|Home | About | Journals | Submit | Contact Us | Français|
Hepatitis E virus (HEV) is a human pathogen that causes acute hepatitis. When an HEV capsid protein containing a 52-amino-acid deletion at the C terminus and a 111-amino-acid deletion at the N terminus is expressed in insect cells, the recombinant HEV capsid protein can self-assemble into a T=1 virus-like particle (VLP) that retains the antigenicity of the native HEV virion. In this study, we used cryoelectron microscopy and image reconstruction to show that anti-HEV monoclonal antibodies bind to the protruding domain of the capsid protein at the lateral side of the spikes. Molecular docking of the HEV VLP crystal structure revealed that Fab224 covered three surface loops of the recombinant truncated second open reading frame (ORF2) protein (PORF2) at the top part of the spike. We also determined the structure of a chimeric HEV VLP and located the inserted B-cell tag, an epitope of 11 amino acids coupled to the C-terminal end of the recombinant ORF2 protein. The binding site of Fab224 appeared to be distinct from the location of the inserted B-cell tag, suggesting that the chimeric VLP could elicit immunity against both HEV and an inserted foreign epitope. Therefore, the T=1 HEV VLP is a novel delivery system for displaying foreign epitopes at the VLP surface in order to induce antibodies against both HEV and the inserted epitope.
Hepatitis E virus (HEV) is a causative agent of acute hepatitis in humans and is primarily transmitted via the fecal-oral route. HEV is thus resistant to the low pH and digestive enzymes associated with the stomach and gastrointestinal tract. HEV regularly causes epidemics in many tropical and subtropical countries. In India, 101 outbreaks were confirmed by serological analysis in the state of Maharashtra in the last 5 years (6), and the lifetime risk of HEV infection exceeds 60% (28). Sporadic cases have also been reported in regions where HEV is endemic, as well as in areas where it is not endemic. Although some of these cases were associated with travel, many cases involved patients without a history of travel to regions where HEV is endemic. Accumulating evidence suggests that sporadic infection occurs through a zoonotic route and is not limited to developing countries. Seroprevalence suggests hepatitis E infection may also be prevalent in high-income countries (21), such as the United States (17), the United Kingdom (3), and Japan (18). The overall mortality rate of HEV infection during an outbreak generally ranges from 1 to 15%, and the highest mortality occurs in pregnant women, with fatality rates of up to 30% (19).
The HEV virion is composed of a 7.2-kb single-stranded RNA molecule and a 32- to 34-nm icosahedral capsid. The HEV genome contains three open reading frames (ORFs). The capsid protein, encoded by the second open reading frame (ORF2), located at the 3′ terminus of the genome, comprises 660 amino acids and is responsible for most capsid-related functions, such as assembly, host interaction, and immunogenicity. Recombinant ORF2 proteins can induce antibodies that block HEV infection in nonhuman primates (12, 27). Four major antigenic domains were predicted to be located within the C-terminal 268 amino acids of the ORF2 protein; one domain was experimentally identified as a neutralization epitope in the Sar-55 ORF2 capsid protein (25, 26). However, the minimal peptide needed to induce anti-HEV neutralizing antibodies contains residues 459 to 607 of the ORF2 protein (33), which is much longer than a linear antigenic epitope, suggesting that the neutralization epitope is conformational. Therefore, the detailed structure of the HEV capsid protein is required in order to understand the organization of HEV epitopes.
Currently, there are 1,600 HEV genomic sequences available through the International Nucleotide Sequence Database Collaboration. They are classified into four genotypes which vary by geographic distribution and host range (10). In contrast, only a single serotype has been identified, suggesting that the immunodominant domain of HEV is highly conserved among genotypes. Antibodies from any one of the four genotypes cross-react with the capsid protein of genotype 1 (7).
Like other hepatitis viruses, HEV does not propagate well in currently available cell culture systems. Hepatitis E preventive strategies so far rely on the use of ORF2-derived recombinant protein (16). When expressed in insect cells, recombinant truncated ORF2 protein (PORF2), with 52 residues deleted from the C terminus and 111 residues deleted from the N terminus, self-assembles into virus-like particles (VLPs) (15). Our previous structural analysis of recombinant HEV VLP by cryoelectron microscopy (cryo-EM) provided the first understanding of the quaternary arrangement of PORF2.
