N-Glycolyl GM1 Ganglioside as a Receptor for Simian Virus 40

Abstract

Carbohydrate microarrays have emerged as powerful tools in analyses of microbe-host interactions. Using a microarray with 190 sequence-defined oligosaccharides in the form of natural glycolipids and neoglycolipids representative of diverse mammalian glycans, we examined interactions of simian virus 40 (SV40) with potential carbohydrate receptors. While the results confirmed the high specificity of SV40 for the ganglioside GM1, they also revealed that N-glycolyl GM1 ganglioside [GM1(Gc)], which is characteristic of simian species and many other nonhuman mammals, is a better ligand than the N-acetyl analog [GM1(Ac)] found in mammals, including humans. After supplementing glycolipid-deficient GM95 cells with GM1(Ac) and GM1(Gc) gangliosides and the corresponding neoglycolipids with phosphatidylethanolamine lipid groups, it was found that GM1(Gc) analogs conferred better virus binding and infectivity. Moreover, we visualized the interaction of NeuGc with VP1 protein of SV40 by molecular modeling and identified a conformation for GM1(Gc) ganglioside in complex with the virus VP1 pentamer that is compatible with its presentation as a membrane receptor. Our results open the way not only to detailed studies of SV40 infection in relation to receptor expression in host cells but also to the monitoring of changes that may occur with time in receptor usage by the virus.

Simian virus 40 (SV40), a nonenveloped DNA virus of the Polyomaviridae family, is one of the most extensively studied animal viruses with respect to transcription, DNA replication, oncogenic transformation, and cell entry (16, 25, 51, 57, 72). It enters animal cells by caveolar/lipid raft-mediated endocytosis and proceeds via organelles termed caveosomes to the endoplasmic reticulum, from where it is released into the cytosol to enter the nucleus through nuclear pores (3, 17, 34, 44, 50, 60).

Like the closely related murine polyomavirus (Py), SV40 has three structural proteins, of which VP1, which forms the icosahedral capsid, is present as 72 homopentamers arranged in T = 7 geometry (38, 65). VP2 and VP3 are minor structural proteins enclosed within the capsid. In recombinant protein expression systems, VP1 alone can assemble into virus-like particles (VLPs) that retain many properties of the infectious virus. They have the same size and icosahedral structure, they bind to the same receptors, and they are endocytosed by host cells (36, 45, 53).

SV40, Py, and the human BK viruses are unusual among viruses in that they use glycolipids as their major cell surface receptors. SV40 uses ganglioside GM1, Py uses GD1a and GT1b, and BK virus uses GD1b and GT1b (39, 58, 68). Crystal structures of Py in complex with oligosaccharide analogs of GD1a and GT1b have illuminated the mode of binding. The disaccharide sequence N-acetylneuraminic acid (NeuAc) α2,3 linked to galactose (Gal), which is common to GD1a and GT1b, binds to a shallow pocket in the exposed surface of VP1 (61, 62). It has been shown that the ganglioside GM1 is required for both attachment and infectivity of SV40 (68), but the structural basis for this binding and specificity remain to be clarified. Binding of SV40 and Py to the cell surface gangliosides is followed by inclusion of the virus in lipid raft domains, transient lateral mobility along the membrane, and induction of tyrosine kinase-mediated signaling, which eventually results in endocytosis of the virus via caveolae or caveolin-free vesicles (17, 18, 20, 51, 52). The gangliosides have an important role in the transbilayer signaling needed to trigger endocytosis (H. Ewers, K. Bacia, W. Chai, G. Schwarzmann, T. Feizi, P. Schwille, A. Helenius, and A. Smith, submitted for publication).

Here we have examined the specificity of binding of SV40 VLPs using carbohydrate microarrays with sequence-defined and lipid-linked oligosaccharide probes that include natural glycolipids and neoglycolipids (NGLs) (22). Having found that the N-glycolylneuraminic acid (NeuGc) analog of the ganglioside GM1 [GM1(Gc)] gave markedly stronger binding signals than the NeuAc analog [GM1(Ac)], we investigated the relative activities of GM1(Gc) and GM1(Ac) as receptors that mediate SV40 infection. In addition, using crystal structures of the homologous VP1 proteins of SV40 and Py and, as a model, the cocrystal structure of Py with 3′sialyllactose, NeuAcα2,3Galβ1,4Glc (PDB code 1SID), we carried out molecular modeling of the VP1/GM1(Gc) complex in order to determine the molecular basis of the favorable interaction of the N-glycolyl analog of GM1 ganglioside with the pentameric VP1 protein.

MATERIALS AND METHODS

Preparation of ganglioside GM1 and its N-glycolyl analog.

Monosialogangliosides were separated from total bovine brain ganglioside mixture by DEAE-Sepharose chromatography, and GM1(Ac) (Fig. 1A) was purified from the monosialoganglioside fraction by silica gel chromatography (41). GM1(Gc) (Fig. 1C) was prepared by exchanging the N-acetyl for N-glycolyl of the sialic acid residue of GM1(Ac) by the following chemical reactions.

FIG. 1.

Structures of the GM1(Ac) and GM1(Gc) gangliosides and corresponding NGLs used in this study. The two major molecular species in each ganglioside are illustrated (A and C), and the ratios of the two ceramides, 100:40 and 60:100, respectively, deduced from MS analyses are shown. The corresponding NGLs (B and D) have the same lipid moiety, DHPE.

GM1(Ac) was deacetylated by alkaline hydrolysis as described previously (69). In brief, the ganglioside was treated with 4 M KOH at 90°C under N₂ for 3.5 h and neutralized with 6 M HCl. After dialysis and lyophilization, the de-N-acetylated GM1 product was purified to over 99% by silica gel column chromatography. Glycolylation was carried out with glycolic acid. Dicyclohexylcarbodiimmide and N-hydroxysuccinimide, both at a final concentration of 0.2 M, were added to a 0.17 M solution of glycolic acid in dry dimethylformamide. After stirring for 30 min at ambient temperature, the de-N-acetyl-GM1 was added to give a final concentration of 0.02 M. The reaction was allowed to proceed at 60°C under vigorous stirring for 1 h. The reaction mixture was cooled to ambient temperature and centrifuged to remove the precipitate formed. Dimethylformamide was evaporated with a rotary evaporator after repeatedly adding a heptane-toluene mixture (2:1 by volume). GM1(Gc) was purified by silica gel chromatography using as the solvent system CHCl₃-CH₃OH-5 M NH₄OH (60:35:5 by volume).

