Research Article

Cryo-EM structure of a herpesvirus capsid at 3.1 Å

See allHide authors and affiliations

Science  06 Apr 2018:
Vol. 360, Issue 6384, eaao7283
DOI: 10.1126/science.aao7283

Focusing in on herpesvirus

The herpesvirus family includes herpes simplex virus type 1 (HSV-1), which causes cold sores, and type 2 (HSV-2), which causes genital herpes. Herpesviruses comprise a large DNA genome enclosed in a large and complex protein cage called a capsid (see the Perspective by Heldwein). Dai and Zhou used electron microscopy to determine a high-resolution structure of the HSV-1 capsid bound to the tegument proteins that occupy the space between the capsid and the nuclear envelope. The structure suggests how these components may play a role in viral transport. Yuan et al. describe a higher-resolution structure of an HSV-2 capsid, providing insight into how the shell assembles and is stabilized.

Science, this issue p. eaao7298, p. eaao7283; see also p. 34

Structured Abstract


Herpes simplex virus type 2 (HSV-2) is a sexually transmitted virus and is the leading causative agent of genital ulcer disease (GUD) worldwide. Patients with HSV-2 have a higher risk of acquiring human immunodeficiency virus (HIV) infection. HSV-2, as well as the closely related herpes simplex virus 1 (HSV-1), are simplexviruses with a natural-host range restricted to humans, belonging to the family of Herpesviridae, whose other members are responsible for a number of diseases, including congenital disorders (e.g., human cytomegalovirus) and even cancers (e.g., Epstein-Barr virus and Kaposi sarcoma herpesvirus). HSVs’ ability to establish a lifelong latent infection within hosts and recurrent reactivation from latency make them highly effective pathogens with seropositivity rates close to 100% in adult populations.


The herpesvirus virion is genetically and structurally one of the largest and most complex viruses known. It has a T = 16 (triangulation number) icosahedral capsid with a diameter of ~125 nm that not only protects the viral genome physically from damage but also plays an important role in the release of viral genome into the nucleus of the host cell. HSV capsid assembly requires the ordered packing of about 4000 protein subunits into the hexons, pentons, and triplexes that comprise the capsid. Previous studies have suggested that the directionality of triplexes on the capsid shell and disulfide bond formation between capsid proteins contribute to HSV capsid assembly, but in the absence of an atomic description of HSV capsids, the molecular basis that drives capsid assembly has remained elusive.


By using a “block-based” image reconstruction approach combined with a Ewald sphere correction, we have visualized the HSV capsid at 3.1-Å resolution by cryo–electron microscopy (cryo-EM) and have built an atomic structure, which includes 28,138 residues in the asymmetric unit, belonging to 46 different conformers of four capsid proteins (VP5, VP23, VP19C, and VP26). These organize into three types of hexons (central, peripentonal, and edge) that contain the major capsid protein VP5 and the small capsid protein VP26, pentons made up of VP5, and triplexes composed of VP23 and VP19C. Acting as core organizers, VP5 proteins form extensive intermolecular networks, involving disulfide bonds (25 per asymmetric unit) and noncovalent interactions, with VP26 proteins and triplexes, that underpin capsid stability and assembly. Together with previous low-resolution structural results, we propose a model for the ordered assembly of the capsid using basic assembly units (a triplex and its covalently linked lasso triangle formed by three VP5s), which then cluster into higher-order structures conforming to twofold symmetry and guide nascent assembly intermediates into the correct T = 16 geometry.


The marked improvement in the resolution of the structure of the herpesvirus capsid determined by cryo-EM allows the first steps toward understanding the drivers of assembly and the basis of stability of the capsid. In addition, the atomic structure could guide rational design of therapeutic agents for treating tumors and therapeutic strategies against HSV.

A 3.1-Å structure of HSV-2 B capsid.

Surface representation of HSV-2’s 1250-Å-wide capsid. Black lines represent particle icosahedral facets.


Structurally and genetically, human herpesviruses are among the largest and most complex of viruses. Using cryo–electron microscopy (cryo-EM) with an optimized image reconstruction strategy, we report the herpes simplex virus type 2 (HSV-2) capsid structure at 3.1 angstroms, which is built up of about 3000 proteins organized into three types of hexons (central, peripentonal, and edge), pentons, and triplexes. Both hexons and pentons contain the major capsid protein, VP5; hexons also contain a small capsid protein, VP26; and triplexes comprise VP23 and VP19C. Acting as core organizers, VP5 proteins form extensive intermolecular networks, involving multiple disulfide bonds (about 1500 in total) and noncovalent interactions, with VP26 proteins and triplexes that underpin capsid stability and assembly. Conformational adaptations of these proteins induced by their microenvironments lead to 46 different conformers that assemble into a massive quasisymmetric shell, exemplifying the structural and functional complexity of HSV.

Herpesviridae is a large family of double-stranded DNA (dsDNA) viruses that cause a number of diseases such as encephalitis, oral and genital blisters, congenital disorders, and even cancers (1). During the course of these infections, a key survival tactic of herpesvirus is to establish a chronic latent infection within the host (2). Consequently, herpesviruses are highly persistent pathogens with seropositivity rates close to 100% in adult populations (3). Herpes simplex virus type 2 (HSV-2) is a sexually transmitted virus and is currently the leading causative agent of genital ulcer disease (GUD) worldwide. Patients with HSV-2 have a higher risk of acquiring human immunodeficiency virus (HIV) infection (46).

