Different Types of nsP3-Containing Protein Complexes in Sindbis Virus-Infected Cells

Abstract

Alphaviruses represent a serious public health threat and cause a wide variety of diseases, ranging from severe encephalitis, which can result in death or neurological sequelae, to mild infection, characterized by fever, skin rashes, and arthritis. In the infected cells, alphaviruses express only four nonstructural proteins, which function in the synthesis of virus-specific RNAs and in modification of the intracellular environment. The results of our study suggest that Sindbis virus (SINV) infection in BHK-21 cells leads to the formation of at least two types of nsP3-containing complexes, one of which was found in association with the plasma membrane and endosome-like vesicles, while the second was coisolated with cell nuclei. The latter complexes could be solubilized only with the cytoskeleton-destabilizing detergent. Besides viral nsPs, in the mammalian cells, both complexes contained G3BP1 and G3BP2 (which were found in different ratios), YBX1, and HSC70. Rasputin, an insect cell-specific homolog of G3BP1, was found in the nsP3-containing complexes isolated from mosquito cells, which was suggestive of a high conservation of the complexes in the cells of both vertebrate and invertebrate origin. The endosome- and plasma membrane-associated complexes contained a high concentration of double-stranded RNAs (dsRNAs), which is indicative of their function in viral-RNA synthesis. The dsRNA synthesis is likely to efficiently proceed on the plasma membrane, and at least some of the protein-RNA complexes would then be transported into the cytosol in association with the endosome-like vesicular organelles. These findings provide new insight into the mechanism of SINV replication and virus-host cell interactions.

The genus Alphavirus in the family Togaviridae contains a number of widely distributed human and animal pathogens. Some of the alphaviruses, including Venezuelan (VEEV), eastern, and western equine encephalitis viruses, constitute a serious public health threat in the United States (53, 63, 65, 66) and cause severe encephalitis in humans and animals that can result in death or neurological sequelae (10, 21, 27, 41). Other family members cause a mild infection, a self-limited febrile illness characterized by fever, skin rashes, and arthritis (21). In spite of differences in their abilities to cause disease, alphaviruses demonstrate strong homology in their encoded proteins and appear to have similar mechanisms of RNA replication (59). Under natural conditions, alphaviruses circulate between mosquito vectors, in which they cause a persistent, life-long infection, and vertebrate hosts, in which the infection is always acute and characterized by a short-term, high-titer viremia that is required for infection of new mosquitoes during blood meals (64). Thus, alphaviruses are capable of replicating in both vertebrate and invertebrate cells and, accordingly, utilize very different intracellular environments for the efficient synthesis of virus-specific RNAs and the production of viral particles.

The alphavirus genome is a single-stranded RNA molecule of positive polarity and is approximately 11.7 kb long (31, 58, 60). The 5′ two-thirds of the genome encodes nonstructural proteins (nsP1 to nsP4) translated directly from the genomic RNA and forming a replicative enzyme complex (RC). The complex functions in viral-genome replication and transcription of a subgenomic RNA, which serves as a template for translation of all of the structural proteins, forming infectious virions (64). Sindbis (SINV) and Semliki Forest (SFV) viruses are the most intensively studied members of the genus. Their nonstructural proteins are synthesized as polyproteins that are sequentially processed into individual nsPs. During SINV infection, two polyproteins, P123 and P1234, are synthesized. nsP4 is cleaved in cis from the polyprotein, and the complex made up of P123 and nsP4 is capable of initiating minus-strand RNA synthesis (57). Further processing of P123, performed by nsP2-associated protease activity (25), transforms the RC into a mature form that functions in both the positive-sense genome and subgenomic-RNA synthesis but no longer produces negative strands (39, 40). To date, our understanding of the composition of this alphavirus RC is still incomplete, and the roles of cellular proteins in RC formation and viral replication have yet to be determined.

In vertebrate cells, SINV and SFV replication leads to the formation of unique membrane structures, namely, type 1 cytopathic vacuoles (CPV-1) (4, 13, 16, 23), having diameters between 600 and 2,000 nm. These vacuoles are believed to be derived from modified secondary lysosomes and endosomes (50) and contain small vesicular invaginations, or spherules, proposed as sites of RNA replication (22, 34). The available data also indicate that viral nsPs are located on the external side of the CPVs, and nsP1- to nsP3-containing protein complexes can be easily dissociated from CPVs (5, 16). Thus, the association of alphavirus nsPs with the membrane-containing organelles is likely not tight.

In previous studies, we and others developed and characterized the SINV variant SINV/nsP3GFP that encoded the recombinant nsP3/green fluorescent protein (GFP) (3, 9, 14, 19). The GFP insertion did not interfere with either virus replication or synthesis of virus-specific RNAs. Most importantly, chimeric nsP3/GFP formed large protein complexes, which had a morphology similar to that previously described for complexes detected in the cells infected with wild-type (wt) alphaviruses (14). The role of nsP3 in SINV replication is not understood. The protein is known to be essential for viral-RNA synthesis, and the mutations might affect both negative- and positive-strand RNA synthesis (24, 36, 37, 62). It contains an amino-terminal so-called X domain, which was found in the nsPs of many other RNA viruses (33) and whose function has yet to be determined. In the infected cells, SINV nsP3 is also highly phosphorylated (11, 42). It was recently demonstrated that SINV and VEEV nsP3s acquire adaptive mutations in response to modifications of the 5′ cis-acting elements in the virus genome (12, 48), which is suggestive of interaction between nsP3 and viral RNA.

By using the GFP-specific antibodies, we isolated from the crude cytosol fraction and identified a spectrum of cellular proteins that interact with nsP3/GFP and which might function in different viral replication processes. A similar set of proteins was isolated by another group in independent experiments (9). Minor variations in the protein spectra were likely due to some differences in isolation techniques. However, after these studies, it remained unclear whether the isolated cellular proteins are directly involved in RC formation. Interpretation of the results was complicated, because SINV nsP2, and possibly other proteins, function in various processes and not only in RNA replication. For example, nsP2 is efficiently transported into the nucleus (15, 51), where it functions as a transcriptional inhibitor (17, 18), and it also binds to ribosomes, suggesting its involvement in translation regulation (3).

In this study, we further dissected the composition of the SINV nsP3-specific protein complexes in different cellular fractions prepared from vertebrate and mosquito cells. BHK-21-specific complexes were isolated from membrane- and nucleus-containing fractions. Besides having different subcellular localizations, these two types of complexes demonstrated variations in their cellular protein contents. Nonetheless, both types contained common proteins, such as G3BP1 and G3BP2 (albeit present in different ratios), as well as YBX1 and HSC70. The lipid membrane-associated complexes isolated from SINV-infected mosquito cells contained the insect G3BP1 homolog Rasputin and HSC70. These results indicate a high conservation of complexes in different cell types. Other experiments demonstrated that only a subset of the nsP3 complexes appear to function as viral RCs. At early times postinfection in mammalian cells, the RCs accumulate on the plasma membrane, where they synthesize viral double-stranded RNA (dsRNA), and then a detectable fraction appears to be transported to the cytosol on the endosome surface.

