Carbohydrate-Carbohydrate Binding of Ganglioside to Integrin α5 Modulates α5β1Function*

Gangliosides GT1b and GD3, components of keratinocyte membranes, inhibit keratinocyte adhesion to fibronectin. Although ganglioside sialylation is known to be important, the mechanism of inhibition is unknown. Using purified insect recombinant α5 and β1 proteins and α5β1 integrin from lysed keratinocyte-derived SCC12 cells, we have shown that GT1b and GD3 inhibit the binding of α5β1 to fibronectin. Co-immunoprecipitation of GT1b and α5β1from SCC12 cells and direct binding of GT1b and GD3 to affinity-purified α5β1 from SCC12 cells and insect recombinant α5β1, particularly the α5 subunit, further suggest interaction between ganglioside and α5β1. The carbohydrate moieties of integrin appear to be critical since gangliosides are unable to bind deglycosylated forms of α5β1from SCC12 and insect cells or poorly glycosylated recombinant α5β1 from Escherichia colicells. The GT1b-α5β1 interaction is inhibited by concanavalin A, suggesting that GT1b binds to mannose structures in α5β1. The preferential binding of GT1b to high mannose rather than reduced mannose ovalbumin further implicates the binding of GT1b to mannose structures. These data provide evidence that highly sialylated gangliosides regulate α5β1-mediated adhesion of epithelial cells to fibronectin through carbohydrate-carbohydrate interactions between GT1b and the α5 subunit of α5β1 integrin.

Human keratinocyte motility on a fibronectin (FN) 1 matrix is critical for the re-epithelialization of healing wounds, for the spread of cutaneous malignancy, and in cutaneous embryogenesis. Although the molecular events that influence this migration are poorly understood, the interaction between keratinocyte ␣ 5 ␤ 1 integrin and the arginine-glycine-aspartic acid (RGD) site of FN is known to be key (1). Increased expression of ␣ 5 ␤ 1 has been shown in keratinocytes at the migrating edge of wounds in vivo, in lesional skin of patients with psoriasis, and in cultured keratinocytes (2)(3)(4). Both receptor clustering on keratinocytes and ligand occupancy of ␣ 5 ␤ 1 are required to activate intracellular signal transduction components (5), including focal adhesion kinase, phosphatidylinositol 3-kinase, protein kinase C, and integrin-linked kinase, leading to cell adhesion to FN (6 -8).
Gangliosides are glycosphingolipids characterized by the presence of one or more sialic acid moieties in the oligosaccharide chain (9). The role of gangliosides, which are localized to the outer leaflet of the plasma membrane of eukaryotic cells, is largely unknown, but studies with cultured keratinocytes and keratinocyte-derived cells suggest that gangliosides are involved in regulating cellular proliferation, differentiation, and adhesion (10 -13). The discovery that gangliosides inhibit cell attachment and spreading on a FN matrix (14) led investigators to consider gangliosides to be the cell receptors for FN before integrin ␣ 5 ␤ 1 was identified (15,16). Although these early studies showed that the terminal sialic acid residues of gangliosides were critical for the inhibitory effect, the mechanism of inhibition was unclear. More than 10 years later, our laboratory used cultured keratinocytes and keratinocyte-derived cell lines to demonstrate that the inhibition of migration and adhesion by ganglioside GT1b is specific to plating on a FN matrix and is competitively inhibited by RGDS peptide (11), suggesting that ganglioside may abrogate the interaction between ␣ 5 ␤ 1 and FN. SCC13 and HaCaT cells, keratinocytederived cell lines with diminished expression of ␣ 5 ␤ 1 in comparison with normal keratinocytes and SCC12 cells, are not inhibited by GT1b in their binding and migration on FN. The demonstration that treatment of HaCaT cells with transforming growth factor-␤1 not only increases the expression of ␣ 5 ␤ 1 , but also restores the responsiveness to the inhibitory effects of ganglioside further supports a putative interaction between GT1b and ␣ 5 ␤ 1 (13).
The possibility of a direct interaction between GT1b and ␣ 5 ␤ 1 as the mechanism for ganglioside action and its relevance for keratinocyte function have not been explored. Using recombinant and affinity-purified forms of ␣ 5 ␤ 1 and novel binding techniques, we provide evidence that GT1b binds directly to the extracellular domain of the ␣ 5 subunit through carbohydratecarbohydrate interactions, thereby inhibiting the interaction of ␣ 5 ␤ 1 with FN.
Determination of Ganglioside Content in SCC12 Cells-Gangliosides were extracted from SCC12 cell membranes using chloroform/methanol as previously described (17). The aqueous phase was separated and desalted, and the bands of gangliosides were separated by thin-layer chromatography in chloroform/methanol/water with 0.02% CaCl 2 (55: 45:10, v/v/v). Gangliosides were detected by resorcinol staining. Band density was quantified by a Storm 800 fluorescence PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, CA). The identity of the four bands of ganglioside was confirmed by immunostaining with anti-GM3 antibody (courtesy of T. T.), anti-9-O-acetyl-GD3 antibody CDw60 (Sigma), anti-GD3 antibody R24 (Calbiochem), and anti-GT1b antibody (courtesy of T. T.). Binding was detected with horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG and enhanced chemiluminescence (ECL).
Construction of ␣ 5 -and ␤ 1 -containing Plasmids and Generation of Recombinant Integrin ␣ 5 and ␤ 1 Subunits in Insect and E. coli Cells-Several human recombinant ␣ 5 and ␤ 1 integrins were generated using both baculoviral and E. coli systems and vectors with an N-terminal histidine tag that allowed cleavage or a C-terminal His tag that could not be cleaved. For generating ␣ 5 integrin in insect cells with an Nterminal cleavable His tag, the pAcHLT-C vector was cleaved by concurrent incubation with StuI and SacI. The ␣ 5 cDNA (GenBank TM /EBI accession number 06256; courtesy of Dr. Erkki Ruoslahti, The Burnham Institute, La Jolla, CA) (18) was cleaved with SacI and SalI, and the SalI site was blunt-ended. The cleaved ␣5 cDNA was ligated into the open pAcHLT-C vector. To prepare integrin ␤ 1 in insect cells with an N-terminal cleavable His tag, the pAcHLT-A vector was cut with StuI and NcoI. The ␤1 cDNA (GenBank TM /EBI accession number 07979; courtesy of Dr. Erkki Ruoslahti) (18) was cut with NcoI and ApaI, blunt-ended at the ApaI site, and ligated into the vector. Sf9 cells (2 ϫ 10 6 ) were seeded into 60-mm Petri dishes and incubated for 2 h at 27°C. BaculoGold virus DNA (Pharmingen) was mixed with the recombinant ␣ 5 or ␤ 1 plasmid DNA and transfected into insect cells following the manufacturer's instructions. For generating recombinant integrins in both the insect system and E. coli with a C-terminal non-cleavable His tag, ␣5 cDNA was cleaved by SalI and HindIII, and the SalI end was blunted-ended. The ␤1 cDNA was cut with ApaI and HindIII, and the ApaI site was blunt-ended. These cDNAs were each ligated into a pTriEX vector, which had been cut at the HindIII site and cut and blunt-ended at the BamHI site. The ␣ 5 -or ␤ 1 -containing plasmid (Life Technologies, Inc.) was transferred into E. coli cells following the manufacturer's instructions or cotransfected with BaculoGold virus DNA into insect cells. The recombinant proteins from both insect cells and E. coli supernatants were purified by passage over a Ni 2ϩ -nitrilotriacetic acid spin column (QIAGEN Inc., Valencia, CA) following the manufacturer's instructions. After purification, the N-terminal His tag was removed from ␣ 5 or ␤ 1 from insect cells, unless desired for use in the nickel-agarose binding (NAB) technique. Thrombin (Pharmingen) was added to the protein at a ratio of 1 unit of thrombin to 10 g of integrin subunit protein, run on a Sephacryl S-200HR column (Amersham Pharmacia Biotech), and dialyzed against phosphate-buffered saline (PBS), pH 7.2, following the manufacturer's instructions. The C-terminal His tags were retained on the E. coli and insect cell integrin proteins. To make recombinant ␣ 5 ␤ 1 complexes, ␣ 5 and ␤ 1 subunits were combined in a 1:1 molar ratio in the presence of 30 M CaCl 2 .