Preparation of pentasaccharides of GM1(Ac) and GM1(Gc) and their neoglycolipids.

The bovine brain monosialoganglioside fraction was dissolved in methanol (5 mg/ml) and subjected to ozonolysis as described previously (26) to cleave the ceramide moiety. Triethanolamine was added to adjust the pH to 10, and the reaction mixture was kept at ambient temperature for 24 h. The released oligosaccharides were isolated by preparative thin-layer chromatography using the solvent system CHCl₃-CH₃OH-50 mM KCl (20:50:13 by volume). The oligosaccharide solution was evaporated to dryness; the residue was suspended in CH₂Cl₂ and extracted three times with water. The aqueous phases were combined and lyophilized. The oligosaccharides were subfractionated by high-pressure liquid chromatography (HPLC) on an amide column (TSK Gel Amide-80). Elution was performed with CH₃CN-H₂O (70:30 by volume) (solvent A) and CH₃CN-H₂O (20:80 by volume) (solvent B), both containing 15 mM KH₂PO₄. A linear gradient of 5 to 15% solvent B was applied for 40 min at a flow rate of 1 ml/min, and the fractions were monitored at 196 nm (Fig. 2). Pooled fractions were desalted twice on a column of Sephadex G10 (1.6 by 36 cm) eluted with water at a flow rate of 20 ml/h. The fractions were analyzed by negative-ion electrospray mass spectrometry (ES-MS) (Fig. 3) to identify the GM1(Ac) and GM1(Gc) pentasaccharides; these were quantified by orcinol assay and lyophilized.

FIG. 2.

Separation of GM1 and GM2 Oligosaccharides by HPLC. Oligosaccharides released by ozonolysis from the monosialoganglioside fraction of bovine brain were fractionated by HPLC on an amide column (TSK Gel amide-80), and the oligosaccharides were identified by ES-MS. Peaks 5 and 6 were identified as GM1(Ac) and GM1(Gc) pentasaccharides, respectively (Fig. 3) and peak 2 as GM2 tetrasaccharide (not shown).

FIG. 3.

Collision-induced dissociation product ion mass spectra of GM1(Ac) and GM1(Gc) pentasaccharides. The molecular ions (M − H)⁻ at m/z 997 and 1013 in panels A and B differ by 16 Da and correspond to those for the GM1(Ac) and GM1(Gc) pentasaccharides, respectively (N-glycolyl is 16 Da larger than N-acetyl). The fragment ions at m/z 290 (A) and 306 (B) are those of Neu5Ac and Neu5Gc, respectively. The fingerprints of other fragment ions from the two oligosaccharides are similar. Fragment ions in the lower-mass region (e.g., m/z <700) derived from nonreducing termini without the sialic acids are identical. Fragment ions in the higher-mass region, containing sialic acid, have mass differences of 16 Da, corresponding to the difference in the N substitution of the sialic acid residues; cf. m/z 773 in panel A with m/z 789 in panel B and m/z 833 in panel A with m/z 849 in panel B.

NGLs of the GM1(Ac) and GM1(Gc) pentasaccharides (Fig. 1B and D) were prepared and quantified as described previously (14). In brief, to 50 nmol of lyophilized pentasaccharide were added water (3 μl), 1,2-dihexadecyl-sn-glycero-3-phosphoethanolamine (DHPE) (50 μl; 5 mg/ml in CHCl₃-methanol [1:1 by volume]), and freshly prepared tetrabutylammonium cyanoborohydride solution (10 μl; 20 μg/μl in methanol). The reaction mixture was incubated at 60°C for 24 h. The NGL products were analyzed by high-performance thin-layer chromatography and purified using silica cartridges (Waters Sep-Pak; 500 mg silica), and their molecular masses were corroborated by ES-MS.

ES-MS.

ES-MS analyses of the gangliosides GM1(Ac) and GM1(Gc) were carried out on both ThermoQuest LCQ_Deca (Finnigan Mat, San Jose, CA) and Q-TOF (Micromass, Manchester, United Kingdom) instruments. Detailed ceramide analysis was as described previously (70). ES-MS analyses of the pentasaccharides and derived NGLs were carried out using the Q-TOF mass spectrometer. The fragmentation fingerprints of the pentasaccharides and gangliosides were obtained by collision-induced dissociation and product ion scanning (CID-MS/MS) (13). Conditions of MS were as follows. For the analyses of oligosaccharides, nitrogen was used as desolvation and nebulizer gas at flow rates of 250 liters/h and 15 liters/h, respectively. The source temperature was 80°C and the desolvation temperature 150°C. The capillary voltage was maintained at 3 kV and the cone voltage at 50 V. Product ion spectra were obtained from CID with argon as the collision gas at a pressure of 1.7 × 10⁵ Pa. For the analyses of gangliosides, ionization was performed under the following conditions: spray voltage, 4 kV; sheath gas flow rate, 50 arbitrary units; capillary voltage, −42 V. The collision energies for fragmentation were 62 V for the gangliosides and 45 V for the pentasaccharides. For the analyses, the oligosaccharides were dissolved in CH₃CN-2 mM NH₄HCO₃ (1:1) and the gangliosides in CHCl₃-methanol-H₂O (25:25:8) at a concentration of 25 pmol/ml, 5 μl of which was loop injected. The respective solvents were also used as the mobile phase and delivered with a Harvard syringe pump (Harvard Apparatus, Holliston, MA) at a flow rate of 5 μl/min.

Microarray analyses.

