Structure of a trimeric variant of the Epstein–Barr virus glycoprotein B

Abstract

Epstein–Barr virus (EBV) is a herpesvirus that is associated with development of malignancies of lymphoid tissue. EBV infections are life-long and occur in >90% of the population. Herpesviruses enter host cells in a process that involves fusion of viral and cellular membranes. The fusion apparatus is comprised of envelope glycoprotein B (gB) and a heterodimeric complex made of glycoproteins H and L. Glycoprotein B is the most conserved envelope glycoprotein in human herpesviruses, and the structure of gB from Herpes simplex virus 1 (HSV-1) is available. Here, we report the crystal structure of the secreted EBV gB ectodomain, which forms 16-nm long spike-like trimers, structurally homologous to the postfusion trimers of the fusion protein G of vesicular stomatitis virus (VSV). Comparative structural analyses of EBV gB and VSV G, which has been solved in its pre and postfusion states, shed light on gB residues that may be involved in conformational changes and membrane fusion. Also, the EBV gB structure reveals that, despite the high sequence conservation of gB in herpesviruses, the relative orientations of individual domains, the surface charge distributions, and the structural details of EBV gB differ from the HSV-1 protein, indicating regions and residues that may have important roles in virus-specific entry.

Herpesviruses are enveloped DNA viruses that infect a wide range of vertebrates and are classified into 3 primary subfamilies (α, β, and γ) (1). The life cycle of herpesviruses consists of lytic and latent phases, during which distinct types of host cells are infected through entry pathways that are virus and cell-type specific. For example, EBV, a γ-herpesvirus, enters epithelial cells and cultured B cells by direct fusion with the cell membrane, whereas endocytosis is required for efficient infection of primary human lymphocytes. HSV-1, an α-herpesvirus, can enter cells by direct cell surface fusion, although pH-dependent and pH-independent endocytic pathways have been reported as well (2). These findings suggest that herpesviruses may have developed multiple entry strategies to ensure efficient infection of various targeted cell types. However, regardless of the entry pathway, fusion of the viral and cell membranes has to occur to allow the release of virion capsids into the cytoplasm prior to the genome being replicated in the nucleus.

The core fusion machinery of herpesviruses is well conserved and consists of the envelope glycoprotein B (gB) and a heterodimeric complex of gH/gL (3). Although numerous studies have indicated that both gB and gH/gL exhibit varying degrees of inherent fusogenic properties (2), the maximal level of fusion is typically achieved only when the 3 proteins act together. In addition to the core fusion glycoproteins, gB, gH, and gL, herpesviruses also express glycoproteins that determine viral tropism and mediate receptor-dependent activation of membrane fusion. HSV-1 gD binding to receptors on target cells serves as a fusion trigger (4), and an analogous role in EBV is contributed by gp42, whose binding to HLA class II molecules on B cells is essential for fusion (5, 6). The mechanism by which receptor engagement of the targeting proteins such as gp42 or gD transmits an activation signal to the gB and gH/gL fusion machinery is not understood.

Structural knowledge of the proteins participating in herpesvirus fusion is essential for understanding the molecular mechanism of the entry process. A cryo-electron tomography study of HSV-1 fusion intermediates described gross changes occurring in the virus and target cell during the course of fusion (7), whereas atomic resolution structures of HSV-1 gD alone and in complex with herpes virus entry mediator receptor (8, 9), and of EBV gp42 alone (10) and bound to HLA Class II (11), have provided more detailed snapshots of the initial stages of the fusion mechanism. Also, the crystal structure of the HSV-1 gB ectodomain revealed an unexpected structural homology with the fusogenic protein G (G) of vesicular stomatitis virus (VSV) (12, 13).

