Three-Dimensional Localization of pORF65 in Kaposi's Sarcoma-Associated Herpesvirus Capsid

Abstract

Of the six herpesvirus capsid proteins, the smallest capsid proteins (SCPs) share the least sequence homology among herpesvirus family members and have been implicated in virus specificity during infection. The herpes simplex virus-1 (HSV-1) SCP was shown to be horn shaped and to specifically bind the upper domain of each major capsid protein in hexons but not in pentons. In Kaposi's sarcoma-associated herpesvirus (KSHV), the protein encoded by the ORF65 gene (pORF65) is the putative SCP but its location remains controversial due to the absence of such horn-shaped densities from both the pentons and hexons of the KSHV capsid reconstructions. To directly locate the KSHV SCP, we have used electron cryomicroscopy and three-dimensional reconstruction techniques to compare the three-dimensional structure of KSHV capsids to that of anti-pORF65 antibody-labeled capsids. Our difference map shows prominent antibody densities bound to the tips of the hexons but not to pentons, indicating that KSHV SCP is attached to the upper domain of the major capsid protein in hexons but not to that in pentons, similar to HSV-1 SCP. The lack of horn-shaped densities on the hexons indicates that KSHV SCP exhibits structural features that are substantially different from those of HSV-1 SCP. The location of SCP at the outermost regions of the capsid suggests a possible role in mediating capsid interactions with the tegument and cytoskeletal proteins during infection.

Herpesvirus virions share a characteristic architecture in which the double-stranded DNA genome is surrounded by an icosahedral protein capsid, a thick tegument layer, and a lipid bilayer envelope (15). The capsid, approximately 1,300 Å in diameter, is a T = 16 icosahedron with 12 pentons forming the vertices, 150 hexons forming the faces and edges, and 320 triplexes interconnecting the pentons and hexons (15, 17). These structural features of the capsid are built from four of the six capsid proteins. In Kaposi's sarcoma-associated herpesvirus (KSHV), a gammaherpesvirus, the pentons and hexons are composed of five and six copies, respectively, of the major capsid protein (MCP), pORF25. The triplexes are heterotrimers containing a monomer of the pORF62 protein and a dimer of the pORF26 protein (20, 23). The fourth and smallest capsid protein (SCP), pORF65, is homologous to the SCPs in other herpesviruses, including the structurally well-characterized herpes simplex type 1 (HSV-1) and cytomegalovirus (CMV), representative members of the alpha- and betaherpesvirus subfamilies, respectively. However, of the six capsid proteins, SCPs share the lowest sequence homology between HSV-1, human CMV (HCMV), and KSHV (23). This low sequence homology of SCPs may be related to their virus-specific functional roles. It has been recently shown that the HCMV UL48.5-encoded SCP is essential for HCMV infection in vivo (3), but its HSV-1 counterpart, VP26, is dispensable for HSV-1 infection (5, 8). While other proteins making up the capsid shell have been shown to have very similar structures located at roughly equivalent positions in HSV-1, HCMV, and KSHV, the exact locations of SCPs have not been explicitly determined in KSHV and HCMV, due to the relatively low resolutions of their three-dimensional (3D) maps and the small size of their SCPs. In HSV-1, where an in vitro assembly system which uses expressed proteins has been developed to generate SCP-minus capsids, SCP has been shown through difference mapping to bind the MCP subunits of the hexons but not those of the pentons (21, 28). HSV-1 SCP has a horn shape (29) with a predominantly β-sheet secondary structure (22, 26) and forms a ring of six subunits that caps the rims of the upper domains of the MCPs of each hexon (21, 28). The hexon-specific association of SCP has also been suggested for HCMV based on subtle differences between the tips of the penton and hexon subunits (4, 19), although a direct confirmation is not yet available.

The difficulties in obtaining large quantities of purified KSHV capsids and the lack of an in vitro KSHV capsid assembly system have limited the resolution of the KSHV capsid structure obtainable by electron cryomicroscopy (cryoEM) and 3D reconstruction and thus prevented a direct localization of the KSHV SCP, pORF65. The first structure of the KSHV capsid at 24 Å resolution (23) did not reveal any recognizable horn-shaped densities that resemble those bound to the HSV-1 hexons, as further confirmed independently by Trus et al. (20). It was not possible to discern the location or shape of pORF65 in either the first reconstruction or that of Trus et al. Although a preliminary immunoelectron microscopy experiment subsequently showed that pORF65 was indeed bound to the capsid (13), it was unable to confirm whether pORF65 binds MCP and whether it binds MCPs in both hexons and pentons of the KSHV capsid. We now demonstrate through antibody labeling and difference mapping that pORF65, despite having structural features substantially different from those of HSV-1 SCP, exhibits a similar pattern of interaction with MCP in KSHV capsid and binds MCPs of the hexons but not of the pentons of the KSHV capsid.

