Characterization of detergent-insoluble complexes containing the cellular prion protein and its scrapie isoform.

Cells infected with prions contain both prion protein isoforms cellular prion protein (PrPC) and scrapie prion protein (PrPSc). PrPSc is formed posttranslationally through the pathological refolding of PrPC. In scrapie-infected ScN2a cells, the metabolism of both PrP isoforms involves cholesterol-dependent pathways. We show here that both PrPC and PrPSc are attached to Triton X-100-insoluble, low-density complexes or “rafts.” These complexes are sensitive to saponin and thus probably contain cholesterol. This finding suggests that the transformation PrPC → PrPSc occurs within rafts. It also reveals the existence of rafts in late compartments of the endocytic pathway, where most PrPSc resides. When Triton X-100 lysates of cells were incubated at 37°C prior to density analysis, PrPC was still found in buoyant complexes, although it now failed to sediment at high speed. This property was shared by another glycophosphatidyl inositol protein, Thy-1, and also by the raft resident GM1. In one ScN2a clone and in the brain of a Syrian hamster with scrapie, Triton X-100 extraction at 37°C permitted resolution of PrPC and PrPSc into two distinct peaks of different densities. This suggests that there are two populations of PrP-containing rafts and may permit isolation of PrPC-specific rafts from those containing PrPSc. Our findings reinforce the contention that rafts are involved in various aspects of PrP metabolism and in the “life cycle” of prions.

Prions are unique proteinaceous pathogens that cause a series of fatal encephalopathies such as Creutzfeldt-Jakob disease of humans, scrapie of sheep, and bovine spongiform encephalopathy (1). Prions seem to propagate in the host by posttranslationally (2, 3) refolding a normal host protein, the cellular prion protein (PrP C ), 1 to an aberrant conformation (4,5). The only known component of prions is the misfolded isoform of PrP C , the scrapie prion protein (PrP Sc ) (6,7). Current evidence argues that direct interaction of PrP Sc with PrP C is a prerequisite for the transformation PrP C ϩ PrP Sc 3 2PrP Sc (8,9). PrP C is a phosphoinositol glycolipid (GPI)-anchored glyco-protein present on the surface of neurons and other cells (10,11). The PrP isoforms appear to be chemically identical (12) but differ in their conformation (4); PrP C contains ϳ40% ␣-helix and is devoid of ␤-sheet, whereas PrP Sc has more than 40% ␤-sheet (4,(13)(14)(15)(16). The two PrP isoforms differ considerably in their properties; PrP C is readily soluble in most detergents and is completely degraded by proteases, whereas PrP Sc is insoluble in detergents, possesses a protease-resistant core termed PrP27-30, and polymerizes into amyloidic structures called prion rods (17,18). Since no isoform-specific PrP antibody has yet been developed, the disparate properties of PrP C and PrP Sc serve as the sole ways to differentiate experimentally between these proteins.
The subcellular sites where PrP Sc is formed, and the trafficking pathways leading to these sites, remain largely unknown. Scrapie-infected mouse neuroblastoma ScN2a cells synthesize both PrP C and PrP Sc , whereas only PrP C is found in uninfected cells (2,19,20). Like other GPI proteins, most PrP C is found in cholesterol-rich, detergent-resistant microdomains of the plasma membrane (21)(22)(23). The PrP isoforms also localize to caveolae-like domains isolated without the use of detergents (70). Plasma membrane PrP C seems to recycle to the interior of the cell in about 1 h (24). Two mutually exclusive posttranslational fates await PrP C in ScN2a cells. Although most PrP C molecules turn over with a t 1/2 of ϳ6 h by a two-step degradation pathway (22,25,26), a small minority of PrP C molecules (ϳ5%) escape degradation, acquire a protease-resistant core, and become PrP Sc (2,3,27). Whether PrP Sc is formed on the plasma membrane or during the internalization of PrP C is unknown (27)(28)(29)(30). PrP Sc is further N-terminally trimmed in an acidic compartment (26,30) and accumulates primarily in lysosomes (20,26,31,32). Interfering with cholesterol-dependent structures or pathways in ScN2a cells inhibits both metabolic fates of PrP C (22). Depriving ScN2a cells of cholesterol inhibited the formation of PrP Sc and also retarded the degradation of PrP C . Dissociating PrP C from detergent-insoluble complexes by replacing the GPI anchor of PrP C with the transmembrane and cytoplasmic regions of mouse CD4 almost completely prevented the formation of PrP Sc (22).
Most, if not all, GPI-anchored proteins become largely insoluble in cold TX-100 while traversing the Golgi stacks (22,23). In a seminal article, Brown and Rose (21) demonstrated that GPI-anchored proteins owe this insolubility to their association with membrane complexes enriched in cholesterol, sphingolipids, and glycolipids. Such membrane microdomains or "rafts" of specialized lipid composition had previously been hypothesized by Simons and van Meer (33) to explain the sorting of lipids in polarized epithelial cells. In these cells, GPI anchors act as a sorting signal that functions in concert with these membrane subdomains (21, 34 -37). In addition to containing GPI proteins, detergent-insoluble membrane complexes are also enriched in several cytoplasmic proteins, including nonreceptor-type tyrosine kinases (38,39). Caveolae, which are cholesteroldependent invaginations of the plasma membrane present in many cell types, appear to constitute a specialized subset of all cellular rafts. Caveolae are involved in the potocytosis of folate, among other putative functions (36,40), and are equipped with components of the machinery of vesicular fusion (41). In cells containing caveolae, detergent-insoluble complexes are also enriched in caveolin/VIP-21, a component of the striated coat of caveolae (37,(42)(43)(44). The existence of caveolae and caveolin in N2a cells remains a matter of debate (23,45,46). However, these cells do contain rafts, and since we were unable to observe caveolae in ScN2a using transmission electron microscopy in thin sections, 2 we have limited the experiments described here to detergent-insoluble rafts. In this article, we further characterize the relationship of the PrP isoforms with rafts.
