A critical step in the productive infection of a virus is its successful entry into a host cell. For many viruses, a single protein is implicated in this complex process that often involves multiple steps and multiple receptors. Such is the case with rotavirus, the major pathogen of infantile gastroenteritis accounting for nearly half a million deaths annually worldwide (
8). Rotavirus, a member of the
Reoviridae family, is a large icosahedral virus with a complex organization consisting of three concentric capsid layers that encapsidate 11 genomic double-stranded RNA segments (
16,
41). The innermost capsid layer is composed of 120 copies of VP2 on a T=1 lattice (
30), and 12 copies each of the RNA-dependent RNA polymerase VP1 (
44) and the guanylyltransferase VP3 (
6,
31) attach as heterodimeric complexes to the inner surface of this layer at the fivefold axial positions (
41). The intermediate capsid layer is composed of 780 copies of VP6 arranged as 260 trimers on a T=13 lattice. The outermost capsid layer contains 780 copies of VP7, with the same icosahedral organization as the intermediate layer, and 120 copies of VP4, which interacts with VP6 and emanates as distinct bilobed spikes through the VP7 capsid layer. Antibody labeling and cryo-electron microscopy (cryo-EM) showed that the spikes on the surface of the triple-layered particle (TLP) are present as 60 dimers of VP4 located near the type II channels surrounding each fivefold vertex (
40,
43).
Although earlier studies implicated VP7 in the cell entry process (
19,
42), subsequent studies have increasingly indicated that VP4 is the major player in this process. VP4 is implicated not only in cell attachment and penetration but also in hemagglutination, neutralization, virulence, and protease-enhanced infectivity of rotavirus (
25,
32,
34). The latter phenomenon is particularly relevant considering that rotavirus replicates in the mature enterocytes of the small intestine, an environment rich in proteases. Proteolytic cleavage of VP4 primes the virus for efficient entry into cells (
1,
17,
26). During proteolysis, VP4 (88 kDa) is cleaved into VP8* (28 kDa) and VP5* (60 kDa), and the cleavage products remain associated in the virion (
18). Our recent structural and biochemical studies have shown that VP4 undergoes a conformational transition from a disordered to an ordered state upon trypsinization, and this transition appears to be responsible for trypsin-enhanced infectivity observed with rotavirus (
10). The X-ray crystallographic structures of VP8* and VP5* have provided strong evidence that the distal globular domain of the VP4 spike is composed of VP8*, with the remaining body of the spike consisting of VP5* (
12,
13,
43).
A consensus opinion that has emerged from recent biochemical studies is that rotavirus entry into cells is a multistep process involving sialic acid (SA)-containing receptors in the initial cell attachment step and integrins such as α
vβ
3, α
4β
1, and α
2β
1 during the subsequent postattachment steps (
9,
22,
23,
45). In this process, the VP8* domain, which has a galectin fold, is involved in the interactions with SA, whereas VP5* is implicated in the interactions with integrins. Involvement of SA during rotavirus infection is not an essential step for all rotavirus strains. For the majority of rotavirus strains, including human rotaviruses, cell entry is SA independent (
7). In these viruses, the majority of neutralizing monoclonal antibodies (MAbs) that recognize VP4 select mutations in VP5* (
27,
28,
38), suggesting that cell entry is mediated mainly by VP5*. It also is thought that cell penetration of rotavirus may require a hydrophobic, fusion domain, which resides on the VP5* cleavage product. These hydrophobic regions could aid in membrane penetration after cell attachment (
11,
14).
As has been observed with other viruses such as influenza virus (
4), flavivirus (
36,
37), alphavirus (
20), and picornaviruses (
3), it is possible that VP4 undergoes distinct conformational changes at various stages during cell entry to mask certain epitopes and reveal others to optimally interact with different receptors and the cellular membrane. We have already seen that VP4 undergoes a drastic conformational change upon trypsinization. Recent X-ray crystallographic studies of VP5* also suggest the possibility of significant structural changes in the spike structure during rotavirus cell entry (
12). Tracking these conformational changes and identifying the epitopes to understand the molecular basis of how VP4 interacts with various receptors during entry is indeed difficult, particularly considering that there is no established reverse genetic system for any member of
Reoviridae, including rotavirus. In lieu of reverse genetics, we have devised a strategy to further our understanding of structure-function relationships in rotavirus. In this strategy, we perturb the particle structure by varying the chemical conditions such as pH, ionic strength, and temperature and evaluate how these perturbations affect virion function. Such a strategy has been useful for gaining insights into the structural organization of the genome in rotavirus (
39). Using a combination of cryo-EM and biochemical techniques, we show here that at elevated pH the VP4 spike undergoes an irreversible conformational change from a bilobed structure to a distinctly stunted trilobed structure to alter the cell binding characteristics of the virus. This structural change is completely abrogated by a neutralizing MAb, allowing us to propose a mechanism of neutralization by this antibody.
