|Home | About | Journals | Submit | Contact Us | Français|
The crystal structure of rotavirus VP7 bound with the Fab from a neutralizing monoclonal shows the mechanism by which members of a large class of neutralizing antibodies inhibit rotavirus infection, indicates how withdrawal of Ca2+ ions becomes an uncoating trigger during cell entry, and provides the “first draft” of a design for subunit immunogens.
Rotavirus outer-layer protein VP7 is a principal target of protective antibodies. Removal of free Ca2+ dissociates the VP7 trimer, releases it from the virion, and initiates penetration-inducing conformational changes in the other outer-layer protein, VP4. We report the crystal structure of VP7 bound with the Fab fragment of a neutralizing monoclonal antibody. The Fab binds across the outer surface of the intersubunit contact, which is stabilized by two Ca2+ sites. Mutations that escape neutralization by other antibodies suggest that the same region bears the epitopes of most neutralizing antibodies. The monovalent Fab is sufficient to neutralize infectivity. We propose that neutralizing antibodies against VP7 act by stabilizing the trimer, thereby inhibiting the uncoating trigger for VP4 rearrangement. A disulfide-linked trimer is a potential subunit immunogen.
Rotaviruses are multi-layered, non-enveloped particles with dsRNA genomes (1). Four structural proteins form a complex, three-layered capsid, which packages two viral enzymes and eleven dsRNA genome segments. A double-layered particle (DLP) assembles in the cytoplasm, buds into the endoplasmic reticulum (ER), receives in this process a transient bilayer membrane, and ultimately acquires an outer layer of protein, viral protein 7 (VP7), in place of the transient envelope. VP7 must be present in sufficient quantity and fold correctly in order to displace the intermediate membrane (2–4). This unusual maturation pathway results in the coating of a cytoplasmically synthesized and assembled inner particle with an ER-synthesized glycoprotein, but with no intervening membrane in the mature virion.
The surface of the DLP is a T=13 icosahedral lattice of the trimeric protein, VP6, anchored on a T=1 inner layer of VP2 (Fig. 1A). VP7 is likewise a trimer, stabilized by Ca2+ ions (5). It forms the outermost virion layer, also with T=13 icosahedral packing, by capping the VP6 pillars (6, 7). Assembly of the VP7 shell locks into place a second outer-layer protein, VP4, which is anchored between VP6 pillars and protrudes above the VP7 layer (8, 9). VP4 spikes mediate attachment to cells and undergo a sequence of conformational changes that lead to endosomal membrane penetration (10, 11). Uncoating of VP7, probably by withdrawal of Ca2+, is necessary for these changes to occur (12). Thus, VP7 participates both in a membrane-displacing assembly step and in a membrane-disrupting entry step.
Rotavirus infection is the principal cause of severe, dehydrating diarrhea in infants (13). Live attenuated vaccines are now being introduced, but the efficacy and practicality of these vaccines in the impoverished settings in which most infant deaths from rotavirus occur have not yet been established (14). VP7 and VP4 are the targets of neutralizing and protective antibodies, and the structures and immunogenicities of these proteins underlie on-going efforts to produce next generation subunit vaccines. Viruses bearing VP7 of at least 15 different serotypes (designated G1-G15) have been isolated, 11 from humans (15, 16). Epitopes of a number of neutralizing monoclonal antibodies (mAbs) have been determined, but lack of a three-dimensional structure has precluded systematic study of neutralization mechanisms.
We have determined the crystal structure of the rhesus rotavirus (RRV, serotype G3) VP7 trimer, in complex with the Fab fragment of neutralizing mAb 4F8 (17). The core of the subunit folds into two compact domains, with disordered N- and C-terminal arms. There are two Ca2+ ions bound at each subunit interface in the trimer. The 4F8 Fab also binds across the trimer interface, stabilizing it even at Ca2+ concentrations that would normally lead to dissociation. Known epitopes map either to the same region of the trimer surface or to a region at the inter-domain boundary within a subunit. We show that the 4F8 Fab fragment neutralizes infectivity, with an IC50 only about 30- to 50-fold higher than that of the intact, divalent IgG, and we conclude that trimer stabilization, which will block uncoating, is the principal mechanism of neutralization by antibodies that recognize epitopes at the subunit interface. In work submitted elsewhere (Chen et al, submitted), which describes a cryoEM image reconstruction at 4 Å resolution of a DLP recoated with recombinant VP7, we show that the N-terminal arms of a VP7 trimer grip the underlying VP6 trimer. A hinge-bending rearrangement at the VP7 intrasubunit domain interface accompanies DLP binding. The 4F8 epitope at the intersubunit contact remains unaltered.
