Reconstitution of Membranes Simulating “Glycosignaling Domain” and Their Susceptibility to Lyso-GM3*

GM3 ganglioside at the surface of mouse melanoma B16 cells is clustered and organized with signal transducer molecules c-Src, Rho A, and focal adhesion kinase (FAK) to form a membrane unit separable from caveolae, which are enriched in cholesterol and caveolin but do not contain GM3 or the above three signal transducers. The GM3-enriched membrane units are involved in GM3-dependent cell adhesion coupled with activation of c-Src, Rho A, and FAK and are termed the “glycosphin-golipid signaling domain” or the “glycosignaling domain” (GSD). In order to assess the essential components that display GSD function, membranes with properties similar to those of GSD were reconstituted using GM3, sphingomyelin, and c-Src, with or without other lipid components. The reconstituted membrane thus prepared displayed GM3-dependent adhesion to plates coated with Gg3 or anti-GM3 antibody, resulting in enhanced c-Src phosphorylation (c-Src phosphorylation response). This response in reconstituted membrane depends on GM3 concentration and was not observed when GM3 was absent or replaced with other gangliosides GM1 or GD1a, precleared Cross-linking between anti-mouse IgG and anti-c-Src IgG at the bead di-methyl octyl glucoside, c-Src glycine-HCl glucoside, immediately addition M The effect of GM3 with gangliosides (GM1, GD1a) was using a lipid m GM1 m of GD1a, as and 3 m g (50 pmol) of c-Src. Membrane reconstitution was performed from lipid/c-Src mixtures with various compositions in 50 m M octylglucoside and 1 m M EDTA by extensive dialysis followed by sucrose density gradient centrifugation, low membrane the as Fr. 5 as

GM3 ganglioside in mouse melanoma B16 cells is clustered and organized with c-Src, Rho A, and focal adhesion kinase (FAK) 1 and causes GM3-dependent cell adhesion coupled with activation of these signal transducers (1)(2)(3), leading to enhanced cell motility and invasiveness (4,5). The term "glycosphingolipid signaling domain" or "glycosignaling domain" (GSD) has been assigned to such structural and functional membrane unit (3,6). GSD appears to be a basic membrane unit found in various types of cells closely associated with GSL function in terms of antigenicity and cell adhesion/recognition coupled with signal transduction (7)(8)(9). However, such units may have been overlooked among (or mixed up with) other membrane units having similar properties such as cholesterolrich, caveolin-containing units (caveolae) involved in endocytosis and signal transduction (10), sphingomyelin (SM)/cholesterol-rich microdomains termed "rafts" (11), or those containing glycosylphosphatidylinositol (GPI) anchors bearing a number of functionally well-defined receptors (12; for review see Ref. 13).

Preparation of c-Src
Mouse c-Src cDNA in pUSE amp was purchased from Upstate Biotechnology (Lake Placid, NY), excised by NotI ϩ HindIII, cloned into a pFastBac HTa NotI-HindIII site, and propagated in insect SF9 cells (ATCC, Rockville, MD) using a baculovirus system (Bac-to-Bac; Life Technologies, Inc.). Infected cells were grown for 4 days, harvested, and washed with phosphate-buffered saline (PBS). Cells were resuspended (10 7 cells/ml) in PBS containing 1 mM diisopropyl fluorophosphate, 10 g/ml chymostatin, 10 g/ml leupeptin, and 100 g/ml aprotinin. Sonication was performed on ice in a sonifier (Sonifier 450; Branson Equip-ment Co., Shelton, CT) 30 times with 0.5-s intermittent pulses (50% "Duty Cycle" at output 2.5). The sonicate was centrifuged at 1000 ϫ g for 5 min, the resulting supernatant was centrifuged at 8000 ϫ g for 20 min, and the second supernatant was centrifuged at 100,000 ϫ g for 60 min to obtain the membrane fraction. The membrane fraction was solubilized by sonicating again for 30 s with 0.5-s intermittent pulses as above in PBS containing 17.1 mM (ϳ0.5%) octyl ␤-D-glucoside, 1 mM diisopropyl fluorophosphate, 10 g/ml chymostatin, 10 g/ml leupeptin, and 100 g/ml aprotinin. The solubilized materials were precleared with anti-mouse IgG-coated Dynabeads (M450; Dynal Inc., Lake Success, NY) and then incubated with Dynabeads conjugated covalently with mouse anti-c-Src monoclonal IgG (B12, Santa Cruz Biotechnology, Santa Cruz, CA) overnight at 4°C. Cross-linking between anti-mouse IgG and anti-c-Src IgG at the bead surface was performed with dimethyl pimelimidate (Sigma). After extensive washing with sodium phosphate buffer (pH 7.4) containing 0.5 M NaCl and 0.5% octyl glucoside, c-Src was eluted with 0.2 M glycine-HCl buffer (pH 3) containing 0.5% octyl glucoside, and pH was immediately adjusted to 7.4 by addition of 1 M Tris. Buffer composition of the c-Src solution was changed by applying it to a NAP-10 desalting column (Amersham Pharmacia Biotech) equilibrated with 50 mM Tris-HCl (pH 7.4), 0.14 M NaCl, 1 mM EDTA, and 50 mM (ϳ1.4%) octyl glucoside. This buffer composition was used for reconstitution of membrane vesicles with properties similar to those of GSD.

