Report

Three-Dimensional Structure of Herpes Simplex Virus from Cryo-Electron Tomography

See allHide authors and affiliations

Science  21 Nov 2003:
Vol. 302, Issue 5649, pp. 1396-1398
DOI: 10.1126/science.1090284

Abstract

Herpes simplex virus, a DNA virus of high complexity, consists of a nucleocapsid surrounded by the tegument—a protein compartment—and the envelope. The latter components, essential for infectivity, are pleiomorphic. Visualized in cryo–electron tomograms of isolated virions, the tegument was seen to form an asymmetric cap: On one side, the capsid closely approached the envelope; on the other side, they were separated by ∼35 nanometers of tegument. The tegument substructure was particulate, with some short actin-like filaments. The envelope contained 600 to 750 glycoprotein spikes that varied in length, spacing, and in the angles at which they emerge from the membrane. Their distribution was nonrandom, suggesting functional clustering.

The molecular architecture of viral capsids has been greatly illuminated by x-ray crystallography and image reconstruction of cryo–electron micrographs (1). Both approaches have relied heavily on exploiting icosahedral symmetry. However, many viruses do not conform to this symmetry, and viruses with icosahedral capsids also have asymmetric elements, such as their encapsidated genomes and surrounding envelopes. Much less structural information has been forthcoming on the latter systems. Electron tomography (ET) provides an approach capable of visualizing individual macromolecular complexes of any or no symmetry at near-molecular resolution (2). This technique involves collecting a tilt series of projection images and combining them computationally to reconstruct a three-dimensional density map: the tomogram (3). When applied to specimens preserved in vitreous ice (cryo-ET), such maps relate directly to the native structure (4).

Herpes simplex virus type 1 (HSV-1) is the archetypal member of the herpesvirus family. A human pathogen, it is responsible for oral cold sores and a rare but severe encephalitis (5). The virion comprises three major structural elements: (i) a nucleocapsid containing the 152-kbp genome; (ii) the envelope, consisting of a lipid bilayer with embedded glycoproteins; and (iii) a proteinaceous region between the capsid and the envelope called the tegument (6). The capsid is icosahedral and has been described in progressively greater detail by cryo–electron microscopy (cryo-EM) (7, 8). It houses the DNA, densely coiled in a “liquid crystalline” arrangement (9)—one of many features that HSV-1 shares with bacteriophages. The envelope accommodates ∼11 viral glycoproteins (10). The tegument, which contains ∼20 proteins, serves as a delivery compartment for proteins that are required early in the course of infection (11). Unlike the capsid, the tegument and envelope have remained largely uncharted territory. Here we describe intact virions as reconstructed by cryo-ET. The observations shed light on tegumentation and envelopment during virion egress (12).

Tilt series were recorded on two 120-keV microscopes that were similar, apart from one being equipped for energy filtering; that is, eliminating inelastically scattered electrons, which degrade images by chromatic aberration (13). For each series, the projections were aligned, and the tomogram was calculated. In all, 30 virion reconstructions were obtained from 14 tomograms. Fourteen of these reconstructions, judged to be of higher quality, form the basis for the observations reported below. The tomograms were appraised both directly and after the application of noise reduction techniques (14).

The virions were seen to be pleiomorphic membrane-bound particles (Fig. 1). They were generally round in shape, although some showed departures from sphericity that appear to be genuine features, not compression artifacts (fig. S1). The bilayer membrane was visualized as a continuous smoothly curved surface, ∼5 nm thick (Fig. 1). Its diameter ranged from 170 to 200 nm, averaging 186 nm. An array of spikes protruded from each virion, making the full diameter, on average, ∼225 nm.

Fig. 1.

Tomographic reconstruction of HSV-1 virions in vitreous ice. (A) Zero-tilt projection from a tilt series. Black dots are 10-nm gold particles used as fiducial markers. (B) Gallery of parallel slices, 15.5 nm apart and 5.2 nm thick, through the virion framed in (A). Each slice represents the average over seven planes. Red arrowheads mark filaments in the tegument. Scale bars, 100 nm.

Inside the envelope (Figs. 1B and 2B), the nucleocapsid occupied an eccentric position: On one side (the proximal pole), it was close to the envelope; on the other side (the distal pole), they were separated by 30 to 35 nm of tegument. The tegument occupied about two-thirds of the volume enclosed within the membrane, and the capsid occupied about one-third. In each tomogram, the tegument had the same cap-like structure; that is, its size and shape were quite well defined, and it was pleiomorphic primarily in substructure. Tegument asymmetry was apparent in some single projections (Fig. 1A), but in others it appeared uniformly thick (asterisk, Fig. 1A). The latter projections represent views along the polar axis.