The herpesvirus virion is one of the largest viruses known and is genetically and structurally one of the most complex. It has a diameter of ~200 nm and comprises four structurally distinct layers: a glycoprotein-bound envelope, an amorphous protein layer called the tegument, an icosahedral capsid, and at the core a large DNA genome. The capsid not only protects the viral genome from damage but also plays an important role in the release of viral genome into the nucleus of the host cell (7). The capsid initially assembles as a precursor procapsid that matures during DNA packing (810). The assembly of a procapsid requires components of the capsid shell and inner scaffold protein, pre-VP22a, a major component of the procapsid that is not present in the mature virion. After that, the procapsid matures through a major structural transformation triggered by limited cleavage of the C-terminal 25 amino acids of scaffold proteins, which are then removed, possibly during entry of DNA into the capsid (8, 11).

Three distinct types of capsids—called A, B, and C capsids—have been detected in lysates of infected cells (12). A capsids are empty and result from abortive DNA packing, B capsids contain a core of scaffold proteins, but no DNA, and C capsids, also known as mature capsids, are fully packed with DNA. B capsids retain a cleaved form of the scaffold proteins, suggesting that the capsid is either in the process of packaging or that the scaffold was not removed in time to allow DNA packaging (13). B capsids may be intermediates in viral assembly or abortive capsid forms generated as a result of unsuccessful proteolytic digestion of scaffold proteins or a failed attempt to package viral DNA. All three types of capsids have icosahedrally symmetric shells (triangulation number of T = 16) composed primarily of 955 copies of the 150-kDa major capsid protein, VP5, arranged as 150 hexons (on the triangular edges and faces) and 11 pentons (on all but one vertex) (14). Other critical components include the dodecameric pUL6 portal complex, which occupies the twelfth vertex; 320 copies of the “triplex” that consists of two copies of VP23 and one copy of VP19C and connects adjacent capsomers located on the capsid (15); 900 copies of VP26, which is one of the components of the outer surface of hexons but not pentons (16). Compared with A and B capsids, C capsids have substantially higher occupancy of the capsid vertex-specific component (CVSC) (a putative heterotrimer consisting of the pUL17, pUL25, and pUL36 proteins) that binds specifically to triplexes adjacent to pentons (17). In addition to the shell proteins mentioned above, B capsids contain a large amount of the scaffold protein VP22a, which is produced by the UL26.5 gene, and smaller amounts of the two proteins coded by the UL26 gene (18). The UL26-encoded protease cleaves itself between amino acids 247 and 248, which separates pUL26 into an N-terminal domain bearing the protease called VP24 and a C-terminal domain called VP21 (19). The pathway for the assembly of mature capsids of herpesvirus has striking parallels with those of dsDNA bacteriophages (20). Furthermore, herpesviruses, although more complex, share the canonical HK97 capsid protein fold with the dsDNA tailed bacteriophage, indicative of common ancestry (21).

Efforts of many investigators over two decades have pushed the resolution of cryo–electron microscopy (cryo-EM) analysis of herpesvirus capsid from 15 to 6 Å (14, 15, 2224). More recently, a 3.9-Å structure of human cytomegalovirus (HCMV) (a member of the Herpesviridae family belonging to the β-herpesvirinae subfamily) capsid was reported (25). We have determined the structure of HSV-2 B capsid to 3.1 Å by using a combination of “block-based” reconstruction and accurate Ewald sphere corrections and have built reliable atomic models for the capsid, which provide insights into its stability and assembly.

Characterization and structure determination

HSV-2 (MS strain) was grown in Vero cells at 37°C and purified by a series of ultracentrifugation steps (see materials and methods). Because the cryo-EM imaging of such large intact HSV virions (>200 nm in diameter) would lead to low signal-to-noise ratios and a depth-of-field gradient through the virions, we purified HSV-2 capsids in the presence of detergents. Three types of capsid of HSV-2 were separated by ultracentrifugation and characterized as 713S A capsid; 924S B capsid and 1205S C capsid (fig. S1). Cryo-EM micrographs of the purified HSV-2 A, B, and C capsids were recorded using an FEI Titan Krios electron microscope equipped with a Falcon detector (fig. S1). A comparison of reconstructions of these three particle types at 8-Å resolution reveals an overall similar morphology for A, B, and C capsids. But, the presence of a scaffold core with a diameter of ~660 Å is observed in the B capsid. Further, in the C capsid, the dsDNA, seen as regularly spaced density shells (~25 Å apart), and the CVSC are observed (Fig. 1A).

Fig. 1 Architecture of the HSV-2 B capsid.

(A) Central slices of HSV-2 A-, B- and C-capsid reconstructions at 8 Å. The insets show dsDNA density and CVSC density. The scaffold core is marked with a red dotted circle. P-Hex (peripentonal), C-Hex (central), E-Hex (edge), and Pen (penton) denote hexons and pentons, respectively. (B) Locations of the various components of the capsid in the cryo-EM map of HSV-2 B capsid. Black lines, particle facets. The threefold icosahedral symmetry axis is marked as a triangle. P (peripentonal), C (center), and E (edge) denote hexons, and Ta to Tf denote triplexes. (C) Cryo-EM map of an asymmetric unit and local electron density maps are shown. The inset shows the density maps (mesh) and atomic models of VP5 and VP23, which illustrate side-chain features. Residues with side chains are labeled; aa denotes amino acids. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

