Crystal Structure of Human Adenovirus at 3.5 Å Resolution

See allHide authors and affiliations

Science  27 Aug 2010:
Vol. 329, Issue 5995, pp. 1071-1075
DOI: 10.1126/science.1187292


Rational development of adenovirus vectors for therapeutic gene transfer is hampered by the lack of accurate structural information. Here, we report the x-ray structure at 3.5 angstrom resolution of the 150-megadalton adenovirus capsid containing nearly 1 million amino acids. We describe interactions between the major capsid protein (hexon) and several accessory molecules that stabilize the capsid. The virus structure also reveals an altered association between the penton base and the trimeric fiber protein, perhaps reflecting an early event in cell entry. The high-resolution structure provides a substantial advance toward understanding the assembly and cell entry mechanisms of a large double-stranded DNA virus and provides new opportunities for improving adenovirus-mediated gene transfer.

Human adenoviruses (HAdV) are nonenveloped double-stranded DNA (dsDNA) viruses that are associated with acute infections (13). Although these infections are generally self-limiting, the reemergence of certain HAdV types has also been linked to potentially fatal respiratory infections in both civilian and military populations (4). Severe disseminated diseases also occur in patients receiving bone marrow–derived stem cells (5, 6). In addition to their disease associations, replication-defective or conditionally replicating HAdVs continue to be evaluated in ~25% of approved phase I to III clinical trials for vaccine and therapeutic gene transfer (7, 8). However, the lack of accurate details of the virus structure limits the reengineering of HAdV vectors and prevents a better understanding of the virus life cycle. High-resolution HAdV structure determination presents a challenge because of the large size (910 Å average diameter, 150 megadalton) and complexity (pseudo-T = 25) of the virus. The crystal structures of the major HAdV capsid proteins, the fiber (9), penton base (PB) (10), and hexon (11), have been solved. The hexon and penton base crystal structures were subsequently used to derive pseudo-atomic models of the HAdV capsid at moderately high resolution (7 to 10 Å) (1214) by cryoelectron microscopy (cryoEM). CryoEM structural analyses provided considerable insight into HAdV organization; however, they did not furnish detailed information on the interactions between the major and accessory (cement) proteins (IIIa,VI,VIII, and IX).

We report here the crystal structure of a recombinant HAdV-5 vector, designated Ad35F, that is equipped with a short and flexible fiber protein derived from HAdV-35 (15). Details of the crystallization (16), diffraction data statistics (table S1), and structure determination of Ad35F at near-atomic resolution (3.5 Å) are described in (17).

The architecture of the HAdV capsid is shown in Fig. 1, A and B. The hexon is the most abundant protein in the capsid, with 720 subunits arranged as 240 trimers on a pseudo-T = 25 icosahedral lattice. Five PB monomer subunits occupy each of the icosahedral vertices and are associated with the trimeric fiber protein. Each of the 20 facets of the capsid contains 12 hexon trimers and a penton at each vertex. The icosahedral asymmetric unit consists of four hexon trimers and one PB monomer. As previously described, each hexon monomer contains two eight-stranded jelly-roll domains, V1 and V2 (11), whereas the PB subunit contains a single jelly-roll domain (10). Three sets of V1 and V2 domains give the hexon trimer a pseudo-hexagonal shape at the base, which in turn gives rise to a pseudo-T = 25 architecture for the HAdV capsid. Large insertions between the strands of the hexon jelly-roll domains form triangular towers on top of the hexagonal base. Representative electron density for a hexon subunit [amino acids (aa) 579 to 582] is shown in fig. S1. Some of the hypervariable region loops (aa 186 to 193 and 250 to 258) that were disordered in the isolated hexon structure are visible in the HAdV capsid crystal structure (fig. S1) because they are involved in multiple, symmetry-related interhexon contacts (figs. S2 and S3). The tertiary structures of the 12 structurally independent hexon subunits are nearly identical, having a root mean square deviation of ~1 Å upon superposition with a few differences found mainly at the amino and carboxy termini.

Fig. 1

Surface rendering and subunit associations of Ad35F. The icosahedral asymmetric unit is composed of one of the PB (pink) subunits and four hexon trimers [cyan (1), orange-red (2), green (3), and yellow (4)]. Cement proteins on the outside of the capsid are shown in blue (four-helix bundle) and magenta (IX triskelion). Twelve hexon trimers occupy one facet of the Ad35F capsid; however, a total of 18 hexons are shown for clarity. A model of the trimeric fiber docked into each penton base is shown in orange. (A) Outer capsid viewed down the particle threefold axis with a facet circumscribed by the white triangle. (B) A zoom-in view of the single facet.

