Reduction of Sphingomyelin Level without Accumulation of Ceramide in Chinese Hamster Ovary Cells Affects Detergent-resistant Membrane Domains and Enhances Cellular Cholesterol Efflux to Methyl-β-cyclodextrin*

We examined the effects of reduction of sphingomyelin level on cholesterol behavior in cells using 2 types of Chinese hamster ovary cell mutants deficient in sphingomyelin synthesis: LY-A strain defective in intracellular trafficking of ceramide for sphingomyelin synthesis, and LY-B strain defective in the enzyme catalyzing the initial step of sphingolipid biosynthesis. Although the sphingomyelin content in LY-A and LY-B cells was ∼40 and ∼15%, respectively, of the wild-type level without accumulation of ceramide, these mutant cells were almost identical in cholesterol content and also in plasma membrane cholesterol level to the wild-type cells. However, density gradient fractionation analysis of Triton X-100-treated lysates of cells prelabeled with [3H]cholesterol showed that the [3H]cholesterol level in the low-density floating fraction was lower in sphingomyelin-deficient cells than in wild-type cells. When cells were exposed to methyl-β-cyclodextrin, cholesterol was more efficiently fluxed from sphingomyelin-deficient cells than wild-type cells. These results suggest that the steady state level of cholesterol at the plasma membrane is little affected by the sphingomyelin levels in Chinese hamster ovary cells, but that sphingomyelin levels play an important role in the retention of cholesterol in the plasma membrane against efflux to extracellular cholesterol-acceptors, due to interaction between sphingomyelin and cholesterol in detergent-resistant membrane domains.

Both cholesterol and sphingomyelin (SM) 1 are preferentially distributed in the plasma membrane of cells (1). Recent devel-opments in membrane biology have demonstrated various lines of evidence that the plasma membrane has microdomains, termed detergent-resistant membrane (DRM) domains or lipid rafts, which are involved in various cellular events, including signal transduction and membrane trafficking (2,3). DRM domains are highly enriched in cholesterol and sphingolipids, and probably exist as liquid-ordered phase, characterized by a conformationally ordered state of the acyl chains of phospholipids that are laterally diffusible (3). Formation of the liquidordered phase reflects the fact that cholesterol interacts favorably with phospholipid acyl chains in an extended conformation. SM and saturated glycerophospholipids readily form the extended acyl chain conformation, while unsaturated phospholipids having cis-configuration of the double bond do not. Many studies with model membranes have indicated that cholesterol interacts with SM more strongly than with unsaturated glycerophospholipids, the predominant forms of natural glycerophospholipids (4 -7). In addition, previous studies with model membrane vesicles consisting of pure lipids have shown that cholesterol enhances detergent insolubility of SM, and that sphingolipids or saturated glycerophospholipids tending to form the lipid-ordered phase are also important for detergent insolubility of cholesterol (8,9).
Only a few reports, however, have addressed the issue of the participation of SM in the formation of DRM domains in cells, whereas many studies have demonstrated various lines of evidence that cholesterol is responsible for formation and functions of DRM domains in cells using specific biological and chemical tools to inhibit cholesterol synthesis and to remove cholesterol from cells (2,3,10). We previously showed evidence that reduction of the cellular sphingolipid level renders glycosyl phosphatidylinositol (GPI)-anchored proteins more sensitive to bacterial phosphatidylinositol-specific phospholipase C (11) and also enhances solubility of GPI-anchored proteins in the non-ionic detergent Triton X-100 (12). The blockage of ceramide synthesis by fumonisin B 1 , a potent inhibitor of dihydrosphingosine-N-acyltransferase, has been shown to inhibit folate uptake via the GPI-anchored receptors (13), and to enhance conversion of prion protein, a GPI-anchored protein, to its scrapie isoform (14). These previous studies have indicated that sphingolipids are important components of DRM domains. However, a mouse melanoma cell mutant defective in glycosphingolipid synthesis has very recently been shown to retain almost normal DRM domains (15). It is therefore imperative that we determine whether SM is a key component of DRM domains.
