Accumulation of Glycosphingolipids in Niemann-Pick C Disease Disrupts Endosomal Transport*

Glycosphingolipids are endocytosed and targeted to the Golgi apparatus but are mistargeted to lysosomes in sphingolipid storage disorders. Substrate reduction therapy utilizes imino sugars to inhibit glucosylceramide synthase and potentially abrogate the effects of storage. Niemann-Pick type C (NPC) disease is a disorder of intracellular transport where glycosphingolipids (GSLs) and cholesterol accumulate in endosomal compartments. The mechanisms of altered intracellular trafficking are not known but may involve the mistargeting and disrupted function of proteins associated with GSL membrane microdomains. Membrane microdomains were isolated by Triton X-100 and sucrose density gradient ultracentrifugation. High pressure liquid chromatography and mass spectrometric analysis of NPC1–/– mouse brain revealed large increases in GSL. Sphingosine was also found to be a component of membrane microdomains, and in NPC liver and spleen, large increases in cholesterol and sphingosine were found. GSL and cholesterol levels were increased in mutant NPC1-null Chinese hamster ovary cells as well as U18666A and progesterone induced NPC cell culture models. However, inhibition of GSL synthesis in NPC cells with N-butyldeoxygalactonojirimycin led to marked decreases in GSL but only small decreases in cholesterol levels. Both annexin 2 and 6, membrane-associated proteins that are important in endocytic trafficking, show distorted distributions in NPC cells. Altered BODIPY lactosylceramide targeting, decreased endocytic uptake of a fluid phase marker, and mistargeting of annexin 2 (phenotypes associated with NPC) are reversed by inhibition of GSL synthesis. It is suggested that accumulating GSL is part of a mislocalized membrane microdomain and is responsible for the deficit in endocytic trafficking found in NPC disease.

Increasing evidence suggests that the membranes of eukaryotic cells are not homogeneously fluid. They contain membrane microdomains, often referred to as lipid rafts and/or caveolae, enriched in glycosphingolipids (GSLs), 1 specific proteins, and cholesterol (1)(2)(3). Being operationally defined, i.e. by detergent insolubility and low buoyant density, the natural (in situ) existence of such membrane microdomains has been and continues to be controversial (4). However, they have recently been visualized by using electron microscopy as small structures covering as much as 35% of the cell surface (5). Although membrane microdomains on the cell surface tend to be transient (6), relatively stable microdomains have been visualized in the endocytic pathway (7) where they are proposed to play a role in protein and lipid sorting (8). Many GSL storage diseases, including Niemann-Pick type C (NPC), are often characterized by enlarged liver and spleen as well as severe neurodegeneration. In these diseases, cholesterol and GSLs accumulate in the endocytic pathway because of reduced breakdown. Hence, they may be a useful tool for determining the relative importance of GSLs in endocytic transport. Fibroblasts from patients suffering from GSL storage diseases do show altered endocytic sorting (9), and it has been suggested that the accumulation of endocytic membrane microdomains may be the basis for altered endocytic sorting (10). It is possible that accumulating membrane microdomains in storage bodies mislocalizes microdomain proteins involved in signaling and transport, precluding normal function (10,11). However, the relative importance of GSLs and proteins in membrane microdomain formation is still not known (12,13).
Annexins 2 and 6 are lipid-binding proteins that are widely involved in facilitating intracellular transport, including endocytic trafficking, and are reported to associate with membrane microdomains (14 -23). Proteins involved in both early and late endocytic transport have been shown already to be inhibited in NPC disease. Annexin 2 normally localizes to early endosomes but has been shown to be mislocalized because of lipid accumulation in late endosomes in NPC disease (15). Rab 7, a protein involved in endocytic transport from late endosomes, is also compromised as a result of lipid storage (24,25).
In this study, we have quantified the levels of GSLs, sphingosine, and cholesterol that occur in lipid rafts (the detergent-insoluble fraction) isolated from various organs of a mouse model of Niemann-Pick C (NPC1) disease by HPLC and mass spectrometry. Raft-associated GSLs increased in all tissues. Cholesterol and sphingosine increased in non-neuronal tissues. GSL-lowering therapy, using imino sugar inhibitors of GSL synthesis, is effective in the treatment of a mouse model of NPC disease (26). In NPC cell culture models, inhibition of GSL synthesis ablated the accumulation of GSLs but had little effect on cholesterol levels. However, phenotypes associated with NPC, such as annexin 2 mislocalization, decreased uptake of fluid phase markers (15), and altered BODIPY LacCer trafficking (9) were all reversed by inhibition of GSL synthesis. At least under conditions of intracellular storage, the accumulation of GSLs appears intimately linked to the perturbation of endocytic transport.

