The Fusion Glycoprotein Shell of Semliki Forest Virus

Cell, Vol. 105, 137–148, April 6, 2001, Copyright 2001 by Cell Press

The Fusion Glycoprotein Shell of Semliki Forest Virus: An Icosahedral Assembly Primed for Fusogenic Activation at Endosomal pH Julien Lescar,1,2,6,7 Alain Roussel,1,6,7 Michelle W. Wien,1 Jorge Navaza,1 Stephen D. Fuller,3 Gisela Wengler,4 Gerd Wengler,4,5 and Fe´lix A. Rey1,5 1 Laboratoire de Ge´ne´tique des Virus CNRS—UPR 9053 1, Avenue de la Terrasse 91198 Gif-sur-Yvette Cedex France 2 European Synchrotron Radiation Facility BP220 F38043, Grenoble cedex France 3 Division of Structural Biology Wellcome Trust Center for Human Genetics Roosevelt Drive University of Oxford Headington, Oxford, OX3 7BN United Kingdom 4 Institut fur Virologie Justus-Liebig Universitat Giessen 35392 Giessen Germany

Summary Semliki Forest virus (SFV) has been extensively studied as a model for analyzing entry of enveloped viruses into target cells. Here we describe the trace of the polypeptide chain of the SFV fusion glycoprotein, E1, derived from an electron density map at 3.5 A˚ resolution and describe its interactions at the surface of the virus. E1 is unexpectedly similar to the flavivirus envelope protein, with three structural domains disposed in the same primary sequence arrangement. These results introduce a new class of membrane fusion proteins which display lateral interactions to induce the necessary curvature and direct budding of closed particles. The resulting surface protein lattice is primed to cause membrane fusion when exposed to the acidic environment of the endosome. Introduction Enveloped viruses contain membrane-anchored surface glycoproteins that are responsible for receptor recognition and entry into target cells through membrane fusion. Understanding the mechanism of viral fusion is relevant not only to the problem of viral entry but also to many other essential processes that need controlled mem5

To whom correspondence should be addressed (e-mail: rey@ gv.cnrs-gif.fr [F. A. R.]; phone/fax: ⫹49 64 12 39 60 [G. W.]). 6 These two authors contributed equally to this work. 7 Present addresses: (J. L.) CERMAV, CNRS, BP53, F38041 Grenoble cedex, France; (A. R.) AFMB UMR-6098, CNRS et Universites d’AixMarseille I et II, 31 chemin Joseph Aiguier, 13402 Marseille Cedex 20, France.

brane rearrangements (Rothman, 1996; Skehel and Wiley, 1998). In particular, vesicular trafficking relies on continuous membrane fusion and budding steps, which are highly regulated by specific cellular proteins (Jahn and Sudhof, 1999). Enveloped viruses provide simplified systems to study such processes. Studies with the alphavirus Semliki Forest virus were the first to establish the receptor-mediated endocytosis entry route, followed by membrane fusion triggered by endosomal pH (Helenius et al., 1980), a strategy used by many enveloped viruses. Since then, work on viral fusion has led to proposals for a mechanism of action of the fusogenic proteins (Weissenhorn et al., 1999). Viral fusion proteins are maintained in the virion in a metastable state, which often results from proteolytic cleavage of a precursor. The polypeptide chain contains a conserved, mostly hydrophobic and glycine-rich, segment, the “fusion peptide,” which is believed to insert into the target membrane during a conformational change of the protein. This conformational change is triggered either by binding of specific receptors in the target cell or by binding of protons in the endosome. The change in conformation of the protein is believed to be thermodynamically coupled to membrane fusion in such a way that the energy released to reach its most stable fold is used to drive the merging of the lipid bilayers. The majority of the viral fusion proteins that have been studied can be divided into two classes. In class I, the fusion peptide is just C-terminal to the cleavage point in the precursor, and becomes—or is near—the N terminus of the mature fusion protein. In class II, the fusion protein, which bears an internal fusion peptide, is folded in tight association with a second protein as a heterodimer. The activating proteolytic cleavage occurs in the second protein, which is believed to act as a chaperone. A wealth of information is available for class I proteins (reviewed in Skehel and Wiley, 1998). The most studied protein in this class is the influenza virus hemagglutinin (HA), (see Skehel and Wiley, 2000 for a recent review). The so-called “hairpin” arrangement of coiled coils, exposing both the N-terminal and C-terminal membrane inserted segments on the same side of a stable protein rod, has been shown to be the most stable conformation of HA and many other class I viral fusion proteins (see Skehel and Wiley, 1998 for a review). The coiled-coil conformation bringing into proximity two different lipid bilayers is also adopted by cellular fusion proteins, the SNAREs, although in this case the two membrane anchoring segments belong to different polypeptide chains (Sutton et al., 1998). Class II fusion proteins are, in contrast, not predicted to form coiled-coils. The only structure determined to date of a class II fusion protein is the envelope protein E of the tick-borne encephalitis (TBE) flavivirus (Rey et al., 1995). Upon synthesis in infected cells, E associates tightly with prM to form a heterodimer. Proteolytic cleavage of prM, which occurs late in the exocytotic pathway, activates the fusogenic potential of E (Stadler et al., 1997). Like influenza virus and alphaviruses, flaviviruses enter cells via the endocytotic route (Rice, 1996), with the E protein undergoing

Cell 138

Table 1. Data Collection and Phasing Statistics Native I Soaking concentration Soaking time; backsoak Resolution range (A˚) Wavelength (A˚) a (A˚) c (A˚) Observed reflections Unique reflections Completeness (%) Rsym (%) Riso (%) (I,II,III) Number of sites Phasing Power (c/a) FOM (c/a)

Native II Native III U-I (I)

U-II (I)

UIII (II) U-IM

Pt-I (I)

Pt-II (I)

Pt-III (I) Pt-IV (I)

Os (III)