Reconstitution of the Membrane with Composition and Properties Similar to Those of GSD
For membrane reconstitution, types of lipids, their proportions, and c-Src quantity were based on those observed originally in GSD fraction (3). Reconstitution of membranes with standard lipid composition, and varying quantities of GM3, other gangliosides, and c-Src, was performed according to the method used for reconstituting the PC-cholesterol membrane with transmembrane receptor (19,20), with some modification as described below.
Varied Composition of GM3, Other Gangliosides, and c-Src in Reconstituted Membranes-To observe the effect of GM3 concentration on c-Src activation response of reconstituted membrane, various quantities of GM3 (from 0.86 to 110 g) were mixed with a fixed quantity of PC (10 g), cholesterol (6.4 g), and SM (22 g) in chloroform-methanol (2:1, v/v), dried, dissolved in TBS containing 50 mM octyl glucoside and 1 mM EDTA as above, and added with a constant quantity of purified c-Src (3 g; 50 pmol). The effect of c-Src quantity on the c-Src activation response of the reconstituted membrane was studied using a lipid mixture containing 55 g of GM3 and fixed PC/cholesterol/SM quantities as described above to which different quantities of c-Src (3 g; 50 pmol; 1.5 g; 25 pmol; or 0.75 g; 12.5 pmol) were added. The effect of replacing GM3 with other gangliosides (GM1, GD1a) was studied using a lipid mixture containing 71.8 g of GM1 or 85.4 g of GD1a, fixed PC/cholesterol/SM quantities as above, and 3 g (50 pmol) of c-Src. Membrane reconstitution was performed from lipid/c-Src mixtures with various compositions in 50 mM octylglucoside and 1 mM EDTA by extensive dialysis followed by sucrose density gradient centrifugation, and the low density membrane component with the same position as Fr. 5 was collected as described above.

Determination of Reconstituted Membrane Components
Lipid Composition of Reconstituted Membranes-Lipids were extracted from reconstituted membrane suspension (1 ml), separated on density gradient centrifugation with 10 ml of methanol and 20 ml of chloroform, sonicated for 15 min, and then centrifuged. The supernatant lipid extract was removed, and the precipitate was extracted again with 10 ml of chloroform/methanol (2:1). The first and second extracts were combined and evaporated, and the residue was dissolved in methanol/water (3:7, v/v), applied to a Bond Elut-packed C18 column (Analytichem International, Harbor City, CA), and washed with the same solvent. Finally, lipids were eluted with chloroform/methanol (2:1). Aliquots of total lipid from reconstituted membrane (10 -20% of total) were separated by thin layer chromatography (either one-or twodimensional). The quantity of lipids and their ratio were determined by densitometry in comparison with a known quantity of standard lipid using the Scion Image program (Scion Corporation, Frederick, MD) as described previously (3,9). Lipid components of reconstituted membrane were compared with those from total detergent-insoluble membranes (DIM) and from the GSD fraction separated by anti-GM3 mAb DH2 (3).