EXPERIMENTAL PROCEDURES
HPLC-grade acetonitrile, methanol, chloroform, and butanol were from VWR Scientific (Poole, UK). Acetone was from Fisher. Tissue culture media and supplements were from Invitrogen. o-Phthaldialdehyde and progesterone were from Sigma. Tetrabutylammonium was from Fluka (Zwijndrecht, Netherlands). C:17-sphingosine 1-phosphate was from Avanti Polar Lipids (Alabaster, AL). U18666A was from Affiniti Research Chemicals (Exeter, UK), and NB-DGJ was from Toronto Research Chemicals (Ontario, Canada). Annexin 2 and 6 antibodies were purchased from TD Labs (Oxford, UK), and secondary antibodies were from Molecular Probes (Leiden, Netherlands).
Cell Culture-Bovine aortic endothelial cells, human umbilical vein endothelial cells, and EC-RF24 endothelial cells were grown to confluency in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Cells were dissociated with trypsin/EDTA and were plated at a density of 50,000 cells/cm 2 in an atmosphere of air/CO 2 (19:1) at 37°C in a humidified incubator in 10-cm dishes. CT43 and RA25 Chinese hamster ovary (CHO) cells were a gift of TY Chang, Dartmouth Medical School. The CT43 cell line is a cholesterol trafficking mutant defective in the NPC1 protein. The RA25 cell line is the parental cell line and has a normal NPC1 allele. Both sets of cells contain a gain of function mutation in SCAP, resulting in resistance to sterol-dependent transcriptional regulation. CHO cells seeded in Ham's F-12 Glutamax-I medium, supplemented with 10% fetal calf serum and 100 units/ml of penicillin/streptomycin, were grown as monolayers at 37°C with 5% CO 2 . RAW 264.1 and J774.A1 mouse macrophages were obtained from the ECACC (Porton Down, UK) and maintained at a density of 5 ϫ 10 5 cells/ml in RPMI supplemented with 10 mM glutamine, 50 units/ml penicillin/streptomycin, and 10% fetal calf serum.
Animals-npc1 NIH spontaneous mutant mice on the BALB/cJ background (27) (denominated as NPC1 Ϫ/Ϫ mice in this paper) were maintained by brother-sister mating of heterozygous animals. At weaning (21 days), tail tips were removed for DNA preparation. PCR was used to determine genotypes at the npc1 locus using the following primer pairs: IMR928 and IMR929, expected fragment size, 173 bp; annealing temperature was 55 or 57°C; IMR928 5Ј CTG TAG CTC ATC TGC CAT CG 3Ј and IMR929 5Ј TCT CAC AGC CAC AAG CTT CC 3Ј. IMR927 and MP258F, expected fragment size, 475 bp. Annealing temperature was 53°C; IMR927 5Ј TGA GCC CAA GCA TAA CTT CC 3Ј and IMP258F 5Ј GGT GCT GGA CAG CCA AGT A 3Ј. NPC1 Ϫ/Ϫ mice developed disease symptoms between 6 and 7 weeks of age, and tissues were harvested at 10 weeks of age.
Membrane Microdomain Isolation-Membrane microdomains were purified by the method of Lisanti et al. (28). Briefly, tissues were homogenized by 10 strokes of a Dounce homogenizer in 0.5% (w/v) Triton X-100 in 25 mM Mes-buffered saline (pH 6.5) and incubated for 30 min on ice. The solution made up to 40% sucrose and transferred to a Beckman Ultraclear 14 ϫ 89-mm tube on ice with a discontinuous gradient of 5 ml of 30% sucrose and 3 ml of 5% sucrose layered on top. Samples were ultracentrifuged in an SW41 rotor at 100 000 ϫ g for 16 h at 4°C, and 2-ml fractions were manually collected.