Glycoprotein B is the most conserved of all of the herpesvirus envelope glycoproteins, and protein sequences of HSV-1 and EBV gB share 29% identity and 43% similarity. Despite the high conservation, gB proteins of HSV-1 and EBV exhibit a number of different functional properties. For example, although EBV gB is predominantly localized in the perinuclear space and exhibits low cell-surface expression and virion incorporation, HSV-1 gB is an abundant envelope protein (14). In addition to being involved in fusion, EBV gB is also essential for viral maturation and egress (15), and gB-null virus cannot be produced (16). In the case of HSV-1, nuclear egress is blocked only if both gB and gH/gL are absent (17). These distinct functional features emphasize the importance of determining structures for even highly conserved proteins such as gB.

EBV is an oncogenic virus, associated with various cancers including Hodgkin's lymphoma, Burkitt's lymphoma, and nasopharyngeal carcinoma. Also, EBV can cause severe proliferative disorders in patients with immune dysfunction. Understanding the EBV fusion mechanism could stimulate the development of therapeutic agents for treatment of EBV infections that can be life-threatening in patients with a compromised immune system. We have previously demonstrated that the hydrophobic EBV gB residues WY^112–113 and WLIW^193–196, which localize to putative fusion loops (FLs) (18), cause the recombinant EBV ectodomains to form aggregates reminiscent of postfusion rosettes observed for class I and II type proteins (19). The HSV-1 gB ectodomain that was used for crystallization (12) did not form rosette structures, but rather simple trimers. By substituting the EBV residues with the corresponding residues from HSV-1 gB, we produced a trimeric EBV gB ectodomain variant, which was then crystallized. Here, we report a 3.2-Å resolution crystal structure of the trimeric EBV gB ectodomain, and the resulting implications for understanding the herpesvirus membrane fusion and entry mechanism.

Results and Discussion

Production and Crystallization of the EBV gB Ectodomain.

The EBV gB ectodomain variant EctoS FL contains residues 23 to 685 (numbers correspond to the residues of immature protein). The truncation at residue 685 was made just before the beginning of 3 hydrophobic segments, the last of which (residues 733–753) serves as the EBV gB membrane anchor. Replacement of wild type, hydrophobic EBV residues that form 2 putative FLs (WY^112–113 and WLIW^193–196) with the HSV-1 residues (HR^177–178 and RVEA^258–261, respectively) [supporting information (SI) Fig. S1], prevents protein aggregation and allowed production of trimeric protein. Expression and purification of the recombinant protein was done as described previously (19).

The EBV gB ectodomain was crystallized and the structure solved by molecular replacement by using the HSV-1 gB structure as a search model. The EctoS FL crystals belong to the P321 space group, they contain 1 gB molecule in the asymmetric unit, and diffract to a resolution of 3.2 Å. The statistics of the data collection, processing, and refinement are given in SI Materials and Methods and Table S1.

Structure of the EBV gB Ectodomain.

The EBV gB ectodomain reveals an elongated rod-like molecule, composed of 5 domains that contain both β-sheet and α-helical secondary structures (Fig. 1A). This architecture is distinct from known class I and II type fusion proteins and their associated structural features. The domain structures found in gB, VSV G, and recently in the baculovirus fusion protein gp64, led to their classification as a novel, class III type of viral fusion protein (20, 21).

Fig. 1.

Structure of the EBV gB Ectodomain. (A) The gB subunit is colored in blue to red, from N and to C terminus, which are labeled N and C, respectively. Numbers indicate N-glycosylated Asn residues, for which a single NAG molecule is modeled. FLs are marked and point in the same direction as the gB C terminus. The ectodomain shown here extends into a fairly hydrophobic, ≈40-residues long, stem region, and a single transmembrane domain, which were removed from the construct used for expression of the recombinant gB due to their high hydrophobicity. The 5 domains of EBV gB are indicated with roman numbers I to V, and are defined and labeled after nomenclature established for the HSV-1 gB ectodomain (Fig. S2 and Table S3). (B) The trimeric gB ectodomain is shown. The subunits are colored as magenta, cyan, and green. The subunit shown on A corresponds to the molecule shown in magenta. The subunits wrap around each other, forming extensive trimerization surfaces.