The normal physiological configuration of rafts prior to cell lysis is still unknown, but they may reflect the lipid microenvironment experienced by their associated proteins in cellular membranes. As criteria for the attachment of macromolecules to rafts, we have used here operational definitions generated by the article by Brown and Rose (21) as well as through the work of other groups. Namely, the PrP isoforms are assumed to be attached to rafts if they fulfill all three of the following criteria: (i) their solubility in TX-100, as assayed by high speed sedimentation, is temperature-dependent, i.e. the protein sediments at 4°C but stays in supernatant if the lysates are warmed to 37°C; (ii) they float up density gradients following extraction in cold Triton X-100; and (iii) their buoyancy is reduced when saponin is added to the extract (47) (indicating the importance of cholesterol in the attachment of GPI to these complexes) or when TX-100 is replaced with n-octyl ␤-D-glucopyranoside (NOG) in the lysis of the cells (21). Throughout this study, we used the ganglioside GM1 as a marker of rafts (48).
We report here that the pathological prion protein isoform PrP Sc is attached to rafts in ScN2a cells. Since the metabolic precursor PrP C also resides in these complexes, the transformation PrP C 3 PrP Sc thus appears to occur within rafts. When cells were lysed in TX-100 at 37°C instead of the "canonical" 4°C (21), PrP C was "solubilized" as judged from its sedimentation properties, but it still floated up density gradients. This property was shared by another GPI protein, Thy-1, which was extracted from EL-4 T-lymphoma cells. In one ScN2a clone, as well as in the brain of a Syrian hamster with scrapie, TX-100 extraction at 37°C permitted resolution of PrP C and PrP Sc into two distinct peaks with different densities. This result suggests that there are two populations of PrP-containing rafts and may permit the isolation of PrP C -specific rafts from those containing PrP Sc . Our findings reinforce the contention that rafts are intimately involved in PrP metabolism and in the "life cycle" of prions. They also reveal the presence of detergent-insoluble membrane complexes in intracellular compartments, probably secondary lysosomes, in which a major portion of PrP Sc resides.
Cells and PrP Preparations-Mouse N2a cells were originally obtained from the American Type Culture Collection (Rockville, MD). Mouse lymphoma EL-4 cells expressing Thy-1 were from Dr. M.
Baniyash (Hebrew University, Jerusalem, Israel). ScN2a is the persistently infected N2a clone described by Butler et al. (19). N2a-c3 and ScN2a-MHM2 express the MHM2-PrP chimeric gene driven by the pSPOX vector (49). N2a-c10 and ScN2a-c10 cells express the same chimeric PrP gene driven by the commercial expression vector pCI-neo (Promega, Madison, WI). All the cells were grown at 37°C in Dulbecco's modified Eagle's medium 16 supplemented with 10% fetal bovine serum. Cells expressing the chimeric MHM2-PrP gene were maintained in 1 mg/ml G418. Syrian hamster prion rods purified by the sucrose gradient method (50) were a kind gift from Dr. S. B. Prusiner (University of California, San Francisco). Enriched preparations of MHM2-PrP carrying the 3F4 epitope were prepared from N2a-c3 cells using ion metal affinity chromatography as modified from the method of Pan et al. (51).
Antibodies-Rabbit antiserum R073 reacts with Syrian hamster (SHa) PrP and mouse (Mo) PrP, as well as with MHM2-PrP, and its specificity in Western blotting has been described elsewhere (20,52). 3F4 is a monoclonal antibody raised against SHaPrP27-30 (53). Its epitope, which includes Met 109 and Met 112 (54), is present on SHaPrP as well as on MHM2-PrP, but is absent from MoPrP. Thus this mAb does not recognize the wild type MoPrP endogenous to N2a cells but does react with the products of the chimeric MHM2-PrP genes (49). Both antibodies were used at a dilution of 1:5000 (of the serum or the ascitic fluid). Thy-1 mAb G7 supernatant was a kind gift from Dr. M. Baniyash.
Flotation Assays-Flotation of detergent-insoluble complexes was performed as described by Brown and Rose (21) with some modifications, as follows ("the standard flotation"). Confluent cells growing in two 10-cm plates (about 3 ϫ 10 7 cells) were incubated with 400 l of lysis buffer (150 mM NaCl, 25 mM Tris-HCl, pH 7.5, 5 mM ETDA, 1% Triton X-100) on ice in the cold room for 30 min. This ratio of lysis buffer volume and cell number was kept constant throughout the experiments described here. In some cases, other detergents were used, or the lysates were further incubated at 37°C, as specified in each experiment. All the subsequent steps were performed at 4°C. Lysates were adjusted to 35% Nycodenz by adding an equal volume of ice-cold 70% Nycodenz prepared in TNE (25 mM Tris, pH 7.5, 150 mM NaCl, 5 mM EDTA) and loaded at the bottom of Beckman Instruments TLS-55 ultracentrifuge tubes. A 8 -35% Nycodenz linear step gradient in TNE was then overlaid above the lysate (200 l each of 25, 22.5, 20, 18, 15, 12, and 8% Nycodenz), and the tubes were spun at 55,000 rpm for 4 h at 4°C in a TLS-55 rotor (g av , 200,000 ϫ g). Fractions were collected from the top of the tube. We found that Nycodenz advantageously replaced sucrose (21) in the flotation assay, permitting us to reduce the centrifugation time needed to reach density equilibrium from 18 to 4 h.
