Cryo-Electron Tomography of the Herpesvirus Procapsid Reveals Interactions of the Portal with the Scaffold and a Shift on Maturation

Herpes simplex virus 1 (HSV-1) infects a majority of humans, causing mostly mild disease but in some cases progressing toward life-threatening encephalitis. Understanding the life cycle of the virus is important to devise countermeasures.

Herpesviridae, is accommodated in a correspondingly massive and complex capsid. During their elaborate life cycle (1), nascent capsids assemble in the nucleus as spherical precursor particles called procapsids (2)(3)(4), which conform to T=16 quasi-icosahedral symmetry and contain a spherical internal scaffold. The scaffold shell consists of the viral proteins pUL26.5 and pUL26, of which the latter has an additional N-terminal module, the viral protease (4). Upon completion of the procapsid, the protease is activated and proceeds to cleave the scaffold, initiating maturation of the capsid. Maturation involves (i) angularization of the capsid; (ii) reorganization of the interactions that engage the triplex proteins, as well as rearrangement of capsomers (1-3) from skew to more symmetric orientations; (iii) conformational changes (2,(5)(6)(7)(8) that establish interactions between capsomer subunits; and (iv) decoration of the mature capsid with additional proteins (9,10). Maturation dramatically increases the capsid's stability, countering the internal pressure exerted by the tightly packed genome (11). This process yields three capsid species, two of which remain in the nucleus. The A-capsid has a mature shell that does not contain any DNA and is essentially empty, while B-capsids retain a shrunken and somewhat disordered scaffold despite having matured into an angular shape. Only mature, DNA-filled capsids, called C-capsids, leave the nucleus, via the perinuclear space (reviewed in references 12 and 13), to undergo the subsequent steps in the viral life cycle.
DNA packaging occurs at a capsid vertex and requires seven proteins, namely pUL6, pUL15, pUL17, pUL25, pUL28, pUL32, and pUL33 (14). pUL6 forms a circular dodecamer similar to the portals of bacteriophages (15) and may facilitate procapsid assembly (16). Heterotrimers of pUL17 (one copy) and pUL25 (two copies) are bound to the pentons and peripentonal triplexes (17)(18)(19). A terminase complex composed of pUL15, pUL28, and pUL33 provides the force for packaging and subsequent cleavage of the DNA (20)(21)(22)(23)(24)(25)(26). The final protein, pUL32, has not been found as a structural component of capsids (14). These proteins replace a penton, presumably at one vertex of the capsid (27). While the structural disposition and transformations of some of these components have been elucidated, many details of assembly, maturation, and packaging remain to be described.
Early attempts to locate the portal-bearing vertex in capsids employed cryo-electron tomography (cryoET) (28,29). The tomograms were of low resolution (55 to 60 Å), complicating analysis. This led to placing the portal either at the level of capsomers (28) or inside the capsid shell (29). Subsequent analysis employing cryo-electron microscopy (cryoEM) and single-particle analysis (SPA), indicated that the portal is located inside the capsid shell (30,31). One of the difficulties in the SPA approach is that the overlap of one vertex with the large mass of the capsid (;100-fold ratio) makes it hard to distinguish a portal vertex from the penton vertices. Nevertheless, the extensive averaging of structural information means that correct assignments would dominate incorrect assignments distributed over the 11 other possibilities and reveal the location of the portal. Recent technological advances in cryoEM enabled the high-resolution reconstruction of various herpesvirus capsids showing the portal on the inside of one vertex (18,(31)(32)(33)(34).
One omission from the existing studies is the predisposition of the portal in the initial assembly product, the procapsid. It is also not clear that only one vertex bears a portal, as assembly can occur without a portal (2). In this study, we returned to cryoET to examine the portal vertices in procapsids and compared them to those in A-capsids as a control. In this effort, we developed the computational capability to find the portal vertices in a more reliable manner, allowing an estimate of the number of portal vertices in each capsid. From subtomogram averages of the portal vertices, we found abundant contacts between the portal and the scaffolding layer in the procapsid. We also found that the portal moves closer to the capsid shell on maturation.

