Structural basis of GM1 ganglioside recognition by simian virus 40

Affinity of Binding.

The structure of the complex between SV40 VP1 and GM1, presented here at 2.25-Å resolution, shows that the virus recognizes its ganglioside receptor via a complex network of interactions. Isothermal titration calorimetry reveals a relatively low affinity of one binding site on VP1 for its ligand. However, affinities in the millimolar range have been observed for several other viral attachment proteins that interact with oligosaccharide receptors, such as Polyoma VP1 or influenza hemagglutinin (18, 21). A lower binding affinity is thought to aid virus spread by facilitating release from the cell surface after lysis. Moreover, the affinity of intact SV40 particles for GM1 would likely be much higher, because many binding sites could be engaged simultaneously.

Ligand Specificity.

Glycan screening demonstrates that binding to GM1 occurs with narrow specificity. Oligosaccharides that contain similar structural motifs in different contexts do not interact with SV40 VP1. Analysis of the crystal structure confirms these findings: addition of carbohydrate residues to either one of the branches would lead to steric clashes with protein residues. Both branches of GM1 are required for interaction with VP1. NMR studies show that GM1 assumes one dominant conformation in solution (20). The GalNAc-(β1,4)-[NeuNAc-(α2,3)]-Gal portion of the oligosaccharide forms a rigid entity because of close contacts between the GalNAc and NeuNAc moieties, which are attached to neighboring carbons on the Gal. This arrangement severely limits the rotational freedom of all three sugar residues. By contrast, unbranched oligosaccharides are more likely to adopt different conformations. This is the case, for example, for the linear α2,3-sialyllactose, whose conformation has been studied both in solution and complexed to different protein ligands (17, 22, 23). Because the conformation of GM1 in the VP1 complex is very similar to its principal structure in solution, we think it likely that interactions of one branch facilitate interactions of the other branch with VP1 residues, and that this bivalent binding mode is a foundation for the high specificity of SV40 for GM1.

Glycan array screening also detected binding signals for two glycoproteins. However, we consider these to be nonspecific because of infection studies with Vibrio cholerae neuraminidase-treated cells (24). Although the NeuNAc in GM1 is protected from cleavage by this neuraminidase (25), the terminal sialic acids on glycoproteins are cleaved. Because neuraminidase treatment did not abolish infection, the glycans on glycoproteins are probably dispensable for binding.

Interaction of SV40 with Its Receptor in Simians, NeuNGc-GM1.

The glycan screen did not include more subtly modified GM1 analogues, such as those containing substituted neuraminic acids. However, it was recently shown that SV40 binds to NeuNGc-GM1 more strongly than to NeuNAc-GM1, indicating that NeuNGc-GM1 is the natural receptor of this virus (16).

The N-acetyl group of NeuNAc faces toward a deep oval cavity, which is formed by the BC1-loop of one subunit and the BC2-loop of its clockwise neighbor (Fig. 3B). The cavity is lined by both polar and hydrophobic residues. The NeuNGc variant carries a CH₂–OH group instead of a CH₃ group on the sialic acid amide group. The additional hydroxyl group of the NeuNGc glycolyl chain does not alter the overall conformational properties of GM1 (26), and a GM1 receptor carrying a terminal NeuNGc residue could be accommodated in our structure without requiring any alteration of either the protein or the ligand. The glycolyl moiety would insert further into the cavity, most likely pointing toward the polar residues at its rim. There are two possible energetically favorable conformations of the CH₂–OH group, both of which would result in additional hydrogen bond formation and stabilizing van der Waals interactions. Because these interactions would occur in a partly hydrophobic environment, they would nicely explain the observed stronger interaction between SV40 and NeuNGc-GM1 compared with NeuNAc-GM1 (16). Although it is, of course, possible that NeuNGc-GM1 and NeuNAc-GM1 have totally different modes of binding, we consider this possibility unlikely.

The binding mode for NeuNGc-GM1 postulated here differs from a recently published model of a NeuNGc–GM1 complex with SV40 (16). The published model, which is based on the known structures of SV40 virions and Polyomavirus–receptor complexes (5, 19), makes the assumption that SV40 and Polyoma bind the sialic acid portions of their ganglioside receptors in generally similar orientations. As discussed below, this assumption is not correct.

Comparison with Polyoma VP1–Receptor Interactions.

Polyoma and SV40 VP1 both bind to ganglioside receptors (7), and in both cases the ligands are recognized at similar locations on the outer surface of VP1. However, the actual ligand structures differ: whereas SV40 recognizes a ganglioside that carries a terminal NeuNAc residue on one branch only (GM1), Polyoma binds to compounds that carry either two (GD1a) or three (GT1b) NeuNAc residues distributed over both branches (Fig. 1B). Nevertheless, Polyoma recognizes only the NeuNAc-(α2,3)-Gal structure on the longer branch, in which the NeuNAc is linked to an unbranched Gal (7, 17⇓–19).

