Epitope Insertion at the N-Terminal Molecular Switch of the Rabbit Hemorrhagic Disease Virus T=3 Capsid Protein Leads to Larger T=4 Capsids

Abstract

Viruses need only one or a few structural capsid proteins to build an infectious particle. This is possible through the extensive use of symmetry and the conformational polymorphism of the structural proteins. Using virus-like particles (VLP) from rabbit hemorrhagic disease virus (RHDV) as a model, we addressed the basis of calicivirus capsid assembly and their application in vaccine design. The RHDV capsid is based on a T=3 lattice containing 180 identical subunits (VP1). We determined the structure of RHDV VLP to 8.0-Å resolution by three-dimensional cryoelectron microscopy; in addition, we used San Miguel sea lion virus (SMSV) and feline calicivirus (FCV) capsid subunit structures to establish the backbone structure of VP1 by homology modeling and flexible docking analysis. Based on the three-domain VP1 model, several insertion mutants were designed to validate the VP1 pseudoatomic model, and foreign epitopes were placed at the N- or C-terminal end, as well as in an exposed loop on the capsid surface. We selected a set of T and B cell epitopes of various lengths derived from viral and eukaryotic origins. Structural analysis of these chimeric capsids further validates the VP1 model to design new chimeras. Whereas most insertions are well tolerated, VP1 with an FCV capsid protein-neutralizing epitope at the N terminus assembled into mixtures of T=3 and larger T=4 capsids. The calicivirus capsid protein, and perhaps that of many other viruses, thus can encode polymorphism modulators that are not anticipated from the plane sequence, with important implications for understanding virus assembly and evolution.

INTRODUCTION

Virus capsids provide model systems for the analysis of the principles of large-scale protein tertiary and quaternary interactions (21). The tertiary structure of capsid proteins has built-in conformational flexibility, allowing them to maintain slightly different intersubunit contacts while facilitating the acquisition of distinct spatial conformations (8, 17). Conformational flexibility in icosahedral viruses with triangulation numbers of T > 1 is essential for the control of particle size (or the T number). In addition, capsids with large T numbers can be generated by one or more auxiliary viral and/or cell proteins (scaffold, minor capsid, or enzymatic proteins) (14). The intrinsic structural polymorphism of capsid proteins is also important, as capsids are dynamic structures whose components have transient conformations related to specific functions in the viral cycle (4).

Using the calicivirus rabbit hemorrhagic disease virus (RHDV) as a model, we addressed the basis of the calicivirus spherical capsid assembly, which allows identical subunits to adopt different conformations and form closed shells of the correct size. Caliciviruses are nonenveloped icosahedral viruses with a polyadenylated positive-sense (+) single-stranded RNA genome of approximately 7.5 kb. The family Caliciviridae, which includes several human pathogens, is divided into four genera, Norovirus, Sapovirus, Vesivirus, and Lagovirus (20). RHDV, the prototype strain of Lagovirus, is the causative agent of an acute disease in rabbits, and it has economic and ecological importance (2). No cell culture has been found to be able to support authentic RHDV, and much of our understanding of these viruses depends on recombinant RHDV virus-like particles (VLP).

The ∼40-nm-diameter calicivirus capsid comprises 180 copies of a single capsid protein (CP). X-ray structures of the norovirus Norwalk virus (NV) VLP (37) as well as San Miguel sea lion virus (SMSV) (9) and feline calicivirus (FCV) (34) virions (both vesiviruses) indicate that the virion comprises 90 CP dimers arranged with T=3 symmetry to form 12 pentamers and 20 hexamers. Each monomer has three domains, an N-terminal arm (NTA), a shell (S) composed of an eight-stranded β-sandwich, and a flexible protruding domain (P), facilitating virus-host receptor interactions (27, 47) at the outermost surface (6). We previously established the mechanism that allows RHDV VP1 (formerly VP60) to switch among quasiequivalent conformational states (5). The NTA region is involved in the control of inherent VP1 polymorphism; a mutant lacking only 29 amino acids loses the ability to acquire different conformational states, and most assemble into a T=1 capsid.

These structural studies showed that VP1 can accommodate insertions of foreign amino acid sequences at both the N and C termini without disrupting VLP assembly, raising the possibility of using RHDV VLP as foreign epitope carriers for vaccine development (5). In addition, specific sites at the outermost region of VP1 were detected as potential insertion sites for foreign epitopes. The chicken ovalbumin T cell epitope was inserted at the N terminus or in a predicted exposed loop; both chimeric VLP induced specific cell responses mediated by cytotoxic and memory T cells (12).