Sequence-defined, lipid-linked oligosaccharide probes were printed on nitrocellulose-coated glass slides (FAST slides; Whatman Ltd.) with Cy3 dye included in the array fluid to enable postarray monitoring of the spots (48). One hundred ninety lipid-linked oligosaccharide probes (see Table S1 in the supplemental material) were arrayed on FAST slides at 2 and 7 fmol per spot, in duplicate, using a noncontact arrayer (Piezorray; Perkin-Elmer LAS, United Kingdom); these were in the forms of natural and synthetic glycolipids and NGLs derived from natural and chemically synthesized oligosaccharides. A total of 201 oligosaccharide positions are indicated in Table S1 in the supplemental material, as some probes were arrayed at more than one position. Binding analysis using biotinylated SV40 VLPs (5 μg/ml) was performed essentially as described previously (48).

In brief, the arrayed slides were overlaid initially for 1 h with blocking solution (5 mM HEPES [pH 7], 150 mM NaCl [HBS] with 3% [wt/vol] bovine serum albumin [Sigma]). The slides were rinsed with HBS and overlaid for 1.5 h with biotinylated SV40 VLPs (5 μg/ml) in blocking solution. The overlaid slides were washed with HBS, and binding was detected using Alexa Fluor 647-labeled streptavidin (Molecular Probes) at 1 μg/ml in blocking solution. After extensive washing with HBS followed by additional washes with distilled water, the slides were dried and scanned using a ProScanArray (Perkin-Elmer LAS), and Alexa Fluor 647-binding signals were quantified using ScanArrayExpress software (Perkin-Elmer LAS). The binding signals were carbohydrate dose related. Microarray data analysis and presentation were carried out using in-house software (M. S. Stoll, unpublished). The results are presented as means of fluorescence intensities of duplicate spots after background subtraction. The error bars represent half of the difference between the two values.

SV40 purification.

The protocols for virus purification were based on those described previously (54). Monkey kidney CV-1 cells were cultured in complete medium, consisting of Dulbecco's modified Eagle's medium (DMEM) (GIBCO) supplemented with 10% fetal calf serum (FCS) (LabForce, Nunningen, Switzerland), 4 mM GlutaMAX, and 50 mM HEPES (GIBCO), at 37°C with 5% CO₂. Forty T175 flasks of subconfluent CV-1 cells were infected with SV40 at a multiplicity of infection of 0.01. Cells were cultured in complete medium for 14 days. To harvest virus, cells were put through three freeze-thaw cycles and then centrifuged at 10,000 × g for 10 min at 4°C. Twenty milliliters of virus-containing supernatant was loaded on a 10-ml cushion of CsCl (1.4 g/ml) in 10 mM HEPES, pH 7.4. Following centrifugation at 76,000 × g for 3 h at 4°C in an SW28 rotor (Beckman), the banded virus in the CsCl cushion was harvested. The density of the CsCl fraction containing SV40 was checked, and the fraction was adjusted to a CsCl density of 1.34 g/ml in 10 mM HEPES, pH 7.4. Following equilibrium centrifugation at 100,000 × g for 16 h at 4°C in a 70.1 Ti rotor (Beckman), the lower virus band was isolated and dialyzed against 50 mM HEPES (pH 8.0)-150 mM NaCl-1 mM CaCl₂ (virus buffer). The purified infectious virus was stored in aliquots at −80°C.

VLP purification.

SV40 VLPs were generated as described previously (37). A lysate of SF9 cells containing expressed SV40 VP1 was a kind gift from Ariella Oppenheim (Department of Haematology, Hebrew University-Hadassah Medical School and Hadassah Hospital, Jerusalem). The lysate contains fully assembled SV40 VLPs as well as assembly intermediates and unassembled VP1 protein. To purify the fully assembled VLPs, the lysate was clarified by centrifugation for 30 min at 10,000 × g in an Eppendorf microcentrifuge using a Beckman SW41Ti rotor. The clarified supernatant (0.5 ml) was centrifuged at 4°C for 2.5 h at 160,000 × g through a 5 to 20% (wt/vol) linear sucrose gradient with a 0.5-ml 60% (wt/vol) sucrose cushion in 10 mM HEPES (pH 7.4) containing 1 mM CaCl₂. After fractionation, 0.5-μl fractions were analyzed by transmission electron microscopy following negative staining, and fractions with homogenous intact particle populations were pooled. The purified particles were dialyzed against virus buffer and stored at −80°C.

Biotin and Cy5 labeling of SV40 VLPs.

Biotin-succimidyl ester or Cy5 (Amersham Biosciences) was covalently coupled to SV40 VLPs in virus buffer using a 10-fold molar excess of the biotin or dye relative to VP1 protein, resulting in fewer than 200 fluorophore or biotin molecules per VLP as determined according to the manufacturer's instructions. Unbound dye was removed by chromatography with a Nap-5 column (Amersham Biosciences). The modified SV40 VLPs were able to bind, enter, and traffic within CV-1 cells to the same extent as nonmodified infectious SV40, as described previously (50) (data not shown).

Analyses of SV40 binding to glycolipid-deficient murine GM95 cells and infection after ganglioside and NGL supplementation.

The glycolipid-deficient murine B16 melanoma GM95 cells (28) and murine 3T6 Swiss albino fibroblast cells were cultured in phenol red-free (PRF) complete medium, consisting of DMEM containing 10% FCS, 4 mM GlutaMAX, and 50 mM HEPES, at 37°C with 5% CO₂.

(i) Supplementation of GM95 cells.

Supplementation of the glycolipid-deficient cells was performed by methods described previously (58) and adapted from reference 21. In brief, GM95 cells were grown in Lab-tek eight-well coverglass chambers for 2 days. Cells were washed twice with serum-free PRF culture medium, consisting of PRF-DMEM (GIBCO) containing 4 mM GlutaMAX and 50 mM HEPES (pH 7.4), and incubated at 4°C for 1 h in 200 μl of serum-free PRF culture medium in the presence of 0.5, 1, or 2 nmol of the GM1(Ac) or GM1(Gc) gangliosides or NGLs. As a negative control, the ganglioside GD1a (U.S. Biologicals) was added. Cells were then cultured at 37°C with 5% CO₂ for 48 h to allow uptake and incorporation of the glycolipids. To remove unincorporated glycolipids, the cells were then cooled to 4°C and washed twice with PRF-DMEM supplemented with 25% FCS, 4 mM GlutaMAX, and 50 mM HEPES (pH 7.4) (lipid wash medium). Cells were then maintained at 4°C in lipid wash medium for 60 min. After washing twice with serum-free medium at 4°C, the cells were ready for binding or infection assays.

