Role of Multiple Drug Resistance Protein 1 in Neutral but Not Acidic Glycosphingolipid Biosynthesis*

Transfection studies have implicated the multiple drug resistance pump, MDR1, as a glucosyl ceramide translocase within the Golgi complex (Lala, P., Ito, S., and Lingwood, C. A. (2000) J. Biol. Chem. 275, 6246–6251). We now show that MDR1 inhibitors, cyclosporin A or ketoconazole, inhibit neutral glycosphingolipid biosynthesis in 11 of 12 cell lines tested. The exception, HeLa cells, do not express MDR1. Microsomal lactosyl ceramide and globotriaosyl ceramide synthesis from endogenous or exogenously added liposomal glucosyl ceramide was inhibited by cyclosporin A, consistent with a direct role for MDR1/glucosyl ceramide translocase activity in their synthesis. In contrast, cellular ganglioside synthesis in the same cells, was unaffected by MDR1 inhibition, suggesting neutral and acid glycosphingolipids are synthesized from distinct precursor glycosphingolipid pools. Metabolic labeling in wild type and knock-out (MDR1a, 1b, MRP1) mouse fibroblasts showed the same loss of neutral glycosphingolipid (glucosyl ceramide, lactosyl ceramide) but not ganglioside (GM3) synthesis, confirming the proposed role for MDR1 translocase activity. Cryo-immunoelectron microscopy showed MDR1 was predominantly intracellular, largely in rab6-containing Golgi vesicles and Golgi cisternae, the site of glycosphingolipid synthesis. These studies identify MDR1 as the major glucosyl ceramide flippase required for neutral glycosphingolipid anabolism and demonstrate a previously unappreciated dichotomy between neutral and acid glycosphingolipid synthesis.

GlcCer 1 is the precursor of the majority of eukaryotic GSLs (1). The UDP galactose ␤1-4 glucosyl ceramide glycosyltransferase (LacCer synthase) within the Golgi lumen then provides the common precursor, LacCer for the synthesis of most complex GSL. Drug-resistant tumor cells have been shown to increase GlcCer synthesis (2,3), and this has been postulated as a mechanism for drug resistance by providing an alternative biosynthetic destination for ceramide, which might otherwise induce apoptosis (4). GlcCer is synthesized from ER-derived ceramide, on the cytosolic cis-Golgi surface (5)(6)(7), and may be transported to the plasma membrane from this site (8). However, the active sites of the glycosyltransferases responsible for further elongation of GSLs are located within the Golgi lumen (9,10). Thus, a mechanism for the translocation of GlcCer from the cytosolic surface of the Golgi to the lumen must exist. We have shown that transfection of cells with the mdr1 gene results in a marked elevation of GlcCer, LacCer, and Gb 3 levels with a concomitant increase in cell sensitivity to the Gb 3 -binding verotoxin 1. Both effects were reversed in the presence of MDR1 inhibitors (11). We proposed that MDR1 can function as a GlcCer translocase to flip GlcCer to the inner Golgi surface to enhance LacCer and subsequent Gb 3 synthesis. MDR1 has been shown to function as a membrane translocase for chemically modified short chain fatty acid GlcCer (8). However such derivatized, indicator GSLs have been considered possible drug substrates for MDR1, and the ability of MDR1 to translocate natural, long-chain GSLs has been questioned on the basis of lack of overt phenotype in MDR1 knockout mice (12). In the present study, we show that fibroblasts from such mice are indeed defective in GSL synthesis, in the manner predicted from the in vitro MDR1 inhibition studies now reported.
In this study, we address the question as to whether MDR1, which is widely expressed in normal tissues (13), is involved in GSL synthesis in cultured cells in general. These studies fill a major gap in understanding the mechanism of GSL biosynthesis and demonstrate a new, unsuspected distinction between acidic and neutral GSL synthesis.