Analysis of SM-deficient cells may yield insight into the biological role of SM. Although treatment of cells with bacterial sphingomyelinase (SMase) is a convenient method for depletion of SM from cell surfaces, this method has the unavoidable flaw that degradation of SM by SMase results in accumulation of ceramide, which serves as a modulator of various cellular functions containing the metabolism and dynamics of cholesterol (16 -18). Thus, when using this method, one must carefully eliminate the possibility that effects of SMase treatment of cells are due to accumulation of ceramide, rather than depletion of SM. Unfortunately, this possibility cannot be eliminated by showing that externally supplied ceramides do not mimic the SMase effects, because long-chain natural ceramide in culture medium is hard to incorporate into the plasma membrane of cells, and short-chain unnatural ceramides that are readily incorporated into cells are quite different in physicochemical properties from natural ceramide produced by treatment of cells with SMase. One way of overcoming this problem is to use cell mutants defective in production of SM without accumulation of ceramide. Recently, we have isolated 2 types of Chinese hamster ovary (CHO) cell mutants defective in sphingolipid biosynthesis (19). One is LY-A strain defective in ATPdependent endoplasmic reticulum to Golgi apparatus trafficking of ceramide directed to SM biosynthesis (20). Thus, the level of SM, but not of glycosphingolipids, in LY-A cells is lower than the wild-type level. Another mutant, LY-B strain, has a defect in the LCB1 subunit of serine palmitoyltransferase, the enzyme catalyzing the first step in sphingolipid biosynthesis, thereby being incapable of de novo synthesis of any sphingolipid species.
In the present study, by analysis with CHO cell mutants, we demonstrated that reduction of cellular SM to ϳ15% of the wild-type level without accumulation of ceramide causes no change in the plasma membrane cholesterol level, but that SM plays a role in the retention of plasma membrane cholesterol against efflux to extracellular acceptors of cholesterol. In addition, this study showed for the first time that SM is implicated in the formation of DRM domains in cells.
Cells and Cell Cultures-The CHO-K1 cell line was obtained from the American Type Culture Collection (ATCC CCL). CHO-K1-derived mutant cell lines, LY-A and LY-B strains, have been previously established by us (19). LY-B/cLCB1, a corrected revertant of LY-B strain, was previously obtained by stable transfection of LY-B cells with the cDNA encoding hamster LCB1 subunit of serine palmitoyltransferase (19). Ham's F-12 medium supplemented with 10% newborn calf serum, penicillin G (100 units/ml), and streptomycin sulfate (100 g/ml) was used as a normal culture medium. Nutridoma-BO medium (F-12 medium containing 1% Nutridoma-SP (Roche Molecular Biochemicals), 0.1% fetal bovine serum, and 10 M sodium oleate-bovine serum albumin complex, and gentamicin (10 g/ml)) was used as a sphingolipid-deficient culture medium (19). CHO cells were maintained in the normal culture medium in the routine manner in a 5% CO 2 atmosphere in 100% humidity at 33°C. For cultivation in a sphingolipid-deficient medium, cells were seeded, incubated in the normal culture medium at 37°C for 1 day, and, after washing twice with serum-free Ham's F-12 medium, cultured in Nutridoma-BO medium for 2 days. In some experiments, Nutridoma-BO medium was supplemented with 1 M D-erythro-sphingosine (Matreya, Inc., Pleasant Gap, PA) as described previously (19).
Assay for Viability of Cells Exposed to Amphotericin B or M␤CD-Cells were seeded on day 0 at 2 ϫ 10 4 cells/well (in 12-well plates) in 2 ml of the normal culture medium. On day 1, each well of cells received 2 ml of Nutridoma-BO medium. On day 3, cell monolayers were incubated in 0.5 ml of Ham's F-12 medium containing amphotericin B or M␤CD at 37°C. After washing with phosphate-buffered saline (PBS), cell monolayers were incubated in 200 l of MTT (5 mg/ml in PBS) at 37°C for 1 h, and the MTT solution was removed. Formazan produced in the cells was dissolved in 400 l of 40 mM HCl in propan-2-ol, and the absorbance at 570 nm of the solution was measured with a spectrophotometer.