The identity of the purified recombinant proteins was confirmed by Western blotting. The purified recombinant proteins were boiled in Laemmli sample buffer (19), and 10 ng/lane was applied to an 8% SDS-polyacrylamide minigel. Protein was transferred to polyvinylidene difluoride filters, and nonspecific binding was blocked by incubation for 1 h with 5% dry milk in 50 mM Tris-HCl, pH 7.6, with 137 mM NaCl and 0.1% Tween 20. Lanes were treated with 50 ng/ml anti-␣ 5 antibody (Transduction Laboratories, Lexington, KY), 100 ng/ml anti-␤1 antibody (Transduction Laboratories), or 250 ng/ml anti-␣ 5 ␤ 1 antibody (Zymed Laboratories Inc., South San Francisco, CA) for 1 h, and bands were detected by chemiluminescence using an ECL kit (Amersham Pharmacia Biotech) with goat anti-mouse secondary antibodies following the manufacturer's instructions.
Generation of Affinity-purified ␣ 5 ␤ 1 from SCC12 Cells-Intact integrin ␣ 5 ␤ 1 was purified from SCC12 cells by affinity chromatography on CNBr-activated Sepharose 4B (Sigma) coupled to RGD polymer and anti-␣ 5 ␤ 1 antibody as a modification of the previously described techniques (20,21). Briefly, RGD polymer (5 mg/ml of gel) was covalently coupled to CNBr-activated Sepharose 4B according to the manufacturer's instructions. SCC12 cells were lysed (20), and the lysate was poured over the RGD column. After elution, the integrin bound to RGD polymer was further purified by affinity chromatography with anti-␣ 5 ␤ 1 antibody coupled to a CNBr-activated Sepharose 4B column as described above for RGD polymer. Fractions were assayed for integrin by Western blot assays using anti-␣ 5 or anti-␤ 1 antibody and chemiluminescence for detection.
ELISA Binding Assays-The ability of ganglioside to block the binding of the ␣ 5 ␤ 1 integrin complex to the RGD-containing fragment of FN in a cell-free system was assayed by ELISA. 5 g/cm 2 RGD-containing FN fragment (Sigma) was cross-linked to the wells of a 96-well plate by treatment with 10 mg/ml 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDAC) (Molecular Probes, Inc., Eugene, OR) in PBS overnight at 4°C. To evaluate the effect of gangliosides on the binding of ␣ 5 ␤ 1 from SCC12 cells to FN, cells were treated overnight with or without GT1b, GD3, GM3, GM2, or GD1a (Calbiochem) at concentrations ranging from 1 nM to 50 M. The SCC12 cells were lysed as described by Pomies et al. (22), and the cell lysates in lysis buffer at protein concentrations of 10 pg/ml to 1 mg/ml were incubated in each FN fragment-coated well for 2 h at room temperature. After washing, the wells were treated with anti-␣ 5 ␤ 1 antibody, followed by HRP-conjugated secondary antibody and 3,3Ј-diaminobenzidine tetrahydrochloride (DAB) reaction substrate. To test the ability of gangliosides to block the binding to FN of a 1:1 molar mixture of insect recombinant ␣ 5 and ␤ 1 , 0.1 pg/ml to 0.1 g/ml integrin (without the His tags) was preincubated with or without gangliosides overnight at 4°C and then applied to the FN fragmentcoated wells. Recombinant integrin bound to FN was immunostained as described for ␣ 5 ␤ 1 from SCC12 cells. Binding was detected in a V max kinetic microplate reader (Molecular Devices, Menlo Park, CA) at A 450 nm . In each case, uncoated wells and omission of anti-␣ 5 ␤ 1 antibody served as negative controls. All studies were repeated at least four times, with triplicate wells for each condition.
Nickel-Agarose Binding Assay-A novel assay, the NAB technique, was used to assess whether the ability of GT1b to interfere with integrin binding to FN specifically involves one of the integrin subunits and to verify that the N terminus of integrin is critical. This assay takes advantage of the fact that the His tag can bind to nickel-agarose; thus, a His tag at the C-terminal end exposes the N terminus with the integrin extracellular domains. 5 g of purified insect recombinant ␣ 5 or ␤ 1 protein with a C-terminal His tag (i.e. the N terminus available) or, as a control, with an Nterminal His tag was mixed with 100 l of nickel-agarose beads (QIA-GEN Inc.) for 1 h at room temperature. After washing, the coated nickel-agarose beads were mixed for 16 h in wash buffer (50 mM sodium phosphate, 300 mM NaCl, and 10% glycerol, pH 8.0) containing 20 g/ml RGD-containing 110-kDa FN fragment with 1-50 M GT1b. Controls included mixing the integrin with gangliosides without FN and mixing the integrin and FN without any ganglioside. The bound bead was then washed and boiled in denaturing Laemmli buffer. Separated proteins were evaluated by Western blotting as described above, except that proteins on polyvinylidene difluoride membrane were detected with a 1:1 mixture of anti-fibronectin antibody (Zymed Laboratories Inc.) and anti-␣ 5 or anti-␤ 1 antibody. As a positive control and to verify results of the ELISAs, a 1:1 mixture of ␣ 5 and ␤ 1 (the ␣ 5 ␤ 1 complex) was bound to nickel-agarose and mixed with FN in the presence or absence of GT1b. Similar techniques were used, except that the nickel-agarose-bound C-terminal His-integrin ␤ 1 subunit was first able to form nickel-agarose-bound C-terminal His-␣ 5 ␤ 1 by incubation with the partner ␣ 5 subunit without its His tag, prior to mixing with FN and ganglioside. To prevent separation of ␣ 5 from ␤ 1 and to allow detection with anti-␣ 5 ␤ 1 antibody, the complex of nickel-agarose-bound His-␣ 5 ␤ 1 and FN was incubated in Laemmli buffer containing 17.4% sucrose for 30 min at room temperature without boiling, and the gel was treated with 2.5% Triton X-100 before transferring to the polyvinylidene difluoride membrane. NAB assays were performed three times.
Co-immunoprecipitation of Gangliosides with ␣ 5 ␤ 1 and Detection of Gangliosides-SCC12 cells were treated for 48 h with or without GT1b or GD3 at concentrations of 1 nM to 50 M or, as a control, with 10, 50, or 200 M GM2, GD1a, or GM3. The cells were lysed in calcium-free buffer containing 20 mM Hepes, pH 7.2, 1% Nonidet P-40, 10% (v/v) glycerol, 50 mM NaF, 1 mM PMSF, 1 mM Na 3 VO 4 , and 10 g/ml leupeptin (23). 5 g of anti-␣ 5 ␤ 1 antibody was added to 2.5 mg of cell-free lysates at 4°C for 2 h. In some studies, 5 g/ml FN was added to determine whether the presence of FN significantly altered the complex formation of ganglioside with ␣ 5 ␤ 1 integrin. After treatment with 5 g of rabbit anti-mouse IgG antibody for 30 min, 10 l of 50% protein A-agarose was added, and the mixture was incubated for another 2 h at 4°C. One-eighth of the mixture was boiled in Laemmli buffer, and the separated integrin ␣ 5 and ␤ 1 subunits were detected by Western blotting with anti-␣ 5 and anti-␤ 1 antibodies. Lipids were extracted from the remaining samples with chloroform/methanol (2:1, v/v) (17), separated by thin-layer chromatography as described above, and identified with anti-GM3, anti-GM2, anti-GM1, anti-GD3, and anti-GT1b antibodies, followed by enhanced chemiluminescence detection as previously described (24). Band density was quantified by the Storm 800 fluorescence PhosphorImager.