Cell Culture
MDCK cells transfected with the human MDR1 cDNA were a gift from Dr. M. Gottesman (National Institutes of Health, Bethesda, MD) (14). The human astrocytoma cell line SF-539, and IOMM Lee and CH157 MN meningioma cell lines were kindly provided by Dr. J. Rutka (Hospital for Sick Children, Toronto, Canada). Vero, HEp-2, SF-539 astrocytoma, ovarian carcinoma SK VLB (MDR variant of the parental SKOV3 cell line (15)), MDR1-MDCK cells were maintained in ␣-minimal essential medium supplemented with 5% or 10% FBS and 40 g/ml gentamicin. MDR1-MDCK medium also contained 80 ng/ml colchicine, and SK VLB medium contained 1 g/ml vinblastine. HeLa cells, NIH 3T3 cells, and meningioma IOMM Lee and CH157 MN cell lines were maintained in Dulbecco's modified Eagle's medium (with 4.5 gm/liter of glucose for HeLa and 3T3, and with 1 gm/liter of glucose for meningioma cell lines) with 10% FBS. Daudi Burkitt's lymphoma and Jurkat E6.1 cells were maintained in RPMI 1640 medium also supplemented with 10% FBS. ECV 304 bladder carcinoma cell line (16) was maintained in M199 medium with 10% FBS and 40 g/ml gentamicin. KOT * This work was supported by the Canadian Cancer Society via Grant 011090 from the National Cancer Institute of Canada and Canadian Institutes for Health Research Grant MT13073. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Verotoxin 1 (VT1) Cytotoxicity
VT1 was purified as described previously (19). Target cells in logarithmic growth phase were cultured in microtiter plates and incubated in triplicate, with increasing dilutions of VT1. After 72 h, the cells remaining attached were stained with 0.1% crystal violet and quantitated using a microtiter plate reader as previously (20). Non-adherent Daudi and Jurkat cell survival data were calculated by 1% Trypan Blue dye exclusion. For MDR1 inhibition studies, CsA or ketoconazole was added to cells 4 or 5 days, respectively, prior to VT1 and maintained during the cytotoxicity assay.

GSL Extraction
GSLs were extracted from 5 ϫ 10 6 exponential growth phase cells. Adherent cells were scraped, and all cells were pelleted by centrifugation at 1000 ϫ g for 10 min. The cell pellet was extracted with 20 volumes of CHCl 3 /CH 3 OH (2:1) as described (21). The extract was partitioned against water, and the lower phase was applied to a silica column in CHCl 3 /CH 3 OH (98:2). The column was washed extensively with CHCl 3 and the glycolipid fraction eluted with CH 3 COCH 3 /CH 3 OH (9:1) (22). Gangliosides were separated from pooled upper and lower phases by DEAE-Sephadex G-25 chromatography and eluted using 0.4 M ammonium acetate in methanol (23).

VT1/TLC Overlay
Equivalent GSL aliquots, based on cell number, were separated by TLC, and the plates were dried and blocked with 1% gelatin in water at 37°C overnight. After washing with 50 mM Tris-buffered saline, pH 7.4, plates were incubated with 0.1 g/ml VT1 for 1 h at room temperature. After washing, the plates were incubated with monoclonal anti-VT1 antibody (24) (2 g/ml) followed, after washing, by peroxidase-conjugated goat anti-mouse antibody. VT1 bound was visualized with 4-chloro-1-naphthol (22).

Microsomal Preparation
Eighty percent confluent MDR1-MDCK cells were washed, scraped into PBS, and centrifuged for 5 min at 300 ϫ g. The pellet was resuspended in buffer (10 mM Tris-HCl, pH 7.4, 10 mM KCl, 1.5 mM MgCl 2 , 10 mM dithiothreitol, 0.5 M sucrose, and 10 mM phenylmethylsulfonyl fluoride) and homogenized using a Dounce homogenizer. The homogenate was centrifuged at 800 ϫ g for 10 min at 4°C, and the supernatant was further centrifuged at 10,000 ϫ g for 10 min at 4°C. The resulting supernatant was centrifuged at 100,000 ϫ g for 10 min, and the pellet was resuspended in 10% of the supernatant (to give a crude, cell free "microsomal" fraction). Protein content was measured by BCA assay (Pierce).

Cell-free Metabolic Labeling
GSLs in MDR1-MDCK microsomes were metabolically labeled with UDP-[ 14 C]galactose. 100 l of the microsomal fraction (100 g of protein) was incubated in Tris-HCl buffer, pH 7.4 (10 mM), and MnCl 2 (10 mM) Ϯ 4 M CsA at 37°C for 2 h, with shaking. 0.2 Ci of UDP-[ 14 C]galactose was then added to a 300-l final volume and incubated for a further 3 h at 37°C with shaking. For some assays, exogenous liposomal GlcCer (60 g sonicated in 20 l of H 2 O) was added during the latter incubation.

