Sphingolipid Depletion Increases Formation of the Scrapie Prion Protein in Neuroblastoma Cells Infected with Prions*

Sphingolipid-rich rafts play an essential role in the posttranslational (Borchelt, D. R., Scott, M., Taraboulos, A., Stahl, N., and Prusiner, S. B. (1990) J. Cell Biol. 110, 743–752)) formation of the scrapie prion protein PrPSc from its normal conformer PrPC(Taraboulos, A., Scott, M., Semenov, A., Avrahami, D., Laszlo, L., Prusiner, S. B., and Avraham, D. (1995) J. Cell Biol. 129, 121–132). We investigated the importance of sphingolipids in the metabolism of the PrP isoforms in scrapie-infected ScN2a cells. The ceramide synthase inhibitor fumonisin B1(FB1) reduced both sphingomyelin (SM) and ganglioside GM1 in cells by up to 50%, whereas PrPSc increased by 3–4-fold. Whereas FB1 profoundly altered the cell lipid composition, the raft residents PrPC, PrPSc, caveolin 1, and GM1 remained insoluble in Triton X-100. Metabolic radiolabeling demonstrated that PrPC production was either unchanged or slightly reduced in FB1-treated cells, whereas PrPSc formation was augmented by 3–4-fold. To identify the sphingolipid species the decrease of which correlates with increased PrPSc, we used two other reagents. When cells were incubated with sphingomyelinase for 3 days, SM levels decreased, GM1 was unaltered, and PrPSc increased by 3–4-fold. In contrast, the glycosphingolipid inhibitor PDMP reduced PrPSc while increasing SM. Thus, PrPSc seems to correlate inversely with SM levels. The effects of SM depletion contrasted with those previously obtained with the cholesterol inhibitor lovastatin, which reduced PrPSc and removed it from detergent-insoluble complexes.

The transmissible spongiform encephalopathies such as scrapie of sheep and Creutzfeldt-Jakob disease of humans are caused by prions (Ref. 1 and reviewed in Ref. 2). Mouse neuroblastoma N2a cells infected with prions produce two distinct isoforms of the prion protein: the normal prion protein PrP C , and its pathological isoform PrP Sc (3)(4)(5). PrP Sc , which is the only known component of the infectious prion (6,7), is formed posttranslationally in the host cell (3,8), perhaps by the refolding of normal PrP C into a ␤-sheet-rich abnormal conformation (9 -14). Although the PrP isoforms appear to be chemically identical (15), their properties are very different: PrP C is readily soluble in most detergents and is completely degraded by proteinase K, whereas PrP Sc is insoluble in detergents and possesses a protease-resistant core termed PrP27-30 (16,17). In suitable experimental conditions, PrP Sc can polymerize into amyloidic structures, the prion rods (18,19). The relationship between rods and scrapie-associated fibrils (20) has not been absolutely established.
The molecular and cellular details of PrP Sc formation in the host cell are not well understood. In particular, the subcellular sites where PrP Sc is formed remain obscure. PrP C is found on the surface of neurons and other cells (21), and is expressed by many cell lines. Both PrP isoforms possess a glycosylphosphatidylinositol (GPI) 1 moiety (22). PrP C is anchored to cellular membranes through its GPI tail and can be removed from the cell surface by the phosphatidylinositol-specific phospholipase C (22). In contrast, the membrane topology of PrP Sc remains unknown (23,24). In prion-infected cells in culture, PrP Sc accumulates mainly, but not exclusively, in secondary lysosomes (25). Previous studies suggested that PrP Sc may acquire its protease-resistant core either on the cell surface or in the endocytic pathway (26 -28). Like many GPI proteins, both PrP isoforms seem to reside within Triton X-100-insoluble, cholesterol-and sphingolipid-rich "rafts" (29 -34), and rafts appear to play a central role in the metabolism of PrP.
The concept of membrane rafts (Ref. 35 and reviewed in Ref. 36) attracted renewed interest when it was found that GPI proteins owe their insolubility in a variety of detergents to their association with membrane areas that are enriched in cholesterol and sphingolipids (34). Brown and Rose (34) introduced the use of flotation gradients to analyze these assemblies and showed that cold Triton X-100 (TX-100)-insoluble rafts buoy toward low density fractions in these gradients. Raft "residents," such as GPI proteins, gangliosides, and caveolin, also migrate to the lighter fractions of these gradients. The relationship between rafts and the structurally defined caveolae has not been entirely elucidated. Because many caveolar components are also insoluble in cold TX-100, and because caveolae are also enriched in cholesterol and sphingolipids (SLs), it seems reasonable to postulate that caveolae are related in some way to membranal rafts. Although rafts and caveolae were first observed on the plasma membrane and the trans-Golgi network (37), it is possible that membranal structures of similar properties exist in the endocytic compartments as well. Indeed, the abnormal PrP isoform PrP Sc , which accumulates mainly in secondary lysosomes, is attached to raft-like domains (32,33). We have previously shown that, in scrapie-infected ScN2a cells, most PrP Sc is attached to "heavy" rafts that can be differentiated from PrP C -containing "light" rafts in Nycodenz density gradients, particularly when TX-100 cell lysates are incubated at 37°C prior to the flotation procedure (33).