Slot Blot Assays-To further evaluate the ability of gangliosides GT1b and GD3 to bind directly to ␣ 5 ␤ 1 or to one of its subunits, slot blot assays were performed. 20 l of 50 M GT1b, GD3, or, as a control, GM3, GM2, GD1a, or 1% bovine serum albumin (BSA) was loaded into each slot of a positively charged nylon membrane (Roche Molecular Biochemicals) on a Bio-Dot SF apparatus (Bio-Rad). The membrane was air-dried and blocked by the addition of 1% BSA in PBS, pH 7.6. After air drying, the membrane was incubated with 50 ng/ml affinity-purified ␣ 5 ␤ 1 from the SCC12 cells, with 5 g/ml insect recombinant protein ␣ 5 or ␤ 1 , or with a 1:1 molar mixture of ␣ 5 and ␤ 1 in PBS, pH 7.6, at 4°C for 4 -6 h. The membrane was washed with PBS, and the binding was detected with anti-␣ 5 , anti-␤ 1 , or anti-␣ 5 ␤ 1 antibody using Western blotting with an ECL kit as described above.
Effect of Ganglioside on the Expression of ␣ 5 ␤ 1 and on the Ability of ␣ 5 to Complex with ␤ 1 -SCC12 cells were incubated overnight with GT1b at concentrations of 1 nM to 50 M GT1b or, as a control, with 50 M GM3, GM2, or GM1 or medium without ganglioside. Cells were lysed, boiled in Laemmli buffer, and run on an 8% SDS-polyacrylamide gel. After transfer to polyvinylidene difluoride membrane, the expression of ␣ 5 and ␤ 1 was detected with anti-␣ 5 and anti-␤ 1 antibodies and an ECL kit.
The NAB technique was also used to investigate whether GT1b interferes with the complex formation between ␣ 5 and ␤ 1 subunits. 100 l of nickel-agarose beads was mixed for 1 h at room temperature with 5 g of insect recombinant ␣ 5 with a C-terminal His tag or with 5 g of insect recombinant ␤ 1 with a C-terminal His tag, leaving the N terminus free to interact. As a negative control, nickel-agarose beads were also mixed with ␣ 5 or ␤ 1 subunits with N-terminal His tags. After washing, the His-␣ 5 -nickel-agarose or His-␤ 1 -nickel-agarose beads were mixed with GT1b at concentrations of 1 nM to 50 M in PBS for 4 h. Subsequently, 5 g of the appropriate partner integrin subunit without its His tag (e.g. ␣ 5 for nickel-agarose-bound His-␤ 1 ) was added for 1 h at room temperature. After washing, the sample was boiled in Laemmli sample buffer for 5 min, and the subunit combination was identified by Western blotting as described above using a 1:1 mixture of anti-␣ 5 and anti-␤ 1 antibodies. In all NAB studies, the presence of GT1b was confirmed by loading an aliquot onto a membrane for slot blot assays as described above, except that anti-GT1b antibody was used to detect GT1b.
Deglycosylation of ␣ 5 ␤ 1 -To assess the role of sugar moieties of ␣ 5 ␤ 1 in the interaction with gangliosides, ␣ 5 ␤ 1 from SCC12 cells and insect recombinant ␣ 5 and ␤ 1 proteins were deglycosylated by a modification of previously described techniques (21,25). In addition, poorly glycosylated recombinant ␣ 5 ␤ 1 was generated in E. coli cells (26) as described above. After reaching 80% confluence, SCC12 cells were switched to serum-free medium with 2 g/ml tunicamycin. 2 units/ml PNGase F (Calbiochem) was added to the medium 22 h later, and the cells were incubated with the combination of PNGase F and tunicamycin for an additional 2 h. After washing with 10 mM Tris-HCl, pH 7.4, 0.15 M NaCl, 1 mM MnCl 2 , and 0.2 mM PMSF to clear cleaved carbohydrate residues and residual enzyme, the deglycosylated proteins were collected by cell lysis with 10 mM Tris-HCl, pH 7.4, 0.15 M NaCl, 1 mM MnCl 2 , 3 mM PMSF, and 0.1 M octyl glucoside. After centrifugation, the pellet was discarded. Recombinant ␣ 5 and ␤ 1 were treated with 0.4 units/ml PNGase F for 1 h at 37°C (21) and then lysed and purified by passage over a Ni 2ϩ -nitrilotriacetic acid spin column as described above. The resultant deglycosylated SCC12 cell ␣ 5 ␤ 1 , insect recombinant ␣ 5 ␤ 1 , and E. coli ␣ 5 ␤ 1 were analyzed in comparison with the original glycosylated forms of ␣ 5 ␤ 1 by electrophoresis on a 7.5% SDSpolyacrylamide minigel. Gels were stained with Coomassie Blue or periodate-Schiff reagent (Sigma) and compared with known molecular mass standards.
Assessment of the Role of Glycosylation in the Ganglioside Interaction with Integrin-ELISAs were performed as described above to examine the effect of gangliosides on the ability of deglycosylated integrins and poorly glycosylated integrin from E. coli to bind to FN. In addition, bead binding assays and slot blot assays were used to determine whether deglycosylation prevents the direct binding of ganglioside to the integrin. Slot blot assays were performed as described above, except that deglycosylated ␣ 5 ␤ 1 from SCC12 cells (50 ng/ml), deglycosylated ␣ 5 ␤ 1 from insect recombinant cells (5 g/ml), and the poorly glycosylated ␣ 5 ␤ 1 from E. coli (5 g/ml) were substituted for the glycosylated forms of ␣ 5 ␤ 1 .
In the bead binding assays, 50 M GT1b or GD3 in PBS was crosslinked to 250 l of 0.2-m amine-modified red FluoSphere beads (1 ϫ 10 5 particles/ml in PBS; F-8763, Molecular Probes, Inc.) by mixing with 5 mg/ml EDAC in PBS overnight at 4°C. 50 M GD1a, GM2, or GM3 in PBS or PBS alone was coated onto FluoSphere beads as negative controls. To detect the attachment of the ganglioside to the FluoSphere beads, 20 l of GT1b-coated, GM3-coated, or uncoated red beads was washed and loaded into the slots of a slot blot apparatus as described above. The colored beads are easily visible under ultraviolet light as red bands. The binding of gangliosides GT1b, GM2, and GM3 to beads was confirmed by incubation with anti-GT1b, anti-GM2 (courtesy of Dr. P. Livingston, Sloan-Kettering, New York, NY), and anti-GM3 monoclonal antibodies, respectively, all applied at a ratio of 1:2 in PBS, and detected by chemiluminescence as described above. Affinity-purified ␣ 5 ␤ 1 integrin protein from SCC12 cells, deglycosylated affinity-purified ␣ 5 ␤ 1 from SCC12 cells, recombinant ␣ 5 and ␤ 1 subunits generated in insect cells and their 1:1 mixture, deglycosylated insect recombinant ␣ 5 and ␤ 1 subunits and their 1:1 mixture, recombinant ␣ 5 and ␤ 1 from E. coli, and 1% BSA were coated onto 96-well plates overnight at 4°C as described for FN ELISAs. To ensure good binding of integrin or integrin subunits to the plate, wells were washed and then treated with anti-␣ 5 , anti-␤ 1 , or anti-␣ 5 ␤ 1 antibody, and the integrin was detected by immunofluorescence. After washing, the beads coated with gangliosides were added to the integrin-coated plates. After a 2-h incubation, the binding of the colored ganglioside-coated spheres to integrin was visualized by immunofluorescence microscopy (Nikon Eclipse TE300 immunofluorescence microscope linked to a computer with Neurolucida software from Mi-crobrightField, Inc., Colchester, VT) at magnification ϫ 200 using a 580-nm filter. The number of beads counted in 13 non-overlapping fields was expressed as a percentage of the number of GT1b-coated beads bound to fully glycosylated affinity-purified ␣ 5 ␤ 1 from SCC12 cells.
