7.8). The cells were then disrupted by passaging 10 times through a 26-G needle, and the postnuclear fraction (600 μl) was isolated by centrifugation for 30 min at 13,000 × g at 4°C. The postnuclear fraction was mixed with 600 μl of OptiPrep (60% iodixanol; D1556; Sigma) and loaded at the bottom of an ultracentrifuge tube. Solutions (1,200 μl) of 20% and 10% iodixanol were prepared by mixing with the hypo-osmotic buffer and were overlaid in the ultracentrifuge tube. The gradients were centrifuged at 50,000 rpm in an SW60Ti rotor for 3 h at 4°C. Nine fractions of 400 μl were collected from the top of the gradient, and 300 μl was precipitated with trichloroacetic acid (TCA) prior to Western blot analysis. The remaining 100 μl of each fraction was titrated by FFA.

Endo-H assay.

Equal amounts of protein lysates were subjected to digestion with endo-β-N-acetylglucosaminidase H (endo-H) (P0702S; New England BioLabs) in the presence of the manufacturer's buffer G5 and protease inhibitors (11836170001; Roche) for 30 min at 37°C. Undigested and digested samples were separated on an 8% Tris-glycine polyacrylamide gel and then analyzed by Western blotting with anti-DV E 4G2 antibody as described above.

Coimmunoprecipitation.

Transfected Huh7 cells were lysed in 50 mM Tris-HCl (pH 7.4), 300 mM NaCl, 5 mM EDTA, 1% Triton X-100 in the presence of protease inhibitors (11836170001; Roche). The postnuclear fraction was isolated by centrifugation for 10 min at 10,000 × g at 4°C. Fifteen micrograms of cell lysates was incubated with 1 μg of anti-DV2 E antibody (MAB8702; Millipore) for 6 h at 4°C under constant rocking. Twenty microliters of an anti-mouse IgG-agarose 50% slurry (A6531; Sigma) was then added to the cell lysate-antibody mixture and was incubated overnight at 4°C under constant rocking. The beads were then pelleted by centrifugation for 1 min at 1,000 × g and washed three times with 1× PBS. The captured proteins were eluted by addition of nonreducing loading buffer and were then analyzed by Western blotting.

Compensatory virus passaging and sequencing.

Viral RNA was extracted from supernatants using a QIAamp viral RNA extraction kit (52906; Qiagen) following the manufacturer's instructions. cDNA was synthesized from 8 μl of viral RNA using the SuperScript III first-strand synthesis system (18080-051; Life Technologies). Part of the DV2 genome was amplified by PCR using Phusion DNA polymerase (M0530L; New England BioLabs) and the primers D2-1 FW and D2-5481 RV. The PCR fragment was then sequenced at the DF/HCC DNA sequencing facility using the primers D2-1 FW, D2-849 FW, D2-2246 RV, D2-1814 FW, D2-2586 FW, D2-4062 RV, D2-3630 FW, and D2-5481 RV (see Table S1 in the supplemental material).

RESULTS

The mutagenesis of the DI/DIII linker in the DV2 E protein impairs viral production.

To examine the role of the E protein DI/DIII linker in DV2 infection, we mutagenized pRS-D2, a plasmid that bears an infectious full-length DV2 cDNA clone (18). We focused on three conserved residues at positions 296, 298, and 299 of the E protein and mutated them to alanine, alanine, and phenylalanine, respectively (Fig. 1B). Three other residues (L292, K295, and C303) are conserved in the DI/DIII linker but were not considered for analysis. The side chains of the leucine and the cysteine at positions 292 and 303 were suspected to interact with residues in domains DI and DIII, respectively. The lysine at position 295 has a charged side chain, which we did not mutate because it could not be replaced conservatively without affecting the flexibility of the DI/DIII linker. The mutagenized plasmids carrying the intended mutations were used in in vitro transcription reactions to produce capped, full-length DV2 genomic RNAs. These transcripts were electroporated into Huh7 cells to produce viral progeny. Transfected cells were examined by immunofluorescence analysis (IFA) for DV2 core (C) protein at 1 day, 2 days, and 3 days postelectroporation. At 1 day postelectroporation, prior to virus spread, a similar number of Huh7 cells were infected (Fig. 2, top), confirming comparable levels of transfection efficiency. Interestingly, at 2 and 3 days postelectroporation, virus spread was not observed for any of the mutant viruses, in contrast to what was observed for the wild-type virus (Fig. 2, middle and bottom).

Fig 2.

Huh7 cells were electroporated with in vitro-transcribed genomic RNAs for wild-type DV2, DV2(E-G296A), DV2(E-S298A), or DV2(E-Y299F) and then analyzed by immunofluorescence staining for the presence of the DV2 core (C) protein on days 1, 2, and 3 postelectroporation. Although comparable numbers of core-positive cells are visible for all viruses on day 1, viruses with mutations in the DI/DIII linker exhibit greatly reduced numbers of cells that are positive for DV2 C on days 2 and 3 postelectroporation relative to the wild type, suggesting the absence of viral spread for these mutants. A representative experiment out of n > 3 repeats is shown. The images were taken at ×200 magnification.

