GM3-enriched Microdomain Involved in Cell Adhesion and Signal Transduction through Carbohydrate-Carbohydrate Interaction in Mouse Melanoma B16 Cells*

Mouse melanoma B16 cells are characterized by the predominant presence of ganglioside GM3 and adhere to lactosylceramide- or Gg3-coated plates through interaction of GM3 with lactosylceramide or Gg3, whereby not only adhesion but also spreading and enhancement of cell motility occur (Kojima, N., Hakomori, S. (1991) J. Biol. Chem. 266, 17552–17558). We now report that the adhesion process is based essentially on a glycosphingolipid-enriched microdomain (GEM) at the B16 cell surface, since >90% of GM3 present in the original cells is found in GEM, and GEM is also enriched in several signal transducer molecules, e.g. c-Src, Ras, Rho, and focal adhesion kinase (FAK). GEM was isolated as a low density membranous fraction by homogenization of B16 cells in lysis buffer under two different conditions (i.e. buffer containing 1% Triton X-100, or hypertonic sodium carbonate without detergent), followed by sucrose density gradient centrifugation. A close association of GM3 with c-Src, Rho, and FAK was indicated by co-immunoprecipitation of GM3 present in GEM by anti-GM3 monoclonal antibody DH2, followed by Western blotting with antibodies directed to these transducer molecules. The following data indicate that GEM is a structural and functional unit for initiation of GM3-dependent cell adhesion coupled with signal transduction. 1) Tyrosine phosphorylation in FAK was greatly enhanced in B16 cells adhered to Gg3-coated plates but was minimal in cells adhered to GM3-coated, GlcCer-coated, or noncoated plates. 2) GTP loading on Ras and Rho increased significantly when cells were adhered to Gg3-coated plates, compared with GM3-coated, GlcCer-coated, or noncoated plates. Since Ras and Rho are closely associated with GM3 in GEM, cell adhesion/stimulation through GM3 in GEM may induce activation of Ras and Rho through enhanced GTP binding.

To understand the molecular mechanism of GSL function, it is essential to clarify the organizational status of GSLs at the cell surface membrane. We found previously that GM3 is particularly enriched in detergent-insoluble substrate attachment matrix (DISAM), and that GSLs in general are enriched in detergent-insoluble material (DIM), when cell monolayer is treated with zwitterionic detergent (27). Around the same time, Young, Tillack, and co-workers observed that GSLs are clustered to form microdomains at the cell surface (28) or even at the liposome surface (29). Increasing evidence for identity or partial overlap of chemical composition and function between DIM and clustered GSLs, i.e. GSL-enriched microdomains (GEM), has emerged from numerous studies during the past decade. For example, Brown and Rose (30) observed that GSLs and glycosylphosphatidylinositol (GPI) anchors, along with sphingomyelin and cholesterol, are enriched in the apical surface (but not basolateral surface) of epithelial cells and are associated with low-density, detergent-insoluble vesicles equivalent to DIM. Association of DIM with Src family protein ki-* This work was supported by NCI, National Institutes of Health, Outstanding Investigator Grant CA42505 (to S. H.). A preliminary note on a part of this study has been published (59). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Supported in part by the Mizutani Foundation for Glycoscience, Tokyo, Japan.
§ Supported by the Associazione per la Promozione delle Ricerche Neurologiche (Italy).
Specific binding of GM3 to epidermal growth factor receptor tyrosine kinase (4) and of GM1 to Trk A kinase (7) indicates a close association of specific GSLs with specific tyrosine kinases. This idea was further supported by the fact that nonreceptor Src family tyrosine kinase Lyn in rat basophilic leukemia cells is co-immunoprecipitated with ␣-Gal-GM1 (38), and Lyn in rat brain is co-immunoprecipitated with ganglioside GD3 (39). See "Discussion" regarding the functional significance of this association.
The current study on mouse B16 melanoma cells indicates that 1) GEM and DIM are enriched in not only GSLs but also multiple transducer molecules (e.g. c-Src, Rho, Ras, FAK, etc.) and 2) GM3-dependent cell adhesion to Gg3-coated plates through GM3-Gg3 interaction induces signal transduction through changes in transducer molecules associated with GEM. This paper is concerned with changes of FAK and GDP/ GTP binding to Ras and Rho.

GSLs, Transducer Molecules, and Antibodies Directed to Them
GM3 was prepared from dog erythrocytes. Its specific mAb, DH2 (IgG 3 ), was established as described previously (40). Specific polyclonal or monoclonal antibodies suitable for Western immunoblotting and directed to various transducer molecules such as c-Src, Ha-Ras, RhoA, FAK, G␣s, and phospholipase C-␤2 were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Antibodies directed to c-Src, FAK, and phospholipase C-␤2 were rabbit polyclonal; antibody directed to RhoA was mouse monoclonal (IgG 1 ); and antibody directed to Ha-Ras was rat monoclonal (IgG 1 ).

Preparation of GM3-enriched Microdomain from B16 Melanoma
Two methods were employed: one with lysis buffer containing 1% Triton X-100, the other with hypertonic sodium carbonate medium.