Many studies have shown that cholesterol and sphingolipids intervene in various functions assigned to caveolae and rafts, as well as in the physical properties of these domains. In MA104 cells, depleting cellular cholesterol impaired the endocytosis of folate (38), a caveolae-related process called potocytosis (reviewed in Ref. 39). Depletion of cholesterol also prevented the clustering of the GPI-linked folate receptor on the surface of MA104 cells (40) and decreased the association of GPI proteins with detergent-insoluble fractions (41). Comparable results were obtained when cells were incubated with agents that complex with cholesterol. Treating epithelial cells with filipin revealed a caveolae-mediated transcytotic transport pathway (42). Preventing synthesis of sphingolipids using fumonisin B 1 (FB 1 ) also disrupted folate endocytosis in MA104 cells (43). The importance of sphingolipids was also demonstrated for the transport of newly synthesized GPI-anchored proteins to the Golgi in yeast (44), for their sorting to the appropriate membrane surface in polarized epithelial cells (Ref. 45 and reviewed in Ref. 46), as well as for their segregation into TX-100-insoluble domains in Chinese hamster ovary cells (47).
Several lines of evidence suggest that the metabolism of the PrP isoforms in general and the formation of PrP Sc in particular are regulated by rafts: (i) both the substrate (PrP C ) of the putative reaction PrP C ϩ PrP Sc 3 2PrP Sc and its product (PrP Sc ) are found within rafts (32,33); (ii) depleting cellular cholesterol rendered most PrP C soluble in cold TX-100, markedly decreased PrP Sc synthesis, and slowed the rate of PrP C degradation (29); and (iii) removing PrP from rafts by replacing its GPI moiety with any of several transmembrane sequences (29,48) completely prevented the formation of protease-resistant PrP in ScN2a cells. How rafts support PrP Sc formation (and thus prion replication) remains uncertain. We have proposed (33) that rafts may function by locally crowding the prion "seed" (PrP Sc ) and the "substrate" (PrP C ) on membranes, thereby promoting their interaction and facilitating the formation of PrP Sc .
In this study, we have investigated the importance of SLs for the metabolism of the PrP isoforms and the formation of PrP Sc . Our findings suggest that SM and glycosphingolipids play disparate roles in the regulation of PrP Sc formation by rafts.

EXPERIMENTAL PROCEDURES
Materials-Cell culture reagents were purchased from Biological Industries (Bet Haemek, Israel). Tissue culture plates were obtained from Miniplast (Ein Shemer, Israel) or Nunc (Roskilde, Denmark). G418 was purchased from Life Technologies, Inc. Cholera toxin subunit B (CTXB) (C9903) and n-octyl ␤-D-glucopyranoside (494459) were from Calbiochem (San Diego, CA). Fumonisin B 1 was from either Sigma (F1147), Calbiochem (San Diego, CA) (344850), or Alexis Biochemicals (350 -017). Sphingomyelinase (nSMase) from Staphylococcus aureus (S7651), bovine brain sphingomyelin (S7004), and GM1 (G7641) were all from Sigma. Lysamine rhodamine C12-ceramide and glucosyl-ceramide, which were used as markers, were a kind gift from Prof. Shimon Gat (Jerusalem, Israel). D-Threo-1-phenyl-2-decanoylamino-3-morpholino-1-propanol (PDMP) was from Matreya Inc. (Pleasant Gap, PA). Other chemicals were from Sigma, unless otherwise stated. Secondary antibodies were from Jackson ImmunoResearch (West Grove, PA Cell Cultures-Mouse neuroblastoma N2a cells and the persistently scrapie-infected subclone ScN2a were obtained as described (4,33). In our hands, cultures of the persistently ScN2a cells often undergo crises during which their PrP Sc production is reduced to undetectable levels. On these occasions, we either thaw fresh cells or clone the ScN2a population to isolate sublines that are better producers of PrP Sc . N2a-C10 and ScN2a-C10 cells (33) stably express the MHM2-PrP chimeric gene (50) that carries the 3F4 epitope. N2a-c and ScN2a-c express human caveolin 1/VIP21 from an expression vector kindly supplied by Dr. Richard Anderson (Dallas, TX). Expression of caveolin does not seem to alter the metabolism of the PrP isoforms in these cells, 2 but the PrP Sc production by these clones appears to be more stable than that of the original ScN2a cells. ScN2a sublines and subclones somewhat vary in their sensitivity to various pharmacological agents, as well as in the detergent insolubility of their rafts (33). Cells were usually grown at 37°C in DMEM16 supplemented with 8% fetal calf serum. In some experiments, we used delipidated fetal calf serum, which was prepared as described (51).
Antibodies-Rabbit antiserum R073 binds to both mouse PrP and to MHM2-PrP (5,52). mAb 3F4 (53) binds to residues Met 109 and Met 112 (54) in chimeric MHM2-PrP but does not recognize the wild type mouse PrP endogenous to N2a cells (50). Both antibodies were used at a dilution of 1:5000 (of the serum or the ascitic fluid, respectively). Rabbit antiserum against caveolin 1 was purchased from Transduction Laboratories (Lexington, KY). Anti-PKC and anti-cholera toxin subunit B was purchased from Sigma.
PrP Isoforms and PrP Analysis-The PrP isoforms were characterized and separated as described (18). PrP Sc was defined as the PrP fraction resistant to proteolysis by proteinase K (20 g/ml, 37°C, 50 min). Fractions from flotation gradients were made 1% with n-octyl ␤-D-glucopyranoside to solubilize rafts prior to incubation with proteinase K (31,34)). SDS-PAGE, Western immunoblotting, metabolic labeling of cells with [ 35 S]methionine, and immunoprecipitation of the PrP isoforms were all carried out as described (29,33).