Next, we analyzed steady-state viral protein expression at 1 and 2 days postelectroporation. We chose to analyze the accumulation of structural (E, C) as well as nonstructural (NS3) proteins. Increasing amounts of NS3, E, and C proteins were observed for the wild-type virus, while the steady-state expression of these proteins remained low for all mutant viruses (Fig. 3A). To determine if the absence of viral spread was a result of the failure to secrete infectious viral particles, we analyzed the culture supernatants collected on 1 and 2 days postelectroporation. As assayed by a standard focus-forming assay (FFA), the wild-type virus-infected cells produced infectious particles, while no infectious virus was detected in the supernatants of the cells electroporated with DV2(E-G296A) and DV2(E-S298A) RNA (Fig. 3C). In some electroporations, we were able to detect low levels of infectious particles in the supernatants collected for DV2(E-Y299F) at late times postelectroporation (Fig. 3C). The presence of infectious particles in this case was found to be linked to a loss of the E-Y299F mutation or the rise of a compensatory mutation (discussed later; see Fig. 9 below). Since the lack of infectious particles in the supernatants could reflect a defect in viral assembly and secretion of viral particles or, alternatively, could indicate that viral particles are produced but are noninfectious due to a defect in viral entry, we analyzed the culture supernatants by Western blotting to determine if viral particles were secreted from the electroporated cells. As shown in Fig. 3B, E protein was undetectable in the supernatants from DV2(E-G296A), DV2(E-S298A), and DV2(E-Y299F), indicating that mutations in the DI/DIII linker prevent the release of infectious virus, even though viral proteins are clearly expressed. Overall, these data strongly suggested that the DI/DIII linker mutants have a defect that prevented the release of viral particles.

Fig 3.

Huh7 cells were electroporated with in vitro-transcribed DV2, DV2(E-G296A), DV2(E-S298A), or DV2(E-Y299F) genomic transcripts. A representative experiment out of n > 3 repeats is shown. (A) The cell lysates were analyzed for the presence of DV2 C, E, and NS3 proteins as well as for the presence of GAPDH by Western blotting at 1 and 2 days postelectroporation. Mutations in the DI/DIII linker are associated with decreased steady-state expression of E, C, and NS3. (B) Examination of the culture supernatants by Western blotting for the presence of DV2 E protein at 1 and 2 days postelectroporation revealed that secretion of viral particles to the culture supernatant was greatly reduced for the DI/DIII mutants relative to the wild-type virus. (C) Quantification of infectious virus in the supernatants at 1 and 2 days postelectroporation by focus-forming assay, moreover, revealed no detectable infectious virus for the DI/DIII linker mutants. The low titer of infectious particles evident for the DV2(E-Y299F) mutant on day 2 postelectroporation was associated with reversion to wild type or the presence of a suppressor mutation. The error bars represent the standard deviation of the assay (titrations were done in duplicate).

Fig 9.

(A) Sequencing data for the compensatory mutation found at amino acid position Q400 in E protein/nucleotide 2136 in the DV2 sequence. Black peak, G; blue peak, C; green peak, A; red peak, T. Translation of the wild-type DV2 sequence is shown on the top right, and the nucleotide changes identified at passage 5 for DV2(E-Y299F) are indicated below (P5). (B) Structure of the DV2 E protein on the virion surface at neutral pH (PDB accession number 1OAN). The stem region (helices H1 and H2, transmembrane domains T1 and T2) was drawn following the crystal structure data (PDB accession number 1P58). The DI/DIII linker is colored in green and its localization is indicated by an arrow. The Q400 residue is located by a green dot in H1, and its localization is indicated by an arrow. The sequences for dengue virus serotypes 1, 2, 3, and 4 (DV1, DV2, DV3, and DV4), Japanese encephalitis virus (JEV), tick-borne encephalitis virus (TBEV), West Nile virus (WNV), and yellow fever virus (YFV) envelope proteins were aligned using ClustalW2. Conserved (*) and semiconserved (:) amino acids are indicated below the alignment. The helix H1 from the E stem region is in gray, and the mutated Q400 amino acid is in green. (C) Huh7 cells were transfected with D2.VLP, D2(E-Y299F).VLP, D2(E-Y299F/Q400H), and D2(E-Q400H) plasmids (DV2, Y299F, Y299F/Q400H, Q400H, respectively) and analyzed at 2 days posttransfection. The cell lysates were analyzed by Western blotting for DV2 E and GAPDH as a loading control. The VLPs released in the supernatants were purified and analyzed by Western blotting for DV2 E [E(sup)]. The presence of the E-Q400H mutation in the context of E-Y299F restored the appearance of VLPs in the culture supernatants. A representative experiment out of n > 3 repeats is shown. (D and E) Huh7 cells were electroporated with in vitro-transcribed DV2, DV2(E-Y299F), DV2(E-Y299F/Q400H), and DV2(E-Q400H) genomic RNAs (DV2, Y299F, Y299F/Q400H, and Q400H, respectively). A representative experiment out of 3 repeats is shown. (D) The cell lysates were analyzed for the presence of DV2 C and E as well as for the presence of GAPDH by Western blotting at 3 days postelectroporation. (E) Quantification of infectious virus in the supernatants at 3 days postelectroporation was evaluated by focus-forming assay. The presence of the E-Q400H mutation fully compensated for the E-Y299F mutation and restored the production of infectious virus from cells electroporated with DV2(E-Y299F/Q400H) genomic RNA to titers comparable to the wild-type. The error bars represent the standard deviation of the assay (titrations were done in duplicate).

