Effect of gangliosides on the distribution of a glycosylphosphatidylinositol-anchored protein in plasma membrane from Chinese hamster ovary-K1 cells.

Glycosylphosphatidylinositol (GPI)-anchored proteins are clustered mainly in sphingolipid-cholesterol microdomains of the plasma membrane. The distribution of GPI-anchored fusion yellow fluorescent protein (GPI-YFP) in the plasma membrane of Chinese hamster ovary (CHO)-K1 cells with different glycolipid compositions was investigated. Cells depleted of glycosphingolipids by inhibiting glucosylceramide synthase activity or cell lines expressing different gangliosides caused by stable transfection of appropriate ganglioside glycosyltransferases or exposed to exogenous GM1 were transfected with GPI-YFP cDNA. The distribution of GPI-YFP fusion protein expressed at the plasma membrane was studied using the membrane-impermeable cross-linking agent bis(sulfosuccinimidyl)suberate. Results indicate that GPI-YFP forms clusters at the surface of cells expressing GM3, or cells depleted of glycolipids, or transfected cells expressing mainly GD3 and GT3, or GM1 and GD1a, or mostly GM2, or highly expressing GM1. However, no significant changes in membrane microdomains of GPI-YFP were detected in the different glycolipid environments provided by the membranes of the cell lines under study. On the other hand, wild type CHO-K1 cells exposed to 100 microm GM1 before cross-linking with bis(sulfosuccinimidyl)suberate showed a dramatic reduction in the amount of GPI-YFP clusters. These findings clearly indicate that manipulating the glycolipid content of the cellular membrane, just by changing the ganglioside biosynthetic activity of the cell, did not significantly affect the association of GPI-YFP on the cell surface of CHO-K1 cells. The effect of exogenous GM1 gangliosides on GPI-YFP plasma membrane distribution might be a consequence of the ganglioside level reached in plasma membrane and/or the effect of particular ganglioside species (micelles) that lead to membrane architecture and/or dynamic modifications.

Glycosylphosphatidylinositol (GPI) 1 -anchored proteins (GPI-APs) are a heterogeneous class of proteins anchored to the membrane via a post-translational lipid modification, the GPImoiety. Biosynthesis of the GPI-moiety occurs in the endoplasmic reticulum via a series of enzymatic steps that sequentially add GPI components. GPI-APs are then transported through the Golgi to the trans-Golgi network, where they are sorted to the cell surface (1). GPI-APs have been found as constituents of the glycosphingolipid-enriched microdomain (GEM) (2)(3)(4), dynamic assemblies of cholesterol, saturated phospholipids, and sphingolipids, characterized by a light buoyant density and resistance to solubilization by Triton X-100 at 4°C (5,6). Experiments of direct visualization of GPI-APs using multicolor imaging of green fluorescent protein (GFP) variants demonstrated that these proteins segregate progressively in Golgi/ trans-Golgi network structures, exclude resident proteins, and exit into particular vesicles before surface delivery in both polarized and nonpolarized cells (7).
Gangliosides are a large family of sialic acid containing glycosphingolipids (GSLs). They are found in the outer leaflet of the plasma membrane of nearly all vertebrate cells, but most concentrate in the central nervous system (8). Like GPI-APs, they have also been reported to reside in GEM (6, 9 -11) and to modulate protein distribution in membrane microdomains. Recently, Jacobson and colleagues (12) demonstrated that GM1, a monosialoganglioside, reduced the abundance of F-Thy-1, a GPI-AP, in lipid clusters reconstituted in a supported model membrane monolayer. Indeed, Kurzchalia and co-workers (3) reported that exogenously added GM1 displaced GPI-APs from the lipid microdomain in living cells. As a result, it was hypothesized that the wide spectrum of biological effects exerted by gangliosides could be explained by an interaction with signaling molecules or by the influence of the structure of plasma membrane, more specifically on GEM (3,5).
Many approaches have been used to investigate ganglioside functions. Numerous studies, as indicated above, consisted of adding exogenous gangliosides to the culture medium and then examining the cellular responses (13). Glycosphingolipid synthesis inhibitors have also been used to modify levels of endogenous glycolipids to investigate glycolipid function (1). Prototype inhibitors are PDMP (14) and its analogous PPPP (15). These two compounds inhibit glucosylceramide (GlcCer) synthase activity blocking the efflux of substrate to ganglioside synthesis. Another approach to diminish the concentration of cell surface glycolipids consisted of hydrolyzing them by treatment of cells with endoglycoceramidase (16), reduction of the neuraminidase-sensitive ganglioside content of cells by stable sialidase expression (17), or by modulation of GM1 surface levels by manipulation of plasma membrane sialidase activity in hippocampal neurons in culture (18). Finally, new strategies such as the targeted disruption or overexpression of genes in the glycolipid biosynthesis pathway or the modification of cellular ganglioside profile by gene transfection of enzymes involved in ganglioside biosynthesis has yielded new insights into the functional and biological roles of glycolipids, particularly ganglioside (19 -21).