Manipulation of Cellular Sphingolipids-In a typical experiment, cells were seeded on day 0 in two 10-cm dishes (approximately 5 ϫ 10 5 cells, for the flotation assay) or in one 6-cm dish (approximately 10 5 cells, for the total lysate and lipid analysis), and cultured for 24 h at 37°C. On day 1, 30 M FB 1 (from a 13.8 mM stock in Me 2 SO) or 70 M PDMP were added to the medium. A 5 mM PDMP stock was prepared as follows: 10 mg were dissolved in 100 l of ethanol, added to 4.6 ml of DMEM, sonicated twice for 3 min, and then dialyzed against HEPES buffer (50 mM HEPES, 150 mM NaCl, 1 mM MgCl 2 , and 0.1 mM CaCl 2 , pH 7.4) for 24 h at 4°C. On day 4, cells were treated with fresh medium supplemented with FB 1 or PDMP and were further incubated to day 7. Treatment with exogenous nSMase was carried out in medium containing delipidated medium. Due to its cytotoxicity, treatment with nSMase was usually limited to 3 days, and therefore the enzyme was added on day 4.
Metabolic Radiolabeling and Analysis of Lipids-Cells (about 10 7 cells for each treatment) were radiolabeled with either [methyl- 3

cells).
Lipids were extracted either from cell suspensions or from low density fractions of flotation gradients. To extract total lipids from cell suspension, radiolabeled cells were washed three times with cold PBS, 1 ml of PBS was added to the plate, and cells were removed by scraping 2 N. Naslavsky and A. Taraboulos, unpublished results. and pelleted (3000 rpm, 5 min in an Eppendorf tabletop centrifuge). Pellets were resuspended in PBS, 20 l were taken for protein quantification, and samples were adjusted to equivalent amounts of protein in equal volume. Total lipids extraction from these cell suspensions or from gradient fractions was carried out with CHCl 3 :CH 3 OH:H 2 O (1:1:1, v/v/v) (55). The lower phase was re-extracted with 3.8 vol. of CH 3 OH: "acid saline" (0.9% NaCl in 10 mM HCl) 2:1.8, v/v. In some experiments involving choline labeling, ester phospholipids were digested by mild alkaline hydrolysis with 0.4 N KOH in CH 3 OH:H 2 O (9:1, v/v) for 4 h at 60°C, prior to further analysis.
Lipid extracts were resolved by TLC on either glass or plastic sheets Silica Gel 60 (Merck, Darmstadt, Germany). The chromatographs were developed in CHCl 3 :CH 3 OH:15 mM CaCl 2 , (60:35:8) (49). The TLC plates were first stained with Coomassie Blue, and the intensity of the staining was quantified by densitometry (see below). The plates were then exposed to a tritium-sensitive imaging plate. Thus, for each lipid species, newly synthesized radiolabeled lipids could be compared with the total mass. The doublet at the bottom of the TLC (Fig. 1B) was assigned to SM because (i) it comigrates with SM markers, (ii) it is positive for organic phosphorous and ninhydrin-negative, (iii) it resists alkaline hydrolysis, and (iv) it labels with both radiolabeled choline and sphinganine.
Coomassie Staining of Lipids-TLC plates were stained with Coomassie Blue as described (47,56) to identify and quantify lipids and to visualize lipid standards. In our hands, only plastic TLC sheets (Merck 1.05748) resisted the procedure, whereas glass plates lost their silica coating. After lipid separation, TLC sheets were air-dried, soaked in 30% MeOH containing 0.03% Coomassie Blue, gently agitated for 3-4 min, and then destained with 30% MeOH for 5 min. Densitometry was then used to quantify the total lipid mass in the Coomassie-stained bands. Because lipids differ in their affinity for the dye, we used this method only to compare identical bands of the same lipid in various experiments. In preliminary calibration experiments using commercial lipid standards, we found that Coomassie staining was linear in the range of 0.05-30 g of lipid/band for SM and PC (data not shown). In our hands, Glc-Cer was not stained by this method.
Detection of GM1 Using Cholera Toxin Subunit B-We detected and quantified cell surface GM1 using its ligand CTXB. Cells were washed with PBS, cooled on ice, and incubated with 5 g/ml CTXB (from a 50 g/ml stock solution in PBS) on ice for 30 min. After extensive rinsing with ice-cold PBS to remove unbound CTXB, cells were lysed, and lysate samples, normalized for protein content, were subjected to a nondenaturing SDS-PAGE analysis (boiling of samples was omitted to prevent dissociation of the pentameric structure of CTXB, which is recognized by the antibody).

