INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Sindbis virus, the prototype of the Alphavirus subgroup of the Togaviridae, has a well-defined T=4 icosahedral structure composed of three structural proteins E1, E2, and C (3, 10, 32, 49). The capsid protein combines with the virus's positive-sense 42S RNA genome in the infected cell cytoplasm and is enveloped in a host-derived membrane bilayer which contains the two virus structural glycoproteins E1 and E2 (7, 8, 10, 47). The E1 and E2 glycoproteins form trimers of heterodimers in the endoplasmic reticulum of the infected cell. The trimers are exported to the plasma membrane, where interaction with nucleocapsids followed by lateral associations between 80 of the trimers produces the icosahedral lattice of spike complexes on the surface of the virus (3, 32, 36, 47). These protein-protein associations are responsible for the precise icosahedral structure of the virus membrane and are stable even in the presence of a nonionic detergent (36). The virus nucleocapsid is situated within the virus membrane and is composed of 240 copies of the capsid protein organized into a T=4 icosahedron matching, precisely, the geometry of the glycoproteins in the virus membrane. The inner icosahedral protein shell is connected to the outer membrane protein shell through specific interactions between the capsid protein and the endodomain of glycoprotein E2 (22, 23).

It has been difficult to study the configuration of the structural glycoproteins in the Sindbis virus membrane bilayer. No crystal structure exists for the membrane glycoproteins, and although the technology has greatly improved, it is unlikely that electron cryomicroscopy will produce images at a resolution sufficient to ascertain protein configurations at atomic resolution. It is important to understand how these proteins are arranged in the lipid bilayer, how they contact one another, which regions of these proteins are exposed on the surface of the spike complex, and which are buried within this structure. This information is important in evaluating domains which are critical for maintaining the structural integrity of the virus and determining the protein domains responsible for virus host interactions, penetration, membrane fusion, and antibody response. Furthermore, Sindbis virus is currently being considered as a vector for human genetic engineering and gene therapy (52). Knowing how its two structural proteins are organized can help produce a more precisely targeted vector and may help prevent unwanted immune reactions.

The precise configuration of the E1 and E2 glycoproteins in the spike complex remains unclear. Enzymatic radiolabeling of exposed tyrosines has indicated that while E1 and E2 have similar numbers of tyrosines in their ectodomains, the tyrosines in E2 are more accessible for labeling than those of E1. Chemical cross-linking studies conducted on the Sindbis virus spike complex in the mature virion have shown that E1-E1 associations stabilize the spike trimer and that the E2 members of the trimer are located on the periphery of the spike, where they form E2-E2 interactions around the fivefold and sixfold axis (3). Electron cryomicroscopy imaging of a PE2-containing Sindbis virus mutant suggested that the E3 region of PE2 lies at the periphery of the spike complex (33). Electron cryomicroscopy of virus complexed with anti-E2 Fab monoclonal antibody fragments showed the Fab fragment to bind on the spike periphery (45). All of these data support an organization of the spike which has E1 centrally located and protected by E2. Others studies have resulted in a model in which the E2 glycoprotein occupies the central portion of the spike and prevents contact between the E1 proteins. This conclusion is based on the observation that multimers of E1 are not detected in detergent-lysed Semliki Forest virus (SFV) until the virus is exposed to low pH. Upon exposure to acid conditions, trimers of E1 can be recovered (39, 50, 51). A morphological study of SFV conducted by electron cryomicroscopy claimed to visualize a process in which the E1 and E2 proteins changed positions in the spike. In this structural reorganization, the E2 protein, which separates the peripheral E1 proteins, moves from the center of the spike to the periphery. This process was called "swiveling" (16).

Several domains of the spike complex have been located on the surface by the technique of escape mutation mapping using monoclonal antibody probes. These studies have shown that (i) the major viral antigenic determinant is on the E2 protein in the region from amino acids 173 to 220 (region 173-220) and (ii) E2 region 186-212 is located on the surface of the spike complex. Also, a charge escape mutation in E1 132 blocked the monoclonal antibody to region 186-212, suggesting that it may be located near this region (46). Finally, it was shown that antibodies responsible for virus neutralization recognized multiple epitopes on E2 and one epitope on E1 (40).