Detergent-containing Conditions-Cells were harvested in 0.02% EDTA (ϳ7 mM), lysed, homogenized, and subjected to sucrose density gradient centrifugation to separate low density light-scattering membranous fraction according to a modification of the method described previously (41). Briefly, 1-5 ϫ 10 7 cells were suspended in 1 ml of lysis buffer containing 1% Triton X-100, 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5 mM EDTA, and 75 units of aprotinin and allowed to stand for 20 min. When a phosphorylation assay was included, 1 mM NaVO 4 was added to inhibit phosphatase. The cell suspension was Dounce-homogenized with 10 strokes, and the lysate was centrifuged for 5 min at 1300 ϫ g to remove nuclei and large cellular debris. The supernatant fraction (postnuclear fraction) was subjected to sucrose density gradient centrifugation, i.e. the fraction was mixed with an equal volume of 85% sucrose (w/v) containing 10 mM Tris buffer (pH 7.5), 150 mM NaCl, 5 mM EDTA, with or without 1 mM NaVO 4 (Tris-NaCl-EDTA buffer). The resulting diluent was placed at the bottom of a sucrose concentration gradient with Tris-NaCl-EDTA buffer. Samples were centrifuged for 18 h at 200,000 ϫ g at 4°C. A white light-scattering band under light illumination (presumably due to the Tyndall phenomenon of the membranous fraction) located at ϳ5-7% sucrose was separated from other fractions at different sucrose concentrations (see Fig. 1). The entire procedure was performed at 0 -4°C (in ice immersion). The protein content of each fraction was determined using a MicroBCA kit (Pierce).
Hypertonic Sodium Carbonate Medium-The original method (42) was modified as follows. Cells were scraped in 500 mM sodium carbonate, pH 11.0 (2-4 ϫ 10 7 cells/2 ml) and homogenized using a loose fitting Dounce homogenizer (20 strokes), a Polytron tissue grinder (three 10-s bursts), and a bath sonicator (three 20-s bursts). 1.5 ml of the cell homogenate thus obtained was placed at the bottom of an ultracentrifuge tube and mixed with an equal volume of 90% sucrose (w/v) in 25 mM MES, pH 6.5, 0.15 M NaCl (MBS) and overlaid with 4 ml of 35% sucrose and 4 ml of 5% sucrose (both in MBS containing 250 mM sodium carbonate). Samples were submitted to ultracentrifugation as described above, and the light-scattering band just above the 5-35% sucrose interface was collected and designated as the GEM fraction (see Fig. 1). The protein content of each fraction was determined as described above.

Determination of Distribution of GM3 and Transducer Molecules in Fractions Obtained from Sucrose Density Gradient Centrifugation
GEM and other fractions were separated by sucrose density gradient centrifugation in medium containing detergent or hypertonic sodium carbonate free from detergent, as described under "Preparation of GM3enriched Microdomain from B16 Melanoma." Each fraction was subjected to analysis of GM3 and various transducers.
Determination of GM3-Each fraction was dialyzed against distilled water to eliminate sucrose and lyophilized. The residue was extracted with chloroform-methanol (2:1), placed on a high performance TLC plate, developed, and immunostained using anti-GM3 mAb DH2 and Vectastain ABC kit (Vector, Burlingame, CA) as described previously (44), i.e. using biotinylated goat anti-mouse Ig as secondary antibody. Final TLC staining was made with metal-enhanced DAB substrate (Pierce). The GM3 band on TLC was also examined by orcinol-sulfuric acid staining.
Determination of Transducers-GEM and other protein fractions were subjected to SDS-PAGE with Western immunoblotting; i.e. protein bands separated on SDS-PAGE were transferred electrophoretically to PVDF membranes (Immobilon-P, Millipore Corp., Bedford, MA) in 25 mM Tris, 192 mM glycine, and 15% methanol at 200 mA for 2 h. The membranes were incubated overnight at 4°C in PBS containing 5% defatted milk, washed in PBS containing 0.05% Tween 20, and incubated for 2 h with various antibodies directed to specific transducers. After incubation with primary antibody with appropriate dilution (see Fig. 2 legend), the membrane was washed twice in PBS containing 0.05% Tween 20 and incubated with goat anti-rabbit, anti-rat, or antimouse Ig conjugated with horseradish peroxidase for 45 min for detection of rabbit, rat, or mouse primary antibodies. The membrane was washed five times in PBS containing 0.05% Tween 20 and developed using the chemiluminescence method with a substrate kit (Super-Signal TM -CL-HRP; Pierce).

Co-immunoprecipitation of GM3 and Transducer Molecules in GEM Fraction
Approximately 500 l of GEM (containing 25-30 g of protein) was mixed with protein A-Sepharose beads (50 l, packed) and stirred by a rotary shaker for 2 h at 4°C to preclear nonspecific binding. After centrifugation (500 ϫ g for 1 min), the supernatant was added with 10 -20 l of anti-GM3 mAb DH2 ascites or with 10 -20 l of mouse myeloma SP-2 or NS/1 ascites as negative control. The mixtures were placed overnight in a rotary mixer at 4°C, added with protein A-Sepharose beads (50 l, packed), and placed again in a rotary mixer for 2 h. Beads were washed three times with PBS containing 0.01% Tween 20, by brief weak centrifugation (500 ϫ g for 1 min) and then suspended with 100 l of sample buffer with mercaptoethanol, heated to 95°C for 3 min, and centrifuged (1000 ϫ g for 2 min). The supernatants were subjected to SDS-PAGE, transferred electrophoretically to PVDF membranes, and incubated with various antibodies directed to specific transducer molecules, under the same conditions as described under "Determination of Transducers." Transducer molecules were detected with secondary antibody by the chemiluminescence method as described above.