Purification of GSLs for HPLC and Mass Spectrometry-GSLs were extracted by addition of 3.2 volumes of CHCl 3 /MeOH (1:2.2) for 10 min at room temperature followed by the addition of 1 volume of CHCl 3 and 1 volume of H 2 O (29). This rapid extraction procedure gives similar recoveries of GSLs to other commonly used methods of lipid extraction FIG. 1. Separation of brain GSLs from NPC1 ؉/؉ and NPC1 ؊/؊ mice by HPLC. Top trace, cerebral GSLs from normal NPC1 ϩ/ϩ mice. Bottom trace, cerebral GSLs from NPC1 Ϫ/Ϫ mice. Representative traces of GSLs after digestion with ceramide glycanase and derivatizing the released oligosaccharides with anthranilic acid before separation by HPLC. Apart from mass spectrometry of intact GSLs, the identity of some GSLs was confirmed by (a) removal of sialic acid residues by trifluoroacetic acid and (b) mass spectrometry of the same samples both before and after derivatization (not shown).
FIG. 2. Sphingosine and dihydrosphingosine but not phytosphingosine associate with membrane microdomains. Endothelial cells (bovine aortic endothelial cells, human umbilical vein endothelial cells, and EC-RF24) were harvested and resuspended in 0.5% Triton X-100, homogenized, and placed at the bottom of a discontinuous sucrose gradient. After centrifugation, fractions were collected, and sphingoid bases were extracted and derivatized before separation and quantification by HPLC. (30,31). Upper phase GSLs were then recovered by SepPak TM C 18 columns (32). Briefly, C 18 SepPak columns (Waters, Milford, MA) were washed with 1 ml of MeOH and 1 ml of water. The CHCl 3 phase was dried down and loaded in 50 l of CHCl 3 /MeOH (1:3). The corresponding aqueous phase was loaded onto the columns and washed five times with 1 ml of water. GSLs were then eluted five times with 1 ml of CHCl 3 /MeOH (1:3) and 1 ml of MeOH, and the samples were dried under nitrogen. At this point the cholesterol and GSLs were quantified as described below. For mass spectrometry, the samples were saponified by the addition of 1 ml of CHCl 3 and 1 ml of 0.2 M NaOH in MeOH, incubated overnight at 37°C, and further purified by silicic acid chromatography (33). Silicic acid columns were pre-equilibrated in CHCl 3 , and the samples were loaded onto the column and washed with 5 ml of CHCl 3 . Neutral GSLs were eluted with 6 ml of acetone/MeOH (9:1). Gangliosides were eluted with 6 ml of CHCl 3 /MeOH (1:3), and the eluates were dried under nitrogen. The samples were resuspended one time with 50 l and then one time with 100 l of CHCl 3 /MeOH (1:2.2) and transferred to a 1.5-ml tube. The samples were dried down in a SpeedVac and resuspended in 20 l of MeOH. Matrix-assisted laser desorption ionization-mass spectrometry was performed as described previously (34).
Quantitation of Cholesterol-Cholesterol was quantitated by using a method similar to that of Franey and Amador (35). Samples were resuspended in 50 l of EtOH and transferred to a microtiter plate along with cholesterol standards (1 mg/ml in ethanol) in duplicate and dried briefly at 80°C. After the addition of 75 l of 3 mM FeCl 3 ⅐6H 2 0 in glacial acetic acid and 50 l of concentrated H 2 SO 4 , the plate was incubated for 10 min in the dark and the absorbance read at 570 nm.
Extraction and Quantitation of Sphingoid Bases by HPLC-Sphingoid bases were extracted and quantified by the method of Merrill et al. (36). Tissues or membrane microdomain fractions were extracted with butanol containing 1% v/v concentrated HCl and dried under nitrogen. Samples were saponified by the addition of 1 ml of chloroform and 1 ml of 0.2 M NaOH in methanol and incubated overnight at 37°C. Lipids were dried under nitrogen, 0.5 ml of water, 1 ml of 1-butanol; 20 l of concentrated HCl was added, and the upper phase was removed. This procedure was repeated several times, and the combined extract was dried under nitrogen. Sphingoid bases were derivatized with o-phthaldialdehyde in boric acid for 20 min at room temperature, centrifuged at low speed, and 300 l of the supernatant transferred to a Waters autosampler vial. 100 l of the derivatized lipids were injected onto a Waters 5 m C 18 8 ϫ 100 mm column, pore size 100 Å (Waters, Milford, MA). Samples were eluted isocratically with filtered and degassed methanol, 10 mM K 2 H 2 PO 4 (pH 7.4), 0.5 M tetrabutylammonium dihydrogen phosphate (83:16:1, v/v) as the mobile phase. A flow rate of 1.2 ml/min was maintained, and the fluorescence was monitored for 120 min at ex 340 nm and em 455 nm.
