Structure of the Prefusion Form of the Vesicular Stomatitis Virus Glycoprotein G

See allHide authors and affiliations

Science  09 Feb 2007:
Vol. 315, Issue 5813, pp. 843-848
DOI: 10.1126/science.1135710


Glycoprotein G of the vesicular stomatitis virus triggers membrane fusion via a low pH–induced structural rearrangement. Despite the equilibrium between the pre- and postfusion states, the structure of the prefusion form, determined to 3.0 angstrom resolution, shows that the fusogenic transition entails an extensive structural reorganization of G. Comparison with the structure of the postfusion form suggests a pathway for the conformational change. In the prefusion form, G has the shape of a tripod with the fusion loops exposed, which point toward the viral membrane, and with the antigenic sites located at the distal end of the molecule. A large number of G glycoproteins, perhaps organized as in the crystals, act cooperatively to induce membrane merging.

The Rhabdoviridae are enveloped bullet-shaped viruses that are widespread among a great variety of organisms, including plants, insects, fishes, mammals, reptiles, and crustaceans (1). This family includes vesicular stomatitis virus (VSV) as well as notable human pathogens, such as rabies virus (RV) and Chandipura virus (2).

The rhabdoviruses enter the cell via the endocytic pathway and subsequently fuse with a cellular membrane within the acidic environment of the endosome (3). Both receptor recognition and membrane fusion are mediated by a single transmembrane (TM) viral glycoprotein (G) that is trimeric and forms the spikes that protrude from the viral surface. The large ectodomain of G (446 out of 495 amino acids for the VSV Indiana strain) is also the target of neutralizing antibodies, and the antigenic sites of G in both VSV and RV have been described in detail (46).

Similar to other viral fusion proteins, G undergoes a fusogenic structural transition during cell entry (7, 8). As for influenza virus hemagglutinin (HA), flavivirus E protein, and Semliki Forest virus E1 protein, the conformational change is triggered at low pH (9). G can adopt at least three conformational states (7, 8, 1014): the native prefusion state detected at the viral surface above pH 7; the activated hydrophobic state, which interacts with the membrane as a first step of the fusion process (11); and the fusion-inactive postfusion conformation that is antigenically distinct from both the native and activated states. There is a pH-dependent equilibrium between the different states of G that is shifted toward the postfusion conformation at low pH (15). Thus, unlike fusogenic proteins from other viral families, the native prefusion conformation is not metastable (9). Indeed, the reversibility of the low pH–induced conformational change is essential to allow G to be transported through the acidic compartments of the Golgi apparatus and to recover its native functional state at the viral surface (16).

We have recently determined the low-pH postfusion three-dimensional structure of the VSV G ectodomain (residues 1 to 422), generated by limited proteolysis of the virions with thermolysin (Gth)(17). In spite of having an unrecognized fold distinct from those of other fusion proteins previously described, the postfusion conformation of G displays the classic hairpin conformation of other viral fusogenic proteins [i.e., an elongated structure with the fusion domain and the TM domain at the same end of the molecule (18)]. As in class I fusion proteins (1921), the postfusion trimer displays a six-helix bundle with the fusion domains at the N terminus of the central helices and the TM domains at the C terminus of the antiparallel outer helices. Each fusion domain bears two fusion loops located at the tip of an elongated β sheet, which is a marked convergence with class II fusion proteins (2224). Unexpectedly, G turned out to be homologous to glycoprotein gB of herpesviruses, the atomic structure of which was published at the same time (25). Because the low pH–induced conformational change of rhabdoviral G is reversible, it remained unclear to what extent the pre- and postfusion conformations differed for this class of fusion proteins.

Among the different crystal forms obtained with Gth (17) (see also the materials and methods in the supporting online material), one of them, which was grown at pH 8.7, appeared to be particularly notable, because the asymmetric unit could not accommodate the postfusion form (125 Å in length) but was consistent with the presence of one protomer of the prefusion form [8.5 nm in length as measured for the RV G ectodomain by electron microscopy (EM) (26)]. This crystal structure of Gth was determined to 3.0 Å resolution by molecular replacement with the use of domains I, III, and IV (Table 1) of the low-pH form as search models. Data collection and refinement statistics are given in table S1. The structure of Gth is depicted in Fig. 1. Its length (88 Å), the location of the antigenic sites, and the comparison with the low-pH structure indicate that this Gth structure corresponds to the prefusion conformation of the molecule. The chain can be traced up to residue 413 (see the electron density for the final model in fig. S1). Clear density is also present for the first residues of both oligosaccharide chains (on N163 and N320) (27), which were disordered in the structure of the low-pH form.