Normal reconstruction strategies yield a reconstruction of ~4-Å resolution with icosahedral symmetry imposed. There are two bottlenecks that limit the cryo-EM resolution of this 1250-Å-diameter virus. First, the complex architecture of the virus does not strictly conform to the icosahedral symmetry. This could be due to the intrinsic structural flexibility associated with the particle or structural perturbation caused by ultracentrifugation during the purification process. Second, the gradient in defocus through the capsid arising from the large size of the virus limits the attainable resolution. To overcome these two problems, we developed a reconstruction method named block-based reconstruction, which together with accurate Ewald sphere corrections (see materials and methods), resulted in a 3.1-Å structure of B capsid (Fourier shell correlation = 0.143 criterion) (26) after processing 45,000 particles (Fig. 1B and fig. S2). The backbone of the polypeptide, as well as many side chains, was clearly defined (Fig. 1C), allowing an atomic model of the HSV capsid shell comprising the four capsid proteins VP5, VP26, VP23, and VP19C to be built de novo (figs. S3 to S5). The model was refined and validated using standard x-ray crystallographic metrics (table S1). An asymmetric unit consists of 16 copies of VP5—which exist in a C-Hex (central hexon), a P-Hex (peripentonal hexon), one-half of an E-Hex (edge hexon) and one-fifth of a Pen—and 15 copies of VP26 and 51⁄3 triplexes (Ta, Tb, Tc, Td, Te, and one-third of Tf, located at the threefold axis) (Fig. 1B). In our atomic model, the asymmetric unit contains all 16 copies of VP5, 15 copies of VP26, and 5 triplexes (Ta, Tb, Tc, Td, and Te) (fig. S2C). The Tf triplex is on the icosahedral threefold axis; thus, densities for Tf were destroyed during the icosahedral averaging (fig. S2C).

VP5 structure and protrusion features

The 150-kDa major capsid protein (VP5) forms both the hexons and pentons and acts as a core organizer in the assembly of the capsids. None of the 16 copies of VP5 in the asymmetric unit are identical (fig. S6), suggesting that the microenvironments surrounding each subunit in the T = 16 icosahedral capsid are not strictly equivalent, which is consistent with the basis of quasi-equivalent packing. Structure-based phylogenetic analysis using Structure Homology Program (SHP) software places P1 and P6 (subunits from C-Hex, P-Hex, E-Hex, and Pen denoted as C1 to C6, P1 to P6, E1 to E6, and Pen1 to Pen5, respectively) between typical hexon- and penton-type VP5s (Fig. 2A) (27). As the major capsid protein in HCMV, each VP5 can be divided into seven domains: upper (residues 482 to 1047), channel (residues 407 to 481 and 1332 to 1374), buttress (residues 1123 to 1331), helix-hairpin (residues 195 to 237), Johnson-fold [residues 71 to 194, 238 to 294, 372 to 406, and 1048 to 1122, a characteristic fold first identified in bacteriophage HK97 gp5 (28)], dimerization (residues 295 to 371), and N-lasso (residues 1 to 70), which constitute three sections: upper, middle (channel, buttress, and helix-hairpin) and lower (Johnson-fold, dimerization, and N-lasso) (Fig. 2B and fig. S7). The upper and middle sections make up the capsid protrusions, whereas the lower section forms the capsid floor (Fig. 2, B and C). A comparison of typical hexon- and penton-type VP5s shows highly conserved upper and middle domains but reveals a conformational change in the lower section that has implications for the differences in assemblies of the penton and the hexon (Fig. 2B). The conformational changes include a ~20° counterclockwise rotation of the central helix and a ~25° clockwise rotation of the E loop, which straightens the subunit and increases the curvature of the lower shell of the penton. In addition, in the penton, the refolding of the N-lasso leads to formation of a β-hairpin (βA2 and βB1), decreasing its protrusion by ~25 Å in comparison with hexon-VP5, whereas reorganization of the dimerization domain creates a 41-Å-long helix (αN), lining the inner capsid surface beneath the penton, yielding a comparatively slim penton with a greater height of ~170 Å and a reduced external diameter of ~160 Å at the “lower” end, relative to those in the hexon (Fig. 2, B and C, and Movie 1).

Fig. 2 Structure and organization of VP5.

(A) Structure-based phylogenetic tree of 16 copies of VP5 present in an asymmetric unit. (B) Ribbon diagram of the typical hexon- and penton-VP5. N and C termini are labeled, and major conformational changes are marked with dashed lines. A model shown in the middle depicts the conformational changes. The upper and middle sections make up the capsid protrusions. (C) Comparison of the surface of C-Hex (up) and the Pen (down); side view (left) and top view (middle) are shown. Each subunit is depicted in a different color. Insets illustrate the differences in the central channel. The residues lining the narrowest region of the central channel are marked and labeled.

Movie 1 Structures of four types of VP5 conformers.

Atomic models of the four types of VP5 conformers are colored by domain. The domains with significant conformational changes are highlighted.

Despite the overall similarities between the channel domains, the axial channels comprising the β-sheet ring within hexons and pentons show distinct differences. The narrowest region of the hexon channel has an internal diameter of ~10 Å, whereas the penton channel is completely blocked by five histidine residues (His 419) (Fig. 2C); this contrasts with the HCMV penton, which has a ~14-Å-diameter channel (fig. S8). Further, the hexon channel is remarkably negatively charged (fig. S8). In contrast, the bottom of the penton channel presents more positive charges, indicating a possible role for this region in interactions with the genomic DNA (fig. S8). The electrostatic surface properties of the channel of the hexon and penton are appreciably different in HCMV, suggesting a distinctive role for the hexon channel in accommodating DNA (25). In addition to the channel domain, the upper domains of hexons and pentons make different contacts (fig. S9).

