Ganglioside GM3 Blocks the Activation of Epidermal Growth Factor Receptor Induced by Integrin at Specific Tyrosine Sites*

The epidermal growth factor receptor (EGFR) can be activated by both direct ligand binding and cross-talk with other molecules, such as integrins. This integrin-mediated cross-talk with growth factor receptors partic-ipates in regulating cell proliferation, survival, migration, and invasion. Previous studies have shown that ligand-dependent EGFR activation is inhibited by GM3, the predominant ganglioside of epithelial cells, but the effect of GM3 on ligand-independent, integrin-EGFR cross-talk is unknown. Using a squamous carcinoma cell line we show that endogenous accumulation of GM3 disrupts the ligand-independent association of the integrin (cid:1) 1 subunit with EGFR and results in inhibition of cell proliferation. Consistently, endogenous depletion of GM3 markedly increases the association of EGFR with tyrosine-phosphorylated integrin (cid:1) 1 and promotes cell proliferation. The ligand-independent stimulation of EGFR does not require focal adhesion kinase phosphorylation or cytoskeletal rearrangement. Stimulation of EGFR and mitogen-activated protein kinase signaling by GM3 depletion involves the phosphorylation of EGFR at tyrosine residues 845, 1068, and 1148 but not 1086 or 1173. The specific blockade of phosphorylation at Tyr-845 with Src family kinase inhibition and at Tyr-1148 with phosphatidylinositol 3-kinase inhibition suggests that GM3 inhibits integrin-induced, ligand-independent EGFR EGFR Phosphorylation— To determine the specificity of the effects of GM3, the effects of modulation of GT1b, another epidermal ganglioside that affects the interaction of FN with cells, were evaluated. GT1b expression was endogenously up-regulated via stably transfected GM2/GD2 synthase cDNA into SCC12 cells using an inducible system as previously described (23). The increase in GT1b after endogenous modification was documented by ganglioside enzyme- linked immunosorbent assay (23). The effects of increased GT1b expression on the integrin-dependent EGFR phosphorylation were assessed by immunoprecipitation and immunoblotting as described above. The effect of increased GT1b on the association of integrin (cid:2) 1 subunit and EGFR was also examined as described.

Integrins are cell surface-adhesive receptors formed by ␣ and ␤ subunits, which bind to extracellular matrix proteins. Integrin-mediated adhesion to extracellular matrix triggers intracellular signaling pathways to modulate cell proliferation, shape, migration, invasion, and survival (for review see Refs. 1 and 2). Integrin signaling is mediated by intracellular molecules, such as c-Src, small GTPases, adaptor molecules such as Shc, the protein-tyrosine phosphatases SHP-2, and phosphatidylinositol 3-kinase (3)(4)(5). Integrins can also cross-communicate with growth factor receptors, enabling growth factor receptor signaling upon extracellular matrix binding of the interacting integrin in the absence of growth factors (1) and culminating in enhanced cell mitogenesis and oncogenesis. Although the mechanism for this cross-talk is poorly understood, this cross-talk does not require cytoskeletal mobility or focal adhesion kinase (FAK) 1 activation (6), suggesting interaction at the membrane level, proximal to FAK and cytoskeletal activation. In fact, certain growth factor receptors have been shown to interact physically with specific integrins, suggesting that the formation of complexes at the membrane level is required for the convergence of signaling pathways (6 -11). EGFR, for example, forms a complex with ␤ 1 integrin after cells attach to fibronectin (6,12), both platelet-derived growth factor receptor-␤ and vascular endothelial cell growth factor receptor 2 show association with ␣ v ␤ 3 integrin (13,14), and ErbB2 associates with ␣ 6 ␤ 1 or ␣ 6 ␤ 4 (15).
Using genetic manipulation of GM3 content, we have investigated the effect of ganglioside modulation on EGFR-integrin cross-talk. To further examine the molecular mechanism of ligand-independent, integrin-promoted EGFR activation modulated by GM3, we have evaluated the effects of GM3 on the integrin ␤ 1 -EGFR association and have addressed the possibility that GM3 expression affects EGFR phosphorylation at specific residues. These studies show that endogenous alterations in GM3 expression influence the association of the EGFR and integrin ␤ 1 subunit and the resultant activation of the EGFR promoted by FN or type I, IV, and VII collagen, all triggers of ␤ 1 integrin activation. Furthermore, these investigations provide evidence that phosphorylation of specific tyrosine residues on the EGFR is impacted by GM3 content modulation, including tyrosine residues that require Src kinase and phosphatidylinositol 3-kinase signaling for phosphorylation.