Although there is a large number of studies on ganglioside function, we have no evidence so far of a comprehensive analysis of the effects mediated by endogenous, exogenous, and depletion of glycosphingolipid in a same cell system. Using a chemical cross-linking approach and density gradient centrifugation in sucrose, in the present work we investigated the effect of particular gangliosides, synthesized using the endogenous machinery of Chinese hamster ovary (CHO)-K1 cells, on the distribution of a GPI-AP in plasma membrane. Furthermore, we also investigated the distribution of GPI-AP in plasma membrane of CHO-K1 cell lines depleted of glycolipids by inhibiting glucosylceramide synthase activity or exposed to exogenous GM1 ganglioside. Our data suggest a differential effect of gangliosides on the distribution of the GPI-AP. At least five different sets of endogenously expressed gangliosides had not significant effect on GPI-AP distribution in plasma membrane. On the other hand, exogenous application of GM1 ganglioside displaced GPI-AP from lipid domains in CHO-K1 cells.
Inhibition of glycolipid synthesis with PPPP (Matreya Inc.) was carried out as described (23). Wild type CHO-K1 cells in culture were treated for 4 days with 2 M PPPP added to the culture medium. Inhibition of glycolipid synthesis was monitored by analyzing the cellular ganglioside content by high performance thin layer chromatography (HPTLC).
Cells were maintained at 37°C in 5% CO 2 in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum and antibiotics. Cells were transfected with 1 g/dish of plasmids carrying cDNAs coding for the total sequence of the yellow fluorescence protein (YFP) fused to a GPI attachment signal (GPI-YFP) or for the vesicular stomatitis virus glycoprotein (VSVG) fused to the cyan fluorescence protein (VSVG-CFP) or for the total sequence of GAP-43 using Lipo-fectAMINE (Invitrogen) or FuGENE 6 (Roche Molecular Biochemicals) as DNA carriers, essentially according to the manufacturers' recommendations. GPI-YFP and VSVG-CFP fusion constructs were provided by P. Keller, Max-Plank Institute, Dresden, Germany (7). The cDNA containing the full coding sequence for GAP-43 was amplified in our laboratory by reverse transcription PCR using, as template, mRNA obtained from mouse brain. The amplified full-length cDNA was subcloned into the pCR 3.1 TA vector (Invitrogen). 15 h later the cells were used for cross-linking experiments (as indicated below) or washed with cold phosphate-buffered saline (PBS) and harvested with 10 mM Tris-HCl buffer (pH 7.2) containing 0.25 M sucrose for Western blot assays.
Triton X-100 Membrane Extraction-Wild type CHO-K1 cells were washed with cold PBS and isolated by scraping. Samples were treated with 0.5 ml of lysis buffer containing 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, 0.1 M Na 2 CO 3 , 5 g/ml aprotinin, 0.5 g/ml leupeptin, 0.7 g/ml pepstatin, and 25 mM Tris-HCl (pH 7.5) (TNE buffer) at 4°C for 1 h and then centrifuged for 1 h at 100,000 ϫ g at 4°C. The supernatant (soluble fraction) was removed, and the pellet (insoluble fraction) was resuspended in 0.2 ml of lysis buffer. Proteins from soluble and insoluble fractions were precipitated with chloroform:methanol (1:4 v/v). The pellets were resuspended in Laemmli buffer (24) and subjected to SDS-PAGE and Western blotting.
Sucrose Gradient-Wild type CHO-K1 cells were lysed in 0.5 ml of TNE lysis buffer at 4°C for 1 h. Lysates were centrifuged for 10 h at 150,000 ϫ g at 4°C on continuous sucrose gradients (5-35%) in TNE buffer without Triton X-100. 12 fractions were collected from the bottom of the sucrose gradient with a fraction collector. Proteins were precipitated with 10% trichloroacetic acid, resuspended in Laemmli buffer, resolved by electrophoresis through 4 -20% SDS-PAGE (Novex Trisglycine gels, Invitrogen) and later analysis by Western blot using the corresponding antibodies. Protein bands in nitrocellulose membranes were visualized by Ponceau S staining and quantified with the computer software Scion Image. Sucrose concentration in each fraction was determined measuring the refractive index (Abbé refractometer, CEQUIMAP, Facultad de Ciencias Químicas, Universidad Nacional de Córdoba).
