Three-Dimensional Structure of Herpes Simplex Virus Type 1 Glycoprotein D at 2.4-Nanometer Resolution

Abstract

Herpes simplex virus type 1 glycoprotein D (gD) is essential for virus infectivity and is responsible for binding to cellular membrane proteins and subsequently promoting fusion between the virus envelope and the cell. No structural data are available for gD or for any other herpesvirus envelope protein. Here we present a three-dimensional model for the baculovirus-expressed truncated protein gD1(306t) based on electron microscopic data. We demonstrate that gD1(306t) appears as a homotetramer containing a pronounced pocket in the center of the molecule. Monoclonal antibody binding demonstrates that the molecule is oriented such that the pocket protrudes away from the virus envelope.

The structure of herpes simplex virus type 1 (HSV-1) is defined by three distinct morphological units. A DNA-containing, icosahedral capsid is surrounded by a layer termed the tegument, which in turn is surrounded by a protein-containing lipid bilayer, the envelope (19). While cryoelectron microscopy and image reconstruction techniques (22) have been used to determine the structure of the capsid and individual capsomeres, structural information about the envelope proteins and the component proteins is sparse. Spikes reported on the surface of HSV-1 (30) have been identified by immunoelectron microscopy as the virus-encoded envelope glycoproteins B (gB), C, and D (24). It was also suggested that gB may form clusters, but no further data have yet been presented about the structure of these glycoproteins.

gD of HSV-1 is essential (10) for the stable binding of the virus to cells and the fusion of the virus to the cell membrane. Three paired disulphide bonds that are essential for the adoption of the native conformation (12) of the molecule and four separate regions that are important for virus entry (2, 14) are contained within amino acid residues 27 to 43 (I), 126 to 161 (II), 225 to 246 (III), and 277 to 310 (IV). Mutations introduced into these regions indicate that they are likely to be involved in different steps during infection and that there is no single functional domain in gD (16).

Two major cell proteins which bind to HSV-1 gD have been identified. The first, termed herpesvirus entry mediator A (HveA, previously designated HVEM), was identified by expression cloning from a cDNA library (13) and is the principal protein responsible for virus entry into human lymphoid cells. Expression of HveA in several cell lines that are nonpermissive for HSV-1 infection allows virus entry and replication, and baculovirus-expressed HveA has been shown to bind directly to baculovirus-expressed gD in vitro (29, 31). The second identified gD binding protein is the poliovirus receptor-related protein 1, termed HveC (6, 11), and is thought to be responsible for HSV-1 and HSV-2 entry into mucosal epithelial cells. A third receptor, HveB, enables HSV-1 entry only after amino acid substitutions in gD at position 27 (6).

Expression of gD truncated prior to the transmembrane anchor at residue 306 by using baculovirus systems and subsequent purification (23, 31) produces biologically functional material of a purity sufficient (>99%) for crystallization trials. Here, we have used transmission electron microscopy in conjunction with image-processing techniques to determine the structure of the protein to a resolution of 2.4 nm.

Recombinant gD1(306t) was expressed in baculovirus and purified by Ni affinity, immunoaffinity and gel permeation chromatography as described previously (23, 31). Negatively stained specimens of the purified protein were prepared (27) by using an aqueous solution of 4% (wt/vol) uranyl acetate (pH 4.5). Samples at a protein concentration of approximately 20 μg/ml were allowed to absorb for 20 s and washed on droplets of double-distilled water prior to staining. Grids were observed by using a Philips CM10 transmission electron microscope operated at an accelerating voltage of 100 kV, and electron micrographs were recorded on Agfa Scientia 23 D 56 electron image sheet film at a calibrated magnification (×50,500). Electron micrographs were digitized by using a LeafScan 45 microdensitometer at a step size of 25 μm, corresponding to 0.495 nm/pixel at the specimen level. Single-particle averaging was performed by using the SPIDER software package (4). Images were contrast normalized, band pass filtered, and aligned by reference-free alignment (18). Classes of particles representing identical orientations were determined by multivariate statistical analysis and hierarchical ascendant classification with complete linkage (5, 20, 28). Resolution was assessed by using the Fourier ring correlation criterion (26). The molecular surface calculations and the rendering of the gD tetramer shown in Fig. 4 were generated by using the GRASP program (15).

FIG. 4.

Three-dimensional model of a gD tetramer viewed from outside the virus looking down onto the envelope, based on the orthogonal averaged projections shown in Fig. 3. The barrel-shaped domains are depicted at the top, and the basal domains forming the interprotomer contacts are shown at the bottom. The tips of the barrel-shaped domains are not involved in interprotomer contact. Note the large pocket in the center of the tetramer.

Although much of the protein aggregated into large precipitate-like assemblies, examples of reproducibly identifiable, nonaggregated gD1(306t) species that demonstrated a high level of homogeneity are shown in Fig. 1. The predominant projection is square (7.5 by 7.5 nm) and is characterized by a pronounced, central stain deficit surrounded by four circular protein domains, suggesting that these particles constitute tetramers. This projection is referred to as the top view. Further examination (see below) revealed a second projection with identical dimensions but with a slightly elongated central stain deficit located between two rectangular protein domains. These observations suggest that the latter projection is orthogonal to the frontal view; this orientation is therefore referred to as the side view. Nearly all the particles depicted were in one of these two orientations relative to the carbon support film and are in close contact with each other. The presence of two principal orthogonal projections was confirmed by correspondence analysis, which is used to distinguish different projections (e.g., top and side views). The outcome of this analysis applied to the different projections of gD (Fig. 1) is shown diagrammatically in Fig. 2. As is evident from the diagram, the two principal components (projections) present in the data set are quite dissimilar, as would be expected for two orthogonal projections. This finding is reflected by the numbers on the y axis. At 1.0, all projections would be in the same class, and no features would allow differentiation between top and side projections. In the present case, only two main projections were revealed from a low level of discrimination (1.0) to a relatively high level (0.20), suggesting that all the gD molecules were in one of two stable orientations on the support film and that these projections can be conclusively distinguished from each other.