Overexpression of GM3 by Treatment with Antisense Oligodeoxynucleotides-Membrane content of GM3 on SCC12 cells was endogenously increased as described previously (24) by treatment concurrently of SCC12 cells with antisense oligodeoxynucleotides of both GM2/GD2 synthase and GD3 synthase, leading to blockade of synthetic pathways downstream of GM3.
Proliferation Assays-Proliferation was assessed by MTT analysis. Briefly, cells were seeded in 96-well plates at 2 ϫ 10 3 cells per well and incubated for 6 h in 10% FBS-containing medium. After starvation of serum, growth factors, and FN for 18 h, cells were grown in the presence of either 10% FBS or 10 g/ml human FN (cellular FN, Sigma). Medium was replaced every 24 h, and cells were stained with MTT reagent per the manufacturer's instructions (Roche Applied Science). Absorbance was read at A 560 nm .
Immunoblotting-Immunoblotting was carried out as described (19,21) using protein extracted from either whole cell lysates or immunoprecipitates and an enhanced chemiluminescence detection system (PerkinElmer Life Sciences). In brief, cells were treated with or without antisense oligodeoxynucleotides or were stably transfected with either sialidase cDNA in a pcDNA3 vector or GM2/GD2 synthase cDNA using an RU-486-inducible system (33, 34) as previously described (18,(22)(23)(24). Cells were plated onto 6-well cell culture plates precoated with or without 5 g/cm 2 FN, 20 g/cm 2 type I or IV collagen, 10 g/cm 2 type VII collagen, or 10 g/cm 2 poly-L-lysine for 0, 10, 30, and 60 min after starvation overnight of serum, FN, and growth factors. Cells were grown in the presence of oligodeoxynucleotides or 100 nM RU-486. In other studies cells were grown in the presence or absence of 250 nM AG1478, an inhibitor of EGFR-related kinase; 20 M LY294002 (Calbiochem), a specific phosphatidylinositol 3-kinase inhibitor (23,35); 3 M PP1 (Alexis Biochemicals), a specific inhibitor of Src kinase (23,36); 0.4 M cytochalasin D (Sigma), a potent inhibitor of actin filament polymerization (37,38); 20 g/ml anti-integrin ␤ 1 subunit blocking or stimulatory antibody (Chemicon); or FAK antisense oligodeoxynucleotide (23) (see details in the legend to Fig. 5). Incubation of cells with oligodeoxynucleotides was continued in cell culture throughout the treatment with inhibitors. After treatment with these inhibitors, starved cells were stimulated with 10 g/ml FN for 10 min. Total protein was quantified by colorimetric assay (Bio-Rad) to ensure equal loading. Ten to 20 g of total protein from the whole cell lysates or an aliquot of immunoprecipitates was boiled in Laemmli buffer (39) and loaded onto SDS-PAGE mini-gels. After transfer to polyvinylidene difluoride or nitrocellulose membranes, the separated proteins were detected by immunoblotting with anti-phosphotyrosine (PY20), anti-EGFR, anti-phospho-EGFR, anti-FAK, or anti-␤ 1 antibodies (Transduction Laboratories), anti-p44/42-mitogen-activated protein kinase (MAPK) or anti-phospho-p42/44-MAPK (Calbiochem), antiphospho-EGFR-845, -1068, -1148, -1086, or -1173 (BIOSOURCE), or anti-phospho-FAK-397 (Upstate) antibody. Blots were reprobed as previously described (21) with anti-actin (Santa Cruz Biotechnology) antibody to confirm equal loading. All blots were repeated in at least three different experiments.
Immunoprecipitation-Cells were prepared as indicated above. After starvation of serum, FN, and EGF overnight, cells were plated on 6-well cell culture plates precoated with or without either 5 g/cm 2 FN, 20 g/cm 2 type I or type IV collagen, or 10 g/cm 2 type VII collagen for 10 min. Cells were harvested and lysed in cold immunoprecipitation buffer as previously described (19,20). Total protein (1 mg) from the cell lysates was mixed with either 5 g of anti-␤ 1 integrin subunit or anti-EGFR polyclonal antibody, and the total reaction volume was adjusted to 1 ml in the immunoprecipitation buffer. After incubation with the antibodies for 2 h at 4°C, protein A-agarose was added, and the immunoprecipitate was incubated for an additional 30 min at 4°C (19).
Effects of Ganglioside GT1b on the Association of Integrin ␤ 1 and EGFR and Integrin-dependent EGFR Phosphorylation-To determine the specificity of the effects of GM3, the effects of modulation of GT1b, another epidermal ganglioside that affects the interaction of FN with cells, were evaluated. GT1b expression was endogenously up-regulated via stably transfected GM2/GD2 synthase cDNA into SCC12 cells using an inducible system as previously described (23). The increase in GT1b after endogenous modification was documented by ganglioside enzymelinked immunosorbent assay (23). The effects of increased GT1b expression on the integrin-dependent EGFR phosphorylation were assessed by immunoprecipitation and immunoblotting as described above. The effect of increased GT1b on the association of integrin ␤ 1 subunit and EGFR was also examined as described.

