Sphingolipid-enriched membrane domains from rat cerebellar granule cells differentiated in culture. A compositional study.

Sphingolipid-enriched membrane domains, characterized by a particular protein and lipid composition, have been detected in a variety of cells. However, limited data are available concerning these domains in neuronal cells. We analyzed the lipid and protein composition of a sphingolipid-enriched membrane fraction prepared from primary rat cerebellar granule cells differentiated in culture. Although the protein content of this fraction was only 1.4% of total cellular protein, 60% of the gangliosides, 67% of the sphingomyelin, 50% of the ceramide, and 40% of the cholesterol were located in this fraction. The protein pattern of the sphingolipid-enriched domain fraction was dramatically different from that associated with the cell homogenate. This fraction contained 25% of the tyrosine-phosphorylated proteins and was enriched in two proteins with apparent molecular masses of 135 and 15 kDa. 12% of cellular glycerophospholipids were located in the fraction, with phosphatidylcholine having the highest enrichment. The molar ratio between proteins, glycerophospholipids, cholesterol, sphingomyelin, ceramide and gangliosides in cerebellar granule cells was 1.6:41.6:6. 1:1.3:0.3:1 in the cell homogenate and 0.04:8.3:4.0:1.4:0.2:1 in the sphingolipid-enriched membrane fraction. These data indicate that selected proteins segregate with sphingolipids in specialized domains in the membrane of cultured neurons.

The increasing body of evidence, obtained by several different experimental approaches from both artificial and cellular models , suggests that lipid and protein components in the cell membrane are not randomly or homogeneously distributed but rather organized in domains with peculiar physicochemical and functional properties, different from those of the surrounding membrane environment, confirming the original prediction of Singer and Nicholson (24).
Sphingolipid-enriched domains that are reported to be enriched in gangliosides, sphingomyelin, and cholesterol (1,(17)(18)(19)(20)(21)(22)(23) are emerging as membrane compartments with relevant biological functions. They are rich in proteins involved in the mechanisms of signal transduction (1,18,(21)(22)(23)(25)(26)(27)(28)(29)(30)(31) and cell adhesion molecules (34). Thus, sphingolipid-enriched microdomains could represent a site within the plasma membrane where different molecules (both lipids and proteins) involved in signal transduction and/or cell adhesion and cell-cell interactions are specifically sorted and concentrated, allowing reciprocal interactions of functional significance. Recent studies have revealed that gangliosides in membrane sphingolipidenriched domains associate closely and specifically with single or multiple signal transducer molecules. Ganglioside GM3 1 is closely associated with c-Src, Rho, FAK, and Ras in B16 melanoma cells (1,21), with c-Src and Csk in neuroblastoma Neuro2a cells (22), and GD3 is associated with Src-family kinase Lyn and the neural cell adhesion molecule TAG-1 in rat brain (33,35). Such structural units seem to be involved in signal transduction in response to glycosphingolipid-mediated stimulation; GM3-mediated cell adhesion of melanoma B16 cells induces c-Src and FAK phosphorylation and Rho and Ras activation (1); treatment of neuroblastoma Neuro2a cells with exogenous gangliosides induces c-Src and mitogen-activated protein kinase activation, leading to neuronal differentiation (22); and treatment of primary cultured rat cerebellar neurons with anti-GD3 antibody induces Lyn activation with consequent phosphorylation of mitogen-activated protein kinases (33). Neurotrophin-induced p75 NTR -dependent sphingomyelin hydrolysis is also localized in a caveolar domain (27).
Procedures for the isolation of sphingolipid-enriched domains have been developed (1)(2)(3)36). They are generally based on the characteristic low density of the domains, which allows their separation by density gradient centrifugation after cell disruption in the presence of detergent (Triton X-100), in hypertonic conditions or by mechanical procedures. Despite the variety of density gradient steps and of cell lysis procedures, 1 Ganglioside and glycosphingolipid nomenclature is in accordance with Svennerholm (58) results obtained with different methods are similar and lead to similar conclusions (1,3).