Fumonisin B 1 Increases PrP Sc but Not PrP C in ScN2a Cells-
Because sphingolipids are important structural (34) and functional (43) components of rafts, we sought to evaluate their contribution to the formation of PrP Sc (29,32,33). We first used the mycotoxin FB 1 to deplete cellular sphingolipids in ScN2a cells. FB 1 inhibits sphingosine N-acyltransferase (ceramide synthase) in a concentration-dependent manner and thereby interferes with the biosynthesis of all sphingolipids beyond the sphingoid base, in a variety of cells (57). Because PrP Sc is very stable in ScN2a cells (3), several days of treatment are necessary to detect changes in its steady state levels by Western immunoblotting. Scrapie-infected ScN2a cells were grown for seven days in the presence of 0, 5, 20, 35, or 50 M FB 1 (Fig. 1). FB 1 treatment (50 M) had only minor effects on the total cellular protein level (Ͻ15% for the maximal FB 1 concentration, as determined by the BCA assay), and we observed no cytopathic effects by microscopic examination. On day 7, the cells were scraped off the dishes in PBS and then divided into two equal aliquots, which were used for analysis of PrP Sc levels and of lipids, respectively. To analyze the effect of FB 1 on SM levels, [ 3 H]choline (2 Ci/ml) was added to the cell medium for the last 24 h of the FB 1 treatment to help assess the influence of the inhibitor on choline incorporation into SM and phosphatidylcholine.
To assess PrP Sc levels, cell lysates were subjected to limited proteolysis with proteinase K (20 g/ml at 37°C for 50 min), and protease-resistant PrP Sc was detected in Western blots. Unexpectedly, PrP Sc increased in a concentration-dependent manner as a result of the FB 1 treatment (Fig. 1A). This result was surprising as it contrasts strikingly with the effect of the hydroxmethylglutaryl-CoA reductase inhibitor lovastatin, which depletes cellular cholesterol and decreases PrP Sc in ScN2a cells (29). Short term (3-4 days) treatment with FB 1 also increased PrP Sc (data not shown), but the optimal effect was obtained after 6 -8 days of treatment, and 30 M FB 1 for 7 days was chosen as our "standard" FB 1 protocol. This rise in PrP Sc was observed in all other ScN2a clones treated with FB 1 , including ScN2a-C10 (not shown) and ScN2a-c cells, which express human caveolin 1 (see Figs. 1C and 5).
To analyze the effect of FB 1 on the incorporation of [ 3 H]choline into lipids, parallel samples were normalized for protein content, extracted, and analyzed by TLC and autoradiography (see under "Experimental Procedures"). Increasing concentrations of FB 1 decreased the amount of choline incorporation into newly synthesized SM down to 10% of control levels (Fig. 1B,  lanes 1-5), whereas the amount of radiolabeled PC remained essentially unchanged (except for the highest concentration of FB 1 ). Coomassie staining of the TLC plates showed that the total amount of SM mass also significantly decreased in FB 1treated cells (Fig. 1B, lanes 6 and 7).
The FB 1 -induced rise in PrP Sc was not due to an increase in the normal PrP precursor, because in uninfected N2a-c cells, the amount of PrP C was unaltered by the inhibitor (Fig. 1C,  lanes 1 and 2), whereas in their scrapie-infected ScN2a-c counterparts, the amount of proteinase K-resistant PrP Sc was once again greatly increased by FB 1 (lanes 3 and 4). Sarkosyl-soluble PrP, which consists primarily of PrP C (17), also remained unaltered in scrapie-infected cells treated with FB 1 (not shown). Finally, FB 1 treatment had no effect on the amount of two proteins that do not reside in rafts, the transmembrane chimera PrP-CD4 (29) (Fig. 1D, lanes 1-4) and the cytosolic protein kinase C ␣ (lanes [5][6][7][8]. Taken together, these results strongly suggest that the metabolism of the pathological PrP isoform is specifically altered by the FB 1 treatment. Decreased Sphingolipids in FB 1 -treated ScN2a Cells-The results shown in Fig. 1 suggest a correlation between decreased SM and enhanced PrP Sc formation in ScN2a cells. To see whether other sphingolipids also correlate with PrP Sc formation, we used the radiolabeled sphingoid base [ 3 H]sphinganine, a precursor of the ceramide backbone of sphingolipids, to radiolabel the cells. In contrast to choline, this precursor is expected to incorporate into all nascent SLs. ScN2a cells were treated with FB 1 (30 M) for 7 days in delipidated medium, and [ 3 H]sphinganine (3 ϫ 10 6 cpm/14 ϫ 10 7 cells) was added to the cells for the final 48 h. Results of the TLC analysis are summarized in Table I. FB 1 clearly reduced the incorporation of [ 3 H]sphinganine into ceramide, SM, and Glc-Cer (Table I). Coomassie Blue staining of the TLC plate showed that the mass of ceramide and SM, was also reduced by FB 1 (Table I). Glc-Cer did not stain with Coomassie Blue.
The ganglioside GM1 localizes to caveolae (58), buoys in density gradients following extraction of cells with TX-100, and has served as a marker of rafts in our previous studies (33). Because SM and GM1 are metabolically related, it was important to see whether GM1 levels are also altered by FB 1 . Cell surface GM1 was measured in a variety of N2a subclones using its ligand, CTXB. Prior to lysis, cells were cooled on ice, incubated with 5 g/ml CTXB for 30 min, and then rinsed thoroughly from the unbound toxoid. Cell lysates were then subjected to nondenaturing SDS-PAGE, and cell surface-bound CTXB was determined by Western immunoblotting developed with a CTXB antibody. Treatment with FB 1 significantly de-creased the level of cell-bound CTXB, and thus presumably of cell surface GM1, in both N2a cells and ScN2a cells (Fig. 1E).