Induction of Endogenous Changes in Ganglioside Content-
Ganglioside GM3 was increased 1.9-fold after SCC12 cells were treated with antisense oligodeoxynucleotides of both GM2/GD2 synthase and GD3 synthase as detected by TLC immunostaining and ganglioside enzyme-linked immunosorbent assays (24). No detectable ganglioside was found on the cell membrane after cells were stably transfected with human plasma ganglioside-specific sialidase (22,24).
Endogenous Modulation of GM3 Regulates Proliferation of Cells-By 72 h after incubation in serum-containing medium, statistically significance differences (p Ͻ 0.05) in proliferation were noted between both sialidase overexpressors and their vector control cells and GM3 overexpressors (antisense oligomer-treated cells) and their control SCC12 cells (Fig. 1A). When cells were grown in the absence of serum or growth factors but in the presence of FN, a statistically significant increase in proliferation of ganglioside-depleted cells versus control cells was noted within 24 h (Fig. 1B); the discrepancy in growth was further accentuated by 72 h of incubation (p Ͻ 0.001). Similarly, proliferation of GM3-overexpressing cells was significantly suppressed by 72 h incubation in the presence of FN (Fig. 1B).
Endogenous Modulation of GM3 Regulates Ligand-independent EGFR and MAPK Phosphorylation-Exposure to FN promoted ligand-independent EGFR phosphorylation ( Fig. 2, middle row, lane 2), which was inhibited by endogenous accumulation of GM3 (Fig. 2, middle row, lane 6) and facilitated by ganglioside depletion (Fig. 2, middle row, lane 8). Modulation of ganglioside GM3 content did not affect EGFR expression (Fig. 2, top row). Blockade of integrin ␤ 1 with anti-␤ 1 -blocking antibody prevented EGFR phosphorylation in the presence of FN, including in ganglioside-depleted cells (Fig. 2, bottom row, lanes 2, 4, and 8). To determine the specificity of the effects of GM3 on ligand-independent, integrin-dependent EGFR activation, the effects of modulation of GT1b, another epidermal ganglioside that affects specifically FN-induced in-tegrin activation, were compared with the effects of changes in content of GM3. Control cells plated on FN (Fig. 3A) or on types I (Fig. 3B), IV, or VII collagen (not shown), but not on poly-Llysine (not shown), showed EGFR phosphorylation between 10 FIG. 1. GM3 overexpression by antisense treatment inhibits cell proliferation, whereas ganglioside-depletion facilitates cell proliferation. To accumulate GM3 endogenously, SCC12 cells in Dulbecco's modified Eagle's medium/F-12 with 10% FBS were treated with antisense oligodeoxynucleotides of both GM2/GD2 synthase (synthesizes GM2 from GM3) and GD3 synthase (synthesizes GD3 from GM3), effectively blocking GM3 metabolism (24). Other SCC12 cells were stably transfected with human plasma ganglioside-specific sialidase gene to eliminate membrane ganglioside as described before (22,24). Parental SCC12 cells, sense-treated cells, and vector-transfected cells served as controls. Cell proliferation was detected by MTT assay as described under "Experimental Procedures." In brief, cells were plated on 96-well plates and starved of serum, growth factors, and FN overnight before incubation in Dulbecco's modified Eagle's medium/F-12 medium with either 10% FBS (A) or 10 g/ml FN (B). The absorbance of each cell line was read at A 560 nm after staining with MTT reagents. Oligodeoxynucleotides were added daily throughout the study. and 30 min (Fig. 3, A and B, second and third rows). Endogenous accumulation of GM3 by antisense treatment inhibited EGFR phosphorylation in the face of either FN or collagen I (Fig. 3, A and B, lane 3 of the second and third rows) in comparison with untreated or sense-treated control SCC12 cells (Fig. 3, A and B, lanes 1 and 2 of the second and third   FIG. 2. GM3 modulates ligand-independent EGFR phosphorylation without affecting EGFR expression. Cells were prepared as indicated in Fig. 1. After starvation of serum, FN, and growth factors overnight in the presence or absence of the oligodeoxynucleotides, cells were plated onto 6-well cell culture plates precoated with or without 5 g/cm 2 FN in the presence or absence of 20 g/ml anti-integrin ␤ 1 subunit-blocking antibody and incubated for 10 min. Cells were then treated with boiled lysis buffer (1% SDS, 1 mM Na 3 VO 4 , 10 mM Tris-HCl, pH 7.4) for 10 min, and insoluble cell debris was removed by brief centrifugation at 1000 rpm for 5 min. Ten (top row) to 15 g (middle and bottom rows) of total protein from the whole cell was applied to a 7.5% SDS-PAGE mini-gel, and EGFR expression (top row) or EGFR phosphorylation (middle and bottom rows) was detected with anti-EGFR or anti-phospho-EGFR monoclonal antibody, respectively.