To eliminate PrP C from fractions we used proteinase K. The fractions were made 1% with NOG (to disrupt rafts and their possible protective effects on PrP C (21,23)) and incubated with 20 g/ml proteinase K (37°C, 1 h). The reaction was stopped with 1 mM phenylmethylsulfonyl fluoride prior to Western analysis. To isolate PrP C from PrP Sc in fractions, we took advantage of the differential solubility of these proteins in Sarkosyl (18). Fractions were diluted three times in TN (150 mM NaCl, 10 mM Tris, pH 7.8) to reduce their density, made 1% with Sarkosyl, and incubated on ice for 30 min to promote the solubilization of rafts and the aggregation of PrP Sc . The fractions were then spun at 45,000 rpm for 2 h at 4°C in a Beckman TLA-45 rotor (g av , 109,000 ϫ g), and the supernatants (enriched in PrP C ) were collected for further analysis.
In some cases, the cells were labeled with CTXB-POD prior to lysis and flotation. The procedure was performed in the cold room on ice. Cells growing on a 10-cm plate were rinsed three times with ice-cold phosphate-buffered saline and then incubated with ice-cold phosphatebuffered saline containing 5 g/ml CTXB-POD (45 min on a rocker) prior to their lysis in TX-100 lysis buffer. The cells were then rinsed three times in ice-cold phosphate-buffered saline to remove unbound toxoid and then directly lysed as described before. In some cases, the toxoid conjugate was added directly to cell lysates rather than to whole cells, as described in "Results." Detection of Cell Surface Thy-1-Cell surface biotinylation of EL-4 cells was performed as follows. Cells growing in a T 75 flask were resuspended in ice-cold buffer A (150 mM NaCl, 1 mM MgCl 2 , 0.1 mM CaCl 2 , 20 mM Hepes, pH 7.4), rinsed three times in the same buffer, and then incubated for 40 min on ice in buffer A containing NHS-X-biotin (watersoluble; 500 g/ml). The cells were then rinsed three times in cold buffer A supplemented with 50 mM NH 4 Cl to quench the biotinylation reagent and then lysed as described in the text. To immunoprecipitate Thy-1 from fractions of Nycodenz gradients, the fractions were made 1% with Sarkosyl, mixed with 1 ml of G7 mAb hybridoma supernatant, and incubated for 18 h at 4°C. Protein A-Sepharose was then added, and 2 L. Laszlo and A. Taraboulos, unpublished results. the mixture was incubated for an additional 30 min at room temperature. The beads were rinsed five times in TNS (100 mM NaCl, 1% Sarkosyl, 10 mM Tris-HCl, pH 7.5) and then boiled in SDS sample buffer prior to analysis by SDS-polyacrylamide gel electrophoresis.
PrP Sc -specific dot immunoblots were performed as described (52). Briefly, aliquots from each fraction were dotted on a nitrocellulose membrane. The membrane was thoroughly air dried, rewetted with TBST, and incubated with proteinase K (50 g/ml, 37°C, 1 h). The membranes were then rinsed in water and sequentially incubated with 1 mM phenylmethylsulfonyl fluoride in TBST (30 min, room temperature) to stop the proteolysis and with guanidine thiocyanate (GdnSCN) (3 M, 5 min, room temperature) to expose PrP Sc epitopes. The membranes were then blocked with 5% milk and further treated for PrP immunodetection as described for Western blots. (21), PrP C clusters on the cell surface (22,31,45) and is insoluble in cold TX-100, as judged by its sedimentation properties (22,23). Because glycolipid/cholesterol rafts and cholesterol-dependent subcellular pathways appear to play such a crucial role in the metabolism of PrP and thus in the biogenesis of prions (22), we sought to better characterize the relationship of the PrP isoforms to these complexes.

Both PrP C and PrP Sc Localize to Low Density Detergentinsoluble Complexes-Like other GPI-anchored proteins
We first examined the solubility of PrP C from N2a cells in cold TX-100, using a flotation assay adapted from the method of Brown and Rose (21). N2a cells (Fig. 1A) were lysed in 1% TX-100 lysis buffer on ice in the cold room, brought to 35% Nycodenz, overlaid with lower density Nycodenz cushions, and then spun for 4 h at 4°C at 55,000 rpm in a TLS-55 rotor (the standard flotation assay). Western analysis verified that PrP C migrated up the gradient as expected from a resident of rafts (Fig. 1A). Identical results were obtained with N2a-c3 and N2a-c10 cells that overexpress the chimeric MHM2-PrP gene (not shown). In contrast, most cellular proteins remained in the bottom fractions (8 -11) of the gradient, as judged from silver staining of the gel (not shown).