MATERIALS AND METHODS

Purification and antibody labeling of KSHV capsids.

KSHV capsids were purified by sucrose density gradient centrifugation from latently infected BCBL-1 cells induced with 12-O-tetradecanoylphorbol-13-acetate (TPA) and γ-butyrate as previously described (23), with slight modifications. In the present study, no mercaptoethanol was used in the cell culture. TPA (20 ng/ml) and 0.3 mM butyric acid were used to induce lytic replication. Capsids were first pelleted by high-speed (60,000 × g) centrifugation for 1 h without the use of polyethylene glycol 6000, purified on a 35% sucrose cushion and then through a 20 to 60% sucrose gradient with 2% NP-40, and resuspended in phosphate-buffered saline.

The cDNAs encoding ORF65 and ORF73 were previously identified by screening a cDNA library from TPA-induced BCBL-1 cells with serum from a patient positive for human immunodeficiency virus and KSHV (1, 30). They were then expressed as a glutathione S-transferase (GST) fusion protein in a baculovirus system, purified, and injected into rabbits. Rabbit anti-ORF65 and anti-ORF73 immunoglobulin G (IgG) antibodies were purified from the sera by affinity chromatography on a protein A-Sepharose 4B column (Amersham Pharmacia Biotech AB, Uppsala, Sweden). To remove nonspecific antibodies, rabbit anti-ORF65 and anti-ORF73 IgGs were passed repeatedly over columns of cyanogen bromide-activated Sepharose 4B covalently coupled with purified GST protein and BJAB cell lysate proteins. The flowthrough from these columns showed no reactivity with GST protein on Western blots and no reactivity with BJAB, CV-1, or COS-1 cells in immunofluorescence and Western blot assays. These purified antibodies were used throughout the present studies.

To generate antibody-labeled capsids, 20 μl of purified anti-pORF65 antibodies at a concentration of 3.16 mg/ml were incubated overnight at 4°C with 20 μl of purified capsids estimated at about 6 × 10¹¹ capsids/ml, in order to saturate the 960 pORF65 binding sites on each capsid. Because KSHV pORF73 is a latent gene product that is not expressed in lytic replication and is thus absent from the KSHV capsids, purified anti-ORF73 antibodies were also incubated with a separate aliquot of purified capsids and served as a negative control. Negative-stain electron microscopy with 2% uranyl acetate was performed on an aliquot of the capsid samples incubated with either anti-ORF65 or anti-ORF73 to evaluate the extent of antibody labeling by monitoring the formation of capsid cross-linkage. For cryoEM imaging of anti-ORF65-labeled capsids, the incubated sample was centrifuged in a Microcon filter at 7,000 × g for 10 min to remove free antibodies in the supernatant and concentrate the labeled capsids. The concentrated sample was then resuspended in phosphate-buffered saline buffer and sonicated twice in a water bath sonicator for 2 min immediately prior to cryoEM sample preparation in order to loosen some of the aggregated capsids.

cryoEM and icosahedral reconstruction.

Sample freezing, electron microscopy, and 3D reconstruction were carried out according to established procedures (23, 29). Micrographs of both antibody-labeled and unlabeled KSHV capsids embedded in vitreous ice were recorded at a magnification of ×30,000 in a JEOL1200 100-kV electron cryomicroscope as previously described (23). Orientation determination and 3D reconstruction were carried out by using program modules in the IMIRS package by using SGI computers with multiprocessors (12, 25). The contrast transfer functions associated with the cryoEM micrographs were determined (27) and corrected (24). The effective resolution of the final map was estimated to be 33 Å, based on the criterion that the Fourier shell cross-correlation coefficient between two independent reconstructions reaches 0.5 (16). Central density slices 9 Å thick were taken perpendicular to the three- and fivefold axes to provide a two-dimensional interpretation of the maps. The difference map was obtained by subtracting the unlabeled capsid structure, scaled to the same resolution of 33 Å, from the antibody-labeled structure and superimposing the resulting difference onto the unlabeled map. 3D visualization was carried out by using Iris Explorer (NAG, Downers Grove, Ill.) with custom-designed modules.

RESULTS AND DISCUSSION

Cross-linking of antibody-labeled KSHV capsids.