Mutagenesis of the DV2 E DI/DIII linker does not perturb entry.

Since a major function of E is to mediate viral entry, we also questioned whether mutations in the DI/DIII linker had an effect on viral entry that might have been obscured in the electroporation experiments by the strong defect in virion secretion. Notably, the flexibility of the DI/DIII linker is thought to be important for accommodating the large-scale rearrangements that these domains undergo during the fusion process (16). To address this, we employed a recombinant viral particle (RVP) system (4) in which successful entry of the recombinant RVPs leads to expression of a luciferase reporter gene. RVPs were produced using a previously established system (4) by transfection of cells with a plasmid that expresses DV2 C, prM, and E proteins along with a plasmid encoding a West Nile virus (WNV) subgenomic replicon expressing the viral nonstructural proteins. Cotransfection of those plasmids results in the production of RVPs composed of DV2 E and prM/M proteins and containing a nucleocapsid formed by DV2 C and WNV RNA (4). A Renilla luciferase reporter encoded by the WNV RNA permits easy quantification of WNV expression in transfected cells as well as the productive entry of the RVP into new cells (4). We observed that the mutations in the DI/DIII linker were associated with a significant decrease in RVP production, as quantified by Western blot analysis of the supernatants for E (Fig. 4B), highlighting an obvious defect in viral particle assembly/release consistent with the phenotype observed in the electroporation experiments. Since we were still able to recover small amounts of RVPs containing the E-S298A or E-Y299F mutations, we examined whether those mutations had an additional effect on entry, taking advantage of the luciferase reporter as an extremely sensitive readout for this process. Accordingly, we infected Huh7 cells with purified RVPs and measured the Renilla luciferase activity at 24 h postinfection. While the luciferase signal that marked productive entry varied among the wild type and mutants, decreased entry activity could be correlated to and explained by decreased yield of RVPs, as reflected by the E content in the RVP preparation (Fig. 4C). This relationship between the amount of E in the RVP preparation (Fig. 4B) and the luciferase activity for those RVPs in the entry assay (Fig. 4C) supported the idea that the DI/DIII linker mutants are still competent for viral entry and that the failure of these mutants to spread (Fig. 2) is not due to an entry defect. Rather, these data suggest that mutations in the DI/DIII linker instead cause a major defect in assembly and release of viral particles.

Fig 4.

HEK293T cells were transfected with D2.CprME, D2(E-G296A).CprME, D2(E-S298A).CprME, and D2(E-Y299F).CprME plasmids (DV2, 296, 298, and 299 respectively) along with the WNV replicon plasmid. A representative experiment out of n > 3 repeats is shown in A and B. (A) At 2 days posttransfection, intracellular expression of DV2 E and C proteins was assessed by Western blotting using blasticidin-S deaminase (BSD), encoded by the D2.CprME and WNV replicon plasmids, as a transfection control. (B) To quantify the yield of RVPs released to the culture supernatants (sup), RVPs were purified from the supernatant and then analyzed by Western blotting for the presence of the E protein. (C) Following infection of Huh7 cells with 10 μl of each purified RVP, intracellular Renilla luciferase activity was quantified at 1 day postinfection as a measure of successful RVP entry. The E levels obtained after Western blot quantification (described in B) averaged from three independent experiments were plotted along with the entry values obtained from the corresponding entry assay. The wild-type DV2 and the DV2(E-S298A) and DV2(E-Y299F) mutants exhibit very comparable steady-state expression levels of both E and C but much larger differences in yield of RVP secreted to the supernatant. Despite the reduced yield of DV2(E-S298A) and DV2(E-Y299F) RVPs, the particles produced appear competent for entry into new cells, as evidenced by the luciferase signal produced when these RVPs are used to infect new cells. Since the DV2(E-G296A) mutant did exhibit somewhat reduced steady-state expression of E, it is less clear whether the defect in RVP secretion was due to limiting expression of E and whether the background luciferase signal detected for this mutant reflected lack of entry. AU, arbitrary units.

Mutations in the DV2 E protein DI/DIII linker prevent the assembly of infectious virions within the cells.

To examine the effect of these mutations in assembly and/or egress of viral particles, we used immunofluorescence to visualize the subcellular localization of structural proteins within infected cells. For simplification, the results for the DV2(E-Y299F) mutant are compared here to those of the wild type. The DV2(E-G296A) and DV2(E-S298A) mutants displayed a phenotype similar to that of the DV2(E-Y299F) mutant (data not shown). At 1 day postelectroporation, colocalization of E and C proteins was observed within cells electroporated with DV2 or DV2(E-Y299F) (Fig. 5A). At 2 days postelectroporation, colocalization of E and C proteins was still observed for wild-type DV2 (Fig. 5B) and appeared at discrete locations resembling sites of viral assembly (20) (Fig. 5B, white arrows). For the DV2(E-Y299F) mutant (Fig. 5B) and other DI/DIII linker mutants (data not shown), the number of cells positive for E protein was reduced relative to the number of cells still positive for C protein on day 2 postelectroporation. Although C protein in these cells was observed to localize at discrete sites in a subset of cells, we did not observe localization of C in sites resembling sites of viral assembly (Fig. 5B, bottom), and the subcellular localization of C in cells lacking E appeared aberrant relative to its localization for the wild-type virus. These data suggest that proper association of viral nucleocapsids with the viral envelope does not occur in the presence of mutations in the DI/DIII linker and that this is consistent with the reduced particle formation observed in Fig. 3 and 4.