GM1 Cell Loading and Quantification of Ganglioside Association-Wild type CHO-K1 cells were incubated for 1 h at 37°C in serum-free DMEM containing 100 M GM1 (Sygen, TRB Pharma; kindly provided by Dr. G. Nores, Universidad Nacional de Córdoba, Argentina). To remove excess lipid, cells were washed extensively with PBS containing 0.2% BSA. Then, cells were subjected to chemical cross-linking (see below), SDS-PAGE, and Western blotting. To evaluate the ganglioside association, cells were exposed to 100 M GM1 in DMEM for 1 h at 37°C, washed extensively with PBS containing 0.2% BSA, and scraped. Cells were pelleted by brief centrifugation, and lipids were extracted from the cell pellet with chloroform:methanol (2:1 v/v). To evaluate the trypsin-stable pool, cells were exposed to 100 M GM1 for 1 h at 37°C, washed extensively with PBS containing 0.2% BSA, and incubated for 10 min at 37°C with PBS containing 0.1% trypsin (3). Cells were pelleted and lipids extracted with chloroform:methanol (2:1 v/v). Lipid purification and chromatography analysis were carried out as indicated below. Gangliosides were visualized by staining with orcinol.
Cross-linking, Electrophoresis, and Western Blotting-We essentially used the procedure of cross-linking described by Friedrichson and Kurzchalia (25). Cells were washed twice with cold PBS and incubated at 4°C with 0.5 mM bis(sulfosuccinimidyl)suberate (BS 3 , Pierce Chemical Co.) for 45 min. Cross-linking was quenched by the addition of 50 mM glycine for 15 min at 4°C. Cells were washed with PBS, collected, and pelleted by 5-min centrifugation at 9,000 ϫ g. Proteins from pellets were resolved by electrophoresis through 4 -20% SDS-polyacrylamide gels under reducing conditions (24) and then transferred electrophoretically to nitrocellulose membranes (26) for 1 h at 300 mA. Protein bands in nitrocellulose membranes were visualized by Ponceau S staining. For immunoblotting, nonspecific binding sites on the nitrocellulose membrane were blocked with 5% defatted dry milk or with 2.5% BSA and 2.5% polyvinylpyrrolidone 40 in Tris-buffered saline (400 mM NaCl, 100 mM Tris-HCl pH 7.5 buffer), depending on the antibody. Anti-GFP polyclonal antibody (Roche Molecular Biochemicals) was used at a dilution of 1:800. Anti-GAP-43 (Santa Cruz Biotechnology) polyclonal antibody was used at a dilution of 1:200. Bands were detected by protein A coupled to horseradish peroxidase combined with the chemiluminescence detection kit (Western Lightning, PerkinElmer Life Sciences) and Kodak Biomax MS films. The molecular weights were calculated based on calibrated standards (Invitrogen) run in every gel. The relative contribution of individual bands was calculated using the computer software Scion Image on scanned films of low exposure images. The statistical analysis was carried out using the Newman-Keuls test from the Statistica software.
Metabolic Labeling, Lipid Extraction, and Chromatography-Briefly, cells in culture (3 ϫ 10 5 cells/35-mm dish) were labeled with 40 Ci/ml [lsqb] 3 H]Gal, PerkinElmer Life Sciences; 34.6 Ci/mmol) for 18 h. For PPPP-treated cells, [ 3 H]Gal was added to the culture medium 18 h before harvesting the cells. After washing with cold PBS, cells were scraped from the plate and pelleted by centrifugation. Lipids were extracted from the cell pellet with chloroform:methanol (2:1 v/v) and freed from water-soluble contaminants by passing through a Sephadex G-25 column. The lipid extract was supplemented with appropriate amounts of standard gangliosides and chromatographed on HPTLC plates (Merck) using chloroform:methanol and 0.25% CaCl 2 (60:36:8 v/v) as solvent. Standard gangliosides were visualized by exposing the plate to iodine vapors. Routinely, 15,000 -20,000 cpm was spotted on each lane. Radioactive gangliosides were visualized using a Fuji Photo Film Bio Imagen analyzer or visualized by fluorography after dipping the plate in 0.4% melted 2,5-diphenyloxazole in 2-methylnaphthalene and exposure to a radiographic film at Ϫ70°C, usually for 4 -6 days (20).
To analyze the pattern of gangliosides in fractions obtained from sucrose gradients, cells in culture (3 ϫ 10 5 cells/35-mm dish) were labeled with 2 Ci/ml of [ 14 C]Gal (PerkinElmer Life Sciences; 329.5 mCi/mmol) for 22 h. Then, cells were lysed and subjected to centrifugation on a continuous sucrose gradient as indicated above. To determinate the ganglioside content of each fraction, they were subjected to Folch-Pi partition (27). The resulting aqueous phases were purified further to eliminate Triton X-100 and sucrose by passing through DEAE-Sephadex and Sep-Pak C 18 cartridge columns (Waters Corp., Milford, MA). Lipid extracts were subjected to HPTLC, and gangliosides were visualized using a Fuji Photo Film Bio Imagen analyzer or by fluorography.