The N and C termini of EBV gB are set ≈10 nm apart. The C terminus is located at the same end of the molecule as the putative EBV gB FLs (Fig. 1A). The 3 gB subunits wrap around each other by interacting through multiple contact surfaces to form a spike-like trimer (Fig. 1B). The trimer is ≈7-nm wide and ≈16-nm long, which is consistent with our previous electron microscopy observations (19).

Electron density was observed for gB residues 42–390 and 448–679. For the N-terminal residues 23–41, C-terminal residues 680–685, and for residues 391–447, which contain an internal furin cleavage site, electron density was not observed, indicating that these regions are disordered. The furin-cleavage site of EBV gB (residues 428–432) is cleaved in the Hi-5 insect cells used for expression of recombinant gB (19), similarly to what has been found in vivo (22). It is likely that the cleavage at the furin site gives rise to 2 flexible segments, consistent with the absence of electron density for residues 391–447. Also, this region is proline-rich, and thus probably unstructured, and it is much longer in EBV gB than in HSV-1 or any other human herpesvirus (Fig. S1). The importance of the EBV segment comprising residues 391–447 for EBV gB function is not known, but the high abundance of Pro, Ser, and Thr residues strongly suggests a flexible region, which could be modified by O-linked carbohydrates. The corresponding and much shorter region in HSV-1 gB (residues 460–499) was also not observed and was cleaved by trypsin before crystallization (12).

In addition to the above regions of the gB structure for which there was no observable electron density, other segments of the gB chain exhibited discontinuous electron density of lower quality. Residues 72–74, 544–549, 562–564, and 580–589 fell into this category, and these residues were omitted from the final structure. The latter 3 segments belong to domain IV, which is located on the top of the trimer spike (Fig. 1B). The packing of EBV and HSV-1 gB trimers in their respective crystal lattices is different, and the EBV domain IV has significantly higher average temperature factors (Table S2) than other EBV domains and the corresponding HSV-1 domain.

EBV gB has 9 predicted N-glycosylation sites, and recent evidence suggests that gB is modified by high mannose and complex N-linked sugars (23). Electron density was observed for Asn residues 163, 290, and 629, and a single N-acetyl glucosamine (NAG) molecule is built at each position (Fig. 1A).

Comparison of the EBV and HSV-1 gB Ectodomain Structures.

HSV-1 gB is composed of 5 domains (12), and the domain organization and domain boundaries of EBV gB are similar (Fig. 2A and Fig. S2). As expected, the domains that have a higher degree of protein sequence conservation show better superposition than the ones with higher sequence variability. For example, domains III and IV, which are the most conserved parts of the gB sequence (≈60% of their residues in EBV gB show conservation of 70% or more when compared with the HSV-1 domains), have rmsd values <2 Å for superposition of their C_α atoms, whereas the least conserved domain I has rmsd of 5 Å (≈50% of residues are conserved 70% or more) (Table S3).

Fig. 2.

Comparison of the EBV and HSV-1 gB Ectodomains. (A) Surface representations of EBV and HSV-1 gB trimers. The most obvious difference in the overall conformation of EBV and HSV-1 gB is the location of domain IV. The domains I and II of the subunit colored in magenta were superimposed and used as a reference point (they occupy the same position in both models). The domains IV assume different locations in EBV and HSV-1 trimers, as emphasized by the arrow. (B) Major structural differences between domains I, II, and III are shown based on their superposition of HSV-1 gB (gray) onto EBV gB (cyan) ectodomain. Domains I, II, and III were superimposed as rigid bodies. The green arrow indicates the rotation of domain IV, also illustrated in A. The yellow arrow points to an additional structural difference in strands β7 and β8 (EBV gB residues 144–157), which are shifted forward and down in EBV gB compared with the HSV-1 molecule. (C) Rotation of the upper part of the central αC (domain III) and domain IV, as seen from a point close to the dashed line drawn in A (coloring is the same as in B). (D) Structural differences in domain I: 2 EBV gB subunits, belonging to the same trimer, are shown in cyan and green, and the superimposed HSV-1 gB is shown in gray. The yellow arrow points to the movement of β7 and β8 of one EBV gB subunit (cyan), which allow the packing of the C-terminal αF of another EBV gB subunit (green). Also indicated is Asn-163 and the attached NAG molecule.