MATERIALS AND METHODS

Cell cultures.

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 supplemented with 10% fetal bovine serum and vitamins. Mosquito C₇10 cells were obtained from Henry Huang (Washington University, St. Louis, MO) and propagated in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum and 10% tryptose phosphate broth.

Plasmid constructs.

The pSINV/nsP3GFP plasmid, containing the infectious cDNA of SINV with a GFP insertion in the nsP3-coding sequence, was described elsewhere (14). pSINV/nsP3Cherry had essentially the same design, but the GFP coding sequence was replaced by a Cherry gene (56). pSINrep/GFP, encoding the SINV replicon with GFP under the control of the subgenomic promoter, was described elsewhere (20). pSINrep(nsP3GFP) encoded the SINV replicon, containing no protein coding sequence in the subgenomic RNA and with a GFP insertion in nsP3. pRab5a/Cherry and pRab7/Cherry were designed by reverse transcription-PCR synthesis of the Rab-coding sequences using RNA isolated from NIH 3T3 cells. These sequences were fused with the Cherry gene by standard PCR-based techniques and cloned into a pTriEx1 vector plasmid (Novagen). Hamster G3BP1 and G3BP2, YBX1, and HSC70 sequences were amplified from the RNA isolated from BHK-21 cells using the same techniques and then fused with a GFP gene and cloned into an episomal pREP4 vector (Invitrogen). The vector was modified by cloning the puromycin acetyltransferase gene under the control of the P_TK promoter. The resulting plasmids were designated pREP/G3BP1GFP, pREP/G3BP2GFP, pREP/YBX1GFP, and pREP/HSC70GFP. Standard recombinant DNA techniques were used for all plasmid constructions. Maps and sequences are available from the authors upon request.

Expression of fusion proteins.

Rab5a/Cherry- and Rab7/Cherry-expressing cell lines were generated by the transfection of pRab5a/Cherry or pRab7/Cherry, along with the pMAM/Neo (Clonetech) plasmid (10:1), into BHK-21 cells using a FuGene 6 reagent according to the manufacturer's instructions (Roche), followed by selection of the G418-resistant cells. To avoid a possible negative effect of overproduction of the proteins, only the cell colonies with a moderate level of expression were used in further experiments. A G3BP1/GFP-expressing cell line was generated by transfecting pREP/G3BP1GFP into BHK-21 cells, followed by selection in the presence of puromycin (5 μg/ml). The clone with a moderate level of G3BP1/GFP expression was used for further experiments. Other fusion proteins demonstrated detectable cytotoxicity or caused cell growth arrest. Therefore, the corresponding plasmids were usually transfected into BHK-21 cells using FuGene 6 or Lipofectamine 2000 reagent according to the manufacturers' instructions (Roche and Invitrogen, respectively), and then, in 18 h, we employed puromycin selection. In standard experiments, ∼80% of the cells were Pur^r, and the untransfected cells died within 24 h after the start of puromycin selection. The following experiments with these cells were performed within the next 48 h, before the cells started to demonstrate morphological changes and cytopathic-effect development resulting from the expression of fusion proteins.

RNA transcription and transfection.

Plasmids encoding SINV genomes, replicons, and helpers were purified by centrifugation in CsCl gradients and linearized by XhoI digestion. RNAs were synthesized with SP6 RNA polymerase (Ambion) in the presence of cap analog (New England Biolabs). The yields and integrity of transcripts were analyzed by gel electrophoresis under nondenaturing conditions. In all of the experiments, the transfections were performed by electroporation of the in vitro-synthesized RNAs under previously described conditions (43).

Fractionation of cell lysates.

Subconfluent BHK-21 or C₇10 cells (1 × 10⁷ cells per 150-mm-diameter dish) were infected with packaged SINrep(nsP3GFP) and SINrep/GFP replicons or recombinant SINV/nsP3GFP and SINV/GFP viruses at a multiplicity of infection (MOI) of 20 infectious units (inf.u.) or PFU per cell, respectively. C₇10 cells were incubated at 30°C in 5% CO₂ for 24 h, and BHK-21 cells were incubated for 6 h at 37°C in 5% CO₂. Then, the cells were washed with phosphate-buffered saline (PBS), scraped, and pelleted by centrifugation at 1,000 × g. Next, they were suspended in hypotonic buffer (10 mM Tris-HCl [pH 7.5], 10 mM NaCl, 5 mM MgCl₂, and 1× protease inhibitor cocktail [Roche]) and, after 15 min of incubation on ice, broken in a tight glass Dounce homogenizer. In the experiments with C₇10 cells, the concentration of sucrose in homogenates was adjusted to 54% by adding 65% sucrose solution, and they were loaded into the bottoms of SW-41 rotor tubes under the layers of 50% (4 ml) and 25% (5 ml) sucrose containing 50 mM Tris-HCl [pH 7.5], 10 mM NaCl, 5 mM MgCl₂, and 1× protease inhibitor cocktail buffer (buffer A). In the experiments with BHK-21 cells, the cell homogenate was adjusted to 51% sucrose and loaded under 3 ml of 45% sucrose and 3.5 ml of 20% sucrose, prepared on buffer A. Samples were centrifuged for 16 h at 34,000 rpm at 4°C or 4 h at 37,000 rpm at 4°C for BHK-21 and C₇10 cells, respectively. Endosomes, microsomes, and plasma membranes formed a highly visible band (the vesicular [VES] fraction) between layers of 25% and 50% (or 20% and 45%) sucrose, and nuclei with residual attached endoplasmic reticulum and cytoskeleton formed a band between the 54% and 50% (or 51% and 45%) sucrose layers (the nuclear [NUC] fraction). Both the VES and NUC bands were harvested and used in further experiments. The cytosol proteins remained under the nuclear band and were not further used.

Isolation of protein complexes and identification of proteins.

For coimmunoprecipitation of the proteins from the VES and NUC fractions, we diluted the samples threefold with buffer A to reduce the concentration of sucrose and added either Empigen BB (Calbiochem) to 2% or Na deoxycholate (DOC) to 0.5%, Triton X-100 to 1%, and NaCl to 0.5 M. Next, we incubated the samples on ice for 20 min and pelleted the undissolved material at 15,000 × g for 20 min at 4°C. The supernatants were mixed with 100 μl of μMACS beads (Miltenyi Biotec) with affinity-purified anti-GFP antibodies. After 1 h of incubation at 5°C, the suspension was loaded on μColumns (Miltenyi Biotec) and washed four times with 200 μl of buffer A supplemented with corresponding detergent. Then, the bound proteins were eluted in 40 μl of protein gel sample buffer. The eluted proteins were separated on sodium dodecyl sulfate-10% polyacrylamide gels (35), followed by staining with Coomassie brilliant blue R-250 or by using a SilverSnap stain kit (for mass spectrometry [MS]), according to the manufacturer's instructions (Pierce). In parallel, the same affinity purification procedure was applied to the cells infected with viruses or replicons and expressing free GFP (not fused with any other protein). The stained bands were excised from the gel and prepared for matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) MS and MALDI-TOF/TOF MS analyses in the Mass Spectrometry Laboratory of the University of Texas Medical Branch Biomolecular Resource Facility, as previously described (3, 14). Applied Biosystems software was used in conjunction with MASCOT to search the NCBInr (rodent, mosquito, and viral) database for protein identification. Protein match probabilities were determined using expectation values and/or MASCOT protein scores.