Adaptations of P-Hex

While surrounding the penton, P-Hex is also adjacent to neighboring hexons (Fig. 3A). Given the fact that the penton exhibits a distinct pattern of assembly compared with the typical hexon, subunits within P-Hex have to adopt conformational changes to allow three distinct interactions: with typical hexons, with pentons, and with themselves (Fig. 3A). Overall, the structure of the P-Hex resembles that of the typical hexon, with an indistinguishable external surface but with differences in the inner protein shell (Fig. 3B). Among the six copies of VP5 of the P-Hex, four are similar to the typical hexon-type VP5, whereas P1 and P6, adjacent to the penton, exhibit notable conformational changes at the N-lasso and dimerization domains (Fig. 3B and Movie 1). A superposition of P1 as well as P6 onto C1 and Pen1 shows that P6 is similar to Pen1 around the dimerization domain, where they both have a 41-Å-long helix, whereas the N-lasso domain of P6 is more similar to that of C1. Conversely, P1 and C1 share common structural features at the dimerization domain, whereas P1 adopts an analogous fold to the N-lasso domain of Pen1 (Fig. 3C). These adaptations in P1 and P6 contribute to enhanced interactions of P-Hex with the penton. Although the penton VP5 (Pen5) loses its lasso to an adjacent P6 (P-Hex), it establishes new quasi-equivalent twofold interactions with a P1 N-lasso domain, which would normally clasp a neighboring penton VP5 (Pen1) but instead adopts a short N-lasso fold (Fig. 3D). In addition, a novel helix pair comprising two long αN helices from the dimerization domain of Pen1 and P6 is observed at the inner surface beneath the penton, forming a quasi-equivalent twofold interaction (Fig. 3D).

Fig. 3 Conformational adaptations of P-Hex.

(A) Schematic diagram of the binding mode of different capsomers. Domains with significant conformational differences (dimerization and N-lasso domains) in the floor of the capsid are highlighted. (B) Ribbon diagram of the side view (left) and the bottom view (right) of P-Hex. The N-lasso and dimerization domains are labeled and highlighted. The red star indicates the region where the conformational adaptations occur. (C) A superposition of C1, P1, P6, and Pen1. At N-lasso domain, P6 is similar to C1, and P1 is similar to penton-VP5. P1 and C1 share common structural features near the dimerization domain; P6 adopts folds similar to those of penton-VP5. (D) Overview of penton, P1, P6, and P5. The insets demonstrate how the local geometry changes near the fivefold axis. The Pen5 VP5 establishes new quasi-equivalent twofold interactions with a P1 N-lasso domain instead of its lasso to an adjacent P6. A novel helix pair is formed by two long αN helices from the dimerization domain of Pen1 and P6.

Extensive interactions at the inner capsid surface

The HSV capsid shell is assembled via extensive interactions of the lower sections of VP5, including both intracapsomer and intercapsomer interactions (Fig. 4 and Movie 2). In the penton, an archetypal example is between Pen4 and Pen5. β strands (βB1 and βA2) from Pen4’s N-lasso domain, together with β strands (βD and βE) from Pen5’s E loop and βL1 from Pen5’s dimerization domain, form a five-stranded β sheet, which is further stabilized by Pen5’s αN (Fig. 4A). In the hexon, βB from P3′ and βD, βE, and βL from P2′ make up a four-stranded β sheet (Fig. 4B). Two types of intercapsomer interactions, Pen–P-Hex and hexon-hexon, are observed. Regarding the former, a helix pair comprising two long αN helices from the dimerization domain of a penton subunit and an opposing P-Hex subunit (P6) is further stabilized by two adjacent central helices (αG), firmly holding the penton and the P-Hex together (Fig. 4C). Additionally, Pen1’s αE and P1’s αF form a small helix pair, as do Pen5’s αF and P6’s αE (Fig. 4C). Moreover, two sets of five-stranded β sheet beneath the helix pair from the Pen (formed by βA2 and βB1 from Pen5 and βD, βE, and βL1 from Pen1) and P-Hex (formed by βA2 and βB1 from P1 and βD, βE, and βL1 from P6) form quasi-equivalent twofold interactions, further strengthening the contacts between the penton and P-Hex (Fig. 4C). Hexon-hexon interactions occur between four VP5s and are characterized by the lassoing of the N-lasso domain of P3 and C3 and the quasi-equivalent dimerization between the two pairs of helices (αM and αN) in the dimerization domain of P2 and C2 (Fig. 4D). The N-lasso (e.g., C3) protrudes out and clasps βB (P3′ N-lasso), and βD and βE (P2’s E loop) across a quasi-twofold axis (Fig. 4D). In the same place, βA joins βB, βD, and βE to form a four-stranded β sheet (Fig. 4D). In addition, three sets of N-lasso interactions form an enclosed triangle around quasi-threefold axes, firmly connecting three hexons (Fig. 4E). Overall, hexon-hexon contact areas are ~35% larger than penton-hexon interaction areas, probably accounting for the vulnerability of pentonal vertices in HSV capsids (29).

Fig. 4 Overview of the interactions at the inner capsid surface.

A schematic representation of the inner surface of the capsid structure is shown. The N-lasso domain and the dimerization domain of each subunit are shown as cartoons, whereas the remaining structure is rendered as a surface. One copy of the Pen, two copies of the P-Hex, and one copy of the C-Hex are shown. The view is from the inside of the capsid. (A to E) Enlarged views of the interactions (boxed) are shown with secondary structural elements colored and labeled. Each subunit participating in the interactions is depicted in a different color and labeled. (A) and (B) are intracapsomer interactions; (C) to (E) are intercapsomer interactions.

Movie 2 Extensive interactions at the inner capsid surface.

One copy of the penton, two copies of the P-Hex, and one copy of the C-Hex are shown. Each subunit participating in the interactions is outlined in different colors. Major secondary structural elements are also labeled in different colors based on the subunits that they come from.