Fig 5.

Huh7 cells were electroporated with in vitro-transcribed DV2 or DV2(E-Y299F) genomic RNA. The coverslips were fixed and analyzed by immunofluorescence staining for the DV2 C and E proteins. Representative images from one experiment out of n > 3 repeats are shown. The top images were taken at ×200 magnification, while the close-up images on the lower strip were taken at ×400 magnification. (A) On day 1 postelectroporation, cells that were positive for DV2 contained both E (red) and C (green) proteins, and both E and C colocalized in the case of both wild-type DV2 and the E-Y299F mutant. (B) On day 2 postelectroporation, the wild-type DV2 exhibited colocalization of E (red) and C (green) proteins at discrete sites in the cytoplasm (white arrow) that resembled sites of viral assembly. In contrast, the E-Y299F protein appeared to have been lost from many of the infected cells, as reflected by a lesser number of cells that are positive for E relative to the number of cells positive for C. Although much of the E-Y299F protein detected appeared to colocalize with C, the staining was distributed throughout the cytoplasm and did not resemble sites of viral assembly. Aberrant localization of C was observed in some cells expressing C but lacking E. E was found not to colocalize with C in a rare subset of cells (purple arrowhead).

To further investigate the idea that mutations in the DI/DIII linker prevent the formation of infectious particles, Huh7 cells were lysed by multiple freeze-thaw cycles at 2 days postelectroporation, and the presence of infectious virus in the lysates was titered by focus-forming assay. Wild-type virus-infected cells had a relatively high titer of intracellular virus, while no infectious intracellular particles were detected for the DI/DIII linker mutants (Fig. 6A). Since no infectious virus was found in the electroporated cells, we questioned if this were due to the failure of the viral structural proteins to associate at assembly sites within the infected cells. To test this hypothesis, we fractionated electroporated Huh7 cell lysates using gradient centrifugation. We analyzed the fractions for the presence of E, prM, and C proteins. Since DV virion assembly sites are associated with endoplasmic reticulum (ER) membranes (25), we used calnexin, an ER resident protein, as a marker. For the wild-type virus, the majority of E and prM accumulated in fractions 4 to 6, whereas the majority of C protein accumulated in fractions 6 to 9 (Fig. 6A). We also measured the viral titer in each fraction and found that, in the case of the wild type, most of the infectious virus accumulated in fraction 6, with a subset of infectious virus in fraction 4 (Fig. 6B, top). As expected, no infectious virus was found in any of the fractions for the DI/DIII linker mutants. Interestingly, although the steady-state abundance of viral proteins was much lower in the case of the mutant viruses, we observed a similar pattern of distribution for the E and prM proteins in the different fractions (Fig. 6B); however, the C protein was noticeably absent from fraction 6, where the infectious virus is expected to accumulate. While this might suggest that the mutant E protein fails to associate with the nucleocapsid, we noted that a minority of E did colocalize with C protein in fractions 7 and 8 of the density gradient for a subset of the mutant viruses and that this colocalized material could reflect a physical association of E with C.

Fig 6.

Huh7 cells were collected 2 days postelectroporation with in vitro-transcribed DV2, DV2(E-G296A), DV2(E-S298F), or DV2(E-Y299F) genomic RNA. A representative experiment out of 2 repeats is shown. (A) The intracellular and extracellular titers were evaluated by focus-forming assay. (B) The cellular contents were analyzed by fractionation on an iodixanol gradient. Nine fractions were collected from the top of the gradient, and the individual fractions were titrated by focus-forming assay. The top diagram represents the distribution of infectious virus in the fractions collected from cells electroporated with wild-type DV2. The majority of the infectious virus was found in fraction 6 and is represented by the darkest shade of gray. The range of FFU/ml for each shade of gray, from the darkest to the lightest, is as follows: 500 to 1,000, 100 to 500, and 10 to 100. No infectious virus was found in cells electroporated with the DV2(E-G296A), DV2(E-S298F), or DV2(E-Y299F) RNAs. Each fraction was also TCA precipitated and analyzed by Western blotting for the presence of calnexin, DV2 C, prM, and E. For wild-type DV2, we observed that all structural proteins coaccumulated mostly in fractions 4 and 6, which contained the ER (as shown through calnexin staining) and coincided with the presence of infectious virus. In the case of the mutant viruses, the C protein was remarkably absent from ER-enriched fractions 4 and 6, despite the presence of E and prM in those fractions.

Mutations in the DI/DIII linker do not affect E protein stability but prevent the formation of VLPs.

To address whether the defect in infectious virus formation was linked to a lack of nucleocapsid incorporation into virions, we chose to use virus-like particles (VLPs), which consist solely of E and prM proteins (24). Another advantage to this system is that it allows us to distinguish a viral assembly defect from a viral entry and/or protein stability defect, since the VLPs are formed independently of viral replication. The E-Y299F mutation was introduced into the VLP expression plasmid, pcDNA3.1-D2.VLP, and transfected into Huh7 cells, using wild-type pcDNA3.1-D2.VLP as a control. E and prM expression were monitored by immunofluorescence at 2 days posttransfection (Fig. 7A). No obvious differences in steady-state expression and subcellular localization of E or prM were observed between wild-type and mutant transfections (Fig. 7A).