Immunoprecipitation Assay-Cells in culture (3 ϫ 10 5 cells/35-mm dish) were labeled with 2 Ci/ml [ 14 C]Gal and then transfected with 1 g/dish of plasmid carrying the cDNA coding for GPI-YFP. After 24 h, cells were washed and treated with 0.2 ml of TNE buffer containing 1% Triton X-100 at 4°C for 20 min. Lysates were centrifuged at 400 ϫ g for 10 min, and the supernatants were mixed with protein G-Sepharose beads and precleared by stirring for 30 min at 4°C. Then, supernatants were incubated with anti-GFP monoclonal antibody and protein G-Sepharose at 4°C for 4 h. After a brief centrifugation, a supernatant was recovered, and beads were washed three times with TNE buffer and 1% Triton X-100 and once with TNE buffer without Triton X-100. Lipids from the supernatant and beads were purified, subjected to HPTLC analysis, and visualized as described above.
Neuraminidase Treatment-Cells from clone 4 transiently expressing GPI-YFP were incubated for 2 h at 37°C in DMEM containing 2 units/ml neuraminidase type V from Clostridium perfringes (Sigma). After this, cells were washed with cold PBS and subjected to crosslinking experiments with BS 3 as indicated above. The activity of the neuraminidase treatment was checked in parallel by metabolic labeling of glycolipid with [ 3 H]Gal for 15 h and incubated for 2 h at 37°C in DMEM containing 2.0 units/ml neuraminidase type V. After the incubation period, cells were washed with PBS, and lipids were extracted and chromatographed following the procedure as indicated above.

RESULTS
Cell Lines and Exogenous Administration of GM1 Ganglioside-Previously, we have established several CHO-K1 cell lines, which express different gangliosides by transfection of GalNAc-T, Gal-T2, or Sial-T2 glycosyltransferases. Wild type CHO-K1 cells express predominantly the ganglioside GM3, as shown in the pattern of radioactive lipids metabolically labeled with [ 3 H]Gal for 18 h (Fig. 1, WT). Cells stably transfected with the cDNA encoding the chicken Sial-T2 under the control of the human elongation factor promoter (clone 2) (20, 22) synthesize mostly GD3 and GT3, accumulate LacCer, and practically do not synthesize GM3 (Fig. 1, clone 2). On the other hand, CHO-K1 cells stably expressing the human full-length Gal-NAc-T cDNA (clone 3) (23) synthesize the a-series ganglioside GM2 and, to a lesser extent, GM1 and GD1a because of the constitutive expression in these cells of the enzymes involved in the synthesis of GM1 and GD1a (28) (Fig. 1, clone 3). Subsequent transfection of mouse Gal-T2 to clone 3 led to the isolation of a new stable cell clone, clone 4 ( Fig. 1, clone 4). This cell clone is characterized by having increased consumption of GM2 for synthesis of GM1 and GD1a.
To get cells with a reduced content of all glycosphingolipid classes, wild type CHO-K1 cells were treated with PPPP, a potent inhibitor of ceramide glucosyltransferase and hence of synthesis of GlcCer and of more complex glycolipids (20,23,29). Exposure of cells to 2 M PPPP in the culture medium for 4 days led to a 90 -95% decrease of GM3 and LacCer content with respect to control cells (Fig. 1, WT-PPPP).
For exogenous GM1 administration, wild type CHO-K1 cells were exposed to 100 M GM1 for 1 h in serum-free medium at 37°C, and then the cells were washed with BSA solution to remove excess GM1. Lipids were extracted from the cells, chromatographed on HPTLC, and revealed by orcinol staining. As indicated above (Fig. 1, WT), CHO-K1 cells express predominantly the ganglioside GM3 (Fig. 2, WT, 0 M GM1); however, a considerable incorporation of GM1 (approximately 3-fold the amount of endogenous GM3) was observed when cells were incubated with the exogenous lipid (Fig. 2, WT, 100 M GM1). After cells were washed with BSA solution, ϳ32% of total associated GM1 could be released by trypsin treatment (not shown). The trypsin-stable pool of GM1 likely corresponds to the molecules inserted into the plasma membrane.