1 mM

1 mM

1 mM 1 mM 0.1 mM 0.1 mM

24 hr

24 hr; 10 mn 24 hr

24 hr

6 hr

6 hr; 30 mn 8hr; 1 hr 30 hr

8 hr

0.1 mM 0.25 mM 0.1 mM

29–3.0 0.99 79.9 334.4 91240

20–3.5 0.72 77.1 334.3 41221

25–3.0 0.93 79.5 335.6 56532

20–3.5 0.72 78.4 338.4 36442

25–4.0 0.96 79.4 340.5 16956

20–3.0 0.72 77.7 335.0 39884

20–3.0 0.93 75.9 332.8 45235

25–3.5 0.99 79.1 333.8 39782

30–3.0 1.07 79.0 334.2 78539

28–3.6 1.07 79.0 334.0 16379

15–4.5 0.96 77.9 335.0 16562

20–3.0 0.93 78.8 335.6 52430

13512 99.0 10.7

8204 98.7 9.2

13107 96.0 7.0

8825 98.2 6.3 32.9 3 .67/.97

4946 82.6 6.5 34.4 2 .83/1.23

8126 98.5 7.1 32.0 1 .53/.78

10798 79.1 7.4 — — —

8408 97.9 13.1 24.3 2 .55/.67

12005 89.7 8.6 29.6 2 .82/1.28

6562 86.5 6.6 23.6 2 .80/1.14

3853 99.7 11.7 36.3 1 .70/1.06

13410 98.2 5.5 18.0 1 .52/.71

— — — — — — 0.73/0.53 0.43/0.30 0.43/0.5

UI-II-III: potassium pentafluoro-oxyuranate, K3UO2F5; Pt-I-II-III: potassium tetrachloro-platinate, K2PtCl4 Pt-IV: potassium tetracyano-platinate, K2Pt(CN)4, Os: K2OsCl6. Values in parenthesis relate to native dataset (I, II or III) to which the derivative crystal is least anisomorphous. Rsym ⫽ ⌺h|⬍I⬎ ⫺ Ih|/⌺h⬍I⬎ where ⬍I⬎ is the mean intensity; Riso⫽ ⌺h|FPH ⫺ FP| / ⌺h | FP |; Phasing power ⫽ ⬍FH⬎/⑀, for centrics (c) and acentrics (a) data respectively (rms isomorphous difference divided by the rms lack of closure). FOM: mean value of figure of merit before density modification (to 3.87 A˚ for form I and II and 4.12 A˚ for form III) (see Supplementary Data at http://www.cell.com/cgi/content/full/105/1/137/DC1 for details).

an irreversible conformational change triggered by low pH (Allison et al., 1995). The E protein is maintained in the virion as a dimer lying flat on the outer surface of the membrane, with the fusion peptide buried at the dimer interface (Allison et al., 2001). Lateral interactions between E dimers stabilize an icosahedral protein shell that covers most of the viral membrane (Ferlenghi et al., 2001). The first step in the conformational change of E is dimer dissociation (Stiasny et al., 1996), thus exposing the previously buried fusion peptide, followed by rearrangement of E into a trimer (Allison et al., 1995). The final conformation of the E trimer is likely to bring into proximity the fusion peptide and the carboxy-terminal

membrane anchor (Ferlenghi et al., 2001), which are located at opposite ends of the molecule in the neutral pH form. Alphaviruses (Strauss and Strauss, 1994) also contain class II fusion proteins. The genus Alphavirus in the Togaviridae family (Schlesinger and Schlesinger, 1996) comprises about 25 different viruses of which Semliki Forest virus (SFV) and Sindbis virus are prototypes. The alphaviral particles have icosahedral symmetry with triangulation T⫽4 (for reviews, see Harrison, 1986; Garoff et al., 1994; Kielian, 1995), with 80 three-lobed projections (often called “spikes”) and a thin protein layer covering most of the viral membrane (Paredes et al., 1993; Figure 1. Experimental Electron Density to 3.5 A˚ Resolution Calculated with Phases Resulting from Averaging Separate MIR, SIRAS, and MAD Maps Obtained for Nonisomorphous Crystals as Explained in the Experimental Procedures Section (A) Overall view of the molecule with contour levels at 3␴ to show the main chain connectivity. The C␣ backbone of E1 is superposed as yellow sticks. (B) The region indicated by a rectangle in (A) was enlarged and contoured to 1.5␴ to show side chains. This region corresponds to the disulfide bridge between residues Cys68 and Cys78. The quality of the electron density map allowed an unambiguous tracing of the polypeptide chain.

Crystal Structure of the Alphavirus Fusion Glycoprotein 139

Figure 2. Overall Fold of E1 and Comparison with TBE Virus E Protein (A) 3D diagram of the folded structure of E1 (top) and of TBE virus E (bottom) in which the three structural domains are displayed in different colors. The same color code is used in all the figures. Disulfide bridges are drawn in green. The conserved stretch of amino acids (83–100 in SFV-E1 and 98–111 in TBE-E), containing the putative fusion peptide, is colored orange. The amino and carboxyl termini are indicated, respectively, by N and C. A yellow triangle marks the glycosylation site (Asn141 in SFV-E1 and Asn154 in TBE-E). (B) 1D diagram to show the overall arrangement of each domain along the primary sequence of E1 (top) and TBE virus E (bottom). Colored regions correspond to the crystallized fragments. The white and the hashed segments correspond to the linker region (also called stem region in TBE virus E protein) and the transmembrane segment, respectively.

Cheng et al., 1995; Mancini et al., 2000). The viral genome, a positive stranded RNA molecule of about 11.4 kb, is packaged within a nucleocapsid formed by the core protein C. The envelope glycoproteins P62 and E1 associate in the endoplasmic reticulum as a heterodimer, which is transported to the plasma membrane of the infected cell. A proteolytic maturation of the P62 precursor into E2 and E3 occurs late during transport to the cellular surface by a furin-like protease. This cleavage causes a change in the conformation of the viral surface (Ferlenghi et al., 1998) and primes it for reactivity upon exposure to acidic conditions. E2 has two functions: its cytoplasmic domain interacts with the nucleocapsid and its ectodomain is responsible for binding a cellular receptor. Most alphaviruses lose the peripheral protein E3, but in SFV it remains associated with the viral surface. The first step upon exposure to low pH is dissociation of E1 from E2 (Wahlberg and Garoff, 1992), followed by formation of an E1 homotrimer (Wahlberg et al., 1992; Gibbons et al., 2000). Electron cryo-microscopy studies (cryo-EM) of SFV have yielded a 9 A˚ resolution picture of the viral particle at neutral pH (Mancini et al., 2000), and the crystal structure of the core protein (Choi et al., 1997) has been placed into the reconstruction. A tentative assignment of the location of the individual polypeptides in the (E1/E2/E3)3 SFV complex assembly by electron microscopy has been attempted (Venien-Bryan and Fuller, 1994). The results presented here, however, revise the interpretation advanced by that work. In this report we present the trace of the SFV-E1 polypeptide chain derived from a 3.5 A˚ electron density map. The overall fold is unexpectedly related to that of the envelope protein (E) of flaviviruses. Each of the three structural domains of SFV-E1 is folded with the same topology as that of the corresponding domain in TBE-E.