Microscopy.

For confocal microscopy, C₇10 or BHK-21 cells were seeded onto Ibidi μ-dishes or eight-well μ-slides (Integrated BioDiagnostics), infected at an MOI of ca. 20 PFU/cell, and incubated at 30 or 37°C in a CO₂ incubator. At the indicated times, they were fixed in 3% formaldehyde in PBS, permeabilized with 0.5% Triton X-100, stained with mouse monoclonal antibodies against dsRNA and AlexaFluor 555-labeled secondary antibodies, and analyzed by using a Zeiss LSM510 Meta confocal microscope with a 63× 1.4-numerical-aperture (NA) oil immersion planapochromal lens. For staining of the lipid-containing cellular structures, cells were incubated in the presence of FM4-64FX dye (Invitrogen; 1 to 5 μg/ml) in complete medium, starting from the beginning of the infection. At the times indicated in the figure legends, they were fixed in 4% formaldehyde in PBS for 10 min, and three-dimensional analysis was performed on the confocal microscope using a 63× 1.4-NA oil immersion planapochromal lens. The image stacks were further processed using Huygens Essential v2.7 deconvolution software (Scientific Volume Imaging) and Imaris v4.2 (Bitplane AG) three-dimensional (3D) rendering software. For staining with LysoTracker DND99 (Invitrogen), cells were incubated for 30 min in complete medium supplemented with dye at 50 nM concentration in a CO₂ incubator at 30 or 37°C. Images were taken of live cells using the confocal microscope with a 63× 1.4-NA oil immersion planapochromal lens.

Transmission electron microscopy (EM).

Cell monolayers were fixed in a mixture of 2.5% formaldehyde, 0.1% glutaraldehyde in 0.05 M cacodylate buffer (pH 7.2) containing 0.03% trinitrophenol and 0.03% CaCl₂ and washed in 0.1 M cacodylate buffer (pH 7.2). They were then scraped off the plastic, pelleted, postfixed in 1% OsO₄ in 0.1 M cacodylate buffer, and stained en bloc with 1% uranyl acetate in 0.1 M maleate buffer, dehydrated in ethanol, and embedded in Poly/Bed 812 epoxy resin (Polysciences). Ultrathin sections were prepared on an Ultracut S ultramicrotome (Reichert-Leica), stained with 2% aqueous uranyl acetate and Pb citrate, and examined in a Philips 201 or CM 100 electron microscope at 60 kV.

Viral-replication analysis.

C₇10 cells (10⁶) were seeded into 35-mm dishes. After 4 h of incubation at 30°C in 5% CO₂, they were infected at an MOI of 1 PFU/cell. At the times indicated in the figures, the media were replaced, and virus titers were determined by plaque assay on BHK-21 cells (38).

RESULTS

SINV nsP3-containing complexes in mammalian cells.

In our previous work, we developed a SINV variant, SINV/nsP3GFP, that encoded the recombinant nsP3/GFP protein (14). The GFP insertion did not interfere with either virus replication or synthesis of virus-specific RNAs. Most importantly, chimeric nsP3/GFP formed large protein complexes, which had a morphology similar to that previously described for complexes detected in cells infected with wt alphaviruses. The morphological appearance of nsP3 complexes argued against their colocalization with endosomes or lysosomes. However, previously published data strongly indicated that at least a fraction of the alphavirus-specific protein complexes were associated with the membrane-containing cellular vacuoles, thought to be late endosomes and/or lysosomes (13, 22, 54). To determine whether nsP3/GFP complexes are associated with vesicular organelles, we applied a lipophilic dye, FM4-64FX. The dye is a vital stain that binds to the plasma membranes of live cells and consequently labels newly formed endosomes and lysosomes (61). The dye was added to live BHK-21 cells at the beginning of the infection with SINV/nsP3GFP virus, and the cells were imaged at 7 h postinfection (see Materials and Methods for details). As demonstrated in Fig. 1, a large fraction of the FM4-64FX-stained vesicles colocalized with the nsP3/GFP-specific protein complexes. However, the major fraction of the nsP3/GFP was not associated with the stained organelles (Fig. 1a and b).

FIG. 1.

Analysis of the colocalization of nsP3/GFP and different, membrane-containing cellular organelles. (a and b) BHK-21 cells were infected with SINV/nsP3GFP at an MOI of ∼20 PFU/cell and stained with FM4-64FX reagent as described in Materials and Methods. (c to f) Rab5a/Cherry (c and d)- and Rab7/Cherry (e and f)-expressing cell lines were infected with SINV/nsP3GFP at the same MOI. (g and h) BHK-21 cells were infected with SINV/nsP3GFP and stained with LysoTracker, as described in Materials and Methods. The images were acquired at 7 to 8 h postinfection. Panels b, d, f, and h represent fragments of cells shown in panels a, c, e, and g, respectively, at higher magnification. The distribution of nsP3/GFP is indicated by green, membrane-containing organelles are shown in red, and yellow indicates colocalization (additionally marked by white arrows).

To determine the type of the membrane-containing cellular organelles, we developed two stable lines of BHK-21 cells that expressed Rab5a or Rab7 (markers of early and late endosomes, respectively). Both proteins were fused with one of the red fluorescent proteins, Cherry (56). Both Rab5a/Cherry and Rab7/Cherry were readily detected in cellular vesicles, but in SINV/nsP3GFP-infected cells, only a small number of Cherry-marked endosomes exhibited colocalization with the nsP3/GFP-specific complexes (Fig. 1c to f). Similarly, only a small fraction of lysosomes, labeled with LysoTracker, were found in association with nsP3/GFP (Fig. 1g and h). The nsP3/GFP complexes were undetectable on mitochondria (data not shown). nsP3/GFP was localized outside of the vesicles, which suggested to us that they represent CPV-1 organelles, for which a similar outer-surface association of viral nonstructural proteins was demonstrated by EM (16). Thus, only a small fraction of SINV nsP3 complexes appear to be located on vesicular organelles, and this result is indirectly supported by the previously published data of Grimley et al. (23), who reported that by 4 h postinfection, CPV-1 vacuoles were observed in only 10 to 20% of the sections of cells infected with SFV at an MOI of 20 to 50 PFU/cell.