Determination of c-Src Phosphorylation Response to Adhesion of Reconstituted Membrane and Effects of Lyso-sphingolipid Derivatives
Basic Procedure for Determination of c-Src Phosphorylation Response-The reconstituted membrane fraction, purified by sucrose density gradient centrifugation, corresponding to Fr. 5 was 10ϫ diluted with kinase buffer (30 mM HEPES (pH 7.5), 10 mM MgCl 2 , 2 mM MnCl 2 , 1 mM CaCl 2 ). The c-Src phosphorylation response of membranes, adhered to 10-cm dishes coated with Gg3 or anti-GM3 antibody, was determined as described previously (3). Briefly, 5-ml aliquots of the diluted solution as above were added to coated dishes, followed by centrifugation at 1500 ϫ g (3000 rpm) at 4°C for 30 min to ensure partial adhesion of membrane to dishes. The dishes were kept on ice for 15 h with gentle rocking (10 strokes per minute). Separate aliquots of diluted Fr. 5 were also mixed with lyso-GM3, its analogues, or other glycosyl-Sph (Me 2 SO as vehicle; final concentration 0.5%). Reconstituted membranes with these reagents, incubated for 30 min at 37°C, were centrifuged and treated in the same way as above.
The c-Src activity of the reconstituted membrane and its response to GM3-dependent membrane adhesion to the Gg3-coated dish were determined as described previously for c-Src activity of "DIM membrane fraction" (3) with slight modification, and activity response was compared with that of reconstituted membrane with or without preincubation in Dulbecco's modified Eagle's medium (DMEM) containing lyso-GM3 or other lyso-compounds, using Me 2 SO as vehicle. Briefly, reconstituted membrane fraction (ϳ5 ml) with or without adhesion to the Gg3-coated dish was added with 2-l aliquots of 20 Ci of [␥-32 P]ATP (corresponding to 370 GBq/mmol, NEN Life Science Products) and 5 l of 1% digitonin in Me 2 SO in order to achieve final digitonin concentration of 0.001%. This digitonin concentration makes membrane vesicles permeable but does not destroy membrane structure. The mixture was then incubated at 37°C for 5 min, and the reaction was stopped by placing on ice and by addition of 5 ml of stop buffer (15 mM HEPES (pH 7.5), 150 mM NaCl, 5 mM EDTA, 1 mM Na 3 VO 4 , 1% Triton X-100, 1 mM PMSF). The soluble fraction was transferred to a centrifuge tube (part 1), and the adhered membrane was desorbed and solubilized in RIPA buffer (part 2). Parts 1 and 2 were combined in the test tube (these were termed "solubilized reconstituted membrane" fractions). The c-Src fraction was separated preliminarily by 10% trichloroacetic acid and further immunoprecipitated in RIPA buffer, followed by SDS-PAGE with autoradiography as described previously (3).
Alternative Method for Comparative c-Src Response in Reconstituted Membrane with Varying Content of GM3, Other Gangliosides, and c-Src-The c-Src phosphorylation response of reconstituted membranes having various quantities of GM3 and c-Src was determined by membrane adhesion to Gg3-coated dish. Each response was simultaneously quantified by 32 P radioactivity present in immunoprecipitate with antic-Src antibody, without separation of c-Src by SDS-PAGE.
The solubilized reconstituted membrane fractions prepared from 32 Plabeled c-Src in the phosphorylation reaction mixture in each dish were precipitated with trichloroacetic acid at a final concentration of 10%. The precipitates were centrifuged, washed twice with acetone to eliminate trichloroacetic acid, dissolved in 1.0 ml of RIPA buffer, pH checked and adjusted to 7.4, mixed with 20 l of protein G-Sepharose, and placed in a rotary mixer at 4°C for 2 h. After centrifugation at 270 ϫ g for 5 min, the supernatants were collected and added with anti-c-Src goat IgG at a final IgG concentration of 1 g/ml, incubated at 4°C overnight, added with 20 l of protein G-Sepharose, and incubated at 4°C for 2 h. Next, Sepharose beads were washed five times with RIPA buffer, and the remaining radioactive counts were measured using a scintillation counter (Beckman LS6500 Scintillation System, Fullerton, CA). Cancer Center, University of Texas, Houston, TX and cultured in DMEM supplemented with 10% fetal calf serum. GM3 clusters and intensity expressed at the melanoma B16 cell surface were observed by immunofluorescence with anti-GM3 mAb DH2 as described previously (17), and the pattern was digitally imaged using a DeltaVision microscope (Applied Precision, Inc.) or Leica TCS-SP confocal laser scanning microscope in the Image Analysis Lab at the Fred Hutchinson Cancer Research Center. The degree of immunofluorescence intensity was also determined by flow cytometry using mAb DH2.