Characterization of Expressed GEM and Non-GEM Resident Proteins-Wild type CHO-K1 cells were transiently transfected with cDNA encoding a fusion protein containing a GPI-anchored signal, a GEM marker (7). The fusion protein contains an endoplasmic reticulum import signal, the total sequence of the YFP, and a consensus N-glycosylation site fused to a GPI attachment signal (GPI-YFP). Also, wild type CHO-K1 cells were transiently transfected with a cDNA encoding the non-GEM marker VSVG fused to the CFP (7). GPI-YFP and VSVG-CFP expression was confirmed by Western blotting with an antibody directed to the fluorescent protein. The antibody detected the GPI-YFP as two bands, one of ϳ39 kDa and the other of approximately 37 kDa (Fig. 3A). To evaluate the nature of the GPI-YFP fraction present in plasma membrane, cell surface proteins were biotinylated with the water-soluble agent Sulfo-NHS-SS-Biotin (EZ-Link biotinylation reagent, Pierce Chemical Co.), purified with streptavidin affinity chromatography, and analyzed by Western blotting (30). Results revealed that the fraction of GPI-YFP with high molecular mass (39 kDa) was the main species present in plasma membrane, indicating that this fraction of expressed GPI-YFP likely corresponded to the fully glycosylated form (results not shown). On the other hand, the analysis of VSVG-CFP expression by Western blotting revealed a band of the expected molecular mass (ϳ90 kDa) (Fig. 3A). The subcellular distribution of GPI-YFP in CHO-K1 cells included a pool at the cell surface and another one with a juxtanuclear distribution (Fig. 3B). This compartment was identified as the Golgi complex by colocalization with the medial Golgi marker, mannosidase II (not shown), as described previously for other epithelial cells (31). VSVG-CFP was distributed evenly at the cell surface of the transfected cells (Fig. 3B).
To characterize further whether GPI-YFP behaved appropriately as a GEM marker and VSVG-CFP as a non-GEM marker, we determined their insolubility in Triton X-100 at 4°C (6). As expected, the GEM marker (GPI-YFP) was 67% insoluble, whereas the non-GEM marker (VSVG-CFP) was less than 22% insoluble (Fig. 3C). Additionally, we also examined the distribution of both proteins in sucrose density gradients (Fig. 4). Such a procedure should separate lipid-rich components from the bulk of the Triton X-100-insoluble material in cell lysates, which would include nuclear remnants, cytoskeleton or extracellular matrix, and any component that binds to such insoluble materials. The insoluble components should pellet, whereas proteins and lipids resistant to Triton X-100 extraction should band at a low density (2). Basically, homogenates from transfected CHO-K1 cells were extracted with Triton X-100 at 4°C, and lysates were subjected to continuous sucrose gradient ultracentrifugation, fractionation, and detection of the fusion proteins and protein marker by Western blotting. GPI-YFP was highly concentrated in low density fractions codistributing with GAP-43, a recognized GEM marker (6,32,33) Essentially the same distribution patterns in sucrose gradient were obtained for GPI-YFP from wild type CHO-K1 cells and clones 2, 3, and 4 (results not shown). In contrast, VSVG-CFP distributed in higher density fractions, behaving essentially as Triton X-100soluble proteins (2) (Fig. 4B). Thus, both protein markers were correctly expressed and sorted in CHO-K1 cells and are useful in sensing the effect of particular gangliosides on their distribution in plasma membrane.
The Association of GPI-YFP on the Cell Surface Is Not Affected by Changing the Expression of Endogenous Gangliosides-To study the association of GPI-YFP on the plasma membrane of living cells we used chemical cross-linking with the membrane-impermeable agent BS 3 (3,25), which possess a spacer arm of 1.14 nm. When wild type CHO-K1 cells transiently expressing GPI-YFP were subjected to chemical crosslinking with the agent BS 3 , a major band of approximately 75 kDa (dimer) and a band of high molecular mass (Ն400 kDa) were detected by Western blotting (Fig. 5A, left panel). The band of high molecular mass likely corresponded to a multimer of ϳ12-15 GPI-YFP molecules. In contrast, CHO-K1 cells transiently expressing VSVG-CFP and subjected BS 3 showed low cross-linking efficiency (less than 8%), as expected for a non-GEM protein marker (Fig. 5A, right panel). In brief, these results are in line with the behavior of the fusion proteins in both sucrose gradient centrifugation and solubilization by nonionic detergent indicated above.
It is assumed that transient transfection of cells results in heterogeneous levels of expression of recombinant proteins. However, wild type CHO-K1 cells expressing different amounts of the GPI-YFP protein, obtained by different times of cDNA transfection, showed almost the same cross-linking efficiency (37-42%) and minor changes in the relative proportion between bands (Fig. 5B and table shown at the bottom of the  figure). This result clearly indicates that the cross-linking pattern is independent of the level of GPI-YFP expression, suggesting that the GPI-YFP molecules are confined in domains with a high concentration (25).
Having shown that GPI-YFP is expressed correctly and that a fraction of the synthesized pool distributed in microdomains at the plasma membrane, we next examined the association of the GPI-YFP in the different glycolipid environment provided by the membranes of the cell lines under study. The efficiency of cross-linking was essentially the same in the different clones analyzed (Table I). Furthermore, wild type CHO-K1 cell expressing GM3, or transfected cells expressing mainly GD3 and GT3, or GM1 and GD1a, or mostly GM2 subjected to chemical cross-linking with BS 3 showed no significant changes in the relative proportion among bands (monomers, dimers, and oligomers) (results of a typical experiment are shown in Fig. 6, and averaged data from three independent experiments are shown in Table I). Similar results were also obtained in CHO-K1 cells with a generalized decrease of glycolipids expression (Fig. 6, WT-PPPP, and Table I). These results clearly indicate that manipulating the glycolipid content of cellular membrane, just by changing the ganglioside biosynthetic activity of the cell, did not significantly affect the association of GPI-YFP on the cell surface of living cells.