VP26 caps hexons only

VP26 (112 residues in length), one of the components of the outer surface of HSV capsids, is ~50% larger than its homolog SCP in HCMV. Unlike HCMV, where SCPs bind both hexons and pentons (30), VP26 only caps hexons (Fig. 5A) in HSV. VP26 of HSV exhibits no significant structural similarity with SCP. VP26 comprises an N-terminal domain of an “M” loop plus αA and a C-terminal domain of αB along with a long insertion loop. The two domains are connected by the extension loop (Fig. 5A). The C-terminal domain plays essential roles in capping hexons through two regions: (i) αB and one half of the insertion loop insert into the cleft formed by two adjacent hexon-VP5 upper domains and make hydrophobic interactions; (ii) the second half of the insertion loop accesses a shallow cleft in the adjacent hexon-VP5 upper domain beneath the extension loop of the adjacent VP26 and interacts through charge and hydrophilic interactions (Fig. 5A). In the N-terminal domain, only αA interacts with the hexon-VP5 upper domain (Fig. 5A). Our structural analysis is consistent with a previously reported mutagenesis study of HSV-1 VP26, which showed that residues 50 to 112 are sufficient for binding hexons (31). To find out if there are any structural constraints that prevent VP26s from capping pentons, we superimposed the structure of the hexon-VP5 upper domain bound to VP26 onto the penton-VP5 upper domain. The analysis suggested a complete loss of interactions between VP26 and the adjacent VP26 and significantly reduced contacts with the adjacent VP5 (Fig. 5B). Therefore, VP26 can only cap hexons, but not pentons. Six copies of VP26 form a ring structure by end-to-end interactions, filling the gaps between the heads of neighboring subunits in a hexon and further stabilizing hexons.

Fig. 5 VP26 structure and its interactions with VP5.

(A) VP26s cap the hexon but not the penton. One set of VP5 and VP26 is colored in light tone; the adjacent set of VP5 and VP26 is colored in bright tone. The rainbow ribbon of VP26 is shown in the leftmost panel of the first row. Enlarged views of interactions (1) to (3) (boxed) are shown. Residues involved in the interactions are shown as sticks in (1) and (3), with VP5 colored in violet (light violet for the adjacent VP5) and VP26 colored in green (light green for the adjacent VP26). In (2), VP5s and VP26s are shown as surface representations and cartoons, respectively, which are colored by residues’ hydrophobic characters: hydrophobic (red) to hydrophilic (white) gradient. (B) The superposed model of penton-VP26s based on the complex structure of hexon-VP26s. The penton is represented as surface, and VP26s are shown as cartoons. The secondary structural elements are colored and labeled. The neighboring VP5s are colored in brown and light brown, and the corresponding VP26s are colored in green and light green, respectively.

The triplex and its interactions with VP5s

The triplex, a structural feature of herpesvirus capsid shells, consists of two VP23 conformers (denoted as VP23-1 and VP23-2) and one copy of VP19C and lies among three adjacent capsomers, linking them together (Fig. 6A and Movie 3). From the bottom view, the triplex adopts a trimeric arrangement to match the quasi-threefold symmetrical environment (Fig. 6A). Overall, the triplex is relatively compact and rigid; the structure being reenforced by four pairs of disulfide bonds (C5-C86 within VP23-1 and VP23-2, C194-C325 within VP19C, and C267-C298 between VP19C and VP23-1) (Fig. 6B and fig. S10). The two copies of VP23 have an α+β fold, comprising an N domain (residues 1 to 144 and 291 to 318) and a C domain (residues 145 to 290), which form the lower and the upper portions, respectively (Fig. 6C). VP23-1 and VP23-2 interact through a number of helices (αF-αJ) in the C domain (Fig. 6D). Interestingly, the conformations of VP23-1 and VP23-2 differ in the C domain due to a ~65° rotation of the αI (αI in VP23-1 becomes two helices, αI and αJ, in VP23-2) helix and a large rearrangement of the αF-αH region. However, the two N domains are very similar (Fig. 6E). VP19C, ~150 residues longer than VP23, makes up the third component of the trimeric fold. A disulfide bond (C267-C298 between VP19C and VP23-1) and helix bundle interactions (including αD and αI from VP19C, αE from VP23-1, and αJ, αD from VP23-2) between VP19C and two VP23 conformers integrates VP19C into the triplex (Fig. 6B). The N-terminal region (residues 1 to 105), comprising three helices, penetrates the capsid floor around the quasi-threefold axis, anchoring the triplex in the capsid. The structure also supports the location of a nuclear localization signal (32) in VP19C at its N terminus (residues 50 to 61), which is presumed to bind DNA in vitro (Fig. 6F) (33), albeit densities for DNA binding to the N-terminal arm of triplex in the C-capsid reconstruction have not been observed yet (23). A number of β strands of the N domains from VP23 and VP19C contact the capsid floor (fig. S11). These contacts contribute the majority of the triplex-capsid interactions. Additionally, triplex-capsid protrusion interactions further fix the orientation of the triplex; these involve charge and hydrophobic interactions with the αt helix, th, and yz loops of the buttress domain of three adjacent VP5s, as well as one intermolecular disulfide bond (C1317-C179 between C2 and VP19C) (Fig. 7A). Previous studies have reported that disulfide bond formation between triplex and VP5 contributes to HSV capsid stability and assembly (34, 35). The disulfide bonds observed in B capsids structurally support these results to some extent. However, the disulfide bonds observed in the structure of B capsids may not be representative of those found in C capsids. We show that covalent as well as noncovalent interactions between triplexes and surrounding VP5s further stabilize the capsid.

Fig. 6 Structural features of triplex.

(A) Overview of Td triplex and its surrounding hexons. Td triplex and the subunits of the hexons that exhibit major interactions with Td triplex are highlighted with different colors (color scheme is the same as in Fig. 1C); the other subunits are shown in gray. (B) Overall structure of the triplex. VP23-1, VP23-2, and VP19C are colored in red, blue, and green, respectively. Electron density maps (blue mesh) for intramolecular and intermolecular disulfide bonds are shown in zoomed-in boxes. (C) Ribbon diagrams of VP23-1 and VP23-2. (D) Surface representation of the dimer consisting of VP23-1 and VP23-2. Shown are the side view (left) and top view (right). The color scheme is the same as in Fig. 4C. (E) Superimposition of VP23-1 (red) over VP23-2 (blue), viewed from the side. (F) Ribbon diagram of VP19C, color ramped from the NH2 terminus (blue) to the COOH terminus (red). The reported nuclear localization signal (NLS) is labeled with a red star.