MATERIALS AND METHODS
Cells, virus, and antibodies.
Rotavirus TLPs (RRV and SA11-4F), grown in the presence of trypsin in MA-104 cells, were purified as described previously and suspended in TNC buffer (10 mM Tris, 150 mM NaCl, 10 mM CaCl
2 [pH 7.5]) (
39). MAbs were purified from ascites by using the MAb-Trap Kit (Amersham Pharmacia Biotech, Piscataway, NJ), and Fab fragments were created by using the ImmunoPure Fab Preparation kit (Pierce, Rockford, IL). Purified Fabs were eluted into 0.1 M phosphate-buffered saline (PBS) at a concentration of 0.7 mg/ml. Prior to addition to rotavirus for structural study, Fabs were dialyzed against 10 mM Tris (pH 7.5) to inhibit clumping of the virus particles. To study the effect of various chemical and pH conditions on the virus, the specimen was dialyzed by using a microdialysis button (Hampton Research, Laguna Niguel, CA) for 30 min in either 250 mM ammonium hydroxide (pH 11.5) or CAPS buffer (pH 11.5). In all biological assays, the specimen was brought back to a normal pH of 7.5 by dialysis in TNC. The term “pH-treated,” unless otherwise stated, refers to particles taken to high pH, to induce conformational changes, and then brought back to normal pH.
Virus radiolabeling.
To quantitate the relative amounts of VP5* and VP8* before and after ammonium high-pH treatment, SA11-4F rotavirus was radiolabeled during infection and viral replication. MA104 cells were starved of methionine and cysteine for 3 h by using Dulbecco’s modified Eagle medium ((1×) lacking l-glutamine, sodium pyruvate, l-methionine, and l-cysteine (Life Technologies, Rockville, MD). The cells were then infected and, after 1 h, l-methionine and l-cysteine (35S-Met/Cys; Amersham Pharmacia Biotech) were added, and the virus was allowed to grow overnight. Virus was purified as described above, and incorporation of 35S-Met/Cys following purification was measured by scintillation counting. Virus concentration was calculated based on the absorbance at 260 nm, and the specific activity of radiolabeled virus was calculated to be ∼7,000 cpm/μg. Both untreated and pH-treated radiolabeled virus particles were repurified again by cesium chloride ultracentrifugation to separate the intact capsid from any potentially soluble protein fragment removed by the pH treatment. Treated virus banded at the same density as untreated virus, and both types of sample were pelleted for analysis. The purified virus samples were then analyzed again to measure protein and radioactivity and found to have no significant decrease in specific activity.
SDS-PAGE.
To determine protein stoichiometry and apparent molecular weights of the protein components of untreated and pH-treated virions, both samples of repurified radiolabeled particles were denatured by boiling and then separated by sodium dodecyl sulfate-12% polyacrylamide gel electrophoresis (SDS-12% PAGE). Gels containing radiolabeled proteins were dried and exposed to PhosphorImager screens, and analyzed in a Storm PhosphorImager (Molecular Dynamics, Sunnyvale, CA). The data were quantitated by using ImageQuant software, and protein bands were boxed out for stoichiometric analysis.
Western blots.
Verification of protein identity and antigenicity was confirmed by Western blotting. A SDS-12% polyacrylamide gel of untreated and pH-treated virus was transferred to a Hybond nitrocellulose membrane (Amersham Pharmacia Biotech) and exposed overnight with primary antibodies to SA11-4F, VP5*, and VP8*. Western blots were developed by using chemiluminescence.
Determination of virus infectivity.