Stimulation of Cells through GM3-dependent Adhesion
B16 melanoma cells, characterized by abundance of GM3, adhere, spread, and show enhanced motility on Gg3-or LacCer-coated plates; this phenomenon is based essentially on interaction of GM3 on B16 cells with Gg3 or LacCer on the plates (17,18) and is referred to here as "GM3-dependent adhesion." To follow time course changes of signal transduction caused by GM3-dependent adhesion, the following experimental design was used. 770 l of GM3, Gg3, or GlcCer solution (containing 100 g/ml ethanol) was added to 6-cm diameter tissue culture plates and dried. As a control, the same volume of ethanol was added to culture plates and dried. Plates were precoated with GM3, Gg3, or GlcCer, and each plate was added with a 3-ml aliquot of B16 cell suspension in serum-free DMEM (10 6 cells). Several plates were stacked in a special plate holder (adaptable to a Beckman centrifuge), centrifuged at 400 rpm for 1 min at room temperature, and then immediately incubated at 37°C in a CO 2 incubator for 5, 10, 20, 30, or 60 min (defined as duration of cell adhesion). Cells were then placed on ice, harvested by rubber, and centrifuged. Cell pellets were analyzed for the levels of FAK and tyrosine phosphate associated with FAK and for the GTP/GDP ratio associated with Rho and Ras (see below).
To block the effect of interaction of Gg3 (coated on plates) with GM3 (expressed on B16 cells), Gg3-coated plates were incubated with 2 ml (3 g/ml) of anti-Gg3 IgM mAb 2D4 (45) or (as negative control) IgM mAb AH6 directed to unrelated structure Le y (46) for 1 h at room temperature. Plates were washed with PBS, and B16 cells were added by centrifugation and incubated in serum-free DMEM at 37°C for 30 min. Cells were detached, and the levels of FAK and its tyrosine phosphorylation were determined as described above.
Since adherence of B16 cells on Gg3-coated solid phase was inhibited by anti-GM3 (18), the effect of anti-GM3 on levels of FAK and its tyrosine phosphate were determined. The detached cells were suspended in serum-free DMEM at 1 ϫ 10 6 cells/3 ml and incubated for 30 min at 37°C in the presence of 2 g/ml DH2 or normal mouse IgG. Subsequently, the reaction was stopped in an ice bath, and an equal volume of ice-cold PBS containing 1 mM NaVO 4 was added. Cells were lysed, and we followed the procedure to determine FAK and FAKassociated tyrosine phosphorylation as described above.

Determination of Levels of FAK and Its Tyrosine Phosphate
Cells stimulated through adhesion to Gg3-coated plates or control cells added on GM3-or GlcCer-coated plates as described above were lysed in 0.5 ml of lysis buffer (50 mM Tris/HCl, pH 7.4, 150 mM NaCl, 2 mM NaF, 1 mM EDTA, 1 mM EGTA, 1 mM NaVO 4 , 1 mM phenylmethylsulfonyl fluoride, 75 units/ml aprotinin, 1% Triton X-100) by sonicating for 10 ϫ 0.1 s (Sonifier 450, Branson Ultrasonic Corp., Danbury, CT). Aliquots of lysate containing the same quantity of protein (ϳ100 g) were boiled for 5 min in SDS-PAGE sample buffer containing 5% 2-mercaptoethanol and electrophoresed on 8% SDS-PAGE, and electrophoresed proteins were transferred to PVDF membranes (Immobilon-P, Millipore). The PVDF membranes were treated with 5% defatted milk in TBS (10 mM Tris/HCl, pH 7.4, 150 mM NaCl) containing 0.05% Tween 20 for 1 h at room temperature and incubated with rabbit anti-FAK polyclonal IgG (Santa Cruz Biotechnology) for 2 h at room temperature. The membranes were washed three times with TBS containing 0.05% Tween 20, incubated with horseradish peroxidase-conjugated goat antirabbit IgG for 1 h at room temperature, washed five times with TBS/ Tween 20, and developed by the chemiluminescence method with a substrate kit (Super-Signal-CL-HRP; Pierce).