Labeling of Cells with [ 3 H]Cholesterol and [ 14 C]Choline
Chloride-Cells were seeded on day 0 at 5 ϫ 10 4 cells/well (in 12-well plates) in 2 ml of the normal culture medium. On day 1, each well of cells received 2 ml of Nutridoma-BO medium. On day 2, cell monolayers were incubated for 16 h in 2 ml of Nutridoma-BO medium containing 1 Ci of [ 3 H]cholesterol or 0.5 Ci of [ 14 C]choline chloride. After washing with serum-free Ham's F-12, the cells were further incubated for 4 h in 2 ml of Nutridoma-BO medium before the following experiments were performed. There were no apparent differences in the levels of cell-associated radioactivity among cell types used. Most of the radioactivity (Ͼ95%) was unesterified cholesterol type in the [ 3 H]cholesterol labeling.
Measurement of Cholesterol Oxidase-sensitive Cholesterol Pool-Treatment of cells with cholesterol oxidase was performed essentially as described previously (21). In brief, cells labeled with [ 3 H]cholesterol were washed with ice-cold PBS and fixed for 10 min on ice with 0.125% glutaraldehyde in PBS. After washing with PBS, fixed cells were incubated in 500 l of PBS containing 2 units/ml cholesterol oxidase with or without 0.1 unit/ml SMase for 30 min at 37°C. Then cells were chilled on ice and washed with PBS. After extraction of cellular lipids from the cells, labeled sterols were separated on TLC plates with a solvent of hexane/diethylether/acetic acid (260/60/3, v/v). The radioactivity in cholesterol and cholestenone separated by TLC was determined by liquid scintillation counting.

Efflux of [ 3 H]Cholesterol and [ 14 C]Choline-containing Phospholipids from Cells to M␤CD-After being rinsed with serum-free Ham's F-12, [ 3 H]cholesterol or [ 14 C
]choline-labeled cell monolayers were incubated in 0.5 ml of Ham's F-12 medium containing M␤CD at 37°C, as described in figure legends. Then the medium incubated with cells was retrieved and centrifuged (5,000 ϫ g, 5 min), and the supernatant was regarded as the medium fraction. Cell monolayers were washed with PBS and lysed with 500 l of 0.1% sodium dodecyl sulfate. The radioactivity of each 100 l of the medium fraction and the cell lysate fraction was counted by liquid scintillation counting.
Flotation of DRM Fractions-DRM fractions were separated as described previously with a few modifications (22). In brief, subconfluent cells labeled with [ 3 H]cholesterol in a 10 cm-dish were extracted for 30 min on ice with 100 l of 1% Triton X-100 in MN buffer (25 mM MES-NaOH (pH 6.5) containing 0.15 M NaCl and a protease inhibitor mixture (Complete TM EDTA-free, Roche Molecular Biochemicals)). The extracts were diluted into 500 l of MN buffer, mixed with 500 l of 85% (w/v) sucrose in MN buffer, and layered under 7 ml of a 10 -30% sucrose gradient in MN buffer. After centrifugation at 75,000 ϫ g for 20 h at 4°C, fractions (0.5 ml) were collected from the bottom of the resulting gradient.
Determination of Content of Various Lipids in Cells-Cells were seeded at 7.5 ϫ 10 5 cells in 30 ml of normal culture medium per 150-mm dish, incubated for 1 day, and then cultivated in 30 ml of Nutridoma-BO medium for 2 days. For supplementation with sphingosine, 150 l of 200 M D-erythro-sphingosine/bovine serum albumin complex was added to the medium at a final concentration of 1 M every day during culture in Nutridoma-BO medium (19). The cells were harvested by scraping, and lipids were extracted from the cells (23). Then phospholipids were separated on TLC plates with a solvent of chloroform/ methanol/acetic acid (65/25/10, v/v), and the content of the individual phospholipid type was determined by the method of Rouser et al. (24), using NaH 2 PO 4 as a standard. The amount of ceramide was determined by a bacterial diacylglycerol kinase method with an assay kit (Amersham Pharmacia Biotech), according to the manufacturer's instructions. The amounts of N-acetylneuraminyl lactosylceramide (G M3 ) and glucosylceramide (GlcCer) were determined by densitometric analysis, as described previously (19). Average molecular weight values of natural G M3 , GlcCer, and ceramide were assumed to be 1198, 727, and 565, respectively. The amount of cholesterol was determined by the cholesterol oxidase method, as described previously (25).