RESULTS
Production of procapsids. Procapsids were produced using the M100 mutant that lacks the viral protease VP24 required for rapid maturation (35). We followed a gentle purification protocol (5,7) to minimize damage to the particles. In addition, we used an antibody binding to the major capsid protein (2) to stabilize the fragile and malleable procapsids. We performed cryoET on the isolated procapsids, collecting and analyzing 80 tilt series, in which the particles were well separated (Fig. 1A).
Production of capsid samples using a pUL25-deficient mutant. We decided to analyze mature A-capsids as a control to avoid interference from internal density features such as scaffolding (cf. B-capsids) or packaged DNA (cf. C-capsids). Note that Huet et al. (31) showed that the portal location and structure is very similar in all mature capsids (A, B, and C). The established protocol for purifying HSV-1 nucleocapsids from a wild-type infection yields only a minor fraction of A-capsids. To increase their yield, we used a mutant lacking the vertex protein, pUL25, that is defective in DNA packaging (36). This virus replicates in a complementary African green monkey cell line transfected with the UL25 gene, but in wild-type cells cannot produce progeny. In our protocol, centrifugation on small-scale sucrose gradients yielded bands for A-capsids and B-capsids but no band for the DNA-filled C-capsids (Fig. 1C). We isolated the UL25-null A-capsid band (here referred to as A-capsids) and collected tilt series for cryoET. Individual capsids in the tomograms were well spread out in the sample, and almost all capsids could be confirmed as A-capsids from their empty appearance (Fig. 1B).
Identification of the portal vertex in cryo-electron tomograms. Our aim was to locate the portal using cryoET to take advantage of the straightforward identification of unique features in three-dimensional (3D) maps. The approach used in an earlier tomographic protocol relied on simple template matching to find the portal vertex (28). Another protocol made use of the comparison of opposite vertices and assessing the density differences between vertices (30). In an effort to improve the detection of portal vertices, we developed a different algorithm aimed at detecting the vertex that is most different from the other vertices. We implemented this algorithm in a new program, called bico, added to the existing Bsoft package (37). An overview of our workflow is given in Fig. 2. From reconstructed tomograms, subtomogram volumes were extracted and aligned to an icosahedral reference, in our case a reconstruction of an HSV-1 C-capsid (Electron Microscopy Data Bank accession number EMD-7472 [38]). Central slice of a tomogram of procapsids. Some particles contain visibly degraded scaffold or show imperfections (red arrowheads) and were excluded from further processing. We took care to include only particles that were separated, spherical, and had a visibly intact scaffold (turquoise arrowheads). (B) Central slice of a tomogram of A-capsids. The individual particles are well defined and separated. Occasionally, a capsid with some internal content-probably some scaffold-was found in the field of view (yellow arrowhead). These particles were not included in further processing. (C) Small-scale sucrose gradient of a capsid mix obtained during initial purification from a UL25-null mutant, showing bands of A-and B-capsids. Bars, 500 Å.
Visualizing the Herpes Portal in the Procapsid ® Next, the 5-fold vertices of each capsid were extracted, and an average of all extracted and correctly oriented vertices was calculated. To focus only on the central part of a vertex, where we expect the biggest difference between penton and portal, we used a soft cylindrical mask in the comparisons of each vertex with the average (see Materials and Methods for mask details). For each set of 12 vertices, bico then classified each vertex set into two subsets. The first subset contained the 11 best-fitting vertices. The second subset contained the vertex with the worst fit, taken to be the portal vertex. The coordinates of the identified portal vertices were then transferred back into the extracted and aligned capsids, and a reconstruction of the capsid featuring the portal vertex was calculated. To reduce noise, this reconstruction was symmetrized 5-fold in the final step and filtered to 30 Å (just beyond the 34-Å resolution estimated by Fourier shell correlation; see Fig. S1 in the supplemental material).