To determine the level of tyrosine phosphate associated with FAK, FAK was isolated from the cell lysate by immunoprecipitation. Briefly, ϳ400 l of cell lysate (containing 20 -30 g of protein) was mixed with protein A/G-agarose beads (10 l, packed; Santa Cruz Biotechnology) and rotated for 2 h at 4°C. After centrifugation (1300 ϫ g for 5 min), the supernatant was added with 2 l of 100 g/ml mouse anti-FAK mAb 77 (IgG 1 , Transduction Laboratories), rotated overnight at 4°C, and then incubated with protein A/G-agarose beads (10 l, packed) for 2 h at 4°C. The mixture was centrifuged at 1300 ϫ g for 5 min, and pelleted beads were washed five times with TBS/Tween 20 and boiled 5 min in SDS-PAGE sample buffer containing 5% 2-mercaptoethanol. The FAK content of each SDS-PAGE sample was measured by Western blotting using an Ultroscan XL densitometer (Amersham Pharmacia Biotech). Samples containing equal amounts of FAK were run on 8% SDS-PAGE and transferred to PVDF membranes. The membranes were probed with horseradish peroxidase-conjugated anti-tyrosine phosphate mAb PY20 (Santa Cruz Biotechnology) and developed using the chemiluminescence method, followed by densitometric analysis.

Determination of the Ratio of GTP/GDP Binding to G-proteins (Ras, Rho)
The original method for determination of GTP and GDP binding to Ras (47) was modified at step 3 as below (treatment of eluate from gel beads), since otherwise its application to Ras and Rho in B16 cells did not give high reproducibility.
Step 1: Metabolic Labeling with 32 P-After culturing of cells in DMEM supplemented with 10% fetal calf serum, medium was replaced with phosphate-free, serum-free DMEM for 4 h. [ 32 P]Phosphate (1 mCi/25-cm 2 flask) was added to the medium under serum-free conditions, and culture was continued for 17 h. Cultured cells were washed five times with TBS, detached with 0.02% EDTA solution, and then resuspended in serum-free DMEM.
Step 2: Cell Adhesion, Cell Lysis, and Adsorption of G-protein on Beads-Three ml of labeled cells (1 ϫ 10 6 cells) were attached by low centrifugation (400 rpm for 1 min) on plates coated with Gg3, GlcCer, GM3, or no coating. The plates were then incubated for 20 min at 37°C, and the reaction was stopped by placement on ice and the addition of 3 ml of ice-cold PBS. Cells were harvested by rubber, suspended in 0.5 ml of HEPES lysis buffer (50 mM HEPES, pH 7.4, 100 mM NaCl, 20 mM MgCl 2 , 10 mM NaF, 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, 1% Triton X-100, 0.5% Nonidet P-40), and Dounce-homogenized 10 strokes. After centrifugation at 1300 ϫ g for 5 min, the supernatant was precleared with 100 l/ml bovine serum albumin-coated charcoal for 10 min at 4°C and then centrifuged again. Supernatants were further precleared by incubation with protein A/G-agarose beads (10 l, packed) for 1 h at 4°C, centrifuged at 1300 ϫ g for 5 min at 4°C, and incubated with 1 g/ml polyclonal or monoclonal anti-Ha-Ras or anti-RhoA antibodies (Santa Cruz Biotechnology), or with 1 g/ml normal mouse IgG as control, overnight at 4°C. The incubation mixtures were added with protein A/G-agarose beads (10 l, packed) and incubated for 2 h at 4°C.
Step 3: Elution and Separation of GTP and GDP-The beads were washed sequentially with buffer A (50 mM HEPES, pH 7.4, 500 mM KCl, 20 mM MgCl 2 ) and buffer B (50 mM HEPES, pH 7.4, 500 mM NaCl, 20 mM MgCl 2 ), resuspended in elution buffer (20 mM Tris/HCl, pH 7.4, 2 mM dithiothreitol, 0.2% SDS, 1 mM GTP, 1 mM GDP), heated at 65°C for 5 min, and centrifuged. The resultant supernatant was treated with 10% trichloroacetic acid and centrifuged at 10,000 ϫ g for 5 min at 4°C. Precipitate was discarded, and supernatant was mixed with 400 l of ether and shaken to extract trichloroacetic acid, and this procedure was repeated twice more. The upper ether layers were discarded, and aqueous lower layers were evaporated by Speed-Vac® (Savant, Farmingdale, NY). The dried residue was dissolved in 10 l of water and spotted on Ecteola cellulose plates, which were then developed with 0.75 M KH 2 PO 4 (pH 3.5). GDP and GTP spots, visualized by fluorescence quenching under ultraviolet light and by autoradiography, were cut, and radioactivity was determined by liquid scintillation counting.

Predominant Presence of GM3 in Low Density Membranous
Fraction-Homogenization of B16 melanoma cells in Tris lysis buffer containing 1% Triton X-100 or hypertonic sodium carbonate solution, followed by sucrose density gradient centrifugation, gave a white light-scattering band in the low density (ϳ5-10% sucrose) region (Fig. 1A). The location of the band was nearly identical for fractions prepared under detergent-containing or detergent-free conditions. Electron micrographs under the two conditions showed the presence of membranous material, although vesicular structures are seen more clearly in GEM fraction prepared by sodium carbonate method, whereas heterogeneous membrane fragments without vesicular structure are seen in GEM fraction prepared by the 1% Triton X-100 method (Fig. 1B).
The pattern of TLC immunostaining of GM3 in each fraction obtained from sucrose density gradient centrifugation in 1% Triton X-100 is shown in Fig. 2. Exactly the same pattern was obtained from sodium carbonate method (data not shown). Over 90% of GM3 in the original cells was found in the low density membranous band prepared under detergent-containing or detergent-free conditions, although the band contained only a small proportion (0.5-2.0%) of total cellular protein.