Metabolic Labeling of Cellular GSLs
MDR1-MDCK Cells-GSLs of exponential growth phase cells were metabolically labeled with [ 14 C]serine (0.5 Ci/ml) for 72 h, Ϯ 4 M CsA. Cells were rinsed twice with PBS and harvested by gently scraping. Cellular lipids were extracted and applied on DEAE-Sephadex as above. Neutral GSLs were unbound, and gangliosides were eluted with ammonium acetate (23). Lipids were separated by TLC and labeled species quantitated by phosphorimaging.
Mouse Wild Type and Knockout (KOT) Fibroblasts-Cells (wild type and KOT11 fibroblasts) were grown, in duplicate, in the presence of 2.5 Ci/ml [ 14 C]acetate or 1 Ci/ml [ 14 C]serine for 60 h in Dulbecco's modified Eagle's medium plus 10% fetal calf serum. Medium was removed and replaced with 3 ml of CHCl 3 :CH 3 OH:H 2 O (1:2.2:0.2, v/v), and extraction was allowed to proceed for 10 min. 3 ml was then transferred to an 11-ml glass tube, and 0.88 ml of CHCl 3 , 0.78 ml of Hanks' balance salt solution, and 0.78 ml of 10 mM acetic acid were added. Cells were then vortexed, and the lower phase was removed. The lower phase was dried and separated by two-dimensional TLC (first dimension, CHCl 3 :CH 3 OH:25% NH 4 OH:H 2 O (65:35:4:4, v/v); second dimension, CHCl 3 :CH 3 COCH 3 : CH 3 OH:CH 3 COOH:H 2 O (50:20:10:10:5, v/v)) and radioactive species quantitated by phosphorimaging and identified relative to standards. Wild type and KOT 51 fibroblasts were similarly labeled with [ 14 C]serine Ϯ 20 M of indomethacin, a selective inhibitor of MRP1 (25). In this case, the dried lipid extract was saponified in 600 l of 0.2 M KOH in MeOH for 3 h at 37°C, neutralized with 1.6 ml of 10 mM HAc, 2 ml of CHCl 3 , 1.8 ml of MeOH, and partitioned. Both phases were recombined, and polar products were removed on a SepPak C18 column. The methanol-eluted fraction was then separated by two-dimensional TLC as above.
Immunostaining of Cell Surface MDR1-HeLa, MDR1-MDCK, Vero, SK VLB, IOMM Lee, and CH157 MN cells were grown on glass coverslips. Washed cells were incubated with MRK-16 monoclonal anti-MDR1 antibody (10 g/ml; Kamiya) for 1 h at 4°C, followed by fluorescein isothiocyanate-conjugated goat anti-mouse secondary antibody, also at 4°C for 1 h. Following extensive washing with 50 mM PBS, the coverslips were fixed in 4% paraformaldehyde in PBS for 30 min at room temperature, mounted with DABCO, and examined and photographed using a Leica DMIRE 2 (Leica Canada, Ontario) fluorescence microscope under incident UV illumination. All images were recorded at the same exposure.