FIG. 3. GM3 blocks ligand-independent EGFR phosphorylation in the face of FN and collagen I. SCC12 cells were treated with or without sense or antisense oligodeoxynucleotides of both GM2/GD2 synthase and GD3 synthase to accumulate ganglioside GM3 endogenously (24). Other SCC12 cells were stably transfected with human plasma ganglioside-specific sialidase gene to eliminate membrane ganglioside or its pcDNA vector alone (22,24) or were stably transfected with human GM2/GD2 synthase cDNA to increase GT1b content endogenously using an RU-486 inducible system (33,34) or its vector, p17 ϫ 4-tkA/pGL-VP, and induced with 100 nM RU-486 as described (23). After starvation of serum, FN, and growth factors overnight in the presence of either oligodeoxynucleotides or RU-486, cells were plated onto 6-well cell culture plates precoated with or without either 5 g/cm 2 FN (A) or 20 g/cm 2 type I collagen (Col I; B) and incubated for up to 60 min. Cells were then treated with cold lysis buffer as described under "Experimental Procedures" at 4°C for 2 h, and insoluble cell debris was removed by brief centrifugation at 1000 rpm for 5 min. EGFR was immunoprecipitated (IP) from the whole cell lysate as described before (18,19), and equal amounts of immunoprecipitates from each group were applied onto a 7.5% SDS-PAGE mini-gel. The phosphorylation of EGFR was detected with anti-phosphotyrosine (PY20) antibody. The purity of the immunoprecipitate was confirmed by applying an aliquot of immunoprecipitate to probe with anti-EGFR monoclonal antibody; equal loading was confirmed by re-probing each membrane with anti-actin monoclonal antibody (not shown). GM3 Disrupts Integrin-induced EGFR Activation rows). Ganglioside depletion stimulated EGFR phosphorylation regardless of matrix (Fig. 3, A and B, lanes 5 and 6 of the second and third rows) in comparison with parental SCC12 (lane 1) and pcDNA mock-transfected (lane 4) cell controls. The effects of GM3 modulation when cells were plated on collagen IV or VII were similar to those of plating on collagen I (data not shown).
Endogenous accumulation of GT1b by GM2/GD2 synthase overexpression inhibited EGFR phosphorylation only in the presence of FN (Fig. 3A, lanes 9 and 10 of the second and third rows) in comparison with p17 ϫ 4-tkA/pGL-VP mock control cells (lane 7) or p17 ϫ 4-tkA/GM2/GD2-pGL-VP 1 clonal cells without RU-486 induction (lane 8); accumulation of GT1b had no effect on phosphorylation of EGFR in the presence of collagens I (Fig. 3B, lanes 9 and 10 of the second and third rows), IV, or VII (not shown). The time course of EGFR activation and return to base-line status was not altered by ganglioside manipulation.
Endogenous accumulation of GM3 inhibited, whereas ganglioside depletion stimulated the activation of MAPK (Fig. 4,  lanes 1, 3, 5, and 7) in cells exposed to FN. Inhibition of EGFR kinase activity with 250 nM AG1478 reversed the increase in phosphorylation of MAPK induced by GM3 depletion (Fig. 4,  lane 6) to the base-line level of activation in SCC12 and pcDNA mock controls (Fig. 4, lanes 2 and 6).
The Effect of Ganglioside Modulation on Ligand-independent EGFR Activation Does Not Require FAK Phosphorylation or Cytoskeletal Reorganization-Treatment of cells plated on FN (Fig. 5) or type I collagen (not shown) with cytochalasin D, a potent inhibitor of actin filament function, for 10 min to 4 h had no effect on the ligand-independent stimulation of EGFR phosphorylation facilitated by ganglioside depletion (Fig. 5A, lanes  7 and 8 of the bottom row) or inhibition of EGFR phosphorylation caused by GM3 accumulation (Fig. 5A, lanes 3 and 4 of  the bottom row). In contrast, treatment with cytochalasin D for as short a duration as 10 min eliminated the phosphorylation of FAK at tyrosine 397 (Fig. 5B, bottom row), which was markedly increased in ganglioside-depleted cells and reduced in cells with increased GM3 expression (Fig. 5B, lanes 3 and 7 of the  bottom row). The expressions of EGFR and FAK were unaffected by ganglioside modulation or treatment with cytochalasin D (Fig. 5, A and B, top rows).
