Atomic resolution structural characterization of recognition of histo-blood group antigens by Norwalk virus

Abstract

Members of Norovirus, a genus in the family Caliciviridae, are causative agents of epidemic diarrhea in humans. Susceptibility to several noroviruses is linked to human histo-blood type, and its determinant histo-blood group antigens (HBGAs) are regarded as receptors for these viruses. Specificity for these carbohydrates is strain-dependent. Norwalk virus (NV) is the prototype genogroup I norovirus that specifically recognizes A- and H-type HBGA, in contrast to genogroup II noroviruses that exhibit a more diverse HBGA binding pattern. To understand the structural basis for how HBGAs interact with the NV capsid protein, and how the specificity is achieved, we carried out x-ray crystallographic analysis of the capsid protein domain by itself and in complex with A- and H-type HBGA at a resolution of approximately 1.4 A. Despite differences in their carbohydrate sequence and linkage, both HBGAs bind to the same surface-exposed site in the capsid protein and project outward from the capsid surface, substantiating their possible role in initiating cell attachment. Precisely juxtaposed polar side chains that engage the sugar hydroxyls in a cooperative hydrogen bonding and a His/Trp pair involved in a cation-pi interaction contribute to selective and specific recognition of A- and H-type HBGAs. This unique binding epitope, confirmed by mutational analysis, is highly conserved, but only in the genogroup I noroviruses, suggesting that a mechanism by which noroviruses infect broader human populations is by evolving different sites with altered HBGA specificities.

Fig. 1.

P domain dimer has the same conformation as in the whole capsid structure. (a) The T=3 capsid structure (showing only the backbone atoms) of NV, determined to 3.4-Å resolution (8), formed by 90 dimers of the capsid protein VP1. Shown are the S domain (amino acids 1–225) and the P1 (amino acids 225–278 and 406–519) and P2 (amino acids 279–405) subdomains (shown in blue, red, and yellow, respectively. (b) A ribbon representation of the VP1 dimer extracted from the capsid structure with S, P1, and P2 domains colored as in a. (c) The structure of the P domain construct (amino acids 225–519) determined to 1.4-Å resolution in the present studies using molecular replacement techniques is identical to that of the P domain in the capsid. The P1 and P2 subdomains are colored as in b. The N and C termini in b and c are denoted. The location of the carbohydrate binding site (see Fig. 2) on one of the dimeric subunits is shown here for reference (rectangular box).

Fig. 2.

NV P domain–HBGA interactions. (a) The electron density (1σ level) of the bound pentasaccharide corresponding to H-type 1 HBGA in the 2F_o − F_c map obtained after MR and before any refinement clearly showed all of the five sugar (labeled) residues (see Fig. S2 for the electron density of the bound A-type trisaccharide). (b) Structural characteristics of the HBGA binding site in the unliganded P domain with four bound water molecules that are displaced by the ligand. The P domain residues, in stick representation, that are involved in hydrogen-bonding interactions with these water molecules are shown. The electron density of the H-type 1 (as a transparent surface) is shown for reference. (c and d) Molecular interactions at the ligand binding site between the P domain residues, in stick representations, and H-type 1 pentasaccharide (c) and A-type trisaccharide (d). The hydrogen-bond interactions are shown in black dashed lines (see Figs. S4 and S5 for more details). The face-to-face stacking interaction (distance = 3.8 Å) between Trp-375 and His-329 can be seen. The hydrophobic interactions involving the α-Fuc in the H-type 1, and the acetamido group of the GlnNAc in the A-type with Trp side chain are indicated by red-filled circles in c and d, respectively. The participating water molecules are labeled as “W.” The glycosidic torsion angles in both of the oligosaccharides are in the energetically preferred region (see Fig. S3).

Fig. 3.

HBGA binding sites in GI and GII NoVs are different. (a) Structure-based sequence comparison of the NV (GI) and VA387 (GII) noroviruses, first two rows, in the P2 domain showing that carbohydrate binding amino acid residues are not conserved in these two groups. The alignment of representative GI NoV sequences (with accession numbers) from each subgroup, in the vicinity of the HBGA binding site, shows the conservation of the residues involved in carbohydrate binding. The β-Gal- and α-Fuc-interacting amino acid residues are denoted by red circles and blue diamonds, respectively. Strictly conserved residues, which include Trp-375 (green square) and His-329, are shaded in blue (least conserved residues in red). (b) Superposition of the P domain structures (Cα-trace) of NV (present study) in green color with H-type 1 carbohydrate, and the VA387 (in yellow) with A-type carbohydrate (21) showing that HBGA binding sites in GI and GII NoVs are distinct. (c) Relative locations of the HBGA biding sites in GI (red circle) and GII (blue circle) mapped on the top surface of the NV P domain dimer. In one of the dimeric subunits, conserved residues are represented in shades of blue as in a. The other dimeric subunit is shown in yellow with twofold related GI (red) and GII (blue) HBGA binding sites denoted by broken circles.

Fig. 4.

Norwalk virus-like particle (VLP) binding to immobilized H-type 1, A, and B trisaccharide carbohydrates detected by surface plasmon resonance. Wild-type NV VLPs bound H-type 1 and A but not B carbohydrates. Point mutant NV VLPs H329A and W375A maintain structural integrity (Inset, electron micrographs of negatively stained mutant VLPs) but lose the ability to bind H-type 1 and A carbohydrates. Multivalent H-type 1, A, and B polyacrylamide-biotin were immobilized to streptavidin-coated Biacore sensor chip SA flow cells. Wild-type, H329A, and W375A NV VLPs were injected through the Biacore sensor chip SA flow cells at 100 μg/ml, 5 μl/min, for 20 min, at 25°C and pH 6.0, and binding was detected as increasing relative response units (RU).