We report the three-dimensional (3D) structure of RHDV VLP at 8-Å resolution. These studies allowed the modeling of the VP1 backbone structure from X-ray structures of other caliciviruses. We hypothesized that RHDV VLP are an optimal vehicle for the presentation of multiple copies of foreign epitopes, for vaccine design, and for production applications. The VP1 pseudomodel was validated by the insertion of foreign sequences, T and B epitopes, in different sites (N- and C-terminal ends and capsid surface loops). The effect of epitope insertion at these sites was analyzed by 3D cryoelectron microscopy (cryo-EM); whereas most chimeric VLP were similar to wild-type RHDV VLP, the insertion of specific FCV-derived epitopes at the N terminus led to polymorphic assemblies, a mixture of T=3 and T=4 capsids. Our findings show a simple mechanism that increases the capacity for VP1 structural polymorphism.

MATERIALS AND METHODS

Recombinant baculovirus transfer vectors.

Baculoviruses were propagated in Trichoplusia ni cells (H5). Based on the plasmid pHAPhSubG (12), we generated the following vectors corresponding to VP1 constructs with foreign amino acid sequences: plasmid pHAVP60-306OVA (L17 construct) has the sequence GSQLESIINFEKLTEGS, which contains a T cell epitope derived from chicken ovalbumin (40) inserted at the N-terminal end, and plasmids pHAVP60-2NeuX2 (N42 construct) and pHAVP60-305NeuX2 (L42 construct), which have a 42-amino-acid insertion ([GSGNDITTANQYDAADIIRN]₂GS) at the N-terminal end or at the exposed loop of the P2 domain, respectively; this segment bears a B cell epitope derived from FCV (Urbana strain) CP. Plasmid pHAVP60-CBT42 (C42 construct) had a 42-amino-acid insertion with a B and a T cell epitope derived from FMDV (13) at the C-terminal end (GSTASARGDLAHLTTTHARHLPAAIEFFEGMVHDSIKEELRP; the B cell epitope is underlined). Recombinant baculoviruses were produced using the BacPAK baculovirus expression system (Clontech) (5). Wild-type (wt) and chimeric VP1 constructs were expressed, and the self-assembled VLP were purified (12). Samples were analyzed by SDS-PAGE (5).

Cryoelectron microscopy.

VLP samples (5 μl) were applied to Quantifoil grids, blotted, and plunged into liquid ethane by following standard procedures (30). Micrographs were recorded under low-dose conditions (∼10 e⁻/Ų) in a Tecnai G2 electron microscope operating at 200 kV and equipped with a field emission gun at a nominal magnification of 50,000×.

Image processing.

Image-processing operations were performed using Bsoft (24), Xmipp (41), and Spider software packages (42). Graphics were produced by UCSF Chimera (35). A Zeiss TD scanner was used to digitize selected micrographs at a 7-mm step size to yield a 1.4-Å pixel size for the specimen. X3d (11) was used to manually select 5,568, 3,964, 8,010, 12,652, and 3,644 individual T=3 capsid images for wt, N42, L17, L42, and C42 micrographs, respectively. For N42, 1,926 T=4 image particles were selected. Defocus was determined with Bshow (http://lsbr.niams.nih.gov/bsoft/bshow/bshow.html). Due to sample heterogeneity, N42 T=3 and T=4 particles were classified using the Xmipp CL2D routine (43) to select 2,541 and 494 particles, respectively.

Using the published RHDV structure (5) low-pass filtered to 30 Å, the Xmipp iterative projection matching routine was carried out to determine the origin and orientation of each particle. For large N42 particles, in addition to a size-scaled low-resolution RHDV T=3 map, the X-ray structures of several T=4 capsids from Nudaurelia capensis omega virus (1ohf [23]), Sindbis virus (1ld4 [49]), and hepatitis B virus (2G33 [7]) were used as starting models in parallel experiments, with equivalent final results. After each refinement iteration, two independent reconstructions were computed using interpolation in Fourier space, and the resolution was assessed by Fourier shell correlation (FSC) between independent half-data-set maps, applying a correlation limit of 0.5 (or 0.3). After the independent refinements, 5,011, 2,287, 445, 7,209, 11,386, and 3,280 particles were included in the wt, N42 T=3, N42 T=4, L17, L42, and C42 capsid three-dimensional reconstructions (3DR), respectively. At the resolutions achieved, wt, L17, and L42 density maps were identical; their particles were combined and reprocessed as described above to generate a merged reconstruction. To determine the resolution of the merged reconstruction lacking spikes (built only by S and NTA domains), the densities corresponding to the P domain were masked in independent half-data-set maps, and FSC between them was calculated. Amplitude decay was calculated using the spatial frequency components from the cryo-EM maps and the X-ray SMSV capsid map (2gh8). The decay profile of the cryo-EM maps was then adjusted to match the profile of the X-ray maps, and the fitted function was applied to cryo-EM maps in a frequency range from 240 Å to the maximum resolution achieved, and a soft low-pass filter was applied. Amplitude decay was also calculated and corrected with Embfactor (16), with similar results.

Although a T=3 capsid is a nonskewed capsid, the dimers showed intrinsic handedness. The enantiomorph map shown in this study was chosen based on our docking analysis. Dimer boundaries were established by contouring the map at various levels, based on its compactness and contacts with neighboring densities and considering the docked structures of the pseudoatomic maps of VP1 and other calicivirus CP.