Conformational changes occur in the alphavirus spike as the process of attachment and membrane fusion takes place. The process of virus membrane-cell membrane fusion can be induced by transient exposure of Sindbis virus-cell complexes to acid pH followed by return to neutral pH (1, 12, 13, 25). We have proposed that this in vitro low-pH-mediated fusion is a two-step event. The initial conformational change required for fusion occurs at acid pH; however, return to neutral-pH conditions is required to complete a second step in the fusion process (12). We have proposed that the return to neutrality is required to establish conditions which allow the reduction of disulfide bridges which stabilize the alphavirus structure (1, 4). E1 region 75-98 is the putative fusion domain, and mutations in this domain in SFV (a related alphavirus) abolish fusion activity (19, 20). There are five cysteine residues between the amino terminus of E1 and the putative fusogenic domain. Evidence suggests that these cysteines are involved in disulfide bridges which would produce a complex three-dimensional structure. It is unlikely that this complex structure could penetrate the host cell membrane to position the fusion domains such that a hydrophobic channel could be created. We have proposed that these disulfides must be reduced before the interaction of the fusogenic domain with the target (host cell) membrane (1).

The precise conformational changes produced in the glycoprotein spike by low-pH exposure are unknown. Reported changes include a decrease in sedimentation velocity (13), a change in sensitivity to proteolytic enzymes (4, 13), an increase in hydrophobicity (30), an increase in the radii of the spike complex (48), and a possible swiveling of the E1 glycoprotein from the periphery to the center of the spike complex (16). A recent study using monoclonal antibody probes determined that upon low-pH exposure, a new epitope is exposed and that a mutation in E1 157 will abolish the binding of the antibody to this epitope (2).

To further elucidate the structure of the alphavirus spike, we have used the techniques of limited proteolysis and selective surface chemical modification in combination with mass spectrometry (MS) to determine which regions of the E1 and E2 proteins are exposed on the virus surface. We have also used this technology to determine how protein conformations change when the virus is exposed to low pH. This technique has been successfully used to determine viral protein dynamics in several non-membrane-containing viruses including flock house virus and rhinovirus (5, 6, 24, 43). This technique is based on the principle that sequences of amino acids that are exposed on the surface of the spike complex will be digested by small concentrations of proteases whereas those buried in the spike complex will be inaccessible. The peptides produced are separated from the remaining virus and identified by MS-based sequencing. This technique is fast and accurate and needs only femtomole amounts of peptide to be accurately identified. This procedure can identify all regions accessible on the surface at the same time, in the same experiment, without the necessity to modify the amino acid sequence or bind anything to the virus that could change its conformation.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Materials. All chemicals were HPLC (high-pressure liquid chromatography) or ultra grade and purchased from Sigma Chemical. ¹²⁵I was purchased from New England Nuclear. Sequencing-grade modified porcine trypsin was purchased from Promega and stored at 80°C until use. Peptides for MALDI-TOF (matrix-assisted laser desorption-ionization time of flight) MS mass calibration, iodination, and trypsin controls were purchased from Sigma. Immobilized pepsin was purchased from Pierce. Iodogen (1,3,4,6-tetrachloro-3,6-diphenylglycoluril) tubes and micro BCA (bicinchoninic acid) protein assays were purchased from Pierce. Amicon 50-kDa-cutoff membrane tubes were purchased from Millipore. C₁₈ Zip-Tips were purchased from Millipore. Poros R3 medium was obtained from PE Applied Biosystems.

Virus and cells. BHK-21 cells were cultured at 37°C in Gibco minimal essential medium supplemented with 10% fetal bovine serum, 5% tryptose phosphate broth, 2 mM L-glutamine, and 50 µM gentamicin in 75-cm² flasks. Heat-resistant Sindbis virus (SVHR) was originally provided by E. R. Pfefferkorn and was passaged in BHK-21 cells as previously described (35).