(ii) Analysis of SV40 binding.

For SV40 binding assays, wells with approximately 5 × 10⁵ cells were selected. Cells were incubated with 200 μl of 0.5-μg/ml Cy5-labeled SV40 VLPs for 1 h at 4°C. Cells were then washed with serum-free PRF culture medium and fixed in 4% (wt/vol) paraformaldehyde in phosphate-buffered saline. Confocal images were acquired with an inverted Zeiss LSM 510 meta (100× objective; numerical aperture, 1.4) under identical acquisition conditions for each experiment. Micrographs were exported to Image J (NIH), and total fluorescence of single cells was quantified, normalized to the values for GM1(Ac)-supplemented cells, and graphed. The results were found to be independent of the pinhole settings. Accordingly, images acquired with an open pinhole were used for quantification, and images acquired at one Airy unit are presented for clarity.

(iii) Analysis of SV40 infection.

For infection assays, cells were inoculated with SV40 at a multiplicity of infection of 178 (as judged by plaque assay in reference host CV1 cells) in 200 μl serum-free PRF culture medium or incubated with culture medium for 2 h at 37°C. Unbound virus was removed, and cells were maintained in PRF culture medium at 37°C with 5% CO₂ for an additional 18 h. After fixation with 4% paraformaldehyde, infected cells were identified by immunodetection of SV40 T-antigen expression (50) by epifluorescence microscopy. Images were acquired on a Zeiss Axiovert microscope with an Achroplan lens (10× objective; numerical aperture, 0.25) with a charge-coupled device camera (Hamamatsu Inc.) using Openlab (Improvision). After export as 8-bit TIFFs, images were processed identically using Photoshop (Adobe). The infected cells were counted, and graphs were generated in Excel (Microsoft). With the virus load used, 72% of 3T6 cells were infected (data not shown). The cell infection rate in the glycolipid-supplemented GM95 cell line, being comparatively low (less than 30%), was scored as an “infection index” normalized as 1 arbitrary unit (AU) for cells supplemented with 1.0 nmol GM(Ac) in 200 μl (5 μM). To evaluate the cholesterol or tyrosine kinase dependence of the infection mediated by the GM1(Ac) and GM1(Gc) gangliosides and corresponding NGLs, the infection experiments were carried out with GM95 cells supplemented with 1 nmol of one or the other of the two gangliosides or the two NGLs. For experiments including pharmacological inhibitors, samples were pretreated with 100 μm genistein (Calbiochem) for 30 min to inhibit tyrosine kinase activity or with 10 μg/ml progesterone (Sigma) and 25 μg/ml nystatin (Sigma) for 16 h to reduce cellular cholesterol. Drugs were maintained in the culture medium throughout the course of the experiments.

Molecular modeling of the interaction between GM1(Gc) and SV40 protein VP1. (i) Selection and building of the virus protein model.

The crystal structure of the SV40 VP1 pentamer at 3.1-Å resolution (62) is available from the Protein Data Bank (code 1SVA) (5). Hydrogen atoms were added using the Sybyl software (Tripos Inc., St Louis, MO) together with partial charges from the Pulman libraries. Positions of hydrogen atoms were optimized with the use of the Tripos force field (15). Calculations of Conolly surface and of electrostatic potential were performed with the MOLCAD software (73).

(ii) Docking of the monosaccharides NeuAc and NeuGc into the VP1 protein of SV40.

One monomer (chain A) of the pentamer of VP1 protein of SV40 was selected for the docking study. The structure of one monomer (chain A) of VP1 protein of Py cocrystallized with 3′siallylactose (PDB code 1SID) (64) was structurally aligned on the SV40 VP1 monomer using Sybyl. The resulting amino acid sequence alignment is shown in Fig. 4. The NeuAc residue bound to Py protein was merged into the SV40 structure and graphically modified to a NeuGc residue. After addition of hydrogen atoms, partial charges, and appropriate carbohydrate atom types (29), the structures of both SV40/NeuAc and SV40/NeuGc were optimized by taking into account the flexibility of the side chains. The loop from amino acid 55 to 70 was also optimized, as it participates in the binding site. Addition of hydrogen atoms and calculations of accessible surface were also applied to the complex between Py and NeuAc for comparison.

FIG. 4.

Amino acid sequence alignment of the SV40 and Py VP1 proteins used to align the crystal structures for homology docking of NeuGc into the binding site of SV40. The amino acids that are highly conserved between the two sequences are shaded in gray. The amino acids involved in NeuAc (or NeuGc) binding are highlighted in yellow, and the ones involved in galactose binding in the model of SV40 VP1 complexed with GMA(Gc) are highlighted in blue. The amino acids from neighboring monomers predicted to bind to glucose are highlighted in green.

(iii) Docking of the GM1(Gc) pentasaccharide head group and the GM1(Gc) ganglioside.

The two major conformations that GM1(Ac) pentasaccharide can adopt in solution (8, 10) were taken into account. These two conformations differ mainly in the φ angle at the NeuAcα2,3Gal linkage, as described in several conformational studies (see references 30 and 59 for reviews). Local energy minimization was performed by optimizing the positions of atoms belonging to both oligosaccharide and protein side chains.

An idealized SV40 pentamer was built by applying the fivefold symmetry observed in the crystal to the A monomer complexed with both possible conformations of the GM1(Gc) carbohydrate head group. The lipid moiety was then added onto the reducing-end glucose residue of GM1(Gc), and conformational analysis was performed on the oligosaccharide-lipid linkage.

RESULTS

SV40 binds more efficiently to GM1(Gc) than GM1(Ac).