Post-embedding Immunogold Cryo-electronmicroscopy-Logarithmic phase MDR1-MDCK cells were scraped and fixed in 4% paraformaldehyde, 0.1% glutaraldehyde in 0.1 M Sorenson's phosphate buffer, pH 7.4, for 2-4 h at room temperature. Cells were then pelleted by centrifugation at 1000 ϫ g for 5 min, washed thoroughly in phosphate buffer, embedded in 15% gelatin, cut into mm 3 -pieces, and infused with 2.3 M sucrose for several hours. The blocks were then mounted on aluminum cryo-ultramicrotomy pins and frozen in liquid nitrogen. Ultrathin cryosections were cut with a diamond knife at Ϫ95°C using a Leica Ultracut R cryo-ultramicrotome (Leica Canada, Ontario). Sections were transferred to Formvar-coated nickel grids in a loop of molten sucrose, and the grids were washed thoroughly in PBS containing 0.15% glycine and 0.5% BSA and PBS containing BSA alone. Sections were then incubated with a polyclonal rabbit antibody against Golgi marker rab6 (32) for 1 h. Following a thorough rinse in PBS/BSA, samples were incubated in a goat anti-rabbit IgG 5-nm gold complex (Amersham Biosciences, Quebec, Canada) for 1 h. This procedure was repeated on the same specimens except MRK-16 anti-MDR1 antibody was used as the primary antibody and goat anti-mouse IgG 10-nm gold complex as the secondary antibody. Sections were then rinsed thoroughly with PBS followed by distilled water and stabilized in a thin film of methylcellulose containing 0.2% uranyl acetate. Controls included the omission of primary and secondary antisera and the use of an irrelevant antibody (either poly-or monoclonal anti-glial fibrillary acidic protein). Samples were then examined in a JEM 1230 transmission electron microscope (JEOL, Peabody, MA), and images were recorded using a charge-coupled device camera (AMT Corp.). Controls were uniformly negative for gold particles. Image analysis was performed on a minimum of 100 images from each group at a nominal magnification of 100,000ϫ using an image analysis program (Image Pro Plus, Media Cybernetics) to determine cell particle density and percentage particle colocalization (where colocalization was defined as 5-and 10-nm particles being within 20 nm of each other).
Data Analysis and Statistics-Differences in CD50 values for each of the cell lines, obtained before and after treatment with CsA and ketoconazole, were compared by 2-tailed non-parametric Student's t test. Means and standard deviations for each experimental point, repeated at least three times, as well as the statistical analysis were performed by GraphPad Prism 3.0 (GraphPad Software Inc., San Diego, CA). A value of p Ͻ 0.05 was considered significant.

MDR1 Inhibitors Inhibit Gb 3 Synthesis and Protect
Cells against VT1 Cytotoxicity-The Gb 3 -containing cell lines, MDR1-MDCK (canine epithelial), Vero (monkey epithelial), Hep-2 (human epithelial), HeLa (human cervical carcinoma), Daudi (human B lymphoid), SF-539 (human astrocytoma), SK VLB (human ovarian carcinoma), Iomm Lee (human meningioma), CH157 MN (human meningioma), and ECV 304 (human bladder carcinoma) cell lines were analyzed for neutral GSL and Gb 3 content and VT1 sensitivity, before and after ketoconazole or CsA treatment ( Fig. 1 and Table I). For each cell line, these drugs were used at the maximum concentration, which alone had no effect on cell viability. Inhibition of MDR1 using either drug protected all cell lines (except HeLa cells) against VT1 cytotoxicity (Table I). The lower phase GSL fraction from control, CsA-treated, or ketoconazole-treated cells was separated by TLC and Gb 3 identified by VT1/TLC overlay. In each case, (except HeLa cells), the Gb 3 (and LacCer and GlcCer) content was significantly reduced in treated cells, correlating with the protection afforded against VT1 cytotoxicity. In HeLa cells, VT1 sensitivity and their Gb 3 content was largely unaffected by either MDR1 inhibitor. The Gb 3 content and VT1 sensitivity of SK VLB and ECV 304 cells was less effectively reduced after CsA, as opposed to ketoconazole treatment, consistent with our finding that the vinblastine sensitivity of these cell lines was affected only by ketoconazole (not shown). Despite the fact that Gb 3 synthesis in ECV 304 and SF-539 cells was significantly inhibited by CsA, a reduced effect on VT1 sensitivity was observed. NIH 3T3 cells are VT1-insensitive and contain only trace levels, if any, Gb 3 (28). The neutral GSLs of mouse 3T3 cells were eliminated by CsA treatment. Similarly, human T cells do not express Gb 3 (29) and the neutral GSLs of the Jurkat T cell line (primarily GlcCer) were lost following culture with CsA (see Figs. 1 and 3a).
Effect of CsA on Cell-free Glycolipid Synthesis-GSL synthesis (LacCer and some Gb 3 ) in a crude MDR1-MDCK microsomal fraction, monitored by [ 14 C]UDP-galactose incorporation, was inhibited by CsA (Fig. 2). Addition of exogenous GlcCer increased LacCer labeling, but CsA inhibition was maintained.