Determination of Protein Concentration-Protein concentrations were determined by the method of Lowry et al. (26) using bovine serum albumin as a standard.

Reduction of the SM Level without Accumulation of Ceramide in CHO Mutant
Cells-The CHO cell mutant LY-B strain defective in the LCB1 subunit of serine palmitoyltransferase is unable to synthesize any sphingolipid species de novo, and another mutant LY-A strain defective in ATP-dependent trafficking of ceramide from the endoplasmic reticulum to the Golgi apparatus for SM synthesis is impaired in SM synthesis but not ceramide or glycosphingolipid synthesis (19,20). As shown in Table I, when cells were cultured in a sphingolipid-deficient medium for 2 days, the SM levels in LY-B and LY-A cells were ϳ15 and ϳ40%, respectively, of the level (25.1 Ϯ 1.0 nmol/mg of protein) in wild-type CHO-K1 cells, being in agreement with our previous study (19). The contents of G M3 , the sole ganglioside in CHO cells, in LY-B and LY-A cells were ϳ20 and ϳ90%, respectively, of the wild-type level (3.2 Ϯ 0.3 nmol/mg of protein), and GlcCer contents in LY-B and LY-A cells were ϳ20 and ϳ110%, respectively, of the wild-type level (1.1 Ϯ 0.1 nmol/mg of protein). Other glycosphingolipids were undetectable even in CHO-K1 cells. The contents of ceramide in LY-B and LY-A cells were ϳ5 and ϳ100%, respectively, of the wildtype level (0.5 nmol/mg of protein). The reduced contents of SM and G M3 in LY-B cells were reversed to the wild-type levels by genetic complementation of the LY-B strain with the hamster LCB1 cDNA (this revertant was designated the LY-B/cLCB1 strain) or by supplementation of the culture medium with D-erythro-sphingosine, which is metabolically converted to complex sphingolipids in cells. Therefore, these CHO cell strains represented an appropriate model cell system for examining whether reduction of SM level without any increase in ceramide level affected behavior of cholesterol in intact cells.
Cholesterol Content in SM-deficient CHO Cells-After culture of cells in the sphingolipid-deficient medium for 2 days, the content of cholesterol in cells was determined. There was no appreciable difference in the content of unesterified cholesterol among LY-A, LY-B, LY-B/cLCB1, and wild-type cells (Table II). Cultivation of cells under sphingosine-supplied conditions did not significantly affect the cholesterol level. The content of esterified cholesterol was too low to be accurately determined in any of the cell types used (data not shown), because cells were cultured under lipoprotein-deficient conditions. These results were consistent with those of our previous study showing that reduction in cellular sphingolipid level through inactivation of SPT did not affect cholesterol level in CHO cells (12,19).
Effect of SM Deficiency in CHO Cells on Amphotericin B Susceptibility-To address the question of whether reduction in cellular SM level affected the level of plasma membraneassociated cholesterol, we first compared amphotericin B sus-ceptibility between SM-deficient and control cells. Cytolytic susceptibility of cells to the polyene macrolide amphotericin B has been suggested to represent a semi-quantitative measure of plasma membrane cholesterol level in intact cells because reduction of the plasma membrane cholesterol level in CHO cells renders the cells resistant to amphotericin B, which exhibits a cytolytic activity after specific binding to sterols at the cell surface (27,28). After cultivation in the sphingolipid-deficient medium for 2 days, cells were incubated with amphotericin B at various concentrations for 1 h at 37°C, and then the viability of the cells was determined. Wild-type cells lost viability after exposure to amphotericin B in a dose-dependent manner with a LD 50 (LD 50 ) of ϳ7 g/ml, and the LD 50 values of amphotericin B susceptibility in LY-A, LY-B, and LY-B/cLCB1 cells were also found to be 4 -7 g/ml (Fig. 1). The observation that LY-A and LY-B cells did not display any higher resistance to amphotericin B than wild-type cells suggested that there was no reduction in plasma membrane cholesterol level in the SM-deficient mutant cells, compared with the control cells. On the other hand, analysis of the binding capability of cells to the cytolysin lysenin, the specific membrane receptor of which is SM, has shown that SM levels in the plasma membranes of LY-A and LY-B cells are lower than the wild-type level (19).    When the fixed cells were not treated with SMase, the apparent pool size of cholesterol oxidase-sensitive [ 3 H]cholesterol was inversely correlated to the SM content in cells (Fig. 2). This observation is explained by the previously elucidated mechanism that the accessibility of cholesterol oxidase to membrane cholesterol is restricted by SM coexisting in the membrane (21). Thus, this result demonstrated another line of evidence that plasma membrane SM levels are lower in LY-A and LY-B cells than in wild-type cells.
