Cryoelectron Microscopy of Protein IX-Modified Adenoviruses Suggests a New Position for the C Terminus of Protein IX

Abstract

Recombinant human adenovirus is a useful gene delivery vector for clinical gene therapy. Minor capsid protein IX of adenovirus has been of recent interest since multiple studies have shown that modifications can be made to its C terminus to alter viral tropism or add molecular tags and/or reporter proteins. We examined the structure of an engineered adenovirus displaying the enhanced green fluorescent protein (EGFP) fused to the C terminus of protein IX. Cryoelectron microscopy and reconstruction localized the C-terminal EGFP fusion between the H2 hexon and the H4 hexon, positioned between adjacent facets, directly above the density previously assigned as protein IIIa. The original assignment of IIIa was based largely on indirect evidence, and the data presented herein support the reassignment of the IIIa density as protein IX.

The adenoviral capsid is a complex assembly of seven polypeptides, organized into an ∼900-Å-diameter icosahedral shell. Twelve trimers of hexon, the major capsid component, are arranged onto each of 20 interlocking triangular facets, with penton capsomeres and their protruding fibers occupying each of the 12 vertex positions (24). The icosahedral lattice of hexons is stabilized by a number of so-called “minor” capsid proteins, including IIIa, VI, VIII, and IX. With the exception of protein IX, these minor components provide vital functions and are absolutely necessary for the adenovirus (Ad) life cycle (35). The presence of protein IX is not required for the encapsidation of Ad genomes of subgenomic size (<36 kb), but virions with capsids lacking IX (ΔIX) are more thermolabile than wild-type virions (5, 6). Previous work (13) suggests that deletion of protein IX prevents the encapsidation of genomes of wild-type or larger sizes (36 to 38 kb). A more recent study indicates that the absence of IX does not prevent encapsidation but rather that the resulting ΔIX virions are defective in infectivity assays (27).

Central to each facet are the groups-of-nine hexons (GONs) that are observed after various methods of capsid disruption (6, 10). The crystal structure of the hexon trimer (25) reveals a pseudohexagonal base with three towers extending upwards, creating a triangular top approximately 113 Å tall. The maximum radius of the hexon trimer is 50 Å. The triangular top is rotated roughly 10 degrees counterclockwise with respect to the hexagonal base. The hexon trimers present within each facet are closely packed along their hexagonal bases, creating a continuous protein shell that is approximately 33 to 44 Å thick. The three towers of each hexon trimer extend roughly 69 Å above the capsid floor, thereby creating cavities between the towers of different hexon trimers. Due to the different relative angles between the bases and towers of individual hexon trimers, each GON contains four large and three small cavities between the nine different hexon trimers.

Structurally, protein IX is thought to function as capsid cement, and trimers of IX are thought to reside at the bottom of the four large cavities of each GON, well below the outermost hexon surface (Fig. 1). This arrangement of IX within the icosahedron results in 12 molecules per facet and 240 molecules per virion, in excellent agreement with the known stoichiometry of IX within the capsid, as calculated from [³⁵S]methionine labeling experiments (32). Protein IX was inferred to reside at this position by difference imaging between a scanning transmission electron microscopy reconstruction of purified GONs and a virtual GON, constructed by placing nine copies of a contour plot of the hexon trimer (derived from X-ray crystal structure coordinates) in the spatial orientation of the native GON. This difference analysis revealed residual mass density arranged as four elongated trimers present within each of the four large cavities of the GON (12). This density was assigned as protein IX based on several indirect pieces of evidence. First, virions lacking IX were shown to be thermolabile and ΔIX virions fail to dissociate into GONs as wild-type virions do (6). Second, IX has been shown to be associated with purified GONs (1, 32). Third, this spatial arrangement agrees with the known stoichiometry of IX and its apparent ability to form multimers (23, 34). Lastly, the mass calculated from the difference imaging of the IX trimers was in close agreement with the actual molecular mass of protein IX: 16.6 to 18.9 kDa by scanning transmission electron microscopy versus 14.3 kDa predicted by sequence analysis (12).