The essential assembly element of the PORF2 protein contained amino acids 125 to 600 (13), and the reconstructed VLP displayed a T=1 icosahedral particle composed of 60 copies of truncated PORF2 (30). Recently, crystal structures were reported for genotype 1 T=1 VLPs (31), genotype 3 T=1 VLPs (32), and genotype 4 T=1 VLPs (8), revealing that PORF2 is composed of three domains, the S domain, M domain, and P domain. The T=1 icosahedral shell is composed of 60 copies of S domains, while the M domain binds tightly to the S domain and interacts with two 3-fold-related M domains to form a surface plateau at each of the 3-fold axes. Two P domains are tightly associated as a dimeric spike that protrudes from each of the icosahedral 2-fold axes. As a result, on a low-resolution cryo-EM density map, the HEV T=1 VLP appears as an icosahedral particle with 30 spikes (30).
Although these VLPs are smaller (270 Å in diameter) than the native HEV virion (320 to 340 Å), oral administration of HEV VLPs to experimental animals can induce anti-HEV antibodies that bind to native HEV (14). When a B-cell tag of 11 amino acids on glycoprotein D of herpes simplex virus was covalently coupled to the C-terminal end of PORF2 (after residue 608), the fusion protein retained the ability of PORF2 to assemble and form chimeric T=1 icosahedral VLPs that were capable of eliciting systemic and mucosal antibodies against both HEV capsid protein and the attached B-cell tag (20). Therefore, the HEV T=1 VLP is a potential carrier for delivering not only HEV antigen but also foreign antigens or antiviral drugs to the host immune system. However, rational design of HEV-based delivery vectors requires detailed information on HEV VLP structure, as well as on HEV immunodominant domains.
Here, we identified antigenic structures using cryo-EM and three-dimensional reconstruction. Our results indicate that the binding footprint of a neutralizing antibody covers the lateral side of the P domain, while a B-cell tag at the C terminus does not alter the assembly of T=1 HEV VLP.
Eight-week-old female BALB/c mice were immunized at 0 and 4 weeks by intraperitoneal inoculation with HEV VLPs (100 μg/ml). Four weeks later, a final boost containing an equal volume of antigen was administered. Three days after the final boost, mouse spleen cells were fused with P3U1 mouse myeloma cells using polyethylene glycol 1500 (50% [wt/vol]) (Boehringer, Mannheim, Germany) essentially as described by Adler and Faine (1). Supernatants from microplate wells positive for hybridoma growth were screened by enzyme-linked immunosorbent assay (ELISA) using recombinant HEV VLPs as the antigen. Hybridomas that secrete antibodies specific for HEV were subcloned three times by limiting dilution, after which they were considered to be monoclonal. Antibodies in the supernatants were isotyped using a mouse monoclonal antibody isotyping kit (Amersham, Little Chalfont, Buckinghamshire, United Kingdom) in accordance with the manufacturer's protocol. Hybridomas were grown in bulk in stationary flasks (Nunc, Roskilde, Denmark) using RPMI 1640 with 15% fetal calf serum. Antibodies were purified from cell supernatants using HiTrap protein G affinity columns (Pharmacia Biotech AB, Uppsala, Sweden) and stored at −80°C. Among all of the antibodies that were generated, MAb224, an immunoglobulin G1 (IgG1) isotype, was chosen for structural analysis.
Isolated Fab224 fragments were prepared from purified mouse monoclonal antibodies by papain cleavage. A reducing l-cysteine buffer was used to activate the papain, and MAb224 was mixed with papain at a molar ratio of 100:1. The mixture was incubated overnight at 30°C. The reaction was stopped by the addition of iodoacetamide, and the product was analyzed by SDS-PAGE. The Fab224 fragments were purified using a 5-ml prepacked protein A chromatography column (Pierce Protein Research) according to the manufacturer's instructions. The Fc fragments and uncleaved MAb224 antibodies were trapped in the column due to their affinity for protein A, while the Fab224 fragments were collected in the flowthrough fraction.