Role of Specific Carbohydrate Moieties in the Interaction with Ganglioside-ELISAs were also used to examine the competitive binding of concanavalin A (ConA) lectin, which recognizes mannose residues, and GT1b or other gangliosides to ␣ 5 ␤ 1 . In these studies, 96-well clear polystyrene plates (Calbiochem) precoated with ConA by the manufacturer were washed, and 200 l of 10 nM GT1b, GM3, or GM2 or 1% BSA was added to the plate concurrent with affinity-purified ␣ 5 ␤ 1 from SCC12 cells at concentrations of 10 Ϫ14 to 10 Ϫ6 g/ml. After washing, the binding of ␣ 5 ␤ 1 integrin to ConA was detected by incubation with anti-␣ 5 ␤ 1 antibody, HRP-conjugated goat anti-mouse secondary antibody, and DAB. Binding was measured as described above at A 450 nm . As a control, the same experiment was performed with Ulex europaeus agglutinin-1 (UEA-1) lectin, which preferentially binds fucose groups rather than mannose, in substitution for the ConA lectin. The experiment was performed at least four times in triplicate.
To further assess a preferential binding of GT1b and GD3 for mannose-containing sugars, ELISAs were performed to examine the binding of gangliosides to high mannose and reduced mannose forms of ovalbumin (Sigma). Purification of "high mannose" ovalbumin protein cleaves mannose residues and reduces content by ϳ30% to generate "low mannose" ovalbumin (27). 50 g/ml high or low mannose ovalbumin was coated by overnight incubation in the presence of 10 mg/ml EDAC in PBS at 4°C onto a 96-well plate. After washing, 200 l of 2-14 nM GM1, GM2, GD1a, GD3 or GT1b or 1% BSA was added to each well and incubated overnight at 4°C. The plate was then washed vigorously with PBS and reacted with anti-GM1, anti-GM2, anti-GD1a, anti-GD3, or anti-GT1b monoclonal antibody, followed by HRP-conjugated secondary antibody and DAB reaction substrate. Binding was detected at A 450 nm . The experiment was performed at least four times in triplicate.
Statistical Analysis-All data were analyzed statistically by Student's t test, with p Ͻ 0.05 considered significant.

RESULTS
GT1b and GD3 Are Ganglioside Components of the Membranes of the Keratinocyte-derived SCC12 Cell-We have previously shown that gangliosides compose 0.1% of the lipids of intact epidermis (17), with GM3 as the predominant ganglioside and smaller amounts of gangliosides GD3 and GT1b. Studies with cultured keratinocytes (data not shown) demonstrate a ganglioside profile that is similar to that of cultured SCC12 cells (Fig. 1), with ϳ62.9% GM3, 16.9% 9-O-acetyl-GD3, 13.7% GD3, and 6.5% GT1b. The presence of all four of these gangliosides in cultured SCC12 cells was confirmed by immuno-staining of the thin-layer chromatography plates with anti-GM3, anti-9-O-acetyl-GD3, anti-GD3, and anti-GT1b antibodies (Fig. 1).
GT1b Inhibits the Adhesion of ␣ 5 ␤ 1 to FN in a Cell-free System-Previous investigations have studied the ability of ganglioside to block the binding and migration of intact cultured keratinocytes and keratinocyte-derived cells on FN. To eliminate the possibility that ganglioside-induced inhibition occurs indirectly through alterations in other cellular components, we studied the ability of GT1b to block the binding of ␣ 5 ␤ 1 from lysed SCC12 cells and of recombinant integrin ␣ 5 ␤ 1 to the RGD-containing cell-binding region of FN. Recombinant ␣ 5 and ␤ 1 integrin proteins were generated using an insect system and a vector with a cleavable His tag. When combined in a 1:1 molar ratio, these purified ␣ 5 and ␤ 1 integrin subunits formed complexes that bound well to FN (Figs. 2B and 3).
In previous studies (11), we bound FN or its RGD-containing fragment directly to plastic wells to show by ELISA that GT1b and GD3 inhibit the binding of cultured keratinocytes and SCC12 cells to native FN and to the RGD-containing fragment of FN. Because the frequent and vigorous washing of the plate led to detachment of variable amounts of FN, even more when the RGD-containing fragment of FN was attached, we studied methods to improve adherence of FN to the plate. When EDAC was used to fix the FN or RGD-containing fragment of FN to the plate, results paralleled those of studies without EDAC fixation, but the colorimetric readings were higher, and the results were beautifully reproducible. Thus, the binding of integrin to FN in the presence of GT1b was tested using an ELISA technique with the RGD-containing fragment of FN fixed to the plate with EDAC and with both ␣ 5 ␤ 1 from lysed SCC12 cells ( Fig. 2A) and the recombinant ␣ 5 ␤ 1 complex (Fig.  2B). 10 nM GT1b significantly decreased the binding to the RGD-containing fragment of FN (p Ͻ 0.05 at 100 pg/ml ␣ 5 ␤ 1 from lysed SCC12 cells and p Ͻ 0.05 at 1 pg/ml insect recombinant ␣ 5 ␤ 1 ). GM3, GM2, and GD1a at concentrations as high as 50 M and 1% BSA did not show a significant inhibitory effect on binding to FN of either ␣ 5 ␤ 1 from lysed SCC12 cells or insect recombinant ␣ 5 ␤ 1 . The effect of 10 nM GD3 on binding was intermediate, but significant for both ␣ 5 ␤ 1 from lysed SCC12 cells (p Ͻ 0.05 at 10 ng/ml ␣ 5 ␤ 1 ) and insect recombinant ␣ 5 ␤ 1 (p Ͻ 0.05 at 10 pg/ml ␣ 5 ␤ 1 ).
Inhibitory Effect of GT1b Requires the ␣ 5 Subunit of ␣ 5 ␤ 1 -The specificity of the inhibitory effect of ganglioside on cell binding to a FN matrix, and not to collagen I or other matrices (11), targeted the ␣ 5 subunit of ␣ 5 ␤ 1 as critical since other ␤ 1 integrin keratinocyte complexes, such as ␣ 2 ␤ 1 and ␣ 3 ␤ 1 , promote adhesion to collagen I. Taking advantage of the known ability of nickel-agarose beads to bind proteins with a His tag, we used a novel technique that directly assesses the ability of the His-tagged protein to bind other proteins, the NAB technique. Using the NAB technique to bind the His tag attached to the intercellular C terminus of ␣ 5 to the nickel-agarose bead (i.e. the extracellular N terminus of integrin is free to interact) and Western blotting, we verified the results of ELISAs, showing that as little as 1 nM GT1b inhibits the binding of both the insect recombinant ␣ 5 ␤ 1 complex (Fig. 3A) and the ␣ 5 subunit (Fig. 3B) to the RGD-containing region of FN, supporting ␣ 5 as the critical subunit for interaction with GT1b. The binding of the ␤ 1 subunit to FN was overall much weaker, but was not diminished by incubation with concentrations of GT1b as high as 50 M (Fig. 3C). ␣ 5 and ␤ 1 subunits with N-terminal His tags (i.e. the intercellular C terminus of integrin is available to interact) did not bind to FN; and thus, Western blots showed only bands of integrin (data not shown).