Fig 7.

Huh7 cells were transfected with D2.VLP and D2(E-Y299F).VLP plasmids (DV2 and Y299F, respectively) and analyzed at 2 days posttransfection. A representative experiment out of n > 3 repeats is shown. (A) The coverslips were fixed and analyzed by immunofluorescence for DV2 prM and E. The images were taken at ×200 magnification. (B) Intracellular contents were analyzed by fractionation on an iodixanol gradient. Nine fractions were collected from the top of the gradient, and each individual fraction was TCA precipitated and then analyzed by Western blotting for calnexin, DV2 prM, and E. (C) The cell lysates were analyzed by Western blotting for DV2 E and GAPDH as a loading control. The VLPs released in the supernatants were purified and analyzed by Western blotting using DV2 E antibody [E (sup)]. Notably, there was a significant reduction in the release of VLPs from cells transfected with the D2(E-Y299F).VLP plasmid despite the accumulation of intracellular E to levels comparable to those observed for the wild-type protein.

Likewise, subcellular fractionation analysis of the transfected cells revealed no differences between the wild-type and mutant constructs (Fig. 7B). A predominant coaccumulation of E and prM proteins in fractions 4 to 6 coincided with the distribution of the ER marker (Fig. 7B). This result was consistent with the subcellular fractionation data obtained with the full-length virus (Fig. 6B) and suggested that mutations in the DI/DIII linker did not affect the association of E and prM with the ER-derived membranes.

Next, we analyzed the release of VLPs to the supernatants. VLPs were concentrated from the supernatants by PEG precipitation and purified by ultracentrifugation on a sucrose cushion. Secretion of VLPs was quantified by monitoring the presence of E protein by Western blotting (Fig. 7C). The E-Y299F mutation resulted in a strong decrease in VLP secretion, which mirrored what was observed for the infectious virus (compare Fig. 3B and Fig. 7C). Interestingly, VLP secretion did not correlate with the steady-state expression level of E, since we observed that intracellular E levels were the same for both wild-type and DI/DIII mutant VLPs (Fig. 7C, top). This suggested that the decreased particle secretion of the DI/DIII linker mutants cannot be attributed to differences in the intracellular abundance of E. More importantly, this showed that the defect in viral assembly was not solely linked to a lack of nucleocapsid incorporation into virions.

Mutations in the DI/DIII linker of E do not affect its interaction with prM or the trafficking and maturation of viral particles.

Although the mutations in the DI/DIII linker did not appear to have a drastic effect on the subcellular localization and accumulation of structural proteins within the cells (Fig. 6 and 7), we wanted to confirm that the mutant E proteins were able to traffic through the secretory pathway. To test this, we collected cell lysates at 2 days postelectroporation with DV2, DV2(E-G296A), DV2(E-S298A), or DV2(E-Y299F) RNA and treated them with endo-H, an enzyme that removes immature high-mannose glycans from proteins. These sugars are added to proteins in the ER and are trimmed and further modified in the Golgi, conferring resistance to endo-H. Sensitivity to endo-H is therefore indicative of proteins that have not advanced beyond the ER in the secretory pathway (9). As shown in Fig. 8A, wild-type E was distributed in roughly equivalent proportions between the Golgi (endo-H resistant) and ER (endo-H sensitive). In the case of the DI/DIII linker mutants, this distribution was shifted toward a greater proportion of protein in the ER, although a significant fraction was found associated with the Golgi (Fig. 8A). Since DV E protein trafficking through the secretory pathway is expected to reflect trafficking of immature virions, we reasoned that the lower accumulation of the E mutants in the Golgi should reflect a reduced steady-state accumulation of immature mutant virions in the cells.

Fig 8.