Selective CsA Inhibition of Neutral Glycolipid as Opposed to
Ganglioside Synthesis-GlcCer is the precursor of both gangliosides and neutral GSLs. We therefore compared the effect of ketoconazole or CsA on the synthesis of Gb 3 and GM3, the simplest ganglioside, in MDR1-MDCK and Vero cells. The effect of CsA on Jurkat cell GlcCer, the only neutral GSL made, and ganglioside content, was compared. In Vero cells cultured with ketoconazole and MDR1-MDCK cells cultured with CsA, Gb 3 levels were inhibited by Ͼ95%, whereas the level of GM3, the only ganglioside expressed, was unaffected. In Jurkat cells, GlcCer was reduced by Ͼ90%, whereas GM3, GM1, and one unidentified, more polar, ganglioside were unaltered after CsA treatment (Fig. 3a).
Autoradiography of the metabolically labeled neutral and acidic GSL fraction of MDR1-MDCK cells cultured with or without CsA, shows that, although Gb 3 synthesis was significantly reduced, the synthesis of GM3 and the more complex gangliosides were not inhibited at all by CsA (Fig. 3b). Sphingomyelin was identified in the [ 14 C]serine-labeled neutral GSL fraction (Fig. 3a), and labeling was not inhibited, even slightly enhanced, in CsA-treated cells.
Glycolipid/Phospholipid Profile of Mouse mdr1a, 1b mrp1 Knockout Fibroblasts-Mice contain two MDR1 homologous genes, which have been knocked out in combination with MRP1 (17,18). Comparison of the metabolic labeling of GSLs of fibroblasts from knockout mice (mdr 1a Ϫ/Ϫ , 1b Ϫ/Ϫ mrp 1 Ϫ/Ϫ ), and wild type fibroblasts ( Fig. 4 1. Effect of MDR1 inhibitors CsA or ketoconazole on cellular globoseries GSL content. In each TLC panel, the orcinol-stained neutral GSL fraction from the lower phase extract (left) and VT1 overlay (not K and L) to identify Gb 3 (right) are shown. GSL standards (from the top) GlcCer, LacCer, Gb 3 , and Gb 4 were run on each TLC (lane 1). A GM3 standard is included in NIH 3T3 and Jurkat panels, but GM3 cannot be quantitated within the lower phase (cf. Fig. 3a, panels E and F) effect on GSL labeling with [ 14 C]serine in wild type or knockout fibroblasts (Table III).
Cell Expression of MDR1-Because HeLa cell GSL biosynthesis and VT1 sensitivity were unaffected by MDR1 inhibitors, we assessed whether MDR1 was expressed by these cells. By anti-MDR1 Western blot (Fig. 5), a 170-kDa MDR1 band was clearly detected in the MDR1-MDCK cell positive control, but no immunoreactivity was found in the HeLa cell extract. HeLa cells showed no surface immunostaining for MDR1 as compared with a selection of cell lines in which MDR1 inhibitors reduce neutral GSL synthesis (MDR1-MDCK, Vero, SF-539, Iomm Lee, SK VLB, and CH157 MN cells), which served as positive controls (Fig. 5). mRNA screening has shown a lack of MDR1 expression in HeLa cells (30). In some cells (e.g. MDR1-MDCK, SF 539), surface MDR1 expression is remarkably punctate, potentially consistent with a caveolar location (31).
Subcellular Localization of MDR1, Rab6 by Immunoelectron Microscopy-Cryo-immunoelectron microscopy of MDR1-MDCK cells showed the majority of MDR1 is intracellular (Fig.  6a) and membrane-associated (Fig. 6, a-h). Little MDR1 is expressed in the plasma membrane as monitored by immunoelectron microscopy (none seen in Fig. 6a) possibly due to the punctate expression of MDR1 (Fig. 5b), which might be missed in electron microscopy sectioning. This would be consistent with the surface staining seen in some electron microscopy sections (not shown). Significant MDR1 staining is clearly associated with the Golgi marker rab6 (32). In some instances, a limiting vesicular membrane can be seen around the colocalized MDR1 and rab6 (Fig. 6, b and c). These vesicles are primarily located at the termini of the Golgi cisternae (Fig. 6, d  and e). MDR1 is also found within these termini (Fig. 6, f and  g) and to a lesser degree, within the lamellar stacks of the Golgi cisternae (Fig. 6, g and h).