Fig. 2.

Segmented surface rendering of a single virion tomogram after denoising. (A) Outer surface showing the distribution of glycoprotein spikes (yellow) protruding from the membrane (blue). (B) Cutaway view of the virion interior, showing the capsid (light blue) and the tegument “cap” (orange) inside the envelope (blue and yellow). pp, proximal pole; dp, distal pole. Scale bar, 100 nm.

At this resolution (∼7 nm), the tegument appeared to consist of a reticulum of particulate density. We saw no indication of regular organization other than the presence of some thin filaments up to 40 nm long, apposed to the membrane (arrows, Fig. 1B). Thus, either some tegument component is polymeric or some cellular filament has been incorporated. The filament width (∼7 nm) suggested actin, which has been reported in other viruses, as a candidate (15). The tegument cap was connected to the envelope by linkers ∼4 nm wide, mainly observed at sites where spikes protruded on the outside of the membrane (Fig. 3, B and C).

Fig. 3.

Glycoprotein spikes in the HSV-1 envelope, shown in a denoised tomogram. (A) Gallery of three distinctive spike morphologies: top, bifurcated spikes; middle, spikes with emergence angle of ∼50°; bottom, curved spikes. Note the close contact of some spikes with tegument densities inside the envelope. (B) Series of slices (at a spacing of 3 nm) through a cluster of spikes, putatively gB. The arrow shows a transmembrane contact between a glycoprotein and the tegument. (C) Segmented surface rendering of the virion portion shown in (B). Tegument is orange, membrane is blue, and spikes are yellow. Scale bars, 20 nm.

The distribution of glycoprotein spikes was evident in spherical sections through the tomograms at a radius just external to the membrane. From Mercator projections of such sections (fig. S2), we counted the number of spikes as 595 to 758 (mean, 659) per virion. Their packing density was variable (Fig. 2A). The average center-to-center spacing was ∼13 nm, but in some patches, it was as little as 9 nm (Fig. 3B), whereas other regions were quite sparsely populated. Dense clusters were usually associated with the distal pole of the tegument and were rarely close to the proximal pole (Fig. 1, slice 6, and Fig. 2B), possibly reflecting a local dearth of tegument material to anchor the spikes.

Spikes ranged in length from 10 to 25 nm (Fig. 3). Typically, they were ∼4 nm wide and many terminated in a globule ∼6 nm across. They also varied in straightness and in the angles at which they protruded from the membrane (the emergence angle). There was at least one kind of spike that emerged from the membrane at an angle of 35 ° to 55° (Fig. 3A, second row). Another spike was bifurcated (Fig. 3A, first row), presumably representing a dimer with divergent ectodomains.

Resolution assessment in tomography tends to be problematic, because the criteria conventionally used in EM, which gauge consistency between images of identical specimens, are not applicable to pleiomorphic specimens. However, the HSV-1 capsid is a consistent feature that is known to relatively high resolution, which allowed us to assess these tomograms. Thus, we calculated a resolution of 7.3 nm for the best individual nonsymmetrized capsids (n = 10 capsids; fig. S3). Taking into account anisotropy introduced by the “missing wedge” effect (3), it followed that the resolution was highest, ∼6.2 nm, in plane and lowest, ∼9 nm, in the dimension normal to it. The resolution of the averaged capsid was 5.6 nm (fig. S3) and isotropic. Thus, the resolution of individual tomograms sufficed to resolve adjacent spikes in the envelope.

In the capsid, the pentons and hexons were clearly visible (Fig. 4A). Their axial channels appeared as dimples and were more clearly demonstrated in transverse sections (Fig. 4D). The tomograms also conveyed indications of the triplexes: heterotrimers of 120 kD (16) at the sites surrounded by three adjacent capsomers (arrows in Fig. 4, D and E).

Fig. 4.