The fit of the crystallographic model into the 9 A˚ cryoEM reconstruction is unique, revealing that E1 also lies as a dimer at the interface between adjacent three-lobed projections in the viral surface. The fusion peptide, at the very tip of domain 2, forms tight contacts with E2. The dimer contacts are therefore unlike those of the flavivirus protein, in which the fusion peptide is buried at the interface of the E homodimer. These results also define unambiguously the location of the two transmembrane proteins E1 and E2 in the viral surface. They show that E1 makes a continuous icosahedral protein shell on the virion, covering most of the lipid membrane. The packing of E1 subunits in the virion thus displays overall similarities to the packing proposed for protein E in flaviviruses (Ferlenghi et al., 2001), but there are also clear differences that probably relate to different modes of receptor interaction. Results Structure Determination Crystals of glycoprotein E1 were obtained as described (Wengler et al., 1999). The diffraction from these crystals is markedly anisotropic, reaching a resolution better than 2.5 A˚ along one axis and only 3.5–3.3 A˚ in the perpendicular directions. Different crystals from the same preparation are rarely isomorphous to each other, complicating the process of phase determination by the multiple isomorphous replacement (MIR) method. We therefore used a combination of MIR and multiwavelength anomalous diffraction (MAD) methods to obtain initial phases to about 4 A˚ resolution using several heavy atom derivatives (listed in Table 1). These phases were improved by density modification and averaging between nonisomorphous crystals (see Experimental Procedures) to extend the phases to about 3.5–3.3 A˚. The

Cell 140

different crystal forms exhibit different hinge angles at the connections between domains of the molecule, which arise from an intrinsic functional flexibility of the rod-shaped E1 molecule. The final experimental electron density map (depicted in Figure 1) shows the majority of the side chains and allows an unambiguous interpretation and tracing of the polypeptide chain. The atomic model is currently being refined (see Experimental Procedures). Overall Fold of the Molecule We describe here the main features that result from the interpretation of our experimental electron density map of E1. The refined structure and the higher resolution details of the model will be published separately. Our trace shows that E1 is remarkably similar to the flavivirus fusion glycoprotein E. It contains three domains, each with a secondary structure consisting almost exclusively of antiparallel ␤ sheets. Figure 2A displays the trace of the polypeptide chain of the two proteins oriented similarly. The 1D diagram of Figure 2B is color coded to indicate the distribution of the domains along the primary sequence. As in flavivirus E, the polypeptide chain of SFV-E1 begins in the central domain (domain I, red in the figure), which is an eight-stranded ␤ barrel with the up-and-down topology. Two long insertions between strands in the central domain form the dimerization domain (domain II, yellow). The fusion peptide lies at the tip of this domain, in the more N-terminal of the two loops that form it (see Figures 2A and 2B, orange). Four disulfide bridges stabilize the conformation of this loop. The C terminus of the E1 ectodomain, which leads to the transmembrane segment (see Figure 2B,) is located at the opposite end of the molecule from the fusion peptide. The C-terminal domain (domain III, blue) is an antiparallel ␤ barrel with an Ig-like topology, stabilized by three disulfide bridges. Each of the three domains is thus in the same relative primary sequence arrangement in both the flavivirus and alphavirus fusion proteins and they share the same topology, despite the absence of detectable sequence similarity. However, the relative orientations of these domains are different in the two molecules. A Crystallographic Dimer The soluble ectodomain of E1 is monomeric in solution (Kielian and Helenius, 1985; Klimjack et al., 1994). Under our crystallization conditions (which require only low ionic strength and polyethylene glycol, 0%–5%, at pH 8), the E1 monomers associate as dimers, which constitute the building blocks of the crystals. The crystallographic dimer is displayed in Figure 3, next to the dimer of TBE virus E glycoprotein. In contrast to the latter, where the fusion peptide is buried at the dimer interface, the fusion peptide of E1 remains accessible. The E1 dimer interaction is also head-to-tail, but the monomers are back-to-back instead of facing each other. The lateral view of the E1 dimer shows that the monomers, which are straight rods instead of being gently curved as in TBE-E, cross at an angle of about 20 degrees, with the fusion peptides exposed at the highest end (furthest from the viral membrane, see below).

Figure 3. The Crystallographic Dimer of E1; Comparison with the TBE-E Protein Dimer (A) The E1 dimer (top) and the TBE virus E dimer (bottom) as viewed from above, (or from outside the virion). The orientation is the same as in Figures 1 and 2. Notice that the fusion peptide (orange) is buried at the dimer interface for TBE-E (and protected by the carbohydrate of the adjacent subunit) and exposed for E1. (B) Lateral view of the dimers: E1 (top) and TBE-E (bottom). The two monomers of E1 cross at an angle of about 20 degrees. The E1 fusion peptide is exposed at the highest end.

Fitting into the Cryo-EM Reconstructions of the Viral Particle The SFV particles, which are about 700 A˚ in outer diameter, exhibit 80 3-fold symmetric projections (see Figure 4A) associated by lateral contacts about local 2-fold symmetry axes of the icosahedral T⫽4 lattice. The 9 A˚ resolution 3D reconstruction reveals that the transmembrane sections of the glycoproteins enter the membrane at the edges of a central cavity located underneath each projection (Mancini et al., 2000). A thin protein layer, called “the skirt,” covers most of the area surrounding the bulky projections, leaving holes only at the icosahedral 2-fold axes (which are also local 6-fold symmetry axes) and at the icosahedral 5-fold axes (Figure 4). The cryo-EM density map unambiguously defines the envelope of E1 and allows the positioning of the model determined by X-ray crystallography (see Experimental Procedures). As shown in Figures 4B, 4C, 4E, and 4F, domains I and III fit in the region of the skirt, about local

Crystal Structure of the Alphavirus Fusion Glycoprotein 141

Cell 142

6-folds and 5-fold axes, with the C terminus immediately adjacent to a region of density underneath the skirt which leads directly to the transmembrane segment. The shape of the EM density fits remarkably well the surface features of these two domains, as shown in Figures 4E–4G. The cryo-EM map shows also that the E1 molecule bends at the connection between domains I and II, as shown in Figure 5, in a region that was proposed as a likely “hinge” region in the flavivirus E-protein (Rey et al., 1995). Domain II packs in a dimer interaction with its local 2-fold related counterpart at the interfaces between projections. As shown in Figure 5, the lateral surface of interaction between domains II in the crystallographic dimer is preserved roughly in the virion, but the angle at which the monomers cross is about 50 degrees instead of 20 degrees, with domains I and III closer to the dyad axis. Placing the 4 independent E1 molecules as two dimers about two independent local dyads of the viral particle allows the generation of the whole E1 layer of the virion by applying the icosahedral symmetry operations (see Figure 6). The hinge angles are slightly different at either side of the local dyad, in order to accommodate nonidentical packing environments in the T⫽4 surface lattice. In addition, the E1 subunit that packs by the 5-fold axis displays a slightly different conformation of domain III with respect to domain I. The flexibility built in at the domain interfaces thus allows the E1 molecule to adapt to alternative interactions during assembly of the viral particle. The resulting model shows that the icosahedral scaffold made by E1 covers most of the viral membrane, in a way similar to that proposed for the flavivirus E protein, as shown in Figure 6C.