Quantitation of GSL Derivatives by HPLC-GSLs were analyzed according to Neville et al. (37). Briefly dried lipid extracts were resuspended in 10 l of incubation buffer (1 mg/ml sodium cholate in 50 mM sodium acetate (pH 5.0)). After vigorous vortexing and spinning in a benchtop picofuge, 10 l of 50 milliunits/10 l of ceramide glycanase (EC 3.2.1.123, Calbiochem) in incubation buffer was added to cleave the glycans. Glucosylceramide is partially digested because of the specificity of the glycanase (37). After 18 h, 10 l of water was added to each enzyme digest followed by 80 l of anthranilic acid (30 mg/ml) and sodium cyanoborohydride (45 mg/ml) to each digest and incubated for 1 h at 80°C. Derivatized oligosaccharides were purified on DPA-6S (Supelco, Ballafonte, PA), and columns were pre-equilibrated two times with 1 ml of CH 3 CN. 1 ml of 97:3 CH 3 CN/H 2 O was added to each sample and vortexed prior to loading. Columns were washed four times with 1 ml of 99:1 CH 3 CN/H 2 O, 0.5 ml of 97:3 CH 3 CN/H 2 O, and the derivatized oligosaccharides were eluted two times with 0.6 ml of water into screwcap Eppendorf and stored at 4°C in the dark until ready for normal phase-HPLC using a Waters Alliance 2695 separations module, an in-line Waters 474 fluorescence detector separation and a 4.6 ϫ 250 mm TSK gel-Amide 80 column (Anachem, Luton, UK).
Immunocytochemistry-J774 and CHO cells were seeded at a density of 5000 cells/coverslip (22 mm) and allowed to adhere overnight. Coverslips were washed twice in PBS prior to fixation with 3.7% paraformaldehyde at room temperature for 10 min. Coverslips were blocked with 1% bovine serum albumin containing 0.1% Tween 20 for 1 h. Monolayers were incubated with 2.5 g/ml primary mouse monoclonal antibodies specific for annexin 2 or 6 in blocking solution for 45 min, followed by three washes in PBS (5 min each wash), and incubation with 5 g/ml Alexa-Fluor 488 (annexin 2) or 564 (annexin 6)-conjugated rabbit anti-mouse secondary antibody for 30 min. Samples were washed three times in PBS (5 min each wash) followed by mounting in Vecta-shield. Fluorescent cells were observed using a Zeiss Axioplan 2 fluorescence microscope. Images were collected using a charge-coupled device camera.
Cholesterol Staining-Monolayers on coverslips were washed three times with PBS and fixed in 3.7% paraformaldehyde for 15 min followed by quenching in medium. Cells were incubated with 125 g/ml filipin (a cytochemical probe specific for cholesterol) for 30 min at 37°C followed by three washes with PBS and mounting in Vectashield.
Fluid Phase Endocytosis-Cellular uptake of HRP was performed as described (15,38). RAW, J774, RA25, and CT43 cells were treated with 2 g/ml U18666A or 10 g/ml progesterone for 18 h in the absence and presence of 100 M NB-DGJ. Cells were then incubated with 3 mg/ml HRP for 2 h at 37°C followed by extensive washing at 4°C, and HRP was assayed using a commercial peroxidase kit (Pierce).

Structural Confirmation and Quantitation of GSLs by HPLC-
The changes in total GSL levels between NPC1 ϩ/ϩ and FIG. 3. A, GSL-lowering treatment reverses altered BODIPY LacCer targeting in an NPC cell culture model. NPC1-containing RA25 cells and NPC1-null CT43 cells were treated with NB-DGJ for 5 days. B, altered GSL trafficking in cell culture models of NPC disease. RAW macrophages or NPC1-containing RA25 cells were treated with 2 g/ml U18666A (U) or 20 g/ml progesterone (Prog) for 18 h. 50% confluent monolayers were grown on coverslips and incubated in the presence of 13 M BODIPY LacCer and pulse-chased as described under "Experimental Procedures." Scale bar indicates 5 m.

TABLE I
Major detergent-insoluble GSLs in membrane microdomain fractions of NPC1 Ϫ/Ϫ brain, liver, and spleen Cerebella, cerebra, liver, and spleen were dissected from normal and NPC1 Ϫ/Ϫ mice and homogenized. The equivalent of 5 mg of tissue was incubated with Triton X-100 before separation of the floating detergent-insoluble and -soluble portions by ultracentrifugation. GSLs from both portions were extracted and purified before digestion with ceramide glycanase before derivatization and quantitation by HPLC. Data are from 3 to 9 experiments and are expressed as pmol/g protein Ϯ S.E. Data in parentheses indicate the concentration in the homogenate (pmol/g protein S.E. Ͻ20%). Total raft protein varied from 210 GM1a was found in the brain but in spleen GM1b predominated.

