INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Sindbis virus is a member of the Alphavirus genus within the Togaviridae family. Alphaviruses are a small group of RNA viruses that are enveloped by a host cell-derived lipid membrane. In contrast to the loose and unorganized envelopes of the orthomyxovirus and paramyxovirus families, the envelope of alphaviruses is a highly organized icosahedral structure of transmembranal virus glycoproteins and a lipid bilayer (18). Sindbis virus is composed of a 49S positive-sense RNA genome and 240 copies each of three structural proteins: two transmembranal glycoproteins (E1 and E2) and an internal capsid (C) protein (41, 45). The structural proteins assemble into two concentric icosahedral shells arranged about the genome in a T=4 icosahedral lattice (6, 22, 37). The inner lattice is the nucleocapsid, which has an outer radius of approximately 200 Å and consists of 240 copies of capsid protein (7, 9, 37, 38). The envelope extends from an inner radius of 200 Å to an outer radius of 325 Å and is composed of 80 trimers of E1/E2 heterodimers which are embedded in a lipid bilayer (37). The (E1/E2)₃ heterotrimers protrude 50 Å from the surface of the virus and flare out into three distinct lobes forming a trimer 230 Å in diameter (37).

The structural proteins are translated from a 26S subgenomic RNA as a 130-kDa precursor protein that is subsequently processed into its individual components. The 30-kDa capsid protein, the first structural protein to be translated, is autocatalytically cleaved from the nascent polypeptide shortly after synthesis. After release from the ribosome translation complex, the capsid protein condenses onto newly synthesized genomic RNA to form nucleocapsids in the cytoplasm of the infected cell. The cleavage of capsid from the growing polypeptide chain exposes a signal sequence that delivers the polypeptide to the endoplasmic reticulum, where protein synthesis continues and glycosylation begins. Translation of protein from the mRNA into the rough endoplasmic reticulum produces PE2, the precursor of E2, and E1 spike proteins, which associate rapidly into PE2/E1 heterodimers and then almost immediately into (PE2/E1)₃ heterotrimers (24, 34, 57). PE2 contains a 64-amino-acid-long E3 sequence preceding the 423-amino-acid-long E2 glycoprotein (42, 46). The cleavage between E3 and E2 occurs in a post-Golgi compartment prior to the delivery of the spike trimers to the plasma membrane. This cleavage, mediated by a cellular protease, occurs at a position immediately after a four-amino-acid motif characterized as basic-X-basic-basic (13, 15, 36, 40, 54). Although the cleaved E3 glycoprotein is not present in Sindbis virus, it is found in the structure of one alphavirus, Semliki Forest virus, as a virion-associated nontransmembranal spike protein.

Despite previous three-dimensional reconstructions of Sindbis virus by electron cryomicroscopy, the organization of the E1 and E2 spikes within the trimer remains unclear (21, 37). Determination of the locations of the functional domains for both of these proteins is important because during specific stages in the virus life cycle, these proteins perform vital functions. In the case of E2, both host cell recognition and attachment have been associated with this protein (14). Antibodies directed against E2 are often neutralizing, and mutations in E2 affect both virus binding and virulence (43, 49). The relative importance of E2 in host cell binding and receptor recognition also has been demonstrated by the identification of mutants in Sindbis virus which fail to bind to chicken embryo fibroblasts (17). An additional function of E2 is the stabilization of spike-nucleocapsid interactions during assembly of progeny and destabilization of these same interactions during the next round of attachment and penetration (8, 28).

E1 appears to maintain the integrity of the icosahedral lattice (2, 3). E1 is folded into a compact metastable structure that is stabilized by intramolecular disulfide bridges, and these are essential for the integrity of the icosahedral lattice of E1-E2 associations (3, 34). In addition to the integrity of the icosahedral lattice, a domain in E1 has been linked with the ability of the virus membrane to fuse with the host cell (23). Thus, the E1 spike protein is probably responsible for the fusion event that follows attachment and introduces the virus nucleocapsid into the cytoplasm.