The major difference in the gross domain arrangement of EBV and HSV-1 gB is reflected in the change of position of the upper part of the trimer (composed of domain IV, and segments of domain III and V) relative to the rest of the molecule (Fig. 2A). Domains I, II, and the beginning of domain III of EBV gB can be considered as a rigid body and can be superimposed with the analogous HSV-1 structures (Fig. 2B). In contrast, the upper part of the domain III, the entire domain IV, and the part of domain V that precedes the C-terminal helices E and F, are rotated clockwise when viewed from the central α helix C (αC) (Fig. 2 B and C). Also, the C-terminal helix αF, belonging to domain V, adopts a more bent conformation in HSV-1 than in EBV gB. Because αF of 1 gB subunit packs in a crevice formed by the other 2 subunits, the straighter EBV gB helix is accommodated by a shift in strands β7 and β8 (residues 144–157) of domain I (Fig. 2 B and D). This segment has poor conservation (Fig. S1) and 1 or 2 residue length variation in human herpesvirus gB sequences. This variability may reflect a requirement to accommodate differently packing C-terminal helices, or may exemplify a part of the gB conformational change required to drive membrane fusion. Additional structures of gB molecules are required to assess these possibilities.

Although the domains of EBV gB share a similar overall fold with the corresponding HSV-1 domains, both proteins exhibit distinct structural details and unique electrostatic landscapes (Fig. 3).

Fig. 3.

Electrostatic surfaces of EBV and HSV-1 gB ectodomains. (Top) View of domain IV from the top. (Middle) Side view of the trimeric spikes. (Bottom) View of domain I and FLs from below. The views shown at the Top and Bottom are obtained by a 90° rotation from the orientations in the Middle toward and away from the reader, respectively. Despite high conservation of the secondary structure elements and folds of individual domains, electrostatic surfaces of the EBV and HSV-1 gB ectodomains are unique. The electrostatic potential of the EBV trimer tip containing the same 6 FL residues (HR and RVEA) as HSV-1 gB (Bottom) carries no resemblance to the HSV-1 electrostatic landscape, due to contribution of the surrounding EBV residues. The center of the 3 positively charged channels in domain I of EBV gB is indicated by the arrow (Bottom). Electrostatic surfaces were generated by PyMol and are contoured on scale from −2 to +2 kT/e (red corresponds to negative charge, and blue to positive).

Description of EBV gB Domains.

Domain I (residues 89–294) contains the gB FLs, and its core region has a fold that resembles that of a plekstrin-homology (PH) domain (Fig. S2). HSV-1 domain I residues 331–337 were not resolved in the structure, because of trypsin cleavage at residue 334 that gave rise to flexible ends. The electron density for the corresponding EBV region (residues 263–269) is observed, and the residues are built as an extended loop that connects strands β15 and β16 (Fig. 2D).

Domain II is composed of residues 77–88 and 295–390, and has a PH domain fold. Although the best characterized members of the PH-domain fold family mediate association with membranes through high-affinity binding to phosphoinositides, the vast majority of PH domain-containing proteins do not function as membrane-binding modules, but rather interact with protein ligands (24). A mutagenesis study of EBV gB revealed that although linker insertions at position 88, 263, 353, and 361 did not affect cell-surface expression of gB, they abrogated the ability of the variant proteins to mediate fusion (23). It is interesting that these charged and hydrophilic residues map to the PH domains and are located at the exposed surfaces (Fig. S3), suggesting that these residues may be defining binding sites for ligands, such as other EBV envelope glycoproteins involved in entry or, possibly, a gB receptor.