Structural studies of KSHV have been hampered by major difficulties with virus isolation. More than 3 liters of extracellular supernatant was used to obtain only a few microliters of capsids with appropriate concentrations for cryoEM, and the methods for purification are inefficient due to the relatively low virus production in chemically induced BCBL-1 cells. Despite this, we managed to obtain adequate amounts of purified KSHV capsids to conduct the antibody labeling and subsequent cryoEM experiments (Fig. 1). Antibodies can cause cross-linking when each Fab arm of an antibody molecule binds a different capsid. This can lead to the formation of visible capsid aggregates. Direct comparison by both negative-stain electron microscopy (Fig. 1a and b) and cryoEM of KSHV capsids before (Fig. 1d) and after (Fig. 1e) incubation with anti-pORF65 antibodies revealed the formation of predominantly capsid aggregates in the presence of anti-pORF65 antibodies, indicating that anti-pORF65 effectively binds capsids and causes cross-linking. In contrast, no cross-linking was observed when the capsids were incubated with purified control anti-ORF73 antibody (Fig. 1c), suggesting that the binding of anti-ORF65 to the capsid is specific. A previous negative-stain electron microscopy experiment used immunogold labeling with a similar anti-pORF65 antibody and a gold-conjugated secondary antibody to decorate the outside of the KSHV capsid, revealing gold clusters bound to capsids (13). While these micrographs were not able to show the exact location and the number of copies of pORF65 on each capsid, they offered the first visual proof that pORF65 is attached to the capsid (13).

FIG. 1.

Electron microscopy of KSHV capsids before and after incubation with anti-pORF65 antibodies and control antibodies. (a to c) Typical areas of negative-stain electron micrographs of KSHV capsids prior to antibody incubation (a), after incubation with anti-pORF65 antibodies (b), and after incubation with negative-control anti-pORF73 antibodies (c). Individual capsids (arrows) are distanced from one another in both the unlabeled capsid preparation (a) and in the capsid preparation incubated with the control antibodies (c). However, capsid aggregation caused by antibody cross-linking is apparent in the anti-pORF65-labeled capsid preparation (b). The arrowhead in panel c points to anti-ORF73 antibodies floating in the background of dispersed capsids. (d to f) cryoEM of unlabeled (d) and anti-pORF65 antibody-labeled (e and f) capsids. In panel e, the cryoEM micrograph of capsids after incubation with excess anti-pORF65 antibodies shows aggregation and piling up of capsids caused by antibody cross-linking. (f) Same as panel e but imaged after water bath sonication. Some bound antibodies are visible as rings of density protruding from capsids (arrows). The cryoEM images in panels d to f were recorded at the same magnification, and their relative scale is indicated by the scale bar in panel d.

To release capsids from their aggregated form caused by antibody cross-linking, we added a water bath sonication step immediately prior to freezing the antibody-labeled capsids for making cryoEM grids. The cryoEM micrographs obtained in this approach show some well-separated antibody-labeled capsids that are adequate for 3D reconstruction analysis (Fig. 1f). In these micrographs, bound antibodies are present as a faint ring of density protrusions extending radially from the capsids (Fig. 1f), similar to micrographs of other antibody-bound viruses (11, 18). Such rings are not present in control micrographs obtained from the capsid preparation prior to antibody labeling (Fig. 1f). The densities around each ring appear to be continuous, indicating that our effort to saturate the binding sites was successful. The extensive overlapping of the capsids in some areas of the micrographs are typical results of antibody cross-linking and serve as another visual confirmation of the binding of the antibodies to the capsid surface.

Localization of pORF65.

A 3D map of the antibody-bound capsids was computed to an effective resolution of 33 Å from 229 particle images selected from 72 micrographs. The central sections of the reconstructions perpendicular to the threefold axis show extra densities attached to the tips of all hexons of the antibody-labeled map (Fig. 2b). These densities are not present at the corresponding locations on the unlabeled map (Fig. 2a). To examine whether such densities exist at the tips of the pentons, we compared central sections extracted perpendicular to a twofold axis, which would show cross-sectional views of pentons, hexons, and triplexes (Fig. 2c and d). This comparison shows that the pentons and triplexes in the anti-pORF65-labeled and unlabeled capsids are essentially identical. The extra densities attached to the tips of the hexons are not present in the vicinity of the pentons (Fig. 2d).

FIG. 2.

Thin slices through the center of unlabeled (a and c) and labeled (b and d) capsid reconstructions perpendicular to the threefold (a and b) and twofold (c and d) axes. Protein density is represented by darkened gray. Solid arrows point to hexons, which were decorated by extra antibody densities in the labeled capsid reconstruction (b and d). Open arrows in panels c and d indicate the pentons, to which no significant extra densities were attached. Arrowheads in panel c denote two triplexes which appear identical in the labeled and unlabeled capsids.