GM3 was detected only minimally (Ͻ10%) in various other fractions (including fraction 12), which contained Ͼ95% of total protein (Fig. 2). Proportions of protein, GM3, and other transducer molecules present in GEM (fraction 5) and in the major high density, protein-containing fraction (fraction 12) are shown in Table I.
Enrichment of Transducer Molecules in GEM-Almost all c-Src and Ras (Ն90%), ϳ50% of Rho, and ϳ10 -25% of FAK originally present in B16 cells were found in GEM fraction prepared under Triton X-100-containing conditions ( Fig. 3A; Table I). The distribution pattern of c-Src, Rho, and FAK was similar but slightly different for GEM prepared in hypertonic sodium carbonate free from detergent (Fig. 3B). In particular, a higher percentage (ϳ25%) of FAK was present in GEM fraction prepared in sodium carbonate. However, Ras was undetectable in any postnuclear fraction prepared by sodium carbonate method, since Ras became insoluble under this method. Enrichment of c-Src, Rho, and FAK in low density GEM fraction, prepared by two different methods, is remarkable in view of the fact that this fraction contained only 0.5-2.0% of total protein.
Other transducers and receptors such as phospholipase C-␤2, epidermal growth factor receptor, and integrins were absent in low density GEM fraction but present in high density fractions (fractions 10 -12) which contained Ͼ95% of cellular protein.

Association of GM3 and Transducer Molecules in GEM-
Aliquots of GEM were immunoprecipitated with anti-GM3 mAb DH2 and protein A-Sepharose beads. GM3 and its possible complex with transducer molecules present in GEM were eluted from the beads and subjected to Western blotting with various anti-transducer antibodies, as described under "Materials and Methods." With this approach, c-Src and RhoA were detected on Western blotting (Fig. 4, A and B). In control experiments, the addition of mouse myeloma ascites or nonspecific mouse IgG to GEM did not result in detectable levels of c-Src or RhoA (Fig. 4). The band corresponding to FAK (M r 125,000) was detected by Western blotting with anti-FAK rabbit polyclonal antibodies of the eluate from DH2-protein A-Sepharose beads (Fig. 4C).
Adhesion of B16 Cells to Gg3-coated Plates but Not to GM3coated Plates Activates FAK; Effect of Anti-GM3 on FAK-Aliquots of samples containing approximately equal quantities of FAK were prepared from B16 cells adhered to Gg3-coated, GM3-coated, or noncoated plates for various durations. FAK quantity was monitored by anti-FAK binding probed by Western blot analysis. FAK-associated tyrosine phosphorylation of the same aliquots was compared by Western blotting with anti-tyrosine phosphate mAb (Fig. 5A). A great enhancement of tyrosine phosphate level was observed in cells adhered to Gg3coated plates. Blotting activity was much lower for cells placed on GM3-coated or noncoated plates (Fig. 5A). The FAK level in these samples was fairly constant regardless of whether cells were adhered to Gg3-coated, GM3-coated, or noncoated plates (Fig. 5B).
Time course changes of FAK-associated tyrosine phosphate level in cells adhered to Gg3-coated plates as compared with GM3-coated or noncoated plates are illustrated in Fig. 6. Enhancement of tyrosine phosphorylation in FAK occurred within 10 min of cell adhesion based on GM3-Gg3 interaction and remained fairly constant even after 30 min.
To observe the specificity of Gg3-dependent enhancement of tyrosine phosphorylation in FAK, FAK activity probed by antityrosine phosphate mAb was compared between B16 cells adhered to Gg3-coated plates preincubated with three different mouse mAbs: 1) anti-Gg3 IgM mAb 2D4, 2) anti-Le y IgM mAb AH6, and 3) nonspecific mouse IgM. Tyrosine phosphorylation of cells adhered to plates preincubated with anti-Gg3 2D4 was greatly decreased (Fig. 7A, lane 1) relative to cells added on plates coated with irrelevant or nonspecific mouse IgM antibodies (Fig. 7A, lanes 2 and 3), although FAK bands of same volume aliquots from experiments 1 and 2 were nearly identical, or the band from experiment 3 was slightly lower than those from experiments 1 and 2 (Fig. 7B). These results indicate that enhancement of tyrosine phosphorylation in FAK resulting from GM3-dependent B16 cell adhesion to Gg3-coated plates is blocked by anti-Gg3 2D4, whereas irrelevant mAb AH6 or nonspecific mouse IgM do not block this enhancement.
Since B16 cell adhesion on a Gg3-coated plate, coupled with FAK activation, was inhibited when the Gg3-coated plate was preincubated with anti-Gg3 as above, and the adhesion process was inhibited in the presence of anti-GM3 (18), the effect of anti-GM3 on FAK-associated tyrosine phosphorylation in B16 cells was examined. Anti-GM3 per se greatly stimulates FAKassociated tyrosine phosphorylation in Fig. 7, C and D. The results strengthen the idea that GM3 is the site triggering signal transduction to activate FAK.