The infectivity of the untreated and pH-treated viruses was determined either by sequential passage of virus in MA104 cells or by plaque assay (
5). Samples were passaged three times in six-well plates to determine whether any infectious virus survived the pH treatment. Initially, 10 μl of 1 mg of untreated and pH-treated virus/ml was incubated with 10 μg of Worthington trypsin/ml. After each passage, the plates were frozen and thawed three times, the samples were treated with Worthington trypsin for 30 min at 37°C, and new cells were inoculated. After each passage, the wells were observed to determine whether the cells exhibited cytopathic effect. The control, untreated sample induced cytopathic effect, whereas the pH-treated virus did not. Samples from both conditions also were subjected to plaque assay after each passage.
Hemagglutination.
To determine whether virus binding to sialic acids was affected by pH treatment or the binding of 2G4 Fabs, hemagglutination assays (HAs) were performed. Untreated RRV, RRV taken to either pH 9.75 or 11.5 and then returned to pH 7.5, RRV pretreated with 2G4 Fabs and then subjected to high pH treatment, or RRV taken to either pH 9.75 or 11.5, returned to pH 7.5 and then treated with 2G4 Fabs were serially twofold diluted in a 96-well plate. The first well-contained a 1:100 dilution of the initial virus concentration of 0.34 mg/ml. Type O human red blood cells were washed three times by diluting with 10 ml of PBS and centrifuged at 1,000 rpm for 10 min each. Then, 50 μl of 0.5% red blood cells in PBS was added to each 50 μl of virus per well. After incubation for 1 h at room temperature, the plate was read for hemagglutination activity. The hemagglutination titer was determined as the reciprocal of the final dilution resulting in hemagglutination.
Cell-binding assays.
Cell-binding assays were performed essentially as previously described (
10). Untreated and pH-treated methionine-labeled TLPs were serially diluted in TNC containing 1% bovine serum albumin, and added to monolayers of MA104 cells in a 96-well plate. The specific activity of the virus was calculated to yield an initial value of 4 μg of virus. Both radiolabeled virus and radiolabeled virus plus a 10-fold concentration of unlabeled competitive virus were added to the plates in duplicate. The cells were incubated on ice for 1 h, washed twice with TNC buffer, and then lysed with 50 μl of 1% SDS. The lysate was counted by using a Beckman LS 3801 scintillation counter (Beckman Instruments, Fullerton, CA). Each assay with the TLPs was performed in duplicate and repeated twice.
Cryo-EM.
Specimen preparation for cryo-EM was carried out by using standard procedures (
10,
15). Each specimen, at a concentration of ∼1.5 mg/ml (5-μl aliquot) was pipetted to a holey carbon-coated copper grid (Quantifoil MicroTools, Jena, Germany). The grid was blotted and flash-frozen in liquid ethane (−190°C) and loaded on a Gatan (Gatan Instruments, Pleasanton, CA) cold-stage cryoholder. Electron micrographs were recorded on a JEOL 1200 transmission electron microscope operating at 100 kV with a magnification of ×30,000 by using an electron dose of 5 e
−/Å
2. From each specimen area in the grid, a focal pair with intended defocus values of 1 and 2 μm was recorded. The images were taken with 1-s exposure to Kodak S0-163 EM film (Eastman Kodak Co., Rochester, NY) and developed for 12 min in Kodak D-19 developer, followed by fixation for 10 min in Kodak fixer.
Antibody labeling of particles.
Decoration of rotavirus TLPs with the Fab fragment of MAb 2G4 was carried out to observe whether the epitopes were conserved. Purified TLPs were pretreated overnight at a stoichiometry of 5 Fabs per VP4 monomer. After Fab addition, some of the samples were dialyzed against ammonium hydroxide (pH 11.5), redialyzed with TNC (pH 7.5), and then flash frozen. In other cases, virus which had been pH treated was then treated overnight with Fabs. Samples were then imaged by cryo-EM as before.
Three-dimensional reconstructions.