Reduced Distribution of Cholesterol to DRM Domains in SMdeficient CHO Cells-We next examined whether the distribution of cholesterol to DRM domains was affected by the reduction in cellular SM level. Lipid and protein components associated with DRM are thought to be fractionated into a Triton X-100-insoluble floating fraction by sucrose density gradient centrifugation (22). Thus, lysates of the [ 3 H]cholesterollabeled cells treated with Triton X-100 were separated in a sucrose density gradient, and the radioactivity of [ 3 H]cholesterol in the fractions enriched in caveolin, a protein marker of DRM domains, was determined. As shown in Fig. 3, the lowdensity floating fraction (fraction number 7-13) contained ϳ60% of total [ 3 H]cholesterol radioactivity in the extract from wild-type cells, whereas the low-density floating fraction in LY-A and LY-B had less than 40 and 30%, respectively, of total [ 3 H]cholesterol radioactivity in the cell extracts (Fig. 3). Western blot analysis showed that the expression levels of caveolin were nearly identical among these 3 cell types (data not shown). In addition, SM and caveolin in these 3 cell types were highly enriched in low-density floating fractions (data not shown). These results indicated that the reduction of cellular SM level affects distribution of cholesterol to DRM domains.
M␤CD serves as a high-affinity acceptor for cholesterol (30 -  (Fig. 3). The reduction in [ 3 H]cholesterol distribution in the DRM domains was more marked in LY-A and LY-B cells than in wild-type cells (Fig. 3). Note that the 3 cell types sustained ϳ100% viability after exposure to 5 mM M␤CD for 5 min (data not shown, and see also Fig. 5B). These results suggested that efflux of cellular cholesterol to M␤CD mainly reduced the level of cholesterol associated with DRM but not non-DRM domains, and that the reduction ratio was inversely correlated with cellular SM content.
Efflux of Cellular Cholesterol and Phospholipids from SMdeficient CHO Cells in the Presence of M␤CD-To confirm that the reduction in cellular SM level affected M␤CD-mediated efflux of cholesterol from cells, we determined time courses of cholesterol efflux from wild-type and SM-deficient mutant cells to a serum-free medium containing M␤CD. After labeling of cells with [ 3 H]cholesterol in the sphingolipid-deficient medium for 16 h, the labeled cells were exposed to 1 mM M␤CD for various times up to 60 min, and the radioactivity of [ 3 H]cholesterol released from the cells to the medium was determined. The M␤CD-mediated [ 3 H]cholesterol efflux rates from LY-A and LY-B cells were ϳ2and ϳ3-fold, respectively, of the wildtype level throughout the time window examined, while there was no significant difference in the efflux rate between wildtype cells and LY-B/cLCB1 cells (Fig. 4A). The enhanced [ 3 H]cholesterol efflux from LY-B cells was suppressed to the wild-type level when cells were cultured under sphingosinesupplied conditions. The enhanced [ 3 H]cholesterol efflux from SM-deficient cells was not due to cell lysis during exposure to M␤CD because almost all cells sustained trypan blue extrusion ability under the conditions used.
To determine the lipid specificity of M␤CD-mediated efflux, the efflux of cellular phospholipids was examined. After metabolic labeling of phospholipids with [ 14 C]choline for 16 h, cells were exposed to 1 mM M␤CD for various periods, and the radioactivity of [ 14 C]choline-labeled phospholipids released to the medium was measured. The fraction of [ 14 C]choline-labeled phospholipids released during 1-h exposure to 1 mM M␤CD was less than 5% of the initial cell-associated level in all cell types examined (Fig. 4B). These control experiments confirmed the specific enhancement of cholesterol efflux via M␤CD through SM deficiency in cells and also the lack of appreciable lysis of cells.