Most data concerning the structure and functional role of sphingolipid-enriched domains have been obtained from studies performed on non-neuronal cells such as Madin-Darby canine kidney cells (17,22,34). The lack of information on sphingolipid-enriched domains in the nervous system is mainly due to the complexity of this tissue at the cellular and compositional level. Moreover, the need to utilize primary cultures for neurons implies difficult work to collect enough cells for the analytical procedures.
With this investigation we studied the composition of sphingolipid-enriched microdomains obtained from rat cerebellar granule cells differentiated in culture. To overcome the analytical problems related to the relatively limited number of cells, we radiolabeled the cellular lipids and proteins using appropriate radioactive precursors, namely [

Materials
Commercial chemicals were the purest available, common solvents were distilled before use, and water was doubly distilled in a glass apparatus. Trypsin, crystalline bovine serum albumin, and all the reagents for cell culture were from Sigma, except for basal modified Eagle's medium and fetal calf serum, which were purchased from Flow Laboratories. Anti-phosphotyrosine mouse monoclonal IgG 2b antibody and protein A/G PLUS-agarose were from Santa Cruz Biotechnology. Sphingosine was prepared from cerebroside (37). Sphingolipids and glycerolipids to be used as standards were extracted from rat brain, purified, and characterized (38). [1-3 H]Sphingosine was prepared by specific chemical oxidation of the primary hydroxyl group of sphingosine followed by reduction with sodium boro[ 3 H]hydride (39) (radiochemical purity, Ͼ98%; specific radioactivity, 2 Ci/mmol). [ 32 P]Orthophosphate (carrier-free) and [ 35 S]methionine (specific radioactivity, 1175 Ci/mmol) were purchased from Amersham Pharmacia Biotech and NEN Life Science Products, respectively. 3 H-Labeled lipids, ceramide, sphingomyelin, phosphatidylethanolamine, and lactosylceramide were extracted from [1-3 H]sphingosine-fed cells, purified, characterized, and used as chromatographic standards.

Cell Cultures
Granule cells obtained from the cerebellum of 8-day-old Harlan Sprague-Dawley rats were prepared and cultured as described (40). It is reported (40,41) that under these conditions, cell culture contains Ͼ90% granule cells, 5% ␥-aminobutyric acid-ergic neurons, and Ͻ5% glial cells. The cells were plated in 100-mm dishes at a density of 9 ϫ 10 6 cells/dish and cultured with 10 ml of basal modified Eagle's medium containing 10% fetal calf serum for 8 days. Experiments were performed between 6 th and 8 th day in culture. Typical protein content at this time was 700 g of protein/dish. Viability was assessed by the Trypan blue exclusion method.

Treatments of Cell Cultures with [ 3 H]Sphingosine or [ 35 S]Methionine or [ 32 P]Orthophosphate
Cells at the 6 th day in culture were incubated in the presence of 3 ϫ 10 Ϫ8 M [1-3 H]sphingosine (5 ml/dish) in cell-conditioned medium for 2 h (pulse). After the pulse, medium was replaced with cell-conditioned medium without radioactive sphingosine, and cells were further incubated for 48 h (chase). Under these conditions all sphingolipids (including ceramide, sphingomyelin, neutral glycolipids, and gangliosides), and phosphatidylethanolamine (obtained by recycling of radioactive ethanolamine formed in the catabolism of [1-3 H]sphingosine) were metabolically radiolabeled (42). Cells at the 7 th day in culture were preincubated in methionine-free medium for 2 h and subsequently incubated in the presence of 25 Ci/ml L-[ 35 S]methionine (5 ml/dish) for 20 h to radiolabel proteins (18,43). Cells at 8 th day in culture were incubated in the presence of 50 Ci/ml carrier-free [ 32 P]orthophosphate (5 ml/dish) in phosphate-free culture medium for 4 h (44). Under these conditions, glycerophospholipids (with the exclusion of phosphatidylethanolamine plasmalogen and including the minor inositol containing compounds) and phosphoproteins were radiolabeled.

Sucrose Gradient Centrifugation
After metabolic radiolabeling, at the 8 th day in culture, cells were subjected to ultracentrifugation on discontinuous sucrose gradient after lysis and homogenization in the presence of 1% Triton X-100 or in hypertonic sodium carbonate medium as described previously (2,3) with minor modifications.