For difference map calculations, spherically averaged radial density profiles were calculated, normalized, and scaled to match the fit between the cryo-EM map and quasiatomic model profiles. A difference map was obtained by the subtraction of two 3DR.

Multiple-sequence alignment and VP1 SSE prediction.

The RHDV VP1 sequence (accession no. Q9YND5) was obtained from the UniProt database (http://www.uniprot.org/). The sequences of SMSV and FCV CP, obtained from the Protein Data Bank, were aligned with MAFFT (http://mafft.cbrc.jp/alignment/server) using a pairwise sequence alignment based on a structure alignment with MATRAS (http://mafft.cbrc.jp/alignment/server) as a guide. The multiple-sequence alignment and the secondary structure elements (SSE) were displayed with EPSpript (http://espript.ibcp.fr/ESPript/ESPript/). SSE predictions were generated with the programs PsiPred (http://bioinf.cs.ucl.ac.uk/psipred/), Jnet (http://www.compbio.dundee.ac.uk/www-jpred/), Porter (http://distill.ucd.ie/porter), Sable (http://sable.cchmc.org), Gor (http://npsa-pbil.ibcp.fr/NPSA/npsa_gor4.htm), Yaspin (http://www.ibi.vu.nl/programs/yaspinwww/), and Profsec (http://www.predictprotein.org/). A consensus prediction was obtained by simple majority at each sequence position.

Homology modeling.

VP1 A, B, and C subunits were modeled using the CP structures of SMSV and FCV (3m8l) as the templates. To adjust predicted SSE with those of the crystal structures of the SMSV and FCV CP, MAFFT multiple-sequence alignment was manually adjusted in some cases. Homology models were obtained with Modeler 9v3 on the Bioinformatics Toolkit server (http://toolkit.tuebingen.mpg.de/modeller), and quality was assessed with Verify3D (http://nihserver.mbi.ucla.edu/Verify_3D/).

Structural analysis.

The structural subunit boundaries were established as described previously (29). The SSE in the subunits were identified and modeled using the AIRS programs (3).

Uro fitting (http://mem.ibs.fr/JORGE/index.html) was carried out on the entire map with X-ray structures of SMSV and FCV CP and the VP1 model (33). A, B, and C conformers were initially treated as independent rigid bodies. Correlation coefficients between atomic models and the cryo-EM density map were 89, 90, and 89% for SMSV, FCV, and RHDV VP1, respectively. P and S (with the NTA region) domain coordinates were then refined as independent rigid bodies. Since the VP1 NTA α-helix is slightly displaced from its corresponding cryo-EM density, its position was refined to optimize local correlation using the Chimera fitting tool. Finally, connecting loops were modeled with Chimera, and the geometry idealization of the model was performed with the REFMAC5 program (32). The electrostatic potential for the T=3 capsid model was calculated with Delphi (http://wiki.c2b2.columbia.edu/honiglab_public/index.php/Software:DelPhi) (38) and surfaced and colored with Chimera. Inter- and intrasubunit contacts determined with the CCP4i package (36) and crosschecked with Chimera showed equivalent results. For T=4 capsid molecular modeling, a so-called superpentamer (five A/B and five adjacent C/C dimmers) from the T=3 capsid was fitted in the N42 T=4 capsid using Chimera to optimize the local correlation. The Situs Colacor routine (http://situs.biomachina.org/) gave similar results.

Accession numbers.

The 3DR for wt, L17, L42, N42 T=3, N42 T=4, and C42 capsids and the merged data set (wt, L17, and L42 capsids) were deposited in the Electron Microscopy Data Bank (EMDB; www.abi.ac.uk/) (accession no. EMD-1933, EMD-1934, EMD-1935, EMD-1936, EMD-1937, EMD-1938, and EMD-1939, respectively). The Cα model for the RHDV VP1 protein was deposited in the Protein Data Bank (www.pdb.org) (PDB code 3zue).

RESULTS

Biochemical and structural analysis of VP1 insertion mutants.

VP1 expression led to the efficient assembly of T=3 capsids that were indistinguishable from virions (5), which prompted us to consider using VLP as a delivery system for immunogenic epitopes. We expressed a set of VP1 chimeras (Fig. 1A). The L17 chimeric protein bears a T cell epitope (17 residues) derived from chicken ovalbumin that is inserted at a predicted exposed loop in the P domain; L42 protein is similar to L17 but has a 42-residue insertion with two copies of a B cell epitope from feline calicivirus CP. The N42 protein bears the same heterologous sequence as L42 but at the N-terminal end. Finally, C42 protein has a 42-residue insertion, with a B and a T cell epitope from foot and mouth disease virus (FMDV) (13) at the C-terminal end.

Fig 1.