An unusual molecular interaction in the HAdV particle occurs between the fivefold symmetric PB and the threefold symmetric fiber protein. The “symmetry mismatch” that occurs between the PB and fiber is of particular interest because of its potential impact on cell receptor interactions as well as subsequent disassembly processes. Previous cryoEM structural analyses of HAdV at moderate resolution and a cocrystal structure of PB with an N-terminal fiber peptide suggested that the fiber protein interacts on the outer surface of the PB (18, 19). Consistent with this model, the crystal structure of recombinant PB alone (Fig. 2) indicated that the central pore of the PB pentamer is too narrow to allow fiber insertion (10). However, in the Ad35F particle the central pore of the PB has a diameter nearly twice that of the isolated (fiberless) PB (10) (Fig. 2B). Hence, the dimensions for the PB pore in the Ad35F structure but not in the isolated PB would allow insertion of the fiber shaft domain into the central cavity of the PB. Consistent with this, Fig. 2, C and D, shows strong FoFc electron density along the icosahedral fivefold symmetry axes, generated by using local threefold symmetry, originating from deep inside the PB pore and extending to its outer surface. We assign this to the shaft region of the fiber protein. This fiber shaft corresponds to a length of 90 Å and accommodates five to six β-spiral repeats that are each about 13 to 15 Å along the fiber axis (Fig. 2, C and D) in agreement with the 5.5 repeats predicted for the HAdV-35 fiber shaft (20). The density corresponding to the fiber knob was not visible. The ability of the penton base to adapt to large changes is striking. Conformational changes observed in the crystal structure may reflect early events in cell entry.

Fig. 2

Structure of the penton complex. (A) The structure of the isolated PB shown in gold (10) with a maximum pore diameter of 28 Å. (B) The structure of the PB in the Ad35F particle shown in magenta with a maximum pore dimension of 50 Å. (C) The threefold averaged difference (FoFc) electron density (in blue, contoured at 2.5σ) corresponding to the fiber molecule along the fivefold axis of the PB, shown as magenta trace. Structure of the HAdV-35 fiber molecule based on the partial HAdV-2 fiber structure (9), represented as ribbon diagram (red) and fitted into the difference density. (D) A vertical view of the fit of the fiber shaft to the difference density (left). The flexible third repeat of the shaft is indicated by an arrow. A view down the fiber axis showing the fit of one of the shaft repeats (right).

The C termini of hexon subunits are involved in two different types of interactions. First, the C termini stabilize the threefold junctions inside the capsid within each group of nine hexons (GONs) in a facet. The C-terminal tail (green) (residue 944 to the last ordered residue—946, 949, or 951 in different subunits) forms an extra strand that interacts with a β sheet in the V2 domain of the adjacent (clockwise; threefold related) hexon subunits in neighboring trimers (Fig. 3, A and B).

Fig. 3

Hexon associations at the inner capsid surface. (A) Inner surface of a facet showing the locations of hexon C termini, shown as spheres according to the hexon color coding used in Fig. 1A. GON hexons in the facet are shown in tan, and the rest of the hexons are shown in gray for clarity. VIII molecules are shown in black. PB subunits are shown in pink and additional cement proteins in purple. Structurally distinct locations (A and B) of VIII molecules as well as the four hexon trimers (1, 2, 3, and 4) of the icosahedral asymmetric unit are indicated. (B) Formation of extended β sheets by the hexon C termini (green) with the EF strands (aa 721 to 730; purple) of the V2 domain of the neighboring (clockwise) hexon subunit at the icosahedral threefold axis (triangle). Hexons related by icosahedral symmetry are indicated with the prime and double prime symbols. (C) Stacking of the hexon C termini (red and green) at the inter-GON interface with a helix from VIII (black). Hexon N termini that interact with VIII are shown in powder blue.

The second type of interaction occurs between two C-terminal tails of hexon subunits, with the two tails coming from adjacent GONs or with one of the tails coming from a peripentonal hexon. The two tails stack on top of each other. Although there are no direct interactions between the C-terminal tails themselves, each interacts with residues in apposing hexon subunits primarily via hydrogen bonds. A highly ordered helix from protein VIII (residues 40 to 60) lies beneath and coplanar with these stacked C termini, further supporting the interface between hexon subunits at the capsid interior (Fig. 3, A and C). In contrast to the C termini, none of the N termini of the hexons are involved in “direct” interhexon interactions. They are, however, involved in critical interactions with the cement proteins, thus indirectly stabilizing the interhexon associations (detailed below).