Hypertonic Sodium Carbonate Method-Cells were harvested as described above, suspended (8 -10 ϫ 10 7 cells/ml) in 500 mM sodium carbonate, pH 11.0, and homogenized using a loose fitting Dounce homogenizer (20 strokes) and a probe sonicator (three 10-s bursts). 1.5 ml of the cell homogenate thus obtained was mixed with an equal volume of 90% sucrose in 25 mM MES, 2 pH 6.5, 150 mM NaCl and overlaid with a discontinuous sucrose gradient (30 -5%) in the same buffer containing 250 mM sodium carbonate. Samples were subjected to ultracentrifugation as described above.
Using both procedures, a white light-scattering band under light illumination located at the interface between 5 and 30% sucrose was detectable. After ultracentrifugation, eleven 1-ml fractions were collected starting from the top of the tube. The light-scattering band corresponded to fraction 5 and was regarded as the sphingolipid-enriched fraction. The bottom fraction (fraction 11) contained a pellet that was carefully homogenized before analysis. The entire procedure was performed at 0 -4°C in ice immersion.

Analysis of Radiolabeled Protein Patterns
Cell lysates and sucrose gradient fractions obtained after labeling cerebellar granule cells with [ 32 P]orthophosphate or [ 35 S]methionine were analyzed to determine protein content and pattern. In the case of [ 32 P]orthophosphate labeling, samples were extensively dialyzed to remove free [ 32 P]phosphate, and phospholipids were removed by chloroform/methanol extraction (45) before analysis. Similar amounts of protein for each sample (0.2-1 g) were subjected to SDS-PAGE on 10% acrylamide gels followed by radioactivity detection.
The qualitative and quantitative determination of phosphotyrosinecontaining proteins was by immunoprecipitation. Aliquots of each fraction obtained from cells labeled with [ 32 P]orthophosphate or [ 35 S]methionine (containing ϳ15-30 g of protein) were diluted 10-fold in immunoprecipitation buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2 mM NaF, 1 mM EDTA, 1 mM EGTA, 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 75 milliunits/ml aprotinin, 1% Triton X-100), mixed with protein A/G-Sepharose beads (40 l packed), and stirred on a rotary stirrer for 2 h at 4°C to preclear nonspecific binding. After centrifugation (270 ϫ g for 2 min), 2 g/ml mouse anti-phosphotyrosine monoclonal IgG 2b or 2 g/ml normal mouse IgG (as negative control) was added to the supernatants. The mixtures were placed on a rotary stirrer overnight at 4°C. Immunoprecipitates were recovered by adding protein A/G-Sepharose beads (40 l packed) and mixing for 2 h. Beads were washed three times with immunoprecipitation buffer, recovered by brief weak centrifugation (270 ϫ g for 2 min), suspended in 50 l of SDS-sample buffer, heated to 95°C for 3 min, and centrifuged (1000 ϫ g for 2 min). Supernatants were subjected to SDS-PAGE, and dried gels were analyzed to determine the radioactive protein patterns. The radioactivity associated with immunoprecipitates was quantitatively determined by liquid scintillation counting. The total amount of immunoprecipitable radioactivity was calculated for each fraction and for the cell lysate. Data were expressed for each fraction as the percentages of total immunoprecipitable radioactivity present in the amount of cell 2 The abbreviations used are: MES, 4-morpholineethanesulfonic acid; PC, phosphatidylcholine; PPC, phosphatidylcholine plasmalogen; PS, phosphatidylserine; PE, phosphatidylethanolamine; PI, phosphatidylinositol; PIP, phosphatidylinositol-4-phosphate; PIP 2 , phosphatidylinositol-4,5-diphosphate; HPTLC, high-performance thin-layer chromatography; PAGE, polyacrylamide gel electrophoresis. lysate loaded on gradient. Radioactivity associated with negative controls never exceeded 5% of radioactivity found in immunoprecipitates.