Fig. 1.

Overall Gth structure in pre- and postfusion conformations. (A) View of the G protomers superimposed on their fusion domains (DIV) and colored by domain (as defined in Table 1) with the fusion loops in green. The two glycosylated asparagines [N163 (labeled “1”) and N320 (labeled “2”)] are displayed as dark green spheres. (B) View of the G trimers, colored by domains as in (A). The trimers were superimposed on the rigid blocks made of DI and the invariant part of DII (RbI-II, defined in Table 1 and highlighted in the boxed inset for one protomer of each conformation). Helix E is indicated on both trimers. (C) Domain architecture of VSV G plotted on a linear diagram, color-coded according to Table 1 with domain boundaries numbered. The unobserved C-terminal segment is in gray, with a checkerboard pattern for the TM domain. The regions that refold in the transition are hatched. All structural figures were generated with PyMOL (38).

Table 1.

Domain nomenclature used in the text. Root mean square deviation (RMSD) is between the pre- and postfusion structures. The number of alpha carbons (Cα) used in superposing the domains is indicated in parentheses.

DomainDomain nameColorResiduesRMSD
DI Lateral domain Red 1 to 17 and 310 to 382 0.42 Å (80 Cα)
DII Trimerization domain Blue 18 to 35, 259 to 309, and 383 to 405 -
DIII PH domain Orange 36 to 46 and 181 to 258 0.40 Å (82 Cα)
DIV Fusion domain Yellow 53 to 172 0.77 Å (94 Cα)
Cter C-terminal part Magenta 406 to 413 -
RbI-II Rigid block - 1 to 25 and 273 to 382 0.56 Å (122 Cα)

The overall architecture of Gth in its prefusion state resembles a tripod (Fig. 1B). Each leg is composed of a fusion domain with the fusion loops pointing toward the viral membrane. The last residues that we can see (including the conserved H407 and P408) pack against the side of the fusion domain. This organization, which is reminiscent of the low-resolution structure of retroviruses' envelope spikes that was recently determined by EM (28, 29), suggests that the TM segments are separate in the membrane. Nevertheless, we cannot exclude the possibility that the missing C-terminal segments of the ectodomain (residues 414 to 446) that lead to the TM segments come together toward the threefold axis.

In the tripod arrangement, the fusion domains are set wide apart, keeping the fusion loops separate (Fig. 1B, left). In contrast to class I and class II fusion proteins, the fusion loops of G are not buried at an oligomeric interface in the prefusion conformation. The hydrophobic residues Y116, A117, W72, and Y73 are exposed (Fig. 1, A and B), even though they cluster near crystal contacts (fig. S3D). The tips of the fusion domains are the most flexible part of the structure (fig. S4) and thus are the least well defined in the electron density maps.

The conformational change involves a dramatic reorganization of the G molecule. Figure S2 shows a comparison of the secondary structure elements of the two conformations with their nomenclature. The pre- and postfusion states are related by flipping both the fusion domain and a C-terminal segment (composed of residues 383 to 413) relative to a rigid block (RbI-II) made by the lateral domain and residues of the trimerization domain that include helix F2 of the prefusion form (Table 1 and Fig. 1B, inset). During the structural transition, both the fusion loops and the TM domain move ∼160 Å from one end of the molecule to the other. Thus, the observed conformational change, although reversible, appears to be similar to that of paramyxovirus F glycoprotein (30). It also suggests that similar intermediates are formed during the fusion-associated refolding of G, HA, and paramyxovirus F glycoprotein (19, 30). In one of these intermediates (Fig. 2C and movie S1), the fusion domain is projected at the top of the spike, allowing the initial interaction with the target membrane.

Fig. 2.

