The SD1 Subdomain of Venezuelan Equine Encephalitis Virus Capsid Protein Plays a Critical Role in Nucleocapsid and Particle Assembly

ABSTRACT

Venezuelan equine encephalitis virus (VEEV) is an important human and animal pathogen, for which no safe and efficient vaccines or therapeutic means have been developed. Viral particle assembly and budding processes represent potential targets for therapeutic intervention. However, our understanding of the mechanistic process of VEEV assembly, RNA encapsidation, and the roles of different capsid-specific domains in these events remain to be described. The results of this new study demonstrate that the very amino-terminal VEEV capsid-specific subdomain SD1 is a critical player in the particle assembly process. It functions in a virus-specific mode, and its deletion, mutation, or replacement by the same subdomain derived from other alphaviruses has strong negative effects on infectious virus release. VEEV variants with mutated SD1 accumulate adaptive mutations in both SD1 and SD2, which result in a more efficiently replicating phenotype. Moreover, efficient nucleocapsid and particle assembly proceeds only when the two subdomains, SD1 and SD2, are derived from the same alphavirus. These two subdomains together appear to form the central core of VEEV nucleocapsids, and their interaction is one of the driving forces of virion assembly and budding. The similar domain structures of alphavirus capsid proteins suggest that this new knowledge can be applied to other alphaviruses.

IMPORTANCE Alphaviruses are a group of human and animal pathogens which cause periodic outbreaks of highly debilitating diseases. Despite significant progress made in understanding the overall structure of alphavirus and VEEV virions, and glycoprotein spikes in particular, the mechanistic process of nucleocapsid assembly, RNA encapsidation, and the roles of different capsid-specific domains in these processes remain to be described. Our new data demonstrate that the very amino-terminal subdomain of Venezuelan equine encephalitis virus capsid protein, SD1, plays a critical role in the nucleocapsid assembly. It functions synergistically with the following SD2 (helix I) and appears to form a core in the center of nucleocapsid. The core formation is one of the driving forces of alphavirus particle assembly.

INTRODUCTION

Venezuelan equine encephalitis virus (VEEV) is a representative member of the New World (NW) alphaviruses, which circulate mostly in Central and South America (1). In nature, VEEV is transmitted by mosquito vectors between vertebrate hosts, in which it causes a highly debilitating disease, often resulting in meningoencephalitis and a frequently lethal outcome (2). Among humans, the mortality rates are below 1%, mostly among the elderly and the very young, but the disease often leads to neurological sequelae (3, 4). Due to its user-friendly characteristics, VEEV has the potential for development as a biological warfare agent. It can be propagated to very high infectious titers in many commonly used cell lines, is very stable in lyophilized form, and is exceptionally efficiently transmitted by aerosol. Therefore, some of the VEEV serotypes are classified as category B select agents by the CDC and require restricted access and high biocontainment conditions. Despite a significant public health threat, no safe and efficient vaccines or therapeutic drugs have been developed against VEEV or other New World alphavirus infections.

Similar to other alphaviruses, the VEEV genome is a positive-sense RNA (5). It mimics cellular mRNAs in that it has a cap structure at the 5′ terminus and a poly(A) tail at the 3′ terminus. The genomic RNA is directly translated into the nonstructural (ns) polyprotein precursors P123 and P1234. Partially and completely processed ns polyproteins mediate replication of the viral genome and synthesis of the subgenomic RNA (6, 7). The latter RNA serves as a template for translation of the viral structural proteins, capsid, E2, and E1, which package viral genomes into released infectious virions (8).

In addition to being a component of infectious viral particles, one of the VEEV structural proteins, capsid protein, is also a critical factor in regulation of the development of the innate immune response. It blocks nuclear pore function (9), interferes with nucleocytoplasmic traffic (10), and ultimately inhibits cellular transcription (11). The detailed knowledge of the mechanism of this interference has already provided the means to generate noncytopathic alphavirus variants that are incapable of inhibiting the antiviral response and, most importantly, the type I interferon (IFN) response (12, 13). Capsid nuclear function is an attractive target for the design of antiviral drugs, which could increase the innate immune response during VEEV replication in vivo.

Interference with particle assembly is another possible avenue to explore in the development of VEEV-specific drugs which do not affect biological functions of the cell. In the last several years, significant progress has been made in understanding the overall structure of alphavirus and VEEV virions, and glycoprotein spikes in particular (14–16). Cryo-electron microscopy (cryo-EM) studies combined with crystallographic analysis of glycoprotein spikes have provided a detailed picture of the structure of the glycoprotein envelope and its interaction with nucleocapsid. However, the mechanistic process of virion assembly, RNA encapsidation, and the roles of different capsid-specific domains in these processes remain to be described.