SM-deficient CHO Cells Are Hyper-susceptible to M␤CD-mediated Cell Lysis-We observed that exposure of CHO cells to M␤CD at higher concentrations caused greater lysis of cells. This observation and the adverse correlation between M␤CDmediated cholesterol efflux and cellular SM content raised the possibility that the reduction in cellular SM level enhanced the cytolytic susceptibility of the cells to M␤CD. This was the case. When wild-type and LY-B/cLCB1 cells were exposed to various concentrations of M␤CD for 1 h at 37°C, the cells maintained ϳ100% viability at 10 mM M␤CD, but lost viability almost completely at 20 mM M␤CD (Fig. 5A). In contrast, SM-deficient cells were much more susceptible to this drug, and almost complete loss of the viability of LY-A and LY-B cells was observed at 5 and 2.5 mM, respectively, of M␤CD (Fig. 5A). When cells were exposed to 5 mM M␤CD for various periods, LY-A and LY-B cells almost completely lost their viability by 90 min, whereas wild-type and LY-B/cLCB1 cells sustained ϳ100% viability at least up to 90 min (Fig. 5B). The enhanced M␤CD susceptibility of LY-B cells was suppressed to near the wildtype level by addition of D-erythro-sphingosine to the culture medium (Fig. 5B).
We also determined whether depletion of plasma membrane SM by SMase treatment affected M␤CD susceptibility. When wild-type cells were treated with 250 milliunits/ml SMase for 1 h at 37°C, the cells showed hyper-susceptibility to M␤CD (Fig. 6), similar to the case with SM-deficient mutant cells. By contrast, treatment of cells with PC-PLC caused no appreciable change in M␤CD susceptibility (Fig. 6). Collectively, these results indicated that SM plays a role in the retention of cholesterol in the plasma membrane against M␤CD-mediated efflux.

DISCUSSION
In this study, we investigated the effects of reduction of cellular SM on cholesterol behavior in the plasma membrane using CHO cell mutants defective in SM production. There were no differences in the content of cellular cholesterol and its distribution to the cholesterol oxidase-sensitive pool between these mutant and wild-type cells, even when the SM content in the mutant cells was 15-40% of the wild-type level (Tables I  and II, and Fig. 2). LY-B cells were also deficient in ceramide and glycosphingolipids, while the ceramide and glycosphingolipid content in LY-A cells was similar to the wild-type levels  (Table I, and see also Ref. 19). Hence, the possibility was eliminated that the failure to affect the cholesterol levels under SM-deficient conditions was due to functional compensation by an increase of ceramide or glycosphingolipid content. These results indicated that the reduction of cellular SM to ϳ15% of the wild-type level has no effect on the steady-state level of cellular cholesterol and its distribution to the plasma membrane. Enrichment of SM in the plasma membrane has been hypothesized to force cholesterol to be also enriched in the plasma membrane, based on the observations that addition of SM to cultured human skin fibroblasts results in an increase in cellular cholesterol content according to the increase in cellular SM level (35), and that cholesterol interacts with SM more strongly than with glycerophospholipids in model membranes (6,36,37). However, our analyses with CHO cells suggest that neither SM, glycosphingolipids, nor ceramide is essential for the preferential distribution of cholesterol to the plasma membrane. Consistent with these results, a recent study has shown that most cellular cholesterol is present in the plasma membrane even after depletion of cell surface SM by treatment of human skin fibroblasts with bacterial SMase (38). It is unlikely that the reduced level of SM in the CHO mutant cells is still sufficient for a stoichiometric association with cholesterol at the plasma membrane, because the estimated amount of SM in LY-B cells (ϳ20 nmol/mol cellular phospholipid) is far less than that of cholesterol (ϳ300 nmol/mol cellular phospholipid) (Tables I and II), and 70 -80% of cholesterol in both LY-B and wild-type cells appears to be distributed in the plasma membrane (Fig. 2). Energy-dependent transport of de novo synthesized cholesterol from the endoplasmic reticulum to the plasma membrane may be primarily responsible for enrichment of cholesterol in the plasma membrane (39,40).