(A) Trafficking of wild-type and mutant E proteins through the secretory pathway was analyzed by endo-H assay of cell lysates 2 days following the electroporation of Huh7 cells with in vitro-transcribed DV2, DV2(E-G296A), DV2(E-S298F), or DV2(E-Y299F) genomic RNAs. Cell lysates were treated with endo-H and then analyzed by Western blotting with anti-DV2 E antibody. A representative experiment out of 3 repeats is shown. In this assay, the pool of E protein resistant to endo-H processing transited through the Golgi apparatus (upper band), whereas the pool of E protein sensitive to endo-H processing was associated with the ER (lower band). Note that due to the decreased steady-state abundance of E proteins bearing mutations in the DI/DIII linker, the exposure time utilized for their detection in the E Western blot was longer than that utilized for the wild-type virus. Relative to the wild-type virus, the DI/DIII linker mutants exhibit a decreased ratio of E in the Golgi versus the ER, suggesting decreased trafficking through the Golgi but no absolute block in this step of viral trafficking. (B and C) To verify that the mutant E-Y299F protein heterodimerizes with prM, Huh7 cells were transfected with D2.VLP and D2(E-Y299F).VLP plasmids (DV2 and Y299F, respectively) and then analyzed on day 2 posttransfection for the presence of the prM-E heterodimer. (B) The coverslips were fixed and analyzed by immunofluorescence using antibodies specific for E and for the prM-E heterodimer (19). A representative experiment out of 2 repeats is shown. The images were taken at ×200 magnification. (C) The steady-state expression of prM and E proteins was monitored in the cell lysates, using GAPDH as a loading control (top). The cell lysates were immunoprecipitated (IP) using an antibody specific for DV E protein. The immunoprecipitates (bottom) were analyzed by Western blotting to confirm that the DV2 E protein had been captured by immunoprecipitation. The DV2 prM protein was also detected in the immunoprecipitates whereas GAPDH was undetectable, showing that the prM coimmunoprecipitation was specific. The levels of IgG in the immunoprecipitation reactions served as a loading control for the immunoprecipitation experiment. The experiment was performed twice with biological duplicates. A representative experiment is shown. In both analyses presented in B and C, the mutant E-Y299F protein was found to associate with prM as efficiently as the wild-type E. (D) To assess whether the E-Y299F mutation affects the maturation of viral particles by preventing the association of the pr peptide with E, LoVo cells, which lack functional furin, were transfected with D2.VLP and D2(E-Y299F).VLP plasmids (DV2 and Y299F, respectively). At 2 days posttransfection, Western blot analysis was utilized to monitor the steady-state expression of E in the cells (using GAPDH as a loading control) and the secretion of immature VLPs to the culture supernatants [E (sup)]. The DV2(E-Y299F) mutant exhibited a strong reduction in immature VLP produced relative to wild-type DV2, indicating that prevention of the processing of prM-E to pr-E does not rescue the phenotype of this mutant. A representative experiment out of 3 repeats is shown.

One explanation for this phenotype might be that the E and prM proteins fail to coordinate the assembly of immature virions at the ER membrane. Since E and prM proteins were not affected in their association with ER membranes, we asked whether the interaction of prM with E was affected by mutations in the DI/DIII linker. It is known that prM/E heterodimerization is important in flavivirus assembly, and mutations that prevent this association have been shown to reduce the assembly of viral particles (3, 26). We used the antibody 2E11, which specifically recognizes DV2 prM/E heterodimers, to address this (19). We transfected Huh7 cells with pcDNA3.1-D2.VLP or pcDNA3.1-D2(E-299F).VLP plasmids and monitored E and prM/E heterodimer expression by immunofluorescence at 2 days posttransfection (Fig. 8B). In both the wild-type and the mutant transfections, we observed that most cells expressing the E protein were positive for the prM/E heterodimer. To further confirm that the DI/DIII linker mutations do not impair the interaction between prM and E, we examined the association of E and prM proteins in lysates obtained from Huh7 cells transfected with pcDNA3.1-D2.VLP or pcDNA3.1-D2(E-299F).VLP. We used an anti-DV2 E monoclonal antibody to pull down DV2 E proteins and then probed the immunoprecipitates for the presence of E and prM proteins (Fig. 8C). As expected, prM was abundantly detected in the wild-type E pull-down (Fig. 8C, DV2 lane). When we analyzed the cells transfected with pcDNA3.1-D2(E-299F).VLP, the ratio of prM associated with E protein was similar to that observed for the wild-type transfection (Fig. 8C). This provides biochemical evidence that the prM/E heterodimerization was not affected by mutation of the DI/DIII linker and was not responsible for the lack of viral particle assembly within the cells.

Next, we hypothesized that the reduced steady-state accumulation of immature virions inside cells might be due to inappropriate or reduced interactions between E and pr, the peptide that remains associated with E at low pH following the furin-mediated cleavage of prM (22, 27). Failure of the DI/DIII linker mutants to bind to pr would be expected to result in premature triggering of E's fusogenic activity in the acidic environment of the trans-Golgi (29). Conversely, prevention of the processing of prM to pr would be expected to rescue the phenotype of these mutants if their phenotype is due to faulty association of E with pr. To test this, we used LoVo cells, which lack functional furin (23) and which have been shown to produce immature DV virions (31). LoVo cells were transfected with pcDNA3.1-D2.VLP or pcDNA3.1-D2(E-Y299F).VLP plasmids, and 2 days posttransfection the supernatants were collected to permit the purification of VLPs. The accumulation of E protein within the cells or the supernatants was monitored by Western blotting (Fig. 8D). The E-Y299F mutation had little or no effect on the intracellular abundance of E but significantly decreased the immature viral particles reaching the culture supernatant (Fig. 8D), essentially mirroring the effect of this mutation on the production of mature virions (Fig. 7C). Together, these data suggest that mutations in the DI/DIII linker lead to a defect upstream of trafficking through the secretory pathway and implicate these residues in the assembly of viral particles at the ER membranes.

A mutation in the stem region of DV2 E protein compensates for a mutation in the DI/DIII linker.

As mentioned earlier, by day 2 postelectroporation, we were able to detect a very low titer of infectious virus released from cells transfected with DV2(E-Y299F) genomic RNA (Fig. 3C). In order to assess whether this infectious virus had retained the E-Y299F mutation, we amplified the virus in Huh7 cells by harvesting, titering, and reinfecting new cells every 2 days. After 5 rounds of amplification, the yield of virus was sufficient to permit the extraction of viral RNA and sequence analysis. Interestingly, the E-Y299F mutation was conserved, but we observed that nucleotide A at position 2136 had been mutated to C or T, resulting in a missense mutation at position 400 in the E protein from glutamine to histidine (Fig. 9A). This residue is located in the stem region of the E protein, in helix 1, and interestingly all Flaviviridae but DV2 conserve a basic lysine or arginine residue at this position (Fig. 9B).