In our previous studies, we have started investigation of the mechanism of VEEV virion formation with particular emphasis on RNA encapsidation and nucleocapsid (NC) assembly (17, 18). Our data demonstrated that assembly of VEEV particles is driven by the carboxy-terminal protease domain of its capsid protein. Deletion mutants that lack the entire amino-terminal domain are capable of efficient formation of RNA-free nucleocapsids, which are released from cells in virus-like particles (VLPs). These VLPs have a VEEV glycoprotein-containing lipid envelope and demonstrate the icosahedral structure typical of an alphavirus virion. Thus, the amino-terminal capsid domain is generally dispensable for VEEV particle release, but its subdomains (SD1 to SD4) orchestrate the integration of viral RNAs into mature viral particles. This RNA interaction with the amino-terminal domain strongly stabilizes virions. Moreover, the most positively charged peptide (SD3) inhibits nucleocapsid (NC) assembly, unless its charge is neutralized by interaction with RNA (17). Two other subdomains located upstream and downstream of SD3 (SD2 and SD4, respectively) appear to determine the specificity of interaction with viral genomic RNA, and mutations in these sequences lead to efficient nonspecific encapsidation of other RNAs (19, 20). So far, the function(s) of the most amino-terminal subdomain (SD1) remains unclear. In the case of VEEV, SD1 is represented by only 33 amino acids (aa), but its deletion has a very strong negative effect on both infectious virus and virus particle release (17). In contrast to other SDs, deletion of SD1 could not be compensated for by acquiring adaptive mutations in other fragments of capsid protein or nsP2 (17, 21). VEEV-specific SD1 contains only a very few positively charged amino acids and so far has not been predicted to have a defined secondary structure. Thus, the available data did not allow for reasonable speculation as to its function.

This new study was focused on furthering the understanding of VEEV capsid protein SD1-specific function in virion assembly. Our new data demonstrate that this small subdomain plays a critical role in the particle formation process. It functions in a virus-specific mode, and its replacement either with a heterologous amino acid sequence or by the same subdomain derived from heterologous alphaviruses has deleterious effects on infectious virus release. Other data suggest that more efficient NC and particle assembly proceeds when the following subdomain, SD2, is homologous to SD1. These two subdomains together appear to form the central core of VEEV NC. Thus, the integrity of SD1 is of critical importance in alphavirus particle formation.

MATERIALS AND METHODS

Cell cultures.

The BHK-21 cells were kindly provided by Paul Olivo (Washington University, St. Louis, MO). They were maintained at 37°C in alpha minimum essential medium (αMEM) supplemented with 10% fetal bovine serum (FBS) and vitamins.

Plasmid constructs.

Standard recombinant DNA techniques were applied for plasmid construction. The PCR fragments containing deletions or mutations in the capsid gene were cloned into pVEEV/GFP to replace the wild-type (wt) capsid sequence. pVEEV/GFP, containing the genome of VEEV TC-83, is described elsewhere (12). All of the modifications are presented in the corresponding figures. In all of the plasmids, cDNAs of modified VEEV genomes were placed under the control of the SP6 promoter. They also contained a green fluorescent protein (GFP) gene under the control of a duplicated subgenomic promoter. GFP expression was used to monitor virus replication and to measure titers, because many of the constructs used were unable to develop plaques or to cause a cytopathic effect (CPE). Helper RNA-containing plasmids, which encoded VEEV capsid protein capable of efficient RNA packaging, have been described elsewhere (18, 22). Sequences of the recombinant genomes can be provided upon request.

In vitro RNA transcription and electroporation.

Plasmids were purified by centrifugation in CsCl gradients. Before the transcription reaction, they were linearized with MluI. RNAs were synthesized by SP6 RNA polymerase in the presence of a cap analog according to the manufacturer's recommendations (Invitrogen). The yield and integrity of the transcripts were analyzed by gel electrophoresis under nondenaturing conditions. Aliquots of transcription reactions were used for electroporation without additional purification. Electroporation of BHK-21 cells was performed under previously described conditions (23). Some of the rescued mutant viruses were unable to replicate to infectious titers sufficient to infect cells in the subsequent experiments using a multiplicity of infection (MOI) of >1 infectious unit (IU)/cell. Therefore, such defective genomes were packaged into infectious virions to titers above 10⁹ IU/ml by coelectroporating their in vitro-synthesized RNA and helper RNA genomes. Viral stocks were harvested within 24 h postelectroporation. Titers of plaque-forming viruses were determined by a standard plaque assay on BHK-21 cells (24). The infectious titers of noncytopathic viruses were determined by infecting BHK-21 cells (5 × 10⁵ cells/well) in six-well Costar plates with 10-fold dilutions of the samples and counting the numbers of GFP-positive cells at 6 h postinfection.

Selection of variants replicating to higher infectious titers.

The original stocks of VEEV variants encoding mutated SD1 domains were passaged 5 times (for VEEV/chikvSD1/GFP and VEEV/eeevSD1/GFP) or 6 times (for VEEV/randSD1/GFP) on BHK-21 cells. To select the most efficiently replicating, evolved variants, at each passage, we infected cells using a 10-fold-reduced volume of virus harvested from the previous passage. Stocks demonstrating a more efficiently spreading infection were titrated on BHK-21 cells to confirm evolution to a large-plaque-forming phenotype. For each variant, several plaque-purified isolates were sequenced using VEEV-specific primers. The identified putative adaptive mutations were cloned back into original constructs, and rescued viruses were tested in terms of rates of infectious virus release, titers, and plaque size.

Analysis of the structural proteins produced in infected cells.