Functional Blockade of FAK with Antisense Oligomer Does Not Affect the Stimulation by Ganglioside Depletion of Ligandindependent EGFR Activation-FAK function of SCC12 cells that overexpress sialidase was blocked with antisense oligodeoxynucleotides of FAK. In the continued presence of FAK antisense oligodeoxynucleotides, ganglioside-depleted cells in the presence of FN (Fig. 5C) or on collagen I (not shown) showed marked inhibition to the absence of FAK phosphorylation (Fig.  5C, top row) but no change in the increased tyrosine phosphorylation of the EGFR (Fig. 5C, bottom row).
GM3 Disrupts the Association of EGFR and the Integrin ␤ 1 Subunit-Ganglioside depletion facilitated the co-immunoprecipitation of EGFR and integrin ␤ 1 subunit when cells were incubated in the presence of either FN (Fig. 6A, lanes 5 and 6, third row) or type I collagen (Fig. 6B, lanes 5 and 6, third row) for 10 min in comparison with parental SCC12 cells or pcDNA mock controls (Fig. 6, A and B, lanes 1 and 4 of the third row). This association diminished after 30 min (Fig. 6, A and B, lanes 5 and 6 of the fourth row) and totally disappeared after 60 min (Fig. 6, A and B, lanes 5 and 6 of the bottom rows). Consistently, endogenous accumulation of GM3 disrupted the co-immunoprecipitation of EGFR and integrin ␤ 1 subunit in the face of either FN (Fig. 6A, lane 3 of the third row) or type I collagen (Fig. 6B, lane 3 of the third row) in comparison with untreated or sense-treated controls (Fig. 6, A and B, lanes 1 and  2 of the third rows). Accumulation of GT1b only affected the association of EGFR and integrin ␤ 1 subunit in the presence of FN (Fig. 6A, lanes 9 and 10 of the third row) but not in the presence of collagen I (Fig. 6B, lanes 9 and 10 of the third row) in comparison with parental SCC12 cells, mock vector controls, and GM2/GD2 synthase-transfected cells without RU-486 induction (Fig. 6, A and B, lanes 1, 7, and 8 of the third row).
Activation of Both the Integrin ␤ 1 Subunit and EGFR Kinase Activity Are Required for the Association of ␤ 1 Integrin and EGFR on Cell Membrane Promoted by Ganglioside Depletion-Blockade of either integrin ␤ 1 subunit activation with anti-␤ 1 blocking antibody or EGFR kinase activity with the kinase inhibitor AG1478 disrupted the association of the integrin ␤ 1 subunit and EGFR (Fig. 7, A and B, third and bottom rows). Activation of the integrin ␤ 1 subunit with anti-␤ 1 stimulatory antibody facilitated this association (Fig. 7, A and B, fourth row) in comparison with untreated cells (Fig. 7, A and B, top  row). Treatment with cytochalasin D for 10 min to 4 h did not affect the association of integrin ␤ 1 and the EGFR that is stimulated by ganglioside depletion in the presence of FN (Fig.  7, A and B, lanes 3 and 4 of the second rows) or collagen I (not shown).
GM3 Modulates EGFR Phosphorylation at Specific Phosphorylation Sites-Endogenous accumulation of GM3 inhibited EGFR phosphorylation at the 845, 1068, and 1148 residues (Fig. 8, lane 3 of the top, second, and third rows) in comparison with untreated or sense-treated control cells (Fig. 8, lanes 1 and  2 of the top, second, and third rows). In contrast, ganglioside depletion facilitated EGFR phosphorylation at the 845, 1068, and 1148 residues (Fig. 8, lanes 5 and 6 of the top, second, and third rows) in comparison with parental SCC12 cells and pcDNA mock control cells (Fig. 8, lanes 1 and 4 of the top,  second, and third rows). Ganglioside expression did not significantly affect the ligand-independent, FN-stimulated EGFR phosphorylation at either the 1086 or 1173 residues (Fig. 8,  fourth and bottom rows). FIG. 4. The increased phosphorylation of p42/p44 MAPK by GM3 depletion is partially ablated by blockade of EGFR signaling. Cells prepared as indicated in Fig. 1 were treated with or without the EGFR inhibitor, 250 nM AG1478, as described under "Experimental Procedures." After starvation of serum, FN, and growth factors overnight in the continued presence of AG1478, cells were plated onto 6-well cell culture plates precoated with 5 g/cm 2 FN and incubated for 10 min. Cells were then treated with boiled lysis buffer as described in Fig. 2 for 10 min, and insoluble cell debris was removed by brief centrifugation at 1000 rpm for 5 min. Twenty g of total protein from the whole cell lysate was applied onto 12% SDS-PAGE mini-gel, and MAPK phosphorylation was detected with anti-phospho-MAPK polyclonal antibody.