Taken together, the data suggested that the SINV-infected cells contained at least two types of nonstructural-protein-containing complexes. One of them appeared to be associated with the membrane-containing vesicular organelles. However, the major fraction of nsP3 formed another type of complexes that exhibited no binding to endosome-like vesicles. These complexes were coisolated with cell nuclei (14), and as we reported previously, this indirect association with nuclei during the isolation procedure is highly resistant to treatment with nonionic detergents and high salt levels (14).

Isolation of the SINV nsP3-containing complexes from different fractions of cell extracts.

In subsequent experiments, we attempted to isolate nsP3-specific protein complexes from different fractions of the cells infected with SINV-specific replicons. SINrep(nsP3GFP) contained a GFP insertion in the nsP3 gene after amino acid (aa) 389 and did not contain any genes under the control of the subgenomic promoter. SINrep/GFP was a control replicon with a GFP gene under the control of the subgenomic promoter. Replicons were used instead of infectious virus to avoid the coisolation of the structural viral proteins, which may be synthesized in close proximity to the sites of RNA replication and are also present on the so-called CPV-2 (16) vesicles. These vesicles contain on their cytosolic surfaces a high concentration of the nucleocapsids and, most likely, on their lumen sides, viral glycoproteins. It is highly unlikely that they are involved in viral-RNA replication, but they could be colocalized with cellular endosomes in the sucrose density gradients. Viral structural proteins are dispensable for RNA replication, and their presence in the samples after affinity purification with nsP- or GFP-specific antibodies (see the experiments with mosquito cells below) is likely not biologically significant.

BHK-21 cells were infected with packaged SINrep(nsP3GFP) or SINrep/GFP replicons at the same MOI. At 6 h postinfection, cells were harvested and homogenized under hypotonic conditions. Then, the sucrose concentration in the lysates was adjusted to 51%. We intentionally avoided any additional fractionation steps, particularly those leading to pellet formation, because they could affect the integrity of the complexes and/or organelles. The direct fractionation of cell lysates by sucrose gradient flotation centrifugation appeared to be the most gentle procedure. Samples were loaded into the bottoms of the tubes under a discontinuous sucrose gradient, and ultracentrifugation and fraction collection were performed as described in Materials and Methods. The membrane organelles and nuclei were clearly separated on the gradient, and the nuclear fraction contained threefold more nsP3/GFP than did the upper fraction, which contained plasma membrane and membrane-containing cellular organelles (Fig. 2).

FIG. 2.

Analysis of nsP3/GFP distribution in different fractions of cell lysates. BHK-21 cells were infected with packaged SINrep(nsP3GFP) at an MOI of 20 inf.u./cell. At 6 h postinfection, cells were harvested and homogenized. A nucleus-containing (NUC) fraction and that containing plasma membranes and vesicular organelles (VES fraction) were isolated by ultracentrifugation in discontinuous sucrose density gradients as described in Materials and Methods. Both fractions were additionally treated with Empigen BB, and both the pellet and the supernatant (Emp-P15 and Emp-S15, respectively), obtained after centrifugation at 15,000 × g, were analyzed for the presence of nsP3/GFP. Equal aliquots of the NUC, VES, and derived S15 and P15 fractions were analyzed by Western blotting. Membranes were developed by GFP-specific rabbit antibody and anti-rabbit IRDye 800-labeled secondary antibody. The intensity of the nsP3/GFP-specific signals was evaluated on a Li-Cor imager.

The nsP3/GFP-containing complexes in the nuclear fraction were not directly localized inside the nuclei or on the nuclear membrane. They were not solubilized by NP-40, Triton X-100, DOC, or their combinations with high salt (14). Therefore, the sucrose gradient-purified membranes and nuclei were next treated with 2% Empigen BB. This detergent is known to dissolve not only membranes, but also intermediate filaments (44) and, more likely, to destabilize weak protein-protein interactions. This treatment released a major fraction of the nsP3-containing complexes coisolated with nuclei and membranes (Fig. 2), and after further centrifugation at 15,000 × g, nsP3/GFP remained in the supernatants (S15). Only a small amount of this fusion protein was found in the pellets (P15). The protein complexes were isolated from the S15 fractions by using magnetic beads loaded with GFP-specific antibodies (see Materials and Methods for details). Identical fractionation and affinity purification procedures were applied to the control BHK-21 cells infected with packaged SINrep/GFP replicon. These samples were used (i) to identify the spectra of proteins nonspecifically binding to magnetic beads and (ii) to detect whether the purified nuclear and membrane fractions were GFP free, indicating that they were not contaminated with soluble cytosol proteins, which were expected to stay in the sucrose gradient below the nuclear band.

The spectra of isolated proteins are presented in Fig. 3. The control samples, prepared from SINrep/GFP-infected cells, contained very low levels of proteins (Fig. 3B, lanes 2 and 5), which were almost undetectable on the gels. Complexes, isolated from the Empigen-treated, membrane-containing (VES) fraction, demonstrated the presence of all of the SINV nsPs (nsP1 to -4), HSC70, YBX1, and G3BP2 (with a minor fraction of G3PB1) (Fig. 3B and C). The samples purified from the nuclear fraction contained all of the above-described proteins (Fig. 3B and C). However, they were detected at ratios that were very different from those in VES-derived samples. The most distinguishing feature was a higher level of G3BP1 than of G3BP2, and nearly all of the large-ribosomal-subunit proteins and histones H1.5/2/4 were also identified. The presence of the H1 histones was most likely an indication that the Empigen treatment affected, to some extent, the integrity of the nuclei, and the coisolation of these proteins might be due to their nonspecific binding to dsRNA or interaction with SINV nsP2. The latter protein is detected at high concentrations in the nuclei of SINV-infected cells (15). The coisolation of the ribosomal subunits is probably not coincidental, because it was very reproducible in the above-described experiments and several others.

FIG. 3.

Identification of the viral and cellular components in the nsP3-containing protein complexes. (A) Schematic representation of the SINV replicons used in the study. (B) BHK-21 cell were infected with the indicated, packaged replicons at an MOI of 20 infectious units/cell. Cells were harvested at 6 h postinfection, and the VES and NUC fractions were isolated by ultracentrifugation in sucrose density gradients, as described in Materials and Methods. Protein complexes were isolated by using GFP-specific antibody from an S15 supernatant of the Empigen-treated NUC and VES fractions. After elution from the affinity columns, proteins were separated on sodium dodecyl sulfate-10% polyacrylamide gels and stained with Coomassie brilliant blue R-250. The stained bands were excised from the gel, and the proteins were identified by MALDI-TOF analysis. The names of the identified proteins are indicated. The band of the protein that was not definitively identified is marked with an asterisk. Lanes 1 and 4 represent 1/540 of the NUC Emp-S15 and VES Emp-S15 fractions, respectively, before affinity purification of the proteins. Lanes 2 and 5 represent five-eighths of the samples isolated from the NUC Emp-S15 and VES Emp-S15 fractions, respectively, derived from the SINrep/GFP-infected cells. Lanes 3 and 6 represent five-eighths of the samples isolated from the NUC Emp-S15 and VES Emp-S15 fractions, respectively, derived from SINrep(nsP3GFP)-infected cells. IgG, immuno- globulin G. (C) Magnified fragments of the gels (lanes 3 and 6), containing HSC70, G3BP1, G3BP2, and nsP1 (boxed in panel B). Note the two forms of G3BP2, which likely represent its differently spliced forms.

nsP3-specific protein complexes in SINV-infected mosquito cells.