Fab4 was prepared by phage display and purified according to the protocol described previously (25). Briefly, chimpanzee 1441 was infected with HEV strain SAR-55. Bone marrow was aspirated from the iliac crest of this animal, and the antibody κ-chain gene and γ1-chain gene were amplified and cloned into the pComb3H phage display vector and pGEM-T cloning vector (Promega), respectively, and transformed into Escherichia coli XL-1 Blue. The bacteria were then amplified and infected with helper phage VCS M13 at a multiplicity of infection of 50 to produce a library displayed on the surfaces of phage particles. Phage was panned on SAR-55 ORF2-coated ELISA wells; four rounds of panning were performed. After amplification of the selected library, the phagemid DNA was extracted and the vector was modified to remove the bacteriophage coat protein III-encoding region of the phage. The phagemid DNAs were religated and transformed into E. coli XL-1 Blue to produce soluble Fabs. The vector pComb3H was constructed to encode a six-histidine tail at the end of the Fab fragment, thus facilitating Fab purification. Fab4 purity was determined by SDS-PAGE, followed by colloidal Coomassie brilliant blue staining.
The production and purification of HEV VLPs were conducted as described previously (13, 15, 20, 30). Briefly, DNA fragments encoding the N-truncated ORF2 protein (for the wild-type VLP) and the chimeric ORF2 protein (for VLP-C-tag) were cloned using the baculovirus transfer vector pVL1393 to yield pVLORF2. Insect Sf9 cells (Riken Cell Bank, Tsukuba, Japan) were used to produce recombinant baculovirus. Tn5 insect cells were infected with the recombinant baculoviruses at a multiplicity of infection of 5 and incubated in Ex-Cell 405 medium (JRH Biosciences, Lenexa, KS) for 6 days at 26.5°C. The supernatant was collected after the removal of cell debris by centrifugation at 10,000 × g for 90 min. The HEV VLPs were pelleted at 100,000 × g for 2 h in a Beckman SW32 Ti rotor and resuspended in 4.5 ml Ex-Cell 405. The VLPs were further purified by centrifugation through a CsCl density gradient (1.31 g/ml) at 110,000 × g for 24 h at 4°C in a Beckman SW 55 Ti rotor. The white virus band was collected and diluted 4 times with Ex-Cell 405 to decrease the CsCl concentration, and then the VLPs were centrifuged for 2 h in a Beckman TLA 55 rotor at 100,000 × g. The VLPs were resuspended in 100 to 500 μl of 10 mM potassium-MES (morpholineethanesulfonic acid) buffer (pH = 6.2) and stored at 4°C. To construct chimeric VLP-C-tag, recombinant baculoviruses were prepared by inserting the B-cell tag epitope from herpes simplex virus glycoprotein D (QPELAPEDPED) at amino acid position 608 (20).
A series of DNA fragments were constructed to encode truncated ORF2 residues 112 to 660, 112 to 608, 112 to 602, 112 to 601, 112 to 600, 112 to 596, and 112 to 589. These recombinant ORF2 genes were inserted into a baculovirus vector and expressed in insect cells using the protocol for VLP production, except that the recombinant proteins were recovered from the cytoplasm after lysis of the cell. Recombinant proteins were heated in 4× Laemmli sample buffer and electrophoresed under reducing conditions in a 10% SDS-polyacrylamide gel. After transfer of proteins to a polyvinylidene difluoride (PVDF) membrane, the membrane was blocked with TBS buffer (20 mM Tris, pH 7.6, NaCl) containing 0.5% Tween 20 (vol/vol) prior to overnight incubation with Fab224 fragments at a 1:10 dilution. After extensive washing with TBS buffer containing 0.05% Tween 20 (vol/vol), alkaline phosphatase-conjugated anti-mouse IgG (Fab specific) was incubated with the membrane for 1 h at room temperature. The blot was then washed and developed with the p-nitroblue tetrazolium-5-bromo-4-chloro-3-indolylphosphate (NBT-BCIP) reaction.
The VLP-Fab complexes were prepared by incubating Fabs with VLPs at a molar ratio exceeding 1:300 (VLP versus Fabs) at 4°C overnight. To reduce the background density in the subsequent structural determination, highly pure VLP-Fab complexes were obtained using a short column containing Sephacryl 300, which resulted in the removal of the unbound Fab from the sample. The fractions containing VLP-Fab complexes were collected based on their optical density readings at a wavelength of 280 nm. The Fab binding occupancy was roughly estimated by performing SDS-PAGE (8-to-25% gradient) on the purified VLP-Fab complexes at a constant voltage using the Phast system (Pharmacia). The particle morphology of VLP-Fab complexes was examined by negative-stain electron microscopy using 2% uranyl acetate.