Analysis of Radioactive Lipids
The cell lysate, postnuclear supernatant, and sucrose gradient fractions obtained after cell metabolic radiolabeling were analyzed to determine the content of radiolabeled lipids. Samples were dialyzed for 4 days against distilled and decarbonated water (changed two times a day). We found that a long dialysis of samples before lipid extraction is necessary for effective removal of sucrose and that the removal of sucrose is critical for the quality of lipid extraction and analysis. After dialysis, samples were lyophilized, and lipids were extracted twice with 0.5 ml of chloroform/methanol, 2:1 (v/v) (45). The total lipid extract was not subjected to partitioning to avoid any possible loss of material. Aliquots of the lipid extract were analyzed by HPTLC as described below, followed by radioactivity imaging for quantification of radioactivity Identity of lipids separated by HPTLC was assessed by co-migration with standard lipids and confirmed by susceptibility of compounds to the following enzymatic and chemical treatments. A sample of the lipid extract was treated at 37°C for 2 h in 50 l of water in the presence of 1 milliunit of Vibrio cholerae sialidase to yield GM1. Sphingomyelin and phosphatidylethanolamine were purified according to the HPTLC blotting technique previously reported (46); they were separated by HPTLC, identified by spraying with primulin, blotted to polyvinylidene difluoride membrane where the corresponding bands were cut, and subjected to elution. Sphingomyelin was treated at 37°C overnight in 30 l of 100 mM Tris-HCl, pH 7.4, 0.5 mM MgCl 2 , 0.05% sodium deoxycholate, in the presence of 11 milliunits of Bacillus cereus sphingomyelinase, to yield ceramide; phosphatidylethanolamine was characterized following its degradation under alkaline conditions. The enzymatic or chemical reaction mixtures were separated by HPTLC, and the reaction products were identified by chromatographic comparison with standard lipids.

Other Analytical Methods
Cholesterol was quantified after separation on HPTLC followed by visualization with 15% concentrated sulfuric acid in 1-butanol. The quantity of cholesterol was determined by densitometry and comparison with known amounts of standard cholesterol using the Molecular Analyst program (Bio-Rad Laboratories). The protein content was determined according to Lowry (47), with the micro BCA assays (Pierce) and as dot spot revealed by Coomassie Blue staining; bovine serum albumin in the presence of sucrose was the reference standard.
The radioactivity associated with cells, with cell fractions, with lipids, and with delipidized pellets was determined by liquid scintillation counting. Digital autoradiography of the HPTLC plates and of the SDS-PAGE gels was performed with a Beta-Imager 2000 instrument (Biospace, Paris, France) using an acquisition time of about 24 h. The radioactivity associated with individual lipids and proteins was determined with the specific ␤-Vision software provided by Biospace. Autoradiography of 32 P-and 35 S-labeled proteins was carried out using Kodak Biomax MR and MS films.

Proteins-Feeding [ 32 P]orthophosphate and [ 35 S]methionine
to cells yielded an incorporation of radioactivity into proteins of 5.20 Ϯ 0.51 ϫ 10 6 cpm/mg cell protein and 44.9 Ϯ 3.60 ϫ 10 6 cpm/mg cell protein in the cell lysate obtained in the presence of Triton X-100, respectively. After discontinuous sucrose gradient centrifugation, fractions were collected from the top of the gradient (fraction 1) to the bottom (fraction 11). The distribution of protein-associated radioactivity into the eleven fractions is reported in Fig. 1. The distribution pattern of 32 P-and 35 S-labeled proteins was very similar. Fraction 5 contained only 1.4 Ϯ 0.4% of the total protein-associated radioactivity, which is predominantly (Ͼ70%) recovered in fractions 10 and 11. The experimental conditions we used for protein labeling were reported to yield a similar specific radioactivity for most proteins (43). The total protein content of the postnuclear fraction subjected to fractionation in a typical experiment was 4.7 mg. Thus, on the basis of the distribution of radioactivity, we calculated that fraction 5 contained 65 g of protein.