To confirm that E-Q400H is a compensatory mutation, we first introduced it into the pcDNA3.1-D2.VLP and pcDNA3.1-D2(E-Y299F).VLP plasmids. As expected, the intracellular abundance of E was unaffected by the mutations (Fig. 9C, top). Next, we purified VLPs released in the supernatants as described above, and we analyzed the VLP production by Western blotting (Fig. 9C, bottom). Interestingly, the E-Q400H mutation compensated for the presence of the E-Y299F mutation and restored VLP production to that of the wild type. It was also noteworthy that the introduction of the E-Q400H mutation alone seemed to increase VLP release to the supernatants (Fig. 9C).

Next, we introduced the E-Q400H mutation into the pRS-D2 and pRS-D2(E-Y299F) plasmids, as described above, to permit characterization of this mutation in the context of the assembly of authentic virions. The E-Q400H single mutant expressed E and C at wild-type levels, demonstrating that this mutation alone does not have a demonstrable effect on the steady-state abundance of these structural proteins (Fig. 9D). Interestingly, however, when introduced in the context of the E-Y299F mutation, the E-Q400H mutation appeared to restore protein expression to near-wild-type levels (Fig. 9D). Analysis of supernatants by focus-forming assay (Fig. 9E) showed that DV2(E-Y299F/Q400H) and DV2(Q400H) mutants produced titers of infectious virus that were comparable to the those of the wild type, whereas no infectious DV2(E-Y299F) virus was detected. Together, these data demonstrate that the E-Q400H mutation suppresses the assembly defect of the E-Y299F mutation on steady-state expression of the structural proteins as well as its effect on infectious particle secretion.

DISCUSSION

Mutagenesis of the DV2 E DI/DIII linker impairs assembly of viral particles.

In this study, we have demonstrated that conservative mutations in the DI/DIII linker of the DV2 E protein lead to severe reductions in virion production. We were able to show that these mutations abolished the formation of viral particles within the cells (Fig. 6A), and we observed a decreased accumulation of E protein in the post-Golgi compartments of the secretory pathway (Fig. 8A). These observations could indicate either that there were no formation of viral particles or that the assembled viral particles were not stable and therefore unable to accumulate in cells after budding from their sites of assembly. Recent studies demonstrated that improper interaction of E with prM can lead to a defect in flavivirus assembly (26, 29). Notably, Zheng and colleagues determined that the mutation E-H244A in DV2 abolished the prM/E interaction and led to the premature fusion of viral particles during transit through the secretory pathway (29). In the case of the DI/DIII linker mutants examined in this study, we were able to show that the prM/E heterodimer was formed in the cells (Fig. 8B and C). Also, when we used cells devoid of furin, we observed a decreased release of immature mutant virions (Fig. 8D). These findings supported the hypothesis that mutant virus particles were not efficiently assembled rather than being unstable or prematurely degraded during their traffic out of the transfected cells.

Immunofluorescence analysis of E and C proteins following electroporation of cells with genomic RNAs encoding wild-type virus and DI/DIII linker mutants indicated that these two structural proteins are expressed and colocalize in the cytoplasm on day 1 postelectroporation (Fig. 5A). At later times postelectroporation, wild-type E protein continued to colocalize with C at discrete sites that resemble sites of DV2 assembly (20) (Fig. 5B). In contrast, E proteins bearing mutations in the DI/DIII linker were no longer detected in many cells still positive for C; moreover, neither E nor C proteins were found to localize in sites resembling those where DV2 assembly occurs for the DI/DIII linker mutants (Fig. 5B). Biochemical fractionation experiments confirmed that despite the association of prM and E with the ER membrane, the association of C protein with mutant E proteins in the gradient was significantly reduced, and in no case was it associated with the appearance of infectious virions (Fig. 6B). Although mutations in the DI/DIII linker may affect the budding of nucleocapsids into prM/E-decorated membranes, we observed that production of VLPs is also impaired in the presence of the E-Y299F mutation (Fig. 7). Since VLPs lack the nucleocapsid, we conclude that DV2 assembly must be affected at a step independent of this interaction. One possibility is that E and prM fail to assemble into a viral particle. An alternate hypothesis would be that there is a defect in the formation of assembly factories at the ER membranes and that the formation of these complexes is coordinated by interactions with the DI/DIII linker.

DV2 particle formation is not limited by steady-state expression of E.

Mutations in the DI/DIII linker might affect DV2 particle assembly simply by promoting degradation of E such that the intracellular supply of E limits particle formation. The mutations we introduced to the DI/DIII linker were not expected to disrupt the structure or stability of the E protein, since the amino acid substitutions were the most conservative changes possible and the structure of the DI/DIII linker is poorly ordered in the prefusion structure (16). It was therefore surprising to us that E proteins bearing mutations in the DI/DIII linker disappeared over time, whereas the C protein persisted in the same cells (Fig. 5B). Decreased availability of E does not appear, however, to explain the assembly defect observed for the DI/DIII mutants, since steady-state expression levels of wild-type and Y299F E proteins appeared to be equivalent in the VLP experiments (Fig. 7C), yet the yield of viral particles produced in the culture supernatants varied drastically (Fig. 7C). This lack of correlation between steady-state expression of E and VLP yield suggests that viral particle production was not limited by degradation of E-Y299F and supports the idea that the DI/DIII linker mutations do not impair protein stability.