Biochemical and structural analysis of VP1 insertion mutants. (A) Scheme of wt and chimeric VP1 proteins used, indicating length (right). Inserts are indicated. N42 and L42 have the same inserted sequence. The C42 insert sequence is shown in Fig. 6. Acidic residues, red; basic residues, blue; polar residues, green; hydrophobic residues, black. (B) Coomassie blue-stained SDS-PAGE gels of wt and chimeric VP1 assemblies used for cryo-EM data acquisition. Molecular size markers (MWM; ×10⁻³ Da) are on the left. (C) Cryo-EM of wt VP1 (top), L17 (middle), and N42 (bottom) capsids. Arrows indicate capsids larger than normal T=3 capsids. Scale bar, 100 nm.

Wild-type (wt) and chimeric VP1 assemblies were purified by ultracentrifugation on a CsCl gradient. Single bands in SDS-PAGE analysis confirmed correct expression (Fig. 1B). Cryo-EM of VLP-enriched fractions showed that L17, L42, and C42 capsids are similar to wt capsids in size and morphology. N42 capsids were a mixture of assemblies ranging from ∼40 to ∼50 nm in diameter (Fig. 1C).

The final resolution of the reconstruction for wt, L17, and L42 VLP was estimated to be ∼9.0 Å based on a Fourier shell correlation (FSC) threshold of 0.3 (Fig. 2A). The central sections of the 3D maps showed no marked structural differences among them in the protein shell; furthermore, difference maps calculated by subtracting L42 or L17 from wt capsids (or vice versa) showed no clear difference at this resolution (not shown). This interpretation was supported when the refinement of wt, L17, and L42 images was combined; the estimated resolution of the combined 3D map was 7.8 Å (Fig. 2B), indicating that these maps are virtually identical. The disappearance of the density corresponding to L17 or L42 insertions is thus a result of their conformational flexibility. The resolution of the merged reconstruction built only by S and NTA domains (see Materials and Methods), in which P domains were masked, was 6.9 Å (Fig. 2B).

Fig 2.

Assessment of the resolution of wt and chimeric VP1 capsids. (A) FSC resolution curves were calculated for wt (blue), L17 (purple), L42 (orange), N42 T=3 (green), N42 T=4 (dashed green), and C42 (brown) capsids. The resolutions at which the correlations dropped below 0.5 and 0.3 are indicated. For the 0.5 threshold, the values for wt, L17, L42, N42 T=3, N42 T=4, and C42 capsid were 10.3, 9.9, 10.3, 20.0, 24.4, and 14.7 Å, respectively; values for the 0.3 threshold were 8.8, 8.6, 8.8, 18.6, 23.2, and 12.0 Å, respectively. (B) FSC resolution curves were calculated for the merged data set (wt, L17, and L42 capsids), with P domain spikes (S+P_wt+L17+L42, green) or without (S_wt+L17+L42, purple), and compared to the wt VP1 capsid (blue). For the 0.5 threshold, the values for the merged data set with spikes or without were 8.9 and 8.1 Å, respectively, and for the 0.3 threshold the values were 7.8 and 6.9 Å, respectively.

The molecular architecture of the capsid is essentially as described previously (5); the most outstanding feature is the presence of 90 dimers protruding from the surface of a T=3 lattice (Fig. 3A). The asymmetric unit of a true T=3 capsid is formed by three quasiequivalent conformations of identical subunits termed A, B, and C (following the nomenclature used initially by Harrison et al. [22]), which are defined by the occupancy of structurally distinct environments. Conformers A, B, and C cluster into two dimer classes, A/B and C/C, as described for other calicivirus capsids and plant viruses, such as tomato bushy stunt virus. A/B dimers are located at a local 2-fold axis, and only the A subunits contribute to the pentamers at the 5-fold axis. C/C dimers are located at the 2-fold axis. The hexamers at the 3-fold axis consist of half of the C/C dimer alternating with B subunits (e.g., B-C-B-C-B-C). The other half of each C/C dimer is part of an adjacent hexamer.

Fig 3.

Three-dimensional cryo-EM of wt VP1 capsid. (A) Surface-shaded representations of the T=3 capsid outer (left) and inner (right) surfaces viewed along an icosahedral 2-fold axis. The positions of two VP1 dimers, A/B and C/C, are indicated. Icosahedral symmetry axes are numbered. Plug-like densities at the 3-fold axis on the inner surface are in red. Six triangle-shaped densities around a 3-fold axis are marked (blue circles; the inset shows a magnified view). Maps are contoured at 2 σ above the mean density (the transparent surface is contoured at 1 σ). (B) RHDV VP1 secondary structural elements (SSE). Segmented dimers A/B (left) and C/C (right) with their SSE: α-helices (red cylinders) and β-sheets (blue planks). Black arrows indicate the α-helix at the basement of the S domain that contributes to the triangle-shaped structures (and corresponds to the NTA α-helix); red arrows indicate the α-helix in the S domain. Note that two additional α-helices are identified at the P1 subdomain in C subunits but not in A or B subunits. To emphasize that the interactions occur only in P2 subdomains, dimers (contoured at 3 σ) are viewed from a different orientation than that in Fig. 5D and F. Inside views are also shown (bottom).