Movie 3 Structure of the triplex and interactions with VP5s.

An overview of a triplex and its surrounding VP5s is shown. Both the covalent and noncovalent interactions within the triplex and between the triplex and the surrounding VP5s are highlighted and labeled.

Fig. 7 Rigid unit interactions and global capsid organization.

(A) Overview of a triplex and its surrounding six VP5s (e.g., Td, top view). A triplex and its covalently linked lasso triangle comprise a structurally rigid unit. The color scheme is same as in Fig. 6A. Insets show the detailed topside interactions within a structurally rigid unit. Enlarged views of interactions (1) to (4) (boxed) are shown. (1) Charge and covalent interactions; the electron density map (blue mesh) for the intermolecular disulfide bond is shown. (2) to (4) Hydrophobic interactions, the triplex, and VP5s are shown as surface representation and cartoon, respectively. Residues involved in the interactions are shown as sticks in (1) to (3). (4) Regions involving hydrophobic interactions are colored by residues’ hydrophobic characters: hydrophobic (red) to hydrophilic (white) gradient; other regions are colored as in Fig. 6A. (B) Two icosahedral facets are shown in surface representation at left with fivefold axis, C-Hex, E-Hex, and P-Hex labeled as 5, C, E and P, respectively, at the corresponding positions. The color scheme is the same as in Fig. 1C. Triplexes (a few sets of Ta to Te) are labeled in different colors according to the asymmetric units they come from. The triplex Tf is colored in purple and labeled without a circle. The local quasi-equivalent twofold axes, where Ta-Tc, Tc-Tb, Tb-Td, and Td-Te exhibit twofold symmetry, are marked with black ellipses. The black triangles represent structurally rigid units (the centers of these three VP5s are connected by black lines). (Right) Pictorial representation of the rigid unit/rigid unit interactions. A triplex [drawn as an inner triangle with red (VP23-1), blue (VP23-2), and green (VP19C) colors and labeled by a circle] and its covalently linked lasso triangle (drawn as an external triangle with different colors according to which hexons the three VP5s come from: C-Hex, violet; E-Hex, cyan; P-Hex, slate) make up a putative assembly unit. The length for each side of the external triangle was calculated by measuring the distance between the centers of two VP5s within a lasso triangle. The associations between assembly units of Tc and Td′ (from the neighboring asymmetric unit) are highlighted in light yellow.

Implications for assembly

In the broader context of the capsid structure, six VP5s from three different capsomers are arranged around a central triplex and three of them form one lasso triangle, upon which the triplex sits (Fig. 7A). Notably, interactions between the central triplex and the lasso triangle (interaction area ~4700 Å2) dominate the contacts with all six VP5s (with an interaction area of ~5300 Å2). A disulfide bond (between the triplex and one VP5 from the lasso triangle) and these anchoring interactions integrate the triplex into the lasso triangle, forming a structurally rigid unit (Fig. 7, A and B). The lasso triangle is not an equilateral triangle (each center of the three VP5s shows variable distances to the other two centers), with an average of ~10% deviations among three sides, and each lasso triangle varies slightly, depending on the microenvironments (Fig. 7B). As suggested previously (36), the triplexes confer a pronounced directionality on assembly of capsid shell. Specifically, VP19C from Ta is oriented outward from the penton, and Ta-Tc, Tc-Tb, Tb-Td, and Td-Te exhibit twofold symmetry, albeit with no direct interactions between them (Fig. 7B). In the absence of scaffold proteins, VP5s and triplex proteins formed capsids with a T = 7 shell (37), suggesting that the internal scaffold proteins play a role in controlling the correct symmetry of the capsid shell. Moreover, increasing evidence indicates that the triplexes are essential for the correct assembly of the capsid (3840). It is quite likely that a structurally rigid unit comprising a triplex, the VP5 lasso triangle, and probably including scaffold proteins, might act as an assembly unit to regulate sequential assembly. The twofold symmetric interactions between these assembly units result from VP5-VP5 interactions, which can explain the triplex directionality (Fig. 7B). One special case is that of the assembly unit Tc (the assembly units are named according to the triplex that they contain) and the assembly unit Td′ from the neighboring asymmetric unit, which are not related by a strict twofold axis. However, due to capsid assembly forces, the neighboring sides (highlighted in light yellow color in Fig. 7B) of the assembly unit Tc and Td′ are adjusted to approximately the same length in order to enhance the interactions and retain the quasi-equivalent twofold axis (Fig. 7B). The assembly unit Tf is another special case due to its location on the threefold axis, so that the imposition of icosahedral symmetry in the reconstruction averages the Tf density.