NPC1
Ϫ/Ϫ mice were initially analyzed by matrix-assisted laser desorption ionization-mass spectrometry (34). No large differences in fatty acid composition could be detected between normal and NPC1 Ϫ/Ϫ brain, liver, and spleen tissues (not shown). GSLs were extracted and digested by ceramide glycanase (37) followed by derivatization, separation, and quantitation of the resultant oligosaccharides by HPLC (Fig. 1). For instance, normal mouse brain GSLs were found to be similar to those reported previously with GM1a, GD1a, GD1b, GT1b, and GQ1b predominating and lower levels of LacCer, GA2, GM3, and GM2 ( Fig. 1) (39). However, in brains from NPC1 Ϫ/Ϫ mice, GSLs increased, including GM2, GA2, GM3, and LacCer in agreement with studies employing thin layer chromatography (40,41). Glycosphingolipids in NPC1 Ϫ/Ϫ Mice Accumulate in Membrane Microdomains-Because GSLs are thought to be important components of membrane microdomains, we investigated whether they accumulated in microdomains in the NPC mouse tissues. All the GSLs were found to be highly enriched in membrane microdomains (Table I). Membrane microdomains from the cerebrum, cerebellum, liver, and spleen were purified by the Triton X-100 detergent procedure using sucrose gradients. In the brains of NPC1 Ϫ/Ϫ mice GSLs increased in agreement with the mass spectrometric and HPLC analysis of the total levels (Fig. 1). The fold increases were broadly similar to that found in total GSL levels in human NPC patients (42). Similar increases were seen in cerebellum (Table I). Table I also shows the levels of membrane microdomain GSLs in NPC1 Ϫ/Ϫ liver and spleen. In liver, the N-glycolyl derivative of GM2 predominated with smaller quantities of neutral lipids. In membrane microdomains isolated from NPC1 Ϫ/Ϫ liver, large increases in GcGM3, LacCer, and GA2 were detected. In NPC1 Ϫ/Ϫ spleen, levels of membrane microdomain LacCer, GA2, GA1, GM3, Gb3, GalNAcGA1, and GM1b increased.
Cholesterol in NPC1 Ϫ/Ϫ Mice Accumulates in Membrane Microdomains-During the last 20 years NPC has been viewed as a cholesterol storage disease. We determined whether the elevated cholesterol that occurred in NPC was enriched in membrane microdomains. Up to 90% of the cholesterol was found in membrane microdomains (the detergent-insoluble low density fraction) in both normal and NPC1 Ϫ/Ϫ tissues (Table II). Table  II shows that membrane microdomain cholesterol levels varied in control mice between different tissues, and large increases (4 -10-fold) in cholesterol could be detected in NPC1 Ϫ/Ϫ liver and spleen, similar to that found in human patients (43). Only a minor increase in cholesterol can be detected in brain.
Sphingosine Preferentially Resides in Membrane Microdomains-Although the occurrence of sphingoid bases in membrane microdomains has not been reported previously, several of the enzymes that were involved in their production, such as acid sphingomyelinase and ceramidase as well as their precursors sphingomyelin and ceramide, have been localized to lipid rafts/caveolae (44 -46). We initially looked in a variety of known caveolae-rich endothelial cell lines to determine whether sphingoid bases were located in membrane microdomains. An HPLC technique was used to separate different sphingoid bases following derivatization with o-phthaldialdehyde (36). Clear separation of sphingosine from sphingosine 1-phosphate, dihydrosphingosine, and phytosphingosine could be obtained. Other unidentified peaks are possibly either lysoether and/or lyso-plasmalogen forms of phosphatidylserine and/or phosphatidylethanolamine and/or lyso forms of complex glycolipids (data not shown). Fig. 2 shows the distribution of sphingosine, dihydrosphingosine (sphinganine), and phytosphingosine over the density gradient fractions from three different endothelial cell lines. Sphingosine and dihydrosphingosine were enriched in the floating membrane microdomain fraction. In contrast, phytosphingosine, which has an extra hydroxyl group, is not en-

TABLE II
Membrane microdomain GSL, cholesterol, and sphingosine concentrations in normal (NPC1 ϩ/ϩ ) and NPC tissues (NPC1 Ϫ/Ϫ ) Brain, liver, and spleen were dissected from normal and NPC mice and homogenized. The equivalent of 5 mg of tissue was incubated with Triton X-100 before separation of the floating detergent-insoluble and -soluble portions by ultracentrifugation. GSLs from both portions were extracted and purified before digestion with ceramide glycanase and before derivatization and quantitation by HPLC. Cholesterol was purified and quantitated by using a cholesterol microassay. Sphingosine from both portions was extracted with acidified butanol before saponification and derivatization and separation and quantitation by HPLC. Data are from 3 to 8 experiments and are expressed as pmol or nmol/g protein Ϯ S.E.