In preliminary experiments performed with standard albumin, we found that sucrose and other components of the gradient buffer interfered with the Lowry, Coomassie Blue, and BCA protein assays. For this reason, the use of a radiolabeling procedure seems to be highly preferable for the precise determination of protein content in fractions prepared by sucrose gradient centrifugation. Fig. 2 shows the patterns of [ 35  fraction 5 and were also enriched in a protein of about 46 kDa. Fig. 3 shows the distribution of phosphotyrosine-containing proteins in the eleven fractions as determined by immunoprecipitation with anti-phosphotyrosine monoclonal antibody. Similar results were obtained using 32 P-and 35 S-radiolabeled fractions. Fraction 5 was highly enriched in phosphotyrosinecontaining proteins. As shown in Fig. 4, phosphotyrosine-containing protein and total protein patterns (Fig. 2) were very similar in fraction 5.
Sphingolipids-Sphingosine is very rapidly and extensively taken up by cells, and it is immediately and efficiently acylated to ceramide, which is then converted into more complex sphingolipids (42,48). Moreover, the great extent of sphingosine recycling for the biosynthesis of complex sphingolipids (42) and the high specific radioactivity of [1-3 H]sphingosine allow incorporation of a large amount of radioactivity into sphingolipids (100.34 Ϯ 17.08 nCi/mg of protein in the homogenate). Thus, with this method the quantitative analysis of sphingolipids in cells or cell fractions is possible with very high sensitivity.
The radioactivity associated to gangliosides GM1, GD3, GD1a, GD1b, O-Ac-GT1b, GT1b, and GQ1b was 9, 5, 26, 12, 13, 35, and 2%, respectively, of the total radioactivity associated to all gangliosides. This distribution is in very good agreement with the ganglioside pattern of granule cells previously reported (49,50). Nevertheless, the chosen pulse-chase feeding conditions did not allow introduction of a similar specific radioactivity in ceramide, sphingomyelin, and gangliosides, because of different turnover of each sphingolipid class (49,50,53,54). The radioactivity ratio between gangliosides and ceramide was about 4-fold higher than the mass ratio, and the radioactivity ratio between ganglioside and sphingomyelin was about 3-fold higher than the mass ratio (Table I). Fig. 5 shows the pattern of radioactive lipids extracted from the cell homogenate after HPTLC separation. Radioactive ceramide, GlcCer, LacCer, sphingomyelin, GM1, GD3, GD1a, GD1b, O-Ac-GT1b, and GQ1b were identified. Radioactive PE was also identified in the lipid extract. The presence of radioactive PE is due to the recycling of the radioactive ethanolamine formed in the catabolism of [1-3 H]sphingosine (42). Fig. 6 shows the distribution of radioactivity incorporated into ceramide, sphingomyelin, glycosphingolipids (including neutral sphingolipids and gangliosides), and PE in each sucrose gradient fraction prepared by the Triton X-100 method. All radioactive sphingolipids were largely associated with fraction 5 (50 -65% of the radioactivity associated with sphingolipids in the cell homogenate), minor quantities being distributed into fractions 4 and 6 -11. In particular, less than 25% of radioactive ceramide and less than 10% of radioactive complex sphingolipids was present in fraction 11, which contained about 60% of cell proteins. Radioactive PE, on the other hand, was predominantly recovered in fractions 9 -11, and only a low amount was detectable in fraction 5. Very similar results were obtained analyzing sucrose gradient fractions prepared from [1-3 H] sphingosine-fed cells after lysis in hypertonic conditions (500 mM sodium carbonate), i.e. the majority of radioactive sphingo- ]methionine (lower panel) followed by immunoprecipitation with anti-phosphotyrosine mAb as described under "Experimental Procedures." Equal amounts of protein for each fraction were subjected to immunoprecipitation. The total amount of immunoprecipitable radioactivity was calculated for each fraction, and data were expressed as percentages of total immunoprecipitable radioactivity present in the homogenate. Fractions were collected as described in the legend of Fig. 1 lipids are recovered in fraction 5 (Ͼ60%), and only a minor amount was present in high density fractions. Thus, sphingolipid distribution in the gradient seems to be independent of the lysis method. Table II  All radioactive sphingolipids were highly enriched in fraction 5 (36.6), whereas enrichments calculated for the high density fractions were much lower (Ͻ8.5 in fraction 9, Х1 in fraction 10, and Ͻ0.5 in fraction 11). Within each fraction, all of them showed a very similar enrichment (ranging from 29.9 for ceramide to 40.2 for radioactive sphingomyelin, in the case of fraction 5).