We compared the distribution of PrP C in the flotation gradient with that of another resident of rafts, the cell surface ganglioside GM1. This glycolipid is enriched in caveolae of A431 cells (48) and partitions with low density, TX-100-insoluble complexes in lymphocytes (57). To detect cell surface GM1 in N2a cells, we used its ligand CTXB. N2a cells were incubated with CTXB-POD (5 g/ml) at 0°C for 45 min, and then the cells were rinsed, lysed in TX-100, and subjected to the standard flotation assay. CTXB-POD was detected in fractions using a simple dot blot. CTXB migrated to fractions identical to those reached by PrP C (Fig. 1B), as expected from a resident of plasma membrane rafts. Interestingly, the migration of PrP C in the Nycodenz gradient was unaltered in cells preincubated with the toxoid (not shown). This observation is in accord with results of Fra et al. (57), who have previously shown that CTXB induces the capping of GM1 but fails to induce the cocapping of the GPI-anchored Thy-1 in T hybridoma cells, suggesting that, although experiencing a similar membrane microenvironment, these two residents of rafts are not necessarily linked.
We presume that PrP C is attached to rafts through its GPI moiety. Lehmann and Harris (58) have recently shown, how-ever, that recombinant PrP molecules carrying one of several pathogenic mutations that cause familial Creutzfeldt-Jakob disease in humans may have membrane attachment sites additional to the GPI anchor (when expressed in Chinese hamster ovary cells). The results of these investigators prompted us to verify whether PrP C in N2a cells (which do not contain any of these pathological mutations) is bound to membranes exclusively through its GPI anchor, or whether there is a subpopulation of PrP C molecules that have an alternative or additional mode of membrane attachment. To check this, we biotinylated N2a-c10 cells with membrane-impermeant NHS-X-biotin and then incubated them with PIPLC (0.1 units/ml, 37°C, 2 h). To prevent the synthesis of new PrP C molecules during the enzymatic treatment, we added cycloheximide (20 g/ml) to the culture medium during the incubation with PIPLC. The cells were then rinsed and lysed, and PrP was immunoprecipitated using the PrP antiserum R073, which recognizes both wild type and MHM2-PrP expressed in these cells. This treatment re- A, PrP C in normal cells. D, protease-resistant PrP Sc in prion-infected cells. Fractions were incubated with proteinase K (20 g/ml, 37°C, 1 h) to eliminate PrP C prior to the electrophoresis. F, Sarkosyl-soluble PrP in prion-infected cells. Fractions were diluted 3 ϫ with TNE, incubated for 30 min with 1% Sarkosyl at 4°C, and then spun at 45,000 rpm for 2 h at 4°C to eliminate most PrP Sc . The supernatant was then subjected to electrophoresis. In B and E, the cells were incubated with CTXB-POD (5 g/ml, 45 min, 4°C) prior to lysis and flotation. Chemoluminescence was used to detect POD-containing fractions in dot blots. Following lysis in cold TX-100, both PrP isoforms float to light fractions also populated by CTXB-POD. In C, biotinylation was used to determine how much cell surface PrP C is removed by PIPLC. N2a-c10 cells were surface-biotinylated with NHS-X-biotin (500 g/ml, 15 min, 10°C) and then treated with PIPLC (0.1 units/ml, 2 h, 37°C) as indicated. Cycloheximide (CHX, 20 g/ml) was added to prevent PrP C synthesis during the PIPLC treatment. Cellular PrP was then immunoprecipitated with the PrP antiserum R073, and the immunoprecipitates were analyzed Western immunoblotting developed with streptavidin-POD. moved all detectable PrP C molecules from the cell surface (Fig.  1C). We thus conclude that these molecules are attached to the cell surface solely through their GPI anchor.
Having shown that PrP C is attached to buoyant TX-100insoluble fractions in N2a cells, we turned to the pathological PrP isoform PrP Sc . Because PrP Sc is inherently insoluble in detergents (18), it is not possible to determine whether it is attached to TX-100-insoluble complexes simply by assessing its sedimentation properties following extraction of cells with TX-100. It was thus hitherto unclear whether PrP Sc , an abnormal GPI protein that accumulates primarily intracellularly (20,31), associates with detergent-insoluble complexes like most normal GPI-anchored proteins. Furthermore, although PrP Sc possesses a GPI tail (11), it is not known whether this moiety serves at all to anchor this pathological protein to cellular membranes. To determine whether PrP Sc is attached to detergent-insoluble complexes, TX-100 lysates of ScN2a cells (not shown), as well as of ScN2a-MHM2 cells that overexpress the chimeric gene MHM2-PrP (Fig. 1D), were subjected to a Nycodenz flotation assay. Gradient fractions were first incubated with proteinase K to degrade PrP C prior to analysis by SDSpolyacrylamide gel electrophoresis. Like PrP C , protease-resistant PrP migrated to the upper part of the gradient (Fig. 1D), as would be expected from a resident of rafts. Identical results were obtained with ScN2a cells (not shown). When ScN2a-MHM2 cells were incubated with CTXB-POD prior to the lysis, the toxoid migrated to fractions identical to those reached by PrP Sc (Fig. 1E).
Prion-infected ScN2a cells contain both PrP Sc and PrP C . Although ScN2a cells do not exhibit any obvious cytopathic effect, they still display a variety of abnormalities (59,60). Thus, there is no a priori reason to assume that the flotation properties of PrP C in these cells must be identical to those of PrP C in uninfected cells. We therefore sought to determine the specific flotation characteristics of PrP C in ScN2a-MHM2 cells.