HSV B capsids structurally resemble C capsids

A feature of the infections caused by herpesviruses is the production of three different types of mature capsid forms (A, B, and C capsids) in the host cell nucleus. Whether A and B capsids are abortive forms or assembly intermediates is still debated (39, 4143). Comparison of the low resolution cryo-EM structures of A, B, and C capsids indicates that except for the presence of capsid-associating tegument proteins in C capsids, structures of all three types of capsids are very similar (13, 44). A high-resolution structural analysis to verify this inference is lacking. To address this issue, we fitted our high-resolution HSV-2 B capsid coordinates with five rigid asymmetric units into the ~7-Å-resolution cryo-EM reconstruction of the HSV-1 virion (EMD-6386), which exhibits ~95%, ~93%, and 83% amino-acid identities in VP5, VP23, and VP19C, respectively, with HSV-2 (23). The fitting gave a correlation coefficient of 0.95, showing that all the capsid proteins, including their secondary structural elements, were well aligned (Fig. 8). The structures and organization of VP5 assembled in the penton and hexons, together with the triplexes, showed no notable differences between the HSV-2 B capsid and the HSV-1 C capsid (Fig. 8). A striking feature of our HSV-2 B-capsid structure is a set of five helix pairs lining the inner capsid surface beneath each penton, which perfectly match densities at equivalent positions of the HSV-1 virion (Fig. 8). Because of the low resolution of the reconstruction of C capsids, it is difficult to identify disulfide bonds of B capsids in C capsids with confidence. The primary sequences of HSV-2 capsid proteins share only ~27% amino acid identity with those of the capsid proteins of HCMV. Despite this, a superposition of our structures of HSV-2 capsid proteins (VP5 and triplex) over the recently reported structure of HCMV reveals overall structural similarities. However, significant conformational differences between the capsids of the two viruses are observed, which is in line with the specific differences between the virus types (fig. S12). Thus, conserved structural features may underlie similar mechanisms of capsid assembly across all herpesviruses.

Fig. 8 Structural comparison of HSV-2 B capsid with HSV-1 virion.

Fitting of HSV-2 B capsid coordinates with five asymmetric units into the ~7-Å resolution cryo-EM reconstruction of HSV-1 virion (emd-6386) (23). The penton, C-Hex, P-Hex, and E-Hex are colored in orange, violet, slate, and cyan, respectively. The five helix pairs lining the inner capsid surface beneath the penton is marked, and a zoomed-in view is shown in the red inset. Zoomed-in views of the penton, C-Hex, and triplex are shown in orange, purple, and green insets, respectively. Each subunit within the penton and hexons is depicted in a different color.


The herpesvirus capsid initially assembles as a precursor procapsid that subsequently matures during DNA packing (810). HSV procapsid assembly is a complex process requiring the ordered packing of some 4000 protein subunits (19). An in vitro assembly system based on a panel of recombinant baculoviruses encoding HSV procapsid proteins has revealed that VP5, triplex, and scaffold proteins are sufficient to produce a capsid with a correct T = 16 shell (8, 45). The imposition of icosahedral symmetry in the reconstruction of the whole capsid likely limited the resolution of the scaffold core. The low-resolution map with a diameter of ~660 Å does not provide information on how scaffold proteins determine the correct symmetry of the particle (Fig. 1A). Together with previous studies on in vitro–assembled procapsids (11), these suggest that the scaffold may not need to adopt a specific organization to fulfill its function, on the condition that its overall dimensions are within the tolerance necessary to ensure the correct shell conformation. Although the portal structures of dsDNA bacteriophages have been extensively studied (46, 47), it has proved difficult to study the portal structure of herpesviruses because there are no sentinel markers for the location of the portal in herpesviruses. Available evidence seems to indicate that the portal associates with scaffold proteins to initiate the capsid assembly in herpesviruses (48, 49). Due to the lack of detailed structural information on the portal and scaffold proteins, we are currently unable to use our model to speculate on the integration of the portal and scaffold proteins into the capsid. However, the observed pattern of directionality of the triplexes and structural adaptations of the capsomers provide some insights into the mode of capsid assembly. A structurally rigid unit comprising a triplex together with its covalently linked lasso triangle might possibly act as one of the basic assembly units, which plays strategic roles in capsid assembly. Such a method of assembly of the capsid conforms to the efficient strategy of producing basic assembly units first, which then assemble into higher-order structures that ultimately lead to the formation of highly complex structures like the capsid of the HSV.

Despite the evolutionary and biological divergence of the eight human herpesviruses that have been classified into three subgroups (50), a large number of core gene products are highly conserved among these viruses (51). Moreover, all herpesviruses share a common virion (capsid) structure, a core genome replication process, and similar entry and egress pathways (52, 53). In summary, this work not only marks a substantial improvement in the resolution of the structure of herpesvirus capsid determined by cryo-EM but also is a step toward understanding the drivers of assembly of the capsid. We have identified extensive networks of intersubunit interactions, including intermolecular disulfide bonds, that underpin capsid stability. In line with structural analysis, B as well as C capsids were robust and could withstand heat treatment at 65°C for 10 min (54) (fig. S13). The structure of HSV described here could provide a basis for development of better antiviral drugs for treating infections caused by herpesviruses.

Materials and methods

HSV-2 capsid preparation

Vero cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) plus 10% fetal bovine serum (FBS). Cells were grown to 90% confluence and infected with herpes simplex virus 2 strain MS at a multiplicity of infection (MOI) of 0.1–1. At 4 days post infection, the cells were collected, resuspended in PBS containing 1% NP-40 and lysed by three cycles of freezing and thawing. After lysis, the solution was centrifuged at 1500 g at 4°C for 15 min to remove large debris. A discontinuous 20% and 60% sucrose gradient (w/v in PBS) was used to further purify the capsids. Fractions containing the capsids at the interface of the two sucrose layers were collected after centrifugation. Crude HSV-2 capsids concentrate [~0.6 mg in 600 μl PBS (pH 7.4)] was loaded onto a continuous 20% to 50% (w/v in PBS) sucrose density gradient and centrifuged at 80,000 g for 1 hour. Three sets of fractions were collected and dialyzed against PBS buffer. Capsids from these fractions were imaged by negative staining electron microscopy and cryo-EM.