Alphaviruses can efficiently replicate in cells of both vertebrate and invertebrate origin. However, in spite of profound differences between the two types of cells, the mechanism of virus replication has been mostly studied in mammalian and avian, but not in insect, cells. Therefore, we intended to compare the compositions of the nsP-containing protein complexes and their compartmentalization in the two fundamentally different environments provided by cells of mammalian and mosquito origin.

In initial experiments, Aedes albopictus-derived C₇10 cells were infected with SINV/nsP3GFP and parental wt SINV Toto1101 at the same MOI. These viruses demonstrated very similar growth rates, and only a small delay in SINV/nsP3GFP replication (Fig. 4A), detected at early times postinfection, cannot be considered an indication that GFP insertion made the results biologically irrelevant. The nsP3-specific protein complexes were readily detected in the C₇10 cells within the first 8 h postinfection. By 12 h, they demonstrated a noticeable association with cellular vacuoles, which suggested an accumulation of nsP3/GFP fusions with endosome-like structures. Staining of live infected cells with FM4-64FX or LysoTracker unambiguously demonstrated that a significant fraction of protein complexes was associated with the lysosomes (Fig. 4B and C). We also analyzed infected mosquito cells to determine whether they contained CPV-1-like vesicles, as previously described in alphavirus-infected mammalian cells (22, 23). These CPVs were readily identified by EM (Fig. 4D). They had numerous invaginations (the distinguishing feature of CPVs) and electron-dense materials on their surfaces, predominantly at the neck of every invagination, previously suggested as nsP accumulation sites (16). Thus, taken together, the results strongly suggested that in the mosquito cells, a significant fraction of SINV nsP-containing complexes might be associated with cellular vacuolar organelles. Notably, nsP3/GFP was also distributed outside of the lysosomes in the form of smaller complexes (Fig. 4B and C).

FIG. 4.

Analysis of nsP3/GFP accumulation in C₇10 cells infected with SINV/nsP3GFP. (A) The schematic representation of wt SINV Toto1101 and SINV/nsP3GFP genomes and replication of these viruses in C₇10 cells. The cells were infected at an MOI of 1 PFU/cell and incubated at 30°C in 5% CO₂. The media were replaced at the indicated time points, and the titers of the released viruses were determined by a plaque assay on BHK-21 cells. (B and C) C₇10 cells were infected at an MOI of ca. 20 PFU/cell and incubated at 30°C in a CO₂ incubator. They were incubated in the presence of either LysoTracker for 30 min at 23.5 h postinfection (B) or FM4-64FX dye for 24 h (C), starting from the beginning of the infection. 3D analysis was performed on the confocal microscope with a 63× 1.4-NA oil immersion planapochromal lens. Panels a demonstrate the distribution of nsP3-GFP; panels b show staining of the cell with LysoTracker or FM4-64FX; panels c represent an overlay of the images; panels d present fragments of the cells shown in panels c at a higher magnification. The bars correspond to 10 μm. (D) CPV-1, detected by EM in C₇10 cells at 24 h postinfection with SINV/nsP3GFP. Staining was performed as described in Materials and Methods. Panel b presents a fragment of panel a at a higher magnification.

In later experiments, we infected C₇10 cells with SINV/nsP3GFP and SINV/GFP variants at the same MOI. The latter virus encoded GFP under the control of the subgenomic promoter (Fig. 5A) and was used to detect the cellular and viral proteins that might nonspecifically coisolate with GFP. At 24 h postinfection, cells were harvested and homogenized under hypotonic conditions, as described in Materials and Methods. Then, the homogenates were fractionated by sucrose gradient flotation centrifugation. SINV nsP2 and nsP3/GFP were readily detectable by specific antibodies in the membrane-containing (VES) fraction on Western blots (data not shown), and nsP3/GFP was also found in association with isolated vesicles by analysis of the samples under a confocal microscope (Fig. 5B). In additional experiments, we attempted to further fractionate endosomes by using gradients with more sucrose layers. However, the nsPs were present in all of the fractions at essentially the same concentrations (data not shown). This could be explained by the finding that, as detected by confocal microscopy, membrane-containing organelles were present in the samples in a highly aggregated form (Fig. 5B); therefore, further experiments with fractionation were discontinued.

FIG. 5.

Identification of the viral and cellular components in the nsP3-containing, membrane-bound protein complexes formed in mosquito cells. (A) Schematic representation of the genomes of SINV variants used in the study. (B) Distribution of nsP3/GFP and lipids in a VES fraction isolated by ultracentrifugation in sucrose density gradients as described in Materials and Methods. Panel a represents the distribution of nsP3/GFP; panel b shows staining with FM4-64FX. (C) Protein complexes were isolated by using GFP-specific antibody from an S15 supernatant of a VES fraction, treated with either 2% Empigen BB or 0.5% DOC, 1% Triton X-100, and 0.5 M NaCl. After elution from the affinity columns, the proteins were separated on sodium dodecyl sulfate-10% polyacryamide gels and stained with a SilverSnap kit. The stained bands were excised from the duplicate gel and stained with Coomassie brilliant blue, and the proteins were iden- tified by MALDI-TOF analysis. The names of the identified proteins are indicated. The asterisks mark the proteins that were not definitively identified. Lanes 1 and 2 represent 1/675 of the VES fractions derived from SINV/GFP- and SINV/nsP3GFP-infected cells, respectively, before affinity purification of the proteins. Lanes 3 and 5 represent three-eighths of the samples isolated from the VES S15 fractions derived from SINV/GFP-infected cells. Lanes 4 and 6 represent three-eighths of the samples isolated from the VES S15 fraction of the cells infected with SINV/nsP3GFP. IgG, immunoglobulin G.