BHK-21 cells in 6-well Costar plates were infected at an MOI of 10 IU/cell with different SD1 mutants for 1 h at 37°C. At 8 h postinfection, cells were harvested and the accumulation of structural viral proteins in each sample was analyzed by Western blotting using polyclonal murine VEEV-specific antibodies (Abs) (generous gift from R. Tesh, University of Texas Medical Branch at Galveston), followed by Alexa Fluor 800-labeled secondary antibodies (Invitrogen). Fluorescence intensities were analyzed on an Odyssey imager (Li-Cor Biosciences) to determine the relative quantities of p62 and capsid protein in each sample. Data were normalized to the level of tubulin or β-actin and the level of viral structural proteins detected in the cells infected with parental VEEV/GFP.

Concentration and analysis of the released viral particles.

BHK-21 cells in 6-well Costar plates were infected at an MOI of 5 IU/cell with different SD1 mutants and evolved variants for 1 h at 37°C. Then, the inocula were replaced with complete medium. After 30 min of incubation at 37°C, cells were washed three times with phosphate-buffered saline (PBS) and further incubated in virus production serum-free (VP-SF) medium (Invitrogen). At 15 h postinfection, before the development of CPE, media were collected and clarified by centrifugation at 15,000 × g for 15 min. Viral particles were pelleted by ultracentrifugation at 54,000 rpm in a TLA-55 rotor (Beckman) through 20% sucrose for 1 h at 4°C. An amount corresponding to 1/20 of the whole-sample volume was analyzed by Western blotting using murine polyclonal VEEV-specific antibodies and Alexa Fluor 800-labeled secondary antibodies. Fluorescence intensities were analyzed and quantified on a Li-Cor imager to determine the relative quantities of E2 and capsid proteins in each sample. Data were normalized to the level of the proteins in particles released from VEEV/GFP-infected cells and purified under identical experimental conditions.

Electron microscopy.

BHK-21 cells were infected with VEEV/GFP or VEEV/SD1-GFP-Cc at an MOI of 20 IU/cell. At 6 h postinfection, cells were washed 4 times with serum-free medium and further incubated in VP-SF medium for 16 h. Supernatants were harvested and clarified by centrifugation at 5,000 × g for 15 min, and particles were concentrated to ∼1 ml using Amicon Ultra 100K centrifugal filters (Millipore). Viral particles and VLPs were further purified by ultracentrifugation in discontinuous sucrose gradients (15%/40%) at 35,000 rpm in a SW-55 rotor (Beckman) for 1 h at 4°C. Bands containing virus particles were identified between the sucrose layers and used for EM analysis. Droplets of virus suspension were placed onto freshly glow-discharged, 400-mesh copper electron microscope grids covered with carbon film. Grids were blotted dry, washed with water, and stained with 1% uranyl acetate. Images were acquired at ×80,000 magnification using an FEI Tecnai F20 electron microscope in the UAB cryo-EM facility.

Confocal microscopy.

BHK-21 cells were seeded in 8-well Ibidi chambers (5 × 10³/well) and incubated overnight at 37°C. Cells were then infected with different VEEV variants at the same MOI. Inocula were removed and replaced with fresh medium, and cells were further incubated at 37°C. At the time postinfection indicated in the figure legends, cells were fixed with 4% paraformaldehyde (PFA) for 15 min, permeabilized with 0.5% Triton X-100 (TX-100) in PBS, and rinsed again with PBS, and nuclei were stained with Hoechst 34580 (Invitrogen). Three-dimensional (3D) stacks were acquired on a Zeiss LSM700 confocal microscope with a 63× 1.4-numerical-aperture (NA) Planapochromat oil objective. The image stacks were deconvolved using Huygens software (Scientific Volume Imaging), and images were assembled in Imaris (Bitplane AG).

RESULTS

Deletion and modification of SD1 affect the release of infectious VEEV.

Our previous study strongly suggested that the very amino-terminal fragment of VEEV capsid protein, SD1, located upstream of helix I, plays an important role in NC formation and in the release of infectious virions (17). To continue investigation of SD1's function, we designed new VEEV variants with modified SD1 sequences (Fig. 1A). One of the mutants, VEEV/R^-SD1/GFP, had all four SD1-specific, positively charged amino acids replaced by those containing no charge. In VEEV/randSD1/GFP, the amino acid sequence of the SD1 domain was randomized. This peptide had the same amino acid content, but the amino acids were placed in a different order. The control VEEV/ΔSD1/GFP variant (17) had the entire SD1 deleted. These SD1 mutants encoded GFP under the control of the duplicated subgenomic promoter to simplify the titration of harvested viruses, because many of the mutants developed in this study were expected to form very small plaques, if any at all. GFP was also required for monitoring the infection spread in cell cultures.

FIG 1.

The SD1 domain of VEEV capsid protein is important for infectious virus release. (A) Schematic representations of the VEEV genome and the domain structure of the capsid protein, with sequence alignments of the SD1 and SD2 capsid domains from wt VEEV and the designed mutants. (B) BHK-21 cells were infected with VEEV/GFP, VEEV/R^-SD1/GFP, VEEV/randSD1/GFP, or VEEV/ΔSD1/GFP at an MOI of 0.01 IU/cell. Media were replaced at the indicated time points, and titers were determined by standard plaque assay on BHK-21 cells. (C) Plaques formed on BHK-21 cells by VEEV/GFP, VEEV/randSD1/GFP, and VEEV/ΔSD1/GFP. Viral stocks were titrated on BHK-21 cells by standard plaque assay and were stained with crystal violet after 2 days of incubation at 37°C. The presented data are representative of three experimental repeats, which generated very similar results.