Imaging of GM3 Expression in Mouse Melanoma B16/F10 Cells
Effects of lyso-GSL compounds on GM3 imaging were determined as follows. Detached cells were washed 2ϫ with DMEM, and 2.5 ϫ 10 6 cells per ml of DMEM were mixed with DMEM solution containing lyso-GM3, Lac-Sph, diLac-Sph, SA-Sph, or lyso-PC (each at 5 M concentration), or 50 M NeuNdcAc lyso-GM3 (Me 2 SO as vehicle; final concentration 0.5%). After 30-min incubation, cells were washed 1ϫ with DMEM and subjected to immunofluorescence followed by microscopy or flow cytometry as described above.

Change of FAK in B16 Cells and of c-Src in Reconstituted Membranes Associated with GM3-dependent Adhesion
The GM3-dependent adhesion of B16 cells to Gg3-coated dishes and the associated change of FAK were determined as described previously (2). The enhanced c-Src activity of the DIM membrane fraction in response to GM3-dependent adhesion of membrane to the Gg3-coated dish was determined as described previously (3). The effects of preincubation with Lac-Sph and lyso-GM3 on FAK response of the B16 cells associated with GM3-dependent cell adhesion were determined (see the legend of Fig. 7A) as well as the c-Src phosphorylation response associated with GM3-dependent adhesion of DIM membrane (see the legends of Figs. 7 and 8).

Reconstituted Membranes
Simulating the Glycosignaling Domain (GSD)-Although the lipid composition and associated signal transducer molecules in GSD of mouse melanoma B16 cells are well documented and distinct from those of caveolincontaining fraction (caveolae) (3), the components essential for GSD function are not clearly identified. To better identify these components, a mixture of membrane lipids and c-Src was processed to obtain reconstituted membranes showing GM3-dependent cell adhesion and associated activation of c-Src, by the procedure described under "Materials and Methods." Lipid composition of reconstituted membranes, separated as low-density fraction, is shown in Fig. 2A, in comparison with the composition of the DIM fraction and the GSD membrane separated by the anti-GM3 antibody (DH2) from 2 ϫ 10 8 melanoma B16 cells. The lipid composition of reconstituted membranes shown in Fig. 2A was based on 55 g GM3, 22 g SM, 10 g PC, 6.4 g cholesterol, and 3 g c-Src. The quantity of GM3 and SM recovered in reconstituted membrane, separated by sucrose density gradient centrifugation, was ϳ22% and ϳ43%, respectively, of the original GM3 and SM used for membrane reconstitution. It is noteworthy that the proportions of GM3, SM, and PC in the reconstituted membrane fraction were very similar to those of the GSD fraction separated by anti-GM3, but distinctively different from those in total DIM (Fr. 5) prepared from B16 cells. Phosphatidylserine and phosphatidylethanolamine were present in the DIM fraction but absent in the reconstituted membrane because we did not include these phospholipids. c-Src was clearly present in reconstituted membrane and was co-immunoprecipitated with GM3 (Fig.  2B). This finding indicates a close association of these components in the reconstituted membrane, similar to GSD.