Virus purification. Medium from infected cells was harvested after an 18-h infection. Cellular debris was removed by centrifugation at 5,000 rpm at room temperature for 15 min in a clinical centrifuge. The virus was then purified twice by gradient centrifugation. The first gradient purification was through a 15 to 35% potassium tartrate step gradient in phosphate-buffered saline (PBS-D). The virus band was collected and then purified through a second 10 to 40% linear potassium tartrate gradient in PBS-D. Both centrifugations were done in a Beckman SW28 rotor at 24,000 rpm for 16 h at 4°C. The purified virus band was collected and concentrated by pelleting at 24,000 rpm in a Beckman SW28 rotor through a 20% sucrose cushion at 4°C. The virus was stored at 4°C under sterile conditions for no longer than 14 days. Protein content of the virus pellet was assayed by a micro BCA assay and checked for purity on a mini-sodium dodecyl sulfate (SDS)-12.5% polyacrylamide gel that was subsequently stained using Coomassie brilliant blue R250. The virus was examined by negative-stain electron microscopy to ensure that the virus was intact.

Predigestion of Pierce immobilized pepsin. To remove any free pepsin and to reduce autodigest products, the immobilized pepsin was allowed to autodigest at 37°C for 45 min; then 200 µl of Pierce immobilized pepsin was placed at 37°C. After 45 min, the immobilized pepsin was washed in 1 ml of PBS-D (pH 4.5) three times using a 0.2-mm syringe filter. The immobilized pepsin was then centrifuged at 1,500 rpm, and the excess PBS-D was discarded. The washed pepsin was then placed on ice until use.

Limited virus digestion and peptide purification. Virus was diluted to 5 µg in 50 µl of reaction buffer with PBS-D at either pH 7.6 or pH 4.5. Proteases were added to the virus suspension: either 2 µl of Pierce sequencing-grade modified trypsin at a concentration of 0.52 mg/ml or 2 µl of washed immobilized pepsin. The trypsin reaction was allowed to proceed for various lengths of time at 37°C and then stopped by adding 4 µl of 100 mM TLCK (N-p-tosyl-L-lysine chloromethyl ketone) in 0.1 N HCl. The pepsin digest was carried out at 37°C in an Eppendorf shaker at 1,050 rpm. Removing the immobilized pepsin from the virus stopped the reaction. The digested peptides and immobilized pepsin were removed from the remaining virus by centrifugation through an Amicon 50-kDa-cutoff membrane tube. The peptides were stored at 80°C until MS analysis could be conducted.

MS and MS/MS analysis for trypsin and pepsin digest products. Liquid chromatography (LC)/MS analyses of aliquots of each digest of Sindbis virus were done using an Ultimate capillary LC system (LC Packings, San Francisco, Calif.; equipped with a Famos autosampler) coupled to a quadrupole time-of-flight (Q-TOF) mass spectrometer (Micromass, Manchester, United Kingdom) with a Z-spray ion source. Aliquots of each digest were preconcentrated and desalted onto a guard column (300 µm [inside diameter {i.d.}] by 1 mm; LC Packings) packed with Pepmap C₁₈ material using the Famos autosampler. After the preconcentration step, the guard column was switched in-line with the analytical capillary column. Peptides were then separated using a 75-µm-i.d. by 15-cm capillary column packed with 3 µm of Pepmap C₁₈ material. Mobile phase A consisted of 0.1% formic acid in a 2% acetonitrile solution, while mobile phase B consisted of 0.1% formic acid in an 80-20% acetonitrile-water solution. Peptides were eluted from the column into the microelectrospray ion source of the Q-TOF mass spectrometer using a gradient of 5% to 50% B in 30 min. The outlet of the capillary column was coupled to a platinum-coated 360-µm-outside-diameter- by 20-µm-i.d. fused silica spray tip (10-µm tip i.d. outlet; New Objective, Inc., Cambridge, Mass.) which made electrical contact through the Picotip holder (New Objective) in the Z-spray ion source. MS survey scans were acquired at a rate of 2 per s from m/z (mass-to-charge ratio) 400 to 2,000. The instrument was operated in a data-dependent MS-to-tandem MS (MS/MS) switching mode where peptide ions detected in MS survey scans triggered a switch to MS/MS for obtaining peptide product ion spectra. To identify peptides contained in the digest, uninterpreted peptide product ion spectra were searched against a protein database containing the sequences of bovine trypsin, porcine pepsin, and Sindbis virus capsid, E1, and E2 proteins using the SEQUEST program (14). Alternatively, product ion spectra were searched against the nonredundant protein database using the Mascot search program (Matrix Sciences, Ltd., London, United Kingdom). Searches were done both with and without the protease specificities option turned on.