VES fractions from SINV/nsP3GFP- and SINV/GFP-infected cells (Fig. 5C, lanes 1 and 2) were further used to isolate the nsP3/GFP- and GFP-specific protein complexes. Two different detergents were applied for protein extraction (2% Empigen BB or 0.5% DOC, 1% Triton X-100, and 0.5 M NaCl), followed by isolation of protein complexes using magnetic beads loaded with GFP-specific antibodies (see Materials and Methods for details). Analysis of the isolated samples by gel electrophoresis demonstrated that, regardless of the detergent used, almost no proteins were present in the samples derived from the SINV/GFP-infected cells (Fig. 5C, lanes 3 and 5). The content of the nsP3 complexes was dependent to a certain extent on the detergent used. Compared to DOC-Triton-NaCl, the Empigen treatment reduced the number of proteins coisolated with nsP3/GFP (particularly the amount of E2 plus E1 glycoproteins) but left a large amount of SINV capsid in the samples. The Empigen treatment-derived samples also contained more large-ribosomal-subunit-specific proteins. We cannot provide a definitive explanation for this detergent dependence during protein isolation, but most likely it was due to endosomes, lysosomes, residual endoplasmic reticulum (microsomes), and CPV-2-like vesicle aggregation in the sucrose gradients and variances in solubilization of the aggregates by the detergents. This phenomenon is now under investigation.

Protein bands were excised from the Coomassie- or silver-stained gels, and proteins were identified by MS (Fig. 5C). All of the SINV nsPs (nsP1, nsP2, nsP3/GFP, and nsP4) were clearly detected in the samples. Two things were notable. (i) nsP3/GFP was present in multiple bands (most likely as a result of the different levels of its phosphorylation) and in the nsP3GFP/nsP4 fusion. The latter protein was previously detected in SINV-infected mammalian cells (Fig. 3) in both our previous experiments (14) and those of other groups (5). (ii) nsP1 was also reproducibly identified in two bands, and the presence of the minor band can likely be explained by either differences in the palmitoylation of the protein or partial initiation of its translation from the downstream methionine. The most abundant cellular proteins were Rasputin, the heat shock protein HSC70, and one of the 14-3-3 proteins. The last two proteins were also isolated from the nsP3/GFP-containing complexes affinity purified from the cytoplasmic fraction of BHK-21 cells (9, 14) and in the experiments presented in Fig. 3. Rasputin is an insect cell-specific homolog of G3BP1 (G3BP) (26) that was previously shown to be a major component of stress granules (29, 30). Rasputin is 224 aa longer than G3BP1 but retains the same nuclear transport factor 2-like and so-called RRM domains (26). Thus, in spite of a small number of identical amino acid residues, it might play a role(s) in SINV replication in mosquito cells similar to that of G3BP1 in cells of vertebrate origin. Three cellular proteins, shown in Fig. 5C, were not identified by MS. Their identities will be defined later by direct protein sequencing.

Thus, the spectra of the proteins isolated from both vertebrate and mosquito cells revealed the possibility of a high degree of SINV nonstructural-protein-containing complex conservation in the cells of vertebrate and invertebrate origin.

G3BPs are colocalized with SINV nsP3-containing complexes.

The detection of G3BPs and their mosquito homolog Rasputin in the nsP3-containing complexes by different methods indicated that this type of cellular protein might play a role(s) in SINV replication. Both G3BP1 and G3BP2 were detected in the protein samples purified from nuclear and membrane fractions; moreover, both isoforms of G3BP2 were isolated (Fig. 3B and C). Therefore, to additionally confirm G3BP interaction with viral nsPs, we synthesized hamster G3BP1 and G3BP2 genes and expressed them as GFP fusions in BHK-21 cells. G3BP1/GFP was not cytotoxic but formed aggregates when expressed at high levels. Therefore, we selected stable cell lines that expressed low levels of the G3BP1/GFP fusion, which, like the endogenous protein, was diffusely distributed in the cytoplasm (Fig. 6). Expression of G3BP2/GFP caused cell growth arrest even at low expression levels, and therefore, the experiments were performed within 3 days after transfection of the plasmid, before the cells started to change morphology and degrade (see Materials and Methods for details). Both fusions were diffusely distributed in the cytoplasm; however, infection with SINV/nsP3Cherry (infectious virus with a Cherry insertion after aa 389 in nsP3) caused a rapid accumulation of G3BP1/GFP and G3BP2/GFP in the sites of nsP3/Cherry complex formation (Fig. 6). This profound relocalization occurred within the first 4 h postinfection and was independent of Cherry insertion, because cells infected with wt SINV Toto1101 demonstrated exactly the same phenomenon of aggregation of G3BP fusions to the sites of nonstructural-protein complex formation (data not shown).

FIG. 6.

Analysis of G3BP1 and G3BP2 relocalization in cells infected with SINV/nsP3Cherry variant. The stable cell line, expressing G3BP1/GFP or BHK-21 cells transiently expressing G3BP2/GFP (see Materials and Methods for details), were infected with SINV/nsP3Cherry at an MOI of ∼20 PFU/cell. The images were acquired at 4 h postinfection on a confocal microscope.

In the experiments following those described above, we also coprecipitated protein complexes from G3BP1/GFP-expressing cells infected with SINV/nsP3Cherry by using magnetic beads loaded with affinity-purified anti-GFP antibody (Fig. 7). In the isolated samples, nsP3/Cherry was readily detected on Western blots by nsP3-specific antibody. Similarly, nsP3/Cherry was coprecipitated from the SINV/nsP3Cherry-infected cells expressing YBX1/GFP and HSC70/GFP proteins by using the same GFP-specific antibody (Fig. 7). Thus, these data supported the notion that G3BPs, YBX1, and HSC70 are components of the SINV nsP3-containing complexes.

FIG. 7.

Coprecipitation of nsP3-Cherry protein from cells expressing G3BP1/GFP, YBX1/GFP, and HSC70/GFP. (A) Schematic representation of the genome of SINV/nsP3Cherry virus used in the study. (B) Cells expressing the indicated GFP fusions were infected with SINV/nsP3Cherry, and protein complexes were isolated by a protocol described elsewhere (14), using magnetic μBeads loaded with affinity-purified anti-GFP antibody. In the isolated samples, nsP3/Cherry was detected by Western blotting. The membranes were developed by SINV nsP3-specific rabbit antibody and anti-rabbit IRDye 800-labeled secondary antibody. The intensity of the nsP3-specific signals was evaluated on a Li-Cor imager.

Replication of SINV-specific RNA occurs in the subset of the nsP3-containing complexes.

The above-mentioned experiments suggested that SINV replication in BHK-21 cells leads to an accumulation of the nsP3-specific protein complexes in (i) the membrane organelle-containing and (ii) the nucleus-containing fractions. However, the present data (Fig. 2) and the results of the previous study (14) demonstrated that, by 8 h postinfection, the nuclear fraction contained more SINV nsP3 than did either the membrane-containing fraction or the postnuclear supernatant. Therefore, the obvious questions were where does viral RNA synthesis proceed and where are the SINV RNA replicative intermediates compartmentalized in the infected cells? It was generally believed that negative-strand RNAs are synthesized as a dsRNA, but it was always a concern that detection of dsRNA might be a result of the RNA isolation procedure. Therefore, in the subsequent experiments, we used dsRNA-specific monoclonal antibodies to define the localization of viral dsRNA intermediates in the infected cells. Starting from 2 h postinfection, a large fraction of nsP3-containing complexes could be seen in association with the plasma membrane and were also found to be colocalized with numerous discrete complexes of dsRNA, which indicated that they are active for RNA replication (Fig. 8A, panels b, c, and d, and 8B). Immunofluorescent staining was performed on the paraformaldehyde-fixed cells (see Materials and Methods for details), thereby excluding any possibility of dsRNA formation by partial annealing of single-stranded RNAs (ssRNAs).