To this end we analyzed the distribution of Sarkosyl-soluble PrP in the flotation gradient. Fractions were diluted three times in TN, made 1% with Sarkosyl, and incubated on ice for 30 min, and then the insoluble PrP was removed by a 2-h spin at 45,000 rpm at 4°C in a TLA-45 rotor. Since most PrP Sc is insoluble in Sarkosyl (18), the supernatant is highly enriched in PrP C . As shown in Fig. 1F, the Sarkosyl-soluble PrP was also found in the lighter fractions of the gradient and was as buoyant as was PrP C in uninfected cells (Fig. 1A). To ascertain that the Sarkosyl-soluble PrP is indeed primarily PrP C , we incubated it with proteinase K (20 g/ml, 37°C, 1 h) and found that this PrP species is indeed completely digested under these conditions (not shown).
Exogenous PrP C and PrP Sc Added to TX-100 Cell Lysate Do Not Migrate with Buoyant Fractions-We wished to further ascertain that the observed buoyancy of the PrP isoforms in the flotation assay was not due to the spontaneous association of these proteins to TX-100-insoluble structures in vitro during the lysis but rather reflected their attachment to insoluble complexes prior to the detergent extraction. GPI proteins are indeed known for their affinity to cholesterol-and sphingolipidcontaining liposomes (in the absence of detergent) (61) as well as to the plasma membranes of whole cells. We therefore added exogenous PrP C and PrP Sc to the lysate of cells prior to the flotation and determined their buoyancy. For PrP C , we used an enriched fraction of recombinant MHM2-PrP C prepared from N2a-c3 cells by ion metal affinity chromatography. This chimeric PrP carries the mAb 3F4 epitope and can thus be differentiated from wild type MoPrP endogenous to N2a cells. Purified MHM2-PrP C was incubated with the TX-100 lysate of N2a cells on ice for 30 min, and the lysate was then subjected to the standard flotation assay. In contrast to endogenous PrP C , these exogeneously added PrP C molecules failed to float up the gradient ( Fig. 2A).
As a control for exogenous PrP Sc , we used purified prion rods obtained from Syrian hamsters. These PrP Sc -containing infectious aggregates are extensively delipidated during their purification (50). To determine whether purified prion rods attach to buoyant complexes in cell lysates, we incubated purified prion rods with the TX-100 lysate of ScN2a cells on ice for 30 min, and we then proceeded with the standard flotation procedure. The rods were detected in the fractions using the PrP Scspecific dot-immunoblotting procedure based on the sequential treatment of the membrane with proteinase K and GdnSCN (52). To differentiate exogenous SHaPrP Sc in the rods from mouse PrP Sc endogenous to ScN2a, we again used the mAb 3F4. As seen in Fig. 2B, these rods did not float up the gradient as did endogenous PrP Sc but instead sedimented to the bottom of the lysate fraction. An identical experiment, performed in the absence of exogenous PrP Sc , verified the specificity of the 3F4 antibody for SHaPrP in the dot blot immunoassay (Fig.  2C). These results show that the buoyancy of endogenous PrP C and PrP Sc in TX-100 cell extracts indeed stems from the physiological association of these proteins with cellular structures prior to the cell lysis.
In contrast to the results with exogenous PrP, when CTXB-POD was added to the lysate of N2a cells and incubated for 30 min (on ice) prior to the centrifugation, most of it did float to the same location as when incubated with whole cells (Fig. 2E). However, if the toxoid was resuspended in TX-100 lysis buffer without cell lysate, it remained at the bottom of the gradient (Fig. 2D). This shows that the toxoid can recognize its ligand within TX-100 cell lysates.
Saponin and NOG Reduce the Buoyancy of PrP C and PrP Sc in Flotation Gradients-To determine whether the buoyant properties of PrP C and PrP Sc in cold TX-100 depend on cholesterol, we added saponin (1%) to the TX-100 cell lysates and incubated them for 30 min on ice prior to the centrifugation (47). This protocol greatly reduced the buoyancy of both PrP C FIG. 2. Exogenous PrP added to cell lysate does not attach to rafts. MHM2-PrP C partially purified from N2a-c3 cells (A) or purified prion rods from SHa (B) were added to the TX-100 lysate of N2a or ScN2a cells, respectively, and incubated for 30 min at 4°C. The lysates were then subjected to the standard flotation assay. The PrP C content of the fractions was analyzed by Western blotting (A), whereas their content in PrP Sc was determined using the prion-specific dot blot immunoassay (B). Both assays were developed with the 3F4 mAb, which detects SHaPrP and MHM2-PrP but does not react with the wild type MoPrP present in N2a and ScN2a cells. In C, rods were omitted from the ScN2a lysates as a control for the specificity of the antibody. In E, CTXB-POD was added to the N2a lysate prior to the flotation assay, whereas in D the toxoid was subjected to the flotation procedure without the lysate.
from N2a cells and PrP Sc from ScN2a-MHM2 cells (Fig. 3, A  and B, respectively). Thus, the buoyancy of both PrP isoforms in TX-100 indeed depends on cholesterol. When NOG was substituted for TX-100 in the extraction of the cells, the buoyancy of both PrP C (Fig. 3C) and PrP Sc (Fig. 3D) was also markedly reduced.
Taken together, our data demonstrate that both PrP isoforms associate with complexes that satisfy our criteria for rafts. This in turn suggests that the transformation PrP C 3 PrP Sc takes place within rafts. Since most PrP Sc is intracellular, these findings also reveal the existence of intracellular rafts in scrapie-infected cells.