The portal-bearing vertex of UL25-null A-capsids. We extracted a total of 164 subtomogram volumes of A-capsids from our tomograms. We identified the likely portal-bearing vertex in the map of each of the A-capsids and calculated an average (Fig. 3), obtaining a resolution of 34 Å (Fig. S1). In line with recent publications (18, 32, 33), we visualized the portal on the inside of the capsid, with its wider end facing FIG 2 Workflow used to identify and average the portal vertices. Individual particles (volumes each containing a capsid) extracted from tomograms and aligned with an icosahedral reference using a frequency space mask to compensate for the missing wedge. This allows the location of the 12 5-fold vertices of each capsid, which are then extracted and averaged. Each vertex is correlated with this average using a missing wedge mask in frequency space and a cylindrical real-space soft mask to isolate the region in a vertex occupied by the penton or portal (cyan dashed line). The vertex with the lowest (worst) correlation is assumed to contain the portal. This is followed by a further refinement to find the best 5-fold orientation around the vertex axis. The coordinates of the portal vertices were then used to transform their corresponding capsids into orientations that placed the portal vertex on the z axis at the top of the volume. toward the center of the particle (Fig. 3A). A superposed atomic model of the HSV-1 portal (PDB identifier [ID] 6OD7 [32]) fits well into the portal density (Fig. 3B). The portal has a narrow end called the "clip domain" that engages the triplexes (green arrow in Fig. 3B) (for the assignment of this density as a triplex, see Huet et al. [31]). The widest part of the portal, the "wing domain," extends toward the floors of the neighboring capsomers (blue arrow in Fig. 3B). The overall appearance and placement of this connection are consistent with a set of b-barrels and spine helices originating from the pUL19 hexons that were identified earlier in a high-resolution reconstruction (32). Knowing what it looks like, we could identify the portal vertex in individual particles in our tomograms (white arrows in Fig. 3C).
The portal in procapsids interacts with the scaffold. We selected tomograms with well-separated spherical particles (Fig. 1A) and extracted 368 subvolumes. As with A-capsids, we identified the portal-bearing vertices, taking care to adjust the mask so that it excluded the scaffold, which might otherwise outweigh the portal density during vertex sorting. After identifying the portal vertex in each procapsid, we calculated an average (Fig. 4) with a resolution of 34 Å (Fig. S1). Consistent with the A-capsids, the portal is located inside the capsid shell ( Fig. 4A) and again agrees with an atomic portal model (Fig. 4B). The portal penetrates deeply into the scaffold layer (yellow arrow in Fig. 4B), almost to the widest part of its wing, a position that allows the formation of extensive interactions. Interestingly, the nominally spherical scaffold shows a degree of perturbation caused by the portal, giving the scaffold an overall teardrop shape (Fig. 4A). We also note that-apart from in the portal region-the scaffold shell shows up in slices as three concentric shells (arrows in Fig. 4A), consistent with previous studies (5,7). Only tenuous contacts (blue arrow in Fig. 4B) are formed between the portal wing domain and the surrounding hexons, but the clip domain appears to interact The portal is located beneath the capsid floor, with its wider end facing the inside of the particle. A high-resolution model (pink; PDB ID 6OD7 [32]) is superimposed on the left half of the portal density. The portal forms connections to the triplex (green arrow) and the capsid floor (blue arrow indicates a link). Note that there is a diffuse density above the portal vertex that could be residual terminase. Bar, 100 Å. (C) Slices from tomograms of A-capsids that traverse the portal in each particle (white arrowheads). Bar, 500 Å.
Visualizing the Herpes Portal in the Procapsid ® with the triplexes (green arrow in Fig. 4B). The vertex overall has a funnel-like shape that likely accommodates DNA insertion by the terminase complex. Finally, the procapsid hexons also display faint residual density on their exterior parts that we attribute to the 6F10 antibody used to stabilize the procapsids during isolation ( Fig. 4A and Fig. 5C).
Conformational changes and movement of the portal during maturation. While the portal orientation is similar in procapsids and A-capsids, contrasts in vertex configuration come to light in a side-by-side comparison (Fig. 5). Upon maturation, the capsid becomes angular, causing all the vertices (both penton and portal) to translocate outwards. We assessed the movement of the pentons by shifting the penton vertex of the procapsid outwards until it overlays the A-capsid penton, yielding a translation of 29 Å. Similarly, we measured the translation of the portal to be 54 Å, thus moving it closer to the capsid shell. The portal in Fig. 5A appears to be somewhat separated from the capsid shell but is much closer in the mature capsid (Fig. 5B). During maturation, the capsid shell stabilizes by closure of the floor and by burying interfaces between the capsomers (8). Our results indicate that the portal also adopts a more stable interaction with the mature capsid shell.