FIG. 8.

Localization of dsRNA- and nsP3/GFP-containing protein complexes in BHK-21 cells infected with SINV/nsP3GFP. (A) BHK-21 cells infected with SINV/nsP3GFP at an MOI of ∼20 PFU/cell. At 4 h postinfection, the cells were fixed with 3% paraformaldehyde, permeabilized with 0.5% Triton X-100, and stained with mouse dsRNA-specific primary antibodies and AlexaFluor 555-labeled secondary antibodies. Panels a and e represent the distribution of nsP3/GFP, panels b and f demonstrate the localization of dsRNA, panels c and g are overlays of two images, and panels d and h show the colocalization of nsP3/GFP and stained dsRNA, which is indicated in white. Colocalization was estimated using Imaris v4.2 software. Panels e to h present enlarged fragments of the cells shown in panels a to d, respectively. (B) Images of BHK-21 cells infected with SINV/nsP3GFP and stained with dsRNA-specific antibody. At 4 and 12 h postinfection (panels a and b, respectively), cells were fixed with 3% paraformaldehyde, permeabilized with 0.5% Triton X-100, and stained with mouse dsRNA-specific primary antibodies and AlexaFluor 555-labeled secondary antibodies. The 3D images were acquired on a confocal microscope equipped with a 63× 1.4-NA oil immersion planapochromal lens. The image stacks were further processed using Huygens Essential v2.7 deconvolution software and the 3D rendering software Imaris v4.2. The central images are presented as multiple-intensity projections of the entire stack, and the (x, z) and (y, z) sections of the reconstructed images are presented as multiple-intensity projections of 2-μm sections denoted on the central image. Note that at 12 h postinfection, the cells started to change morphology due to cytopathic-effect development. Panels c and d present enlarged fragments of the (x, z) sections boxed in panels a and b, respectively.

Further analysis of 3D images by using deconvolution and rendering software demonstrated that, at early times postinfection, the dsRNAs were almost exclusively localized to plasma membranes, where dsRNA-containing complexes formed punctuated stripes and large arrays. The enlarged images of the cell, representing a fragment of the plasma membrane with nsP3/GFP-containing complexes and dsRNA, are demonstrated in Fig. 8A, panels e to h. nsP3/GFP protein was distributed on the plasma membrane around dsRNA sites (Fig. 8A, panel g). However, a significant amount of this protein was directly colocalized with dsRNA (Fig. 8A, panel h). To confirm the interaction of dsRNA with viral nsPs, in another experiment, we isolated complexes from the NP-40-treated VES fraction using dsRNA-specific antibody and protein G-loaded magnetic μBeads. In these samples, nsP3/GFP, nsP2, and nsP1 were unambiguously identified in the Coomassie-stained gels by MS (data not shown), which suggested that nsPs are in direct contact with SINV dsRNA replicative intermediates.

The number of dsRNA molecules, detected on the plasma membrane at early times post infection (Fig. 8B, panels a and c) continued to grow, and at later times, a detectable fraction of dsRNA relocalized to the cytoplasm (Fig. 8B, panels b and d), most likely by endocytosis (34). It was previously suggested that the presence of alphavirus-specific spherules on the plasma membrane either is the result of fusion of CPVs with the membrane or is mediated by direct, nsP1-mediated binding of RCs to the anionic phospholipids, followed by CPV formation via endocytosis (23, 34). Our data indicate that the plasma membrane is not only the site of RC formation, but is the place of viral dsRNA synthesis, and many dsRNA molecules remain associated with the plasma membrane even at late times postinfection. However, most importantly, a large fraction of nsP3/GFP fusion protein, even at early times postinfection, was found in the cytoplasm, where it formed complexes in which dsRNA was undetectable (Fig. 8B, panels b and d). These complexes either had a very different structure, and dsRNA was no longer accessible to the antibodies, even after permeabilization with nonionic detergent, or they contained no dsRNA and had functions different from those of RNA replication. These possibilities are now under investigation.

DISCUSSION

Our previous results and those of other research groups demonstrated that, in infected cells, SINV-specific protein complexes are composed not only of viral nonstructural proteins, but also of proteins having a cellular origin (2, 9, 14). We identified a number of RNA-binding and other proteins in samples isolated from the postnuclear supernatant of SINV/nsP3GFP-infected BHK-21 cells by using affinity-purified GFP-specific antibodies. However, it was reasonable to expect that virus replication could lead to the formation of a variety of complexes having different functions in virus replication and virus-host cell interactions. Thus, the application of the affinity purification procedure to the crude cytoplasmic fraction might result in the detection of a mixture of the components belonging to different types of complexes. Moreover, the crude cytoplasmic extract contained a very high concentration of the proteins that might bind to viral ds- and ssRNAs during isolation procedures after treatment of the lysate with nonionic detergents. In addition, the experimental data also indicated that some of the SINV nsP3-containing complexes were associated with membrane-containing organelles (9, 14, 16, 22, 23, 34) but that others (containing the major fraction of nsP3) were copurified with cell nuclei and resisted treatment with nonionic detergents (14). Previous EM-based studies suggested the presence of nsP3 in two types of intracellular structures. This protein was identified on the surfaces of CPV-1 vesicles (16), but at the same time, it formed large complexes not associated with any vesicular organelles (9, 54). Thus, we had many reasons to expect to find at least two types of nsP3-containing complexes that formed in the cells during SINV replication.

Therefore, we first analyzed the association of SINV nsP3 with cellular organelles. In SINV/nsP3GFP-infected cells, a small fraction of nsP3/GFP was found on the surfaces of early/late endosomes and lysosomes, and higher association of nsP3 with cellular membrane-containing organelles was detected when the cells were labeled with FM4-64FX dye. Since FM4-64FX dye was applied at the beginning of infection, it labeled only newly formed endosomes and ultimately marked the lysosomes. Thus, the nsP3/GFP and FM4-64FX colocalization suggested to us that vesicle-bound nsP complexes originated on the plasma membrane as previously proposed (34). Importantly, a large fraction of nsP3 was not associated with membrane-containing organelles and formed another type of complex.