To gain insights into the range of carbohydrate sequences to which SV40 can bind, we used biotinylated SV40 VLPs to probe microarrays of 190 lipid-linked oligosaccharides, selected from our library of carbohydrate probes, printed at 201 positions (some having been printed more than once). Natural and synthetic glycolipids and NGLs representative of diverse mammalian glycans were included in the microarrays (see Table S1 in the supplemental material). Among them were N-glycans (neutral and sialylated, high-mannose and complex types), O-glycans, blood group-related sequences (A, B, H, Lewis^a, Lewis^b, Lewis^x, and Lewis^y) on linear or branched backbones and their sialylated and/or sulfated analogs, gangliosides, oligosaccharide fragments of glycosaminoglycans, and homo-oligomers of sialic acid. They ranged in size from 2 to 20 monosaccharide residues. SV40 bound most strongly to ganglioside GM1(Gc) (Fig. 5, position 185). Ganglioside GM1(Ac), position 199, which was shown previously to bind SV40, was a considerably weaker binder. Gangliosides GD1a and GT1b, which are known to be bound by Py, were included in the array at positions 111 and 108, respectively; they did not show significant binding. Thus, the binding of SV40 to its receptor gangliosides was highly specific, with GM1(Gc) being the most efficient ligand.

FIG. 5.

The N-glycolyl analog of GM1 [GM1(Gc)] is revealed in screening microarray analysis to be more strongly bound by SV40 than the N-acetyl analog [GM1(Ac)]. Numerical scores are shown for the binding signals (means of duplicate values at 7 fmol/spot, minus background) for oligosaccharide probes at array positions 101 to 201, including GM1(Gc) and GM1(Ac), which gave binding signals with SV40 VLPs, and GD1a and GT1b, which are known ligands for Py but gave no binding. Error bars represent half of the difference between duplicate values.

The GM1 analogs bear the same lipid moieties but in different proportions.

Both the carbohydrate and the lipid moieties contribute to the interactions of proteins with glycolipids. Although the saccharides comprise the actual ligands, the lipid moieties can be important in the presentation (31, 32, 67). To characterize and compare the lipid moieties of the GM1(Ac) and GM1(Gc) gangliosides, we carried out MS analyses. Two major molecular ion species [(M − H)⁻] were detected in both gangliosides (Table 1): m/z 1544 and 1572 in a ratio of 100:40 for GM1(Ac) and m/z 1560 and 1588 in a ratio of 60:100 for GM1(Gc). The 16-Da difference between the two pairs of molecular ions corresponds to the difference between NeuAc and NeuGc.

TABLE 1.

ES-MS and ES-CID-MS/MS analyses of gangliosides GM1(Ac) and GM1(Gc)

Ganglioside	(M − H)−a	Other major fragment ionsb							Ceramide ions, Y0c
		Y2α	Y2β	DY2α-2β	Y1	Z1	B2β	B1α
GM1(Ac)	1,544 (100)	1,253	1,179	888	726	708	364	290	564 (d18:1 18:0)
	1,572 (40)	1,281	1,207	916	754	736	364	290	592 (d20:1 18:0)
GM1(Gc)	1,560 (60)	1,253	1,195	888	726	708	364	306	564 (d18:1 18;0)
	1,588 (100)	1,281	1,223	916	754	736	364	306	592 (d20:1 18:0)

^aNominal mass is given, and relative intensity is shown in parentheses.

^bThe nomenclature for the fragment ions is based on that described in references 12 and 19.

^cSphingosine and fatty acid assignments in parentheses are based on detailed ceramide analysis (Table 2).

MS/MS product ion scanning showed that the two gangliosides contained the same lipid moieties. The molecular ion species at m/z 1544 from GM1(Ac) and m/z 1560 from GM1(Gc) both gave the same ceramide fragment ion at m/z 564. Similarly, the molecular ion species at m/z 1572 from GM1(Ac) and m/z 1588 from GM1(Gc) gave a ceramide ion at m/z 592. The 28-Da difference between the two ceramide ions corresponds to a two-carbon (CH₂CH₂) difference in the lipids. The ions at m/z 564 and m/z 592 represent ceramide comprising sphingosine d18:1 and sphingosine d20:1, respectively, each linked to a C_18:0 fatty acid chain as determined by detailed ceramide analysis (Table 2). Thus, although the two gangliosides had the same lipid moieties, the ratios of the two ceramides (m/z 564 and m/z 592) were different for GM1(Ac) and GM1(Gc), as depicted in Fig. 1.

TABLE 2.

ES-CID-MSⁿ analyses of ceramides from gangliosides GM1(Ac) and GM1(Gc)

Ganglioside	Mass											Ceramide assignment
	Ceramide ions					Fatty acid ionsa			Long-chain base ions
	Y0	—H2O	—CH2O	—CH4O	—CH4O2	A	B	C	D	E − H2O	F
GM1(Ac)	564	546	534	532	516	282	308	324	237	263	268	d18:1 18:0
	592	574	562	560	544	282	308	324	265	291	296	d20:1 18:0
GM1(Gc)	564	546	534	532	516	282	308	324	237	263	268	d18:1 18:0
	592	574	562	560	544	282	308	324	265	291	296	d20:1 18:0

^aNomenclature for the fatty acid and long-chain base fragment ions is as described in reference 42.

To determine the role of the lipid and whether the difference in binding efficiency between the two gangliosides reflected the difference in sphingosine and fatty acid moieties or the N substitution of the sialic acid residue, the pentasaccharide moieties were isolated and converted to NGLs. They were produced by conjugating the oligosaccharides (by reductive amination) to the aminoglycerophospholipid DHPE, which has two C₁₆ alkyl chains (23). In this way, their lipid moieties were identical (Fig. 1).

GM1(Gc) analogs are more effective in supporting SV40 infection than GM1(Ac) analogs.