FIG. 1.

Current model of the adenovirus capsid. (A) Isosurface rendering of the 10-Å wild-type capsid (EBI accession number EMD-1113) described by Fabry et al. viewed along the icosahedral threefold axis (11). (B) Enlarged view of a single facet with the fivefold, threefold, and twofold axes marked by black pentagons, triangle, and ellipses, respectively. IIIa density is outlined in red. (C) Schematic of the facet, illustrating the organization of the hexons and minor components. Pentons are represented with black pentagons. Hexon trimers are represented by hexagons; those that belong to the group of nine are colored gray, and those that do not are colored white. The three groups of four unique hexons within each facet are designated by the numeric labels H1, H2, H3, and H4 in yellow, green, and magenta numbering. H1 designates peripentonal hexons, H2 hexons lie along the twofold axis, H3 hexons surround the threefold axis, and H4 hexons are at the fourth nonequivalent position. Hexons labeled by H1′ and H2′ are from the neighboring facets. Density assigned as protein IIIa (red circles) lies between the H4 hexons and the H2′ hexons of the adjacent facets. Trimers of protein IX (blue) are thought to be present within the GON. The revised model proposed by Saban et al. (26) places the N-terminal portion of IX at the light-blue positions, while the C-terminal α-helical domains of 3/12 of the monomers (pink) lie between the nearby peripentonal H1 trimers. In this model, the remaining 9/12 of the monomers of IX have C termini that radiate outwards from the capsid and are disordered and therefore are not visible by cryoEM.

Three separate structural determinations of the Ad capsid were recently reported, each providing insights into and interpretations of protein IX and its location within the virion. Fabry et al. describe the capsid structure of both wild-type adenovirus type 5 (Ad5) and a ΔIX mutant at 10-Å (Fig. 1) and 15-Å resolution, respectively (11). Surprisingly, difference imaging of these two structures reveals that deletion of protein IX results in the disappearance of the densities previously assigned to both the IX and the IIIa proteins. The authors speculate that the absence of IX inherently destabilizes the capsid and results in the release of the IIIa protein upon freeze-thawing of the specimen. Contrary to this interpretation, protein gels of the virions did not demonstrate complete loss of protein IIIa but rather a slight decrease in its band intensity. This reduced intensity of the IIIa band is not consistent with the complete disappearance of its presumed density in the cryoelectron microscopy (cryoEM) reconstruction. Similarly, a separate report by Scheres et al. analyzing both wild-type and ΔIX capsids found that deletion of IX results primarily in the loss of IIIa density, with minimal density loss at the positions assigned to protein IX, within the GON cavities (28). In their discussion, Scheres et al. even raise the possibility that the weak density currently attributed to IX may be an artifact.

A third report by Saban et al. reconstructs a chimeric Ad5/35 chimeric capsid to a resolution of 9 Å and reports visualization of both the N and C termini of protein IX (26). In their model (adapted in Fig. 1), protein IX can adopt different conformations, with 9/12 of the monomers within each facet having disordered C termini radiating outwards from the capsid surface while the remaining 3/12 of the monomers have C termini positioned between the peripentonal H1 hexon trimers near the penton base. The authors of that report do not mention any connections between IX and IIIa.

Recent work has described the successful display of peptides and proteins fused to the C terminus of protein IX, with negligible effects on virion structure, stability, and virion infectivity. These studies include fusions ranging from small epitope tags such as short RGD and poly-lysine peptides to larger domains such as the 71-amino-acid biotin acceptor peptide (BAP) and the 240-residue enhanced green fluorescent protein (EGFP) (4, 8, 16, 20, 33).