The newly designed variants were viable, but VEEV/ΔSD1/GFP and VEEV/randSD1/GFP stocks harvested after electroporation of the in vitro-synthesized RNA demonstrated very low infectious titers. Therefore, high-titer stocks were generated by coelectroporation of the in vitro-synthesized viral and helper RNAs. The helper RNA provided wt capsid protein in trans, allowing efficient encapsidation of the defective genomes. The helper RNA itself was packaged into released virions very inefficiently (18). This approach produced infectious titers of approximately 10⁹ IU/ml for both VEEV/ΔSD1/GFP and VEEV/randSD1/GFP.

Next, cells were infected at the same MOI with the designed variants, and the rates of infectious virus release were analyzed. Mutation of the positively charged amino acids produced a very small but reproducible negative effect on the infectious titers (Fig. 1B) compared to the original VEEV/GFP variant. In contrast, randomization of the SD1 sequence, which retained all of the amino acids but altered their relative positions in SD1, led to a 3- to 4-order-of-magnitude decrease of virus titers that was observed consistently throughout the experiments. This negative effect was not as pronounced as that observed in the SD1 deletion mutant, suggesting that capsid protein with a strongly modified SD1 sequence still functioned better in infectious virus release than the mutated variant lacking SD1 at all. Both VEEV/ΔSD1/GFP and VEEV/randSD1/GFP developed dramatically smaller, pinpoint plaques on BHK-21 cells (Fig. 1C).

Taken together, the data additionally demonstrate the importance of SD1 at the amino terminus of VEEV capsid for infectious virus release. Positively charged amino acids appear to play only a minor role in this process, but the overall sequence of SD1 appears to be critical for its function.

SD1 sequences derived from heterologous alphaviruses have negative effects on the infectious titers of VEEV.

Deletion and randomization of SD1 resulted in dramatic changes in capsid protein sequence, and their effects on the mechanism of NC assembly were not easy to interpret. Therefore, to further analyze the activity of SD1, we replaced this fragment in VEEV/GFP with SD1 sequences derived from heterologous alphaviruses: chikungunya virus (CHIKV), the distantly related Old World alphavirus, and the North American strain of eastern equine encephalitis virus (EEEV), the more closely related New World alphavirus. SD1 sequences cloned into VEEV/chikvSD1/GFP and VEEV/eeevSD1/GFP and their alignments with VEEV SD1 are presented in Fig. 2A. Both of them strongly differ from VEEV SD1, with CHIKV SD1 demonstrating a particularly low level of identity.

FIG 2.

SD1 sequences derived from heterologous alphaviruses have negative effects on VEEV capsid's function in infectious virus release. (A) Schematic representations of the VEEV genome and the domain structure of the capsid protein, with sequence alignments of the SD1 and SD2 capsid domains in VEEV/GFP, VEEV/eeevSD1/GFP, and VEEV/chikvSD1/GFP. (B) BHK-21 cells were infected with VEEV/GFP, VEEV/eeevSD1/GFP, or VEEV/chikvSD1/GFP at an MOI of 0.01 IU/cell. Media were replaced at the indicated time points, and titers were determined by standard plaque assay on BHK-21 cells. The experiments were repeated twice at different MOIs and produced similar results.

The in vitro-synthesized RNAs of VEEV/chikvSD1/GFP and VEEV/eeevSD1/GFP were coelectroporated with the VEEV capsid-encoding helper RNA to efficiently package mutated genomes into infectious virions. Infection of naive BHK-21 cells with electroporation-derived stocks showed that the newly designed mutants replicated less efficiently than parental VEEV/GFP (Fig. 2B). The rates of infectious VEEV/chikvSD1/GFP virus release were similar to those of VEEV/randSD1/GFP (Fig. 1B and 2B). This indicated that heterologous CHIKV SD1 functions as inefficiently as a randomized SD1 containing the VEEV-specific amino acids but in the wrong order (Fig. 1). VEEV/eeevSD1/GFP exhibited higher replication rates than VEEV/chikvSD1/GFP (Fig. 2B), probably due to a higher sequence similarity between EEEV SD1 and VEEV SD1. However, its replication remained less efficient than that of the parental VEEV/GFP.

As the introduced modifications of SD1 could also have an impact on the amount of capsid protein production during infection, we analyzed the accumulation of mutated VEEV capsid proteins and glycoproteins in the infected cells (Fig. 3A). The small differences detected were insufficient to explain strong decreases in virus titers.

FIG 3.

VEEV SD1 mutants produce similar levels of structural proteins but release viral particles less efficiently than VEEV/GFP. (A) Analysis of viral protein expression in BHK-21 cells infected with VEEV/GFP and indicated SD1 mutants. Cells were infected at an MOI of 10 IU/cell. They were harvested at 8 h postinfection, and accumulation of structural viral proteins was analyzed by Western blotting using VEEV-specific antibodies, as described in Materials and Methods. Relative levels of p62 (E3+E2) protein in each sample are indicated. Data were normalized to tubulin levels and to the level of protein detected in VEEV/GFP-infected cells. Numbers at left are molecular masses in kilodaltons. (B) BHK-21 cells were infected at an MOI of 5 IU/cell and then incubated at 37°C in serum-free medium as described in Materials and Methods. Particles were collected at 15 h postinfection and concentrated by ultracentrifugation as described in Materials and Methods. Samples were analyzed by Western blotting using VEEV-specific antibodies. Relative quantities of E2+E1 proteins in each sample are indicated. Data were normalized to the level of proteins in the particles released from VEEV/GFP-infected cells.