Sample preparation and cryo-EM were performed following previously described, well-established procedures (13, 30). Briefly, a drop containing 3.5 μl of the sample was applied to a glow-discharged holey carbon-coated copper grid, blotted with a piece of filter paper for 3 s to remove the extra liquid, and quickly plunged into liquid ethane cooled by liquid nitrogen. Samples were frozen in a thin layer of vitrified ice. The grid was then transferred into a Gatan 626DH cryo holder and kept at a low temperature (−178°C) during the subsequent data collection. Micrographs were collected under low-dose conditions (<10 e−/Å2) using Kodak SO163 film at a magnification of ×45,000 on an FEI CM-120 electron microscope operated at 120 kV, and particles were photographed at a defocus range of 1,000 to 3,000 nm. Micrographs were visually inspected and selected based on a suitable particle concentration, optimal ice thickness, and minimal specimen drift. Only micrographs fulfilling these criteria were analyzed.
Selected micrographs were digitized using a Heidelberg Primescan D8200 (Heidelberg, Germany) at a 14-μm scanning step size, corresponding to 3.11 Å per pixel of specimen space. Particles were manually picked and centered by cross-correlating each one against the circular average image. The astigmatism and defocus value were evaluated by the superimposed power spectra from all particles within a single micrograph. The contrast transfer function's first zero was approximately within the range of 17 to 20 Å−1 for the data used for the structural determination. The self-common-lines algorithm (4) was used to yield the initial models for VLP-C-tag, VLP-Fab4, and VLP-Fab224. The origin and orientation search for each particle was carried out iteratively using the polar Fourier transformation (PFT) algorithm running on an AMD MP1800 MHz dual-processor Linux workstation (2). Three-dimensional reconstructions were computed by combining a set of particles with orientations that spread evenly in an icosahedral asymmetric unit using the Fourier-Bessel algorithm and by superimposing 5-3-2 icosahedral symmetry. To examine the reliability of the three-dimensional reconstruction, the data set was evenly divided into two parts at the final refinement step and two three-dimensional reconstructions were computed. The resolution was estimated using Fourier shell correlation (FSC) by assessing the agreement between these two reconstructions in Fourier space. Using a coefficient value of 0.5 as the criteria, the estimated resolutions of the three-dimensional reconstructions of VLP-C-tag, VLP-Fab224, and VLP-Fab4 were computed as 17.5 Å, 18.5 Å, and 24 Å, respectively.
The three-dimensional reconstructions were rendered and visualized using the Chimera program (22). The contour level was chosen at a value corresponding to 100% of the mass of the PORF2 protein. The electron density map was displayed in the isosurface mode, which builds a barrier to contour the density about a certain threshold.
The density of the bound Fab molecule was determined from a difference density map, which was calculated by subtracting the cryo-EM map of unbound HEV T=1 VLP from the density map of the Fab-VLP complex. The cryo-EM map of unbound HEV VLP was published previously (30). Because the cryo-EM data for unbound VLP and the Fab-VLP complex were collected with the same FEI CM-120 electron microscope under similar imaging conditions, the difference density map was calculated by direct subtraction of the density of unbound VLP from the reconstruction of the Fab-VLP complex after normalizing the contrast between the two maps. The calculated difference map was used as a constraint in model fitting. Manual fitting was carried out by translational and rotational movement of the three-dimensional crystal structure of the PORF2 protein (PDB ID 2ZZQ) (31) into the cryo-EM density maps using program O (9). To obtain the best fit, the atomic model of the PORF2 subunit was treated as a rigid body. The fitting was first manually refined by minimizing the crashes between symmetry-related PORF2 molecules and then evaluated based on the cross correlation coefficient (CC value) between the cryo-EM density and the density computed from the fitted PORF2 coordinates. Fitting was halted when the CC value reached 80%. The figures were prepared using the program PyMOL (5), and the surface stereographic projection of the HEV VLP was prepared using the program RIVEM (29).