Effect on Ras and Rho Activity of B16 Cell Adhesion to Gg3-coated Plates-GTP/GDP binding of Ras and Rho, reflecting the status of signal transduction through these G-proteins, was measured in B16 cells adhered to Gg3-coated plates in comparison with GM3-coated, GlcCer-coated, and noncoated plates, since both Ras and Rho are found in GEM and associated closely with GM3. GTP binding to Ras and Rho was increased greatly in cells adhered to Gg3-coated plates and was minimal in cells added to GlcCer-coated or noncoated plates (Fig. 8A). The ratio of GTP/GDP binding to both Ras and Rho, quantitated by a scintillation counter, was Ͼ2.0 when cells were adhered to a Gg3-coated plate but was Ͻ1.0 when cells were added to GlcCer-coated plates. The GTP/GDP binding ratio associated with Rho varied extensively for cells adhered to GM3-coated plates and showed a mean value higher than that for cells adhered to GlcCer-coated plates; in contrast, the GTP/GDP binding ratio associated with Ras was similar for GM3-versus GlcCer-coated plates (Fig. 8B). Thus, for B16 cells adhered to GM3-coated plates, the Rho response was different from the Ras response. DISCUSSION Two categories of glycosphingolipid function in cell membrane have been described: 1) GSLs modulate function of adhesion receptors (e.g. integrin receptor) or are directly involved in cell adhesion as targets for various carbohydrate-binding receptors (e.g. selectin, galectin, Ig family receptors) or for binding by complementary GSL (carbohydrate-carbohydrate interaction); 2) GSLs modulate key molecules involved in signal transduction, such as growth factor receptor-associated kinases and protein kinase C (for a review, see Ref. 1). These two functions occur simultaneously and in close association with each other. Interaction of GSLs with other membrane components is a prerequisite for all the above processes and is strongly affected by the organizational state of GSLs in membrane.
a Percentage quantity of proteins and GM3 present in fraction relative to quantity present in total cellular extract. 5-10% of total GM3 was found in fractions 7-9.
b Relative intensity of immunoblotted band in Western blotting analysis given by the same quantity of protein present in each fraction. The quantity of protein used in a typical analysis was 2.5 g. c ND, not detected. bottom of each panel). Panel A, prepared in medium containing 1% Triton X-100. Panel B, prepared in hypertonic sodium carbonate.

FIG. 3. Detection of transducer molecules in GEM fraction (fraction 5 of Fig. 1) and fraction 12 (the major protein-containing fraction) by Western blotting with respective antibodies (indicated at the
There is increasing evidence that GSLs are not randomly distributed at the cell membrane but rather are self-assembled to form clusters among themselves or with other membrane molecules; these assemblies have hereby been termed GEM (see above). A few lines of study as below support this idea.
1) The first line of study revealed morphological clusters of GSLs at the cell surface membrane, as detected by electron microscopy with the freeze-fracture technique. These GSL clusters are separated from glycoprotein clusters (28) and are also observed in the surface membrane of liposomes (29,48).
2) The second line of study provided a method for isolation of low density membranous fractions, which may include GSL clusters. We showed previously that GSLs are present in baby hamster kidney cells in DIM, and that GM3 is enriched in DISAM (27). Interestingly, purified GSLs are soluble in aqueous buffer containing nonionic or zwitterionic detergent, but they become detergent-resistant when they are organized in membrane, particularly at the substrate attachment site (or focal adhesion site). The majority of GSLs of typical epithelial cells (e.g. intestinal epithelia, Madin-Darby canine kidney cells) are located at the apical surface; a smaller quantity of GSLs are located at the basolateral surface (30,35,49). Nascent protein linked to the GPI anchor is soluble in aqueous buffer containing detergent, but when it is incorporated in apical membrane it becomes detergent-insoluble. GPI-anchored proteins are associated with sphingomyelin, cholesterol, and various other GSLs (30). Thus, DIM or DISAM may include GEM.
3) A third line of study has developed along with studies of caveolae, invaginations of the plasma membrane (50, 51) that play an essential role in endocytosis and signal transduction (33,34). Many studies since 1992 have indicated that caveolae are enriched in GPI-anchored protein, the characteristic pro- FIG. 4. Co-immunoprecipitation of transducer molecules present in GEM fraction with anti-GM3 mAb DH2 (IgG 3 ). Freshly prepared GEM fraction was incubated with Sepharose-protein A conjugate to preclear nonspecific components, incubated with DH2 (or control NS/1 mouse ascites). Sepharose-protein A conjugate was added to and incubated with the reaction mixture. The bound components, eluted and subjected to SDS-PAGE, were Western blotted as described under "Materials and Methods." Panel A, lane 1, immunoprecipitated with DH2 and probed by anti-c-Src (rabbit polyclonal); lane 2, treated with control mouse ascites and Sepharose-protein A conjugate and probed by anti-c-Src; lane 3, control c-Src from GEM fraction. Panel B, lane 1, treated with control mouse ascites and Sepharose-protein A conjugate and probed by anti-Rho (mouse IgG). Lane 2, immunoprecipitated with DH2 and probed by anti-Rho. Lane 3, control RhoA from GEM fraction. Bands with M r ϳ30,000 and above in lanes 1 and 2 in panel B correspond to mouse IgG revealed by secondary anti-mouse IgG antibody, since anti-RhoA mAb used in this experiment was IgG 1 . Panel C, lane 1, treated with control mouse ascites and Sepharose-protein A conjugate, probed by anti-FAK rabbit polyclonal antibodies; lane 2, immunoprecipitated with DH2 and Sepharose-protein A beads, probed by anti-FAK rabbit polyclonal; lane 3, immunoprecipitated with anti-FAK mAb 77 and Sepharose-protein A conjugate, probed by anti-FAK rabbit polyclonal. Weaker bands below ϳ80 kDa may represent mouse IgG H-chain cross-reacting with rabbit polyclonal antibodies. Since the level of FAK present in GEM fraction is much lower than the levels of c-Src and RhoA, conditions necessary to observe FAK bands are more stringent. The weak mouse IgG H-chain cross-reacting bands appear only under these more stringent conditions; they are not seen under less stringent conditions used for rabbit polyclonal anti-c-Src. In panels A, B, and C, locations of c-Src, RhoA, and FAK (respectively) are indicated by labels and arrows at the right.
