Single-particle virus reconstructions are typically generated with icosahedral symmetry imposed, meaning that any features that are not icosahedrally arranged will be lost. As the portal is believed to be located at just one of the 12 5-fold vertices in the capsid, it cannot be definitively resolved in icosahedral reconstructions, necessitating the use of a nonicosahedral, symmetry-free approach. Unfortunately, reconstructions of viruses free of icosahedral enforcement, which are ideal for resolving portal structures, are limited by difficulties in identifying the orientation of the unique vertex (2, 3). In the case of the HSV-1 capsid, there are no sentinel markers for the location of the portal, such as those provided by the external tail structures in some double-stranded DNA bacteriophages (1, 4, 9, 10, 20). The lack of a nonicosahedrally arranged protein density that protrudes from the surface of the HSV-1 capsid presents a computational challenge in determining, from the noisy electron cryomicroscopy (cryo-EM) image, which of the 12 vertices is the portal vertex. An additional complication is that the portal is approximately the same size (~814 to 1,036 kDa in total, assuming 11 to 14 copies of UL6 [19]), as the pentons (745 kDa) that occupy the other 5-fold vertices and, consequently, does not show up as an obvious excess or absence of density in projection images of capsids. Even though the HSV-1 capsid lacks a tail density, a feature that has proven to be an efficient marker of the portal vertex in bacteriophages (4, 10, 13), we posited that recent advances in Zernike phase-contrast EM (ZPC-EM), which produces dramatic low-resolution contrast enhancement (5, 6), would prove useful in helping to identify this unique vertex. Furthermore, when used in conjunction with an advanced reconstruction algorithm (12), our ability to identify this vertex should be greatly enhanced. Identification of this vertex would then make it possible to reconstruct HSV-1 capsids to a moderate resolution using single-particle cryo-EM without enforcing any symmetry (14).

ZPC-EM relies on using a thin carbon film with a small hole in the middle, suspended at the back focal plane of the microscope's objective lens, to retard the phase of scattered electrons by π/2 with respect to the unscattered beam (7). The resulting shift in phase is manifested in dramatically enhanced low-resolution image contrast compared to that of conventional imaging methods.

HSV-1 B capsids, purified by sucrose gradient sedimentation (16), were applied to thin-carbon-film-coated Quantifoil grids (14) and plunge-frozen in liquid ethane using a Vitrobot (FEI, Eindhoven). The grids were kept at 100K in a JEM2200FSC field-emission electron microscope (JEOL, Tokyo, Japan) operating at 200 kV, equipped with an in-column energy filter and a Zernike-type phase plate. Images were recorded on a Tietz 4,000- by 4,000-pixel SlowScan charge-coupled-device (CCD) camera (TVIPS, Germany) at ×56,000 detector magnification, with a total dose of 20 e/Ų. The ZPC-EM image was targeted at close to zero defocus in order to obtain maximum contrast enhancement. The contrast of the resulting images is substantially higher than that of conventional cryo-EM images (Fig. ​(Fig.11 a and b). The white halo around each particle in Fig. ​Fig.1b1b is a consequence of the phase-contrast optics and is characteristic of ZPC-EM images (6, 14).

The particle images were boxed with the e2boxer.py program (18) and reconstructed using the Multi-Path Simulated Annealing algorithm (11-13). From 353 CCD frames, 6,033 single-particle images of HSV-1 B capsids were collected. Under ZPC-EM conditions, we were able to reconstruct an icosahedral map (Fig. ​(Fig.1c)1c) without the need for contrast transfer function correction (6, 14). The resolved icosahedral orientation of each particle was subsequently used to initiate the process of determining its asymmetric orientation (12). In effect, the algorithm compares an iteratively refined de novo model of the portal density to specific regions of the raw data in order to determine which one of the 60 equivalent icosahedral orientations is most likely to be the “true” asymmetric orientation. This procedure does not require a priori knowledge of the portal structure or its location in the capsid (12). To enhance the specificity of this selection process, we chose only those particles where the mean score for one vertex was statistically higher (P value < 0.05) than the score for the other 11 vertices. The asymmetric model was refined until no further improvements were observed. The 2,308 particles whose asymmetric orientations satisfied these statistical conditions were used for reconstructing the final asymmetric map, estimated at ~25 Å by the 0.5 Fourier Shell Correlation criterion (Fig. ​(Fig.22 a). The seemingly low resolution, in comparison to the total number of particles used (14), is a reflection of the fact that the B capsid's portal vertex is not readily apparent in the raw images, and as a result, the asymmetric orientations for some of the particle images might be incorrectly assigned.