Having defined the E1 portions of the electron density in the 9 A˚ 3D reconstruction of the viral particle, we can also assign density to E2. The location of the peripheral E3 protein (which is only 62 amino acids long) in the SFV projections was inferred earlier from comparative analyses of cryo-EM reconstructions with alphaviruses that do not contain this protein (Venien-Bryan and Fuller, 1994) and also from comparison between wild-type Sindbis virus and a mutant that retains E3 (Paredes et al., 1998). Both studies hinted to locations of E3 as a protrusion at the highest radii of the particle. Most of the region surrounding domain II of E1 (see Figure 4) therefore corresponds to E2. It is clear in the cryo-EM density that the region of E2 indicated by an orange star in Figure 4D is connected to the end of the C-terminal domain of E1 (blue domain) in the skirt (see Figures 4D–4F). The C-terminal portion of the two polypeptide chains (E1 and E2) thus meet after the end of the blue domain of E1 to form a short stem that leads directly to the transmembrane segment. There is only density for one transmembrane path per E1/E2 heterodimers in the reconstruction, so that the two polypeptides are likely to traverse the lipid bilayer in tight association. The distribution of E2 that we have obtained is in agreement with both its nucleocapsid binding activity through its C-terminal cytoplasmic domain, and with its receptor binding role in the outside. Its shape is more globular than that of the long rod of E1, and connects down to the membrane in a more “vertical” way. The location of the C-terminal portion of E2, and so its cytoplasmic domain, are determined by its interaction with E1. This interaction therefore presents the three nucleocapsid binding sites of each (E1/E2)3 assembly in exactly the

Figure 4. Positioning of the E1 Dimer into the Cryo-EM Reconstruction (A) Surface representation of the SFV structure as determined by cryo-EM to about 9 A˚ resolution (Mancini et al., 2000) depth-cued according to the radial distribution color code of the left. The outer diameter of the viral particle is 706 A˚. The 80 projections lie roughly between 280 and 350 A˚ in radius and the skirt, in the light green region, is at a radius of about 270 A˚. The black rectangle indicates the area enlarged in (B). (B) View down a local dyad of the viral particle as indicated in (A), relating an icosahedral 3-fold axis (bottom left) to a local 3-fold (top right). The cryo-EM density is represented by a white net and the icosahedral symmetry axes are indicated with white numbers. Two E1 monomers have been positioned in the cryo-EM map, with domains I and III (red and blue) in the skirt region. When the red and blue domains of each monomer (at each side of the local dyad) are positioned as shown in the figure, the corresponding domains II come out of EM density and run straight into each other if the conformation found in the crystals were maintained. The long, finger-like domain II is clearly outlined at the edges of the 3-lobed projections, leading to an unambiguous fit, with the fusion peptide (colored orange) positioned near the lateral end of the projections. The dashed lines indicate the slab zone used for (C). (C) View rotated 90 degrees to that in (B) about a horizontal axis, bringing the local dyad into the plane of the figure. A letter C marks the carboxyl terminus of the E1 ectodomain. S indicates the “stem” region connecting the C terminus of domain III to the transmembrane region (TM in the figure) (this region is seen more clearly in panel F). Two arrows at the lower-right end of the panel, blue and red, indicate the outer and the inner leaflet of the viral membrane, respectively. The base of the projection forms an arched cavity over the membrane. The roof of the cavity around the local dyad matches the crossed fit of the two yellow domains as indicated. This fit unambiguously assigns the region just above domain II as being E2 and E3. The region labeled “E2” indicates the putative receptor binding region of alphavirus E2 as visualized by cryo-EM (Smith et al., 1995). (D) View down a local 3-fold axis of the T⫽4 icosahedral lattice. The 5-fold axis and the 2-fold (quasi 6-folds) icosahedral axes are labeled. This figure shows that domain II forms the lower edges of the projections with E2 lying in the region marked by an orange star, which extends to cover E1 as well, as seen in (C). E2 projects down, passing underneath the fusion peptide to reach the skirt and enter the transmembrane region, as shown below. (E) Enlargement of the skirt area around a quasi 6-fold axis, with domains I and III from three adjacent projections. The orange star marks density corresponding to E2, directly connected to the region marked by the same star in (D). The white arrow shows the direction of the view in (F). (F) View orthogonal to (E) along the white arrow. The orange star shows the position of the same density corresponding to E2 indicated in panel (E). It merges with density coming from the C terminus of E1 (at the blue domain) which enters the “stem” region, marked by S. There is a single density crossing the viral membrane, corresponding to both the transmembrane regions of E1 and E2 (marked by TM). (G) Tilted view, in between the orientations shown in (E) and (F), to show roughly the path of the E2 polypeptide. The same E2 subunit occupies the space marked by the orange star within each of the two adjacent projections displayed. It surrounds the region of the tip of domain II to reach down and contact domain III from a different E1 subunit belonging to the same (E1/E2) heterohexamer, with which it interacts through the transmembrane segment as well. This is summarized in the diagram of Figure 6D.

Crystal Structure of the Alphavirus Fusion Glycoprotein 143

right spacing, about 5-folds and quasi 6-folds, for a matching interaction with the core. This is in agreement with reports showing that the presence of E2 alone is not enough for particle production (Barth and Garoff, 1997). At the other end of the E2 molecule, the putative receptor binding region, as visualized by cryo-EM (Smith et al., 1995), lies directly above the region contacting the E1 fusion peptide, about 30 A˚ outward in the radial direction, as seen in Figure 4C. Discussion One major conclusion we can draw from the results described above is that the fusion proteins from flaviviruses and alphaviruses appear to have evolved from a common ancestor. It is possible that this similarity between fusion proteins can be extended to other regular enveloped viruses. Our results indeed support the proposed model of the hepatitis C virus envelope protein E2 which was based on the 3D structure of the flavivirus envelope protein E (Yagnik et al., 2000), especially because these viruses belong to the same viral family and so are likely to have diverged more recently. The alphavirus and the flavivirus fusion proteins share the same fold and their internal fusion peptides lie at the end of the molecule furthest from the transmembrane segment. They both display lateral dimeric contacts that provide stability to the icosahedral protein lattice covering most of the viral membrane. The nature of the contacts is different in the two cases, however, with the fusion peptide buried at the E homodimer interface in flaviviruses and at the E1/E2 heterodimer interface in alphaviruses. These features are in agreement with biochemical data. The flavivirus E homodimer has been shown to dissociate upon exposure to low pH (Stiasny et al., 1996), in the first step of its fusogenic conformational change, and dissociation of the E1/E2 heterodimer is the first step in the conformational change of the alphavirus surface (Wahlberg and Garoff, 1992). In both cases the initial step is therefore exposure of a previously buried fusion peptide. Mutations of residues in the E1 fusion peptide result in decreased stability of the E1/E2 heterodimer (Duffus et al., 1995), in agreement with its interaction with E2. Another conclusion drawn from fitting E1 into the reconstruction of the viral particle is that the fusion protein acts as the major assembly element of the icosahedral shell, through its lateral interactions at the 5-fold and local 2-fold and 6-fold axes. Indeed, E1 exclusively makes the regular scaffold connecting the 80 projections at the viral surface, with interactions between E2 subunits limited to 3-fold related contacts within each projection. The model for the icosahedral shell of E1 presented in Figure 6C is reminiscent of the SFV particles containing E1 only obtained by (Omar and Koblet, 1988). Such particles were shown to be infectious and able to induce cell fusion. Electron micrographs of negatively stained E1-only particles exhibit a rather smooth surface, in agreement with our model. Interestingly, in some alphaviruses, smaller particles with presumably T⫽3 symmetry have also been reported (Rumenapf et al., 1995), suggesting that the built-in E1 flexibility allows it to adapt to alternative packings to form different sur-