Iodination of intact virus. Sindbis virus was iodinated with ¹²⁵I or ¹²⁷I using Pierce iodotubes. Iodination with ¹²⁵I was carried out with 1 mCi of ¹²⁵I per reaction sample. The iodination reaction was stopped with the addition of saturated (0.453 mg/ml, 25°C) tyrosine. Excess ¹²⁵I was removed by gel filtration using Sephadex G-25. Radiolabeled virus was stored at 4°C until use. Virus was iodinated with ¹²⁷I using 50 µM K¹²⁵I in PBS-D (pH 7.4). The reaction was stopped by the addition of excess tyrosine and stored at 80°C until use. The virus was denatured by the addition of 1% BME and heated at 99°C for 25 min. The denatured virus was then digested with trypsin for 18 h at 37°C; the peptides were not removed from the remaining virus after the digest was completed.

Digestion with protease V8. The digestion conditions and virus concentration were the same as for the trypsin digest of iodinated and noniodinated virus. V8 was added at 4 µg of V8 per 100 µl of SVHR. The iodinated and noniodinated SVHR was denatured and reduced by heating at 99°C in the presence of 2-mercaptoethanol (BME) (same conditions as for trypsin digest). The digestion was allowed to proceed 16 h at 37°C. Digestion was stopped by freezing the sample at 80°C.

MALDI-MS of iodinated virus peptides. Viral tryptic peptides iodinated with ¹²⁷I were analyzed using a Bruker Proflex linear TOF mass spectrometer equipped with a MALDI source, delayed extraction, and a 1-GHz digitizer. The peptides were desalted with either C₁₈ Zip-Tips or homemade micro-desalting columns using Poros R3 medium; 1.5 µl of desalted peptides were combined with 1.5 µl of saturated -cyano-4-hydroxycinnamic acid in 50% acetonitrile with 0.1% trifluoroacetic acid. The above mixture was applied to the MALDI target using the dried droplet method. Mass calibration was obtained first through an external calibration and then through an internal calibration using diagnostic capsid tryptic masses. The mass spectrometer used in this study typically has a mass accuracy of 0.05 to 0.1% if calibrated internally. Iodinated peptides were compared to noniodinated peptide controls, and differences in the spectra were used to identify the iodinated masses. Comparing these shifted masses to masses computed from a theoretical trypsin digest of the iodinated Sindbis virus structural proteins then identified the peptides with the modified tyrosines. Cases where masses obtained by MALDI-MS coincidentally match more than one possible Sindbis virus peptide are given less weight than masses that have only one possible Sindbis virus peptide match.

SDS-polyacrylamide gel electrophoresis of iodinated and noniodinated virus proteins. Iodinated and noniodinated viral proteins were separated on a 12.5% polyacrylamide gel. The gel was dried and visualized by autoradiography (29). Nonradiolabeled viral proteins were identified by silver staining as previously described (42).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Protein domains exposed on the surface of the virus spike complex at neutral pH. To determine which domains of E1 and E2 are accessible to proteases in an aqueous neutral-pH environment, we performed a number of limited proteolysis experiments on intact Sindbis virus. Limited proteolysis when combined with MS can be used to probe for higher-order structure of proteins and protein complexes (for a review, see reference 21). Freshly prepared Sindbis virus was treated with trypsin or pepsin as described in Materials and Methods. The peptides released from the virus by protease treatment were separated from the partially digested virus and identified by MS as described in Materials and Methods. The released peptides were separated by nanoscale capillary LC which was coupled through a microelectrospray interface to a Q-TOF mass spectrometer. Peptides detected by the mass spectrometer were subjected to MS/MS in order to generate product ion spectra. MS/MS is a technique by which a selected precursor ion is fragmented, and the resulting fragment ions are mass analyzed to yield a product ion (fragment ion) spectrum. Peptide primary sequence can be derived from product ion spectra (31, 37), and computer algorithms can rapidly search protein databases with uninterpreted product ion spectra for matching peptide sequences (14).