Structural changes in the protomer between the pre- and postfusion conformations and relative movements of domains. In (A) and (B), fragments of the pre- and postfusion conformations are displayed to the left and right, respectively. Secondary structure elements of the prefusion form that refold are named and numbered according to fig. S2. (A) Relative movement of PH (DIII, orange) and fusion (DIV, yellow) domains. The protomers are superimposed on DIII. Hinge residues 47 to 52 (prefusion helix A0) and 173 to 180 (postfusion helix C) are colored in cyan and gray-blue, respectively. (B) Domain II refolding. DI and DIII are omitted in the top panels for clarity but are shown in the bottom panels to provide the relative orientations in the two forms. The protomers are superimposed on the invariant part of DII, which is indicated in dark blue, whereas the three segments that refold and/or relocate are indicated in shades of green. In the prefusion form, strands a1 and y1 form an interchain β sheet. The DIII-DIV hinge (bottom panels) is displayed and colored as in (A), with the two segments connected by a yellow bar to mark the location of the fusion domain. (C) Cartoon representation of the relative organization of domains with respect to the viral membrane during the conformational change. The one-sided black arrows indicate the relative movements of domains. The N- to C-terminal orientations of helices F2 (blue; left), F (blue; middle and right), and H (dark blue; right) are indicated with white arrows. Pre- (left) and postfusion (right) conformations are shown. The trimer axes are indicated. The middle cartoon shows how the fusion loops (in green) would be projected after the refolding of both the DIII-DIV hinge and the DII-DIII connection and before the C-terminal refolding of helix H.

In spite of large rearrangements in their relative orientation (Fig. 2, A and B), domains I, III, and IV retain their folded structure (Table 1 and Figs. 1A and 2). In this and the following paragraphs, we describe the conformational change of a protomer by considering RbI-II as invariant (Fig. 1B, inset). The flippings of both the fusion domain and the TM segment relative to RbI-II occur through a concerted rearrangement of distinct regions of the molecule. Although we have only snapshots of the initial and final states, analysis of the two structures (see the description of movie S1 in the supporting online material) suggests a plausible sequence of events leading from pre- to postfusion conformations.

The fusion domain is projected toward the target membrane through the combination of two movements (Fig. 2C): a 94° rotation around the hinge between the fusion and pleckstrin homology (PH) domains (Fig. 2A) and the repositioning of the latter domain at the top of the trimerization domain (Fig. 2B). The rotation involves the reorganization of two segments (residues 47 to 52 and 173 to 180) of the polypeptide chain. In the former segment, helix A0 unfolds whereas, in the latter segment, helix C forms (Fig. 2A). Mutations M44 → Vor I in RV G, which kinetically stabilize the native conformation (31), map to this region. Their location suggests that they impede the slight distortion of strands b and j of the PH domain that accompanies the movement.

The trimerization domain undergoes a major refolding event during the transition between the pre- and postfusion structures (Fig. 2B). This refolding drives the repositioning of the PH domain and the flipping of the C-terminal part and involves all three segments of the trimerization domain (Fig. 1C).

As a first step, central helix F2 (residues 276 to 294) is lengthened by the recruitment of a segment (made up of residues 263 to 275) to form the long helix F. The second segment that refolds is composed of residues 26 to 35, which, in the prefusion conformation, is buried in a groove of RbI-II that is closed by residues 263 to 275. A sharp bend is introduced right after the conserved motif C24P25: The peptide bond between P25 and S26 flips, which redirects the polypeptide chain at an 80° angle, and short helix A (residues 24 to 29) is formed. The conformation of short strand a1 (residues 22 to 24), involved in the interchain β-sheet a1y1 in the prefusion conformation, is unchanged, although it is not paired to strand y1 of the adjacent protomer in the postfusion conformation (Fig. 2B).

The small β-sheet q1y2 of the native form is then broken, although the individual strands q1 and y2 retain their β conformation in the postfusion form, and residues 384 to 400 (including helices H1 and H2 and strand y1) refold into helix H. This helix then positions itself in the grooves of the central core in an antiparallel manner to form the six-helix bundle. This movement repositions the TM domains at the same end of the molecule as the fusion domains (Figs. 1B and 2B). Finally, residues 259 to 261 and 403 to 405, which are distant by ∼30 Å in the prefusion conformation, form sheet qz that zips together helices F and H in the postfusion state (Fig. 2B). Strands q and z are already in an extended β structure in the native conformation, primed to form sheet qz in the postfusion form.