Cryo-electron micrographs were selected for correct defocus, ice quality, contrast, and particle concentration and were scanned on a Zeiss SCAI microdensitometer (Carl Zeiss, Inc., Englewood, CO), with a scanning interval of 14 μm corresponding to 4.67 Å in the object. Reconstructions were carried out by using the data from closer-to-focus images to a resolution within the first zero of the contrast transfer function with appropriate corrections. The defocus values were −1.34 μm (using 136 particles to a resolution of 22.3 Å) for pH 7.5, −1.34 μm (using 175 particles to a resolution of 22.3 Å) for pH 11.5, and −1.46 μm (using 118 particles to a resolution of 22.8 Å) for Fab pretreatment, followed by high-pH treatment, and −1.46 μm (using 75 particles to a resolution of 22.8 Å) for Fab labeling after high-pH treatment. Determination of the defocus values, orientation determination, refinement, and three-dimensional reconstructions were carried out by using the ICOS toolkit software (
29), and resolution assessment and choice of the appropriate contour levels were conducted as described previously (
39). For each reconstruction, the number of particles with unique orientations was found to be adequate by examining the spread of inverse Eigen values, which was <0.1 for 99% of the data. After the final refinement, in each case, the average phase residual between the images and their corresponding projections was <45°. Threshold values for the reconstructions were chosen to account for 780 molecules of VP6 between radii of 250 and 350 Å in all reconstructions.
Difference maps and antibody fitting.
The maps of rotavirus before and after addition of Fab and/or high-pH treatment were scaled, and differences were computationally calculated. Maps without Fabs were subtracted from maps containing Fabs, and differences (representing Fab densities) were color coded to indicate regions of interaction with VP4.
DISCUSSION
The study presented here, together with other structural studies on rotavirus, clearly shows a conformational flexibility inherent in VP4 spikes. Our previous studies showed trypsinization transforms flexible spikes into rigid 120 Å bilobed structures that are clearly visible in cryo-EM reconstructions of trypsinized virions at physiological conditions (
10). Cryo-EM reconstructions with VP4 specific antibodies have shown that these spikes are dimers of VP4 (
40,
43). However, recent crystallographic studies have shown that the tryptic fragment VP5*, which constitutes the central body of the spike, forms a trimer (
12). These crystallographic studies have further suggested that VP4 may undergo a transformation from a dimeric to a trimeric structure during the cell entry process. Considering that several biochemical studies have increasingly suggested that cell entry of rotavirus may involve sequential interactions with multiple receptors (
33), such a conformational flexibility is perhaps not surprising.
In our studies presented here, we have shown that VP4 spikes at high pH undergo a drastic irreversible conformational change resulting in a stunted structure with a pronounced trilobed appearance. In contrast, the overall structure of the virion, including the spikes, is unaffected by low or moderate pH conditions consistent with its ability to survive in the low-pH environment of the gastrointestinal tract. The mass density calculations of the spikes altered by high pH clearly indicated significant mass loss. Our initial conjecture was that the tryptic fragment VP8*, which is localized to the distal globular domains of the native spike (
13), was dislodged from the spike. However, Western blot and stoichiometric analyses of radiolabeled particles unambiguously indicated that both VP5* and VP8* are present on particles in the same stoichiometric proportions before and after high pH treatment. This indicates that the interactions, possibly hydrophobic, between these two cleavage products in the spike structure are strong enough to withstand high-pH treatment and remain associated with the capsid. These observations further suggest that a significant portion of the spike structure becomes disordered or flexible at high pH and therefore is not visible in the reconstructions.
What is the chemical composition of the visible portion of the altered spike? The distal portion of the altered spike is recognized by a VP5*-specific antibody 2G4, suggesting that the altered spike consists of VP5*. Because this antibody recognizes only a conformational epitope (
35), this observation further indicates that, despite high-pH treatment, the structural integrity of the antigenic epitope in VP5* is intact. The observations that pH-treated particles have lost the ability to hemagglutinate and interact with SA indicate that VP8*, which is responsible for these two properties, is likely the portion that is disordered and/or flexible and not visible in the reconstruction. Three Fab densities are seen attached to each of the trilobed spikes. There are two possible interpretations to this observation. One possibility is that the trilobed spike represents a trimerized VP5*. Recent crystallographic studies clearly indicate that VP5* forms trimers with strong interactions between the monomeric subunits (
12). The authors of that study proposed a model in which each spike is a trimer of VP4 and, upon trypsinization, two of the monomers form the visible spike in cryo-EM reconstruction rotavirus particles, while the other monomer being floppy is not visible in the reconstruction. During cell entry, by yet unknown entry associated events, the floppy VP4 monomer, together with the other two molecules, trimerize as seen in the VP5* crystal structure. It is plausible that high-pH treatment renders VP8* disordered and triggers dimer-to-trimer transition of VP5*, perhaps mimicking a post-SA-attachment step during cell entry. Such a drastic conformational rearrangement, although first proposed for nonenveloped viruses, has been visualized for enveloped viruses such as flavivirus (
36,
37) and alphavirus (
20). Another possible interpretation, which cannot be ruled out, is that the trilobed spike represents one stable monomer and another monomer in two alternate conformations. Higher-resolution cryo-EM structural analysis of the pH-treated particles may be required to resolve this question unambiguously.