Figure 5. The E1 Dimer in the Virion (A) Conformational rearrangement about a hinge region. Domain II of the molecule before and after fitting were superposed. The “moved” domains I and III are displayed in a shaded tone of red and blue, respectively. The blue arrow shows the direction of the movement, which results in a shift of about 12 to 15 A˚ of C␣ carbons at the tip of domain III. (B) The dimer interaction viewed down the local dyad in the viral particle. Comparison with Figure 3A shows that the blue domains have moved inward with respect to the crystallographic dimer, so that they are at about the same distance to the dyad axis as the fusion peptide. (C) Side view of the dimer, showing that the monomers cross at about 50 degrees instead of 20 degrees as in the crystallographic dimer (compare with Figure 3B). The dimer contacts between yellow and red domains as found in the crystal are not present here, but the area corresponding to the interactions between yellow domains is roughly preserved.

face lattices. The arrangement of E1 subunits in the T⫽4 surface lattice is indeed similar to the E/M protein arrangement in flaviviruses in the T⫽3 lattice (Figure 6C), where the M proteins have been mapped to the center of triangles formed by three E dimers (Ferlenghi et al., 2001). The main difference derives from the fact that in alphaviruses, the receptor binding and membrane fusion activities reside in two different proteins, whereas in flaviviruses the E protein is responsible for both. The important difference in the kinetics of fusion, which are much faster for flaviviruses than for alphaviruses (Corver et al., 2000), is likely to be due to the presence of bulky E2 at the center of the (E1/E2)3 complex. In a somewhat analogous way to the HA1 top domains in influenza virus, the ectodomain of E2 has to move out of the way

Cell 144

Crystal Structure of the Alphavirus Fusion Glycoprotein 145

to allow homotrimerization of E1 and membrane fusion to occur. It is clear that the stage of the process visualized by the time-resolved cryo-EM study (Fuller et al., 1995) was a very early step. Our results indicate that the detected rearrangement of the viral surface must be reinterpreted as movements of E2. Dissociation of E2 would lead to an unstable state of E1 in which the fusion peptides are transiently directed toward the target membrane, corresponding to a prefusion intermediate state as postulated for HIV gp41 and influenza HA (Weissenhorn et al., 1997; Skehel and Wiley, 1998). Our results show that there is flexibility at the hinge regions of E1. This flexibility is used for the purpose of assembly into the T⫽4 icosahedral particle in which the four independent subunits are necessarily in nonidentical environments. It is noteworthy that the hinge region between domains I and II lies at the center of the molecule, with about the same distance from the hinge to the fusion peptide and to the C terminus leading to the membrane anchor. The E1 homotrimer (Wahlberg et al., 1992) that is formed as a consequence of the conformational change was shown to be very stable as judged by its thermal denaturation properties and protease resistance (Gibbons et al., 2000). It is possible that the hinge region is used in a similar foldback mechanism that leads to the coiled-coil hairpin of class I fusion proteins, with the fusion peptide and the transmembrane segment at the same end of a stable protein rod. The important enhancement in stability of the final trimer suggests that the conformational change is likely to involve more than mere rearrangements about hinge regions, however. Whatever the nature of the final conformation of E1, the transition would result in arrangements of five or six fusogenic trimers surrounding an initial fusion pore, similar to the rings proposed for influenza HA (Chernomordik et al., 1999). A similar model can be proposed for the flavivirus E protein trimer resulting from an analogous conformational change.

A major common property of the alpha- and flavivirus fusion proteins is their function as a scaffold to drive budding and viral assembly. Contacts between adjacent proteins in the surface lattice introduce curvature in the budding particle, as shown by the flavivirus capsidless subviral particles (Russel et al., 1980; Schalich et al., 1996). In the case of alphaviruses, capsidless particles have not been observed and budding appears to rely on contacts between the cytoplasmic domain of E2 and the core protein. Nevertheless, the recent observation that lateral interactions between glycoproteins at the viral surface can overcome defects of nucleocapsid assembly in a core protein mutant (Forsell et al., 2000) points to an important role of the glycoprotein layer in budding. The class II fusion proteins have thus a role similar to that of the cellular COP proteins, which allow budding of vesicles by introducing on its outer surface the necessary curvature to close the final vesicular assembly. In this respect, the regular enveloped viruses act as extracellular coated vesicles that can deliver their cargo, the viral genome, from cell to cell. The SNAREs are the only known cellular fusion proteins today, and are functional analogs of the viral class I fusion proteins. Whether cellular homologs of viral class II fusion proteins exist, with a dual coating and fusogenic role, remains to be determined. Experimental Procedures Data Collection and Structure Determination This work derived from previous observations on the solubilization of the Semliki Forest virus membrane proteins (Helenius and Von Bonsdorff, 1976; Kielian and Helenius, 1985), which prompted us to purify and crystallize protein E1 (Wengler et al., 1999). The crystallization conditions and the characterization of the crystals have already been described (Wengler et al., 1999). The space-group is P6422 and the cell parameters are listed in table 1 (lines 5 and 6). Heavy-atom derivatives were prepared by soaking the crystals into solutions of different heavy-atom salts. All crystallographic measurements were done at 100 K. Table 1 gives a summary of the