The buried interface between two subunits in the trimer is 1600 Å2 per protomer, as calculated by the Protein Interfaces, Surfaces, and Assemblies server (32). This value is less than half of that of the buried interface in the low-pH form. This explains the increased stability of the oligomeric structure of G at low pH (8). The interactions between protomers are all located in domain II (fig. S5) but are different from those observed in the postfusion form (Fig. 3, A and B). Not only is prefusion helix F2 shorter than postfusion helix F, it is also tilted and its C-terminal end is kept away from the trimer axis (Fig. 3A). This results from repulsive forces between the carboxylates of the three E286 amino acids (Fig. 3C). In contrast to the postfusion form, the main contribution to trimer stability is not due to the central helix bundle but appears to come from interchain β-sheet a1y1 [which must break during the fusogenic transition, before the formation of helix H (Fig. 2B)] and its environment, burying 1250 Å2 per protomer (Fig. 3D). The conformational change occurs at the viral surface even in absence of the target membrane. This seems to be topologically impossible without transient dissociation of the trimer. This hypothesis is in agreement with the large differences in the trimeric interfaces between the native and the postfusion conformations of G.

Fig. 3.

The trimeric interface of the prefusion conformation [(A), (C), and (D)] as compared to that of the postfusion conformation (B). For clarity, only DII [the only domain involved in the interface in the prefusion conformation (fig. S5)] is represented, and the three protomers are colored in three shades of blue. Secondary structure elements that refold and/or relocate are labeled. (A) Top view (orientation as in fig. S5, looking down toward the viral membrane) of the trimeric interface of the prefusion conformation. The arrow indicates the viewpoint used in (D). (B) Trimeric interface of the postfusion conformation, superimposed on the invariant parts of DII in (A). The view therefore would now be from the membrane. (C) Zoom of image in (A) showing only the three helices F2 and the side chains involved in their interactions, which are colored by atom type (oxygen, red; nitrogen, blue; sulfur, yellow; carbon: green, magenta, or dark blue, depending on the protomer) and labeled. As in the postfusion state, V275 and L279 contribute to hydrophobic stabilizing interactions at the center of the molecule, but L283 now makes a lateral interaction with I278. The three E286 amino acids in the center are 4 Å apart in native crystals. In the YbCl3-derivative crystal used for refinement of the model (table S1), they chelate an ytterbium ion (not shown), bringing their side-chain oxygen atoms within 3.5 Å. (D) Close-up view of the outer region of the prefusion trimeric interface seen from the side. Contact residues are colored as in (C), with main-chain atoms included only when they participate in the contacts. Besides the canonical hydrogen bonds of the β sheet, the interface is stabilized through extended van der Waals contacts and a hydrogen bond between the imidazole ring and the carboxyl group of T265 of the neighboring protomer. Finally, carboxyl groups of L384 and I387 make two hydrogen bonds with the guanidium group of R277 of the other chain. These three hydrogen bonds are displayed as magenta dashed lines.

A number of the few conserved residues (fig. S2) are involved in key networks of interactions that are different in the two forms (Fig. 4). This set of residues includes amino acids D137, Y139, H407, and P408 that cluster together in the postfusion conformation to stabilize β-sheet qz of the trimerization domain (Fig. 4B). In the prefusion conformation, the qz sheet does not exist: D137 and Y139 remain associated with the segment corresponding to the q strand and contribute to a network of hydrogen bonds that also involves conserved W236 of the PH domain (Fig. 4A, top). This network is disorganized during the rotation of the fusion domain relative to the PH domain (Fig. 2A). Conserved histidines—H407 [involved in a salt bridge with D137 in the low-pH structure (Fig. 4B)], H162 [previously shown to be involved in the interactions between fusion domains in the low-pH conformation (17)], and H60—cluster together (H60 is absent in RV G, but H86, which corresponds to S84 in VSV G, replaces it) (Fig. 4A, bottom). Protonation of these residues at low pH is likely to destabilize the interaction between the C-terminal segment of Gth and the fusion domain in the prefusion conformation, priming the initial movement of the fusion domain toward the target membrane. Conversely, the acidic amino acids that were either buried at the trimer interface (D268) or brought close together (D274 with D395 and E276 with D393) in the postfusion acidic conformation (17) are solvent-exposed in the prefusion state (not shown). Thus, the histidines in the prefusion form and the acidic residues in the postfusion form appear to constitute two pH-sensitive molecular switches.