c-Src Phosphorylation Response of Reconstituted Membrane
Coupled with GM3-dependent Adhesion-The c-Src phosphorylation response was observed in three types of reconstituted membrane containing GM3 adhered to the Gg3-coated dish, i.e. membrane consisting of GM3/SM/c-Src (Fig. 3A, lane 3 9). In all cases, c-Src phosphorylation was not (or minimally) observed when the membrane was added on the GlcCer-coated dish (Fig. 3A,  lanes 2, 4, 6, 8), or placed in nonadherent polypropylene tube (lane 1). The degree of c-Src phosphorylation induced by adhesion of reconstituted membrane to Gg3-coated versus GlcCer-coated dishes was quantified by densitometric scanning of autoradiography bands (Fig. 3B). The c-Src phosphorylation response occurred only in membrane containing GM3/SM/c-Src and was higher following addition of PC or PC/cholesterol. Optimal response was observed for the composition GM3 (55 g), SM (22 g), PC (10 g), and cholesterol (6.4 g) (referred to as "SGPC" and hereby termed "standard composition"). There was no response for the membrane with LacCer in place of GM3 (SLPC).
Strong c-Src phosphorylation was also observed when GM3/ SM/PC/cholesterol/c-Src-reconstituted membrane was adhered to the DH2-coated dish (Fig. 3C, lane 3). No response was observed for membrane without GM3, i.e. LacCer/SM/PC/cholesterol/c-Src (lane 5), or placed in a nonadherent polypropylene tube (lane 1), or on a dish coated with normal mouse IgG (lanes 2 and 4).

Effect of Varying Quantities of GM3 and c-Src, and Substitution of GM3 by Other Gangliosides, on c-Src Phosphorylation
Response of Reconstituted Membrane-The c-Src phosphorylation response of the reconstituted membrane was optimal with the standard composition and c-Src quantity of 3 g (50 pmol) as described above. The response decreased significantly when GM3 quantity was decreased, and decreased slightly when GM3 quantity was doubled (from 55 to 110 g). When the c-Src quantity was decreased to 1.5 g (25 pmol) or 0.75 g (12.5 pmol), response also decreased significantly (Fig. 4A).
When GM3 in reconstituted membrane (with a fixed quantity of other lipids and c-Src) was replaced with GM1 or GD1a and the membrane was placed on the Gg3-coated dish, c-Src phosphorylation response was very low (Fig. 4B). This is consistent with the previous observation that only GM3, not GM1, interacts with Gg3 (4).
Effects of Lyso-GM3 and Its Analogues on GM3 Expression Pattern and on GM3-dependent B16 Cell Adhesion-In a search for specific compounds that disrupt GSL clusters, which play a central role in maintenance of structure and function of GSD, lyso-GSLs and their derivatives were considered to interrupt cis-interaction between GSLs (see "Discussion"). The effects of various lyso-compounds (lyso-GM3, NeuNdcAc lyso-GM3, SA-Sph, Lac-Sph, diLac-Sph, psychosine, lyso-PC) on GSD function, in terms of GM3-dependent B16 cell adhesion associated with FAK enhancement and c-Src phosphorylation response, were studied initially.
Preincubation of B16 cells with 0.5-10 M synthetic lyso-GM3 followed by washing with DMEM strongly inhibited the subsequent adhesion of cells to the Gg3-coated dish. Similarly, preincubation of cells with NeuNdcAc lyso-GM3 inhibited this adhesion at much higher concentrations (5-50 M). In contrast, psychosine, lyso-PC, Lac-Sph, and diLac-Sph had no inhibitory effect even at 10 M (Fig. 5A). Filipin at 0.07-0.47 M and nystatin at 16 -54 M, which at these concentrations destroy caveolae function, did not inhibit this GM3-dependent cell adhesion (Fig. 5A). Based on the reduced viability of cells determined by Trypan Blue exclusion test, lyso-GM3 had no cytotoxic effect up to 10 M but became cytotoxic at Ͼ20 M. NeuNdcAc lyso-GM3 had no cytotoxic effect even at 100 M (Fig. 5B).