In addition, we observe a shift in the triplex network surrounding the portal vertex, which narrows the vertex channel. These changes become even more apparent when the vertices are viewed from the top ( Fig. 5C and D). While triplexes are somewhat weakly defined in the procapsid portal vertex, they become much more visible in the A-capsid and even extend toward the 5-fold symmetry axis. It is possible that some of this extending density could be ascribed to components of the tegument, particularly pUL17. We recall that, upon maturation, the hexons rearrange from a skewed orientation to true hexameric symmetry (8).
We also compared the triplexes of the portal vertex with those of the opposing penton vertex in both maps. The triplexes in the procapsid portal vertex are poorly defined (Fig. 6A), making it hard to ascertain whether they are in the same conformation as those at the opposing penton vertex (Fig. 6B). In contrast, the orientations of the triplexes in the A-capsid are clearly the same for both vertices (Fig. 6C and D).
The number of portal vertices per capsid. While we implicitly assumed that there is exactly one portal per capsid, that is not necessarily the case. To determine the number of portal vertices in each capsid, we examined the distribution of correlation coefficients for the 12 vertices. Our aim was to generate two clusters: the few vertices with the worst correlation coefficients indicating portal vertices, and the rest as penton vertices. To separate the clusters, we took the biggest difference between successive correlation coefficients in a sorted list as the distinction between penton and portal vertices. Because we expected only a few portal vertices, only the six with the lowest coefficients were assessed. However, this strategy assumes there is at least one portal vertex. We compared the difference between the two worst vertices with the next three differences and assumed that when it is the lowest, it indicated a capsid without a portal vertex. The resultant distributions are shown in Fig. 7, giving average portal vertices per capsid of 1.07 for procapsids and 1.25 for A-capsids, consistent with previous biochemical estimations (16,27) but higher than the 0.88 derived from single-particle analysis of virions (calculated from the information given for Fig. S3 from reference 18). In our subtomogram averages, we included all of the capsids, but we did not consider the minor contribution of the capsids without portals to significantly change our conclusions. About 65% of the capsids have a single portal vertex, in agreement with the 73% reported for A-capsids (28). A still significant number (;20%) have two portal vertices, while those with three are rare. If the incorporation of portal vertices follows a random pattern, the expected distribution would be binomial in nature (blue lines in Fig. 7). The predominance of a singly incorporated portal indicates that the assembly is more specific than a random process. In virions, we would expect at least one portal per capsid, but the analysis of McElwee et al. (18) recovered only 88% of the expected portal vertices. We believe that the analysis of tomograms in 3 dimensions provides a more robust way to identify specific substructures.

DISCUSSION
Localization of the portal in procapsids compared to A-capsids. In a critical step of the HSV-1 life cycle, newly synthesized DNA is threaded into the capsid, which undergoes extensive conformational changes as it matures (5, 8), although the latter process can occur without DNA packaging (8) and may precede DNA packaging in vivo Visualizing the Herpes Portal in the Procapsid ® (39). In both herpesviruses and tailed bacteriophages, DNA is inserted through a 5-fold vertex via the dodecameric portal protein and driven by a terminase complex. While the overall structure of the HSV-1 portal, pUL6, has been known for some time (15), its orientation and placement in the context of the capsid shell have been subject to debate. The portal's overall shape makes it difficult to distinguish visually from capsomer pentons in micrographs, making consistent alignment of two-dimensional (2D) projections a challenge. In an earlier study, we used cryoET, aiming to characterize the portal vertex of HSV-1 (28). Here, we refined this approach by developing a new program designed to identify the portal vertex on individual capsids, thereby allowing coherent averaging. Applying it to HSV-1 procapsids and A-capsids, we located the portal on the inside of the particle, consistent with several recent SPA-based studies (18,30,(32)(33)(34). In confirmation, an atomic model of pUL6 (32) fits well into our portal density (Fig. 8). This density extends beyond the clip domain of the model, presumably representing some of residues 308 to 515, which were not modeled (corresponding   residues are also absent from a reconstruction of the Epstein-Barr-virus portal [40]). Interestingly, a corresponding turret-shaped density was discovered in the portal vertex of Kaposi's sarcoma herpesvirus (KSHV) (34); it was theorized that this helical turret is an extension of the portal's clip domain and could be a docking site for the terminase complex. If such a component is indeed an additional protein, it may only be added to mature particles upon the successful retention of DNA.