Mechanism of Inhibition of Adhesion by Ganglioside Involves   FIG. 1. Ganglioside content of keratinocyte-derived SCC12 cells. Gangliosides were extracted from SCC12 cell membranes using chloroform/methanol (2:1). The aqueous phase was separated and desalted, and the bands of gangliosides were separated by thin-layer chromatography in chloroform/methanol/water with 0.02% CaCl 2 (55: 45:10, v/v/v). Gangliosides were detected by resorcinol staining (not shown). The identity of each ganglioside was confirmed by immunostaining with a mixture of anti-GM3, anti-9-O-acetyl-GD3, anti-GD3, and anti-GT1b antibodies, followed by HRP-conjugated goat anti-mouse IgG and ECL (second lane). The first lane shows ganglioside standards GM3, GD3, and GT1b.
FIG. 2. Effect of ganglioside on adhesion of ␣ 5 ␤ 1 to FN. 5 g/cm 2 RGD-containing fragment of FN was cross-linked to the wells of a 96-well plate by treating with 10 mg/ml EDAC in PBS overnight at 4°C. A, to test the effect of gangliosides on the binding of ␣ 5 ␤ 1 from SCC12 cells, cells were treated overnight with or without 10 nM GT1b or GD3 or 50 M GM3, GM2, or GD1a (concentrations based on previous studies showing no inhibition or toxicity at concentrations of 1 nM to 50 M GM3, GM2, and GD1a) and then lysed. Lysates at protein concentrations of 10 pg/ml to 1 mg/ml were added to each FN fragment-coated well, followed by anti-␣ 5 ␤ 1 antibody, HRP-conjugated secondary antibody, and DAB reaction substrate. B, to test the ability of gangliosides to block the binding to FN of a 1:1 molar mixture of insect recombinant ␣ 5 and ␤ 1 , 0.1 pg/ml to 0.1 g/ml recombinant ␣ 5 ␤ 1 was preincubated with or without gangliosides overnight at 4°C and studied as described above for ␣ 5 ␤ 1 from lysed SCC12 cells. Binding was detected in a V max kinetic microplate reader at A 450 nm . A: Ⅺ, SCC12 ␣ 5 ␤ 1 without ganglioside; ‚, SCC12 the Direct Binding of GT1b to ␣ 5 ␤ 1 Integrin, Particularly to the ␣ 5 Subunit-The complexing of GT1b with affinity-purified ␣ 5 ␤ 1 from SCC12 cells was shown by co-immunoprecipitation of GT1b and ␣ 5 ␤ 1 (Fig. 4A). Even without the addition of supplemental purified GT1b, the presence of GT1b (Fig. 4B) and GD3 (data not shown) was detected in the complex with ␣ 5 ␤ 1 by thin-layer chromatography immunostaining using anti-GT1b and anti-GD3 antibodies, respectively. The addition of FN to the mixture neither significantly increased nor decreased the content of GT1b in the complex with ␣ 5 ␤ 1 (Fig. 4B). As little as 1 nM exogenous GT1b, added to SCC12 cells to increase membrane GT1b content, significantly increased the detection of GT1b after extraction from the immunoprecipitated complex and separation by thin-layer chromatography (Fig. 4B). The addition of 50 M GD3 to cells before immunoprecipitation of ␣ 5 ␤ 1 resulted in the detection of a strong band of GD3 in the complex (Fig. 4A), but even the addition of 200 M GM3 did not lead to the detection of GM3 in the complex with immunoprecipitated ␣ 5 ␤ 1 , as shown by immunostaining with anti-GM3 antibodies (Fig. 4A). Similarly, GM2 and GM1 did not immunoprecipitate with ␣ 5 ␤ 1 . Western blotting of the immunoprecipitated complex confirmed the presence of the ␣ 5 ␤ 1 proteins (Fig. 4C).
Slot blot assays also showed significant binding of both affinity-purified ␣ 5 ␤ 1 (Fig. 5A) and the recombinant ␣ 5 subunit (Fig. 5B) to GT1b (both p Ͻ 0.01 when compared by densitometric measurements with BSA and ganglioside controls) and to GD3 (both p Ͻ 0.05 compared with controls). No significant binding of integrin to ganglioside was detected with GM3, GM2, or GD1a.
GT1b Does Not Affect the Expression of Integrins or the Ability of Integrin Subunits to Form the ␣ 5 ␤ 1 Complex-Western blot analysis showed no detectable alteration in the expres-sion of ␣ 5 ␤ 1 from lysed SCC12 cells that had been incubated in the presence of GT1b at concentrations of 1 nM to 50 M (Fig.  6A). Incubation of the SCC12 cells with other gangliosides at 50 M, including GM3, GM2, and GM1, similarly did not alter SCC12 cell expression (data not shown). To address the question of the ability of ganglioside to block integrin subunits from FIG. 3. NAB technique used to demonstrate the effect of ganglioside GT1b on the binding of FN to recombinant ␣ 5 ␤ 1 subunits. 5 g of purified insect recombinant ␣ 5 or ␤ 1 protein with a C-terminal histidine tag was mixed with 100 l of nickel-agarose beads for 1 h at room temperature. Binding of the C-terminal His tag leaves the N-terminal extracellular portion of the integrin available for binding. After washing, the coated beads with the free extracellular N terminus of integrin were mixed for 16 h in wash buffer with 20 g/ml 110-kDa FN cell-binding fragment in the presence of up to 50 M GT1b. For evaluating the ␣ 5 ␤ 1 complex, the beads with integrin were first mixed with the partner integrin subunit without its histidine tag before incubation with FN. The bound bead was then washed and incubated with Laemmli buffer containing 17.4% sucrose for the ␣ 5 ␤ 1 complex at room temperature or boiled in denaturing Laemmli buffer for ␣ 5 and ␤ 1 subunits. Bands were separated on an 8% SDS-polyacrylamide gel and transferred to polyvinylidene difluoride membrane for Western blotting. The presence of integrin or integrin subunit and FN was detected with a 1:1 combination of anti-fibronectin antibody and anti-␣ 5 ␤ 1 (A), anti-␣ 5 (B), or anti-␤ 1 (C) antibody.

FIG. 4.
Ability of GT1b to co-immunoprecipitate with ␣ 5 ␤ 1 from SCC12 cells. SCC12 cells were treated for 48 h with or without gangliosides at concentrations as high as 200 M. The cells were lysed in buffer, and 5 g of anti-␣ 5 ␤ 1 antibody was added to 2.5 mg of cell-free lysates at 4°C for 2 h. In some studies, 5 g/ml FN was added to determine whether the presence of FN significantly altered the complex formation of ganglioside with the ␣ 5 ␤ 1 integrin. After treatment with 5 g of rabbit anti-mouse IgG for 30 min, 10 l of 50% protein A-agarose was added, and the mixture were incubated for another 2 h at 4°C. Lipids were extracted from seven-eighths of the samples for thin-layer chromatography with chloroform/methanol (2:1, v/v). The aqueous phase was separated and desalted, and the bands were separated by thin-layer chromatography in chloroform/methanol/water with 0.02% FIG. 5. Direct binding of GT1b and GD3 to ␣ 5 ␤ 1 and the ␣ 5 subunit. 20 l of 50 M GT1b, GM3, GD3, GM2, or GD1a or 1% BSA was loaded into each slot of a positively charged nylon membrane on a Bio-Dot SF slot blot apparatus. After blocking with 1% BSA, the membrane was incubated with 50 ng/ml affinity-purified ␣ 5 ␤ 1 (A) or 5 g/ml recombinant protein ␣ 5 (B) at 4°C for 4 -6 h. Binding was detected with anti-␣ 5 ␤ 1 or anti-␣ 5 antibody by Western blotting with enhanced chemiluminescence detection.
forming an ␣ 5 ␤ 1 complex, we again used the NAB technique. Regardless of whether the nickel-agarose beads bound the His tag of the C-terminal domain of insect recombinant ␤ 1 or the His tag of the C-terminal domain of ␣ 5 (in both situations, the extracellular N-terminal domains are free to interact), treatment with concentrations of GT1b from 1 nM to 50 M had no effect on the ability of the integrin subunit to recognize and bind to its partner (i.e. ␣ 5 to ␤ 1 and ␤ 1 to ␣ 5 ), as detected by equally dense bands of ␣ 5 and ␤ 1 on Western blots after subunit separation (Fig. 6B).