Glycerophospholipids-Incorporation of radioactivity into cellular glycerophospholipids, by metabolical labeling with [ 32 P]orthophosphate, was 0.94 Ϯ 0.05 ϫ 10 6 cpm/mg protein in the cell lysate obtained in the presence of Triton X-100. Each glycerophospholipid has specific turnover because of different and multiple biosynthetic pathways. The experimental conditions we used allowed the introduction of radioactivity in the majority of species (Table I), thus making it possible to also follow those species that are minor or very minor cell components, like PI, PIP, and PIP 2 (52). Fig. 7 shows the patterns of 32 P-labeled lipids from the cell homogenate and gradient fractions after HPTLC separation. Radioactive PE, PS, PC, PI, PPC, PIP, and PIP 2 were identified. Phosphatidylethanolamine plasmalogen and the phosphosphingolipid sphingomyelin were present in trace amounts. The radioactive glycerophospholipid patterns in the sucrose gradient fractions were remarkably different (Fig. 7). Radioactive PC was by far the most abundant radioactive glycerophospholipid in fraction 5, followed by radioactive PI and PE. Fig. 8 shows the distribution of each radioactive glycerophospholipid into fractions prepared by sucrose gradient centrifugation. The majority of radioactivity (70 -95%) for all radioactive glycerophospholipid species was largely recovered in fractions 9 -11. Fraction 5 contained less than 10% of the radioactivity associated with each glycerophospholipid, with the exception of PC: about 22% of radioactivity associated with PC was detected in fraction 5. In the case of PE, the distribution of radioactivity in the fractions was remarkably similar to that determined after labeling with [1-3 H]sphingosine (Fig. 6). The enrichment of single radioactive glycerophospholipids in each fraction (calculated as described above for sphingolipids) is reported in Table II. The highest values were associated with fraction 9, whereas enrichments in fraction 5 were modest. However, enrichments for the different radioactive glycerophospholipid species were dramatically different from each other even within the same fraction. In fraction 5, radioactive PC and PPC showed the highest (13.2) and lowest enrichment (1.1), respectively.
Quantitative Determination of Endogenous Sphingolipids and Glycerophospholipids-The experimental conditions used  I Lipid and protein contents in rat cerebellar granule cells differentiated in culture Data on the mass sphingolipid and glycerophospholipid contents are from Refs. 49 -51, 53, and 54. By these data and by the radioactivity contents in fraction 5 (F 5) and in homogenate (Hom), we calculated the mass content of each class in fraction 5 and molar ratios of proteins and different lipids in fraction 5 and in homogenate. Data on cholesterol have been determined in this study. Molar ratios were normalized to total ganglioside content (nmol of component/nmoles of gangliosides in that fraction). Protein radioactivity is from 35 S labeling (as cpm). Sphingolipid radioactivity is from 3 H labeling (as dpm). Glycerophospholipid radioactivity is from 32 P labeling (as cpm). Protein molar content was calculated on the average of the protein molecular mass and must considered an approximate value. for cell lipid metabolic radiolabeling were designed to incorporate a significant amount of radioactivity in each lipid, allowing the detection and analysis of all species, including the minor components. However, because of the different turnover of each species, it is impossible to achieve the same specific radioactivity for each component. Under our experimental conditions, only gangliosides incorporated radioactive sphingosine proportionally to their endogenous content (Table I). Thus, the percentage of distribution of radioactivity between the 3 H-or 32 P-labeled lipids within each fraction does not reflect the mass lipid composition in the fraction. This has been calculated on the basis of the endogenous lipid content in the homogenate and the distribution of radioactivity in each lipid within the gradient (Table I). Quantitative data on the total ganglioside, sphingomyelin, ceramide, and glycerophospholipid contents in rat cerebellar granule cells differentiated in culture previously reported (49 -51, 53, 54) were used for the calculation. The data we calculated are reported (as molar ratios) in Table I. Comparison of the lipid molar ratio between fraction 5 and the homogenate clearly shows the great difference in their relative composition. In fact, the molar ratio between glycerophospholipids and gangliosides is about 42:1 in the homogenate, and it is reduced to about 8:1 in fraction 5.