Exogenous GM1 Ganglioside Displaces GPI-YFP from GEM Microdomains in CHO-K1 Cells-CHO-K1 cells exposed to exogenous GM1 showed an inhibition in the formation of GPI-YFP dimer and oligomer. Basically, wild type CHO-K1 cells transiently expressing GPI-YFP protein were exposed to 100 M GM1 for 1 h in serum-free medium at 37°C and subjected to chemical cross-linking with BS 3 . As indicated above, the cross-linking efficiency in untreated CHO-K1 cells was ϳ46% (Fig. 7, Ϫ100 M GM1), whereas the exogenous addition of GM1 reduced the cross-linking efficiency significantly to 14% (Fig. 7, ϩ100 M GM1).
GPI-AP Distribution in CHO-K1 Cells Highly Expressing GM1 Ganglioside-In the light of results with exogenous GM1 ganglioside, it was important to analyze the association of GPI-AP in cells expressing mainly GM1. To obtain cells highly expressing GM1 ganglioside, cells from clone 4 (see Fig. 1) were incubated with C. perfringes neuraminidase type V to convert GD1a to GM1 (34). Cells from clone 4 express predominantly tGD1a and, to a lesser extent, GM1 as is shown in the pattern of radioactive lipids metabolically labeled with [ 3 H]Gal for 18 h (Fig. 8A, ϪNANase). After neuraminidase treatment ϳ90% of GD1a was converted to GM1, the main ganglioside expressed under this conditions by cells from clone 4 (Fig. 8A, ϩNANase). Thus, cells from this clone transiently expressing GPI-YFP were incubated with neuraminidase, and then the association of GPI-AP was analyzed by cross-linking with BS 3 . Results shown in Fig. 8B clearly indicate that even in cells highly expressing the ganglioside GM1 there was no any change in the association of GPI-APs on the plasma membrane of the cells.

Distribution Patterns of Gangliosides in Fractions Obtained
from Sucrose Gradient Centrifugation-Lipids from wild type CHO-K1 cells and cells from clones 2, 3, and 4 were metabolically labeled with [ 14 C]Gal. Homogenates from different cells were extracted with 1% Triton X-100 at 4°C, and lysates were centrifuged to equilibrium on a continuous sucrose gradient. 12 fractions were collected from the bottom of the sucrose gradient, and gangliosides were purified and analyzed by HPTLC as described under "Experimental Procedures." As reported previously in mouse melanoma B16 cells (5) and in Neuro2a cells (9) Ͼ90% of GM3 from wild type CHO-K1 cells banded at low density on a sucrose gradient (fraction 10) (Fig. 9, WT). Similarly, ϳ95% of GM2 extracted from clone 3 was recovered in fraction 10 (Fig. 9, clone 3). The protein content of fraction 10 represented only a small portion of total protein amount loaded on gradient (1.5%, see Fig. 4A). Significant amounts of GM3 and GM2 were also recovered in heavier fractions. Interestingly, we observed a differential enrichment of the lower band of GM3 in fractions 5, 6, and 12. More than 68% of gangliosides GD3 and GT3 from clone 2 was recovered in fractions 10, 11, and 12 ( Fig. 9, clone 2). 13% of GD3 and GT3 was also found in fraction 5. Lower, but significant amounts of GD3 and GT3, were also recovered in the remaining fractions. ϳ63% of GD1a, the major ganglioside expressed by cells from clone 4, and GM1 were mainly found in low density fractions 9 and 10 ( Fig. 9, clone 4). The rest of the GD1a was distributed uniformly in the remaining fractions. In experiments not shown we also investigated the distribution pattern of GPI-YFP from different clones in continuous sucrose gradients. We essentially found the same distribution patterns for GPI-YFP from wild type CHO-K1 cells and clones 2, 3, and 4 (see Fig. 4). Thus, more than 70% of GPI-YFP partitions into fractions 8 and 9, whereas the rest of the chimera protein was distributed in fractions 7 and 10. Thus, GPI-YFP was found to codistribute in sucrose gradient mainly with GD1a and, in minor proportion, with GM3, GM2, GD3, and GT3 (fraction 10). Taken together, the results indicate that manipulating the glycolipid content of cellular membrane did not significantly affect the distribution of GPI-YFP on the cell surface, analyzed by sucrose gradient centrifugation. These results are in line with those obtained using the chemical cross-linking BS 3 (see Fig. 6).