View larger version (20K): [in this window] [in a new window]   FIG. 1.   Representative product ion spectra of a peptide released from the Sindbis virus spike complex. Sindbis virus was digested with trypsin for 15 min as described in Materials and Methods. The released peptides were removed from the remaining virus, separated by capillary HPLC, and analyzed by LC/MS with data-dependent MS/MS switching. This peptide was matched to E2 31-44 using the Mascot search engine.

View larger version (50K): [in this window] [in a new window]   FIG. 2.   Tryptic peptides released from the Sindbis virus spike complex at neutral pH. Virus was digested at 1, 5, and 15 min, and the released peptides were removed from the remaining virus, separated by capillary HPLC, and analyzed by LC/MS with data-dependent MS/MS switching. The x axis indicates the amino acid number from the amino to carboxy terminus. No Sindbis virus peptides were found in the 1-min digests.

View larger version (30K): [in this window] [in a new window]   FIG. 3.   Purified Sindbis virus is intact, and capsid protein is not iodinated unless it is lysed (+) with 0.1% SDS prior to iodination. In the nonlysed viral sample (), no capsid is seen, and E2 is iodinated to a greater extent than E1, which is consistent with previous studies.

Protein domains exposed on the surface of the virus spike complex at low pH. Having identified the amino acid sequences accessible to proteases at neutral pH, we next conducted experiments to determine how the spike proteins change conformation when the virus is exposed to low pH. When Sindbis virus is exposed to low pH, a conformational change takes place that is the first step in a two-step process which can lead to the fusion of the virus membrane with a target membrane (12). The second step in membrane fusion requires a return to neutral-pH conditions. We have presented evidence that this step is necessary to allow reshuffling of disulfide bridges which maintain the structure of the virus membrane (1, 13, 25). Fuller et al. (16) have indicated that the conformational change occurring at low pH involves a dramatic reorganization of the virus membrane in which the E1 and E2 proteins change their protein-protein contacts and their physical positions in the protein lattice, a process referred to as swiveling. Swiveling results from the breaking of the heterodimer of E1 and E2, allowing the E1 proteins to move from the periphery of the spike to the center, forming E1 homotrimers (16). These homotrimers then initiate the fusion reaction which in SFV involves the putative fusion domain E1 75-98 (20). This low-pH conformational change renders Sindbis virus noninfectious while largely preserving the overall virus structure (4, 13). If reducing agents are then added to low-pH-treated virus, the virus will disassemble, releasing its internal, capsid protein-complexed, RNA genome (4).

View larger version (46K): [in this window] [in a new window]   FIG. 4.   Peptides released from the Sindbis virus spike complex at low pH. Sindbis virus was digested with immobilized pepsin at pH 4.5 for 1, 5, and 15 min. The reaction was stopped, and the released peptides were removed from the remaining virus, separated by capillary HPLC, and analyzed by LC/MS with data-dependent MS/MS switching. The x axis indicates the amino acid number from the amino to carboxy terminus.