Increased PrP Sc Correlates with Decreased SM in Cells
Treated with nSMase-To further explore the relationship of specific sphingolipids and PrP Sc formation, we used two pharmacological agents that selectively affect either SM or glycosphingolipids, respectively. To directly reduce the level of SM, we added a neutral bacterial sphingomyelinase (nSMase) to the cell medium. Because the enzyme caused cytopathic effects when applied for more than 4 days, we limited the experiments with this enzyme to 3 days. Because changes in PrP Sc formation are usually difficult to detect by Western analysis in short term experiments (due to very large background signals contributed by preexisting PrP Sc ), we used metabolic radiolabeling to follow the formation of PrP Sc in nSMase-treated cells. ScN2a-C10 cells were incubated with or without nSMase (0.5 or 2 units/ml) for 3 days in medium supplemented with delipidated fetal calf serum, pulsed with [ 35 S]methionine for 1 h, and then chased for another 6 h to permit the formation of PrP Sc (3). Newly formed PrP Sc increased in the cells treated with the higher concentration of nSMase ( Fig. 2A), similar to the results obtained with FB 1 (Fig. 1). Other ScN2a sublines were more sensitive than ScN2a-C10 to nSMase and displayed increased PrP Sc at lower enzyme levels.
The influence of nSMase (0.4 units/ml, 3 days) on the level of SM, Glc-Cer, and ceramide in ScN2a cells is summarized in Table I. ScN2a cells were incubated in the presence or absence   1 and 2) and ScN2a-c (lanes 3 and 4) cells lysates were analyzed in Western blots developed with R073. D, lanes 1-4, expression of the chimeric transmembrane-anchored CD4-PrP in N2a-CD4 and ScN2a-CD4 cells analyzed in Western blots developed with the mAb 3F4; lanes 5-8, N2a-c and ScN2a-c cells. Protein kinase C␣ was detected utilizing a commercial antiserum. E, cell surface GM1 was assessed utilizing its ligand CTXB on intact N2a, ScN2a, and ScC10 cells. Following FB 1 treatment, cells were incubated on ice with CTXB (5 g/ml, 30 min), rinsed, and lysed, and cell lysates were analyzed by nondenaturing SDS-PAGE. Western blot was developed with a CTXB antiserum. FB 1 decreased cell surface GM1. of nSMase for 3 days in medium containing delipidated serum, and [ 3 H]sphinganine was added to the cell medium for the last 48 h of the experiment. The nSMase treatment resulted in the almost complete loss of both radiolabeled and Coomassie stained SM (to 10 and 20% of control levels, respectively; see Table I). In contrast, nSMase significantly increased newly synthesized Glc-Cer (155% of control level). Ceramide remained essentially unchanged by this enzymatic treatment.
Thus, whereas nSMase reduced SM in the cells, its effect on Glc-Cer and ceramide were distinctly different from that of FB 1 . Hence, the results with nSMase suggest that elevated formation of PrP Sc correlates with a decrease in cellular SM but not with changes in either ceramide or Glc-Cer.
PDMP Increases SM and Reduces PrP Sc in ScN2a Cells-Several glycosphingolipids, and especially gangliosides, are enriched in TX-100-insoluble fractions (34, 43, 58 -60). To further investigate the relationship between glycosphingolipids and PrP Sc formation, we utilized the ceramide analog PDMP, which blocks glycosphingolipid synthesis by inhibiting the first step in the pathway of glycosphingolipid biosynthesis, which is catalyzed by UDP-glucose:ceramide glucosyl transferase (61)(62)(63)(64). Although the inhibition of cellular glycosphingolipids by PDMP is well documented, its effect on SM seems to depend on the cell type examined (65), and in some cases it has been shown to produce an increase in cellular SM (61,62). We studied the effect of PDMP on SM in several ScN2a sublines. ScN2a cells were treated with various concentrations of PDMP (0 -70 M) for 7 days, and radiolabeled choline was added for the last 72 h of the treatment (Fig. 2, B and C). Cell lysates were analyzed as described above. TLC analysis revealed that [ 3 H]choline incorporation into SM increased in a dose-dependent manner reaching a maximum at 70 M of PDMP (to 267% of control level). In another experiment, both [ 3 H]sphinganine incorporation into SM and total SM mass were again increased in cells treated with PDMP (Table I). Under these conditions, radiolabeled ceramide also increased, but Glc-Cer decreased (Table I). SM mass also increased in cells treated with PDMP (based on Coomassie Blue staining), but to a smaller degree than the newly synthesized lipid (121% of control levels, Table I). Cell surface GM1 was also reduced in PDMP treated cells, based on CTXB binding (not shown).
Having established that PDMP increases SM while decreasing glycosphingolipids such as Glc-Cer and GM1 in ScN2a cells, we examined the effect of this inhibitor on PrP Sc levels. The level of proteinase K-resistant PrP Sc in aliquots from the experiment described in Fig. 2C were assessed by Western analysis (Fig. 2B). PDMP reduced the levels of PrP Sc in a dose-dependent manner (see also Fig. 5). This result further supports an inverse correlation between levels of SM and PrP Sc .