In this paper, we examined the organization of the glycoproteins within the trimer by three-dimensional structural analyses of wild-type Sindbis virus and two mutants which fail to cleave PE2, leaving E3 as an unambiguous 64-amino-acid structural tag covalently associated with the amino terminus of E2. From this and the work of others, we have inferred the probable locations of E2 and E1 (as well as E3) functional domains within the trimeric spike complex and used these observations to propose a model for virus attachment and penetration.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Sindbis virus and PE2 mutants. In the PE2 mutants of Sindbis virus, PE2 cleavage was prevented by mutation of the +1 position relative to the cleavage site (25). Conversion of this residue from an Arg in the Sindbis virus TRSB wild-type [TRSB (wt)] background to an Asn created a new site for N-linked glycosylation, and the added carbohydrate abrogated PE2 cleavage. As a result, PE2 of the mutant (TRSB-N) was incorporated into virions. The TRSB-N mutation, however, was lethal in that these PE2-containing particles were unable to infect other cells. Revertants of the TRSB-N mutation were isolated after transfection of BHK cells with infectious in vitro transcripts from pTRSB-N (the cDNA clone of the mutant). One such revertant, TRSB-NE2G216, retained the PE2 cleavage defect but acquired a second site mutation (Glu to Gly at E2 position 216) which conferred infectivity on the PE2-containing virion.

Virus purification. The growth and purification of TRSB, TRSB-N, and TRSB-NE2G216 have been previously described. Briefly, infectious virus was harvested from cultures of infected BHK-21 cells and then purified by isopycnic density gradient centrifugation on linear 20 to 35% potassium tartrate gradients as described elsewhere (25, 39). Noninfectious TRSB-N particles were harvested after electroporation of pTRSB-N transcripts into BHK-21 cells. Virus preparations were then concentrated to titers of approximating 10¹² to 10¹⁴ PFU/ml, using Centricon filters.

Radiolabeling and polyacrylamide gel analysis. Viral proteins were ³⁵S labeled by growing virus in the presence of [³⁵S]methionine. BHK-21 cells were infected with virus at a multiplicity of infection of 10 PFU/ml or electroporated with the transcripts from the mutant clone pTRSB-N as described elsewhere (25). The cells were grown for 5 h in complete minimal essential medium and starved in methionine-free minimal essential medium supplemented with 5% donor calf serum and 1% penicillin-streptomycin for an additional 3 h. Once the cells were adequately starved for methionine, [³⁵S]methionine was added to the medium to a final specific activity of 20 µCi/ml. The cells were incubated at 37°C for 12 h in the presence of [³⁵S]methionine prior to virus purification. Labeled virus was pelleted, resuspended in 30 µl of 2× sample buffer, and heated to 100°C for 5 min. Approximately 1.5 × 10⁵ cpm each of wild-type and mutant virus was added separately to two lanes of a sodium dodecyl sulfate (SDS)-10% polyacrylamide gel prepared as described previously (25). The gel was electrophoresed for 20 h at 16-mA constant current, fixed in 40% methanol-10% acetic acid, soaked in Amplify fluorographic solution (Amersham), dyed, and exposed to film.

Specimen preparation. Unlabeled virus specimens were prepared for electron cryomicroscopy by freezing them on copper grids coated with a film of perforated holey carbon, using established protocols (16, 20).

Electron cryomicroscopy. Frozen grids were transferred to a Gatan cryo-specimen holder cooled to approximately 155°C and examined in a JEOL 1200 electron cryomicroscope, where micrographs were recorded at a nominal magnification of 30,000 operating at a voltage of 100 keV, using flood-beam imaging. To prevent radiation damage to the specimens, the micrographs were recorded at low electron doses of 4 to 5 electrons per Ų. Two images per specimen area were recorded, one aimed at approximately 1.5-µm underfocus and the second aimed at 2.2-µm underfocus. Since images viewed closer to focus in an electron microscope contain higher resolution detail, the 1.5-µm micrographs were taken first to reduce the amount of damage caused by electron irradiation. The higher-defocus micrograph, as a consequence, recorded double the exposure, or 8 to 10 electrons per Ų of the specimen area. The second micrographs were taken because higher-defocus images have higher contrast. By processing both negatives separately, we were able to use the orientations from the higher-contrast images to identify the orientations of the closer-to-focus noisier images. Closer-to-focus images were used in the three-dimensional reconstructions.