The inhibitory effect of 1-5 M lyso-GM3 or 50 M NeuNdcAc lyso-GM3 on GM3-dependent B16 cell adhesion is ascribable to reduced anti-GM3 mAb binding, as was observed clearly by confocal microscopy (Fig. 6A) and by fluorescent microscopy with digital imaging (data not shown). The expression pattern was not changed when cells were pretreated with 1-20 M lactosyl-Sph or 1-10 M lyso-PC. Pretreatment of cells with 50 M NeuNdcAc lyso-GM3 clearly reduced GM3 expression ob-served by flow cytometry (Fig. 6B), and anti-GM3 mAb DH2 did not cross-react with this compound (Fig. 6C). The great reduction of GM3 expression by pretreatment of cells with even higher concentrations of lyso-GM3 (5-10 M) was not due to release of GM3 from the cell surface, because this treatment did not change the cell GM3 content (Fig. 6D).
Effects of Lyso-GM3 and Its Analogues on Cellular FAK Activity and on Membrane c-Src Activity Associated with GM3dependent Adhesion-An increase of FAK activity associated with the GM3-dependent adhesion of B16 cells to the Gg3coated dish was observed as described previously (2). This  4 g), and c-Src (3 g; 50 pmol) and with a varying quantity (as indicated) of GM3 (solid bar) were adhered to dishes coated with Gg3 or GlcCer, and c-Src phosphorylation response was determined as described in the text. Ordinate, % response relative to response of reconstituted membranes with standard composition (i.e. SGPC, containing 55 g of GM3). Response of standard reconstituted membranes with reduced c-Src quantity, 1.5 g (25 pmol) or 0.75 g (12.5 pmol), is shown by striped bars. Each bar shows the mean Ϯ S.D. of three experiments. B, response of membranes containing GM1 or GD1a instead of GM3 with fixed quantity of other lipids and c-Src as above. The binding substrate is GlcCer or Gg3, as indicated. Ordinate as the same as that in A. effect was blocked by preincubation of B16 cells with 1-5 M lyso-GM3 or 25-50 M NeuNdcAc lyso-GM3 followed by washing and adhesion but not by preincubation with 10 M lactosyl-Sph (Fig. 7A).
The great enhancement of c-Src activity in low density membrane fraction (DIM fraction 5) upon adhesion of the membrane to the Gg3-coated dish was inhibited by 1-10 M lyso-GM3 or 50 M NeuNdcAc lyso-GM3 but was unaffected by 10 M lactosyl-Sph (Fig. 7B).
Susceptibility of Reconstituted Membrane to Lyso-GM3 Is Similar to That of GSD of B16 Cells-Analogous to GSD of native B16 cells (see above), c-Src phosphorylation response in the reconstituted membrane was strongly blocked in the presence of 1 or 5 M lyso-GM3 (Fig. 8, lanes 6 and 7), 50 M NeuNdcAc lyso-GM3 (lane 5), or 1 or 5 M SA-Sph but was maintained when the membrane was incubated with GM3 or Lac-Sph (lanes 3 and 4) or psychosine (lane 8).
These results indicate that the reconstituted membrane containing GM3, SM, and c-Src as basic components was capable of c-Src activation response to GM3-dependent adhesion and the response was disrupted by lyso-GM3 and its analogue, similarly to naturally occurring GSD in melanoma B16 cells. After incubation with a second antibody, cells were washed 3ϫ with PBS, fixed with 1% paraformaldehyde (4°C, 30 min), washed 1ϫ with PBS, placed on a chamber slide (Nunc Inc., Naperville, IL), affixed by centrifugation (900 rpm, 3 min), mounted with Fluorogard reagent (Bio-Rad), and subjected to confocal microscopy. B, flow cytometry of B16 cells treated with 50 M NeuNdcAc lyso-GM3 (peak 2) as compared with untreated cells (peak 3) and control cells without antibody (peak 1). Cells treated in the same way as for the adhesion assay were washed 3ϫ with PBS, incubated with PBS containing mAb DH2 (1 g/ml) at 4°C for 1 h, washed 2ϫ with PBS, incubated with PBS containing fluorescein isothiocyanate goat anti-mouse IgG at 4°C for 1 h, and subjected to flow cytometry. C, reactivity of anti-GM3 mAb DH2 with GM3 (spot 1), lyso-GM3 (spot 2), NeuNdcAc lyso-GM3 (spot 3), Lac-Sph (spot 4), and diLac-Sph (spot 5). Dot blot analysis was performed using a "Bio-dot" apparatus (Bio-Rad) as described previously (3). D, B16 cells (2.5 ϫ 10 6 ) were incubated in 1 ml of DMEM containing 0.5% Me 2 SO without (as control) or with 10 M lyso-GM3 (final concentration) for 30 min at 37°C. Cells were centrifuged, and the cell pellet (of the same protein quantity) was extracted using chloroform/methanol (2:1). The total GSL fraction was developed on thin layer chromatography with orcinol-sulfuric acid spray. Note that both control and lyso-GM3-treated cells show the same GM3 duplet. Other bands are not identified.