Increased Formation of Protease-resistant PrP Sc in Cells Treated with FB 1 -The effect of FB 1 on the steady state levels of PrP Sc could be due either to the increased formation of PrP Sc or a decrease in the rate of its clearance from the cells. To decide between these alternatives, we performed kinetic (pulsechase) experiments in N2a-C10 and ScN2a-C10 cells (Fig. 3). Eighteen hours prior to the pulse radiolabeling, as well as during the pulse period, tunicamycin (0.75 g/ml) was added to the cell medium to reduce the M r heterogeneity of the PrP and permit a better electrophoretic resolution of its isoforms (Nlinked glycosylation is not necessary for the formation of protease-resistant PrP Sc (8)). N2a-C10 and ScN2a-C10 cells were treated with FB 1 (30 M, Fig. 3, lanes 1, 3, and 6) for 12 h prior to radiolabeling, as well as during the pulse and the chase periods. Cells were pulse-radiolabeled with [ 35 S]methionine for 1 h and then either chased for 6 h in unlabeled medium, to permit the formation of protease-resistant PrP Sc (Fig. 3, lanes  5 and 6), or immediately lysed at the end of the pulse (lanes  1-4), and cell lysates were then processed by PrP immunoprecipitation. In Fig. 3, lanes 3-6, lysates were digested with proteinase K (20 g/ml at 37°C for 50 min) and then treated  C10 (lanes 1 and 2) and ScC10 cells (lanes 3-6) were treated for 18 h with tunicamycin (0.75 g/ml) to reduce the glycosylation of PrP. Cells were then pulsed for 1 h with [ 35 S]methionine (0.3mCi/4 ϫ 10 7 cells). Radiolabeled cells were either immediately lysed (lanes 1-4) or chased for 6 h. FB 1 (30 M) or carrier Me 2 SO were added to the cell medium for 12 h prior to the radiolabeling, and then maintained throughout the pulse and the chase periods. PrP C (lanes 1 and 2) or protease-resistant PrP Sc (lanes 3-6) was then immunoprecipitated from the cell lysates with rabbit antiserum R073, followed by SDS-PAGE and autoradiography.
with GdnSCN (3 M for 10 min at room temperature) to permit PrP Sc immunoprecipitation. The amount of radiolabeled newly synthesized PrP C was either unchanged or slightly decreased in cells treated with FB 1 (Fig. 3, lanes 1 and 2). In contrast, the formation of PrP Sc was enhanced, and it seemed to proceed at an accelerated rate (lanes 3-6) in cells treated with FB 1 . The accelerated metabolism of PrP under SM depletion is the opposite of the effect observed in cells deprived of cholesterol, in which both the degradation of PrP C and the formation of proteinase K-resistant PrP Sc were slowed (29).

and nSMase Decrease SM in Low Density Fractions but Do Not Alter the Buoyancy of PrP and Several Other Raft
Residents in Flotation Gradients-Next we set out to examine the effect of FB 1 , nSMase, and PDMP on the properties of rafts in N2a, ScN2a, and in a variety of their subclones. In particular, we used N2a-c and ScN2a-c cells, which express human caveolin-1/VIP21, so that we could assess the effects of these inhibitors on caveolin as well. Expression of caveolin-1 in these cells appears to have no effect on the formation of PrP Sc . 2 Results obtained with these cells are presented in Figs. 4 and 5. Cells were treated with FB 1 (30 M, 6 days), nSMase (2 units/ ml, 3 days), or PDMP (70 M, 6 days) as indicated. [ 3 H]Choline (2 Ci/ml) was added to the medium for the last 72 h of the experiment. The cells were then lysed in either cold TX-100 or cold CHAPS (see below) and then analyzed by a standard flotation procedure in Nycodenz (33).
To assess the effect of FB 1 , nSMase, and PDMP on the SM content of rafts, we analyzed gradient fractions resolved on TLC. Because TX-100 interferes with the separation of lipids by TLC, we replaced this detergent with CHAPS (1%) (34) in these experiments. PrP flotation results obtained with CHAPS were similar to those obtained with TX-100 (not shown). ScN2a-c cells treated with the lipid modulators were radiolabeled with [ 3 H]choline for the last 72 h of the experiment. Under these conditions, the radiospecific activity of the choline lipids reaches a steady state and thus represents the mass of these molecules (66). Lipids were extracted from the low density fractions (fractions 1-7) of the density gradients, subjected to mild alkaline hydrolysis to eliminate radiolabeled phosphatidylcholine, and separated on TLC plates, which were then examined by autoradiography. FB 1 and nSMase (Fig. 4, B and C) reduced the amount of [ 3 H]SM in low density fractions to 53 and 46% of control levels (Fig. 4A), respectively. In contrast, PDMP increased the amount of [ 3 H]SM in these fractions to 127% of control levels.
We next examined the flotation properties of the PrP isoforms in cells treated with the various effectors (Fig. 5). In untreated N2a-c cells, PrP C floated to the lightest fractions of the gradient, confirming that PrP C is indeed attached to buoyant TX-100-insoluble rafts in these cells (Fig. 5A). When the cells were treated with FB 1 , the amount of PrP C remained unaltered, and its migration pattern was unchanged (Fig. 5B). This result contrasts with that obtained with lovastatin, which renders most PrP C soluble in TX-100 (29), and correlates well with the opposing influence of these two inhibitors, FB 1 and lovastatin, on the formation of PrP Sc . We next analyzed PrP Sc from ScN2a-c cells in flotation experiments. Proteinase K-resistant PrP Sc characteristically peaked around fractions 4 -8 that are denser than those containing PrP C (33). In cells treated with FB 1 , the amount of PrP Sc increased substantially (compare Fig. 5, C and D), but its flotation pattern remained unchanged. A similar pattern was obtained in experiments with nSMase (Fig. 5E). PDMP caused a dramatic drop in PrP Sc (Fig. 5F), although the remaining PrP Sc still migrated to fraction 5, where it usually peaks. Interestingly, PDMP also somewhat reduced PrP C in some but not all experiments (by up to 30%) but the distribution of PrP C along the Nycodenz gradient remained unchanged. The pronounced differences in PrP Sc levels caused by the effectors in flotation gradients stand in contrast to unchanged content in total protein as measured in each fraction by BCA analysis (Fig. 5K).