The binding of the monoclonal antibody Fab224 to PORF2 was examined via immunoblot analysis. A series of recombinant ORF2 proteins with C-terminal truncations were separated by SDS-PAGE on a 10% gel under reducing conditions and blotted with Fab224 (Fig. (Fig.11 A). Fab224 recognized both reduced and denatured recombinant ORF2 proteins that contained amino acids 112 to 660, 112 to 608, 112 to 602, and 112 to 601. In contrast, recombinant ORF2 proteins composed of residues 112 to 600, 112 to 596, and 112 to 589 did not bind to Fab224. These data indicate that residues 597 to 601 are critical for Fab224 binding to PORF2. Because the recombinant ORF2 proteins were recovered from cell cytoplasm where multiple forms of PORF2 were reported (15), the positive bands observed at a low molecular weight may be the proteolytic products or degraded forms of ORF2 that contain the Fab224 binding sequence.
The chimeric VLPs (Fig. (Fig.1C)1C) and the Fab224-conjugated VLP complex (Fig. (Fig.1D)1D) showed circular profiles with spike-like densities that extended from the surface. As we observed previously (15, 30), they appeared to have a white, contrasting center, indicating that they are empty particles lacking RNA (data not shown). The sizes of both VLPs were approximately 27 nm without taking into account the extra densities that extended from the VLP-Fab224 surface (Fig. (Fig.1D1D).
The cryo-EM structure of HEV-Fab224 was reconstructed from 615 images of individual particles and displayed T=1 icosahedral symmetry with 60 protein subunits that were arranged into 30 dimeric protruding spikes located at each icosahedral 2-fold axis (Fig. (Fig.22 A). Sixty Fab molecules were observed around each VLP particle, bound to the shoulder of the P domain. The Fab density extended ~57 Å radially away from the spike surface. The density corresponding to the Fab was approximately equal in magnitude to that of the HEV VLP, indicating that most or all of the 60 binding sites were occupied by a Fab molecule. The density corresponding to the VLP capsid was removed from the cryo-EM map, producing a Fab differential density map that was used to pinpoint the binding site of the Fab224 antibody (Fig. (Fig.33 A and B).
In addition, the structure of HEV VLP in complex with the neutralizing antibody Fab4 was determined by combining 264 individual images. Fab4 precipitates both the native HEV virion and recombinant PORF2 peptides, but the reaction depends on the presence of amino acids 597 to 607 (26). Three-dimensional reconstruction of the VLP-Fab4 complex showed 60 Fab molecules bound to each HEV VLP. Unlike the VLP-Fab224 complex, the density corresponding to Fab4 was about one-third of that of the capsid (Fig. (Fig.2A),2A), suggesting that only 30 to 40% of the binding sites were occupied by the Fab. Moreover, the binding of Fab4 appeared to be deeper on the side wall of the protruding domain toward the capsid shell, leaving its Fc domain exposed above the surface of the plateau (Fig. (Fig.2A).2A). In contrast, the entire Fab224 molecule stood mainly on the top of the P domain surface. The Fab224 and the Fab4 molecules extend along the long axis of the P domain. In both cases, no steric hindrance of the Fab on the P domain with the neighboring Fab molecules at either the 5-fold or the 3-fold axes was apparent. The orientation of the Fabs relative to the plateau appeared different at a radius of 135 Å. The long axis of Fab224 tilted toward the neighboring spike, while the long axis of Fab4 pointed to the 5-fold axis (Fig. (Fig.2A2A).
To further analyze the Fab and HEV VLP binding interface, the crystal structure of genotype 1 PORF2 was docked onto the VLP-Fab224 cryo-EM density map. The genotype 1 PORF2 crystal structure (PDB ID 2ZZQ) is composed of three domains (31), and these domains are in good agreement with those of genotype 3 and genotype 4 PORF2 (PDB ID 2ZTN and 3HAG, respectively) (8, 32). The coordinates fitted very well with the cryo-EM density map without any adjustment (CC value of 80%). The atoms on the surface of the HEV VLP capsid were plotted and colored according to their radial distance and overlapped with the density of the Fab at the surface plateau of the protruding spike (Fig. (Fig.2B2B).