In both SV40 and Polyoma, a terminal α2,3-linked NeuNAc serves as a major contact point, and in both cases, VP1 residues at equivalent locations are used for interactions with NeuNAc (Fig. 4 A and B). Despite these similarities, the NeuNAc moiety is bound in different orientations by the two proteins. In Polyoma, the NeuNAc carboxylate faces away from the fivefold axis of the pentamer, forming a key salt bridge with Arg-77, and the glycerol side chain points away from the virion into solution (Fig. 4 B and D). By contrast, the glycerol chain of NeuNAc in GM1 faces toward VP1, whereas its carboxylate group faces toward the fivefold axis and does not engage in a salt bridge (Fig. 4 A and C). Lys-67, which is the SV40 residue equivalent to Arg-77 in Polyoma, does not directly contact NeuNAc but instead forms the ridge that separates the binding pockets for the two branches of GM1. The conformation of Arg-77 in Polyoma is stabilized by Gln-59 and Tyr-72, whereas Lys-67 in SV40 is held in place by a salt bridge to Asp-81. The conformation of NeuNAc and Gal in the Polyoma binding site does not support a branch at the axial O4 of the Gal residue, whereas this branch and the rigid conformation it imposes are recognized by SV40. Thus, residues at equivalent positions in the sequence and in space have different contacts in SV40 and Polyoma, leading to different surfaces and ligand specificities.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

SV40 and Polyoma bind their ganglioside receptors in different orientations and conformations. (A and B) Comparison of specific interactions between VP1 from SV40 (A) and Polyoma (B) with their ligands. Polyoma was crystallized with an oligosaccharide that contained a fragment also present in GD1a and GT1b (shown in color). The additional α2,6-linked NeuNAc present in the ligand, but not on a ganglioside receptor, is shown in gray. The same view was used for both proteins, and residues at equivalent positions in both proteins have the same color. The Arg-Gly motif of Polyoma and its SV40 counterpart are colored yellow, and the residues holding them in place are shown in green. Hydrophobic residues lining the deep cavity in SV40 and their charged Polyoma equivalents are turquoise. Additional residues contacting NeuNAc in both complexes are gray. (C and D) Different binding surfaces of SV40 (C) and Polyoma VP1 (D) for their ligands. The same view from top of the pentamer was used for both complexes, and the color scheme is as in A and B. The terminal Glc was omitted from the SV40 VP1-GM1 complex for clarity. (E) Alignment of the VP1 sequences of SV40, Polyoma, BKV, and JCV. A sequence alignment of SV40, BKV, and JCV VP1 was performed with T-Coffee (www.ebi.ac.uk/t-coffee), then the aligned sequences were arranged according to a structure-based alignment of SV40 and Polyoma (18). SV40 residues identical to both BKV and JCV are marked with an asterisk; those identical to either BKV or JCV are marked with a dot. The color scheme is as in A and B, with additional SV40 residues in contact with GM1 in gray.

Implications for Receptor Binding of Human BKV and JCV.

Both JCV and BKV attach to sialic acids (27, 28), and their VP1 proteins are highly homologous in sequence to SV40 VP1. Although the Polyoma and SV40 VP1 surface loops differ in length and sequence, the surface loops of BKV, JCV and SV40 VP1 all have the same length (Fig. 4E). Furthermore, 6 and 7 of the 10 NeuNAc-contacting residues of SV40 are identical in BKV and JCV, respectively, whereas none are identical in Polyoma. This level of conservation suggests that SV40 can serve as a model for BKV and JCV, and that the three proteins will have similar structures for their receptor-binding sites. We therefore propose that the human polyomaviruses bind NeuNAc in a similar position and orientation as SV40. Furthermore, studies with cells expressing modified sialic acids showed that the chain attached to the amide nitrogen of neuraminic acids contributes to BKV attachment and suggest that it inserts into a pocket on VP1 (29). Sequence conservation suggests that JCV and BKV contain cavities similar to the one in SV40.

Although JCV can bind sialic acids on several glycolipids, only GT1b can inhibit infection (27). In addition, JCV attachment and entry into human glial cells proceed via engagement of serotonin receptors (30). The closely related BKV can use gangliosides GD1b and GT1b as receptors (6). Both GD1b and GT1b share the rigid GalNAc-(β1,4)-[NeuNAc-(α2,3)]-Gal oligosaccharide core structure of GM1 (Fig. 1B), and this part of the carbohydrate has a structure similar to GM1 (31). Both GD1b and GT1b carry an extra NeuNAc residue attached via an α2,8-linkage to the glycerol chain of the NeuNAc found in GM1. Adding a second NeuNAc in this manner to the NeuNAc bound by SV40 would not be possible due to steric clashes.

Comparison with Cholera Toxin.

Very few infectious agents enter cells via lipid rafts and the ER. Bacterial toxins such as cholera toxin or enterotoxin are known to use this pathway, and they share some interesting properties with polyomaviruses. Both systems consist of a pentameric structure (VP1 in polyomaviruses, the B5 pentamer in toxins) that binds a monomeric protein (VP2 in polyomaviruses, protein A in toxins) at its center (32, 33); thus they both exhibit an unusual AB₅ symmetry mismatch (SI Fig. 6). Furthermore, both systems are thought to undergo a conformational change during entry that exposes the central monomeric protein and leads to membrane penetration. Finally, both systems use gangliosides as functional receptors. In fact, the receptor for cholera toxin is GM1, and a comparison of the cholera toxin B protein with SV40 shows that both bind GM1 in essentially the same conformation (SI Fig. 6), which is the principal conformation in solution (20, 32). Despite these similarities, there is no structural conservation of the individual interactions (SI Fig. 6 B and D), and cholera toxin has a much higher affinity for GM1 (34). However, the many structural and functional similarities between the two systems suggest a distant link in evolution.