Although the pH-treated particles loose the ability to hemagglutinate, cell-binding assays indicate that they retain the ability to specifically bind to cells in a SA-independent manner. This is in contrast to the untreated virions, which exhibit SA-dependent cell binding. For SA-dependent rotavirus strains, the initial step in cell entry constitutes interactions with SA-containing receptors. However, cell binding for most rotavirus strains, including human rotavirus, is SA independent (
7). In these strains, the majority of the neutralizing MAbs that recognize VP4 select mutations in VP5* (
27,
28,
38), suggesting that cell entry is mediated mainly by VP5*. Thus, a likely interpretation is that the altered spike represents a transitional state geared to interact with one of the downstream receptors after SA attachment in the multistep cell entry process of rotavirus. Possible involvement of VP7 in the cell binding of pH-treated particles is ruled out because the cell binding is inhibited by the VP5*-specific 2G4-Fab. In its cell attachment, the pH-treated particles appear to resemble the nar3 mutant of RRV (
21,
45). This mutant, which exhibits SA-independent cell binding in contrast to its parental strain, has been shown to attach to the cell surface by interacting with integrin α
2β
1 through the DGE motif in VP5*. In this mutant also, just as in our pH-treated particles, the 2G4 antibody inhibits cell binding (
46). A distinct possibility is that the DGE motif (residues 308 to 310) becomes exposed in the pH-treated particles, and the 2G4-Fab inhibits cell binding of the pH-treated particles by sterically hindering the accessibility of this motif. Although pH treatment exposes a domain on the spike that can interact with a cellular factor, it appears to have destroyed the other determinants required for successful internalization, since pH-treated particles are not infectious.
A remarkable observation made during the course of these experiments is that 2G4, a VP5*-specific neutralizing antibody, can completely protect the spike from pH-induced conformational changes. Equally remarkable is its ability to stay attached to the spikes despite drastic changes in pH (from ca. pH 7 to 11 and back to pH 7). To our knowledge, no other studies have provided a structural demonstration of such a strong antibody-antigen interaction. In general, an antibody-antigen interface is less hydrophobic with relatively more intermolecular hydrogen bonds than found at the protein-protein interface of a homodimer (
24). Because 2G4-VP5* interactions are not sensitive to variations in pH, we can argue that they are predominantly hydrophobic. In this respect, 2G4-VP5* interactions perhaps are an exception to the general trend found in other antibody-antigen interactions. Studies with 2G4 escape mutants indeed show involvement of a hydrophobic region between residues 384 and 401 of VP5* (
35).
The observation that 2G4 protects the structural and functional integrity of the VP4 spike allows us to propose a mechanism of neutralization by this antibody. It is known that 2G4 does not interfere with cell binding but inhibits internalization (
34). We propose that the mechanism by which 2G4 neutralizes rotavirus is by inhibiting a postattachment conformational change that is required for interacting with downstream receptors. This is also consistent with our observation that 2G4 binding to pH-treated particles inhibits cell binding. The ability of a certain antibody to prevent conformational changes necessary for effective internalization of a virus has indeed been demonstrated in yet another case. An antibody that binds to membrane-distal domains of the influenza virus hemagglutinin can prevent the low-pH transition required for fusion activity (
2).
Despite the lack of a reverse genetics system for rotavirus, which would have been ideal to dissect the role of VP4 in cell entry process, our “structural mutagenesis” strategy has provided useful insights into conformational properties of VP4 in relation to the rotavirus cell entry process. A hypothesis that emerges from the various observations we have made is that high-pH treatment induces a conformational change in the VP4 spikes that mimics a post-SA-attachment step in the rotavirus cell entry process, which involves sequential interactions with multiple receptors. This conformational change is inhibited by 2G4 MAb to neutralize the virus.