Our previously published data have demonstrated that modifications in SD1 affect PFU/genome equivalent (GE) ratio (17). This negative effect, which characterizes particle infectivity, is relatively small, within 4-fold, and does not explain infectious titers a few orders of magnitude lower. Therefore, to show that extensive changes in SD1 had also a strong negative effect on particle release, we directly analyzed their accumulation in the medium (Fig. 3B). The lower infectious titers detected through plaque assays strongly correlated with the inefficient accumulation of extracellular viral particles.

Thus, taken together, the data suggested that the replacement of VEEV SD1 with heterologous or synthetic versions or the deletion of SD1 affected viral particle formation but not the production of structural proteins. Moreover, since the CHIKV- and EEEV-specific SD1s were functional in their native context, but not in VEEV capsid, this suggested that alphavirus SD1 may function cooperatively with other capsid subdomains in NC formation.

VEEV SD1 mutants evolve to a more efficiently replicating phenotype.

One of the characteristic features of alphaviruses is their extraordinary ability for evolution toward a more efficiently replicating phenotype. The identification of adaptive mutations, which increase virus replication to higher infectious titers, is an important means of understanding protein function and interactions between proteins or domains of the same protein. Thus, in the subsequent experiments, we assessed the ability of VEEV variants with heterologous SD1s to evolve to higher infectious titers. To select the most efficiently replicating variants, a number of passages of VEEV/ΔSD1/GFP, VEEV/chikvSD1/GFP, VEEV/eeevSD1/GFP, and VEEV/randSD1/GFP were performed in BHK-21 cells using lower volumes of medium for infection at each subsequent passage. For all viruses, except VEEV/ΔSD1/GFP, after five or six passages, plaques that formed on BHK-21 cells became larger.

Capsid genes of the variants isolated from randomly selected plaques were sequenced. As expected, no mutations were identified in VEEV/ΔSD1/GFP. VEEV/chikvSD1/GFP contained two mutations in all plaque-purified variants. One was located in the CHIKV-specific SD1, and the second was found in the VEEV SD2 (helix I) sequence (Fig. 4A). Plaque-isolated variants of VEEV/eeevSD1/GFP contained a single mutation in either SD1 or SD2 (Fig. 4B). VEEV/randSD1/GFP was the least capable variant in acquiring adaptive mutations. During serial passaging, its large plaque-forming variant appeared noticeably later than others, after 6 passages. The reason for its slower adaptation was that in this case, a greater change in the capsid-coding sequence was required (Fig. 4C): a duplication of a 7-aa-long SD2 fragment, which most likely extended the α-helix.

FIG 4.

VEEV/GFP SD1 mutants accumulate adaptive mutations in SD1- and SD2-coding sequences. The originally designed SD1 mutants were passaged several times on BHK-21 cells as described in Materials and Methods. Evolved variants were isolated by plaque purification, and the capsid-coding sequences were analyzed. The identified mutations are indicated under sequence alignments in each panel. Duplicated sequence in adapted VEEV/randSD1/GFP is indicated by green and blue font colors (C). The identified mutations were introduced into the original constructs, and rescued viruses were analyzed in terms of replication rates and sizes of plaques formed on BHK-21 cells. BHK-21 cells were infected at an MOI of 0.01 IU/cell. Media were replaced at the indicated time points. Titers and plaque sizes were determined by standard plaque assay on BHK-21 cells as described in Materials and Methods. All of the plaques were stained with crystal violet after 2 days of incubation at 37°C.

To confirm the positive effects of the mutations, they were transferred into the originally designed variants. The newly designed VEEV/chikvSD1ev/GFP, VEEV/eeevSD1ev1/GFP, VEEV/eeevSD1ev2/GFP, and VEEV/randSD1ev/GFP viruses demonstrated 10- to 100-fold-higher replication rates than the parental constructs (Fig. 4). At all times postinfection, titers of the adapted variants were closer to those of VEEV/GFP.

Thus, in all of the selected second-site mutants, adaptation was associated with additional direct modification of the amino-terminal fragment of capsid protein in either SD1 or SD2 domains. This was an indication that these two subdomains likely have synergistic or additive functions in formation of infectious virions. The more efficient particle release by cells infected with VEEV/randSD1ev/GFP (Fig. 3B) than by cells infected with the original VEEV/randSD1/GFP suggested that assembly of the virions rather than their infectivity was the reason for higher infectious titers. SD1s derived from heterologous alphaviruses appeared to be more easily adaptable to their new viral host than the randomized VEEV SD1 sequence, suggesting that they partially retained their function in the context of VEEV capsid.

SD1 function in virion assembly depends on the SD2 (helix I) sequence.