riched in the membrane microdomain fraction and therefore serves as a control to show the specificity of the localization of sphingosine and sphinganine to rafts/caveolae. Membrane Microdomain Association of Sphingosine in NPC1 Ϫ/Ϫ Mice-Because sphingosine was localized to rafts/ caveolae in endothelial cell lines, we determined whether it localized to membrane microdomains in NPC disease; large increases in sphingosine have been reported in the liver and spleen of NPC patients as well as mouse models of NPC disease (47,48). Table II shows large increases in microdomain sphingosine in liver and spleen of NPC1 Ϫ/Ϫ mice. Modest changes were detected in brain microdomains, similar to whole tissue (48). Low levels of microdomain dihydrosphingosine also accumulated in NPC1 Ϫ/Ϫ tissues (not shown); sphingosine 1-phosphate levels were too small to be detected reliably. Overall, the results clearly show that a significant proportion of the sphingosine, cholesterol, and GSL that accumulates in NPC disease was located in membrane microdomains.
NPC Cell Cultures Show Increases in Both GSL and Cholesterol-Although increases in cholesterol in NPC cell culture models are well documented (49 -51), the study of GSL is not as extensive. We determined whether increases in GSL occur in NPC cell culture models, and we compared this to the increases in cholesterol (Table III). By using U18666A and progesterone treatment to induce an NPC phenotype (52,53), total GSL levels increased (Table III). LacCer, GA2, GM3, GM2, GD1a, and GM1a were detected in RAW macrophages, and all increased to similar degrees. Increases in GSLs (LacCer, GM3, GcGM3, and GM1b) were observed when NPC1-containing RA25 cells were compared with NPC1 null CT43 cells (Table  III). Cholesterol levels were larger but showed similar fold increases to GSL levels. However, administration of NB-DGJ decreased GSL levels without significantly affecting cholesterol levels (Table III). To decrease cholesterol levels further sphingomyelin may also have to be depleted.

Altered Endocytic Trafficking in an NPC Cell Culture Model Is Corrected by GSL-lowering Drugs-When normal RA25
CHO cells were pulse-labeled with BODIPY-LacCer, the fluorescent lipid was targeted primarily to the Golgi complex ( Fig.  3A) (9). In NPC1-null CT43 cells, fluorescence accumulates in punctate endocytic structures, reminiscent of previous results in NPC patient fibroblasts (24). Because GSL-lowering therapy was effective in a mouse model of NPC disease (26), we investigated whether GSL depletion can also reverse altered endocytic trafficking. Administration of NB-DGJ, an inhibitor of GSL synthesis, for 5 days leads to a reversal of altered trafficking and perinuclear localization is restored (Fig. 3A); shorter incubations were not as effective (not shown). These results suggest that NPC is similar to other GSL storage disorders; altered BODIPY LacCer sorting in a Gaucher cell culture model is also reversed by GSL-lowering drugs (54).
Altered GSL Trafficking Occurs in U18666A and Progesterone-induced NPC Cell Culture Models-Because clear increases in GSL levels were seen in U18666A and progesteronetreated cells, we determined whether GSL trafficking was also affected. Fig. 3B shows that BODIPY LacCer is targeted to a perinuclear region in RAW macrophages (54). Treatment of both the RAW macrophages or NPC1 containing RA25 cells with U18666A or progesterone alters BODIPY LacCer trafficking to punctate endocytic structures.