Figure 6. Overview of the E1 Glycoprotein Shell (A) Lateral view displaying all of the different layers of the viral particle. The icosahedral 2-fold axis lies vertically in the plane of the figure, at the center. The nucleocapsid, labeled N, surrounds the genomic RNA region (labeled R). The viral membrane is indicated by letter M, with the skirt (S) layer and the projections (P) above it. The C␣ trace of E1, filling the skirt shell and the base of the surface projections, is displayed within the wire-frame representing the cryo-EM density. (B) View down a local 2-fold axis of the T⫽4 icosahedral lattice. The icosahedral 2-, 3- and 5-fold axes are indicated by yellow numbers. A “3” thus marks one of the 20 (E1/E2/E3)3 assemblies at the icosahedral 3-folds. Local dyads relating an icosahedral 3-fold to a local 3-fold axis are indicated by blue ovals and those relating two local 3-folds by red ovals. Two E1 dimers form the icosahedral asymmetric unit of the T⫽4 particle, one lying on a “blue” local dyad, the other on a “red” one. This figure shows that there are no 3-fold related contacts between E1 subunits within each projecting glycoprotein complex. (C) The T⫽4 icosahedral protein layer formed by E1 (left) and the T⫽3 layer proposed for TBE virus E protein (right, from Ferlenghi et al., 2001). White numbers denote the icosahedral symmetry axes. Local 3-folds in each lattice are indicated by solid white triangles. Each polypeptide chain is drawn as a C␣ trace colored according to the scheme of Figures 2 and 4. The most important difference in the two cases is that in SFV the fusion peptides stick up to pack against E2 whereas in the flavivirus the fusion peptides are held down and pack against the domain I/III interface of the dyad-related monomer. Another difference is that in the flavivirus particle the E protein makes 3-fold related contacts, which is not the case for E1. The bar is 100 A˚. (D) Diagrammatic representation of the contacts. Drawn is one of the 20 triangular faces of the icosahedron enclosing the SFV viral particle and the distribution of E1 and E2 subunits. The T⫽4 surface lattice implies a subtriangulation of this face into four quasi-equivalent triangles. Black numbers denote strict icosahedral symmetry axes. Open triangles mark local 3-folds and open ovals local 2-folds. Blue and red ovals are the same as those described in (B). Each E1 monomer is represented as a rod made of three domains, color-coded as in Figure 2. The orange stars mark the fusion peptide. E2 is drawn in white: solid and dashed lines outline E2 molecules disposed about strict or local 3-fold axes, respectively. For clarity, only the outline of the projecting domain of E2 is shown in the center and the top-right subtriangle. A black asterisk marks the C-terminal end, next to the transmembrane region, of each subunit. A full white ellipse indicates the connection of the top of E2 with its transmembrane region. The subunits have been labeled in the top-right subtriangle, as (A), (B), and (C) in black and white letters for E1 and E2, respectively. This is to indicate that subunit A of E2 contacts subunit A of E1 around the fusion peptide and subunit B of E1 in the C-terminal region, and so on, within the same heterohexameric complex.

Cell 146

X-ray data collection statistics. All diffraction images were processed using program DENZO (Otwinowski and Minor, 1996). Integrated intensities were scaled using either SCALEPACK (Otwinowski and Minor, 1996) or SCALA from the CCP4 program suite (CCP4, 1994). The diffraction pattern of the E1 crystals is anisotropic, extending to better than 2.5 A˚ along the c* hexagonal axis, and only to about 3.5 A˚ in radial directions. To partially compensate for this effect, we applied an empirical correction factor to the structure factor amplitudes of the native data, based on the observed difference between the intensity fall-off along c* and the perpendicular directions. We screened over 25 heavy-atom compounds and collected over 60 diffraction data sets. The lack of isomorphism between the different data sets, as indicated by the variations of the unit cell dimensions (Table 1), hampered the determination of the heavy-atom binding sites from the “isomorphous”-difference Patterson function. Data collected from a crystal soaked in K3UO2F5 salt at the LIII Uranyl absorption edge (UI, see Table 1), yielded the first interpretable anomalous-difference Patterson map, revealing two major sites, one of them located at a special position, on a 2-fold axis. Phases calculated from this derivative (with program MLPHARE; CCP4, 1994) were then used to identify the sites in the other derivatives by cross-Fourier calculations. A molecular envelope was determined by visual inspection of the 4 A˚ resolution MIRAS electron density map. Program DMMULTI (Cowtan, 1994) was used to combine phase information from different nonisomorphous crystals and phase extension, using a total of 8 crystal forms for which three contained experimental phases. The averaging transformations between the crystal forms were determined for each domain with program AmoRe (Navaza, 1987). Program O (Jones et al., 1991) was used to inspect and modify the solvent masks used for averaging. An example of the resulting averaged electron density map which was used for building the polypeptide chain (with program TURBOFRODO; Roussel and Cambillaud, 1991) is shown in Figure 1. An atomic model for E1 was built in the electron density map and is currently being refined with program CNS (Bru¨nger et al., 1998) against reflections measured at higher resolution. This model contains amino acids 1–381, with two loops connecting strands in the C-terminal domain (amino acids 310–314 and 346–351) which are disordered and therefore not present in the model. For clarity, in all figures we have included arbitrary connections to cover the space occupied by the disordered loops, to show roughly the path of the polypeptide chain. The current free R factor (Bru¨nger, 1997) is 33.7% (for 10% of the diffraction data—1166 reflections—randomly chosen) and an R factor of 27.5% for 10,353 reflections between 15 and 3.0 A˚ resolution and a tightly restrained stereochemistry.

Fitting into the Cryo-EM Reconstruction of the Viral Particle We tried initially to fit the crystallographic dimer (Figure 3) from crystal form I about the local 2-fold symmetry axes of the T⫽4 particle as a rigid body. The quality of the 9 A˚ cryo-EM reconstructions revealed immediately that there were adjustments in the conformation of the molecule, which were different to the changes found in the different crystal forms. Because the EM density defines remarkably well the envelope of the region of E1 corresponding to domains I and III (red and blue, respectively, in Figure 1), we fitted these domains manually into the corresponding density. The outline of domain II is also very clear in the reconstruction. This domain was manually rotated about the hinge region to fit its density, as indicated in Figure 5A. Fitting one monomer allowed placing of its local dyad partner by applying the corresponding symmetry operation. The hinge rotation was found to be slightly different in the second monomer, likely to accommodate nonidentical environments for the dyad-distal domains. The initial fit of the molecule was optimized by maximizing the correlation between structure factors calculated from the cryo-EM reconstructions and those calculated from the model in the 400–10 A˚ resolution range, using the procedure described in Mathieu et al. (2001). For this purpose, a mask was calculated from the model at the initial position to envelop out density not corresponding to E1. Each E1 subunit was divided into two rigid bodies, one containing domains I and III and the other containing domain II. The fourth subunit, with domain III about a 5-fold axis, was divided into three rigid bodies, each corresponding