On the inner capsid surface, 60 inwardly protruding triangular structures surround the hexameric centers (Fig. 3A, circles; inset). This structure is formed by three rod-like densities that correspond to three α-helices, as determined by the secondary structure element (SSE) Hunter program (Fig. 3B). Both hexameric and pentameric regions are concave, forming an internal cavity; only hexameric cavities are nearly closed by a perforated plug-like density (Fig. 3A, red).

Pseudoatomic model of VP1 protein.

Although there is considerable X-ray structural detail available for calicivirus CP, no lagovirus CP has been characterized at atomic resolution. VP1 protein was modeled using SMSV and FCV CP as the templates based on their sequence identity and similarity levels. The average identity between CP sequences of RHDV versus SMSV or FCV was 25.65% (average similarity, 45.25%). When only S domain sequences were compared, identity increases to 40.75% (average similarity, 62.25%). P domain sequences showed 21% identity (38.75% similarity). These values are relatively high, considering that a lack of sequence similarity is common among viral structural proteins. Docking analysis further supported this selection (below), and SSE predictions were compatible with the template models (Fig. 4). Our 3D cryo-EM map was used as a constraint for the computationally derived model of VP1 protein, which was fitted in two steps. Flexibility was allowed between the P and S domains by splitting them at the linker region. Whereas P is rotated along its longitudinal axis by ∼30° for A and B subunits and ∼20° for C subunits, the S domains bent toward the interior of the capsid by ∼5° for A and B and ∼8° for C subunits. The local correlation of the N-terminal α-helix was then optimized at the inner surface of domain S (see Movie S1 in the supplemental material) to yield the quasiatomic structure of the T=3 capsid of RHDV (Fig. 5A and B).

Fig 4.

Multiple-sequence alignment and homology model of VP1. Multiple-sequence alignment of VP1, SMSV, and FCV CP amino acid sequences. Identical residues, white on red background; partially conserved residues, red. VP1 SSE reflects the consensus of several SSE predictions (see Materials and Methods), and those of SMSV (2gh8) and FCV (3m8l) CP are assigned based on their crystal structures. Color coding for SMSV CP is as initially shown by Prasad's group (9); N-terminal arm, green; S domain, blue; P1 subdomain, yellow; P2 subdomain, orange. The FCV CP X-ray structure was obtained from strain FCV-5 (GenBank accession no. DQ910790). Our study was performed with FCV Urbana strain (GenBank no. NC_001481). The two sequences have 90.42% identity and are the same size. The thick red line indicates the region equivalent to the N-terminal hypervariable loop used in this study (GSGNDITTANQYDAADIIRN).

Fig 5.

Quasiatomic model of RHDV T=3 capsid. (A) Complete T=3 capsid (white) with the VP1 quasiatomic model docked. The two VP1 dimer types are indicated (A/B, blue/red; C/C, green/green); white triangles define two icosahedral asymmetric units (ABC). Subunits with subscripts are related to A, B, and C by icosahedral symmetry (e.g., B to B₅ by a 5-fold rotation). The locations of 5-fold, 3-fold, and 2-fold axes are indicated. (B) A 50-Å-thick RHDV VLP slab viewed along an icosahedral 3-fold axis. P2 subdomain-mediated interactions can be seen in some A/B and C/C dimer sections. (C) A quarter of a 50-Å-thick RHDV VLP slab viewed along an icosahedral 2-fold axis. VP1 domains are color coded (P2, orange; P1, yellow; S, blue; NTA, green). Red densities indicate 2 of the 20 different densities at the 3-fold axis on the capsid inner surface, which were calculated by subtracting the capsid quasiatomic model from the wt cryo-EM capsid. (D and E) A/B contact, viewed along the line joining the 3- and 5-fold axes (D) and from the inside (E). Bent contact between S domains (D) and molecular swapping of NTA domains (E) are shown. (F and G) C/C contact, viewed along the line joining two 3-fold axes (F) or from inside (G). Planar contact between S domains (F) and molecular swapping of NTA domains (G) are shown. Icosahedral symmetry axes, black symbols. Diagrams of bent and planar contacts and the NTA exchanges are shown.

The VP1 model has three domains (Fig. 5C), an N-terminal arm (NTA, residues 26 to 70, green), a shell domain (S domain, residues 71 to 227; blue) and, connected by an 8-residue linker (residues 228 to 235), a protruding domain (P domain) subdivided into two subdomains, P1 (residues 236 to 285 and 448 to 571; yellow) and P2 (residues 286 to 447; orange). The S domain fold is the classical jelly roll β-barrel. The P2 subdomain, a large insertion in P1, is the most exposed region of the dimer. The VP1 model nevertheless lacked the first 25 N-terminal residues, as this region was invisible in the X-ray templates. A difference map obtained by subtracting the capsid quasiatomic model from the wt capsid showed plug-like densities at the 3-fold axis, which remained unoccupied by the VP1 pseudoatomic model (Fig. 5C, red).