To further analyze the possible interaction of SD1 with other subdomains of capsid protein, we designed a set of variants encoding chimeric capsids, in which either the SD1+SD2, SD1+SD2+SD3, or SD1+SD2+SD3+SD4 subdomains in VEEV capsid protein were replaced by those derived from the capsid of the most distantly related CHIKV. The resulting VEEV/chikvSD1/GFP, VEEV/chikvSD12/GFP, VEEV/chikvSD123/GFP, and VEEV/chikvSD1234/GFP constructs are presented in Fig. 5A. wt VEEV SD2 also functions as a supraphysiological nuclear export signal (supraNES) (9, 10). In a previous study, its deletion led to accumulation of adaptive mutations in capsid protein, which caused inactivation of the downstream nuclear localization signal (NLS) and prevented accumulation of capsid in the nucleus (17). So far, there is no experimental evidence that CHIKV-specific SD2 can function as an NES. Therefore, to avoid accumulation of VEEV/chikSD12/GFP-specific capsid protein, which retains VEEV NLS, in the nucleus, we introduced additional point mutations into the SD3 NLS sequence.

FIG 5.

The presence of SD2, but not other subdomains, homologous to SD1 has a positive effect on infectious virus release. (A) Schematic representations of the domain structures of different chimeric capsid proteins, in which the subdomains SD1, SD2, SD3, and SD4 of VEEV were replaced with the corresponding CHIKV-specific SDs. (B) BHK-21 cells were infected with the indicated mutants at an MOI of 0.1 IU/cell. Media were replaced at the indicated time points, and virus titers were determined by standard plaque assay on BHK-21 cells. (C) Analysis of viral protein expression in BHK-21 cells infected with VEEV/GFP or the indicated capsid mutants. Cells were infected at an MOI of 10 IU/cell. Cells were harvested at 8 h postinfection. Accumulation of structural viral proteins was analyzed by Western blotting using VEEV-specific antibodies, as described in Materials and Methods. The relative quantities of p62 (E3+E2) protein in each sample are indicated. Data were normalized to the β-actin levels and to the level of protein detected in VEEV/GFP-infected cells.

All of these variants were viable, and analysis of their replication rates demonstrated that the presence of CHIKV SD2 in addition to SD1 in VEEV/chikvSD12/GFP had a readily detectable positive effect on virus replication rates (Fig. 5B). Further addition of CHIKV-specific SD3 did not cause a noticeable increase in the rates of VEEV/chikvSD123/GFP virus release. Furthermore, VEEV with all four CHIKV-derived capsid subdomains, VEEV/chikvSD1234/GFP, replicated less efficiently than the original VEEV/GFP or variants containing two or three CHIKV-derived, upstream-located subdomains (Fig. 5B). In the infected cells, the newly designed capsid proteins accumulated to similar levels (Fig. 5C), suggesting that the difference in infectious virus release was not a result of protein translation or stability.

The results of the previous study demonstrated that the sequence of SD3 had no effect on the efficiency of NC and virion formation, if this subdomain remained highly positively charged (17). This observation suggested a plausible explanation as to why replacement of VEEV SD3 with CHIKV SD3 in addition to the upstream-located capsid subdomain substitutions had no stimulatory effect on the rates of infectious virus release. The reduction in titers produced by SD4 replacement was likely the result of SD4's critical function in defining the specificity of capsid protein interaction with the virus-specific packaging signal (PS) and, thus, packaging of viral genomes into nucleocapsids (20, 25). In the viruses of VEEV and SFV serocomplexes, PSs are located in different fragments of the genomes (25, 26), and thus, CHIKV-specific SD4 probably bound inefficiently to the heterologous VEEV PS. Moreover, interactions of SD4 and other SDs with the carboxy-terminal, protease-containing domain are currently understudied, and the potential negative effect of heterologous sequences on virus release cannot be ruled out. Taken together, the data from these experiments additionally suggested that the function of VEEV SD1 in particle assembly is strongly dependent on the origin of the SD2 (helix I) but not on other following SDs.

SD1 plays a critical role in VEEV particle assembly at the plasma membrane.

To further demonstrate the critical role of SD1 in particle assembly and release, we applied another experimental system. A sequence encoding GFP was cloned into different positions of the capsid gene in the VEEV genome to replace the natural subdomains. VEEV/SD1-GFP-Cc and VEEV/SD12-GFP-Cc constructs encoded SD1 and SD1+SD2, respectively, followed by GFP and the carboxy-terminal domain of capsid (Cc) (Fig. 6A and 7A). VEEV/GFP-Cc had no SD1 and SD2 upstream of GFP, and VEErep/SD1-GFP-Cc replicon encoded SD1-GFP-Cc but had all of the glycoprotein genes deleted. We expected that insertion of this relatively long GFP into the VEEV capsid would cause a strong reduction in the efficiency of virion budding, but on the other hand, this could promote accumulation of assembling NCs in particular cellular compartments and make their visualization by 3D imaging possible. Thus, these constructs could help us to understand some critical steps of the VEEV budding process. GFP insertion was also useful for visualization purposes as the available VEEV capsid-specific antibodies (Abs) did not interact with formed NC and probably with partially assembled NC as well. Thus, their detection by immunostaining could be problematic.

FIG 6.

Cells infected with VEEV/SD1-GFP-Cc produce GFP-containing VLPs. (A) Schematic representation of the VEEV/SD1-GFP-Cc genome and SD1-GFP-Cc protein. (B) BHK-21 cells in 8-well Ibidi chambers were infected with VEEV/SD1-GFP-Cc at an MOI of 20 IU/cell. At 8 h postinfection, they were fixed and the 3D images were acquired by confocal microscopy. One of the representative images is shown. Bar, 10 μm. (C) The released VEEV/SD1-GFP-Cc-specific VLPs were concentrated as described in Materials and Methods, adsorbed to the glass slides, and visualized by confocal microscopy. (D) EM analysis of the negatively stained particles released from cells infected with VEEV/GFP or VEEV/SD1-GFP-capsid. Bars, 50 nm.