The association of the portal with the scaffold in the procapsid. Applying our workflow to procapsids led to several novel observations. The procapsid portal is located inside the vertex forming contacts with the scaffolding shell and deforming the scaffold proteins' spherical arrangement at the interaction site. Conserved tryptophan residues have been identified in HSV-1 pUL6 (41) (see also Fig. 2 in reference 42) that play crucial roles in interaction between portal and, most likely, conserved scaffold residues 143 to 151 of pUL26, particularly YYPGE (43,44). Figure 8B shows that these critical tryptophan residues arrange in a ring in the wing domain (orange arrow in Fig. 8B), as well as a ring in the base of the portal (blue arrow in Fig. 8B). We know the scaffold C terminus is bound to the capsid (45), which puts it in proximity to the portal wing domain. Conserved tryptophan residues are a feature not only of HSV-1 but also of some bacteriophages (see reference 46 and reviewed in reference 42). The scaffold probably shares the symmetry of the surface shell with which it coassembles, but it is not highly ordered and its apparent sphericity is enhanced by averaging. From previous studies we know that the protease at the N-terminal part of about 10% of the scaffold proteins is located toward the center of the procapsid (4), and the C-terminus interacts with the capsid shell (47) and portal (43). This means the protein is in an extended radial conformation consistent with a side-by-side packing of coiled coil dimers, as inferred by Pelletier et al. (48). In the procapsid, we see three peaks in the concentric density shells (Fig. 4), in agreement with the three domains proposed by Trus et al. (7).
Our subtomogram average of procapsids has a clear association between the scaffold and portal. In the study by Rochat et al. (30) on B-capsids, they raised the possibility that density close to the portal could be ascribed to the scaffold. However, the scaffold is cleaved and retracted, and the portal is closer to the capsid shell than in the Visualizing the Herpes Portal in the Procapsid ® procapsid. If there are remaining connections, they are much less extensive than in the procapsid.
Incorporation of the portal during procapsid assembly. The stage at which the portal incorporates into the procapsid is still unknown, with possibilities that it initiates assembly, inserts at some intermediate point, or inserts at the end (16). Procapsids can assemble in vitro without the portal present (2,3), and some of the capsids we analyzed appear to lack a portal vertex (Fig. 7). We also found an average copy number slightly larger than 1 and consistent with previous studies (16,27). If one assumes a random model of incorporation, it is expected to follow a binomial distribution. However, Fig. 7 shows it clearly deviates from a binomial pattern, suggesting a predisposition toward a single portal vertex per capsid. This suggests that the portal may facilitate procapsid assembly and thus ensure that a majority has at least one portal vertex. However, if it initiates assembly, it does not preclude incorporation of one or two additional portals.
Morphological changes upon capsid maturation. Cleavage of the C-terminal 25 residues of the scaffolding protein by the protease (47) accelerates maturation of the capsid shell. We used a virus defective in the protease to allow us to purify procapsids, which then slowly mature (8). On cleavage, the scaffold changes from extended in the procapsid (i.e., large-core particles) to somewhat contracted in the B-capsid (i.e., small-core particles). At the same time, the vertices of the capsid shell shift outwards by ;29 Å during angularization, with the portal itself moving even further by ;54 Å. The portal therefore disengages from the scaffold to some extent (as evident from the B-capsid map of Rochat et al. [30]), while strengthening interactions with the capsid shell. The portal movement is accompanied by the establishment of a support network with the newly established capsid floor (termed "suspension network" in another study [32]).