Dimer assembly and T=3 capsid-stabilizing interactions.

The capsid quasiatomic model was analyzed in detail. Whereas A/B dimers have an inwardly bent conformation, C/C dimers are relatively flat (Fig. 5D and F). Both classes of dimers are stabilized by three main intradimeric interactions: (i) NTA are interchanged between icosahedral and local 2-fold axis-related subunits (Fig. 5E and G), (ii) the ¹⁰¹PFTAVLSQMY¹¹⁰ α-helices at the S domain contact each other, and (iii) P2 subdomains interact at the tips of projections (Fig. 5D to G).

Interdimeric contacts occur at the S and NTA domains. Interactions among the highly conserved β-barrel interfaces are mainly hydrophobic; the NTA α-helix forms a network of interactions beneath the shell that increases capsid stability. As in SMSV, three NTA α-helices from adjacent subunits have hydrophobic interactions at the local 3-fold axis, which colocalizes to the inner center of the asymmetric icosahedral unit.

Finally, whereas there are many interactions between the P base and the upper area of its related S domain in the A subunit, these contacts are almost absent from B and C subunits.

Visualization of the VP1 molecular switch.

Tilted and planar contacts, corresponding to A/B and C/C dimers, respectively (Fig. 5D and F), involve a structural polymorphism for the CP. The molecular switch responsible for these conformations is inherent to VP1 and is found in the first 29 N-terminal residues (5).

The docked VP1 structure indicated the approximate paths of VP1 N-terminal segments beneath the shell (Fig. 6A). This interpretation is supported by the proximity of the NTA α-helices, an unambiguous reference on the inner shell surface. Whereas the N terminus of B subunits extended toward the empty cavity at the 5-fold axis (Fig. 6A, red), N termini of A and C subunits converged at the 3-fold axis, where the plug-like density (∼18 Å thick, ∼38 Å diameter) is found (Fig. 6A, blue and green). This implies that the molecular switch is ordered, at least in part, as it obeys the icosahedral symmetry in A and/or C subunits, while it is disordered (invisible) in B subunits. The plug at the local 6-fold axis was the most important feature in the difference map calculated between the cryo-EM map and the capsid quasiatomic model (Fig. 5C).

Fig 6.

VP1 molecular switch. (A) T=3 VP1 capsid viewed down a 3-fold axis from inside, showing NTA domains. N termini of B subunits (red) end at the 5-fold axis; A and C N termini converge at the 3-fold axis into a conspicuous unoccupied density. The last visible N-terminal residue is indicated by a sphere. Black symbols indicate icosahedral symmetry axes (asterisks indicate local 3-fold axes in which three α-helices interact). (B) View as described for panel A, with accessible inner surfaces represented with electrostatic potentials, showing the distribution of negative (red) and positive (blue) charges. Hexagons mark the absent densities at the 3-fold axis in the pseudoatomic T=3 capsid. Note the negative grooves defined among the NTA segments.

The electrostatic potential on the capsid inner surface showed the distinct local environments around the 5- and 3-fold axes that could explain the two conformations, ordered and disordered, of the VP1 N termini (Fig. 6B). This representation indicated an unpredicted feature, as it showed that the hydrophobic surfaces of the swapped NTA delimited highly negatively charged grooves at the 2-fold and local 2-fold axes.

C42 capsid structure.

The chimeric C42 protein, containing FMDV B and T cell epitopes covalently bound to the C-terminal end, assembles into empty T=3 particles with an extra density at the 5- and 3-fold axes on the outer surface (Fig. 7A, red). This result is compatible with the VP1 model, whose C terminus faces the pentameric and hexameric cup-shaped depressions. The extra density did not account for the entire insertion, indicating that the insert was partially disordered. To confirm the location of these densities, we calculated a difference map by the arithmetic subtraction of wt capsid from C42 3DR (Fig. 7B and C).

Fig 7.

C42 capsid structure. (A) Surface-shaded representation of the C42 T=3 capsid outer surface viewed along an icosahedral 2-fold axis. Additional protruding densities, corresponding partially to the inserted epitope at the 3- and 5-fold axes, are red. Bottom, inserted sequence at the VP1 C terminus is shown (FMDV B and T epitopes, gray- and yellow-shaded regions, respectively). (B and C) Difference map calculated by subtracting wt VP1 from C42 capsid. The resulting difference map is shown in red on the outer surface of a C42 capsid viewed along icosahedral 3-fold (B) and 5-fold (C) axes. Fitted pseudoatomic models of the A, B, and C subunits of VP1 are color coded as described for Fig. 2 (A, blue; B, red; C, green). C-terminal ends are indicated by spheres.

N42 capsid structure.