Recombinant RRV VP7, expressed in insect cells as described previously (5), and the Fab fragment of mAb 4F8, form a 1:1 complex that can be isolated by size exclusion chromatography (Fig. S1) and crystallized in space group P4132 (a=244.18 Å, one VP7 subunit plus one Fab fragment per asymmetric unit) from PEG 4K in a pH 5.6 sodium citrate buffer with 0.1 mM CaCl2. Previous efforts to crystallize the VP7 trimer had yielded only very disordered crystals. We recorded diffraction to a minimum Bragg spacing of 3.4 Å resolution using beamline ID-24C at the Advanced Photon Source (Argonne National Laboratory) (Table S1). We obtained starting phases by carrying out a molecular replacement search with a library of 244 antibody fragment structures. We performed the rotation function calculations with an integration radius of 35 Å, using the program MOLREP (18), with each of the 244 structures as probe. Of these calculations, 8 yielded promising solutions, as judged by a distinct difference between the ratio of rotation-function score to sigma (RF/σ) for the best solution and that for the next best. Inspection showed that the Fab fragments that yielded these solutions all had similar elbow angles and that the positions and orientations of the fragment in the unit cell all had similar values. One of the top solutions was 1dbm, a murine IgG1-κ Fab, like 4F8. We therefore used this model, modified to display the correct residues of the 4F8 variable domains (sequenced from the hybridoma cell line), for further calculations and model building. Most of the VP7 polypeptide chain was evident in the molecular-replacement electron density, with the exception of N- and C-terminal segments, and we could build a preliminary model without difficulty. The high solvent content (87%) allowed us to calculate a very clear map by solvent flipping (Fig. S2), despite the modest resolution (3.4 Å).
The VP7 subunit is a “Rossmann fold” (domain I), with a jelly-roll beta sandwich (domain II) inserted between α-helix D and β-strand 11 (Figs. 1B,C). There are four disulfide bonds, one within domain I, and the other three within domain II (Fig. 1b). After initial refinement, a difference map revealed two strong peaks at the subunit interface. Addition of Ca2+ ions to the model at each of these positions improved the Rfree. The final model contains residues 78 to 312 (Table S1). Asn69, the single glycosylated residue in RRV VP7, is in the disordered, N-terminal arm (Fig. 1D). Three subunits assemble into a thin triangular plate with a central depression, a variable surface that faces outward on the virion, and a more conserved, somewhat negatively charged, inward facing surface.
The two Ca2+ sites both have side-chain carboxylate and main-chain carbonyl neighbors appropriate for divalent cations, and the contributing side chains are conserved among Group A rotaviruses (Fig. 2). There are six protein-derived ligands at site 1 and four at site 2; one or two water molecules, not included because of the limited resolution, presumably complete the coordination sphere at the latter site. Site 2 is close to the Fab interface. Presence of the bound ion might influence the conformation of VP7 loops in contact with the antibody. Mutants in strain RF that confer resistance to low calcium have a P75L substitution, sometimes accompanied by a P279S mutation (19). The former site is in the N-terminal arm, at a location where a leucine might enhance the grip on VP6; the latter, on the inward-facing surface of VP7, is at a point of contact with the outward-facing surface of VP6 (19). The resistance to low Ca probably arises from the indirect effect of either mutation in stabilizing the contact with VP6, rather than from a direct effect on the Ca2+ sites.
The Fab binds across the intersubunit junction, on the surface of the trimer that will face outward when VP7 coats the virion (Chen et al, submitted)(Fig. 2). The heavy chain carries a majority of the contacts (78% of total buried surface area at the Fab/VP7 interface), to domain I of one subunit and domain II of the other. The one clear light-chain contact is at the site of a neutralization escape mutation (N96D)(20). Like all well-characterized, VP7-directed neutralizing antibodies, 4F8 binds only trimeric VP7 and not the Ca2+-free monomer (21). A number of neutralizing antibodies, including 4F8 and 159, block VP7 uncoating from virions, even in the presence of Ca2+ chelators (21, 22). By locking down the subunit interface, these antibodies probably stabilize the trimer, even at very low divalent cation concentrations.
If 4F8 prevents uncoating by stabilizing individual trimers, then the Fab fragment should also neutralize. Fig. 3 shows the results of neutralization assays with intact 4F8 antibody and with 4F8 Fab. The IC50 for the Fab is about 30- to 50-fold higher than for the intact, bivalent mAb, presumably because of an avidity effect, but the neutralization activity of the Fab is unambiguous. Earlier experiments that came to the opposite conclusion generated the Fab by papain digestion of virion-antibody complexes (22, 23); the concentration of Fab thus produced was probably not sufficient to give a measurable effect.