The overall shape of the portal vertex differs significantly between procapsids and mature capsids (Fig. 6). In procapsids, we note that the center of the vertex is essentially open, which would, in a biological context, be advantageous for the insertion of DNA. Triplex density is present around the portal vertex, but it is somewhat fainter than that around a penton vertex, which could result from a lack of interaction between the portal and the procapsid itself. In the mature capsid, the vertex is narrower, and it appears that density originating at the triplexes extends toward the vertex axis. This extended density may indeed be contributed by components of the tegument, particularly pUL17. While our A-capsids lacked pUL25, it has been shown that procapsids contain practically no pUL25 (25), and, given that the procapsid undergoes substantial conformational changes during maturation, it is likely that no tegument component protein is present on these immature particles. The extended, star-shaped density we see on the A-capsid portal vertex (Fig. 5D) may therefore act as a docking platform for copies of pUL25 that would be added in the course of a wild-type infection and a-helices contributed by pUL36 at a later stage. This platform, however, is evidently insufficient to retain DNA in the capsid, as the absence of pUL25 alone is enough to completely prevent the emergence of propagation-competent, DNA-filled C-capsids (36), underscoring pUL25's functional versatility as a DNA retention protein and a key player in nuclear egress (49).
Conclusion. Our routine for processing cryo-electron tomograms enabled us to visualize the portal in procapsids in comparison with mature A-capsids. Particularly revealing are the interactions of scaffold with the portal, likely to be involved in procapsid assembly. Quantitation of the number of portal vertices per capsid further solidifies a role in assembly. We also gained insight into the intricate maturation-induced changes in the portal vertex, visualizing an outward movement of the portal exceeding vertex movement and thus stabilizing the interaction with the capsid shell.

MATERIALS AND METHODS
Cell handling and propagation of virus stocks. Stocks of the UL25 deletion mutant (UL25-null) (36) were generated by using a complementary African green monkey cell line (both gifts from Fred L. Homa, University of Pittsburgh, Pittsburgh, PA). At roughly 75% confluence, monolayers in 175-cm 2 flasks were infected with virus stock at a multiplicity of infection (MOI) of 0.1 in 1 ml of phosphate-buffered saline (PBS) without calcium or magnesium per flask for 40 min at room temperature, followed by 5 min at 37°C. The cells were then overlaid with 15 ml virus growth medium (1Â minimal essential medium [MEM] Alpha, 1.5% [vol/vol] penicillin-streptomycin; Corning) and incubated at 37°C. At 48 h postinfection, virus-containing medium was clarified by centrifugation at 1,500 rpm for 5 min, and crude cell debris were removed from the supernatant by pelleting at 6,000 rpm for 15 min in a Beckman SW28 rotor. Virus was pelleted for 45 min at 19,000 rpm in a Beckman SW28 rotor. The virus pellets were resuspended in small volumes of PBS without calcium or magnesium, and aliquots of 500 ml were frozen at 280°C. The virus titer was assessed in complementary cells 48 h after infection.
Stocks of the procapsid-producing M100 line (35) were generated similarly using the complementary F-3 cell line. Stock titers were assessed as described earlier (5).
Production of capsids. Roller bottles of African green monkey cells were grown to about 75% confluence and infected with UL25-null virus stocks at an MOI of 10. By 14 h postinfection, cells had started to round up, but not detach from the surface. Cells were harvested by centrifugation at 1,500 rpm for 5 min, and any virus still attached to the cell surface was removed by incubation with TNE (20 mM Tris [pH 7.4], 500 mM NaCl, and 1 mM EDTA) on ice for 15 min. Cells were again pelleted for 5 min at 2,000 rpm, resuspended in TNE, and broken up by sonication. TX-100 (1.5% [vol/vol]) was added to solubilize nuclear membranes, followed by pelleting of crude cell debris for 30 min at 6,000 rpm in a Beckman SW28 rotor. The resulting supernatant was supplemented with a cushion of 30% (wt/vol) sucrose in TNE, and a mix of A-and B-capsids was pelleted for 1 h at 20,000 rpm in a Beckman SW28 rotor. The pellets were resuspended in TNE, supplemented with complete protease inhibitor cocktail (Roche), and applied to a preparative gradient containing a range of 20 to 50% (wt/vol) sucrose in TNE. The gradient was run for 1 h at 22,000 rpm in a Beckman SW55Ti rotor. Distinct bands containing the different capsid species were isolated, pelleted for 50 min at 22,000 rpm in a Beckman SW55Ti rotor, and resuspended in small amounts of TNE to reach an appropriate concentration for cryoET.