Src and PI3 Kinase Signaling Differentially Impact the Phosphorylation of Specific EGFR Phosphorylation Sites Induced by
Ganglioside Depletion-Blockade of Src kinase activity with 3 M PP1 dramatically decreased ganglioside depletion-facilitated EGFR phosphorylation at the 845 residue (Fig. 9, top row), whereas blockade of PI3 kinase activity with 20 M LY294002 significantly inhibited ganglioside depletion-promoted EGFR phosphorylation at the 1148 residue (Fig. 9, bottom row). Neither Src nor PI3 kinase activation was required for ganglioside depletion-promoted, ligand-independent EGFR phosphorylation at the 1068 residue (Fig. 9, middle row). DISCUSSION Cross-communication between integrins and growth factor receptors is thought to be required for maximal activation of the Ras-MAPK signal transduction pathway that drives cell proliferation. How integrin and growth factor receptor signaling are integrated proximal to MAPK is largely unknown. In this study we demonstrate a regulatory role for ganglioside GM3 in ligandindependent, matrix-dependent inhibition of EGFR and MAPK phosphorylation leading to modulation of epithelial cell proliferation. The matrix-dependent effects on EGFR phosphorylation of manipulation of GM3 expression are seen when cells are plated on a variety of matrices that activate integrin ␤ 1 , suggesting modulation of ␤ 1 integrin-EGFR cross-talk.
Existing models for Src and FAK function in integrin signaling have located FAK upstream of Src kinase in integrin signaling (40). FAK autophosphorylation at Tyr-397 has been proposed to recruit Src kinase through its SH2 domain, with stabilization of the Src-FAK interaction through the SH3 domain of Src kinase. However, mutant cells lacking Src kinases show little induction of tyrosine phosphorylation of FAK after integrin stimulation, suggesting that Src kinase is upstream of FAK (41). Furthermore, sites on Src SH2 and SH3 domains, although key for binding to FAK and other proteins involved in cell motility, appear to have little effect on FAK tyrosine phosphorylation (42). Our studies show that the effect of ganglioside depletion on integrin-induced EGFR phosphorylation persists despite either lack of functional FAK or inhibition of cytoskeletal rearrangement. In contrast, blockade of EGFR kinase activation by AG1478 significantly diminishes the increased phosphorylation of MAPK induced by ganglioside depletion to the level of that of the cell without ganglioside depletion, providing further evidence that extracellular matrix-induced growth factor receptor activation plays a significant role in FIG. 5. Cytochalasin D disrupts integrin-dependent phosphorylation of FAK but not EGFR. A and B, cells were prepared as described in Fig. 1. After starvation of serum, FN, and growth factors overnight in the presence of oligodeoxynucleotides, cells were treated with 0.4 M cytochalasin for 10 min to 4 h; 10 g/ml FN was added in the final 10 min. Cells were then treated with boiled lysis buffer as described in Fig. 2 for 10 min, and insoluble cell debris was removed by centrifugation at 1000 rpm for 5 min. Ten (top rows) to 15 g (bottom rows) of total protein from the whole cell lysate was applied onto 7.5% SDS-PAGE mini-gel for EGFR detection (A) or 10% SDS-PAGE mini-gels for FAK detection (B). The expression of EGFR or FAK was detected with anti-EGFR (A, top row) or anti-FAK (B, top row) monoclonal antibody. The phosphorylation of EGFR or FAK was detected with anti-phospho-EGFR (A, bottom row) or anti-phospho-397-FAK (B, bottom row) antibody. C, cells were incubated with antisense oligodeoxynucleotides of FAK to block the function of FAK as described before (23). In brief, synthetic oligodeoxynucleotides were generated complementary to either the sense or the antisense strand of the 20 nucleotides encoding human FAK at the 5Ј end (68), including the initiator codon ATG (5Ј-ATGGCAGCTGCTTACCTTGA-3Ј, FAK sense oligodeoxynucleotide; 5Ј-TCAAGGTAAGCAGCTGCCAT-3Ј, FAK antisense oligodeoxynucleotide). Cells grown in 6-well cell culture plates were incubated with 30 M sense or antisense oligodeoxynucleotides of FAK in serum-free Dulbecco's modified Eagle's medium/F-12 medium for 30 min before 10% FBS was added. The oligodeoxynucleotides were refreshed every other day for 5 days. Cells were then trypsinized and transferred into new 6-well cell culture plates precoated with 5 g/cm 2 FN. After 10 min of incubation with FN in the continued presence of sense or antisense oligodeoxynucleotides of FAK, cells were treated with boiled lysis buffer as described in Fig. 2, and 20 g of post-nuclear lysates were applied for immunoblotting. Phosphorylation of FAK (top row) and EGFR (bottom row) were determined as described for A and B.
adhesion-induced MAPK activation and that the activation of ERK by ganglioside depletion is EGFR-dependent. These results are consistent with the demonstration by Moro et al. (6) that treatment with cytochalasin D of ECV304 human endothelial cells plated on collagen I had no effect on adhesioninduced tyrosine phosphorylation of the EGFR, although it dramatically reduced phosphorylation of FAK.