Glycosylation of ␣ 5 ␤ 1 Integrin Is Critical for Interaction with GT1b-Native affinity-purified human ␣ 5 ␤ 1 and recombinant proteins generated in insect cell systems are relatively well glycosylated (28). In contrast, recombinant proteins generated in an E. coli system are poorly glycosylated (26). To assess the importance of the integrin carbohydrate groups for function and for binding to ganglioside, the carbohydrate groups were stripped both from the insect recombinant subunits and from ␣ 5 ␤ 1 from SCC12 cells by treatment with PNGase F alone or with PNGase F and tunicamycin, respectively. In addition, ␣ 5 and ␤ 1 were generated in an E. coli system. The deglycosylated forms of ␣ 5 ␤ 1 were compared with the poorly glycosylated E. coli integrin and with the original glycosylated forms of integrin. Treatment with PNGase F and tunicamycin decreased the molecular mass of ␣ 5 from SCC12 cells from 145 to 100 kDa and that of ␤ 1 from SCC12 cells from 135 to 90 kDa (Fig. 7A). Similarly, deglycosylation with PNGase F decreased the molecular mass of insect recombinant ␣ 5 from 135 to 100 kDa and that of ␤ 1 from 130 to 90 kDa. The molecular mass of E. coli ␣ 5 was 100 kDa, and that of ␤ 1 was 92 kDa. Use of a periodate-Schiff reagent stain to detect carbohydrate groups showed that deglycosylation drastically reduced or eliminated detectable carbohydrate residues on ␣ 5 ␤ 1 (Fig. 7B) and verified the poor glycosylation of the E. coli integrins. Deglycosylation decreased the ability of ␣ 5 ␤ 1 from lysed SCC12 cells and recombinant ␣ 5 ␤ 1 from insect cells to bind to the RGD-containing fragment of FN. As detected by ELISAs, deglycosylation reduced the binding of ␣ 5 ␤ 1 from SCC12 cells to FN by 70% and that of a 1:1 mixture of insect recombinant ␣ 5 and ␤ 1 integrins to FN by 65%. Similarly, a 1:1 mixture of ␣ 5 and ␤ 1 generated in an E. coli system bound weakly to FN (71% less than glycosylated ␣ 5 ␤ 1 from SCC12 cells), verifying the importance of integrin glycosylation for interaction with the RGD-binding region of FN (21).
Fluorescent bead binding assays were used to detect binding of ganglioside to integrin by immunofluorescence. In this technique, the ganglioside is initially bound to an amine-modified fluorescent FluoSphere bead with the carbodiimide reagent EDAC. Because the fixation by this method requires loss of a carboxyl group of the ganglioside, we were concerned about alteration in ganglioside function and ability to bind. In several trials, whether binding GM1 to the ␤ subunit of cholera toxin, GM3 to the epidermal growth factor receptor, or GT1b to ␣ 5 ␤ 1 , binding was beautifully reproducible, and controls with other gangliosides and receptor proteins were consistently negative (data not shown). Fluorescent bead binding assays showed strong binding of gangliosides GT1b and GD3 to native affinitypurified ␣ 5 ␤ 1 from SCC12 cells (Table I), with 293.67 Ϯ 18.65 beads bound per 13 non-overlapping fields within the wells. Consistent with the results of other experiments, the binding of the GT1b-coated beads to native recombinant ␣ 5 ␤ 1 was 84% that of affinity-purified ␣ 5 ␤ 1 from SCC12 cells, and the binding of GT1b to the ␣ 5 and ␤ 1 subunits was 64 and 35%, respectively. The binding of the GD3-coated beads was 66 -88% that of the GT1b-coated beads. In contrast, the binding of beads coated FIG. 6. Lack of effect of GT1b on the expression of integrin ␣ 5 ␤ 1 or the ability of the ␣ 5 and ␤ 1 subunits to complex with each other. A, to study the effect of ganglioside on the expression of ␣ 5 and ␤ 1 , SCC12 cells were incubated overnight with or without GT1b at concentrations of 1 nM to 50 M. Cells were lysed, run on an 8% SDS-polyacrylamide minigel, and transferred to polyvinylidene difluoride membrane, and the expression of ␣ 5 and ␤ 1 was detected with a mixture of anti-␣ 5 and anti-␤ 1 antibodies and an ECL kit. B, to determine the ability of GT1b to interfere with the complex formation between ␣ 5 and ␤ 1 subunits, a NAB sandwich technique was performed. 100 l of nickel-agarose beads was mixed with 5 g of one of four recombinant integrin subunits: ␣ 5 with a His tag at the extracellular N terminus, ␣ 5 with its His tag at the intracellular C terminus, ␤ 1 with its His tag at the N terminus, or ␤ 1 with its His tag at the C terminus. After washing, the His-␣ 5 -nickel-agarose or the His-␤ 1 -nickel-agarose beads were mixed with 1-100 M GT1b in PBS for 4 h, and 5 g of the appropriate partner integrin subunit without its His tag (for example, recombinant ␤ 1 integrin for ␣ 5 ) was added for 1 h at room temperature. After washing, the sample was boiled in Laemmli sample buffer for 5 min, and bands were identified by Western blotting as described above using a 1:1 mixture of anti-␣ 5 and anti-␤ 1 antibodies.
FIG. 7. Deglycosylation of ␣ 5 ␤ 1 . SCC12 cells were treated with 2 g/ml tunicamycin for 24 h, and 2 units/ml PNGase F was added to the medium for the last 2 h of incubation. After washing with 10 mM Tris-HCl, pH 7.4, 0.15 M NaCl, 1 mM MnCl 2 , and 0.2 mM PMSF to clear cleaved carbohydrate residues and residual enzyme, the deglycosylated proteins were collected by cell lysis with 10 mM Tris-HCl, pH 7.4, 0.15 M NaCl, 1 mM MnCl 2 , 3 mM PMSF, and 0.1 M octyl glucoside. After centrifugation, the pellet was discarded. Sf9 insect and E. coli recombinant ␣ 5 and ␤ 1 were treated with 0.4 units/ml PNGase F for 1 h at 37°C and then lysed and prepared as described for ␣ 5 ␤ 1 from SCC12 cells. The resultant deglycosylated forms of ␣ 5 ␤ 1 (ϩ lanes) were analyzed in comparison with the original glycosylated ␣ 5 ␤ 1 forms (Ϫ lanes) by electrophoresis on a 7.5% SDS-polyacrylamide minigel. Gels were stained with Coomassie Blue (A) or periodate-Schiff reagent (B) and compared with known molecular mass standards. with GM3, GD1a, and GM2 was Ͻ9% that of GT1b to affinitypurified ␣ 5 ␤ 1 with all forms of glycosylated integrin.