Decreased Fluid-phase Uptake Is Reversed by the Inhibition of GSL Synthesis in NPC Cell Culture Models-Annexins are of particular interest because they are thought to be important in endocytic trafficking and are associated with membrane microdomains (16,21,55,56). NPC cell culture models show altered annexin 2 localization and a decrease in the intracellular accumulation of the fluid phase marker HRP (15). The intracellular accumulation of cholesterol is involved in this process (15), so do GSLs also play a role? Similar to previous results in U18666Atreated baby hamster kidney cells (15), when RAW or J774 macrophages were treated with U18666A the amount of intra-

TABLE III
Effect of a GSL-lowering drug on an NPC cell culture model RAW cells were treated with U18666A (2 g/ml) for 24 h, or NPC1containing RA25 cells were compared to NPC1-null CT43 cells. RAW were either simultaneously treated with NB-DGJ (100 M) for 24 h, or RA25 and CT43 cells were treated with NB-DGJ (50 M) for 5 days. GSLs and cholesterol were extracted and measured as described under "Experimental Procedures" and expressed as pmol/g protein Ϯ S.E. (n ϭ 3-4).
cellular accumulation of HRP was found to be reduced (Fig. 4).
Simultaneous treatment with NB-DGJ led to a reversal in HRP uptake (Fig. 4). These results suggest that the level of GSL can influence this process. GSL depletion with NB-DGJ also stimulated the uptake of HRP by about 100% in both macrophage cell lines (not shown) suggesting that GSLs may play a role in endocytic trafficking in some cell types, as has been reported previously (54). Comparison of RA25 cells with NPC1-null CT43 cells showed a large decrease in the amount of HRP uptake (Fig. 5), similar to what was reported in U18666A-treated cells. In contrast to macrophages, treatment of NPC1 containing RA25 cells with NB-DGJ did not significantly affect the level of HRP uptake (Fig. 5). When NPC1-null CT43 cells were treated with NB-DGJ, a time-dependent increase in the levels of HRP uptake was observed; complete reversal took 10 days (Fig. 5). We next investigated whether the changes in fluid phase uptake correlated with annexin 2 distribution. In untreated RA25 cells annexin 2 was found to be distributed throughout the cytoplasm with a punctate peripheral distribution (Fig. 6), consistent with previous studies showing that annexin 2 partly localizes to early endosomes (57). No large changes were seen following administration with NB-DGJ in RA25 cells. A prominent decrease in peripheral staining was observed in CT43 cells (Fig. 6) consistent with redistribution to late endosomes, similar to other NPC cell culture models (15). Treatment of CT43 cells with NB-DGJ for 5 days caused a marked reversal in the altered annexin 2 distribution (Fig. 6). These results suggest that GSL levels and annexin 2 distribution are interlinked, at least under conditions of GSL storage. It is also possible that inhibition of GSL synthesis leads to increases in ceramide levels. However, addition of ceramide analogues to NPC cell culture models does not restore the altered BODIPY LacCer trafficking (not shown). One further possibility was that inhibition of GSL synthesis decreases the levels of downstream catabolites such as sphingosine. However, addition of a ceramidase inhibitor (58) also fails to restore altered BODIPY LacCer targeting in NPC cells (not shown).

Annexin 6 Shows an Altered Localization in NPC Cells-
NPC cells show changes in late endosomal transport (59,60) as well as changes in early endosomal transport. We looked at annexin 6 because it has been implicated in late endosomal transport and cholesterol trafficking (22,23). In RA25 CHO cells, we localized cholesterol by using filipin to the perinuclear region (Fig. 7A). However, in NPC1-null CT43 cells and in RA25 cells treated with U18666A (22), a more punctate cholesterol-filipin distribution is seen throughout the cell (Fig. 7A). When a similar experiment was performed by using an antibody to annexin 6, RA25 cells show a perinuclear distribution (Fig. 7B), whereas CT43 (NPC1 Ϫ/Ϫ ) and U18666A-treated cells show a punctate distribution throughout the cell. Similar changes have been shown for annexin 6 overexpressing CHO cells treated with U18666A (22). Hence, annexin 6 appears to be relocalized to punctate endocytic structures in NPC cells in a similar fashion to that previously observed for annexin 2 (15).

DISCUSSION
In this study, we found that NPC disease is associated with increased levels of GSLs, cholesterol, and sphingosine in membrane microdomains. Although our study uses detergent insolubility as an operational definition of membrane microdomains, a good correlation between detergent resistance and the mobility of lipid probes visualized by single molecule microscopy has been reported (6). A previous study (51) reported evidence that the level of membrane microdomains is increased in an NPC cell culture model; however, relatively small changes in the detergent insolubility of cholesterol were detected. Given that the present data disrupted endocytic trafficking may only engage a small percentage of the total cholesterol pool. This is consistent with the yeast NPC model where altered localization of sphingolipids occurred without a large increase in ergosterol levels (61).