to each of the three domains defined in Figure 1, resulting in a total of 9 rigid bodies refined simultaneously. To minimize errors introduced by the original enveloping out of adjacent subunits, the procedure generates half a shell of the virion for the calculation of the structure factors. Each rigid body is forced to obey the icosahedral symmetry. The calculations included all coefficients between 400 A˚ and 10 A˚ resolution (about 350,000 coefficients). The final optimized values were R ⫽ 33.8% (48.7%) and a correlation coefficient of 89.6% (83.8%) for data between 400–15 A˚ and (in parenthesis) data between 400–10 A˚ resolution for the superposition shown in Figure 4. Graphic Representations for Figures Figure 1 was prepared with program TURBO-FRODO (Roussel and Cambillaud, 1991); Figures 2A, 3, 4B–4G, and 5 with program RIBBONS (Carson, 1987); Figure 4A with program AVS (Sheehan et al., 1996); and Figure 6 with program O (Jones et al., 1991). Acknowledgments We thank S. Harrison, M. Kielian, F. Heinz, and Y. Gaudin for comments on the manuscript; S. Bressanelli, F. Coulibaly, M. Mathieu, S. Duquerroy, and R. Kahn for help with data collection; J. Janin for support and encouragement; M.-C. Vaney and K. Cowtan for help modifying DMMULTI to work with 8 crystal-forms simultaneously; B. Arnoux and A. P. Jaudier for assistance with the computer work; M. Carson for adapting program RIBBONS to display virus structures; members of the ESRF/EMBL “Joint Structural Biology Group” for help and useful discussions. Diffraction data were collected at the synchrotron sources of ESRF (Grenoble, France), DESY (Hamburg, Germany) and LURE (Orsay, France). This work was supported by CNRS-ATIPE, PCV, SESAME, ARC, and FRM grants as well as HFSP grant RG102/96 to F. A. R. in addition to grants by the Deutsche Forschungsgemeinschaft (Projekt We 518) and by the Fonds Der Chemishen Industrie to G.W. J.L. is a scientific collaborator with the European Synchrotron Radiation Facility. Received December 26, 2000; revised March 15, 2001. References Allison, S.L., Schalich, J., Stiasny, K., Mandl, C.W., Kunz, C., and Heinz, F.X. (1995). Oligomeric rearrangement of tick-borne encephalitis virus envelope proteins induced by an acidic pH. J. Virol. 69, 695–700. Allison, S.L., Schalich, J., Stiasny, K., Mandl, C.W., and Heinz, F.X. (2001). An internal fusion peptide in the flavivirus envelope protein E. J. Virol, in press. Barth, B.U., and Garoff, H. (1997). The nucleocapsid-binding spike subunit E2 of Semliki Forest virus requires complex formation with the E1 subunit for activity. J. Virol. 71, 7857–7865. Bru¨nger, A.T. (1997). Free R Value: Cross-validation in crystallography. Methods in Enzymol. 277, 366–396. Bru¨nger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography & NMR system: A new software suite for macromolecular structure determination. Acta Crystallogr. D54, 905–921. Carson, M. (1987). Ribbon models of macromolecules. J. Mol. Graphics 5, 103–106. CCP4 (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D50, 760–763. Cheng, R.H., Kuhn, R.J., Olson, N.H., Rossmann, M.G., Choi, H.K., Smith, T.J., and Baker, T.S. (1995). Nucleocapsid and glycoprotein organization in an enveloped virus. Cell 80, 621–630. Chernomordik, L.V., Leikina, E., Kozlov, M.M., Frolov, V.A., and Zimmerberg, J. (1999). Structural intermediates in influenza haemagglutinin-mediated fusion. Mol. Membr. Biol. 16, 33–42. Choi, H.K., Lu, G., Lee, S., Wengler, G., and Rossmann, M.G. (1997). Structure of Semliki Forest virus core protein. Proteins 27, 345–359. Corver, J., Ortiz, A., Allison, S.L., Schalich, J., Heinz, F.X., and

Crystal Structure of the Alphavirus Fusion Glycoprotein 147

Wilschut, J. (2000). Membrane fusion activity of tick-borne encephalitis virus and recombinant subviral particles in a liposomal model system. Virology 269, 37–46. Cowtan, K.D. (1994). DM program. Joint CCP4 ESF-EACBM Newsletter 31, 34–38. Duffus, W.A., Levy-Mintz, P., Klimjack, M.R., and Kielian, M. (1995). Mutations in the putative fusion peptide of Semliki Forest virus affect spike protein oligomerization and virus assembly. J. Virol. 69, 2471– 2479. Ferlenghi, I., Gowen, B., de Haas, F., Mancini, E.J., Garoff, H., Sjoberg, M., and Fuller, S.D. (1998). The first step: activation of the Semliki Forest virus spike protein precursor causes a localized conformational change in the trimeric spike. J. Mol. Biol. 283, 71–81. Ferlenghi, I., Clarke, M., Ruttan, T., Allison, S.L., Schalich, J., Heinz, F.X., Harrison, S.C., Rey, F.A., and Fuller, S.D. (2001). Molecular organization of a recombinant subviral particle from tick-borne encephalitis virus. Mol. Cell 7, 593–602. Forsell, K., Xing, L., Kozlovska, T., Cheng, R.H., and Garoff, H. (2000). Membrane proteins organize a symmetrical virus. EMBO J. 19, 5081– 5091. Fuller, S.D., Berriman, J.A., Butcher, S.J., and Gowen, B.E. (1995). Low pH induces swiveling of the glycoprotein heterodimers in the Semliki Forest virus spike complex. Cell 81, 715–725. Garoff, H., Wilschut, J., Liljestrom, P., Wahlberg, J.M., Bron, R., Suomalainen, M., Smyth, J., Salminen, A., Barth, B.U., Zhao, H., et al. (1994). Assembly and entry mechanisms of Semliki Forest virus. Arch. Virol. Suppl. 9, 329–338. Gibbons, D.L., Ahn, A., Chatterjee, P.K., and Kielian, M. (2000). Formation and characterization of the trimeric form of the fusion protein of Semliki Forest Virus. J. Virol. 74, 7772–7780. Harrison, S.C. (1986). Alphavirus structure. In The togaviridae and flaviviridae, S. Schlesinger and M. Schlesinger, eds. (New York: Plenum Press), pp. 21–34. Helenius, A., and Von Bonsdorff, C.H. (1976). Semliki Forest Virus membrane proteins. Preparation and characterization of spike complexes soluble in detergent-free medium. Biochim. Biophys. Acta 436, 895–899. Helenius, A., Kartenbeck, J., Simons, K., and Fries, E. (1980). On the entry of Semliki forest virus into BHK-21 cells. J. Cell Biol. 84, 404–420. Jahn, R., and Sudhof, T.C. (1999). Membrane fusion and exocytosis. Annu. Rev. Biochem. 68, 863–911. Jones, T.A., Zhou, J.-Y., Cowan, S.W., and Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47, 110–119. Kielian, M. (1995). Membrane fusion and the alphavirus life cycle. Adv. Vir. Res. 45, 113–151. Kielian, M., and Helenius, A. (1985). pH-induced alterations in the fusogenic spike protein of Semliki Forest virus. J. Cell Biol. 101, 2284–2291. Klimjack, M.R., Jeffrey, S., and Kielian, M. (1994). Membrane and protein interactions of a soluble form of the Semliki Forest virus fusion protein. J. Virol. 68, 6940–6946.