GPI-YFP and Ganglioside Association in CHO-K1 Cells-CHO-K1 cells and cells from clones 2, 3, and 4 transiently expressing GPI-YFP were metabolically labeled with [ 14 C]Gal. Homogenates from different cells were extracted with 1% Triton X-100 at 4°C, and lysates were immunoprecipitated by adding anti-GFP monoclonal antibody and protein G-Sepharose beads. Gangliosides were eluted, purified, and subjected to HPTLC analysis as described under "Experimental Procedures." As shown in Fig. 10, GPI-YFP protein expressed in clone 4 was able to coimmunoprecipitate with GD1a. In addition, a small fraction of GM3 from wild type CHO-K1 cells and GM2 from clone 3 also coimmunoprecipitated with GPI-YFP. Contrarily, GD3 and GT3 fail to coimmunoprecipitate under these conditions. It should be noted that the percentage of association between GPI-YFP and GD1a, GM2, or GM3 could be higher considering that GPI-YFP was transiently expressed with a typical efficiency of transfection of 15-20%. More than 95% of GPI-YFP was recovered in the immunoprecipitate, and the expression levels of GPI-YFP in all clones were almost the same (not shown). DISCUSSION On the basis that the bulk of cellular GSLs are membranebound, it has been speculated that they participate in cell surface events such as signal transduction and cell adhesion (4, 35, 36). By targeted disruption of genes it was demonstrated that the essential role of GSL structures is in embryonic development and the differentiation of some tissues, supporting the concept that GSLs are involved in crucial cell processes (37). GSLs have been reported to reside preferentially in GEM at the cell surface (6, 9 -11), thereby GSLs segregated in microdomains could modulate signal transduction cascades (4,5,38), concentrate membrane proteins and associated molecules or exclude membrane components (12). The aim of the present study was to investigate the effect of particular GSLs, synthesized using the endogenous machinery of CHO-K1 cells, on the distribution of a GPI-anchored fusion protein (GPI-YFP) in plasma membrane. Furthermore, the distribution of the GPI-AP in membrane of CHO-K1 cell lines depleted of glycolipids, by inhibiting glucosylceramide synthase activity, or exposed to exogenous GM1 ganglioside was studied.
Here, we demonstrated that the GEM marker GPI-YFP is expressed correctly in CHO-K1 cell and that an important fraction of the synthesized pool was distributed in microdomains at the plasma membrane, revealed both by biochemical studies and by the membrane-impermeable agent BS 3 . Interestingly, the association of the GPI-AP in microdomains did not change significantly in the different glycolipid environments provided by the membranes of the cell lines under study. Thus, wild type CHO-K1 cells expressing mainly GM3 or transfectant cells expressing mainly GD3, GT3 and LacCer, or GM1 and GD1a, or mostly GM2 subjected to chemical cross-linking with BS 3 showed no significant changes in the relative proportion among bands (monomers, dimers, and oligomers), which is indicative of GPI-AP membrane distribution. Similar results were also obtained in CHO-K1 cells with a generalized decrease of glycolipid expression. Supporting these results, manipulating the glycolipid content of cellular membrane did not significantly affect the distribution of GPI-YFP on the cell surface analyzed by sucrose gradient centrifugation. Taken together, results obtained under our experimental conditions suggest that particular sets of endogenously expressed gangliosides, besides lactosylceramide and glucosylceramide, are not essential for GPI-AP distribution in plasma membrane from CHO-K1 cells or, possibly, the effect of changing the glycosphingolipid composition can be suppressed by other membrane lipids, maintaining the correct plasma membrane physical properties (39).
Analysis of the distribution patterns of gangliosides in fractions obtained from sucrose gradient centrifugation indicates that gangliosides are present mainly at low density fractions. However, they showed an interesting segregation in different fractions. As reported previously (38,40), these results probably reflect the heterogeneous composition of GEM domains present in membranes from different CHO-K1 cell clones. Immunoprecipitation of GPI-YFP resulted in coimmunoprecipitation of GM3 in wild type CHO-K1 cells, GM2 in cell from clone 3, and GD1a in cell from clone 4, indicating a TABLE I GPI-YFP distribution in plasma membrane from CHO-K1 cell lines revealed by BS 3 treatment Wild type cells or cells from clone 2, 3, 4, or wild type cells treated with PPPP (WT-PPPP) transiently expressing GPI-YFP were incubated with BS 3 , quenched by glycine, and collected. Proteins were resolved by electrophoresis through 10% SDS-polyacrylamide gels, electrophoretically transferred to nitrocellulose membranes, immunoblotted with anti-GFP, and visualized by the chemiluminescence technique. The relative contribution of individual bands was calculated using the computer software Scion Image on scanned films of low exposure images. Results are the mean Ϯ S.D. of at least three independent experiments. A representative Western blot experiment is shown in Fig. 6 Table I. physical association among these molecules. However, results shown in Fig. 6 and Table I suggest that this association is not critical, at least for GPI-AP distribution in plasma membrane.