The FCV CP hypervariable region (HVR) localizes at the outermost tip of its P2 subdomain, corresponding to a neutralizing B cell epitope. The chimeric N42 protein has two tandem copies covalently bound at the VP1 N terminus, and it assembles into ∼40- and ∼50-nm-diameter VLP, with preference for the smaller capsid assembly (∼2:1 ratio) (Fig. 1C). Whereas 50% of the initial T=3-like particles were selected by our classification analysis (see Materials and Methods) for further structural analysis, only 25% of the original set of larger capsids was selected; this suggests that these larger assemblies were unstable or obeyed limited icosahedral symmetry. The 3DR of larger N42 capsids showed a T=4 lattice, while smaller particles were similar to T=3 capsids (Fig. 8A to D). Both VLP types were assembled with partially preserved structural integrity, as the resolution of their maps were limited to 20 and 25 Å. To confirm the correct determination of image orientations, T=4 and T=3 maps were reprojected in appropriate viewing geometry and compared to original images (Fig. 8D).

Fig 8.

NT42 T=3 and T=4 capsid structures. (A) Surface-shaded representations of the outer (top) an inner (bottom) surfaces of the NT42 T=3 capsid, viewed along an icosahedral 2-fold axis at 20-Å resolution. A, B, and C subunits are shown as ribbons. Inner surfaces of NT42 (bottom left) and wt (bottom right) T=3 capsid show the differences at the 3-fold axis (arrows indicate the sectioned 3-fold axis). (B) Outer (top) and inner (bottom) surfaces of the NT42 T=4 capsid, viewed along an icosahedral 2-fold axis at 25-Å resolution. The icosahedral asymmetric unit is shown (A, blue; B, red; C, green; D, yellow). (C) S domains in the T=4 (top) and T=3 (bottom) icosahedral shells. Interacting surfaces between A, B, C, and D β-barrels are quasiequivalent. (D) Images of T=4 (row 1) and T=3 (row 4) capsids taken directly from original cryomicrographs compared to the projected views (T=4, row 2; T=3, row 3) of the 3DR in the corresponding orientation. Selected capsids are oriented close to a 2-fold (column 1), 3-fold (column 2), and 5-fold (column 3) symmetry axis.

P domains in the T=4 capsid nonetheless appear distorted, probably due to the steric hindrance of these domains. Dimers appear very close to each other and are poorly distinguished. The comparison of wt and NT42 T=3 capsids at the same resolution showed the outward displacement of the local 6-fold axis internal plug-like density in the NT42 capsid, filling the cavity rather than plugging it, as in the wt T=3 capsid (Fig. 8A, arrows).

Molecular modeling of the T=4 VP60-based capsid.

A larger VP1-based T=4 capsid requires an asymmetric structural unit of four quasiequivalent conformations. The most conservative approach to building a T=4 capsid based on dimeric capsomers would consist of two types of dimers, A/B and C/D, similarly to the tetravirus procapsids (46). To model the T=4 capsid from T=3 dimers, a superpentamer constituted by the five pentameric A/B dimers and the five adjacent C/C dimers was extracted from the T=3 capsid and fitted in the N42 capsid density (Fig. 8C, boldface line). Freedom was allowed to maximize the surrounding interactions at the S domain, and the optimal fit was obtained when each superpentamer was treated as an independent rigid body (but not with dimers or monomers). A pseudoatomic T=4 capsid was generated in which C/C dimers became C/D dimers (although they are indistinguishable at the resolution of our cryo-EM map). A difference map calculated between the T=4 and pseudoatomic maps showed small differences. Subunit environments are quasiequivalent between T=3 and T=4 capsids; A, B, and C subunits form local 3-fold interactions, while D subunits are related by an icosahedral 3-fold axis.

DISCUSSION

Our structural studies constitute a framework for the design of RHDV recombinant capsids as a delivery system for B and T cell epitopes (1, 12), as high-resolution structures of viruses in the genus Lagovirus currently are not available. In addition to defining RHDV particle structure, our studies contribute to establishing the determinants of its assembly process. We studied the mechanism that allows VP1 to switch among quasiequivalent conformational states. Mutants lacking the first 29 N-terminal amino acid residues lose the ability to acquire distinct conformations and assemble mostly into a T=1 capsid (5). The VP1 molecular switch that controls RHDV structural polymorphism is located in the NTA region, which faces the inner surface of the capsid shell. Here, we found that the FCV capsid protein primary structure has cryptic segment(s) that are able to increase structural polymorphism and, thus, its cargo size (or T number).

VP1 protein can accommodate insertions of foreign sequences at the N and C termini, as well as in a predicted exposed loop of the P domain, without disrupting VLP formation (5, 12). With the exception of the chimeric VLP with the N-terminal insertion, insertions at other sites do not affect T=3 capsid architecture. Reconstructions of the VLP with loop insertions were identical to wt VLP T=3 capsid organization, as the density for the loops was not visible in cryo-EM reconstructions. This indicates that surface insertions were dynamic and suggests great flexibility of the loops at the tip of the P domain; this feature might facilitate the interaction of this tip with its host receptors, in accordance with other studies (27, 47). For the C-terminal insertion, the extra density at the center of the hollows at the 6- and 5-fold locations is consistent with the C termini of the subunits facing these hollows.