Image processing and three-dimensional reconstruction. Following electron cryomicroscopy, micrograph focal pairs were selected for computer processing. The criterion for selection was based on virus concentration (greater than 100 per micrograph), uniform ice thickness, and absence of specimen drift and image astigmatism. Selected pairs of electron micrographs were scanned on a Perkin-Elmer microdensitometer with a pixel size of 16 by 16 µm, corresponding to 5.33 by 5.33 Å in the specimen. The digitized images were displayed on a Silicon Graphics workstation, and in the case of the wild-type virus and infectious mutant, individual particle images were interactively boxed out into 256- by 256-pixel² areas. The same particles from both micrographs were boxed out in the same order and given similar file names to identify them as image pairs of the same particles. During the processing of the last data set, that of the noninfectious mutant, an automated particle selection method which enabled us to acquire the boxed images of virus particles in a very short amount of time became available (48). The final images were boxed with the same parameters as the previous data sets. Image defocus was estimated by using the sum of the Fourier intensities of the individual selected particles (55).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Glycoprotein composition of the viral envelope. The protein compositions of both the wild type and the PE2-containing mutants were examined by SDS-polyacrylamide gel electrophoresis. To confirm the presence of PE2 in the mutant and to determine the background amount of E2 that might also be present, we ran ³⁵S-radiolabeled TRSB (wt) together with the infectious revertant on SDS-polyacrylamide gels (Fig. 1). In the lane containing wild-type virus, four bands, corresponding to the capsid protein, the two spike proteins E1 and E2, and a weak PE2 band, were resolved. The wild-type PE2 band is typical and is due to a small amount of uncleaved PE2 that escapes cleavage to E2. Table 1 provides a complete summary of the protein composition, molecular weights, and glycosylation sites of Sindbis virus. As can be seen in Table 1, the E1 and E2 bands have similar molecular weights. Consequently, the E1 and E2 bands were not clearly resolved in the wild-type lane in Fig. 1.

View larger version (36K): [in this window] [in a new window]   FIG. 1.   Protein composition of wild-type and noninfections PE2 mutant (PEwt and PEmut) Sindbis virus. Samples were 35S radiolabeled and analyzed by electrophoresis on an SDS-10% polyacrylamide gel. Lane 1, wild-type Sindbis virus; lane 2, TRSB-NE2G216 (infectious PE2 mutant). Positions of the capsid protein, the two spike proteins E1 and E2, the PE2mut, and a weak PE2wt band are shown. The PE2wt band is typical and is due to a small amount of uncleaved PE2 that escapes cleavage to E2. PE2mut runs as a higher-molecular-weight band due to an extra glycosylated site. The E1 and E2 bands run as a smear in the wild-type lane because of their similar molecular weights (Table 1).

Electron cryomicroscopy. Figure 2 shows selected regions of the higher-defocus micrographs of the wild-type (Fig. 2A), noninfectious (Fig. 2B), and infectious (Fig. 2C) specimens. Figures 2A and B demonstrate that superficially the PE2 mutants and the wild-type virus are not significantly different from one another at the level of virus images on the micrographs. The images of the infectious particles also display characteristics similar to those seen in Fig. 2A and B (higher-magnification comparison of the infectious mutant not shown). However, in the case of the infectious specimen, empty and collapsed envelopes (Fig. 2C, white arrows) and free nucleocapsids (black arrows) are seen significantly more often than in the other two specimens. Free nucleocapsids were observed less frequently than empty or collapsed envelopes, presumably because they quickly disintegrate in the environment.

View larger version (74K): [in this window] [in a new window]   FIG. 2.   Selected regions of 100-kV flood-beam electron cryomicrographs of TRSB (wt) (A), TRSB-N (noninfectious mutant) (B), and TRSB-NE2G216 (infectious mutant) (C). In the case of the infectious mutant, the magnification is lowered to display a larger field of the specimen area. The white arrows indicate collapsed virus envelopes, while the black arrows indicate free nucleocapsids. Black scale bars, 700 Å; white scale bar, 1,000 Å.