Deglycosylation of insect recombinant ␣ 5 ␤ 1 eliminated the ability of ganglioside GT1b or GD3 to bind the integrin to the basal level of binding of other gangliosides (Table I) and of beads coated with albumin alone. Similarly, both deglycosylated ␣ 5 ␤ 1 from SCC12 cells and the poorly glycosylated ␣ 5 ␤ 1 generated in E. coli cells were not able to bind gangliosides in fluorescent bead binding assays (data not shown). Slot blot assays also showed no detectable binding of E. coli ␣ 5 ␤ 1 or its subunits to gangliosides. These data suggested the importance of the carbohydrate moieties of ␣ 5 in the interaction with the carbohydrate components of ganglioside.
Extracellular N-terminal Region of ␣ 5 Is Recognized by GT1b-To determine the site on ␣ 5 bound by ganglioside, NAB assays were performed in which insect recombinant ␣ 5 integrin with either an intracellular C-terminal His tag or an extracellular N-terminal His tag was linked to the nickel-agarose bead. After the addition of ␤ 1 without its His tag (for ␣ 5 ␤ 1 ) or without the addition of ␤ 1 (for studying the ␣ 5 subunit), the free extracellular N-terminal region or intracellular C-terminal region was allowed to interact with GT1b and with FN in solution. Recovery of protein was determined by Western blotting, showing the binding of the integrin to FN. The interaction of GT1b and the truncated termini of ␣ 5 ␤ 1 or the subunit was determined by thin-layer chromatography immunostaining with anti-GT1b antibody. GT1b was detectable only when the C-terminal His-␣ 5 ␤ 1 or C-terminal His-␣ 5 was studied, demonstrating the direct interaction of GT1b with the extracellular N-terminal region of ␣ 5 .
Gangliosides Preferentially Bind Mannose Groups-To address the possibility that gangliosides recognize specific carbohydrate moieties of the integrin glycoprotein, competition experiments were performed between gangliosides and ConA and between gangliosides and UEA-1 regarding their ability to bind affinity-purified ␣ 5 ␤ 1 from SCC12 cells. Although ␣ 5 ␤ 1 bound well to both ConA and UEA-1, presumably by recognition of the integrin's mannose and fucose residues, respectively, GT1b blocked the binding of ␣ 5 ␤ 1 only to ConA (p Ͻ 0.001) (Fig. 8A), and not to UEA-1 (Fig. 8B). Gangliosides GM3 and GM2 did not inhibit the binding of ␣ 5 ␤ 1 to either lectin. To further consider that GT1b and GD3 recognize and bind to mannose residues, ELISAs were performed with ovalbumin molecules with varying contents of mannose. GT1b and GD3 selectively bound the high mannose form of ovalbumin (p Ͻ 0.01 for GT1b and p Ͻ 0.05 for GD3) (Fig. 9, A and B). Gangliosides GD1a, GM2, and GM1 showed weak binding, and no difference in binding in relation to the mannose content of the sugars was noted with these control gangliosides. DISCUSSION During the last decade, several studies have shown the importance of carbohydrate-carbohydrate interactions as the basis for cell adhesion and recognition. For example, asparaginelinked oligosaccharides of ␣ 6 ␤ 1 integrin are known to interact with metaperiodate-and N-glycanase-sensitive components of laminin (29). Although the greatest interest has focused on glycoprotein-glycoprotein interactions, glycolipids may also participate in these carbohydrate-carbohydrate relationships and have gained increasing attention as possible key modulators. Le X -Le X glycolipid interactions are thought to mediate at least the primary recognition and adhesion process among mouse pre-implantation embryo cells (30). In addition, ganglioside GM3 can interact with several glycosphingolipids with a terminal N-acetylgalactosamine and with lactosylceramide. GM3-Gg3 binding is thought to participate in the interaction of B16 melanoma cells, which express high levels of GM3, and mouse L5178 lymphoma cells, which express high levels of Gg3 (asialo-GM2) (31). The interaction of GM3 with Gg3 or lactosylceramide, expressed on endothelial cells, may also initiate metastasis (32). Most recently, Zheng and Hakomori (33) demonstrated that the interaction of a "disialyl-I" carbohydrate epitope on soluble FN and cell membrane glycosphingolipid Gg3 promotes binding of the cell to soluble fibronectin, suggesting that glycosphingolipid aggregates may interconnect FN and the cell surface through an additional binding mechanism.
Although the initial observation that polysialylated gangliosides inhibit the binding of cells to FN was described in 1979 (14), the mechanism of this effect has been unknown. The demonstration that binding of ␣ 5 ␤ 1 to FN is inhibited in a cell-free system provides further evidence of a direct role of ganglioside interaction with integrin, rather than triggering an Gangliosides GT1b, GD3, GM3, GD1a, and GM2 were coated onto amine-modified 0.2-m FluoSphere beads. Affinity-purified integrin protein from SCC12 cells, recombinant proteins generated in insect cells, or recombinant proteins deglycosylated with PNGase F were coated onto 96-well plates by cross-linking with EDAC. After washing, the beads coated with gangliosides were added to the integrin-coated plate. After a 2-h incubation, binding of the colored ganglioside-coated spheres to integrin was visualized by immunofluorescence microscopy. The number of beads counted in 13 non-overlapping fields was expressed as a percentage of the number of GT1b-bound beads, with 100% representing 293.67 Ϯ 18.65 beads/13 non-overlapping fields. inhibitory effect through an additional cellular component. The strong correlation between the results of slot blot, ELISA, and bead binding assays with affinity-purified ␣ 5 ␤ 1 from SCC12 cells and the results with insect recombinant ␣ 5 ␤ 1 suggests that the recombinant proteins generated in insect cells represent a good model of ␣ 5 ␤ 1 generated endogenously in human keratinocytes. Furthermore, the preferential binding of ganglioside to the ␣ 5 subunit and, similarly, the specific inhibition of ␣ 5 binding to FN by ganglioside are consistent with the previously demonstrated specificity of the inhibitory effect of ganglioside GT1b on binding to a FN matrix and not to several other matrices that are recognized by keratinocyte integrins that share the ␤ 1 subunit (11). The ability of a particular ganglioside to bind to ␣ 5 ␤ 1 correlates well with the inhibitory effect of that ganglioside on ␣ 5 ␤ 1 binding to the RGD-containing fragment of FN. Thus, GT1b binds most strongly and shows the greatest inhibitory effect, whereas GD3 shows both weaker binding to integrin and a lesser effect on inhibition. Other tested gangliosides, including GM3, did not show any inhibitory effect or binding. Given that the carbohydrate moieties distinguish these gangliosides, these data not only show a direct interaction between the gangliosides and integrin, but also suggest that differences in both the degree of sialylation and the pattern of glycosylation affect the ability to bind integrin. Zheng et al. (21) clearly demonstrated that glycosylation of integrin ␣ 5 ␤ 1 is critical for binding to FN; however, the importance of integrin glycosylation for binding of ganglioside to ␣ 5 ␤ 1 has not been addressed. To consider the role of sugar moieties on the integrin in the interaction with ganglioside, we deglycosylated native ␣ 5 ␤ 1 from SCC12 cells prior to extraction from the cells and also cleaved the sugar moieties from recombinant ␣ 5 and ␤ 1 that we generated in insect cells. In addition, we generated poorly glycosylated recombinant ␣ 5 and ␤ 1 in E. coli. All three of these sugar-poor forms of ␣ 5 ␤ 1 showed poor to no staining with periodate-Schiff reagent and an ϳ30% reduction in molecular mass after deglycosylation.