Paredes, A.M., Brown, D.T., Rothnagel, R., Chiu, W., Schoepp, R.J., Johnston, R.E., and Prasad, B.V. (1993). Three-dimensional structure of a membrane-containing virus. Proc. Natl. Acad. Sci. USA 90, 9095–9099. Paredes, A.M., Heidner, H., Thuman-Commike, P., Prasad, B.V., Johnston, R.E., and Chiu, W. (1998). Structural localization of the E3 glycoprotein in attenuated Sindbis virus mutants. J. Virol. 72, 1534–1541. Rey, F.A., Heinz, F.X., Mandl, C., Kunz, C., and Harrison, S.C. (1995). The envelope glycoprotein from tick-borne encephalitis virus at 2 A resolution. Nature 375, 291–298. Rice, C.M. (1996). Flaviviridae: The viruses and their replication. In Virology, B.N. Fields, ed. (Philadelphia: Lippincott–Raven), pp. 931–959. Rothman, J.E. (1996). The protein machinery of vesicle budding and fusion. Protein Sci. 5, 185–194. Roussel, A., and Cambillaud, C. (1991). Turbo Frodo. In Silicon Graphics Geometry (Mountain View, CA: Silicon Graphics). Rumenapf, T., Brown, D.T., Strauss, E.G., Konig, M., Rameriz-Mitchel, R., and Strauss, J.H. (1995). Aura alphavirus subgenomic RNA is packaged into virions of two sizes. J. Virol. 69, 1741–1746. Russel, P.K., Brandt, W.E., and Dalrymple, J.M. (1980). Chemical and antigenic structure of flaviviruses. In The Togaviruses, R.W. Schlesinger, ed. (New York: Academic Press), pp. 501–529. Schalich, J., Allison, S.L., Stiasny, K., Mandl, C.W., Kunz, C., and Heinz, F.X. (1996). Recombinant subviral particles from tick-borne encephalitis virus are fusogenic and provide a model system for studying flavivirus envelope glycoprotein functions. J. Virol. 70, 4549–4557. Schlesinger, S., and Schlesinger, M. (1996). Togaviridae: the viruses and their replication. In Virology, B.N. Fields, ed. (Philadelphia: Lippincott–Raven), pp. 825–841. Sheehan, B., Fuller, S.D., Pique, M.E., and Yeager, M. (1996). AVS software for visualization in molecular microscopy. J. Struct. Biol. 116, 99–106. Skehel, J.J., and Wiley, D.C. (1998). Coiled coils in both intracellular vesicle and viral membrane fusion. Cell 95, 871–874. Skehel, J.J., and Wiley, D.C. (2000). Receptor binding and membrane fusion in virus entry: the influenza hemagglutinin. Annu. Rev. Biochem. 69, 531–569. Smith, T.J., Cheng, R.H., Olson, N.H., Peterson, P., Chase, E., Kuhn, R.J., and Baker, T.S. (1995). Putative receptor binding sites on alphaviruses as visualized by cryoelectron microscopy. Proc. Natl. Acad. Sci. USA 92, 10648–10652. Stadler, K., Allison, S.L., Schalich, J., and Heinz, F.X. (1997). Proteolytic activation of tick-borne encephalitis virus by furin. J. Virol. 71, 8475–8481. Stiasny, K., Allison, S.L., Marchler-Bauer, A., Kunz, C., and Heinz, F.X. (1996). Structural requirements for low-pH-induced rearrangements in the envelope glycoprotein of tick-borne encephalitis virus. J. Virol. 70, 8142–8147. Strauss, J.H., and Strauss, E.G. (1994). The alphaviruses: gene expression, replication, and evolution. Microbiol. Rev. 58, 491–562.

Mancini, E.J., Clarke, M., Gowen, B.E., Rutten, T., and Fuller, S.D. (2000). Cryo-electron microscopy reveals the functional organization of an enveloped virus, Semliki Forest virus. Mol. Cell 5, 255–266.

Sutton, R.B., Fasshauer, D., Jahn, R., and Brunger, A.T. (1998). Crystal structure of a SNARE complex involved in synaptic exocytosis at 2.4 A resolution. Nature 395, 347–353.

Mathieu, M., Petitpas, I., Navaza, J., Lepault, J., Kohli, E., Pothier, P., Prasad, B.V.V., Cohen, J., and Rey, F.A. (2001). Atomic structure of the major capsid protein of rotavirus: implications for the architecture of the virion. EMBO J., in press.

Venien-Bryan, C., and Fuller, S.D. (1994). The organization of the spike complex of Semliki Forest virus. J. Mol. Biol. 236, 572–583.

Navaza, J. (1987). On the fast rotation function. Acta Crystallogr. A 43, 645–653.

Wahlberg, J.M., and Garoff, H. (1992). Membrane fusion process of Semliki Forest virus. I: Low pH-induced rearrangement in spike protein quaternary structure precedes virus penetration into cells. J. Cell Biol. 116, 339–348.

Omar, A., and Koblet, H. (1988). Semliki Forest virus particles containing only the E1 envelope glycoprotein are infectious and can induce cell-cell fusion. Virology 166, 17–23.

Wahlberg, J.M., Bron, R., Wilschut, J., and Garoff, H. (1992). Membrane fusion of Semliki Forest virus involves homotrimers of the fusion protein. J. Virol. 66, 7309–7318.

Otwinowski, Z., and Minor, W. (1996). Processing of X-ray diffraction data collected in oscillation mode. In Methods Enzymol., C.W. Carter and R. M. Sweet, eds. (London: Academic Press), pp. 307–326.

Weissenhorn, W., Dessen, A., Harrison, S.C., Skehel, J.J., and Wiley, D.C. (1997). Atomic structure of the ectodomain from HIV-1 gp41. Nature 387, 426–430.

Cell 148

Weissenhorn, W., Dessen, A., Calder, L.J., Harrison, S.C., Skehel, J.J., and Wiley, D.C. (1999). Structural basis for membrane fusion by enveloped viruses. Mol. Membr. Biol. 16, 3–9. Wengler, G., Wengler, G., and Rey, F.A. (1999). The isolation of the ectodomain of the alphavirus E1 protein as a soluble hemagglutinin and its crystallization. Virology 257, 472–482. Yagnik, A.T., Lahm, A., Meola, A., Roccasecca, R.M., Ercole, B.B., Nicosia, A., and Tramontano, A. (2000). A model for the hepatitis C virus envelope glycoprotein E2. Proteins Struct. Funct. Genet. 40, 355–366. Protein Data Bank ID Code The coordinates of the C␣ trace of glycoprotein E1 have been deposited and the PDB accession code is 1I9W.