Ganglioside depletion specifically phosphorylates and GM3 overexpression inhibits phosphorylation of three specific EGFR tyrosines in the absence of EGFR ligand. Two of these, Tyr-1068 and Tyr-1148, are major EGFR autophosphorylation sites (43)(44)(45)(46), yet two other EGFR autophosphorylation sites (Tyr-1086 and Tyr-1173) are not impacted by modulation of GM3 content. The third site at which ligand-induced EGFR phosphorylation is regulated by ganglioside is Tyr-845, a unique site that is not phosphorylated by EGFR ligands (47)(48)(49) but is the site of binding and phosphorylation of the EGFR by Src kinase (48). Trans-activation by Src kinase of this EGFR Tyr-845 site is considered critical to the matrix-induced stimulation of mitogenesis and tumorigenesis (48 -57). Consistently, the   7. The association of integrin ␤ 1 and EGFR requires both integrin ␤ 1 and EGFR kinase activities. Cells that overexpress ganglioside-specific sialidase (22), pcDNA vector controls, and parental SCC12 control cells were treated with or without 0.4 M cytochalasin D for 30 min, 20 g/ml anti-integrin ␤ 1 blocking or stimulatory antibody for 4 h, or 250 nM AG1478 for 12 h. Integrin ␤ 1 subunit or EGFR was immunoprecipitated (IP) from undenatured cell lysate with either anti-integrin ␤ 1 subunit or anti-EGFR polyclonal antibody as described before (18). The association of integrin ␤ 1 with EGFR was detected with either anti-EGFR (A) or anti-integrin ␤ 1 subunit (B) monoclonal antibody.

GM3 Disrupts Integrin-induced EGFR Activation
absence of Src kinases profoundly suppresses matrix-induced tyrosine phosphorylation (41). Although phosphorylation by c-Src at Tyr-845 occurs with both EGF-and integrin-mediated EGFR activation (47), Tice et al. (49) has recently shown that mutation at Tyr-845 does not affect its kinase activity in response to EGF, suggesting that phosphorylation at the Tyr-845 site may play a more critical role in integrin-induced EGFR activation. Given a role for Src kinase binding to the EGFR in its ganglioside-modulated effects on EGFR phosphorylation and induction of mitogenesis, these data suggest that Src kinase functions upstream of FAK in ligand-induced EGFR activation and that Src kinase binding to FAK is not required for EGFR phosphorylation.
Adhesion to matrix induces the formation of a macromolecular complex that includes the EGFR, ␤ 1 integrin, Src kinase, and the adaptor protein p130Cas, but not FAK (10). This coimmunoprecipitation strictly requires matrix adhesion and the presence of EGFR; it is disrupted by treatment with PP1, a specific Src kinase inhibitor, and is markedly diminished in c-Src Ϫ/Ϫ fibroblasts, suggesting that Src kinase and EGFR are both required for complex formation. The activation of the EGFR by ganglioside depletion at sites that are activated by Src kinase (Tyr-845) and phosphatidylinositol 3-kinase (Tyr-1148) and the associated increase in EGFR-␤ 1 integrin coimmunoprecipitation suggests that a complex of the EGFR and ␤ 1 integrin together with the Src kinase and phosphatidylinositol 3-kinase is required for cross-talk. Our data showing that increased GM3 expression inhibits the association of ␤ 1 integrin and the EGFR further provides evidence that ganglioside acts as a disruptor of complex formation. The lack of effect on the EGFR-␤ 1 integrin association of treatment of cells with cytochalasin D further suggests that cytoskeletal reorganization through FAK signaling is not necessary for EGFR-␤ 1 integrin complex formation. The complex formation, however, requires that both ␤ 1 integrin and EGFR kinase be activated given that treatment with either anti-integrin ␤ 1 blocking antibody or AG1478 to block EGFR kinase function prevents the association of ␤ 1 and EGFR.