To directly visualize how peptide and protein ligands might be displayed on the C terminus of IX, cryoEM imaging and reconstruction of the previously described Ad-IX-BAP (4) were performed. Comparison of Ad-IX-BAP and wild-type Ad5 revealed additional mass density above the capsomeres, positioned between the H2 and H4 trimers along the outer edges of the GON (data not shown). Unfortunately, statistical analysis was unable to assign significance to this mass difference, perhaps due to incomplete IX-BAP occupancy, the small mass of BAP, and/or the engineered flexibility designed in the fusion between the IX and BAP proteins.

To achieve a more significant mass difference, the larger 27-kDa EGFP protein was fused to the C terminus of IX, linked via a 45-Å α-helical spacer (33), to help alleviate any steric hindrance which might be imposed by the EGFP domain during capsid assembly. This vector (Ad-IX-45-EGFP) was produced by previously described methods (3) and displays EGFP on the capsid, with negligible effects on virion thermostability and infectivity (Fig. 2), in good agreement with similar IX-GFP vectors described recently (16, 20). Provided that the 45-Å spacer is not flexible enough to render the EGFP domain too disordered to appear in the reconstruction, the larger size of EGFP than that of BAP (27 kDa versus 7.4 kDa) was expected to result in a much stronger signal.

FIG. 2.

Characterization of the Ad-IX-45-EGFP specimen. (A) CsCl density gradient of Ad-IX-45-EGFP virions. Clarified 293A cell lysate was centrifuged through a discontinuous CsCl gradient and visualized under UV light. Green fluorescence was observed with Ad-IX-45-EGFP but not control Ad-IX-BAP, demonstrating proper encapsidation of the IX-45-EGFP fusion. The bottom band, containing mature virions, was collected and used for cryoEM. (B) Infectivity and thermostability of Ad-IX-45-EGFP. CsCl-purified vectors expressing the dsRed2 fluorescent protein were diluted in Dulbecco modified Eagle medium and heated to 48°C for the indicated amounts of time before HeLa cell transduction was measured by flow cytometry at 48 h postinfection. AdΔIX, previously described in reference 4, lacks protein IX and was used as a control. Infections were performed at a multiplicity of infection of 1,000 viral particles/cell. wt, wild type.

To test this, an Ad-IX-45-EGFP sample was purified from 293 cell lysate by CsCl banding, concentrated to 3 × 10¹² viral particles/ml with an Amicon Ultra-4 centrifugal concentrator, and resuspended in phosphate-buffered saline (PBS) for specimen preparation. The sample was kept in PBS at 4°C for 8 h prior to grid preparation and freezing. CryoEM specimens were prepared by applying 2 to 3 μl of purified vector directly to Quantifoil copper grids with holey carbon film, followed by blotting with filter paper and rapid plunging into liquid ethane. The frozen hydrated specimen grid was kept at −170°C in a Gatan cryoholder during the microscope examination. Micrographs were acquired with a defocus ranging from 0.5 to 2.0 μm under focus and at a nominal microscope magnification of ×50,000 directly on a Gatan Ultrascan 4000 charge-coupled-device camera connected to a JEOL 2010F transmission electron microscope. The Semi-Automated Virus Reconstruction software suite (14) was used to process approximately 800 particles and reconstruct a final map to a resolution of 22 Å, as estimated by the 0.5 Fourier shell correlation criterion (data not shown). The reconstruction was then low-pass filtered to 22 Å for subsequent visualization and analysis.

Surprisingly, the reconstruction of the Ad-IX-45-EGFP specimen (Fig. 3) shows lower density for the pentons than for the hexons, suggesting that a large fraction of pentons had dissociated from the capsid. Previous work has described the poor stability of Ad vectors in PBS, with viral titer decreasing by 10% every 15 min at 4°C (7). Ad-IX-45-EGFP samples stored in the more stable A195 buffer (9) display minimal loss of infectivity and thermostability compared to unmodified Ad vectors (Fig. 2), suggesting that fusion of EGFP to protein IX does not inherently destabilize the pentons. Unfortunately, the high level of sucrose and other solutes in the A195 buffer confounded attempts to prepare a high-contrast specimen for cryoEM.

FIG. 3.