FIG 7.

SD1-containing GFP-capsid fusions, expressed by VEEV, accumulate at the plasma membrane. (A) Schematic representation of the recombinant viral and replicon genomes and the encoded GFP-capsid fusions. (B) BHK-21 cells in 8-well Ibidi chambers were infected with the indicated constructs at an MOI of 20 IU/cell. They were fixed at 6 h postinfection, and the 3D images were acquired on a confocal microscope. The top panel in each image shows a 1-μm xy section of the cell close to the basal membrane; the bottom panel shows a 1-μm xz section through the center of the nucleus. All images are presented as multiple-intensity projections. Bars, 10 μm. White arrows indicate the locations of GFP-capsid accumulation at the plasma membrane.

All of the constructs were not viable in terms of RNA packaging and infectious virus production (data not shown). Thus, the genomes of all of the designed recombinant constructs were synthesized in vitro and packaged into infectious virus particles by coelectroporation with helper, VEEV capsid-encoding RNA. The VEErep/SD1-GFP-Cc replicon RNA was packaged using the helper encoding both VEEV capsid and envelope glycoproteins. Next, BHK-21 cells were infected with the indicated constructs to analyze particle release. Cells infected with VEEV/SD1-GFP-Cc demonstrated development of filopodia and membrane fragmentation (Fig. 6B and data not shown) and released GFP-positive particles. They were released less efficiently than those formed by replicating VEEV/GFP but, after concentration, were readily detectable under a confocal microscope (Fig. 6C). Their further analysis by EM revealed that they have an irregular structure, different from icosahedral wt VEEV virions (Fig. 6D). No particle release was detected from cells infected with either VEErep/SD1-GFP-Cc, expressing no glycoproteins, or VEEV/GFP-Cc, which encoded a GFP-Cc fusion that lacked SD1 and SD2. Thus, SD1 is playing an important role in particle assembly and budding, which becomes particularly evident in the absence of other factors, such as RNA-capsid protein interaction.

Next, BHK-21 cells were infected with the designed constructs (Fig. 7A) at the same MOI, and 3D images were acquired (Fig. 7B). The recombinant capsid proteins expressed by VEEV/SD1-GFP-Cc and VEEV/SD12-GFP-Cc constructs demonstrated strong accumulation at the plasma membrane, likely in the site of VLP budding. We did not detect differences in membrane-specific SD1-GFP-Cc and SD12-GFP-Cc accumulation, but 3D confocal microscopy is not a sufficiently quantitative method. The GFP-Cc protein, which had no SD1 upstream of GFP, was diffusely distributed in the cytoplasm with no detectable accumulation at the plasma membrane. Importantly, expression of SD1-GFP-Cc in the absence of glycoproteins, achieved from the VEErep/SD1-GFP-Cc replicon, also did not lead to accumulation of this protein near the plasma membrane. This was an indication that interaction of VEEV capsid with glycoprotein spikes is absolutely essential for its accumulation/relocalization to the plasma membrane.

Taken together, the data suggest that SD1 function is critical for assembly of VEEV NC. It appears to be as important for the virion budding at the plasma membrane as (i) lateral interactions between the carboxy-terminal capsid's domains, which mediate assembly of icosahedral NC, and (ii) capsid interactions with glycoprotein spikes, which drive the envelope formation (Fig. 8).

FIG 8.

VEEV particle assembly requires several protein-protein and RNA-protein interactions. Schematic representation of VEEV particle assembly and budding. Arrows indicate critical protein-protein and RNA-protein interactions, which drive the budding process and stabilize NC and virion.

DISCUSSION

The data accumulated in the last few years strongly suggested that NC formation during VEEV assembly is a complicated process, which is tightly regulated and determined by more than specific interaction of capsid protein with the virus-specific PS in the genomic RNA (17, 26, 27). The studies performed on different alphaviruses also indicated the existence of two mechanisms in virion assembly. The first mechanism suggests preformation of the nucleocapsids in the cytoplasm, followed by their migration to the plasma membrane either by passive diffusion or at the surface of previously described cytopathic vacuoles (CPVII) (28–31). Such preformed NCs and CPVII vesicles are readily detectable by EM in the cytoplasm of virus-infected cells. Following migration to the plasma membrane, NCs attain their glycoprotein-containing lipid envelope during budding through the plasma membrane.

The second mechanism describes a process in which NCs form directly during particle budding (17, 18, 32, 33). The latter pathway of virion assembly and release appears to be very efficient, at least for VEEV, but was previously underscored in the experiments with wt viruses producing large amounts of intracellular NCs. However, engineering of VEEV variants with defective capsid proteins, which are incapable of forming stable cytoplasmic NCs, allowed the dissection of this mechanism and demonstrated its efficiency (18).

VEEV and other alphavirus capsid proteins contain two large domains. The carboxy-terminal domain functions as a self-protease in processing of the structural polyprotein (34). Expression of this individual domain together with viral glycoproteins was found to be sufficient for formation of the icosahedral VEEV NC scaffold, which interacts with glycoprotein spikes and mediates VLP release (17, 18). However, VLPs with NC formed by the carboxy-terminal domain are less stable than wt virions, and no cytoplasmic NCs are detected upon expression of the domain in the cells (17, 18).