The most prominent feature of domain III is the 42-residue long αC helix, which wraps around the helices from other 2 subunits in a left-handed twist. Domain III is composed of residues 52–68, 455–527, and 617–624 (Fig. S2).

Domain IV is made of residues 528–616 and a short N-terminal region consisting of residues 42–51. The segment containing residues 42–46 was not resolved in the HSV-1 structure, and, in EBV gB, these 4 N-terminal residues form an extended structure that points downwards (Fig. 1A) and makes multiple contacts with domain III residues 511–514, and residues 64–65 of domain III of the neighboring subunit. Domain IV of EBV gB appears to be more disordered and potentially dynamic as compared with the HSV-1 gB structure (Table S2). It is interesting that the analogous, albeit smaller, domain IV of baculovirus gp64 seems to be poorly defined as well (21), suggesting that a certain degree of disorder in domain IV might be a common feature of class III fusion proteins.

Last, residues 625–679 form the extended domain V, which reaches across and inserts into the cavity formed by other 2 subunits, tying together the gB trimer.

Putative FLs of EBV gB.

Two segments located at one tip of the gB structure are analogous to the bipartite fusion peptide of VSV G, and were proposed to form the FLs of HSV-1 gB (residues 173–179 and 258–265) (12). We previously demonstrated that substitution of the analogous EBV residues, WY^112–113 and WLIW^193–196, abolishes the ability of gB to mediate fusion, suggesting strongly that these 6 residues function as EBV gB FLs (18). Mutational studies of HSV-1 gB showed that the gB residues Trp¹⁷⁴, Tyr¹⁷⁹, and Ala²⁶¹ are also essential for the function of gB (25).

It is interesting to note that the electrostatic surfaces of the crystallized trimeric EBV and HSV-1 gB spikes are very different, despite the grafting of HSV-1 FL residues onto EBV (Fig. 3 Bottom). The EBV trimer exhibits 3 highly positive charged channels, which spread radially from the center, between the exposed FLs. It is possible that while the putative FL residues of EBV penetrate the membrane, the positively charged crevices may have a role in association with negatively charged phospholipid heads. The electrostatic differences between EBV and HSV-1 may also explain why simple grafting of the fusions loop sequences from HSV-1 to EBV gB disrupts the fusion activity of the protein. The FL regions may interact with target membranes in different ways in the HSV-1 and EBV gB proteins, which may be sensitive to the local structural context in each protein.

When domains I of EBV and HSV-1 gB are superimposed, the C_α atoms of residues forming the tips of putative FL 1 (EBV gB HR^112–113) occupy similar positions (Fig. S4A). However, there is a marked difference in the conformation of the extended loop region comprising residues 190–193, which just precedes the second FL (RVEA^193–196) (Fig. S4B). This difference is likely due in part to the insertion of a Gly residue at position 192 in EBV gB. Gly¹⁹² in EBV gB may allow EBV FLs to adopt an optimal conformation for membrane interactions that is nonoptimal for the HSV-1 loop residues. These observed structural differences may thereby also contribute to the lack of activity in the EBV gB FL mutants. In addition to these comparisons, which are based on the experimentally observed data, a theoretical model of EBV gB containing the wild type residues WY^112–113 and WLIW^193–196 was also created and illustrates the hydrophobic patches that these residues would form at the tip of the wild-type molecule (Fig. S4C).

Structural Homology of the EBV gB Ectodomain and VSV G.