Icosahedral reconstruction of Ad-IX-45-EGFP viewed along the threefold axis. The EGFP domain was segmented manually using Amira (Mercury Computer Systems, Chelmsford, MA). Surface renderings of reconstructions were carried out with UCSF Chimera (San Francisco, CA). (A) Orthoslice of reconstruction showing reduced intensity for penton (black arrow) and EGFP (green arrow). (B to D) Radially colored isosurface renderings of reconstruction at successively lower contours. (B) A conservative high contour shows a void at the penton position. (C) At a lower contour, the EGFP density appears, and the first sign of penton is visible. (D) At this contour, a larger volume appears for EGFP. (E) When the contour is lowered significantly, the noise of the reconstruction is clearly visible.

Further comparison of Ad-IX-45-EGFP to the wild-type capsid reveals two more prominent differences. First, the density previously attributed to IIIa is absent in Ad-IX-45-EGFP, a phenomenon that was also observed with the Ad-IX-BAP reconstruction. Second, there is a presence of density attributable to the extra mass of the C-terminal EGFP fusion (Fig. 4). Quite surprisingly, these regions of density were present at an unexpected location of the icosahedral capsid. Rather than being within the vicinity of the large cavities of the GON, the extra densities are present at locations on opposite and adjacent sides of the twofold axis: above the capsomeres, positioned between the H2 and H4 trimers, just along the outer edges of the GON (Fig. 3). Interestingly, this extra density is located directly above the position attributed to protein IIIa, which is positioned between the borders of different facets, perhaps functioning to rivet the facets together (Fig. 4). The putative EGFP density reconstructs at a contour intensity similar to that of penton, suggesting either incomplete occupancy or disorder due to flexibility of the IX-45-EGFP fusion. The EGFP volume is visible in isosurface rendering when the contouring threshold is lowered beneath the threshold necessary for normal capsid rendering. When the threshold is successively lowered, both penton and EGFP density become visible (Fig. 3).

FIG. 4.

Isosurface rendering of Ad-IX-45-EGFP after EGFP was manually segmented and rendered at lower contouring thresholds. (A) Entire icosahedral capsid showing one EGFP per asymmetric unit. (B) A close-up of the features highlighted by the yellow box in panel A. The density previously attributed as IIIa in wild-type capsids has been added in red. (C) Another perspective of the close-up in panel B visualized from the perspective of H1′ looking towards H3. The crystal structure of the hexon has been docked into hexons H2′ and H4.

Our data show the appearance of extra density approximately 50 Å above the position of IIIa upon fusion of EGFP to the C terminus of protein IX (Fig. 4), consistent with the presence of the 45-Å spacer at the same position where added BAP density was observed in the Ad-IX-BAP reconstruction (data not shown). These data support several possible models for the location of IX within the virion. One possibility follows the rationale of Saban et al. (26), in that the N termini of protein IX may be present within the GON while some of the C termini may be positioned within the space occupied by IIIa. In this scenario, the fusion of domains to the C terminus of IX may result in the destabilization or ejection of IIIa from the capsid, thereby explaining the loss of IIIa density and the appearance of new density above its previous position. Although this interpretation cannot be excluded, it seems improbable since the current published models have protein IIIa spanning the capsid and residing at positions both above and below the hexon shell (24, 31), making ejection of this density unlikely.

A much more plausible scenario is that the density currently believed to be IIIa is in fact attributable to protein IX or at least the C-terminal portion of protein IX. Multiple-structure prediction algorithms (15, 19, 36) suggest a high probability for α-helical structures with heptad repeats characteristic of coiled coils (23) towards the C terminus of protein IX (residues 94 to 136). Fusion of the BAP or EGFP to these helical structures could potentially disrupt the folding of such helical coiled-coil domains, thereby causing their disappearance in the cryoEM reconstruction. The reassignment of IX to the IIIa density is also strongly supported by the recent cryoEM data observed with ΔIX virions exhibiting a clear loss of IIIa density in the absence of IX (11, 28).