MDR1 Plays a General Role in Neutral GSL Biosynthesis-
GlcCer synthase is the only GSL glycosyltransferase located at the cytosolic surface of the Golgi, rather than within the lumen (6). This requires a mechanism, until now undescribed, for GlcCer translocation into the Golgi for synthesis of the more complex GSLs. GlcCer synthase has been implicated in drug resistance, because its activation may metabolically prevent ceramide accumulation via the action of cytotoxic drugs (3,33) and thereby avoid apoptosis induction. MDR1-mediated GlcCer Golgi translocation has been implicated to augment this process (34). In addition to its drug pump activity, MDR1 has been shown to be a phospholipid/GSL "flippase," monitored using non-physiological short-chain lipid probes (8). MDR1 can traffic natural short-chain phospholipids (35), but defining a role for MDR1 in long-acyl-chain lipid transport has proved more elusive (12). Knockout mice are viable, and no GSL or phospholipid phenotype has been reported. The related MDR2 and MDR3 are phosphatidylcholine translocators (36), and MDR3, at least, has no activity for GSLs (8). Bacterial MDR1 homologues function as lipid flippases in phospholipid and lipid A trafficking (37,38).
We suggested that MDR1 flipping GlcCer from the cytosolic to the luminal Golgi face mediated the increased GlcCer, Lac-Cer, and Gb 3 observed following MDCK cell transfection with MDR1 (11). This increase was reversed by CsA. Our present studies show that MDR1 is within Golgi stacks/vesicles and carries out this function in most cultured cells. This defines a new function for "normal" cell MDR1 and fills a significant gap in understanding the steps involved in GSL biosynthesis.
MDR1 inhibition resulted in the loss of lower phase neutral  50 ) calculated from the dose response was increased for all cells (except HeLa), after culture of cells with CsA or ketoconazole as described in Fig. 1. For each experiment, the effect on cell sensitivity to vinblastine was simultaneously assayed to confirm efficacy to prevent MDR1-mediated drug efflux.  GSLs, GlcCer, LacCer, Gb 3 , and Gb 4 (if present), in all but (MDR1-negative) HeLa cells. (GM3, detected in some cell extracts, appears reduced, but GM3 partitions into both upper and lower phases and cannot therefore be quantitated in the lower phase. Isolation of the ganglioside fraction showed GM3 levels were unaffected.) The reduction in GlcCer following MDR1 inhibition can be attributed to cytosolic glucocerebrosidase (39). CsA inhibition of GlcCer-dependent LacCer biosynthesis in a microsomal assay supports a direct translocase role for MDR1. This assay is complex however, requiring transfer of the GlcCer from the exogenous liposome to the outer microsomal membrane for MDR1 translocation, as well as nucleotide luminal transit (40). The microsomal inhibitory effect of CsA is more rapid than in cell culture, consistent with a more immediate effect on GlcCer translocation. LacCer synthesis is affected in Ͻ5 h, whereas several days of culture are required for intact cells.
An alternative explanation might be that MDR1 inhibitors, in some way, promote neutral GSL catabolism. However, CsA was found to have no effect on cellular ␣-galactosidase activity in vitro (not shown) and did not affect LacCer synthesis from C 6 -GlcCer (41). Moreover, it is unlikely that the lack of MDR1 in the KOT fibroblasts would have a similar effect on GSL catabolism.   3 and 4). Labeled species are quantitated in Table II. MDR1 inhibitors protected sensitive cells from VT1 cytotoxicity (except for HeLa cells). However the effect was not always proportional to the reduction in Gb 3 , likely due to the differential efficacy of Gb 3 lipid isoforms to bind VT (42,43), to be effectively presented within the phospholipid bilayer (44,45) and traffic within the cell (46). Significantly, verapamil, another MDR1 inhibitor (47), has been shown both to protect cells against VT1 (48) and reduce GSL synthesis (49). In the latter study on tamoxifen, verapamil, and CsA MDR1 inhibition, the levels of GlcCer and LacCer, but not GM3 ganglioside, were primarily reduced. This is consistent with our results, as is the anecdotal report of the independence of GM3 synthesis and MDR1 (50).