FIG. 5. Effect of B16 melanoma cell adhesion to Gg3-coated and noncoated plates on levels of FAK-associated tyrosine phosphate, based on equal quantities of FAK (probed in blotting activity with anti-FAK). B16 cells were added by centrifugation (400 rpm for 1 min) on multiple plates precoated with Gg3 or GM3 or on plates treated with ethanol and evaporated (noncoated plates) and incubated for 10, 20, 30, or 60 min (indicated as 1, 2, 3, or 4, respectively, in each panel) at 37°C. Cells were harvested, extracted, and immunoprecipitated with anti-FAK mAb. Protein containing an equal FAK quantity extracted from cells incubated for the durations above was subjected to SDS-PAGE followed by Western blotting. A (top three blots), blotted with anti-tyrosine phosphate mAb. B (bottom three blots), blotted with anti-FAK mAb. The captions none, GM3, and Gg3 apply to both A and B and mean that cells were added to noncoated, GM3coated, and Gg3-coated plates, respectively, and forced to adhere by weak centrifugation as described under "Stimulation of Cells through GM3-dependent Adhesion." Note that FAK-associated tyrosine phosphorylation was stimulated strongly by adhesion to Gg3-coated plates but only minimally by adhesion to GM3-coated plates. tein caveolin (32), and Src family and GTP-binding proteins in addition to GSLs and sphingomyelin/cholesterol (31, 52) (for reviews see Refs. 36 and 37). However, the composition of caveolae is inconsistent. Caveolae from endothelial cells are enriched in GPI-anchored protein but devoid of caveolin; others are enriched in GM1 and GPI-anchored protein (53). Some GEM may be independent of microdomains associated with caveolae, since they are present in cells not containing caveolae (54). GSLs in GEM are involved in cell adhesion, while those in caveolae are not involved in cell adhesion.
The possibility that GEM might be an artifact caused by homogenization in detergent-containing medium has been ruled out by the development of procedures for isolation of GEM or caveolae using detergent-free medium (42,55). An interesting finding of the present study is the nearly identical composition in terms of GM3, c-Src, Rho, FAK, and other transducer molecules found in GEM isolated from B16 melanoma cells in medium containing Triton X-100, compared with GEM isolated from cells in detergent-free hypertonic sodium carbonate medium (Table I). Therefore, GEM is a consistent structural unit representing a microdomain of plasma membrane.
A remarkable feature of GSLs in GEM is their ability to perform two different functions: 1) binding to various GSL ligands (e.g. lectins, antibodies, complementary GSLs); and 2) initiation of signal transduction through activation of GEMassociated transducer molecules. These two functions are closely coupled through organization of GSLs in GEM. As an example, we now demonstrate that binding of GSLs in GEM to complementary GSLs coated on plates initiates activation of a few transducers present in GEM.
A close association of transducer molecules with GSLs was indicated by a few studies; e.g. p53/56 lyn is co-immunoprecipitated with ␣-galactosyl-GD1b (but not other gangliosides, including GD3) in rat basophilic leukemia cells (38) and with GD3 (but not other gangliosides) in rat brain (39,56). The functional significance of the association between gangliosides and p53/56 lyn is largely unclear, although a possible connection was suggested by a transient enhancement of autophosphoryl-ation of Lyn following brief (30 -60-s) exposure of cerebellar cells to anti-GD3 (56).