Structures of VSV G in both a pre and a postfusion state have been determined (13, 26), and the structure of HSV-1 gB (12) and EBV gB are likely to represent a postfusion conformation based on the structural homology with the latter form of G. Despite the lack of protein sequence homology between gB and G, the folds of their domains are remarkably similar, as well as their 3D arrangement (Fig. 4Upper). Although the significant structural homology indicates a potential common evolutionary origin, the distinct structural characteristics of gB and G point to the features each protein may have acquired to perform their functions in virus-specific contexts. For example, each of the first 4 gB domains is larger than the corresponding domain of G, and contains additional secondary structural elements (G does not have a domain analogous to the gB domain V). As a result, the gB ectodomain is composed of ≈710 residues, whereas the G ectodomain is ≈440-residues long. The larger gB structure might be related to the more complex entry machinery of herpesviruses that requires cooperation of gB, gH/gL, and receptor-binding proteins, unlike the fusion apparatus of VSV, which is comprised solely of G.

Fig. 4.

Comparison of VSV G and EBV gB structures. (Upper) Postfusion conformations of VSV G and EBV gB. The structures are colored blue to red from their N to C termini. Domains of gB and G in their postfusion conformations have similar secondary structure topologies, although each of the gB domains is larger, and gB has an additional domain V, which is not present in G. Dashed arrows indicate the missing stem regions, and connect the C termini of the models (marked with letter C) with the expected transmembrane domain position. The prominent αF of VSV G, and the analogous αC of gB, are colored in yellow and indicated with arrows. Disulfide bridges of G, Cys²⁴-Cys²⁸⁴, and gB, Cys⁶⁸-Cys⁴⁸⁴, are marked and shown as magenta sticks, connecting the αF and αC, respectively, with the corresponding N-terminal segments (shown in blue). (Lower) The prefusion structure of VSV G and a theoretical model of the prefusion form of EBV gB. In the prefusion conformation of G, only domain III undergoes a significant refolding event (thus, marked as domain III*), which involves breaking of the long αF, whereas the other 3 domains relocate to a different position, but preserve their primary folds. Domain III of gB was modeled by using the domain III* of G as a guide. Dashed arrows represent the direction of stem regions in case of G, and domain V and stem regions for gB.

Domain I presents the bipartite fusion peptides of VSV G, WY^72–73, and YA^116–117, and EBV gB (wild type residues are WY^112–113 and WLIW^193–196). Both G and EBV gB loops contain hydrophobic residues consistent with a role in membrane penetration, and these residues are located at the tip of domain I. However, the relative orientations of these FLs are distinct, with the gB domain I and FLs tilted at an angle with respect to the trimer axis.

Domain IV of gB and G is made entirely of β-sheets, and is located to one side of αF of domain III of G, whereas in gB it assumes a position on top of αC of domain III. In gB, the β-domain is composed of 6 and 3 strands, and the corresponding domain in G contains 4 and 5 strands. In HSV-1 gB, there are 4 conserved cysteines, which form 2 disulfide bonds (Cys¹¹⁶-Cys⁵³⁷ and Cys⁵⁹⁶-Cys⁶³³). In EBV gB, the region comprised of residues 580–589, and containing Cys⁵⁸⁸, is not built, but, because the 4 cysteines are conserved in EBV gB, the same disulfide bonding pattern would be expected. In domain IV of G, there are no corresponding disulfide bonds.

Domain V of gB is made of an extended segment and C-terminal helices αE and αF. The analogous domain does not exist in G, and, in the postfusion conformation, domain IV of G continues into a 17-residues long αH, which is assigned as part of domain III. Domain V of gB inserts into a groove formed by 2 other gB subunits contributing significantly to the trimer interface. This arrangement is different in postfusion VSV G, where the C-terminal segment flips back more locally to interact within the same subunit of the G trimer. It is possible that the postfusion trimer of VSV G has fewer trimerization contacts (which are mostly formed through the helices of domain III) to preserve some conformational flexibility, so it can revert to the prefusion form. In gB, the packing of the final αF helix appears to tie together the gB C terminus and FLs into closer proximity. This interaction may be functionally analogous to the N-cap structure observed in the influenza hemagglutinin HA2 trimer, which forms at the end of the HA2 coiled coil and is energetically important for driving membrane fusion. Peptides mimicking the C-terminal helix of human cytomegalovirus were shown to inhibit viral entry (27), possibly by preventing the formation of a postfusion gB trimer. Analogous EBV gB peptides could serve as a basis for design of therapeutic agents that might prevent EBV spread.