ELISAs with the deglycosylated forms of ␣ 5 ␤ 1 and the poorly glycosylated ␣ 5 ␤ 1 from E. coli reiterated the findings of Zheng et al. (21), demonstrating that binding to FN of deglycosylated integrin decreases by at least 80% in comparison with glycosylated ␣ 5 ␤ 1 . We have now shown that the ganglioside-integrin complex formation, which may participate in the regulation of cell-matrix interactions, also requires the recognition by ganglioside of integrin glycosylation sites in the extracellular Nterminal region, the region of ␣ 5 ␤ 1 involved in RGD binding. Although deglycosylation of integrin also dissociates the ␣ 5 and ␤ 1 subunits from each other (19), GT1b does not affect the ability of integrin subunits to associate.
The interference by GT1b in the binding of ␣ 5 ␤ 1 to Con A, which is known to bind mannose residues (34), and not to UEA-1, which is known to bind to fucose residues, suggests that GT1b recognizes a mannose-containing region of the extracellular domain of N-glycan. The demonstration that GT1b and GD3 bind preferentially to high mannose ovalbumin, but not to ovalbumin with a lower content of mannose, further supports the recognition by these gangliosides of mannose structures. Although the carbohydrate structures of ␣ 5 ␤ 1 from human keratinocytes are unknown, at least 35 different types of N-linked oligosaccharides have been separated and characterized from the ␣ 5 ␤ 1 receptor from human placenta (28). The most common sugar group is the biantennary di-␣-(2,3)-sialylfucosyl residue, and Ͼ50% of the sugars are fucosylated at the N-acetylglucosamine residue at its reducing end; all contain an oligomannose core typical of N-glycans. High mannose residues compose only 1.5% of the sugars of human placenta ␣ 5 ␤ 1 . In contrast, the N-glycosylation patterns of recombinant glycopro- FIG. 9. Ganglioside binding to ovalbumin with different mannose contents. ELISAs were performed to assess the ability of gangliosides to bind the reduced mannose (A) and high mannose (B) forms of ovalbumin. 50 g/ml ovalbumin was coated onto 96-well plates (50 l/well) in the presence of 10 mg/ml EDAC in PBS overnight at 4°C and washed. 2-14 nM GM1, GM2, GD1a, GD3, or GT1b or 1% BSA was added to each well and incubated overnight at 4°C. The plate was then washed vigorously and reacted with anti-GM1, anti-GM2, anti-GD1a, anti-GD3, or anti-GT1b monoclonal antibody (or no primary antibody for BSA), followed by HRP-conjugated secondary antibody and DAB reaction substrate. Ⅺ, GM1; ‚, GD1a; छ, GM2; ϫ, GD3; *, GT1b. were washed, and 10 nM GT1b, GM3, or GM2 or 1% BSA was added to the plate concurrent with affinity-purified ␣ 5 ␤1 from SCC12 cells at concentrations of 10 Ϫ14 to 10 Ϫ6 g/ml. Binding was measured at A 450 nm after treatment with anti-␣ 5 ␤ 1 antibody, HRP-conjugated secondary antibody, and DAB reaction substrate. Ⅺ, ␣ 5 ␤ 1 without ganglioside; ‚, ␣ 5 ␤ 1 ϩ GM3; छ, ␣ 5 ␤ 1 ϩ GM2; *, ␣ 5 ␤ 1 ϩ GT1b. teins expressed in insect cells, although variable, show mostly high mannose-type (Man 5-9 -GlcNAc 2 ) and short truncated structures with fucose ␣1,6-linked to the asparagine-bound GlcNAc residue (35). The binding of ConA to mannose does not distinguish among different manno-oligosaccharides. In contrast, other lectins show greater specificity. For example, Tulipa gesneriana lectin prefers manno-oligosaccharides with Man(␣1-6)Man linkage rather than ␣1-2 or ␣1-3 linkage (36), Galanthus nivalis agglutinin prefers terminal Man(␣1-3)Man (37), and Crocus sativus lectin prefers Man(␣1-3)Man(␤1-4)GlcNAc (38) in the N-glycan core structure. Further study, including competition assays with these lectins that recognize more specific mannose structures, will be required to explore the specificity of ganglioside binding and to explain the recognition by specific gangliosides of the carbohydrate moieties of specific glycoproteins.
Carbohydrate-carbohydrate interactions variably require the presence of divalent calcium. For example, autoaggregation of mouse F9 embryonic carcinoma cells is based on Le X -Le X interaction in the presence of divalent calcium (30), but the carbohydrate-carbohydrate interaction of globoside (Gb4) with nLc4 (precursor of Le X ), GalGb4, and Le X does not require the presence of divalent cation (39). Although divalent calcium is critical to the binding of ␣ 5 ␤ 1 and the ␣ 5 subunit to fibronectin and to the ability of the insect recombinant ␣ 5 and ␤ 1 subunits to complex, 2 our co-immunoprecipitation studies of ganglioside and ␣ 5 ␤ 1 were not performed in the presence of calcium, and ELISA studies in the presence of 1 mM EDTA have shown no effect of depletion of divalent cation on the ability of GT1b to bind to either the 1:1 mixture of insect recombinant ␣ 5 and ␤ 1 subunits or to the ␣ 5 subunit itself. 2 Although these results suggest that the carbohydrate-carbohydrate interaction of complex gangliosides with ␣ 5 ␤ 1 integrin does not require divalent cation, further studies are in progress.
The potential role of membrane gangliosides as "cofactors" that are able to complex with receptors and facilitate or block function is also suggested by ganglioside-receptor interactions other than GT1b (GD3)-␣ 5 ␤ 1 . For example, GD3 and GD2 coimmunoprecipitate with ␣ v ␤ 3 (40); given the structural similarities between the ␣ 5 ␤ 1 and ␣ v ␤ 3 integrins, it is likely that the ganglioside-␣ v ␤ 3 interaction similarly involves carbohydratecarbohydrate recognition of the extracellular N-terminal region of ␣ v . Studies in 1988 first demonstrated that ganglioside GM3 co-immunoprecipitates with the epidermal growth factor receptor (EGFR) (41), and this ability to complex was further suggested by ELISAs (42). In fact, we have recently provided evidence that the relationship between GM3 and the EGFR that is required for inhibition of EGFR activation involves carbohydrate-carbohydrate interaction. 3 Furthermore, we have determined that the basis for the inhibitory effect of GM3 on EGFR activity in SCC12 cells involves, at least in part, the decreased availability of receptors for ligand binding, resulting in decreased phosphorylation of several components of the EGFR signal transduction pathway (43). The lack of any direct interaction of GM3 with integrin ␣ 5 ␤ 1 suggests that the regulation by GM3 of ␣ 5 ␤ 1 /FN-mediated adhesion as described by Zheng et al. (44) relates to other effects of GM3, such as those on signal transduction pathways, rather than a direct binding of ganglioside to the integrin.
In contrast to the strong evidence that keratinocyte GM3 participates in modulating EGFR activity, the role of the more complex gangliosides GD3 and GT1b, present in lesser amounts in intact epidermis (17), cultured keratinocytes, and keratinocyte-derived SCC12 cells, has been unclear. The specific carbohydrate-carbohydrate interactions between gangliosides GD3 and especially GT1b and integrin ␣ 5 ␤ 1 indicate a potentially critical role for these gangliosides in the function of human keratinocytes, which are highly dependent upon the interaction of ␣ 5 ␤ 1 and FN for regulation of cell adhesion, motility, and the ability to undergo apoptosis. Given the potential importance of these sialylated gangliosides in modulating biologic behavior in vivo, further studies on the role of gangliosides in disorders with aberrant adhesion and motility on fibronectin, such as psoriasis and cutaneous carcinomas, are warranted.