Fig. 4.

Alternative networks of conserved residues in the pre- (A) and postfusion (B) conformations. The orientation is as that in Fig. 2A. Conserved residues are displayed in stick representation (main-chain atoms are not shown unless they participate in interdomain contacts). Hydrogen bonds are displayed as magenta dashed lines. [(A) and (B), top] Close-up views of the DIII-DIV connection are shown. The prefusion hydrogen bonds of Y139 to the main chain of W236 are relocated to the postfusion qz sheet, whereas D137 switches from making a bidentate hydrogen bond to the main chain to engaging in a salt bridge with H407. [(A), bottom] A close-up view of the prefusion DIV-Cter interface that has to be disrupted for DIV to move is shown. Note the cluster of conserved histidines, including H407.

The major antigenic sites of rhabdoviruses are located in the lateral and PH domains (46) (Fig. 5). The accessibility of antigenic sites to antibodies has been studied in detail for RV G. Antibodies directed against RV G site II are unable to recognize the protein in its low-pH conformation (7, 15). Indeed, during the structural transition, this site moves from the top of the molecule to a less accessible location at the surface of the virus. Conversely, the N-terminal epitope of RV G (NS) is only accessible in the low-pH conformation at the viral surface (31). Finally, RV G minor site a is recognized in both conformations (7). As for monoclonal antibody 17D2 (33) that binds only the prefusion conformation (34), its epitope is located in the segment of helix F that is unfolded in the native structure.

Fig. 5.

Antigenic sites of Rhabdoviridae mapped onto the surface of the pre- (A) and postfusion (B) VSV G trimers. Sites are colored on both forms and labeled on the form(s) in which they are recognized. VSV sites are labeled in bold, and RV sites are labeled in italics within parentheses. VSV sites A1 (residues 37 to 38, corresponding to RV antigenic site II located on segments composed of residues 34 to 42 and 198 to 200) and A2 (located at the surface of helix E indicated in Fig. 1) are indicated in shades of red. The RV G site recognized by antibody 17D2 (between residues 255 and 270) is in orange. NS (extending from amino acid 10 to 15) is in dark blue. VSV site B (extending from amino acid 341 to 347), corresponding to RV G minor antigenic site a (amino acid 340 to 342), is in magenta. In the prefusion conformation, the cleft between DI and DIII is colored black. It is flanked by residues 331 and 334, in gray, whose counterparts in RV affect virulence.

The cellular receptor of VSV G has not been identified. Nevertheless, a canyon located between the lateral and PH domains is exposed at the top of the molecule and could be involved in ligand binding (Fig. 5A). In support of this, residues 330 and 333 of RV G, which are involved in the recognition of the putative viral receptor p75 (low-affinity nerve growth factor receptor) (35) and which affect viral pathogenesis, align with residues 331 and 334 of VSV G, which are located at either end of the canyon.

In a previous study, we estimated the minimal number of trimeric spikes involved in the formation of a RV fusion complex as about 15 (15). At the viral surface, a local organization of the spikes resembling the P6 lattice found in the crystal (in which all the spikes are oriented identically, with the major antigenic sites exposed at their tops) (fig. S3) might organize the glycoproteins in an optimal manner for a concerted conformational change. It might also facilitate the formation of the initial intermediates on the fusion pathway. Indeed, the initial lipidic deformations leading to the formation of the stalk and the initial fusion pore (36) can form inside the inner rim of such a hexagon. Reinforcing the idea that the P6 organization may reflect the structure of a fusion relevant complex, a local hexagonal lattice of spikes of similar dimensions has been observed at low temperature under mildly acidic conditions at the surface of some RV G mutants that were affected in the kinetics of their low pH–induced structural transition (31).

It is often considered that fusogenic proteins drive membrane fusion by coupling irreversible protein refolding to membrane deformation (37). At least for rhabdoviral G, this is not the case. Rather, it appears that a concerted cooperative change of a large number of glycoproteins (perhaps organized in a hexagonal lattice, like the one present in the crystals) is used to overcome the high energetic barrier encountered during fusion.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Table S1


Movie S1

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article