GSLs at the cell surface (22) and the insolubility of GSLs in buffer solution containing neutral or zwitterionic detergent (12,23). A number of subsequent studies indicates that sphingolipids and cholesterol are associated with GPI anchors and Src family kinases to form structural and functional domains (for reviews see Refs. 11 and 13), similar to membranes derived from caveolae, which are involved in endocytosis and signal transduction (for review see Ref. 10). However, little attention has been paid to the functional notion of GSLs in either sphingolipid/cholesterol domains or caveolae.
Our previous studies provide evidence that clustered GM3 or other GSLs, organized with c-Src, Rho A, and FAK at the surface of mouse melanoma B16 cells, form structural and functional units involved in GM3-dependent cell adhesion as well as signal transduction (1,2) to promote cell motility (5). Such structural units coexist in detergent-insoluble, low density membrane fraction (DIM) with, but can also be immunoseparated from, cholesterol-enriched, caveolin-containing membrane units originating from caveolae (3). Similar GSL signaling/adhesion domains have been observed in the human embryonal carcinoma 2102 (7,8), mouse neuroblastoma Neuro2a (9), and rat cerebellar cells (24) and are proposed to be termed "glycosignaling domain" (GSD) (3, 6).
These previous studies suggest that GM3, SM, and c-Src may be essential components of GSD in B16 melanoma cells. The purpose of the present study was to further assess essential lipid components and their functional effect on c-Src activity in GSD, using two approaches: (i) reconstitution of membranes displaying adhesive properties similar to those of GSD, coupled with signal transduction through c-Src activation; (ii) comparative effects of lyso-GM3 and other lyso compounds on GSD function in B16 cells and in reconstituted membranes. We successfully reconstituted membranes simulating GSD composition and function based on a classic concept proposed by Racker (14), employing a modified method used previously for reconstitution of PC/cholesterol membrane vesicles with inserted transmembrane proteins, particularly growth factor receptors or integrin receptors (19,20). Reconstituted membranes are characterized by the following properties: (i) lipid composition (major lipids GM3 and SM; much smaller quantities of PC and cholesterol) is very similar to that of naturally occurring GSD membranes immunoseparated from caveolae fraction present in DIM fraction; (ii) c-Src in the membrane is closely associated with GM3 as shown by co-immunoprecipitation; (iii) the membranes bind to the solid phase coated with Gg3 or anti-GM3 mAb, with consequent activation of c-Src phosphorylation, but do not bind to the GlcCer-coated solid phase, and hence no c-Src phosphorylation occurs; (iv) optimal adhesion and c-Src phosphorylation response are observed with a certain composition of GM3, c-Src, and other lipid components, but the response is lower when the quantity of GM3 or c-Src is reduced; (v) the reconstituted membrane in which GM3 is replaced by GM1, GD1a, or LacCer does not show binding to Gg3 or anti-GM3 mAb, or consequent c-Src phosphorylation response; (vi) addition of cholesterol and PC significantly increases the c-Src phosphorylation response; and (vii) binding to solid-phase Gg3 and c-Src phosphorylation response is blocked by lyso-GM3 or its derivative, but the binding is not affected by lyso-PC, psychosine, or Lac-Sph. All these properties of reconstituted membranes are very similar to those of GSD in B16 cells (2,3), indicating that GM3, SM, and c-Src are essential components of GSD, whereas PC and cholesterol are auxiliary components that enhance or stabilize GSD function. Enhancement of c-Src activity in response to adhesion to the Gg3-coated dish was determined using a reconstituted SGPC membrane (see Fig. 3). The effect on this process by various lyso-GSLs known to affect structure and function of GSD was tested as described in the text. Lane 1, membrane added to the GlcCercoated dish. Lanes 2-9, membrane adhered to Gg3-coated dishes in the presence of various GSLs and lyso-GSLs (Me 2 SO as vehicle; final