We previously showed that sphingolipids are involved in the detergent insolubility of a GPI-anchored protein in CHO cells (12). However, another research group has recently shown that glycosphingolipids are dispensable for formation of DRM domains in murine melanoma cells (15), although the original concept of lipid-raft is based on the assumption that the hydrogen bond interaction between the sugar head groups of glycosphingolipids and GPI-anchored proteins is important for formation of detergent-resistant rafts (41). Ostermeyer et al. (15) also demonstrated that there is no difference in M␤CD-mediated efflux of cholesterol between glycosphingolipid-depleted cells and control cells. We herein demonstrated that CHO mutant LY-A cells having a lower level of SM, but not of glycosphingolipids, had a lower level of DRM-associated cholesterol than wild-type cells (Fig. 3). These results indicate that SM plays a role in the formation of DRM domains in CHO cells. This conclusion is consistent with the observation that the liquid-ordered phase of lipid bilayers forms in model membranes containing cholesterol and SM (42). In model membranes, total mol % of lipids tending to form the liquid-ordered phase has been shown to be important for the formation of DRM domains (8,9,42). Considering that levels of total sphingolipids in SM-deficient CHO mutant cells are lower than the wild-type level (Table I), the total level of SM and glycosphingolipids, but not the level of SM itself, might be a key factor for the formation of DRM domains in cells (see also below).
Various investigators have demonstrated that treatment of cells with exogenous SMase enhances efflux of cellular cholesterol to the extracellular cholesterol acceptors, such as high density lipoproteins and cyclodextrins (31,(43)(44)(45)(46). Nevertheless, these previous studies did not eliminate the possibility that the enhanced efflux of cholesterol after SMase treatment was due to accumulation of ceramide. We herein demonstrated that the reduction of cellular SM content without any increase in ceramide level renders plasma membrane cholesterol prone to flux to the extracellular acceptor M␤CD (Fig. 4). Thus, we conclude that SM play a role in the retention of cholesterol in the plasma membrane against efflux to the extracellular cholesterol-acceptor, probably due to favorable van der Waals and hydrogen-bonding interactions between cholesterol and SM (47,48). This conclusion does not exclude the possibility that total sphingolipids (SM, ceramide, and glycosphingolipids) or complex sphingolipids (SM and glycosphingolipids) are responsible for the retention of cholesterol. However, treatment of cells with bacterial SMase not only enhances cholesterol efflux from cells (Refs. 31 and 43-46, and Fig. 6) and also reduces cholesterol distribution to DRM domains (46), although total sphingolipid levels most likely remain unchanged after SMase treatment. Therefore, levels of SM itself or total complex sphingolipids in cells are likely to be important for the formation of DRM domains and the retention of cholesterol against efflux.
Interestingly, M␤CD-mediated efflux of cholesterol from cells results in reduction of cholesterol level in DRM, but not non-DRM domains (Fig. 3). The simplest explanation for this observation is that M␤CD-mediated efflux of cholesterol preferentially occurs at the DRM domains in cells. Fielding and Fielding (22,49) have suggested that cholesterol efflux from the cell surface to high density lipoproteins occurs largely at DRM domains of the plasma membrane. Alternatively, M␤CDmediated efflux of cholesterol may occur at non-DRM domains of the plasma membrane, but cellular cholesterol is rapidly redistributed from DRM to non-DRM domains for homeostasis of the cholesterol level of non-DRM domains. A previous study has shown that de novo synthesized cholesterol at the endoplasmic reticulum is delivered to DRM domains of the plasma membrane by a caveolin-dependent mechanism, and that the cholesterol delivered to DRM domains then rapidly flows to non-DRM domains at the plasma membrane (40).
It is also noteworthy that cytotoxic susceptibility of CHO cells to M␤CD is enhanced when cellular SM level is reduced (Figs. 5 and 6). Because M␤CD-mediated efflux of cholesterol is enhanced in SM-deficient cells (Fig. 4), it is likely that cholesterol is depleted below the level necessary for sustaining viability more easily from SM-deficient cells than from control cells. The higher toxicity of M␤CD to SM-deficient cells than control cells may allow the use of this drug for a positive selection of revertants of SM-deficient mutant cells.