Chemical modification of tyrosines on the surface of the spike complex at neutral pH. Chemical modification of amino acids and the identification of the protein domains containing those modified amino acids by MS have proven to be a powerful tool for the determination of protein conformation. Using iodine activated by immobilized Iodogen, we have modified tyrosines on the surface of the Sindbis virus spike complex (described in Materials and Methods). Activated iodine can also modify histidines if a high concentration of iodine is used and/or the reaction is done at pH 9.0 (38). We used a very small concentration of activated iodine, 50 µm KI at pH 7.4. At these conditions, no histidine is iodinated and only the surfaces of cells and viruses are iodinated (26). Iodine has a mass of 126.9 Da, and the sequence of the virus glycoproteins is precisely known. It is therefore possible to precisely identify those tyrosines which are exposed on the surface of the spike complex. Purified Sindbis virus was iodinated with ¹²⁷I using the lodogen procedure (described in Materials and Methods). The reaction was quenched with an equal volume of cold tyrosine, and the iodinated virus was denatured by adding BME to a final concentration of 1% and heating the mixture at 99°C for 25 min. The denatured virus was then proteolytically digested with trypsin or V8 protease. The resulting digests of iodinated virus were compared to noniodinated control digests by MALDI-TOF MS. In these experiments, aliquots of each digest mixture were analyzed without any separation of the resulting peptides. Peptides containing iodotyrosines were identified by identifying masses in the iodinated sample that had shifted by multiples of 126 Da from the noniodinated controls.

View larger version (19K): [in this window] [in a new window]   FIG. 5.   Representative MALDI-TOF spectrum comparing iodinated and noniodinated Sindbis virus peptides. (A) Comparison of a tryptic digest of Sindbis virus iodinated with 127I (red) to a noniodinated Sindbis digest control (green). (B) Angiotensin III control iodinated with 127I.

View larger version (70K): [in this window] [in a new window]   FIG. 6.   Tyrosines are digested from the viral spike surface, while the spike proteins remain largely intact. Sindbis virus was iodinated with 125I. The iodinated virus was then digested with trypsin for 1, 5, 15, and 90 min, and the resulting mixture was separated on SDS-polyacrylamide gels. Radiolabeled peptides were visualized by autoradiography and compared to the nonradiolabeled viral proteins, which were visualized by silver staining.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The stable and precise icosahedral structure of Sindbis virus makes it an excellent model for studies on the structure of virus membranes. We have used limited proteolysis in combination with MS to determine which amino acids are accessible on the surface of the glycoprotein spike complex and how the conformation of these protein domains change when the virus is exposed to low pH. This MS-based mapping technique has determined the conformational dynamics of two non-membrane-containing viruses, flock house virus and rhinovirus (5, 24). The MS-based mapping not only matched the crystal structure for each virus but also identified protein domains of the viruses that dynamically "breathe." These dynamic conformational changes were not seen or predicted from the viral crystal structures. MS-based mapping is, therefore, able to assay flexible conformational changes in viral proteins that were previously unknown. The study presented here is similar to these previous studies with one major exception: Sindbis virus is a membrane-containing virus, and no crystal structure exists for any of its membrane components.

The data presented above indicate that the configuration of E1 and E2 in the spike complex at neutral pH is most likely one in which E1 occupies the center of the structure and is almost entirely shielded by E2 (Fig. 7). The E2 protein domains found on the surface of the spike are consistent with the observation that E2 173-220 is the major antigenic site on E2 (46). Strauss and coworkers (46) also identified E2 domain 186-212 as exposed at neutral pH. We found that regions 158-186 and 231-260 are located on the surface of the spike. Our experiments may have missed the region between E2 amino acids 186 and 231 because there is a carbohydrate binding site at E2 position 196. Any peptides containing carbohydrate would be missed in our study because of the unknown mass additions and heterogeneity of the carbohydrate. In addition, this region may not have been cleaved efficiently due to steric hindrance caused by the carbohydrate moiety. The three domains of E1 that are accessible and not shielded by E2 are 1-21, 161-176, and 212-220. E1 contains a number of cysteines between amino acids 49 and 114. It is possible that these domains are on the surface of the spike complex, even though these peptides were not detected. It is possible that disulfides constrain the structure in this region to such an extent that trypsin was not able to access potential cleavage sites. E1 and E2 both contain two carbohydrate binding sites, E1 139 and 245 and E2 196 and 318, and as indicated above, peptides containing this modification would have escaped detection. The technique of tyrosine modification by iodination before protease digestion can overcome some of these limitations although iodinated tyrosines located in the same peptide as a carbohydrate will still be missed.