Buoyant PrP C and Thy-1 following TX-100 Extraction at 37°C-Since the primary subcellular localizations of PrP C and PrP Sc in neuroblastoma cells are different, we surmised that the rafts to which they are attached may be of different lipid or protein compositions, so that perhaps experimental conditions may be devised to differentiate between these putatively disparate PrP-containing complexes. In particular, we sought to determine whether PrP Sc and PrP C might behave differently in a flotation assay when lysed in TX-100 at temperatures higher than the "canonical" 4°C.
We first studied PrP C in uninfected cells. PrP C , like other GPI proteins, fails to sediment at high speed following lysis in TX-100 at 37°C (22,23). A simple sedimentation experiment performed with TX-100 lysates of N2a cells confirmed this property (Fig. 4A). It also confirmed that the addition of saponin to cold TX-100 lysates greatly increased the solubility of PrP C (Fig. 4A, lanes 5 and 6). We next sought to determine whether PrP C would still migrate up a Nycodenz density gradient if the TX-100 lysate was incubated at 37°C prior to the centrifugation. N2a cells were lysed on ice in cold TX-100 lysis buffer as before. The lysate was then further incubated at 37°C for 1 h, cooled back on ice, and then subjected to the standard flotation procedure. To our surprise, we found that the incubation at 37°C did not prevent the migration of most PrP C molecules toward the lighter fractions of the Nycodenz gradients (Fig. 4B). This property was also shared by cell surface GM1, another resident of rafts, as judged by the migration of CTXB-POD in the standard flotation assay following lysis of N2a cells at 37°C (Fig. 4C). In some experiments PrP C floated to even lighter fractions than when the lysis was performed at 4°C (not shown). The results of the flotation assay thus contrasted sharply with the increased apparent solubility of PrP C in TX-100 at 37°C, as judged by the sedimentation method (compare with Fig. 4A, lanes 1-4; Ref. 22).
To see whether other GPI proteins also remain insoluble and buoyant after lysis at 37°C, we examined the flotation properties of Thy-1 extracted from EL-4 T-lymphoma cells with TX-  3 and 4). In lanes 5 and 6, 1% saponin was added to the lysate prior to the incubation on ice. The lysates were then subjected to a high speed spin in a TLA-45 rotor (45,000 rpm, 1 h, 4°C), and the pellets (p) and supernatants (s) were collected and analyzed for PrP C by Western blotting. B and C, flotation assays. N2a cells were lysed in TX-100 at 4°C. The lysates were then incubated for 1 h at 37°C, cooled back on ice, and subjected to the standard flotation assay. In B, the PrP C content of the fractions was assessed by Western blotting using the R073 antiserum. In C, the cells were incubated with CTXB-POD (5 g/ml for 45 min at 4°C) prior to the lysis, and the content of the fractions in CTXB-POD was assayed by a direct dot blot. D and E, EL-4 cells were biotinylated with water-soluble NHS-X-biotin (500 g/ml, 40 min on ice) in buffer A. The cells were then rinsed in buffer A containing 50 mM NH 4 Cl and lysed in TX-100 on ice. One-half of the lysate was then warmed to 37°C for 1 h and then cooled back on ice, whereas the other half of the lysate remained on ice during the whole period. The lysates were subjected to the standard flotation. Thy-1 was immunoprecipitated from the fractions using the mAb G7. Biotinylated proteins in the immunoprecipitates were then detected in a Western blot developed with streptavidin conjugated with peroxidase. Although PrP C exhibited a decreased sedimentability following the lysis at 37°C (A, lanes 1-4), both PrP C and Thy-1 remained attached to low density complexes in these conditions (B and E, respectively). 100 at 37°C (Fig. 4, D and E). Because the mAb G7 does not recognize Thy-1 on Western blots, we resorted to cell surface biotinylation followed by immunoprecipitation to detect Thy-1. EL-4 cells were biotinylated using a membrane-impermeant NHS-X-biotin. They were then lysed in TX-100, and one-half of the lysate was warmed to 37°C for 1 h and then transferred back to ice (Fig. 4E), whereas the other half stayed on ice for the whole period (Fig. 4D). The lysates were subsequently subjected to a standard flotation procedure, and Thy-1 was immunoprecipitated from the gradient fractions using the mAb G7. The immunoprecipitates were then analyzed in Western blots developed with streptavidin-POD. As seen in Fig. 4E, Thy-1 indeed floated up the gradient even when the lysates had been subjected to an incubation at 37°C, a property that is thus not exclusive to PrP C in N2a cells.
PrP C and PrP Sc Localize to Complexes of Different Densities in 37°C Lysates of Some ScN2a Clones-Having shown that most PrP C still localizes to light fractions in density gradients even when the cells are lysed at 37°C, we turned our attention to the behavior of PrP Sc in lysates at this temperature. When the TX-100 lysate of ScN2a cells was incubated at 37°C for 1 h prior to the standard flotation assay, the protease-resistant PrP floated to fractions 3-7, similar to the distribution of PrP C in N2a cells (compare Figs. 5A and 4B). Thus, PrP Sc extracted from ScN2a cells stayed in detergent-insoluble buoyant rafts even following lysis at the higher temperature, therefore behaving like PrP C in this particular experiment. In some experiments (not shown), the peak of PrP Sc was slightly heavier than that of PrP C , but this difference was not very accentuated in these cells.