Analytical ultracentrifugation

Sedimentation velocity experiments were performed on a Beckman XL-I analytical ultracentrifuge at 20°C. HSV-2 capsids were diluted with PBS buffer to 400 μL with A280nm absorption of about 0.8. Samples were loaded into a conventional double-sector quartz cell and mounted in a Beckman four-hole An-60 Ti rotor. Data were collected at 3000 rpm at a wavelength of 280 nm. Interference sedimentation coefficient distributions were calculated from the sedimentation velocity data using the SEDFIT software program (

Negative stain

For examination by negative staining, preparations of capsids were diluted in PBS (pH 7.4) to a suitable concentration, applied to a freshly glow-discharged grid, washed twice with PBS, and stained with 1% phosphotungstic acid (pH 7). All samples were examined on a 120-kV electron microscope.

Cryo-EM and data collection

For cryo-grid preparation, a 3-μL aliquot of purified HSV-2 capsids was applied to a fresh glow-discharged 400-mesh holey carbon-coated copper grid (C-flat, CF-2/1–2C, Protochips). Grids were blotted for 3.5 s in 80% relative humidity for plunge-freezing (Vitrobot; FEI) in liquid ethane. Cryo-EM data sets were collected at 300 kV with a Titan Krios microscope (FEI). Movies (25 frames, each 0.2 s, total dose 25 eÅ−2) were recorded using a Falcon3 detector with a defocus range of 0.8 to 2.3 μm. Automated single-particle data acquisition was performed with SerialEM, with a nominal magnification of 59,000×, which yields a final pixel size of 1.38 Å.

Image processing

A total of 5527 micrographs were recorded. Out of these, 4600 micrographs with visible CTF rings beyond 1/5 Å in their spectra were selected for further processing (fig. S1). The defocus value for each micrograph was determined using Gctf (55). More than 55,000 particles from the 4600 micrographs were boxed using EMAN package (56). The RELION (57) and jalign (58) were used to process the HSV capsid data, which led to a reconstruction of ~4-Å resolution with icosahedral symmetry imposed. There were two obstacles that limited the resolution of this 1250-Å-diameter virus. One was the structural complexity resulting from the fact that the capsid was not strictly following the icosahedral symmetry (59). The reason for this anomaly could be the intrinsic structural flexibility associated with the particle or some form of structural perturbation (local deformation) caused by ultracentrifugation during the purification process. The other was the Ewald sphere effect, which is equivalent to the consequence of the defocus gradient on this large virus. Conventional programs available for determining cryo-EM structures cannot deal with these two problems. We developed a reconstruction method named block-based reconstruction. In this method, the asymmetric unit of the icosahedral virus was divided into 4 blocks and each block was refined and reconstructed separately. During refinement, in each boxed cryo-EM image, there were 60 icosahedral-symmetry–related copies for each block. The initial orientation and center parameters for the copies could be calculated based on the icosahedral symmetry virus reconstruction by RELION or jalign. A local search for better rotation and translation parameters of each copy was performed to overcome the capsid structure complexity. After 4 blocks being refined and reconstructed, a program was used to combine the 4 blocks into an asymmetric unit. The resolution as determined by Golden standard Fourier shell correlation was 3.6 Å using the 0.143 threshold. To solve the Ewald sphere problem, knowing the rotation and translation parameters of a virus, the distance d between the center of one copy in the 3D virus and the center of the virus along the z axis (parallel to the incident electron beam) was calculated. Assuming the defocus obtained by fitting the Thon ring representing the distance between the center of the virus and focused point of objective lens, the local defocus of each copy in the 3D virus was the sum of d and the defocus value. This local defocus of each copy instead of the uniform defocus obtained by fitting was used to reconstruct the blocks. After combining these blocks, the final resolution of HSV capsid approached 3.1-Å resolution. We used the program of post process in RELION (57) to evaluate the resolution and did the map sharpening (estimate the B-factor automatically).

Model building and refinement

The ab-initial atomic models of the C-Hex, P-Hex, Pen, VP26, and Triplex were built de novo into density using COOT (60). Among these, the crystal structure of the upper domain (61) was manually fitted into the maps of the capsomers using CHIMERA (62) and further manually corrected in COOT (60). Models were further improved by iterative positional and B-factor refinement in real space using Phenix (63), rebuilding in COOT (60), and evaluated by Molprobity (64) and Refmac (65). Refinement statistics are listed in table S1.

Supplementary Materials

Figs. S1 to S13

Table S1

References (6669)

References and Notes

Acknowledgments: We thank X. Huang, B. Zhu, Z. Guo, and J. Lei for cryo-EM data collection; and the Center for Biological Imaging (CBI), Institute of Biophysics, and the Tsinghua University Branch of China National Center for Protein Sciences for EM work. Funding: Work was supported by the National Key Research and Development Program (2017YFC0840300, 2014CB542800, 2014CBA02003, 2016YFA0501100, and 2017YFA0504700), the Strategic Priority Research Program (XDB08000000), National Science Foundation of China (813300237, 31570717, 91530321, 31570742, and 81520108019), and Natural Science Foundation of Hunan Province (2017RS3033). X.W. was supported by Young Elite scientist sponsorship by CAST and program C of “One Hundred Talented People” of the Chinese Academy of Sciences. X.Z. received scholarships from the “National Thousand (Young) Talents Program” from the Office of Global Experts Recruitment in China. Author contributions: S.Y., J.-L.W., D.Z., N.W., Q.G., and X.W. performed experiments; X.W., H.L., X.Z., J.-Z.W., and Z.R. designed the study; all authors analyzed data; and X.W., H.L., X.Z., J.-Z.W., and Z.R. wrote the manuscript. Competing interests: All authors have no competing interests. Data and materials availability: Cryo-EM density map of HSV-2 B capsid has been deposited in the Electron Microscopy Data Bank under accession code EMD-6907, and the atomic coordinates of the asymmetric unit have been deposited in the Protein Data Bank under accession code 5ZAP.

Stay Connected to Science

Navigate This Article