View larger version (22K): [in this window] [in a new window]   FIG. 7.   Amino acid regions of E1 and E2 accessible to proteases at pH 7.4 and 4.5.

Exposure of virus to low pH produced dramatic changes in the outcome of these experiments (Fig. 7). The amount of E1 protein accessible to protease increased from 13% to 35% of the total protein in the E1 ectodomain. The domains of E1 exposed by low pH included regions 86-91 and 110-119, which are near to and overlap the putative fusion domain E1 75-98 described for SFV (20). E1 region 145-158 also became accessible to protease after exposure to low pH. A recent study showed that a charge mutation in E1 157 will block the binding of an E1 low-pH-specific monoclonal antibody (2), indicating that this amino acid, and the region around it, are not exposed in a neutral pH environment. Our data are consistent with these findings and further indicate that region 145-158 is exposed at low pH while E1 region 161-170 is exposed at both neutral and low pH. Region 160-175 is also conserved among the alphaviruses and has a conserved hydrophobic region from amino acids 161 to 166. E1 region 273-287 was also exposed during low-pH exposure. This domain is conserved among the alphaviruses and contains a number of hydrophobic amino acids between E1 276 and E1 287. It is interesting that region E1 161-166 is exposed at both neutral and low pH whereas region 273-287 is exposed only at low pH.

Little is known about the fusion mechanism of Sindbis virus. It is possible that Sindbis virus has a novel fusion motor, a contention supported by its dissimilarity to other membrane-containing viruses. Many viruses such as influenza virus and human immunodeficiency virus employ hydrophobic coiled-coil domains located at the amino termini of glycoproteins in their fusion motors. The sequence of protein around the putative fusion domain in the alphavirus SFV does not contain a predicted coiled-coil region. The E1 glycoproteins containing the putative fusion domain also have 74 disulfide cross-bridged amino acids between the amino-terminal ends of the protein and the fusion domain. Clearly this globular structure cannot be forced through the membrane of a host cell to allow integration of the fusion domain. Thus, the alphaviruses may not follow the influenza virus paradigm regarding the fusion of membranes by virus glycoproteins. The data presented above also indicate that other conserved hydrophobic regions of the E1 protein, probably region 273-287 and possibly region 161-166, may play a part in this fusion process.

Our data indicate that the E2 protein actually becomes less exposed at low pH. This would seem to contradict previous studies that have shown that upon low-pH exposure E2 becomes more accessible to trypsin (3, 13). An explanation for this disparity is that we are identifying only the most superficial domains of the spike complex. Studies indicating an increase in E2 sensitivity at low-pH conditions typically used a large amount of enzyme for a considerable period of time. It is probable that under these conditions, additional conformational changes take place in the E2 protein as a result of the protease digestion itself and that these conformational changes then lead to the complete digestion of E2.

The data presented above support the contention that the organization of the alphavirus spike is one in which the E1 glycoprotein is centrally situated and exists as E1 trimers. The E1 trimers are surrounded by, and protected from the environment by, E2. This model is in agreement with numerous other studies utilizing chemical cross-linkers, antibody binding, electron microscopy, and genetics. It greatly simplifies proposed mechanisms for the interaction of virus glycoproteins with host cells during attachment and penetration. In this model, the virus glycoproteins are not required to change position in the envelope, a difficult process considering the multiple protein interactions which hold the two nested icosahedra together (see Introduction). The observation that E1 trimers are recovered after detergent lysis of virus exposed to low pH (the basis for proposing the alternative model) is likely the result of conformational changes occurring low pH and stabilizing the hydrophobic domains in the centrally located E1 trimer. The stabilized E1 trimers can then be recovered after detergent lysis. Furthermore, the morphological studies involving electron cryomicroscopy which purported to visualize the proposed change in position of E1 and E2 (the second observation leading to the alternative model) compared images of low-pH- and neutral-pH-exposed virus at greatly different resolutions (16), and the conformational changes reported may be due to differences in the resolution of the data sets.