In contrast to these results with ScN2a cells, well separated PrP C and PrP Sc fractions were observed when the transfected ScN2a-MHM2 subclone was used in similar solubilization experiments. When TX-100 lysates of ScN2a-MHM2 cells were incubated at 37°C and then subjected to the standard flotation assay, protease-resistant PrP Sc migrated to the middle of the gradient (Fig. 5B). In contrast, Sarkosyl-soluble PrP (made up primarily of PrP C ; Ref. 18) migrated to the top of the gradient (Fig. 5C). Thus, these lysis conditions did generate fractions of clearly different buoyancy for PrP C and PrP Sc when applied to ScN2a-MHM2 cells.
We sought to determine whether the enhanced separation of the PrP C and PrP Sc peaks in ScN2a-MHM2 cells could be due to the overexpression of chimeric MHM2-PrP in these cells. To verify this, we used another subclone of ScN2a (ScN2a-c10), which expresses MHM2-PrP at yet higher levels than ScN2a-MHM2 cells (but driven by another expression vector). In contrast to the results with ScN2a-MHM2 cells, PrP Sc extracted from this clone at 37°C (Fig. 5D) reproducibly comigrated with PrP C (Fig. 5E) to the lighter fractions of the gradient. Thus, the separation of the PrP isoforms in the flotation gradient is not a result of chimeric PrP overexpression in these cells but rather seems to depend on clonal differences between the cells. We surmise that details in the lipid composition of cellular membranes (and in particular of rafts) may vary in these two transfected lines and determine disparate buoyant characteristics of PrP rafts.
PrP C and PrP Sc from SHa Brain Also Partially Separate following Lysis at 37°C-Having devised lysis conditions that permit the separation of PrP C and PrP Sc rafts in ScN2a-MHM2 cells, we moved on to determine the flotation characteristics of the PrP isoforms in the brain following lysis with TX-100 at 37°C. Since PrP is found in very diverse cells in the brain, in which lipid composition can be expected to vary widely, it was not obvious that the PrP isoforms would separate in these conditions. To explore the flotation characteristics of the PrP isoforms, we used the brain of a SHa clinically sick with Sc237 scrapie. To keep the lysis conditions as close as possible to the standard flotation assay, we added a small piece of this brain (ϳ1 mg of wet tissue) to the TX-100 lysate of N2a cells. The lysate was then warmed to 37°C for 1 h, returned to ice, and then subjected to the standard flotation assay. We then used Western blots (developed with 3F4 to detect SHaPrP above the MoPrP background) to determine the distribution of proteinase K-resistant PrP Sc and of Sarkosyl-soluble PrP in the gradient. A major portion of brain PrP C was buoyant (Fig. 5F) and separated well from the PrP Sc peak (Fig. 5G). Similar results were obtained when brain samples were lysed directly in TX-100 lysis buffer without mixing with lysates of cultured cells (not shown).
FIG. 5. In some cell lines and in the brain, PrP C and PrP Sc segregate to peaks of different densities following lysis in TX-100 at 37°C. ScN2a (A), ScN2a-MHM2 (B and C), and ScN2a-c10 (D and E) cells were lysed in TX-100 at 4°C. In F and G, a small piece of the brain of a Syrian hamster with scrapie was added to the lysate of N2a cells. The lysates were then incubated at 37°C for 1 h, cooled back on ice, and subjected to the standard flotation assay. One-half of the fraction was then subjected to proteinase K-catalyzed proteolysis (20 g/ml, 37°C, 1 h), and their PrP Sc content was then analyzed by Western blotting (A, B, D, and G). The other half was diluted three times with TN, made 1% with Sarkosyl, incubated on ice for 30 min, and subjected to a high speed spin (45,000 rpm, 1 h, 4°C) (C, E, and F). The supernatant was then concentrated by methanol precipitation, and its PrP content was analyzed by Western blotting. R073 was used to develop all immunoblots except in F and G, in which 3F4 was used to detect the SHaPrP above the MoPrP background endogenous to N2a cells.
Taken together, these results suggest that in some cells PrP Sc molecules distribute to a population of rafts that differ in both solubility and density following detergent extraction at 37°C. Alternatively, it is possible that the extractability of PrP Sc from its carrier rafts differs from that displayed by PrP C , perhaps due to a different anchoring mode. Further studies will be needed to discern between these and other possibilities. DISCUSSION The finding that most PrP Sc cofractionates with rafts in ScN2a clones has several important implications. First, since the metabolic precursor PrP C also localizes to rafts (22), it is probable that the transformation PrP C 3 PrP Sc occurs within this differentiated lipid environment. This conclusion is in line with our previous data that cholesterol-dependent pathways, as well as attachment of PrP C to rafts, are essential for the efficient conversion of PrP C into PrP Sc (22). How would these membrane domains facilitate the formation of PrP Sc ? It is possible that rafts favor the interaction of PrP C with existing PrP Sc "seed" molecules, maybe by concentrating PrP molecules within confined stretches of the plasma membrane or aligning them in a manner propitious for their interaction. Another possibility is that rafts contain some indispensable machinery engaged in the formation of PrP Sc , such as protein X (62) or other possible facilitators of PrP Sc formation. Analysis of rafts from ScN2a cells or rodent brains could thus conceivably enable us to discover such putative cofactors engaged in the formation of PrP Sc . It is possible that such rafts could yield a favorable environment to study the formation of PrP Sc in vitro (9). It will be interesting to see, for instance, whether proteaseresistant PrP can be formed in rafts isolated from ScN2a cells. Finally, it is also possible that rafts serve merely as vehicles that target PrP C to as yet unidentified subcellular sites hospitable to the conversion to PrP Sc .