These extra densities can also be seen protruding from the hexons of the 3D reconstruction of the labeled capsid (Fig. 3a), whereas the triplexes and pentons remain practically identical to those of the unlabeled capsid (Fig. 3b). The difference map confirms that the only significant extra densities are attached to the hexons (Fig. 3c). The small residual difference densities over the pentons and some triplexes are due to slight surface variations between the two reconstructions, probably reflecting the statistical fluctuations expected in 3D maps reconstructed from relatively small numbers of particles at a relatively low resolution. The hexon-associated densities appear from the top view as single or double connected beads approximately 20 to 80 Å wide and 40 to 60 Å tall. Since these are not large enough to constitute an entire IgG molecule, we attribute these densities to the Fab portion of the antibodies. The bound antibodies were not visualized in their entirety in our reconstruction most likely for two reasons. First, the polyclonal antibodies used in this study, though specific for pORF65, may not have bound to exactly the same epitope on pORF65; thus, the densities shown in our structure would represent only the average densities of the antibodies at slightly different orientations. Secondly, the flexibility of the Fc hinge regions of the antibodies may have disrupted the icosahedral symmetry, thereby preventing these regions from being reproducibly reconstructed. Steric hindrance may also have prevented saturation of all possible binding sites on the hexons since the antibody molecules are ostensibly larger than the distance between hexon subunits. In any case, we can definitively conclude from these observations that pORF65 binds the tips of hexon subunits of the KSHV capsid and does not bind the pentons. The lack of a horn-shaped density near the KSHV hexon tip suggests that pORF65 either interacts more extensively with or is integrated into the MCP or that its distal regions are more disordered than the HSV-1 SCP and thus could not be visualized in our icosahedral reconstruction.

FIG. 3.

3D localization of pORF65 on the capsid by difference imaging of antibody-labeled (a) and unlabeled (b) KSHV capsids. The 3D structures of both the antibody-labeled (a) and unlabeled (b) capsids were reconstructed at 33 Å resolution. (c) Superposition of the difference map (red), obtained by subtracting panel b from panel a, on the unlabeled capsid. Significant density differences (red), attributed to bound antibodies, were only present on the hexons and not on the pentons (one is circled in black).

The SCPs are a very diverse group of proteins that share almost no sequence homology between subfamilies (23), although at least within the gammaherpesviruses there is significantly more similarity (13). SCP homology was assigned by local genetic colinearity, but the proteins also share basic pI, stoichiometry, and small size. In HSV-1, HCMV, and KSHV, the SCP is the smallest capsid protein that is present in stoichiometric amounts, usually in a 0.8 to 1:1 ratio with the MCP as estimated from sodium dodecyl sulfate gels. However, they do vary widely in size: the KSHV pORF65 is more than twice as large as the HCMV SCP, with HSV-1 VP26 in between. Thus, it is not surprising that they might have different shapes and binding properties. HSV-1 VP26 is a two-domain protein that forms interconnected, hexameric rings crowning the hexon (28). SCPs in HCMVs and simian CMVs attach more towards the tips of the hexon subunits, giving them an elongated appearance (4, 19). Although it is not possible to discern the exact location or shape of pORF65 from our reconstruction, antibody-attributed densities on the difference map appear to bind the center of the upper domain of the hexon subunits, similar to HSV-1 VP26. Interestingly, the radial density plot for the KSHV capsid more closely resembles that of simian CMV than that of HSV-1 (20).

Implications of SCPs in viral infection.

The SCPs are all highly basic and thus positively charged. Despite this common feature, SCPs appear to serve very different functions in their respective viruses. The HSV-1 SCP VP26 is not assembled at the same time as the other capsid proteins, which simultaneously and instantaneously form the procapsid (7, 14). Accordingly, VP26 has been shown to be dispensable for HSV-1 capsid assembly in vitro but necessary for infectious virus production in the nervous system in vivo (8). In contrast, HCMV SCP is absolutely required for growth in vitro (3). Channel catfish virus, a distant relative of the human herpesviruses, has no identifiable SCP homolog but retains the characteristic herpesvirus morphology (2). The function of KSHV pORF65 is currently unknown, but its position on the outer regions of the capsid suggests that, in addition to possible DNA interaction, it may also regulate capsid-tegument interactions. pORF65 may be involved in tegumentation and/or uncoating, as well as tegument-host cell protein interactions. However, it should be noted that the absence of the SCP VP26 from HSV-1 virions did not affect the conformation of the tegument proteins bound to the capsid (5). In HCMV, the tegument proteins are bound to both pentons and hexons (6). Although these observations do not rule out a role for SCP in tegument interactions, they suggest at least that steric hindrance of MCP-tegument binding sites by SCP is not a factor. A recent study revealed that HSV-1 VP26 associates with ribosomes and may regulate host protein translation (9). A similar role for pORF65 is possible, as evidenced by another recent study in which pORF65 was shown to relocate from the nucleus to the cytoplasm of primary effusion lymphoma cells following TPA induction (10). However, there is currently no conclusive evidence. Studies of pORF65-minus capsids will be necessary to determine the precise shape and location of pORF65 on the capsid and elucidate its role in assembly and infection.