The sites of mutations in VP7 that permit escape from neutralization by various monoclonal antibodies map to two regions, 7-1 and 7-2, on the exposed surface of the protein. Each region includes several “epitopes” previously designated by letters (see Fig. 4 and Table S2, in which we group the published epitopes into the two structurally defined regions). Region 7-1, which spans the intersubunit boundary, is immunodominant. It contains the positions of escape mutations selected by 58 of 68 tested neutralizing mAbs, including 4F8. Two mAbs select escape mutations in both regions 7-1 and 7-2 (Table S2). Modification at position 211 by an oligosaccharide, which would block antibody binding to region 7-1, confers resistance to neutralization by hyperimmune anti-rotavirus serum (20). Most or all of the antibodies that bind region 7-1 probably neutralize by a mechanism similar to the one proposed for 4F8 in the preceding paragraph. Region 7-2 is at the interdomain boundary within a single VP7 subunit. Antibodies that bind region 7-2 may neutralize by a different mechanism. Analysis of the high resolution cryoEM reconstruction of VP7-coated DLPs, described in the accompanying paper (24), shows that the subunit undergoes a conformational change when it binds the DLP, with a substantial dislocation at the domain interface but little or no change at the intersubunit contact. We suggest that antibodies that bind in region 2 stabilize the trimer either by fixing the virion-associated conformation, by cross-linking subunits within a trimer, or by cross-linking adjacent trimers on the surface of the virus particle.
The sequences of amino-acid residues at positions 87-101 and 208-211 are conserved among strains within a G serotype but not across different serotypes. These sequences are “signatures” that can be used to predict the serotype of a new isolate (25). They correspond to surface ridges on either side of the subunit interface, in regions 7-1a and 7-1b, respectively. Because both homotypically and heterotypically neutralizing mAbs select mutations in region 7-1 (and also in region 7-2), some aspect of antibody binding other than epitope location must determine breadth of neutralizing capacity (20, 26, 27).
Regions 7-1 and 7-2 together cover much of the outward-facing surface of VP7. Analysis of sequences from all 11 human G serotypes shows that most of the variability is in residues on the outward-facing surface of the VP7 trimer (Fig. S3). Because there is no extensive conserved patch on this surface, any potential cellular receptor that binds VP7 on the virion must have a very small binding footprint or its identity and site of interaction must vary among serotypes or isolates. It has been suggested that VP7 may interact with αXβ2 integrins after attachment (28). The site proposed as an integrin-binding motif (GPR, resides 253–255) is on the inward-facing surface of the trimer and would only be available to interact with integrins after uncoating.
We have designed a disulfide-linked variant of the VP7 trimer by substituting cysteines for Thr276 and Gln305, which face each other across the subunit contact. The resulting disulfide is largely buried at the interface. The secreted product of insect-cell expression is indeed stably trimeric (Fig. S4). Unlike the highly infectious particles obtained by recoating DLPs with wild-type recombinant outer capsid proteins (9), DLPs recoated with VP4 and the disulfide-stabilized VP7 variant are not infectious, nor do they mediate α-sarcin co-entry, an assay for membrane penetration (29, 30). The arm-grip mode of association with VP6 thus allows the VP7 trimer to bind and to lock VP4 in place, but cross-linking of the core subunits retards dissociation, probably by several orders of magnitude. These observations support the notion that withdrawal of Ca2+ is the uncoating trigger in an endosome.
Rotavirus infection and parenteral immunization with virions both induce a strong VP7-specific neutralizing antibody response (1). Recombinant VP7 elicits neutralizing antibodies only inefficiently, probably because the free trimer dissociates; the response can be enhanced by adding a C-terminal membrane anchor, which presumably increases trimer stability by immobilizing the subunit in two dimensions on the cell surface (5, 31). The disulfide cross-linked VP7 trimer, which binds neutralizing antibodies such as mAb 159, is a good first candidate for a more effective, stable, structurally engineered subunit immunogen.
We thank Marina Babyonyshev for help with protein preparations, the staff of NE-CAT (APS, Argonne National Laboratory) for assistance with x-ray data collection, Robyn Stanfield (The Scripps Research Institute) for the library of Fab structures, and John Patton (NIAID) for help with sequence variability analysis. The work was supported by NIH Grant CA-13202 to SCH and by an Ellison Medical Foundation New Investigators in Infectious Diseases Award to PRD. SCH is an Investigator in the Howard Hughes Medical Institute. PRD and ECS are employees and shareholders of Novartis Vaccines and Diagnostics, Inc.