The amino-terminal domain is highly positively charged and mostly disordered. It can be divided into four smaller subdomains, SD1 to SD4, of which SD3 and SD4 contain almost the entire repertoire of positively charged amino acids and are responsible for specific interaction with the viral genome. Without interaction with RNA, the high positive charge of SD3 prevents NC assembly (17). The functions of the other two amino-terminal subdomains, SD1 and SD2, are not very well explored. So far, it was shown that SD2 forms a short α-helix, which mediates dimerization of Sindbis virus (SINV) capsid protein (19, 35), which is essential for NC assembly. Deletion of this fragment in VEEV capsid protein also had a detectable negative effect on virus replication rates and final virus titers (17). SD1 is the very amino-terminal peptide, which is typically ∼35 aa long in different alphaviruses. It does not contain a high concentration of positively charged amino acids, is highly diverse among alphavirus species, and has not been predicted to have a defined secondary structure. Therefore, it was unclear whether this polypeptide plays any significant, specific role in virion assembly.

The results of this study demonstrate that despite its short length, SD1 is critical for this process. Its deletion has a deleterious effect on virion release, and in contrast to other subdomains, the negative effects of such deletion could not be compensated for by accumulation of adaptive mutations in other fragments of capsid protein. The replacement of VEEV SD1 by a peptide containing the same amino acids in a different order also reduced viral titers by a few orders of magnitude, suggesting the sequence-specific mode of SD1 function. Surprisingly, the replacement of SD1 by the same domains derived from CHIKV and EEEV (Fig. 2) demonstrated that these heterologous SD1 sequences, which certainly function very efficiently in their native context, are very inefficient in the context of VEEV capsid protein. Notably, the SD1 sequence of the more closely related New World (NW) alphavirus EEEV was more efficient in infectious virion release than the SD1 sequence derived from the more distant VEEV relative CHIKV. This was further evidence that VEEV SD1 is a critically important subdomain, which has been specifically optimized for efficient virion assembly by independent evolution of the different alphavirus species. Indeed, VEEV variants with chimeric capsid proteins rapidly evolved and accumulated adaptive mutations in CHIKV and EEEV SD1 sequences. Interestingly, compensatory mutations were found not only in SD1 itself but also in SD2 (helix I). VEEV with randomized SD1, VEEV/randSD1/GFP, developed the most extensive modification in SD2. All of the detected mutations strongly increased the rates of infectious virus release and indicated that SD1 and SD2 might synergistically function in virus assembly. Indeed, combinations of CHIKV-derived SD1 and SD2 peptides in VEEV/chikvSD12/GFP resulted in more efficient release of infectious virus particles compared to VEEV/chikvSD1/GFP. VEEV/chikvSD12/GFP and VEEV/chikvSD123/GFP variants replicate at higher rates and to higher titers than VEEV/chikvSD1/GFP.

The results of the experiments with VEEV variants containing GFP in their capsid protein also support the possibility that SD1 is critically involved in NC assembly and virion building. GFP/capsid chimeric proteins strongly accumulated at the plasma membrane and caused release of at least partially assembled particles, VLPs, only if they contained the amino-terminal SD1. Without this subdomain, the GFP/capsid protein was diffusely distributed in the cytoplasm.

Cryo-EM reconstruction studies have previously described the existence of a central core within NC (16). This core is thought to contain the first amino-terminal residues of the capsid protein, corresponding to the SD1 and SD2 subdomains, and it was suggested that its formation could initiate the rest of NC assembly. Thus, our functional data support this structural model and suggest that SD1, in cooperation with SD2, plays a crucial role in NC core assembly by mediating capsid multimerization, probably through SD1-SD1 or SD1-SD2 interactions. They also support the hypothesis that the latter core formation is critically important for the generation of both cytoplasmic NCs and those formed at the plasma membrane at the places of virion budding.

As was expected based on published data (28, 29), no accumulation of the protein at the plasma membrane was detected in the absence of glycoprotein expression, indicating that capsid interaction with glycoprotein spikes is a critical driving force in particle formation.

Taken together, the accumulated data suggest the existence of a variety of capsid-specific checkpoints in the VEEV assembly process. They are summarized in Fig. 8. (i) SD3 and SD4 mediate interaction with the viral genome. This interaction neutralizes the high positive charge of the capsid's amino terminus, makes further assembly possible, and strongly stabilizes the assembled NCs. (ii) Lateral interactions between carboxy-terminal domains of capsid proteins mediate formation of icosahedral NCs. (iii) Interaction of capsid with the cytoplasmic peptide of E2 in viral glycoprotein spikes is a driving force of virus budding. (iv) SD1 functions additively or synergistically with SD2 in NC central core formation.

Thus, despite a reasonably short length, ∼100 aa, the amino-terminal domain of VEEV capsid protein exhibits a wide variety of sequence-specific functions. (i) It contains both supraphysiological nuclear export and nuclear localization signals required for inducing transcription inhibition (9, 11). (ii) It has an RNA-binding activity that stabilizes NC and virions (18). (iii) It determines specificity of RNA packaging into virions (20). (iv) The amino-terminal SD1+SD2 fragment appears to form a central core in VEEV virions, which is critical for virus assembly and budding.