The contiguous shell of the Ad35F capsid contains four cement proteins, IIIa, VI, VIII, and IX, that play important roles in stabilizing the virion. Their locations have been tentatively assigned to the outer and inner surface of the capsid on the basis of their copy number, cryoEM, and biochemical analysis (12, 14, 21). We did not observe significant side-chain densities needed to definitively assign their sequence with the exception of VIII (see below). The triskelion-like structures, formed by N-terminal regions of three protein IX subunits (12, 13), are located at four different positions per facet (Fig. 1) and stabilize the threefold GON junctions on the capsid exterior. The Ad35F crystal structure revealed a very-well-ordered four-helix bundle, 80 Å in length, that is located on the capsid exterior between hexon-4 of one GON and hexon-2 of the neighboring GON (molecule in blue; Fig. 1, A and B). Early cryoEM analyses (12, 22) tentatively assigned density at this location to IIIa, which is the largest cement protein (63 kD) in the HAdV capsid shell. However, a recent 6 Å resolution cryoEM analysis revealed the coiled coil shape of this density and led to the alternate hypothesis that this region might correspond to the C termini of four IX molecules (14). This assignment to protein IX was also suggested by cryoEM analysis of Ad vectors with peptide-tagged IX (23). Interestingly, the x-ray structure revealed that two of the helices in this four-helix bundle appear to be connected at one end (fig. S4a inset). This finding, as well as the different distances observed from the visible ends of the (trimeric) triskelions of protein IX to the nearest helix of the (tetrameric) four-helix bundle, raises the question as to whether the helical bundle is formed by four C-terminal helices of IX. Another possibility is that this helical bundle represents a domain of IIIa as originally proposed. A higher-resolution density map will be required to definitively assign an amino acid sequence to this region.

The Ad35F structure also revealed multiple helices, organized primarily as two domains (fig. S4b), on the inside of the virus capsid at the vertex region. The connectivity between these helices is not clear. Although this region has been tentatively assigned as IIIa on the basis of a cryoEM analysis (14), it is possible that at least some of these helices could correspond to the membrane lytic VI molecule (24) that participates in endosomalysis and whose N-terminal 80 residues have a high degree of α-helical content. The helical domain, closest to the vertex, “cements” the contacts between two peripentonal hexons, whereas the second domain appears to interact with the base of one peripentonal hexon. There is also a helix layered on top of one of the molecules of VIII (A) (fig. S4b). This helix, along with two other short helices, could belong to yet another cement protein. Interestingly, difference maps also showed that all of the hexon cavities are filled with electron density (fig. S5), suggesting that at least part of the putative protein VI molecules could be sequestered inside the hexon trimers, as previously suggested (25). These densities could account for over 200 VI molecules present in the HAdV capsid.

The HAdV particle contains 120 copies of protein VIII, a key cement protein that has been tentatively assigned to the interior of the capsid on the basis of its copy number and predicted α-helical content (12, 14). Our x-ray structure confirmed the location of VIII and provided more detailed information on its interactions with nearby hexon proteins (Fig. 4). We fit the VIII sequence into the electron density on the basis of multiple criteria, including the presence of a single predicted α helix, the location of key proline residues associated with a prominent U-shaped bend in the central portion of the VIII segment, and the direction of the densities for the residues with long side chains (facing toward the N terminus of the helix) in the sharpened maps. The ordered part of VIII molecule has an extended fold with an end-to-end span of 83 Å, starting with a helix of ~20 residues, followed by a distinct inverted U-shaped structure, and succeeded by an extended polypeptide chain (Fig. 4A). Furthermore, a large portion of the aligned sequence (residues 37 to 90) exhibited favorable side chain interactions with the residues from the nearby hexon subunits (Fig. 4, C and D). Fifty-four residues of the mature form of VIII, resulting from cleavage by the virus cysteine protease, are visible at two independent locations in the electron density maps. The two structurally distinct VIII molecules (A and B) closely trace the base of individual hexons (inside) along the outer edge of the GON hexons and mediate the interactions between GONs as well as between peripentonal hexons (Fig. 4B). Each VIII molecule in the Ad35F particle interacts with the N termini of three subunits of hexon trimers related by local threefold symmetry. One of the protein VIII molecules, designated A, is located close to vertex region and interacts with two peripentonal hexons, whereas the second one, designated B, closely interacts with hexon-4 and hexon-3′ (Figs. 3A and 4B). The ordered parts of each of these molecules closely associate with three hexon trimers.