The present study indicates that GM3-dependent adhesion of B16 cells to Gg3-coated plates induces two remarkable consequences in signal transduction: 1) FAK activation (greatly increased level of FAK-associated tyrosine phosphate) and 2) enhanced GTP binding to Rho and Ras. FAK activation was observed soon (within 30 min) after B16 cells were adhered to Gg3-coated plates but was not observed after the addition of cells to GM3-coated, GlcCer-coated, or noncoated plates. GTP binding to Rho was significantly enhanced when cells were added to GM3-coated plates, to a slightly lesser extent than when added to Gg3-coated plates. These observations suggest that GM3 in GEM, upon interaction with Gg3, may induce signal transduction, although the exact mechanism and consequent phenotypic changes (in motility and morphology) remain to be studied. The major transducer associated with GM3 in  1 and 2) or on Gg3-coated plates (lane 3). Cells were homogenized, precleared with agarose beads, and added with normal mouse IgG (as control) or with anti-Ha-Ras or anti-RhoA antibody together with protein A/G-agarose beads (see "Step 2: Cell Adhesion, Cell Lysis, and Adsorption of G-protein on Beads"), followed by elution (see "Step 3: Elution and Separation of GTP/GDP"). Preparations used for lanes 1 and 2 were protein A/G-agarose beads added with normal mouse IgG (lane 1) or with anti-Ha-Ras or anti-RhoA antibody (lane 2). The preparation used for lane 3 was protein A/G-agarose beads added with anti-Ha-Ras or anti-RhoA antibody, but cells adhered on Gg3-coated plates. In panel A, the location of GTP and GDP spots in the autoradiogram was determined by the addition of cold GTP and GDP in samples, and spots corresponding to GTP and GDP, separated on Ecteola cellulose plate were detected by fluorescent quenching spot. The area corresponding to GTP and GDP is indicated to the right of each autoradiogram. In panel B, 32 P radioactivities associated with fluorescent quenching spots of GTP and GDP were determined by a scintillation counter after cutting the spots, and the ratio of the activities of GTP/GDP was calibrated.
GEM of B16 cells is c-Src. However, changes of c-Src upon GM3-dependent cell adhesion were not studied because of technical difficulties in distinguishing active versus inactive forms of c-Src. It is possible that the majority of c-Src associated with GEM is in inactive form, since Csk, the inhibitor tyrosine kinase for c-Src (57), is detectable in GEM and in fraction 12. Therefore, it is also possible that c-Src may be activated upon interaction of GM3 with Gg3, leading in turn to activation of FAK. The mechanism by which cell adhesion through GM3 in GEM alters transducers in GEM is another major topic to be studied. It may be analogous to the mechanism proposed by Ullrich and Schlessinger (58) for activation of the transducer domain of growth factor receptor tyrosine kinase.
Involvement of GEM in GM3-dependent cell adhesion of B16 cells is obvious because Ͼ90% of GM3 is located in GEM, and the process requires multivalency. Spreading and enhanced motility of B16 cells following GM3-dependent adhesion (18,24) depends on enhanced signal transduction, which is now characterized as enhanced FAK activity and enhanced GTP binding to Rho and Ras. An obvious effect of anti-GM3 DH2 on FAK-associated tyrosine phosphorylation in B16 cells further supports the general idea that GM3 in GEM is the site involved in cell adhesion as well as in initiation of cell signal transduction. Thus, cell adhesion coupled with signal transduction clearly operates through GEM, a structural and functional unit of plasma membrane. Importantly, GM3-dependent adhesion is coupled with motility change (23), and this coupled process mimics B16 cell adhesion to mouse endothelial cells and is regarded as the initial step in B16 melanoma metastasis (24,25).
The function of GSL in GEM may be different from that of caveolae, which are involved in endocytosis and signal transduction, but not in cell adhesion because of their invaginated structure. Integrins and Ig family receptors involved in cell adhesion are not found within caveolae. GSLs in GEM may be targets of carbohydrate-binding proteins (lectins) and antibodies, which may also induce signal transduction. A synergistic effect between cell adhesion based on GSL-GSL interaction and adhesion based on integrin-fibronectin or integrin-laminin interaction has been described (23). It is possible that integrins or Ig family receptors may translocate close to GEM when GSLs in GEM are involved in cell adhesion, whereas such a mechanism may not occur in caveolae. A proposed scheme for function of GEM in comparison with caveolae and in connection with other adhesive receptors is shown in Fig. 9. GEM as a structural unit of adhesion coupled with signal transduction is analogous with growth factor receptor. Whereas growth factor receptor is capable of binding only to a single, specific ligand and transducing a signal based on activation of a specific cytoplasmic tyrosine kinase, GEM may be capable of binding to multiple factors (complementary GSLs, antibodies, selectins, galec- FIG. 9. Schematic illustration of proposed GEM function in cell adhesion coupled with signaling. Top, GSLs are self-assembled in microdomains, which constitute a major component of GEM, and transducers (TDa, TDb, etc.) are associated with these GSL microdomains. Glycoprotein patches (Gp) are physically separate from GSL microdomains, based on electron microscopy with the freeze-fracture technique (28). GSL microdomains are subject to binding by lectins, antibodies (Ab), and complementary GSLs. Bottom, A, GEM is associated with inactive TD (1). Stimulation of GSLs in GEM by GSL ligands (antibody, complementary GSL, and lectin) induces conformational change of TDa, TDb, and TDc, which triggers signal transduction (2). Caveolae (invaginations of plasma membrane) may contain GEM-like components (G), transducers (TDa, TDb, and TDc), growth factor receptors (x), and caveolin (y). They are involved mainly in growth factor receptor-mediated signal transduction and not in cell adhesion (3). B, GEM is involved in cell adhesion through GSL-GSL interaction, which induces conformational change of TDa, TDb, and TDc, triggering signal transduction (1). Integrin receptor and Ig family receptor are involved in cell adhesion, which is synergistic with the adhesion controlled by GEM (2). tins, etc.) and transducing multiple signals (some stimulatory, some inhibitory).