We compared the activities of the GM1(Gc) and GM1(Ac) gangliosides and their NGL analogs as cell receptors for SV40 binding and infection by performing glycolipid supplementation experiments using the glycolipid-deficient GM95 cell line. Fluorescently labeled VLPs (Cy5-SV40) showed little or no binding to the nonsupplemented GM95 cells or to the GD1a-supplemented cells when analyzed by confocal microscopy. Binding was observed, however, to cells supplemented with GM1(Ac) and GM1(Gc) gangliosides and their NGL analogs (Fig. 6A). Quantification of Cy5-SV40 fluorescence (Fig. 6B) showed that binding was greater to the cells supplemented with the GM1(Gc) ganglioside (1.55 ± 0.18 mean ± standard error of the mean [SEM]) than to those supplemented with the GM1(Ac) ganglioside (1.00 ± 0.12 [mean ± SEM]). There was also a similar difference, albeit less marked, between the two NGL analogs (1.42 ± 0.17 and 1.14 ± 0.16 [mean ± SEM], respectively, Fig. 6B).

FIG. 6.

Enhanced SV40 binding to GM95 cells supplemented with GM1(Gc) gangliosides and NGLs leads to enhanced infection compared with cells supplemented with the GM1(Ac) analog. (A) Binding of Cy5-SV40 VLPs to GM95 cells that were either nonsupplemented or supplemented with the ganglioside GM1(Ac) or GM1(Gc), the corresponding NGLs, or, as a negative control, GD1a, all at 1 nmol per 5 × 10⁵ cells. (B) Quantification of micrographs of Cy5-SV40 VLP fluorescence of GM95 cells treated as for panel A. (C) Detection of SV40 infection (viral T-antigen expression in GM95 cells that were either nonsupplemented or supplemented with gangliosides and NGLs as for panel A). AU, arbitrary units. (D) Quantification of SV40 infection in cells treated as for panel C except that glycolipid supplements were 0.5, 1, or 2 nmol per 5 × 10⁵ cells.(E) Quantification of SV40 infection of drug-treated or untreated GM95 cells supplemented with the ganglioside GM1(Ac) or GM1(Gc) and the corresponding NGLs at 1 nmol per 5 × 10⁵ cells. The indicated cells were treated with the drug genistein (Gen), which inhibits tyrosine kinases, or with a combination of progesterone (an inhibitor of cholesterol biosynthesis) and nystatin (a sterol chelator) (N/P) or with no drug. Prefixes g and n designate gangliosides and NGLs, respectively. Bars in panels A and C, 10 μm. Error bars indicate standard errors of the means for three independent experiments with at least 50 cells in panel B and 200 cells in panels D and E. Color coding for the cell transfectants in panel D is the same as in panels B and E.

GM95 cells nonsupplemented or supplemented with GD1a showed no detectable infection by SV40 when assayed for the expression of T antigen. In contrast, cells supplemented with the GM1(Ac) and GM1(Gc) gangliosides or their NGL analogs were permissive to infection. Results with cells supplemented with 1 nmol of the glycolipids are shown in Fig. 6C. Fourteen percent of cells supplemented with 1 nmol of GM1(Ac) ganglioside were infected (results not shown); this was taken as an infection index of 1 arbitrary unit. The infection indices of the cells supplemented with 0.5, 1.0, and 2.0 nmol of the gangliosides or NGLs are shown in Fig. 6D. Not only the cells supplemented with the GM1(Gc) ganglioside but also those supplemented with the corresponding NGL analog had greater infection indices than cells supplemented with the GM1(Ac) ganglioside and NGL.

As SV40 infection in host monkey kidney cell lines is dependent on tyrosine kinase activity and cell membrane cholesterol (49), we evaluated infection in cells supplemented with the GM(Ac) and GM(Gc) gangliosides as well as the NGLs when cells were treated with the tyrosine kinase inhibitor genistein or with agents that reduce free cellular cholesterol, progesterone, and nystatin. The infection indices of drug-treated cells supplemented with 1.0 nmol of the gangliosides or NGLs are shown in Fig. 6E. There was almost complete inhibition of infection in the presence of the pharmacological inhibitors.

Modeling of the binding site verifies the favorable interaction of the N-glycolyl analog.

To understand the structural basis for the carbohydrate recognition by SV40, we performed molecular modeling of the interaction between NeuAc and NeuGc and SV40 protein VP1. Comparison of the amino acid sequences and the accessible surface of the proteins indicated that the NeuAc binding pocket of Py has its counterpart in SV40 (Fig. 4 and 7). Using homology with the orientation of NeuAc in the structure of the Py VP1/siallylactose complex (64) (Fig. 7B and C), docking was performed by orienting NeuAc and NeuGc in the equivalent binding pocket of the SV40 VP1 monomer (Fig. 7D and 8).

FIG. 7.

Docking of the monosaccharide NeuGc into the VP1 monomer of SV40 and comparison with NeuAc in the VP1 monomer of Py. (A) Architecture of both SV40 and Py capsids as taken from the Viper database (55). (B) Enlargement of the pentamer surface of the VP1 of Py complexed with sialyllactose (64) (PDB code 1SID); only amino acids 1 to 320 are represented for each monomer. (C) Hydrogen bond network of NeuAc in the binding site of VP1 of Py as observed in the crystal structure. (D) Molecular modeling of NeuGc into the binding site of a SV40 monomer. Hydrogen atoms are not displayed for the sake of clarity.

FIG. 8.

Graphical representation of the VP1-sialic acid complexes of Py and SV40 obtained by molecular modeling. The proteins are represented with their Connolly surfaces color coded according to electrostatic potential (from red for positively charged areas to blue for negatively charged areas). (A) VP1of Py complexed with NeuAc (corresponding to terminal NeuAc in the crystal structure of Py VP1 complexed with sialyllactose, PDB code SID). (B) VP1 of SV40 complexed with NeuAc. (C) VP1 of SV40 complexed with NeuGc.