The VP1 N terminus is the most critical site for the insertion of short heterologous peptides through genetic engineering. The N-terminal hypervariable region (NT HRV, or 5′HRV) of calicivirus capsid proteins, which maps in an exposed loop of the P2 subdomain, has a major role in antigenicity and receptor interactions in mammalian caliciviruses (9). The FCV HRV Urbana strain corresponds to a neutralization B epitope, and the double insertion of this epitope at the VP1 N-terminal end leads to larger T=4 capsids, increasing VP1 capacity for structural polymorphism. The preliminary analysis of a single-insertion mutant showed similar assemblies (unpublished data). These results, together with those including the VP1 N-terminal deletion mutants (5), show that assembly is perturbed by manipulating the switch region of this subunit. Nevertheless, heterologous epitopes at the N terminus do not always interfere with the VP1 switching mechanism; other insertion mutants, with either a 12-residue (MENYTDVFDDTQ) epitope from the transmissible gastroenteritis virus nucleoprotein (31) or a T cell epitope (SIINFEKL) derived from chicken ovalbumin, showed that chimeric assemblies corresponded to quasinative T=3 capsids (12). Specific interactions must take place between the inherent VP1 molecular switch and the exogenous peptide.

High-resolution X-ray structures of several caliciviruses show that the CP three-domain organization (NTA-S-P) is conserved (9, 34, 37). The NTA are similar and have an important role in capsid assembly; they act as molecular switches that alter subunit interfaces and/or conformations needed to assemble the quasiequivalent T=3 capsid. N-terminal deletion mutants have been generated for a number of T=3 viruses, with results analogous to those for RHDV, that is, T=1 capsids are assembled (15, 19, 25). This effect is not specific to T=3 capsids; in papillomavirus, the truncation of 10 residues from the major capsid protein N terminus leads to the assembly of T=1 rather than capsid-like T=7 structures (10).

The availability of SMSV/FCV capsid subunit structures facilitated the interpretation of our subnanometer cryo-EM data; the domain organization of the RHDV capsid protein is similar and the NTA is also located on the inner surface of the wild-type capsid. The remarkably close correspondence between the VP1 NTA α-helix (residues 47 to 56) and its cryo-EM density map suggest that the N-terminal 29-residue switches (not included in our VP1 model) contribute to the plug-like density at the 3-fold axis (unoccupied by the VP1 model after docking analysis) for A and/or C subunits. The molecular switch has two conformations: ordered (visible), to establish the flat 2-fold C-contacts, and disordered at the bent local 2-fold A/B contacts where it was not visible, at least for the B subunits. The plug-like density is similar to that of the β-annulus structure of many plant T=3 capsids (22, 44, 48).

The β-barrel motif is a very common fold in CP of many spherical viruses (39), and it is suitable to build unrelated capsids of various sizes (by increasing T number). A notable exception is the capsid protein of hepatitis B virus (HBV), which forms cores of various T numbers and does not share this fold. The proportion of HBV T=3 and T=4 particles is affected by modifying the length of the C terminus, which is located on the inner surface (50).

RHDV virions purified from infected rabbits consist of two particle types, intact and spikeless T=3 capsids (26), highlighting the importance of S domain interactions. P domains in the T=4 capsid nonetheless provide a distorted appearance, probably due to steric hindrance. These alterations are apparent at the 2-fold axis, in which hexamers are skewed, and at the 3-fold axis, where three P-P spikes from C/D dimers converge. Our results show that N42 T=3 and T=4 capsids use similar quasiequivalent subunit-subunit interactions at the shell, with the required change in the location of a quasi-6-fold axis from icosahedral 3-fold axes (T=3) to 2-fold axes (T=4). Both N42 particle types were assembled concomitantly in dynamic equilibrium and might represent a set of conformationally metastable particles, which is consistent with their relatively low preservation of icosahedral symmetry and their coexistence with smaller, broken assemblies.

Although T number polymorphism is not uncommon among viral capsid proteins (28, 45), most previous manipulations using genetic techniques to study the molecular basis of CP inherent structural polymorphism led to the loss of structural plasticity. The characterization of cryptic modulators not apparent from the sequence might help to decipher the molecular mechanisms used to create spacious capsids. Our ongoing studies use defined genetic modifications and a combinatorial approach to optimize the self assembly of chimeric VP1 proteins into T=4 (or larger) particles or the construction of other nanostructures. We anticipate that specific simple or combined changes in the HVR sequence will induce equilibrium displacement between T=3 and T=4 capsid assembly, and T=4 assembly be favored. This VP1-based design could have great effect when applied to the engineering of macromolecular assemblies (18).

Finally, the VP1-based system shows that the evolutionary barriers that must be overcome by CP-encoding genes might be smaller than anticipated. The incorporation of small determinants at critical sites could promote the formation of larger viral capsids; this would in turn allow the incorporation of larger genomes and the concomitant packaging of specific viral components, such as polymerases and other enzymes, reducing virus dependence on host cell machinery.