Our data suggest that the accumulation of stored lipids is linked to mislocalization of membrane microdomain-associated proteins such as annexin 2 and 6 (16 -18, 20, 21, 56). Recently, using BODIPY LacCer, GSL membrane microdomains have been visualized in the early endocytic pathway, even in normal cells (7). Both early and recycling endosomes as well as the trans-Golgi network are enriched in membrane lipids that are expected to form membrane microdomains (62)(63)(64). In contrast, late endosomal and lysosomal membranes are normally unlikely to accumulate microdomains because of the activity of sphingolipid hydrolases (10). In NPC cells, however, the situation may be different because the GSL and sphingosine microdomain content is very high (65). Although the precise localization of these membrane microdomains is a subject for future study, GSLs have been shown to accumulate in endocytic compartments (51,66).
Apart from the accumulation of microdomains, the mislocalization of GSL may have specific effects on cell physiology. Particular GSLs have also been found to be important in the transport of toxins and viruses (67,68), and it has been suggested that GSLs are employed for the trafficking of endogenous proteins, perhaps with specific GSL-binding domains (69,70). Intracellular transport from the plasma membrane to the endoplasmic reticulum by these specific GSL-dependent pathways is also reliant on their ability to form membrane microdomains (71,72). GSL-dependent trafficking may be particularly sensitive to disruption in NPC and other GSL storage diseases (68). It is still unclear why particular GSLs predominate in NPC, as possibly specific GSL recycling pathways are affected (73). Calcium mobilization is also necessary for endosomal fusion (74), and defective calcium transport may form the basis for related storage disorders such as mucolipidosis type IV (75,76). Calcium transport has been shown to be influenced by GSLs and their lyso derivatives (77) and is defective in a mouse model of Sandhoff disease, a phenotype corrected by GSL depletion (78). It should also be noted that GSLs have also been implicated in signaling (79,80), and their accumulation may well disrupt the function of specific signaling processes in NPC either directly or by altering plasma membrane phospholipid saturation (65,81,82). The accumulation of GSLs may disrupt cell physiology by more than one mechanism, in a cell typespecific fashion. It has been reported that NPC can be protective against viral infection (83). However, given the severe phenotypes that NPC and most other sphingolipid storage diseases present in the homozygote, identification of heterozygote phenotypes will be necessary to envisage a protective role against infection in particular ethnic populations (84). A heterozygote phenotype has already been described in NPC disease (85). In this study NPC1 ϩ/Ϫ mice showed less macrophage apoptosis in a mouse model of atherosclerosis, suggesting that the immune system may be particularly affected in the heterozygote, as neuronal function is overtly normal.
The identification of sphingosine in microdomains is a novel finding. It is unknown why sphingosine accumulates in NPC mice, and only small increases have been detected in NPC cell culture models (48). Sphingosine, fatty acids, and cholesterol are products of membrane microdomain degradation. How these lipids are transported from the late endosome to the endoplasmic reticulum for reutilization or esterification is not known, but NPC2, NPC1, MLN64, and late endosome/endoplasmic reticulum contact sites may be involved (86 -90).
It is unclear why phytosphingosine was not found to be enriched in the floating microdomain fraction; however, it may be a contaminant from bovine serum (91). Other experiments suggest that exogenously added sphingosine is also not enriched in the floating microdomain fraction. 2 The principal advantage of the HPLC technique employed in this study for measuring GSL levels is that it is very sensitive and allows clearer separation and identification of different GSL species when compared with thin layer chromatography. Unfortunately, glucosylceramide and sphingomyelin cannot be accurately quantified because of the specificity of ceramide glycanase. Increases in glucosylceramide and sphingomyelin have been reported by using other techniques in several tissues isolated from NPC1 Ϫ/Ϫ mice (41,92). It is suggested that the intracellular accumulation of GSL is causally linked to changes in endocytosis in NPC disease, through changes in annexin distribution. How this relates to the disease pathology has not been addressed in this study, and the success in treating NPC1 Ϫ/Ϫ mice with imino sugar inhibitors of GSL synthesis (26) suggests GSLs are important. The current observation of large increases in microdomain components in NPC disease, and presumably other GSL storage diseases, indicates that these diseases may be good models for the role of membrane microdomains in intracellular trafficking.