Although CsA and ketoconazole are inhibitors of MDR1 (51,52), these inhibitors have other effects (53), most notable are the immunosuppressive effects of CsA (54). However, the lack of effect of these inhibitors on neutral GSL biosynthesis in HeLa cells may, in effect, provide a specificity control in which the effects of CsA and ketoconazole are MDR1-mediated. HeLa cells show that alternative mechanisms for GlcCer translocation must exist. Such redundancy may explain the lack of a gross GSL phenotype in knockout mice, although individual tissues, such as skin, from which the fibroblasts studied were derived, may exhibit a defect.
The lack of effect of MDR1 inhibitors on ganglioside biosynthesis provides evidence of a previously unrecognized, differential biosynthetic route for acidic and neutral GSLs. This chal-

TABLE III
Effect of indomethacin on ͓ 14 C͔ serine GSL metabolic labeling in wild type and knockout fibroblasts To confirm that the lack of MRP1 in KOT cells plays no role in the reduced neutral GSL synthesis in these cells, the effect of cell culture with indomethacin, an MRP1-selective inhibitor, on GSL metabolic labeling with ͓ 14 C͔serine was determined in wild type and KOT cells in duplicate. After growth in ͓ 14 C͔serine, lipids were extracted from cells, saponified, separated by two-dimensional TLC, and quantitated by PhosphorImager analysis. Indomethacin was found to have no inhibitory effect on neutral (GlcCer) synthesis.  Table II.  Fig. 4) The metabolically radiolabeled lipids extracted from cells (in duplicate) and separated by two-dimensional TLC (Fig. 4) were quantitated by PhosphorImager analysis. By both ͓ 14 C͔acetate (A) and ͓ 14 C͔serine (B) labeling, the synthesis of neutral GSLs GlcCer and LacCer was severely reduced, whereas GM3 ganglioside synthesis was not inhibited in the KOT cells, as compared to the wild type control. Phospholipid synthesis was relatively unaffected except for increased SM synthesis and a marked decrease in PC labeling from ͓ 14 C͔serine. Neutral lipid labeling was reduced in KOT cells. lenges the paradigm of a single lactosyl (hence glucosyl) ceramide pool as the common precursor for neutral and acidic GSLs. Despite the fact that GlcCer (ϩLacCer, ϩGb 3 if present) levels were depleted to background values by CsA, GM3 synthesis was unaffected within the same cells. Higher gangliosides were not degraded to maintain GM3 levels. In the KOT cells, GlcCer and LacCer synthesis were reduced by 80 -90%, yet the level of GM3 in the same cells was not reduced, compared with wild type cells. Both these data sets infer that GM3 and Gb 3 (neutral GSLs) must be synthesized from separate LacCer (and hence GlcCer) pools. Moreover, because the GlcCer/LacCer pool from which GM3 is synthesized in the presence of CsA must be small, the LacCer substrate K m for the sialyl transferase must be low or the GM3 synthase associated with GlcCer/LacCer synthase to provide a topological product 3 substrate advantage (55). In Jurkat cells, LacCer was undetectable, yet the ganglioside content was significant, confirming that GM3 synthase requires only a low substrate pool. Inhibition of cellular GlcCer synthase using product analogues (56,57) inhibits both neutral and acidic GSL biosynthesis (58 -61) and only a single GlcCer synthase transcript has been detected (62). Thus, the separate regulation of ganglioside and neutral GSL synthesis is not at the level of GlcCer synthesis but must occur via some later sorting process. CsA has no direct effect on the activity of enzymes involved in Gb 3 biosynthesis (11). The LacCer K m for Gb 3 synthase is 50 M (63), whereas that for GM3 synthase is 270 M (64), suggesting that, in the absence of other factors, when LacCer is limiting, Gb 3 should be synthesized in preference to GM3. This is consistent with studies, with fumonisin B inhibition of de novo GSL biosynthesis, that showed the selective enhancement of Gb 3 synthesis from the remodeling of other GSLs (28). Gb 3 synthesis was favored over GM3 at low LacCer levels. The synthesis of LacCer may be uncoupled from GlcCer synthesis in MDR cells (41); i.e. not all GlcCer is available for LacCer synthesis, indicating different GlcCer pools. Gb 3 and GM3 synthases could be located within different Golgi stacks, using different LacCer pools, differentially affected by MDR1 inhibitors. GM3 synthase has been located to the medial Golgi/trans-Golgi network (65 ,66). The location of Gb 3 synthase has yet to be reported, but a cis/medial Golgi location would not be unreasonable. However, our location of MDR1 primarily in the medial and cis-Golgi would argue against gross spatial separation. The different LacCer pools could require different GlcCer translocators. Our HeLa cell results indicate the existence of such species. Distinct precursor GSL pools might arise from differential ER ceramide delivery within the Golgi stacks, as proposed (67). In addition, because inhibition of GlcCer synthesis can result in relocation of LacCer (68), such redistributed LacCer pools may be more effectively available for GM3 than Gb 3 synthesis.