As reported for other proteins (41)(42)(43)(44)(45), the possibility that changes in glycolipid composition might have affected the intracellular trafficking of GPI-YFP protein was considered. Because the efficiency of cross-linking was essentially the same in the different clones and PPPP-treated cells, normal packaging and sorting of GPI-YFP into transport vesicles does not seem to require GSLs or at least the ganglioside species expressed in membranes of the CHO-K1 cell lines. These data support a prior report that glycosphingolipids are not necessary for formation of detergent-resistant membrane lipid domains and for biosynthetic transport of a GPI-AP in melanoma cells (46).
In contrast with results obtained using the set of CHO-K1 cell lines, wild type CHO-K1 cells exposed to exogenous GM1 showed an inhibition in the clustering of GPI-YFP in plasma membrane. It is known that exogenous gangliosides are incorporated into cells in different forms: micelles adherent to surface membrane, micelles attached to binding protein, or monomers inserted into the membrane (13). In consequence and according to the procedure of our experimental conditions, the action of the exogenous GM1 on the distribution of GPI-YFP in membrane from CHO-K1 cells could be attributed to the fraction of monomers inserted into the plasma membrane. However, high endogenous expression of GM1 in neuraminidasetreated cells from clone 4 did not significantly affect the distribution of GPI-YFP in plasma membrane. The results suggest that the difference observed between endogenous and exogenous GM1 on GPI-YFP distribution might be a consequence of the ganglioside concentration reached in plasma membrane of CHO-K1 cells exposed to exogenous GM1, which probably led to membrane architecture and dynamic modifications. Lastly, we cannot completely rule out a perturbing effect on GPI-YFP protein distribution by the small fraction of GM1 micelles attached to membrane proteins (trypsin-labile pool). In a recent report, it was described that the neoexpression of GM1 in Swiss 3T3 cells results in the dispersion of platelet- CHO-K1 cells and cells from clones 2, 3, and 4 transiently expressing GPI-YFP were metabolically labeled with [ 14 C]Gal. Lysates were immunoprecipitated by adding anti-GFP monoclonal antibody and protein G-Sepharose beads. Gangliosides were eluted, subjected to HPTLC analysis, and visualized using a Fuji Photo Film Bio Imagen analyzer. St, total 14 C-labeled glycolipids from clone 4 (run as standard); I, gangliosides recovered from beads (coimmunoprecipitated with GPI-YFP); S, gangliosides recovered from the supernatant. The positions of the gangliosides are indicated. derived growth factor receptor from GEM microdomains (47). These results, together those showed in this work, suggest that the effect of endogenous GM1 on protein distribution might be a cell-and protein-dependent process.
How does exogenous GM1 exert its effect on GPI-YFP distribution? The GEM hypothesis suggests the existence of dynamic assemblies of cholesterol and sphingolipids structured in a liquid-ordered phase (10,47). Lipid microdomains can include or exclude proteins in a highly dynamic process, which could be regulated by negative modulators and/or by removal or addition of GEM components (6,47). Recently, results from the laboratory of Ken Jacobson (12) showed that exogenous GM1 displaced the GPI-anchored cell surface protein Thy-1 from lipid microdomains reconstituted in a supported model membrane monolayer. The authors suggest that this behavior may result from a competition between the two molecules for lipid microdomain occupancy or structural changes of lipid domains sufficient to affect Thy-1 partitioning. If this behavior is similar in natural membranes from living cells, this could explain the effect of exogenous GM1 on GPI-YFP distribution in plasma membrane from CHO-K1 cells. Although further work will be required to determine the molecular basis of the exogenous GM1 effect, it should be noted that a small fraction (15%) of the BSA-resistant pool of exogenous GM1 was found to be resistant to extraction with Triton X-100 at 4°C, indicating that exogenous GM1 accumulated mainly in non-GEM microdomains, whereas the endogenous GM1 present in cells from clone 4 (both treated or untreated with neuraminidase) was almost completely labile (97%) to detergent extraction. 2 To identify molecular events where glycolipids are participating are fundamental to an understanding of the physiological, pathological, and pharmacological function of these lipids. A number of clinical studies have been carried out or are under way to evaluate the potential of GM1 as a therapy for treatment of neuronal diseases (48). The exogenous administration of high doses (1,000 mg) of GM1 has been used in clinical trials to evaluate its effect on human Parkinson's disease (49). It was proposed that the mechanism of promoting neuronal survival by GM1 is that it either mimics or potentiates neurotrophic factors by activation of transmembrane tyrosine kinase receptors (48). However, in light of results shown in this work it is particularly attractive to investigate whether the exogenous administration of high doses of GM1 also alters the distribution of endogenous GPI-APs present mainly in apical plasma membrane of epithelial cells or in membranes from lymphocytes (1), which could have important consequences on the cellular physiology.