In order to agree with the known stoichiometry of IX, this revised model of the Ad facet would place four molecules of protein IX at each of the current IIIa locales along the edges of the GONs between hexons H4 and H2 of neighboring facets (Fig. 1). Placement of IX at this position could conceivably agree with its role as a cementing protein, but the model does not explain why virions that contain IX dissociate into GON structures while those that lack IX do not (6). Recent work suggests that IIIa has little surface accessibility because antiserum raised against IIIa reacts only with broken viral particles (28), and an older study on Ar⁺ plasma etching of Ad2 virions concludes that IIIa is located somewhere between the core and the outer capsid shell, similar to protein VI (21). Taken together, these data suggest that IIIa may not be exposed on the virion surface and that protein IX may be responsible for the density previously assigned to protein IIIa.

Review of the literature reveals that the IIIa density was originally misassigned to dimers of the small 14.5-kDa protein VIII (30), another hexon-associated protein that is thought to have a stoichiometry of 120 molecules per virion (32). Shortly thereafter, this density was reassigned to monomers of the 63-kDa protein IIIa (31), which are present at 60 copies per virion (32). The rationale for the original reassignment from VIII to IIIa was never fully explained, but the current assignment is based on comparisons between the measured and predicted volumes of the cryoEM mass density and the molecular mass of IIIa. The measured volume of the density in question is 238 voxels, very close to the predicted volume of protein IIIa, which is 245 voxels, as based on molecular weight (32). Interestingly, the molecular mass of four molecules of IX (4 × 14.3 kDa = 57.2 kDa) is close to that of a monomer of IIIa (63 kDa). Therefore, the volume argument for IIIa assignment also supports the assignment of four protein IX molecules to the same location. Although protein IX has been described as a trimer, the only direct evidence of multimerization has been coimmunoprecipitation data with IX-epitope fusions (23, 34), as the protein consistently runs as a monomer on nondenaturing polyacrylamide gels (data not shown). Deletion of the coiled-coil multimerization domain has no effect on the encapsidation of IX or its virion stabilization activity (34). It is therefore feasible that IX could be organized as clusters of four monomers within the virion.

While the above rationales have intellectual value, our cryoEM reconstruction of protein IX-modified capsids represents the first direct visual evidence for the localization of this small protein within the context of the large adenoviral icosahedron. These results contradict all previous reports on the location of IX within the GON and suggest that protein IX has previously been mistaken for protein IIIa. At this new position, IX is more surface exposed and may function to confer capsid stability by bridging hexons between the interlocking facets. This surface localization is consistent with the ability of C-terminally displayed proteins to bind to cell surface receptors or avidin, in contrast to protein IX's previous assignment deep in the valleys between hexon trimers. The fact that ΔIX virions cannot package large genomes into functional particles (27) and that nonfunctional virions with unusually small genomes fail to incorporate IX (29) implies that the interactions of protein IX with other capsomeres during assembly and genome encapsidation are much more complicated than previously thought. The new assignment of IX raises many questions about the molecular architecture of the adenoviral capsid and the organization of the minor components therein. Clearly, this study highlights the need for additional biochemical and structural analyses of the adenoviral icosahedron.

Protein IX has garnered recent attention in the adenoviral vector-targeting community as a scaffold for the attachment of large targeting or reporter molecules to the Ad capsid (22). Although protein IX appears to have great potential for capsid labeling and noninvasive monitoring of Ad particle dynamics (17, 18), recent work in our laboratory has shown that retargeting Ad transduction through protein IX-anchored ligands is surprisingly ineffective (2). A thorough knowledge of the higher-resolution structure of protein IX and its orientation in the Ad capsid may enable rational design of effective cell-targeting adenoviral vectors based on protein IX.

Protein structure accession number.

The reconstruction described here has been deposited under accession number EMD-1272 in the Electron Microscopy Data Bank (EMDB), hosted by the European Bioinformatics Institute (EBI).