To determine the difference in protein composition of the two types of nsP3 complexes, we extensively modified the previously used immunoprecipitation procedure. The membrane- and nucleus-containing fractions were purified by ultracentrifugation of lysates of SINrep(nsP3GFP)-infected BHK-21 cells in a sucrose gradient. Then, to compare the compositions of the membrane- and nucleus-associated complexes, both fractions were treated under the same conditions with the zwitterionic detergent Empigen BB. This reagent not only is capable of solubilizing the membranes, it also dissociates the intermediate filaments and some protein complexes (44). As expected, it released protein complexes from the membrane fraction and at least partially solubilized the nucleus-associated, virus-specific complexes. Next, the nsP3/GFP-bound proteins were affinity purified by using a GFP-specific antibody. Both types of complexes displayed similarity in protein composition, in that they both contained all of the SINV nsPs, as well as HSC70 and YBX1. Other major components were represented by G3BP1 and G3BP2, which were found in different ratios. Similarly, in mosquito cells, the G3BP1 homolog Rasputin was coisolated with nsP3/GFP from the sucrose gradient fraction containing endosomes, lysosomes, and, most likely, plasma membranes. This conservation of the G3BP component in the complexes isolated from such varied cell types suggested the importance of G3BPs and Rasputin, either for SINV replication or for virus-specific modification of cell biology.

G3BP1 was previously described as a protein interacting with RanGTP and was shown to regulate cell proliferation and RNA degradation. It was also identified as a major component of stress and RNA granules, as well as P bodies (places of cellular mRNA degradation) and RNA granules (29, 30). Although the compositions of all granule types have exhibited strong similarities, the granules themselves have different functions. The stress granules are the sites of stalled translation, and their formation is stress dependent. The RNA granules are, in turn, the means of mRNA transport in the neuronal axons and dendrites (1). The cellular function(s) of G3BP2 is less understood, but it is also resident in stress and RNA granules (26).

During SINV replication, G3BP1/GFP and G3BP2/GFP were found to change their distribution, at first being diffuse and then accumulating in the nsP3-containing complexes (Fig. 6). This phenomenon provides a plausible explanation as to why the classical stress granules were no longer formed in response to arsenite treatment in SFV-infected cells (47). Therefore, it is reasonable to speculate that by sequestering G3BP1 (and G3BP2), SINV developed a means of interfering with classical stress granule formation and might use it to modify cellular translation and to redirect it to the synthesis of virus-specific proteins. Interestingly, binding of G3BPs to nsP-containing complexes was detected during SINV, but not VEEV, infection (data not shown). This correlates with the slower inhibition of translation in VEEV-infected cells and led us to suggest that G3BPs might not be directly involved in RNA replication.

Besides G3BPs, the complexes coisolated with cell nuclei appeared to contain ribosomes, and their presence might also indicate that they have a function in the preferential translation of virus-specific templates. Such a function was recently described for G3BP1-containing complexes in the translation of viral templates in vaccinia virus-infected cells (28). However, the ribosome colocalization may be also a result of their binding to nsP2. The latter protein is known for its ribosome-binding activity (3, 52), and it was definitively identified in the isolated samples.

Another identified protein, YBX1 (or mYb-1b protein), belongs to the evolutionarily conserved family of nucleic acid-binding proteins. Y-box proteins have been described in bacteria, plants, and animals, and all of them contain a “cold shock” domain (CSD). Three members of this family were identified in vertebrates (45, 46), but only YBX1 is expressed ubiquitously in adults. In vertebrate cells, the Y-box proteins are comprised of three domains: an alanine- and proline-rich N-terminal domain, followed by a CSD domain and a C-terminal domain that consists of repeated sequences of predominantly basic or acidic amino acids (B/A repeat) (32). Two domains, the CSD and the B/A repeat, have been shown to mediate binding to RNA, ssDNA, and DNA; however, information about the sequence specificity of RNA binding is very limited. YBX1 has been implicated in the regulation of transcription, translation, and RNA processing. It is mostly present in the cytoplasm, where it can be found as a component of ribonucleoprotein complexes, and has recently been reported to bind to the dengue virus 3′ untranslated region and to have an antiviral effect (49). Moreover, in our experiments, YBX1 was detected in SINV nsP3-containing complexes isolated from mammalian, but not mosquito, cells. Thus, it might function as a component of the antiviral cell response in cells of vertebrate origin. This hypothesis needs further experimental support.

The heat shock HSC70 protein was the third cellular factor reproducibly isolated from both BHK-21 and C₇10 cells, and SINV replication led to its partial relocalization to sites of nsP3 accumulation (14). The growing body of evidence indicates that heat shock proteins play important roles in the replication of many viruses (6-8, 55, 67), and our results suggest alphaviruses are not an exception. The mechanism of HSC70 function in SINV replication has yet to be determined.

Another important result from our study was that SINV nonstructural-protein complexes synthesize virus-specific dsRNAs on the plasma membrane, where at early times postinfection, the dsRNAs were found to be organized in large punctuated arrays. On the plasma membrane, the dsRNA was definitively colocalized with small, nsP3-containing complexes, and their size suggested to us that they were likely to be in the early stages of development. At late times postinfection, a small but detectable fraction of dsRNA was transported into the cytosol. The number of dsRNA-containing structures correlated with low concentrations of nsP3/GFP-containing endosomes found in the cytosol even at late times postinfection (Fig. 1 and 8B, panel b). A high concentration of dsRNA on the plasma membrane indicates that its role in viral dsRNA and most likely in ssRNA synthesis might be strongly underestimated. The dsRNA is not accessible to specific antibody (even if cells are physically broken) unless the cells are treated with nonionic detergent (data not shown). This finding suggests that, on both the endosomes and the plasma membrane, these molecules are localized inside lipid membrane-containing compartments that prevent the dsRNA from efficiently functioning in triggering signaling cascades, such as those mediated by MDA5 or RIG-I, and inducing the antiviral response. During fractionation in sucrose gradients, the dsRNA was mainly present in the fraction containing membrane organelles (endosomes and plasma membrane), rather than that associated with nuclei, which appeared to indicate that the latter complexes play a less important role, if any at all, in viral-RNA synthesis (data not shown).

In conclusion, we demonstrated the following. (i) SINV infection in BHK-21 cells led to the formation of at least two types of nsP3-containing complexes, one of which was found in association with endosome-like vesicles and the plasma membrane while the second was coisolated with cell nuclei and could be solubilized only by treatment with filament-destabilizing detergent, but not with standard, nonionic detergents. (ii) Both types of complexes contained as common components G3BP1 and G3BP2 (in different ratios), YBX1, and HSC70. (iii) Rasputin, an insect cell-specific homolog of G3BP1, was identified in the membrane-associated, nsP3-containing complexes isolated from mosquito cells, a finding that suggested that the complexes were highly conserved in the cells of vertebrate and invertebrate origin. (iv) The membrane-associated complexes contained high concentrations of dsRNAs, which indicated their efficient function in viral RNA synthesis. (v) dsRNA synthesis efficiently proceeds on the plasma membrane, followed by at least partial transport of dsRNA into the cytosol with endocytotic vesicles. These findings provide new insights into the mechanism of SINV replication and virus-host cell interactions.