Cholesterol-The content of cholesterol in rat cerebellar granule cells differentiated in culture was 64 Ϯ 9 nmol/mg of protein in the cell homogenate. In this range of cholesterol concentration, S.D. for the method was Ϯ15%. The distribution of cholesterol in fractions prepared by sucrose gradient centrifugation is shown in Fig. 9. About 40% of cellular cholesterol was associated with fraction 5 after sucrose gradient centrifugation. Tables I and II show the relative cholesterol content and the cholesterol enrichment in fraction 5 and in the homogenate, respectively.

DISCUSSION
The sphingolipid-enriched membrane fraction prepared by the Triton X-100 method (1-3) contains 50 -65% of cellular sphingolipids and 40% of cholesterol but only a minor part of the glycerophospholipids (about 10%). The distribution of sphingolipids within the gradient was identical using the sodium carbonate method. On the basis of the distribution of each lipid within the gradient and of the cellular content of each lipid, we calculated the molar composition of the sphingolipidenriched domain, i.e. of fraction 5 of the sucrose gradient fractionation, and compared it to that of the homogenate, as reported in Table I. Among the plasma membrane complex lipids, gangliosides are the most characteristic of neuronal cells. Thus, we normalized these data (Table I)   content. The comparison of the molar composition of the homogenate and fraction 5 allows some interesting conclusions. Molar ratios calculated for different sphingolipid species are very similar in homogenate and fraction 5, indicating that all sphingolipids concentrate in the membrane domains to a similar extent. These data strongly suggest that spontaneous segregation of sphingolipids within the membrane is the driving force leading to the formation of lipid domains, independent of the presence of particular scaffolding proteins. Spontaneous segregation of gangliosides and other sphingolipids has been well described in artificial and cellular models (4 -14). Clustering of glycosphingolipids in microdomains could provide a favorable membrane geometry to accommodate their large hydrophilic headgroups (55). The possibility of forming hydrogen bonds because of the amide and hydroxy group of ceramide (56) should further stabilize their clustering. All of these interactions contribute to the formation of a very rigid microenvironment (57), providing a clear explanation for the resistance of sphingolipid-enriched domains to detergent solubilization, extraction with sodium carbonate, or mechanical disruption. Cholesterol is also predominantly recovered in the sphingolipidenriched fraction prepared from cerebellar granule cells. However, its molar ratio to gangliosides varies from about 6:1 in homogenate to about 4:1 in fraction 5, indicating that it segregates together with sphingolipids, contributing to the formation of a rigid phase, but without representing a major contributor to this phenomenon, as confirmed by studies on model membranes (6,10). The segregation process of sphingolipids excludes the majority of cellular glycerophospholipids; their molar ratio to gangliosides dramatically decreases from 42:1 in the homogenate to about 8:1 in fraction 5. Interestingly, not all glycerophospholipids are excluded to the same extent from the sphingolipid-enriched domain. In particular, a higher portion of PC (that is the glycerophospholipid bearing the largest headgroup), with respect to other glycerophospholipids segregates together with sphingolipids. Even more dramatic is the exclusion of proteins from the sphingolipid-enriched domain (a rough protein to ganglioside molar ratio, calculated on the basis of the average protein molecular mass, ranges from about 1.6:1 in the homogenate to 0.04:1 in fraction 5). This strongly indicates that only selected protein molecules are associated with these domains, as confirmed from their particular protein patterns and enrichment in phosphotyrosine containing proteins. Actually, many proteins reported to be present in sphingolipid-enriched domains bear some structural features, such as lipid modifications, GPI anchors, or transmembrane domains, which would explain their segregation with sphingolipids. However, not all proteins bearing these features associate with sphingolipid-enriched domains, suggesting that other elements are relevant in determining their segregation. Specific interactions with lipid components or with unknown adapter proteins cannot be excluded.