The Fab224 interacted with the residues on the side of the ORF2 spike rather than with those residues on the spike's plateau surface (Fig. (Fig.3C).3C). The contact footprint did not overlap with the dimeric interface of the PORF2 spike. As expected, Fab224 recognizes a conformational epitope, and its binding site covers a surface composed of three loops, including amino acids 470 to 493 in AB loop, amino acids 539 to 569 in CD loop, and amino acids 581 to 595 in EF loop (Fig. (Fig.3D).3D). Residues E479, D481, T484, Y485, and S487 from the AB loop and residues Y532, S533, and K534 from the CD loop were in close contact with the Fab molecule.
Chimeric HEV VLP-C-tag was constructed using a PORF2 fusion protein in which a B-cell tag of 11 amino acids was incorporated into the C terminus of PORF2 (Fig. (Fig.1B).1B). A total of 782 images of individual particles were used to reconstruct the final three-dimensional model of VLP-C-tag. In agreement with the previously published cryo-EM VLP structures, the surface of VLP-C-tag can be divided into two distinct layers, an icosahedral shell and a protruding spike (Fig. (Fig.44 A). The spike projects outward from the icosahedral shell and is composed of a PORF2 dimer. The distance between two adjacent spikes was ~76 Å as measured between the centers of the surface plateaus. These results are consistent with the measurements of VLPs obtained either from Tn5 insect cells (30) or from Sf9 insect cells (13), and no detectable density was added onto the outer surface of the spike. No RNA density was detected within the chimeric VLP-C-tag.
The crystal structure fit very well within the VLP-C-tag density map (Fig. (Fig.4B),4B), indicating that the insertion of the C-terminal 11 amino acids inhibits neither the dimer-dimer interactions nor the formation of T=1 VLP. When the density maps were contoured to cover 100%, the radii of the S domains were roughly the same for both the VLP-C-tag and the VLP-Fab224 map, and the heights of the protruding spikes appeared similar. No density difference was observed from the docking (Fig. (Fig.5),5), suggesting that the inserted B-cell tag is flexible and less ordered. However, model fitting revealed that coordinates with unoccupied density appeared at the lateral side of the spike and underneath the Fab224 binding site (Fig. 5A and B), which may correspond to the inserted peptide.
HEV T=1 VLP is a vaccine candidate that induces protective immunity in nonhuman primates (12). It can also be used as an antigen carrier to deliver foreign epitopes through oral administration (20). Therefore, structural analysis of the antibody recognition sites is essential to suppress the neutralization effect of host vector-specific antibodies. For this purpose, we determined the structure of HEV VLP in complex with antibodies Fab224 (VLP-Fab224) and Fab4 (VLP-Fab4) and the structure of chimeric HEV VLP carrying a B-cell tag at the C terminus of PORF2 (VLP-C-tag). Docking the PORF2 crystal structure provides spatial information on the HEV antigenic domain and structural guidance to better design foreign epitope insertion.
The antigenic properties of HEV and the mechanisms by which it is neutralized are difficult to characterize due to the lack of adequate cell culture replication systems. Therefore, our understanding of HEV immunology is mainly based on studies using recombinant proteins expressed in E. coli (23) and recombinant proteins or HEV VLPs generated using the baculovirus expression systems (15, 24). Data from these studies indicate that the C-terminal region of PORF2 participates in the immune response against HEV and that the HEV neutralization epitope is conformational. The minimum peptide required to induce HEV-neutralizing antibodies corresponds to a region of 148 residues in PORF2, from amino acids 459 to 607 (33). This peptide coincides with the P domain revealed in the crystal structures of PORF2. The density of the Fab in our cryo-EM structure interfaced entirely with the spikes, thus confirming that the P domain is primarily responsible for HEV antigenicity. Fab4 is a chimpanzee antibody that recognizes the ORF2 protein and was isolated from a cDNA library by using phage display (25). Fab4 binds to native HEV virions and recombinant PORF2 peptides containing amino acids 597 to 607 (26). We performed fitting with the VLP-Fab4 structure; however, the Fab4 density was too weak to conclusively determine the Fab4 binding site on the surface of HEV VLP. However, the density corresponding to the Fab4 molecule did cover amino acid 606 (data not shown). It is not clear why Fab224 appeared not to interact with peptides lacking amino acids 599 to 608 in immunoblot analysis. However, the Fab224 binding site is consistent with the critical antigenic residues determined previously using mutagenesis. It was found that double mutations that changed residues E479 and K534 or Y485 and I529 to alanine selectively abrogated PORF2's reactivity with neutralizing antibodies (11). Experiments with another set of mutants defined the same region as the HEV antigenic domain, with antibody recognition residues spreading over the AB, CD, and EF loops (32). The antibodies used in both experiments were neutralizing antibodies; therefore, the Fab224 binding surface is part of the dominant neutralization site, suggesting that the monoclonal antibody Fab224 is a neutralizing antibody. This neutralization site partially overlaps with the receptor binding site (32), and antibody binding may create spatial hindrance that prevents HEV VLPs from attaching to the cell surface.