Intraviral nucleocapsid structure depicted by cryo-ET. (A) Surface rendering of the outer surface after averaging 11 nucleocapsids and applying icosahedral symmetry. The color map represents radii from 61 nm (blue) to 69 nm (orange). Additional density is associated with the pentons: white/orange in (A) and circled in (D), a central section of the map shown in (A). Note the absence of density in the corresponding position (E) from a cryo-EM reconstruction of empty capsids at the same resolution. (B) and (C) show the densities respectively associated with the pentons on the tegument-distal (B) and tegument-proximal (C) sides of the capsid. Arrows in (D) and (E) mark the triplexes (16), which are marginally visible. Scale bar in (A), (D), and (E), 50nm; in (B), and (C), 15 nm.

Capsid-tegument contacts have been described by cryo-EM in cytomegalovirus (17) and HSV-1 (18). ln both studies, icosahedral symmetry was imposed. However, our observation that the tegument is asymmetric raises the possibility that the interactions adjacent to the proximal pole may differ from those on the distal side. With tomograms, this question may be addressed. We oriented the reconstructed capsids by aligning their respective fivefold axes that lie closest to the axis between the proximal and distal poles, then averaged them and performed fivefold symmetrization. When compared to the nontegumented capsid at the same resolution, additional density was seen associated with both the distal and the proximal pentons (Fig. 4, B and C), in the same location as was detected by icosahedral reconstruction of whole virions (18). Thus, either the same protein engages the pentons on both sides, or different proteins bind to similar sites on the respective pentons.

Envelopes are studded with 600 to 750 spikes, of which several morphological types were distinguished. Hitherto, there have been few observations of noncrystalline membrane proteins in situ, and those molecules tend to be diagrammed as oriented perpendicular to the membrane (emergence angle, 90°). Departing from this paradigm, one kind of spike emerged at an angle of ∼45° (Fig. 3A), a property also attributed to the poliovirus receptor (19). Of the ∼11 glycoproteins, a crystal structure has been determined only for gD (20). We are not yet in a position to associate observed structures with specific glycoproteins. However, one relatively abundant spike, ∼20 nm long and terminating in a globule (Fig. 3B), resembled a spike visualized by negative staining and determined by immunolabeling to be gB (21).

The distributions of spikes were nonrandom, tending to be sparse at the proximal pole and densely packed around the distal pole. The observed partitioning is consistent with the proposal that lipid rafts participate in envelope assembly (22) (fig. S4). An intriguing possibility is that patterns within these distributions reflect functional associations, such as local clustering of glycoproteins destined to serially engage different receptors during cell entry (23).

It is now widely accepted that herpesvirus egress is a two-stage process whereby nucleocapsids are released into the cytoplasm by envelopment/de-envelopment at the nuclear membrane, and undergo secondary envelopment at a Golgi-derived compartment (12, 24). Although some tegument components may be acquired during exit from the nucleus, the main assembly phase accompanies secondary envelopment. With respect to establishing the asymmetric cap, a key question is whether the thin proximal pole or the thick distal pole is membrane-associated as budding proceeds. The latter is suggested by in vivo observations (25). A hypothetical pathway is diagrammed in fig. S4.

Tegument assembly is a challenging proposition, as a large number of components must be accommodated. Because many of them are dispensable, the design must be adaptable enough to survive their absence. We envisage that assembly initiates at a site that will become the distal pole, with contacts between capsid pentons and membrane glycoproteins mediated by linking protein(s). As assembly proceeds and tegument proteins accrete between the capsid and the membrane, the linkers extend, providing a framework for other components to bind. In this way, assembly continues even if dispensable components are unavailable, terminating with the binding of shorter or less extended linkers to pentons on the proximal side.

Some intriguing parallels exist between tegument assembly and the formation of signaling complexes (26) and focal adhesions (27). These processes are all nucleated on membranes; involve the specific recruitment of multiple protein species but without the formation of a geometrically regular structure; and in all cases, phosphorylation appears to play an important role.

We see considerable prospects for cryo-ET to characterize the entry and exit pathways of herpesviruses and other viruses, and in particular to explore the “nanoecology” of spike clustering in the HSV-1 envelope. Such studies may exploit comparisons with mutants and virions complexed with receptors and antibodies. Enhanced resolution may allow identification of the unique vertex occupied by the UL6 protein, which serves as a phage-like portal (28). These in vitro investigations can be complemented with tomograms of sections containing viral particles in situ at various stages of these pathways (29).

Supporting Online Material

www.sciencemag.org/cgi/content/full/302/5649/1396/DC1

Materials and Methods

Figs. S1 to S4

References and Notes

View Abstract

Navigate This Article