Fig. 4

Structure and interactions of pVIII with the hexons. (A) Display of 2FoFc electron density (contoured at 1.0σ) of the ordered part of an VIII molecule. (B) An interior view of the facet showing the locations of proteins VIII (black) and additional cement proteins (purple). An outline of the GONs is shown by a thick blue line. A line trace (in white) corresponding to the location of three P30 molecules from PRD1 (9) is shown in comparison with protein VIII and in the context of the Ad35F structure. Hexons are color coded and numbered according to the scheme described in Fig. 1A and (A), respectively. (C) Nonpolar interactions between an VIII helix (black) with the N termini (powder blue) of two neighboring hexons viewed from the inside of the virus. I indicates Ile; P, Pro; R, Arg; and M, Met. (D) Close approach of the inverted U-shaped bend in VIII (black) to the N terminus of a hexon subunit. However, the interactions of the U-shaped structure are not as close and specific as those mediated by the helix in (C).

Specifically, residues Ile41 and Ile54 in the N-terminal leg of the VIII molecule (A) mediate nonpolar interactions with Pro7 and flanking residues at the N terminus of a peripentonal hexon (hexon-1′) (Fig. 4B) and hexon-4 in the reference icosahedral asymmetric unit, respectively (Fig. 4, B and C). The inverted U-shaped kink comprising VIII residues 63 to 77 interacts with N-terminal residues 1 to 7 of one of the subunits of another peripentonal hexon (hexon-1) (Fig. 4D). Thus, VIII molecule-A staples two peripentonal hexons with hexon-4. Likewise, the second VIII molecule-B mediates associations between three hexons, hexon-2′ of a neighboring GON with hexon-3′ and hexon-4 from the reference GON, by using the quasi-equivalent interactions as described for VIII molecule-A (Fig. 4, B to D).

The Ad35F structure provides insights into the virus-host interactions. Although interhexon contacts are extensive and are augmented by the cement proteins, the interactions between the penton base and peripentonal hexons are rather tenuous (fig. S6). This would favor efficient capsid disassembly and release of the penton complex during cell entry. The crystal structure of the entire HAdV particle also provides an initial assessment of the folds and interactions of cement proteins with the hexon subunits. Improved knowledge of HAdV assembly from its individual capsid proteins could lead to the development of novel antiviral strategies to block infection at multiple cell entry steps and facilitate the development of HAdV vectors with improved tissue targeting.

Supporting Online Material

Materials and Methods

Figs. S1 to S6

Table S1


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. The authors express their deep gratitude to J. E. Johnson for advice and guidance during all phases of the structure determination and for reading the manuscript; T.-M. Mullen and L. Gritton for the outstanding technical support for vector production; S. Venkataraman for help with the early crystallization setups of HAdV; I. Mathews of the Stanford Synchrotron Radiation Lightsource for initial help with the data collection; W. Minor and I. Minor for generously providing the larger version of the SCALEPACK program; E. Ollmann-Saphire and J. G. Smith for comments on the manuscript; T. McCarthy for preparation of this paper; and R. Fischetti, N. Sanishvili, and other members at General Medicine and Cancer Institutes Collaborative Access Team (GM/CA CAT) for discussions and technical support. GM/CA CAT has been funded in whole or in part with federal funds from the National Cancer Institute (Y1-CO-1020) and the National Institute of General Medical Science (Y1-GM-1104). The use of the Advance Photon Source was supported by the U.S. Department of Energy, Basic Energy Sciences, Office of Science (contract no. DE-AC02-06CH11357). The coordinates of Ad35F crystal structure have been deposited in the Protein Data Bank (PDB) with the identification code 1VSZ. This work was supported by NIH grants R01 AI070771 to V.S.R., HL054352 and EY011431 to G.R.N., and AI042929 to P.L.S. This is manuscript no. 20538 from the Scripps Research Institute. G.R.N. holds a patent titled “Adenovirus vectors, packaging cell lines, compositions, and methods for preparation and use” (patent no. 7,232,899).
View Abstract

Stay Connected to Science

Navigate This Article