C-terminal helices of gB and G are followed by ≈40 residues that are hydrophobic, membrane-proximal stem regions. For VSV G, these residues have an important role in the fusion process (28). The stem regions are absent from the gB and G ectodomain structures determined to date.

Comparative Analysis of the Domains III of EBV gB and VSV G Indicates gB Residues that May Have a Role in Conformational Switching.

The long, central coiled-coil helix (αC) of domain III in gB is 10 residues longer than the corresponding helix αF in VSV G, lengthening the gB molecule significantly (16 nm) compared with G (12.5 nm). A disulfide bridge (Cys⁶⁸-Cys⁴⁸⁴) in gB ties the extended N-terminal region (between strands β2 and β3) and αC together (Fig. 4 Upper), and an analogous disulfide bridge is present in VSV G (Cys²⁴-Cys²⁸⁴). This disulfide bridge remains intact during the transition of G from pre to postfusion conformations, whereas domain III undergoes significant refolding. The long αF of the postfusion form of G (analogous to the αC of gB) is broken into 2 shorter helices in the prefusion form. The point at which the VSV αF breaks (residues 272–274) is ≈10 residues N-terminal to Cys²⁸⁴, and the analogous region in EBV gB contains Gly⁴⁷⁷, a residue that destabilizes helical structures. Gly⁴⁷⁷, which is partially conserved in other gB sequences, might correspond to an analogous point of structural change in gB, allowing a prefusion conformation in which αC is divided into 2 shorter helices. Also, the EBV gB Cys⁶⁸ in the N-terminal region is followed by Pro⁶⁹, and the analogous VSV G Cys²⁴ is followed by Pro²⁵. Cys²⁵ and Pro²⁵ in VSV G are located at a junction of secondary structural changes in pre and postfusion forms, corresponding to the loss of an N-terminal β-strand and the gain of a C-terminal α helix in the transition (26). It is possible that Pro⁶⁹ and Gly⁴⁷⁷ of EBV gB act as “switch” residues, which might be important for an analogous conformational change. Functional studies are required to determine the involvement and importance of these residues in a gB conformational transition.

Mechanism of Herpesvirus Fusion.

Although the full mechanism of herpesvirus entry remains unclear, some important insights into the interplay of the proteins involved in membrane fusion have been gained recently. The HSV-1 gH/gL complex was shown to cause lipid mixing indicative of hemifusion, whereas gB was required to resolve the hemifusion intermediate and allow contents mixing (29). Also, it was shown that both gB and the gH/gL complex can bind to gD, but interactions between gB and gH/gL could be observed only in the presence of gD (30, 31). The latter findings suggest strongly that the fusion machinery of HSV-1 is assembled only on gD activation. Whether the same applies to other herpesviruses remains to be determined.

VSV G, an abundant envelope glycoprotein, which mediates low pH-induced fusion in the endosomal compartment, is so far the only example of a fusion protein that exists in an equilibrium of pre and postfusion states and undergoes a reversible pH-induced conformational change (32). It was proposed that the initial movement in which domain I of VSV G separates from the C terminus is a result of protonation of His⁶⁰, His¹⁶², and His⁴⁰⁷, and the resulting disruption of a hydrogen bond network (26). These residues show a high degree of conservation among animal rhabdoviruses, and histidine residues have an important role in low-pH-induced conformational changes of class I and II type fusion proteins as well (33). pH-dependent entry of herpesviruses into some cell types has been reported (2), but the gB residues important for this entry mechanism have not been identified. An analysis of gB sequences (Fig. S1) indicates that there are no histidine residues at the positions analogous to the proposed His switch residues in VSV G, and none of the histidines found in gB are well conserved among human herpesviruses. The absence of His conservation suggests that pH-dependent entry in herpesviruses may be mediated by other residues or could be virus-specific in nature.