The relevant amino acids are not fully conserved between the two structures, but both binding pockets contain a basic amino acid (Arg77 in Py and Lys67 in SV40) that can establish a salt bridge to the carboxyl group of NeuAc or NeuGc. Hydrogen bonding with O4 in SV40 is provided by Thr276 and Gln278, which replace His298 in Py VP1. In the Py structure, the N-acetyl group is bound by Tyr72 and Asn293 (64), which are replaced by Gln62 and Ser274, respectively, in the SV40 VP1 structure. In addition to this network of salt bridges and hydrogen bonds conserved in the two structures, the N-glycolyl group of NeuGc establishes two additional contacts with the nitrogen atom of the side chains of two residues, Gln62 and Asn272, in SV40 VP1. These two polar amino acids at the edge of the binding site likely favor an interaction with the hydroxymethyl present in the N-glycolyl pendant group rather than with the methyl of the N-acetyl group.

The Gln62 of SV40 VP1 plays a special role, as it permits two strong cooperative hydrogen bonds (NH—O) between its side chain and the N-glycolyl tail (Fig. 7D). Such a role cannot be played by the equivalent Tyr72 in the Py VP1 structure (Fig. 7C). The Asn272 equivalent in the Py VP1 is an isoleucine (Ile290), which would not make any additional contact. It was previously proposed that Py would display high-affinity NeuGc-containing saccharides by establishing hydrogen bonds to Tyr72 and to Asp85 of neighboring monomers (63). We examined this by molecular modeling and observed that the hydroxyl group of Tyr72 is not accessible for the hydroxymethyl of NeuGc. Therefore, only one additional hydrogen bond would be established between Py and NeuGc (compared to Py and NeuAc), whereas two strong hydrogen bonds are established between the hydroxymethyl group of NeuGc and SV40.

Accommodating the GM1(Gc) pentasaccharide in the VP1 monomer and pentamer.

Starting from the anchoring point of the NeuGc monomer described above, the conformational analysis of the GM1(Gc) pentasaccharide head group was performed by considering the main energy minimum of each glycosidic linkage by using the energy maps available at the Internet sites Glyco-3D (http://www.cermav.cnrs.fr/glyco3d) and Glycosciences.de (http://www.dkfz.de/spec/glycosciences.de/). In the GM1(Ac) analog, most of the flexibility originates from the NeuAcα2,3Gal linkage. A similar conformational behavior is predicted for the GM1(Gc) pentasaccharide, because the N-glycolyl pendant group does not have much influence on the conformational behavior of the adjacent glycosidic linkage (9).

GM1 has been found in nuclear magnetic resonance studies to adopt a predominant conformation in solution that is similar to the lower-energy conformation predicted by molecular modeling (1, 8). This also corresponds to the single conformation observed in the complex of the cholera toxin B pentamer with the GM1(Ac) pentasaccharide (43). However, a higher-energy conformation predicted by molecular modeling has also been described recently in nuclear magnetic resonance analyses of the complex of galectin 1 with GM1(Ac) pentasaccharide (56). Both conformations correspond to possible conformations of GM1(Gc), as the glycolyl group is far away from the glycosidic linkage and does not affect the conformational behavior (9).

Starting with the localization of the NeuGc binding site as determined above, the two conformations of the pentasaccharide could be fitted onto the VP1 monomer surface without generating steric conflict. After geometry optimization of the oligosaccharide and the protein side chains, it appeared that in both cases, stabilizing contacts could occur between the protein and the ligand (Fig. 9). Due to the differences in the orientation of the NeuGcα2,3Gal linkage (φ = −38° and ψ = 92° [designated conf_1] and φ = 89° and ψ = 104° [conf_2]), the reducing-end glucose residue points in completely different directions in the two models. Figure 9A displays the docking mode obtained with the conformation of the NeuGcα2,3Gal linkage that corresponds closely to the orientation observed in the GM1/cholera toxin complex (conf_1). The other docking mode (conf_2) is displayed in Fig. 9B.

FIG. 9.

Graphical representation of the two possible conformations of GM1(Gc) pentasaccharide docked on the SV40 VP1 monomer surface. The proteins are represented with their Connolly surfaces color coded according to electrostatic potential (from red for positively charged areas to blue for negatively charged areas). (A) Conf_1; (B) conf_2.

However, in the SV40 capsid the VP1 molecules occur as tightly associated pentamers in which contacts arise between the GM1(Gc) oligosaccharide and a loop in the neighboring VP1 molecule, Thr135-Gly141 (Fig. 10A). When the lipid moiety is added to the reducing end of the oligosaccharide, only the docking model defined for conf_1 enables the lipid moiety to point perpendicularly to the pentamer surface (Fig. 10B). Thus, conf_1 is the only conformation compatible with GM1 presentation both in the microarray binding experiments and at the cell surface. Therefore, in the complex of SV40 with GM1(Gc) ganglioside, the oligosaccharide moiety has to be in the lowest-energy conformation, as observed in the GM1(Ac) pentasaccharide complex with cholera toxin. The model shown in Fig. 10 can be considered compatible with both the requirement of the binding pockets and membrane topology, although it would be necessary to obtain a crystal structure of the complex for defining the details of the interaction.

FIG. 10.

Docking of the GM1(Gc) onto the surface of the SV40 VP1 pentamer. (A) Top view of the SV40 VP1 pentamer with the Connolly surface of the five chains (amino acids 1 to 320) and the five pentasaccharides bound. (B) Side view of the pentamer with five GM1(Gc) ganglioside molecules.

DISCUSSION

Microbes often use host cell carbohydrates as receptors for adhesion and cell entry, and many of these are sialylated oligosaccharide sequences (33, 40, 71). The identification and elucidation of their structures are important for understanding the pathobiology of infection. The advent of carbohydrate microarrays has transformed studies of carbohydrate-protein interactions in glycobiology, leading to high-throughput analyses of biomedically important systems that operate through oligosaccharide recognition (2, 6, 22, 24, 48, 74). Pathogen-host interactions are notable among these. Carbohydrate microarrays not only are well suited to screening large numbers of oligosaccharide probes but also provide access to deeper insight into the spectrum of carbohydrate specificities. They have, for example, revealed that receptors for infective agents may comprise families of sialo-oligosaccharide sequences with particular capping motifs, such as those for the hemagglutinins of influenza viruses (66) or a broad range of motifs recognized by the micronemal protein 1 of Toxoplasma gondii (7). At the other extreme, the receptor structures can be highly restricted, as is apparent from the present investigation of carbohydrate binding by SV40. In either case, the specificities constitute key determinants of host range, tissue tropism, and pathogenicity of the infective agents.