Wild-type and mutant structures. The three-dimensional reconstructions (Fig. 3) were each determined from single micrographs corresponding to approximately 1.5-µm underfocus for the wild type, 1.0-µm underfocus for the noninfectious PE2 mutant, and 1.5-µm underfocus for the infectious PE2 mutant. All reconstructions were performed to a nominal resolution of 25 Å, which was within the first contrast transfer function zero for all images. Examination of the mean inverse eigenvalue distribution shows that at 25 Å, all of the mean inverse eigenvalues were less than 0.1 for all three of the reconstructions. Furthermore, the percentage of mean inverse eigenvalues less than 0.01 range from 81 to 85 for the three sets of data. Thus, Fourier space is sufficiently sampled for all three reconstructions.

View larger version (159K): [in this window] [in a new window]   FIG. 3.   Surface representation of three-dimensional reconstructions shown from a threefold view: TRSB (wt), TRSB-N (noninfectious mutant), and TRSB-NE2G216 (infectious mutant). The appropriate contour level for surface rendering was determined for the T=4 virus envelopes by assuming a protein density of 1.325 g/cm3 and by using the published molecular weights of the viral components compiled from the sources shown in Table 1.

View larger version (81K): [in this window] [in a new window]   FIG. 4.   Superimposition of computationally isolated trimers from the wild type and the noninfectious PE2 mutant. The mutant trimer is displayed in wire frame with the blue density denoting 1.325 g/cm3 and the red indicating a denser region of the protein. The wild-type virus trimer is surface rendered in semitransparent yellow. We have assigned the major mutant protrusion that extends outside the wild-type trimer to the E3 region of the mutant PE2.

Differences in wild-type and mutant structures. It is in the morphology of the trimer appendages that the mutant viruses differ significantly from the wild-type virus. Superimposed maps of the wild-type and noninfectious mutant viruses reveal that the mutant appendages have a protrusion, measuring 27 by 22 by 37 Å, which emerges from approximately the middle of each appendage (Fig. 4). The density corresponding to the protrusion can also be observed in twofold equatorial sections of the three-dimensional density maps (Fig. 5B, arrowheads). As can be seen in both Fig. 4 and 5 and as confirmed by difference imaging (data not shown), the protrusion is not present in the wild-type virus structure but is present in both mutant structures. This protrusion is the major difference observed between the wild-type and mutant structures. Comparison of the difference in volume between the mutant structures and the wild type agrees with the volume estimated for the E3 peptide (using the calculated molecular mass of 9.8 kDa). We deduce from this that the E3 regions of the mutant PE2 proteins are the protrusions observed in the trimers of the mutant structures.

View larger version (101K): [in this window] [in a new window]   FIG. 5.   Equatorial sections of the wild-type (A) and infectious mutant (B) structures shown along the twofold axis. The arrows indicate the densities present in the mutant which we have attributed to the E3 region of the PE2 mutant.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Specimen morphology. The collapsed or empty envelopes shown in Fig. 2C (white arrows) were observed in abundance in all micrographs of the infectious (TRSB-NE2G216) revertant. Also present were an unusual number of free nucleocapsids (Fig. 2C, black arrows). The presence of both free nucleocapsids and empty envelopes may be the result of a weakening in the association between the E2 endodomains and the underlying capsid proteins of the revertant, because it is presumably that interaction which is responsible for attachment of the envelope to the nucleocapsid. The putative altered E2-core association might result from a transmembranal effect caused by the second resuscitating mutation on the E2 ectodomain of this mutant. An analogous effect in which a mutation in the capsid protein resulted in the alteration of the stability of the spike ectodomains outside the virus membrane has been demonstrated (28).

Three-dimensional reconstruction. The structures shown here have resolutions similar to those of the recently published structures of Ross River virus and Semliki Forest virus (6, 22). A comparison of these structures reveals that the structures of our PE2-containing mutants are more similar to the structures of both Ross River virus and Semliki Forest virus than to the parental wild-type Sindbis virus (6, 22). Also significant is the observation that E3 cleavage was not required for infectivity of the infectious mutant. It had been proposed that PE2 cleavage is necessary to make the spike trimer complex fusion competent (26). In addition, the trimer pore observed in our reconstructions is unique to our structures and is absent in the structures of Ross River virus and Semliki Forest virus, whose structures were determined to a resolution comparable to that for the structures in this study. Note, however, that the absence of the pore in the previous studies may be the result of different imaging conditions and therefore does not exclude the possibility of the pore's presence in those structures. The observation of a new structural feature such as a pore is not unexpected since with both more accurate refinement of the data and higher resolution of the reconstructions, finer details begin to emerge.