FIG. 1.

(a) Electron micrograph of negatively stained recombinant gD1(306t). Particles viewed from the top are marked by small arrows, side views are highlighted by arrowheads. Bar, 50 nm. (b) A typical complex formed between gD1(306t) (side view), and the monoclonal antibody DL11 is shown at high magnification, together with a side view of gD1(306t) without antibody (lhs). The binding is also presented in diagrammatic form in panel c. Note that the antibody (IgG) is bound to the basal domain (B) of gD. The binding site is marked by an arrow.

FIG. 2.

Diagram showing the division of the data into two major classes (top and side views on the left and right, respectively) corresponding to their preferred orientation on the support film. The cut-off threshold (y axis) was optimized for maximum resolution. Further details are provided in the text.

Averaging single particles present in each orientation allowed the mapping of top and side views (Fig. 3). The resolution for each map, determined by Fourier ring correlation, is 2.4 nm. Viewed from the top (Fig. 3a), each promoter has a roughly triangular shape. Together with the density distribution, which is apparent in the side view (Fig. 3b), these data reveal the general shape of one gD promoter to be a cylinder, comprising a bulky basal domain extending into a slightly tapered barrel. Estimation of the molecular mass from the volume (9, 25) of a single promoter consisting of a basal domain of roughly 33 nm³ and a barrel-shaped domain of approximately 28 nm³ yielded a value just under 45 kDa, in good agreement with the expected mass for a single gD1(306t) polypeptide chain (31). While the basal domains of the four adjacent protomers are connected, the tapered, barrel-like domains are parallel across their entire length, creating a pronounced pocket that opens toward only one side of the tetramer. The volume enclosed by the four protomers is a little less than 16 nm³, meaning that the pocket could house a peptide with a mass of approximately 10 kDa.

FIG. 3.

Averaged top (a) and side (b) views. Bar, 10 nm. (a) The circular protein deficit is surrounded by four protomers; (b) Two protomers are separated by an elongated stain channel that terminates at the basal domains (see text for additional details).

To determine conclusively which of the identified domains houses the C-terminal region, gD1(306t):DL11 immunoglobulin G (IgG) complexes were isolated and imaged by electron microscopy. Complexes of gD1(306t):DL11 monoclonal antibody were isolated essentially as described previously (1). Samples of gD1(306t) and DL11 (21), each at a protein concentration of approximately 100 μg/ml, were mixed in buffer containing 200 mM sodium chloride–50 mM Tris-HCl (pH 7.5) and incubated for 30 min at 30°C. After incubation, large immune complexes were removed by centrifugation for 5 min at 10,000 × g. The resulting supernatant was applied to a Superdex 200 column (Pharmacia SMART system) equilibrated with 200 mM sodium chloride–50 mM Tris-HCl (pH 7.5). Analysis of the major peak containing gD1(306t)-DL11 complexes by electron microscopy revealed that DL11 binds to the basal domain, as depicted by a typical projection in Fig. 1b and c. Since DL11 recognizes the central region of gD, which in the truncated form of the protein is closer to the C-terminal region of gD (17), this region must form part of the basal domain. We propose that the barrel-shaped domains of gD extend into the extracellular space away from DL11 binding domain and the envelope of the virion and that, subsequently, the aforementioned pocket opens toward the extracellular space.

Although gD-1(306t) lacks the membrane-spanning and cytoplasmic domains, it has many attributes in common with native gD-1, in that it reacts identically with a panel of conformationally dependent monoclonal antibodies (23), blocks virus infection of cells (3, 16), and binds to HveA and HveC in vitro (11, 31). However, the protein exists as a homodimer in solution (31) and cross-linking experiments with purified virus as the source of native gD reveal gD monomers, homodimers, and homotrimers (7, 8). It is possible that the conditions used for sample preparation prior to electron microscopy encouraged protein-protein association and that the cross-linking experiments are indicative of a somewhat random distribution of gD molecules within the virus envelope. Therefore, at present it is difficult to determine whether the gD tetramer (Fig. 4) is the fully functional form of the protein.

If the gD tetramer is indeed the fully functional form of the protein, then the pocket created within it can accommodate a protein with a mass of approximately 10 kDa (see above), implying that this large protein deficit could house portions of binding site(s) for HveA and HveC. Both HveA and HveC have been shown to bind directly to gD (11, 29). As HveA has a mass of 20 kDa, a large portion of the protein could be accommodated inside the binding pocket. Similar considerations hold true for HveC. This model of gD-receptor binding is particularly attractive because gD can form complexes with other envelope glycoproteins without impeding the virus-to-cell binding process. Complex formation of gD with gB, gC, gH, and gL into the essential fusion complex within the native envelope has been reported previously (7, 8). While portions of both binding domains, for HveA and HveC, may be present in the pocket, sterical constraints would prevent the simultaneous binding of HveA and HveC. Relevantly, HveA and HveC compete for overlapping but distinct binding sites (6, 11, 21).

An alternative and possibly more likely explanation is that the process of HSV-1 binding and the penetration of a cell is a dynamic process involving multiple protein-protein interactions. It is also possible that a gD dimer interacts with other proteins to form a binding complex with an affinity much higher than those for gD binding to HveA and HveC. To establish whether the functional form of gD is dimeric or tetrameric requires further study, but the overall structure described here would remain the same.