The role of the DI/DIII linker in related viruses.

Our conclusion that the DI/DIII linker of E is important for DV2 assembly is consistent with other studies in which mutations in the corresponding region of other class II viral envelope proteins have also been reported to affect viral assembly. In the E2 protein of the closely related hepatitis C virus (HCV), the IgVR structurally resembles the DV DI/DIII linker and has been shown to be essential for HCV assembly (1, 14). More recently, a study of the DI/DIII linker of the Semliki Forest virus (SFV) E1 protein reported that a phenylalanine-to-alanine mutation at position 287 caused a defect in assembly (30). The authors suggested that residue F287 is part of a network of interactions that connect the DI/DIII linker with domain I and that it could be important for viral assembly. These findings support the idea that the DI/DIII linker is critical for viral assembly across the alpha- and flaviviruses, although the structural details of how the DI/DIII linker functions in the assembly process may differ between alpha- and flaviviruses, since the structure of the DV E protein in its prefusion conformation (15) revealed no interaction of the linker with another domain.

Interestingly, although our experiments suggest that mutations in the DV2 DI/DIII linker do not affect viral entry (Fig. 4), Zheng and coworkers found that residues in the SFV DI/DIII linker contribute to the network of interactions that stabilize the E1 postfusion trimer (30). Most notably, the defect in viral fusion caused by mutation of the highly conserved H3 residue in the SFV E1 protein was rescued by a D284A mutation in the SFV E1 DI/DIII linker (30); these authors also showed that another mutation in the E1 DI/DIII linker, R289A, could affect the fusion step of SFV entry in the absence of the E1-H3A mutation (30). In contrast, the DI/DIII linker was shown by the Kielian group not to affect the trimerization of the DV2 DI/DII protein in biochemical assays (12). Thus, although the data supporting a network of interactions implicating the E1 DI/DIII linker in E1 trimerization and SFV fusion are strong, the existence of an analogous interaction network in the DV E protein remains to be determined. We note that Q400H, the compensatory mutation for the DV2(E-Y299F) virus, is located in a region of the E protein stem that was previously shown to be important in trimerization of the tick-borne encephalitis (TBE) virus E protein (3). This suggests that an interaction network may exist and play a role in DV entry, although the molecular details of these interactions and their function in viral entry and assembly may differ within and perhaps among different groups of alpha- and flaviviruses.

A mutation in the E protein stem region fully compensates for the E-Y299F mutation.

We found that the E-Q400H mutation fully compensated for the defect in assembly of E-Y299F (Fig. 9). This compensatory mutation should be a useful tool in understanding the mechanism responsible for the assembly defect imposed by the DI/DIII mutations. Residue Q400 is located in the first helix of the DV2 E protein stem and appears to be buried in the virion lipid bilayer (28). Interestingly, this amino acid is not conserved among flaviviruses (Fig. 9) or even among the different serotypes of DV. Other residues in helix 1 of the DV E stem were previously found to have a role in entry and assembly (13). In the work by Lin et al., several mutations in helix 1 of the DV4 E (such as I398P and T405P) greatly affected the assembly of VLPs, presumably by preventing their release from the membrane to the lumen of the ER (13). This role of helix 1 would fit well with our current hypothesis that the DI/DIII linker mutations prevent the formation of assembly complexes at the ER membrane and would explain how the Q400H mutation could compensate for such a defect. It is interesting to note that the sole introduction of E-Q400H led to a marked increase in the release of viral particles in the context of infectious virus (Fig. 9E) and VLPs (Fig. 9C). Additional experiments are needed to determine whether this mutation enhances viral particle assembly via a general mechanism versus via a mechanism that specifically suppresses the assembly defect of E-Y299F.

In conclusion, our study demonstrates that mutations in the flexible linker between domains I and III of the DV2 E protein can significantly affect the production of DV2 virions and VLPs. In contrast to what has previously been reported for the prototype alphavirus SFV, mutations in this linker did not appear to affect viral entry, which fits well with the demonstration that the DV2 DI/DIII linker is not necessary for E protein trimerization in vitro (12). While these mutations do not appear to affect the interaction of prM and E, they do appear to affect the assembly of particles in which core is associated with E and prM in both immunofluorescence and biochemical fractionation experiments. Combined with the additional observation that VLP production was not limited by steady-state expression levels of E in our experiments, these data are collectively consistent with a model in which mutations in the DI/DIII linker affect budding of viral particles from the ER membrane. The identification of the DV2 E-Q400H mutation as a suppressor of the assembly defect caused by mutation of the DI/DIII linker is likewise consistent with this model. Given the unstructured appearance of the DI/DIII linker in the high-resolution crystal structures of E, the structural role of this region of the protein in viral assembly is not obvious. Selection of additional second-site mutations that suppress the assembly defect caused by mutation of the DI/DIII linker may provide an alternative approach to elucidate the molecular interactions underlying these phenomena.