Localization of E3 in alphaviruses. Difference imaging between the wild-type and mutant virus structures shows an additional protrusion on the trimer appendages in the mutant structures. We assign this extra density in the mutants to the E3 protein sequence. Assuming a protein density of 1.325 g/cm³, the volume of the extra density observed in both mutants is consistent with the estimated mass of 9.8 kDa for E3. Discounting this presumptive E3 mass, the remaining density closely conforms to the TRSB (wt) structure. In comparing the structures, we find that there are no other major differences between the structures from the core to a radius of 302 Å. Therefore, the simplest explanation for the additional density is that this represents the E3 amino acids covalently linked to the amino terminus of E2.

The alphavirus spike. The E1-E2 glycoprotein lattice of alphaviruses is a highly organized structure. Structural integrity for this lattice is derived from both the lateral interactions between trimers and the association between the endodomain of the E2 glycoprotein with the capsid proteins of the nucleocapsid (2). Chemical cross-linking studies have revealed that the alphavirus spike is a trimer of E1/E2 heterodimers and that the E1 glycoproteins exist as homotrimers within the complex (2). Fuller et al., however, have suggested a different arrangement where the E2 glycoproteins occupy the center of the spike complex as a triplex and the E1 proteins occupy the spaces between them (22, 51). Their model was derived from the observation that trimeric aggregates of the E1 protein could be recovered from virus after exposure to acid pH. This observation led to a model where the central position of the E2 proteins in the spike prevented interactions among E1 spikes until the virion was exposed to the low-pH environment of an endosome. Upon exposure to low pH, as the Fuller model suggests, the functional domains of the E1 and E2 proteins would "swivel" their position in the spike complex, allowing the E1 glycoproteins to form trimers that would initiate membrane fusion (22). The three-dimensional structure of the virus lattice, however, is not significantly affected by treatments which inactivate virus infectivity, such as low pH and brief treatment with dithiothreitol (3, 22). This finding suggests that although some reconfiguration of the distal portions of the spike trimer may occur after exposure to low pH, the complete reversal of the positions of E1 and E2 is unlikely without the loss of lateral and transmembranal protein-protein interactions that give the virion its integrity.

View larger version (28K): [in this window] [in a new window]   FIG. 6.   Proposed model for the functional domains of the spike proteins and membrane fusion. (A) In the spike trimer, the E2 functional domain is the most outwardly exposed structural protein, while the E1 fusigenic domain is located in the center of the spike complex. E3 is on the outer edge of the E2 spike protein. (B) During attachment, the tripod-like complex forms a specific and stable three-point interaction with the host cell through interactions with cell receptors. These interactions induce a conformational change in the structure of the trimer in which the E2 tips are separated, bringing the center of the spike and the E1 homotrimer closer to the host membrane. As the trimer tips spread apart, conformational changes that expose the fusigenic peptides of E1 occur in the spike proteins. At the same time, the center of the spike containing the E1 trimer is brought close to the host membrane, where the fusigenic peptides of E1 are exposed and initiate fusion.

Conclusions. Difference imaging between the three-dimensional structures of Sindbis virus TRSB and the mutant viruses TRSB-N and TRSB-NE2G216 indicated that the E3 protein is attached to the outer side of the flared portions of the spike trimer, between the distal tips and the center of the spike complex. This observation, along with the epitope mapping studies of Smith et al. (44) on Sindbis and Ross River viruses, strongly suggests that the flaring portions of the spike are composed predominantly of the E2 glycoprotein. By comparison, we also find that the E3 protein in Semliki Forest virus is in a location comparable to the E3 density present in our PE2 mutants and that a similar density exists on the spike of Ross River virus. These observations suggest that Ross River virus may, like Semliki Forest virus, contain E3.