A hypothetical prefusion model of EBV gB (Fig. 4 Lower) was created by using the prefusion conformation of G as a model (described in detail in SI Materials and Methods). We superimposed gB domains I, II, and IV to best fit the fold and localization of the corresponding domains of G (domain V was not modeled). Domain III of gB was modified, so that it resembles the prefusion conformation of domain III of G: αC is split into 2 shorter helices. The prefusion gB trimer (Fig. S3B) could be generated from the prefusion monomer without steric clashes, despite the fact that each of the gB domains are larger. According to this model, the gB PH domain II is located on the top of the molecule, similar to the domain II of G, and would be a likely candidate for receiving an activation signal from the other viral glycoproteins involved in herpesvirus entry. The prefusion gB model provides some testable hypotheses, but whether the structural homology between G and gB extends into functional homology and a similar fusion mechanism remains to be shown.

Final Remarks.

Human herpesvirus infections are associated with various severe complications, especially in individuals with immune dysfunctions. EBV can cause lymphoid cancers and HSV encephalitis. Approaches to limiting viral spread in infected hosts may impact virus-associated pathology. However, there are no therapeutics directed at the entry process. Structural information obtained for EBV gB and presented here reveals differences in domain organization, FL conformations, and surface electrostatics between EBV and HSV-1 gB proteins. These distinct features likely influence how EBV and HSV-1 gB proteins may interact with other members of the fusion complex and with lipid membranes to mediate fusion. A comparison with the VSV G protein suggests specific regions of EBV gB that may undergo conformational switching during virus entry, and we propose a hypothetical model of a prefusion form of gB. The structural information on EBV and HSV-1 gB proteins advances our understanding of the herpesvirus entry mechanism and provides a foundation for development of entry inhibitors.

Materials and Methods

Crystallization of the EBV gB Ectodomain Variant EctoS FL.

To produce a nonaggregating EBV gB ectodomain variant (EctoS FL), the wild type EBV residues WY^112–113 and WLIW^193–196 were mutated to HR^112–113 and RVEA^193–196. Mutagenesis, expression, and purification of the EctoS FL were performed as previously described (19). Purified protein was concentrated to 5 mg/mL in 10 mM Tris, 100 mM NaCl pH 8. The EBV gB ectodomain was crystallized by using the hanging drop method. Drops were made by mixing equal volumes of 5 mg/mL protein solution and precipitating solution containing 0.1 M N-Cyclohexyl-3-aminopropanesulfonic acid (CAPS) pH 10.5 or 11.5, 1.40 M (NH₄)₂SO₄ (AS) and 0.2 M Li₂SO₄. Crystals were harvested in 1.6 M AS, 0.1 M CAPS, pH 10.5 or 11.5, 0.2 M Li₂SO₄, 10 mM Tris pH 7.4, 100 mM NaCl. For freezing purposes, the crystals were soaked in the harvest buffer containing 15% glycerol for 0.5–2 min, and then flash-frozen in liquid nitrogen. Data collection and processing information are described in detail in Table S1.

Structure Determination and Refinement.

Molecular replacement, performed by using Phaser (34), and the HSV-1 gB monomer (PDB code 2GUM) as a search model, provided a solution for one gB monomer in the asymmetric unit. The initial solution was refined using CNS (35), and composite omit maps calculated by CNS were used for manual building in Coot (36). Domain IV was added in the later stages of refinement, when the electron density corresponding to the domain IV became interpretable. Last cycles of refinement were done using Refmac (34), and the final model had R_free of 28.3% and R_work of 24.2%. The details of model refinement, final refinement, and geometric statistics are provided in SI Materials and Methods and Table S1.