Both the icosahedral and the asymmetric reconstruction of the HSV-1 B capsid (Fig. ​(Fig.1c1c and ​and2a)2a) show the characteristic T=16 capsid shell. The asymmetric reconstruction shows a large density underneath one of the 12 vertices (Fig. ​(Fig.2a).2a). In addition, the capsid shell at this vertex is markedly different from the other 11 vertices and lacks the typical penton density (Fig. 2b and c) seen there. This finding is consistent with the tomographic analysis of chemically treated HSV-1 B capsids, which concluded that the outer surface of the capsid at this position is formed by the top of the portal (3). The relatively weak capsid shell density seen at this vertex may arise due to difficulty in accurately assigning the asymmetric orientations, which would result in the inclusion of some penton density at this vertex. Even though the portal sits beneath the proteins that comprise the capsid shell, a portion of it is externally exposed and would be accessible to other viral or cellular proteins, as well as to antibodies specific to the portal protein UL6 (2, 15).

Figure ​Figure2d2d shows a cross-section of the capsid passing through the region of density beneath the unique vertex. The cross-section clearly shows that this density (Fig. ​(Fig.33 a) has an apparent 12-fold symmetry, as confirmed by rotational correlation analysis (Fig. ​(Fig.3b).3b). It is important to note that no symmetry was imposed during the generation of this map and that both the 5-fold nature of the surrounding capsid shell and the 12-fold symmetry of the internal density arise directly from the data. The simultaneous visualization of these two symmetry-mismatched features is only possible if they are not free to adopt different rotational positions but are fixed relative to each other. As the B capsids used in this study contained cleaved scaffolding proteins, we cannot eliminate the possibility that these proteins contribute to the density we see. However, the presence of this structure at only one vertex and its 12-fold nature, which is consistent with previously described portal structures for tailed bacteriophages (4, 10, 13), gives us confidence that we are detecting the HSV-1 portal. Although a 12-subunit structure was one of several forms found for isolated in vitro-assembled HSV-1 portals (19), this is the first definitive demonstration that the portal in a virus of eukaryotes shares the same stoichiometry as that of bacteriophages.

To assess the general structural features and arrangement of the HSV-1 portal subunits, we applied c12 symmetry to our asymmetric map and extracted the 12-fold-averaged portal density (Fig. ​(Fig.3c).3c). The organization and overall dimensions of our portal structure resemble those previously published for the 12-subunit form of the in vitro-assembled HSV-1 portal. However, comparison with the published figures appears to show subtle differences, for example, in the dimensions and features of the peripheral flanges (19). It is not clear whether this reflects genuine variation between the two structures in biochemically isolated and capsid-resident states or differences in the methods used for the two reconstructions.

Previous studies, which have reported two incompatible positions for the portal complex, one inside (3, 8) and the other outside the floor of the capsid shell (2), cast uncertainty over its true location. These studies, which used cryoelectron tomography to examine the nonicosahedral features of the capsid, were unable to resolve the subunit organization and provide unambiguous identification of the portal, leading to the confusion over its true location. The description of the portal in HSV-1 B capsids presented here resolves this uncertainty. The 12-fold symmetry of the internal density (Fig. ​(Fig.3)3) convincingly identifies it as the portal and establishes the location inside the capsid shell as being correct.