Procapsids of the M100 mutant in the viral protease were produced as described earlier (5). They were precipitated with the 6F10 antibody (raised against the triplex protein pUL19) for purification and capsid stability purposes (4).
Grid preparation and cryo-electron tomography. Purified capsids were mixed with Aurion anionic 10-nm bovine serum albumin (BSA) gold tracers (EMS) at a 1:1 ratio, and 3 ml were applied to Quantifoil copper grids covered with 300-mesh holey carbon (EMS), which were previously plasma-cleaned for 12 s with an argon-oxygen mixture (25% oxygen) in a model 1020 plasma cleaner (Fischione, Export, PA). Using a Leica EM GP plunger (Leica Microsystems, Buffalo Grove, IL), excess liquid was blotted off for 2.5 s, and grids were flash-frozen in liquid ethane. For the A-capsids, 21 tilt series ranging from 257°to 157°at 3°increments were collected at Â20,000 magnification with a defocus target of 22.5 mm on a JEM-2200FS cryo-electron microscope (JEOL USA, Peabody, MA) operating at 200 kV and equipped with an in-column energy filter and a K2 Summit detector (Gatan, Warrendale, PA). For procapsids, 80 tilt series ranging from 257°to 157°at 3°increments at Â81,000 magnification with a 22 mm defocus target were acquired on a 300-kV Titan Krios (Thermo Fisher Scientific, FEI, Hillsboro, OR) equipped with a postcolumn model 967 GIF Quantum LS energy filter with a K2 Summit detector (Gatan, Warrendale, PA). In both cases, data were acquired using SerialEM (50).
Tomogram reconstruction and identification of the portal-bearing vertex. Image processing and tomogram reconstruction were carried out using Bsoft (50,51). Tomograms were binned 2-fold, resulting in final pixel sizes of 3.66 Å for A-capsids and 3.61 Å for procapsids. For both capsids, promising particles were extracted in 450-Â 450-Â 450-pixel boxes (program bpick) and aligned (program bfind) to a reference of an HSV-1 C-capsid (38), using an appropriate frequency space mask to compensate for the missing wedge (created with the program bmissing). Given the correct alignment parameters, the program bpick was used to extract the 12 vertices for each capsid (150-Â 150-Â 150-pixel boxes) offset 140 pixels outward from the center of the particle on the 5-fold symmetry axes. Using the new Bsoft program, bico, each set of 12 vertices was cross-correlated against an average calculated from all vertices, and the vertex with the lowest correlation was deemed the portal-bearing vertex. For the whole capsid alignment, a frequency space mask was used in the correlation calculation (created with the program bmissing). In addition, a real-space mask was used to isolate a cylindrical region occupied by the penton at the vertex. The mask was created with the program beditimg, specifying a cylinder centered at 75, 75, 45 with a radius of 20, a height of 90, and a soft edge with a Gaussian width of 3 pixels. The softness of this mask was tuned to avoid high-frequency correlations in the comparison of vertex volumes. To address the symmetry mismatch between the 5-fold vertex and the 12-fold portal, the best 5-fold orientations of the portal-bearing vertices were found and refined (program bico). The coordinates corresponding to the portal vertices were then transferred to the parameter file referencing the full capsid volumes (program bico). Finally, the average of the whole capsids was calculated with the portal vertex located at the top of the volume. Graphics were prepared using PyMOL (52) and UCSF Chimera (53).
Data availability. Maps for the UL25-null A-capsid and the procapsid, both showing the portal vertex, were deposited in the Electron Microscopy Data Bank under accession numbers EMD-22378 and EMD-22379, respectively.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. FIG S1, PDF file, 0.1 MB.
Visualizing the Herpes Portal in the Procapsid ®