Thus, a reduction of as much as 50% in the level of SM in the low density fractions (Fig. 4) did not alter the buoyancy of either PrP C or PrP Sc in flotation gradients. However, these changes did correlate with significant changes in the amount of PrP Sc formation. To determine the influence of changes in SM on other raft residents, we chose to analyze the migration pattern and the amount of two non-GPI raft-resident molecules, caveolin and GM1. For this purpose, we used fractions from the gradients depicted in Fig. 5, C-F. Caveolin was detected in Western immunoblots developed with a caveolin antiserum (Fig. 5, G-J). No change in the flotation pattern or in the amount of caveolin was recorded in cells treated with either FB 1 , nSMase, or PDMP. Our results are in line with previous observations obtained by Hanada et al. (47), who showed that depletion of cholesterol and/or sphingolipids in Chinese hamster ovary cells did not alter the expression levels on caveolin. Finally, we assessed the effect of FB 1 and nSMase on the buoyancy of cell surface GM1 (Fig. 4E). The vast majority of CTXB was found in the light fractions of the gradient, and this distribution was undisturbed by either FB 1 or nSMase. However, FB 1 reduced the binding of CTXB to rafts by about 50%, whereas nSMase treatment did not change CTXB binding significantly. PDMP caused a similar decrease in CTXB binding, as seen with FB 1 , but its distribution along the gradient again remained unchanged (results not shown). It is unlikely that CTXB itself could have influenced the clustering of GM1 within detergent-insoluble rafts, because this toxoid acts as a monovalent ligand to GM1 and does not trigger its aggregation on the plasma membrane (67).
Thus, the level of SM as well as that of GM1 within rafts can be modified without altering the flotation characteristics of several raft resident molecules, suggesting that these sphingolipids do not significantly modulate the buoyant density of rafts in TX-100 (see under "Discussion"). This contrasts with the critical role played by cholesterol in maintaining the insolubility of PrP in TX-100 (29). DISCUSSION The data presented here support the view that cholesteroland sphingolipid-rich rafts regulate the formation of PrP Sc , and thus presumably the propagation of prions. Viewed through combining information at the physicochemical and cell biology levels, these data may also point to possible mechanisms by which rafts contribute to PrP Sc formation.
Increased PrP Sc in SM-depleted Cells-Comparison of the results obtained with FB 1 , nSMase, and PDMP shows that PrP Sc formation correlates inversely with the amount of SM in the cells, but is relatively insensitive to the amount of GM1 (Fig. 5), ceramide, and Glc-Cer (Table I). For instance, the levels of GM1 (and presumably of other glycolipids as well) could be either decreased (with FB 1 ) or left unchanged with nSMase, resulting in both cases in an increased formation of PrP Sc . Similarly, FB 1 and nSMase both reduce SM but had an opposite influence on ceramide levels, and yet both treatments led to increased PrP Sc . This difference between SM and other sphingolipids could originate in disparate roles played by SM and glycosphingolipids in regulating raft structure and function. Although the detergent insolubility of rafts may be determined in part by the acyl tails of sphingolipids (68,69), it is certainly possible that head groups could modulate essential aspects of rafts (70). We cannot rule out, however, the possibility that the determining lipid parameter of PrP Sc formation is the total level of sphingolipids in rafts, rather than the identity of their head group.
The mechanisms by which rafts regulate the formation of PrP Sc in prion-producing cells are not yet understood, but several models have been proposed (33). One possibility is that rafts provide physical platforms that laterally concentrate PrP C substrate and PrP Sc seed within cellular membranes, thereby enhancing the opportunity of these molecules to interact (29,33). Lateral phase separation and domain formation in membranes is known to affect the rate of many reactions in membranes through the control of crowdiness of interacting species. The crowding of raft residents on the cell surface has been observed for other GPI proteins, such as the folate receptor (40), and has now been substantiated both by cross-linking experiments performed in live cells (71) and by measuring energy transfer between GPI-linked folate receptor molecules loaded with a fluorescently labeled ligand (72). Putative accessory molecules, such as Protein X (73), that have been envisaged to play a role in the formation of PrP Sc , could also be made more available for interaction with PrP if they reside within rafts.
Whether such a clustering might be amplified upon SM depletion remains to be seen. Even when present in membranes in low molar fractions, SM induces the formation of plane domains (Refs. 74 and 75 and references therein). Reducing SM in cells causes the concomitant removal of cholesterol from membranes, probably because the capacity of membranes to solubilize cholesterol is impaired (76,77). Conceivably, the coordinate depletion of these two lipids would result in a diminution in the total cell membrane area devoted to rafts and a further crowding of interacting PrP species.