Because they are highly organized capsids that mimic the overall structure of virus particles, VLPs are a robust means by which to simultaneously carry small molecules, peptide antigenic epitopes, and DNA vaccines from heterogeneous sources to target disease sites. However, this rational vaccine design relies on excellent VLP structural information so that epitopes can be effectively conjugated to the VLP surface. In a previous study, rather than selecting PORF2 insertion sites on the basis of structural information, six insertion sites were selected according to restriction enzyme sites located either internally (four sites) or in the N or C terminus of PORF2. The internal sites are located after residues A179, R366, A507, and R542. Fusion proteins carrying insertions at sites A179 and R336 completely failed to produce VLPs, and insertions at A507 and R542 greatly reduced VLP production (20). Crystal structure data revealed that the spatial position of these sites is disadvantageous. Residue A179 is located in the S domain in the middle of an α-helix, which is necessary for the integrity of the S domain and its interaction with the 2-fold-related neighboring subunit. R366 is located in the M domain and favors electrostatic interaction with residue E386 from the 3-fold-related neighboring subunit. Although located within the P domain, the side chain of R542 is within the dimeric interface and guides the hydrophobic interaction of the two monomers. Replacement of R542 may misalign the orientation between two P domains and weaken the dimeric interaction between PORF2 proteins. Residue A507 in the P domain plays an important role in maintaining P domain orientation by fixing the angle of the long proline-rich hinge. Moreover, none of the four residues are exposed on the surface of VLPs, although some of them are located on the surface of individual PORF2 subunits (Fig. 4C and D). Therefore, the insertion of a foreign sequence at these sites does not interfere with the expression of individual proteins but, rather, hinders the assembly of HEV VLPs. The crystal structure revealed that the C terminus is exposed on the surface of VLPs, while the N terminus points toward the VLP center. Therefore, insertion at these two sites does not inhibit VLP assembly; however, the C terminus is more suitable for tethering bulky foreign antigenic sequences, as was shown in a previous report (20).
The cryo-EM structure of the chimeric HEV VLP-C-tag suggested that the B-cell tag was located at the lateral side of the spike, not far from residue A606 (C-terminal end in the crystal structure) (Fig. (Fig.5A).5A). This density is located beneath the Fab224 binding site but nonetheless overlaps with the potential binding site of Fab4. As a result, the insertion of the 11-amino-acid B-cell sequence may leave the HEV antigenic site partially open and accessible to the host immune system. This explains why mice can develop antibodies against both HEV and the foreign epitope after oral administration of VLP-C-tag (20).
In conclusion, the cryo-EM structures of VLP-Fab224 identified the lateral surface of the P domain as the recognition site for anti-HEV neutralizing antibodies. The insertion of a B-cell epitope at the PORF2 C terminus does not interfere with T=1 VLP assembly. Thus, T=1 HEV VLPs are a novel tool for oral vaccine delivery, as they constitute nonreplicating entities that can induce mucosal immunity without adjuvant. The induction of antibodies against both HEV and the target disease is an additional advantage of this delivery system.
We thank K. Kato for assistance with antibody preparation and N. Miyazaki for initial model fitting of the P domain structural density.
This project was supported in part by grants from the STINT Foundation, the Medical Research Council, and the PIOMS Institutional Program to R.H.C. This study was also partly funded by a grant from the Swedish Research Council to L.X. J.C.W. and L.X. were supported by grants from the Cancer Research and Discovery Programs, respectively. J.C.W. was initially supported by a grant from NSC as an exchange student under the cosupervision of D. M. Liou and Y. J. Sung.
Published ahead of print on 10 November 2010.