concentration, 0.5%) as described in the text. A major question remaining to be elucidated is the mechanism by which c-Src incorporated in the lipid bilayer is activated when external GM3 is stimulated by its ligand. In view of recent findings on crystal structure, c-Src activation is thought to be based on disordering of autoinhibitory interaction between intramolecular subunits, leading to Tyr 416 autophosphorylation. This is associated with distortion of an ordered helical loop located between the "N-lobe" and the "C-lobe" of the kinase domain (25). Activation does not necessarily always depend on dephosphorylation of Tyr 527 at the C terminus, which was previously considered to lock the molecule in an inhibited conformation (26). Tyr 527 -independent activation of Src kinase through sulfhydryl group modification was proposed recently (27). c-Src produced in SF9 cells using the baculovirus system is an inactive form that lacks Tyr 416 phosphorylation. Such inactive c-Src is activated by external ligand bound to SH3 and SH2 domains (25,28). It is crucial to elucidate the mechanism causing activation of c-Src, bound to the lipid bilayer, by enhanced clustering of GM3. The mechanism may be analogous to activation of cytoplasmic tyrosine kinase induced by growth factor-dependent clustering of its receptor (29).
Glycosyl-Sph, having sialic acid and N-unsubstituted Sph, was found to disrupt GSD function, as indicated by inhibition of GM3 immunostaining by anti-GM3 DH2, and inhibition of GM3-dependent adhesion and associated FAK activity and c-Src phosphorylation response. The simplest compound of this type is SA-Sph, as previously found (30). In the present study, two other compounds, lyso-GM3 and NeuNdcAc lyso-GM3, were found to also disrupt GSD function. Other lyso-GSLs and lyso-PC had no effect at similar concentration. The effect of glycosyl-Sph on GSD is in striking contrast to the effect of cholesterol-binding reagents that disrupt caveolae structure and function. Significantly, the disruptive effect of lyso-GM3 and its analogue on GSD in reconstituted membrane is similar to that in native B16 cells.
The effect of lyso-GM3 on GSD function may well be due to disruption of cis interaction of GM3 and consequent dispersion of GM3 clustering, although further extensive study is needed to clarify the exact mechanism. Preincubation of cells with 1-5 M lyso-GM3, followed by washing, greatly reduced staining with anti-GM3 mAb DH2 and GM3 clustering, without changing the cellular level of GM3. The reduction of DH2 staining was not due to competitive inhibition of DH2 binding to the cell surface by lyso-GM3, because cells were washed before the DH2 was added, lyso-GM3 cross-reacts with DH2 (Fig. 5C), and a large quantity of lyso-GM3 was inserted in the membrane even though the GM3 content was unchanged (Fig. 5D). Lyso-GM3, therefore, may cause a change in GM3 organization (e.g. dispersion of GM3 clustering as above). The blocking effect of lyso-GM3 and its analogue on FAK activity and c-Src phosphorylation response must be considered a consequence of GM3 organizational change in GSD, in both B16 cells and in reconstituted membrane.
Activated c-Src often appears as a doublet band on SDS gel, analogous to membrane-bound c-Src (31) (see legend of Fig.  3A). An effect on GSD similar to that observed for lyso-GM3 was found previously for sialyl␣231Sph (30). Neither filipin nor nystatin had an effect on GM3-dependent adhesion and associated FAK activation (3) at doses that disrupt caveolae structure and function (32,33). Interestingly, the reconstituted membrane showed the same susceptibility to lyso-ganglioside analogues as the original B16 cells or isolated GSD membrane fraction.
In summary, reconstituted membranes show all the characteristic properties of GSD from melanoma B16 cells. GSD from other cell types is characterized by different major GSLs in combination with different signal transducers (7)(8)(9)24). A close association and interaction between clustered GSLs and signal transducers, particularly Src family kinases, small G-proteins, and FAK, play a central role in cellular interactions coupled with signal transduction in general.