Another possibility is that SM depletion would alter other properties of rafts that are relevant to the chances of interaction between their resident molecules. For instance, interaction of PrP C with PrP Sc could be increased due to an altered fluidity of these domains. Direct measurements of GPI protein mobility in cell membranes support the notion that various physical properties of rafts are altered by pharmacological intervention in sphingolipid metabolism. Hanada et al. (78) demonstrated recently that short term (16 h) incubation of cells with FB 1 rendered a GPI-linked protein (a CD14 chimera) more sensitive to phosphatidylinositol-specific phospholipase C and that long FIG. 5. Flotation profiles of the PrP isoforms and of caveolin unchanged by FB 1 and nSMase. N2a (A and B) and ScN2a-c (C-J) cells were treated with the indicated inhibitors in delipidated medium (FB 1 , 30 M for 6 days; nSMase, 2 units/ml for 4 days; or PDMP, 70 M for 7 days). Cells were extensively washed with cold PBS, cooled on ice and lysed with flotation lysis buffer (containing 1% TX-100). Prior to flotation, lysates from ScN2a-caveolin cells were incubated with 0.5 g/ml CTXB on a rotator at 4°C (20 min) to allow binding of the toxoid to both intracellular and plasma membrane ganglioside GM1. Lysates were then subjected to the standard flotation procedure. Gradient fractions were then subjected to various assays. One half of each fraction was used to determine either PrP C (A and B) or proteinase K-resistant PrP Sc (C-F) in Western immunoblots developed with the rabbit PrP antiserum R073. Fractions from ScN2a-c cells were also assayed for their content in caveolin 1 in Western blots (G-J). The protein content of fractions depicted in C-E was determined by BCA (K), and the distribution of CTXB in these gradients was determined by a Western immunoblot (Fig. 4E). chain SM reversed this effect when administered to the cell medium. This again suggests that depletion of sphingolipids may cause a spatial alteration in the structure of rafts, hence affecting its interaction with resident GPI-anchored proteins.
Finally, rafts could encourage the formation of PrP Sc by providing propitious trafficking and sorting pathways. The formation of PrP Sc seems to be restricted to specific subcellular locations, as it appears to occur either on the cell surface or in the endocytic pathway (26 -28, 79), but does not happen in the ER-Golgi of cells treated with brefeldin A (80). Thus, variations in either exocytosis or endocytosis could modify the formation of PrP Sc . Results from many laboratories have shown that altering sphingolipid synthesis often modifies vesicular traffic. Transport of both SL and vesicular stomatitis virus G protein from the Golgi apparatus to the plasma membrane is retarded following treatment of cells with PDMP (81). Export of SM is believed to proceed via vesicles that contain raft-like, detergent-insoluble domains (37,82). Endocytosis is also vastly modified by SL inhibitors. In J774 macrophages, as much as 30% of the plasma membrane was internalized within minutes of treatment with nSMase, through 400-nm vesicles lacking a distinguishable coat (83). These observations suggest that large portions of bilayers may become unstable when cell surface SM is hydrolyzed by exogenous nSMase. Whether rafts residents are involved in this enhanced internalization has not been addressed by these investigators. Increased recycling of PrP C could well be accompanied by an augmentation of PrP Sc formation, as well as by the observed acceleration in the degradation of PrP C . Finally, it is interesting to note that brefeldin A, which blocks the formation of PrP Sc (80), also alters the metabolism of SM in cells (84).
Because ceramide and many of its metabolites play important roles in signal transduction, it is also possible that the PrP Sc increase could be caused indirectly by altered levels of lipid second messengers. Because ceramide could either be decreased (with FB 1 ) or increased (with nSMase), resulting in identical changes in PrP Sc synthesis, it is unlikely that the total amount of this lipid in the cell could directly regulate PrP Sc formation. However, the subcellular distribution of ceramide was not addressed in our experiments and could play a crucial role in regulating PrP Sc .
Contrasting Influence of FB 1 and Lovastatin on PrP Metabolism-Several studies have shown that both SLs and cholesterol are essential for the proper functioning of caveolae. For instance, the potocytosis of folate was markedly decreased when MA104 cells were treated with compactin to inhibit the synthesis of cholesterol (38). A similar effect on folate potocytosis was observed when Caco-2 cells were treated with FB 1 to decrease cellular SLs (43). Because PrP Sc formation drops dramatically in cholesterol-depleted cells (29), we were expecting that reducing SLs in ScN2a cells would also hamper the formation of this pathological PrP isoform. That FB 1 and nSMase both enhance PrP Sc formation thus stands in striking contrast with results obtained in cells starved from cholesterol. Indeed, depletion of cells from cholesterol or from SL had diametrically opposed influence on two additional aspects of PrP metabolism. First, the metabolic degradation of PrP was accelerated in cells treated with either FB 1 alone or with a combination of FB 1 and nSMase, 2 whereas in cells treated with lovastatin, PrP turnover was slowed (29). This result suggests that PrP C degradation and PrP Sc formation may be coupled in ScN2a cells. Second, the opposing action of lovastatin and of SL inhibitors on the formation of PrP Sc also correlates well with their contrasting influence on PrP C -containing rafts. Whereas in cholesterolstarved cells, most PrP C becomes soluble in TX-100 (29), both PrP C and PrP Sc remain attached to TX-100 buoyant rafts (Fig.   4) in ScN2a cells depleted from SLs (with FB 1 or nSMase). This correlation thus supports the thesis that association of PrP C with rafts is essential for the efficient pathological conversion of this protein into PrP Sc . Our results also indicate that cholesterol and SM play entirely different roles in determining the properties of rafts. That rafts support the formation of the abnormal PrP Sc (and thus the replication of prions) is perhaps the first example of their involvement in a pathological process. Whether the lipid composition of rafts could be a future therapeutic target in prion diseases remains to be seen.