In the microarray system used, the glycans were presented to the VLPs as glycolipids or NGLs. It is likely that this mode of presentation is optimal for detecting multivalent binders such as SV40. The glycans may be oriented in a manner similar to that of glycolipids in a lipid bilayer membrane. Furthermore, the glycolipids may be laterally mobile and therefore able to cluster in a way similar to receptor clustering under the bound virus as occurs in the plasma membrane. Binding of a single NeuAc-Gal-Glc trisaccharide to the VP1 of Py has been shown to be weak; the binding constant is 5 to 10 μM (64). A single contact is thus not enough to anchor the virus particle to the cell surface. Although the presentation of the sugar moieties in the microarray is predicted to be membrane like, the relative difference in SV40 binding between the two GM1 analogs was larger in the array than at the surface of GM95 cells. This may have reflected complexities in the plasma membrane such as additional interactions between the virus and other cell surface components and the presence of lipid microdomains.

Previous studies have shown that SV40 can use the ganglioside GM1(Ac) for host cell attachment and infectivity (58, 68). While confirming the binding to ganglioside GM1(Ac), our microarray analyses revealed more efficient binding to the GM1(Gc) analog. Ganglioside GM1(Gc) was also more strongly bound by SV40 VLPs than ganglioside GM1(Ac) when incorporated into the ganglioside-deficient GM95 cells, and it restored infectivity more efficiently. Although we do not have data on the actual affinities, this suggests that ganglioside GM1(Gc) is a better receptor.

Knowing that the ceramide moieties of natural glycolipids are heterogeneous and that they can influence presentation and accessibility of the glycan moiety (32, 67), we carried out MS analyses of the GM1(Ac) and GM1(Gc) gangliosides. Since differences were observed in the ratio of major ceramides, NGLs were constructed using the homogenous aminoglycerophospholipid DHPE as the common lipid tail. When added to cells, GM1(Gc) NGL supported binding and infection equally as well as the natural GM1(Gc) ganglioside. The results collectively indicate that, irrespective of the lipid moiety, the Gc analogs were better receptors than the Ac analogs and that the superior receptor activity is conferred by the NeuGc moiety.

Our data show that the higher efficiency of SV40 binding conferred by the NeuGc moiety of GM1 ganglioside and NGL correlates with the virus infection in the live cell analysis. After cell binding, SV40 induces transmembrane signaling which leads to endocytic internalization and infection (51). In the cell infection system we showed that DHPE in the NGLs can effectively substitute for the ceramides in the gangliosides. Moreover the drug inhibition data indicate that the infection pathway for the GM1(Ac) and GM1(Gc) gangliosides as well as their NGL analogs is tyrosine kinase and cholesterol dependent, as it is for infection in cells of simians, the primary hosts. This is analogous to the observations of Pacuszka et al. in cell signaling experiments using cholera toxin (46, 47). In those studies, glycolipid-deficient cells were also supplemented with GM1 and the NGL analog. Here the NGL, like the natural ganglioside, served as a functional receptor in the supplemented cells and effectively mediated a rise in cyclic AMP levels. We conclude, therefore, that the greater SV40 binding and infectabilty of the cells supplemented with the GM1(Gc) analogs are a reflection of their superior receptor activities.

The functionality of differently N-substituted sialic acid derivatives has been investigated for murine and primate Pys. The biosynthetic inclusion of a more hydrophobic group than N-acetyl at C5 resulted in different infection rates depending on the virus and the host cell (35). In the case of Py infection of 3T6 cells, the substitution of N-acetyl with N-propionyl resulted in inhibition of infection and led to the hypothesis that the virus cannot accommodate an additional methylene group in the sialic acid binding groove (27). In contrast, NeuGc had an effect similar to that of NeuAc for inducing conformational change of Py (11).

NeuGc is expressed on most mammalian cells, including those of simians, but unlike NeuAc, which is ubiquitous in mammals, very little NeuGc is detectable in human cells (71). The reason is an inactivating mutation in the gene encoding human CMP-N-acetylneuraminic acid hydroxylase, the rate-limiting enzyme in NeuGc synthesis. The traces of NeuGc that are observed in human tissues are believed to result from the uptake of NeuGc from dietary sources of animal origin. The lack of NeuGc may explain the different susceptibility of humans to certain microbial pathogens. However, this issue has not been thoroughly studied (71).

Most of the published work on SV40 entry has been performed with simian cells such as CV-1 and Vero, in which we now know that the entry is GM1(Gc) mediated. A rather detailed picture of entry via this receptor is therefore already available. At least 50-fold more virus must be added to HeLa cells to obtain a level of infection comparable to that in simian cells (M. Schelhaas and A. Helenius, unpublished observations). Judging by inhibitor and microscopy data, the viruses do, however, follow the same pathway(s). Although less efficiently, GM1(Ac) is thus able to support productive infection in human cells.

Our results have considerable relevance to understanding the tropism of SV40 in simians, the natural hosts, and in other mammals, including humans, on which there have been many studies (16, 25, 72). As tissue tropism has not been correlated in detail with receptor expression, it has been difficult to explain the apparent preferential tropism of SV40 for nonhuman cells and to understand details of the pathobiology of the infection process in African green monkeys, rodents, and humans. It is tempting to speculate that the high rate of infection of simians with SV40 is due to the prevalence of GM1(Gc). The paucity of this ganglioside in humans may well be an example of evolutionary genetics, whereby humans have evolved to develop at least a partial barrier to virus infection. There is, however, increasing evidence for the presence of SV40 in a variety of human tumors (4, 75). Our results now open the way not only to detailed studies of SV40 infection in relation to receptor expression in host cells but also the monitoring of changes in receptor usage by the virus that may occur with time.