An intriguing possibility is that the synthesis of gangliosides and neutral GSLs from LacCer might be cell cycle-selective. The receptors for cholera toxin (GM1 ganglioside) and verotoxin (Gb 3 ) have been shown to be synthesized at the G 1 and G 2 /M phases, respectively (69). Although LacCer is a common precursor, Gb 3 and GM1 synthesis were not competitive (69).
Although the flippase activity of MDR1 for natural lipids has been questioned, largely on the basis of a lack of a phospholipid phenotype in knockout mice, the present studies with KOT cells demonstrate a selective role for MDR1 in GlcCer translocation for neutral GSL synthesis. Although mrp1 is also lacking in KOT cells, inhibitors of MRP1 do not affect wild type (or KOT) fibroblast (present work) or MDR1-MDCK cell (11) GSL synthesis. The difference in phosphatidyl choline labeling we observe in KOT cells implies a role for MDR1 in endogenous phospholipid translocation, which has been overlooked. The lack of an overt phospholipid phenotype in knockout animals (70) is likely a function of redundancy such as we see for HeLa cells in the present study. The increased labeling of sphingomyelin in KOT cells could result from an increased ceramide content, subsequent to inhibition of neutral GSL synthesis.
The metabolic labeling of the neutral lipid fraction was sig- nificantly reduced in the KOT cells. This fraction should contain, among other species, cholesterol esters. The reduction may therefore relate to the role for MDR1 proposed in cholesterol homeostasis (71)(72)(73) and trafficking (74).
In our earlier report (11), we showed MDR1 activity correlated with increased globoseries GSLs, primarily containing C16, and C18 fatty acids. In the present study, we see that for cells expressing Gb 3 as a doublet by TLC, due to lipid heterogeneity, both bands are susceptible to MDR1 inhibitors, indicating that endogenous MDR1 may flip GlcCer without lipid chain length preference. In the KOT cells, however, the more slowly migrating (shorter fatty acid chains) GlcCer and CTH were more severely reduced, consistent with an MDR1 bias for the more polar GlcCer lipid isoforms.
The involvement of MDR1 in neutral, but not acidic, GSL synthesis is relevant to studies on the regulation of ceramide synthesis by the oug1 gene (75). Transfection of cells with this gene resulted in the selective increase in C18 fatty acid-containing dihydroceramide. This increased ceramide pool was utilized in neutral GSL, but not ganglioside, biosynthesis. This was proposed to result from an unusual fatty acid specificity of the synthetic enzymes (75), but this also requires, as do our studies, that gangliosides and neutral GSL are synthesized from different GlcCer/LacCer pools. It is possible that MDR1 is preferentially responsible for the translocation of C16 and C18 GlcCer, whereas long-chain fatty acid GlcCer is translocated by another mechanism in the later Golgi/endomembrane fractions, in which the cholesterol content and, hence, bilayer thickness, is increased. Gangliosides containing short fatty acid chains may be synthesized via a recycling, as opposed to a de novo pathway (76).
The present studies show that MDR1 is a Golgi component involved in neutral GSL biosynthesis. This shows for the first time, that gangliosides and neutral GSLs can be synthesized via separate pathways that can be differentially regulated. The involvement of MDR1 in Gb 3 synthesis supports the contention (77-81) that VT1 can be an effective antineoplastic.