What is the mechanism of ganglioside inhibition of EGFR-␤ 1 integrin cross-talk, and how is it stimulated by ganglioside depletion? The assembly of transduction complexes is thought to involve caveolin-1, a transmembrane protein that acts as a scaffolding protein to aggregate growth factor receptors, integrins, and Src kinases, thus promoting signaling. Caveolin-1 has been shown to participate in integrin-induced Shc tyrosine phosphorylation, leading to Ras-MAPK activation (58). Furthermore, depletion of caveolin-1 has been shown to disrupt the association of Src kinases with ␤ 1 integrins, resulting in loss of adhesion (59). Trans-activation of the EGFR by another stimulus, angiotensin II, has recently been shown to occur in caveolar domains together with tyrosine-phosphorylated caveolin-1 and c-Src (60). We have noted by coimmunoprecipitation studies that GM3 is able to complex with caveolin-1, EGFR, and Src in SCC12 cells. 2 We propose that GM3 interferes with integrininduced EGFR phosphorylation through the effect of ganglioside on caveolin-1 function and disruption of complex formation. Four means of ganglioside interference with caveolin-1 function are possible. First and least likely is that ganglioside may be able to bind to caveolin-1 or a complex component directly, thus sterically preventing signaling. Second, ganglioside GM3 may shift caveolin-1 out of its complex with EGFR, ␤ 1 integrin, and Src kinases, thus inhibiting signaling. We have previously shown that, in the face of EGF, overexpression of GM3 shifts caveolin-1 from the detergent-insoluble caveolar domains to detergent-soluble domains (24). Shifts in caveolin-1 localization and the effect of ganglioside modulation on caveolin association with other molecules in the absence of EGF, but 2 X.-Q. Wang, P. Sun, and A. S. Paller, unpublished results.
FIG. 8. GM3 modulates ligand-independent EGFR phosphorylation at 845, 1068, and 1148 residues. Cells were incubated in the presence of FN for 10 min after starvation of serum, FN, and growth factors, and total protein extracts were prepared as described in Fig. 1. Twenty g of total protein from the whole cell lysate was applied onto 7.5% SDS-PAGE mini-gels. The phosphorylation of EGFR at specific residues was detected with anti-EGFR-phospho-845, -1068, -1086, -1148, and -1173 antibodies.
FIG. 9. Src kinase or phosphatidylinositol 3-kinase activity is required to trigger GM3 depletion-promoted ligand-independent EGFR phosphorylation at 845 or 1148 residues. SCC12 cells were stably transfected with or without human plasma ganglioside-specific sialidase gene in pcDNA vector (22). Cells were treated with or without either 3 M Src kinase inhibitor PP1or 20 M phosphatidylinositol 3-kinase inhibitor LY294002 as described before (23). After starvation of serum, FN, and growth factors overnight in the presence of either PP1 or LY294002, cells were stimulated with 10 g/ml FN for 10 min. Total protein from whole cell lysates was prepared by treating cells with boiled lysis buffer as indicated in Fig. 2 for 10 min and centrifuging to remove insoluble cell debris. 20 g of total protein from the whole cell lysate was applied onto a 7.5% SDS-PAGE mini-gel, and the phosphorylation of EGFR promoted by ganglioside depletion at specific residues was detected with anti-EGFR-phospho-845, -1068, and -1148, respectively. presence of matrix, have not been explored. Third, GM3 may facilitate the phosphorylation of caveolin-1 at its Tyr-14 site, which has been shown to recruit Csk (C-terminal Src kinase) and thereby inactivate Src kinase to interrupt signaling (61). Although the possibility of alteration of caveolin-1 phosphorylation by GM3 has not yet been explored for ligand-induced EGFR phosphorylation, we have previously demonstrated that overexpression of GM3 in the presence of EGF leads to increased Tyr-14 phosphorylation of caveolin-1 (24). Finally, human ganglioside-specific sialidase, which cleaves the sialic acid residue of GM3 and, thus, regulates its content, has recently been shown to colocalize with and bind to caveolin-1 as well (62). Although the function of sialidase with respect to caveolin-1 action remains unknown, GM3 may disrupt the caveolin-1 association with sialidase, whereas overexpression of the sialidase itself may abet the potential caveolin-modulated signaling between integrin and the EGFR. These alternative possibilities deserve further exploration.
The up-regulated expression of EGFR and integrins on tumors has been correlated with accelerated tumor cell proliferation and with increased tumorigenesis (7,63,64). GM3 ganglioside has previously been shown to inhibit ligand-induced EGFR phosphorylation and proliferation of both normal keratinocytes and, more so, squamous carcinoma cells (18,20,24,(65)(66)(67). Given the important role of matrices, particularly a fibronectin substratum, in encouraging tumor cell spread, the demonstration that gangliosides also control matrix-induced cell proliferation, inhibiting Src kinase-modulated phosphorylation of the EGFR, gives further impetus to the development of targeted anti-ganglioside therapy for cancers, particularly epithelial carcinomas.