The exact subcellular compartments in which PrP C is degraded and PrP Sc forms and the trafficking pathways leading to these sites have not been characterized. Several lines of evidence suggest that an internalization step may be involved in the formation of PrP Sc in ScN2a cells. First, when ScN2a cells are metabolically radiolabeled and chased at 18°C, even for long periods, radiolabeled PrP C does reach the plasma membrane (PM), but PrP Sc is not formed. However, if the cells are warmed to 37°C for 30 min and then transferred back to 18°C, protease-resistant PrP Sc now forms at the lower temperature (27). The existence of an 18°C block that is relieved by a short exposure at 37°C suggests the involvement of the endocytic pathway. Second, nascent PrP Sc , most of which is formed from PM PrP C , is trimmed in an acidic compartment shortly after its formation and thus has to be first endocytosed (27)(28)(29)(30). However, it is still possible that some or most PrP Sc is formed on the PM. Indeed, small amounts of PrP Sc seem to be present on the cell surface, 3 and they could function as seed molecules for the reaction. A central question is, thus: What internalization pathway, if any, is involved in the formation of PrP Sc ? Because PrP is attached to glycolipid rafts, the emerging nonclathrin endocytic pathways (63; reviewed in Ref. 64) are attractive candidates for such internalization mechanisms. Both caveolae and some GPI proteins are internalized through mechanisms that depend on actin and protein kinases (65)(66)(67). These may involve vesicles that contain various members of the general fusion machinery, such as NSF (41), as well as additional specific proteins, which may be homologous to the apical transport vesicle-specific annexin XIIIb (68).
In contrast to the localization of PrP C to rafts, Shyng et al. (46) have recently provided evidence that PrP C is enriched in clathrin-coated pits in N2a cells. Whether their finding presents a real contradiction to the localization of most PrP C in rafts is not clear. First, the data of Shyng et al. (46) pertain only to a small minority of cell surface PrP C molecules that localize to clathrin-coated pits. Second, association with coated pits and attachment to rafts may not be mutually exclusive. Whether clathrin-coated pits could contain some raftlike subdomains is not known. Based on the inhibition of PrP C endocytosis by hypertonic sucrose, these authors also suggested that PrP C internalizes via clathrin-coated pits. However, whether this drastic treatment specifically inhibits coated pit endocytosis without affecting other internalization pathways is unknown. In addition, because the experiments described here were designed to address only the steady state solubilization properties of PrP, we would have been unable to detect possible transient events such as, for instance, a temporary removal of PrP from rafts. For instance, Rijnboutt et al. (69) have recently shown that although most (GPI-anchored) folate receptor is detergentinsoluble in KB cells, it is the soluble minority that is endocytosed and thus presumably participated in the potocytosis of folate. Clearly, more work is needed to decipher the trafficking pathways used by PM PrP C .
That PrP C and PrP Sc rafts could be separated from each other in some neuroblastoma lines as well as in the brain has both theoretical and practical implications. There are at least two potential mechanisms that could lead to such a separation. One possibility is that most PrP Sc molecules indeed reside in rafts of different lipid and protein compositions that are denser, after TX-100 extraction, than those that carry PrP C . This is certainly plausible, since the primary localization of these proteins is entirely different. Whereas PrP C is found mainly on the cell surface, PrP Sc is primarily intracellular and accumulates mainly (but not only) in secondary lysosomes of infected cells (20,31). Our 37°C extraction procedure paves the way for the isolation and characterization of both types of rafts.
It is also possible that the reason for the segregation of PrP C and PrP Sc in different fractions of the flotation gradient is that at 37°C these proteins are extracted from the membranes with different efficiencies. The mode of association of PrP Sc with membranes has never been elucidated. Although our results suggest that most cellular PrP Sc is attached to saponin-sensitive buoyant rafts, probably through its GPI moiety, the details of this association are unknown. It is possible, for instance, that PrP Sc exists as aggregates that attach to cellular membranes only through a small fraction of the GPI moieties present in the aggregate. Such aggregates would be more tenuously attached to membranes and therefore extracted from them more easily, thus becoming less buoyant after extraction at 37°C. Additional work will be needed to determine the mode of attachment of PrP Sc to membranes. Finally, because the Sc237 strain of scrapie used to inoculate this Syrian hamster induces sizable amounts of extacellular PrP Sc plaques, it is possible that some of the PrP Sc detected in Fig. 5G is a priori extracellular and not bound to membranes.
What is the reason for the apparent contradiction between results obtained with the sedimentation and the flotation assays when lysates are incubated at 37°C? It is possible that the insoluble complexes generated by the lysis of cells in TX-100 at 37°C may be smaller than those produced at 4°C but may retain a similar density. This would decrease their sedimentability (Fig. 4A) while maintaining their buoyancy in Nycodenz gradients (Fig. 4, B and E). Further studies will be needed to clarify this point.
Since most PrP Sc is intracellular, the finding that it is attached to rafts reveals the existence of TX-100-insoluble rafts in intracellular organelles, probably in late compartments of the endocytic pathway. Whether this is a pathological property of prion-infected cells or a general attribute of many cell types remains to be determined. Various investigators have shown that the internalization pathway of GPI proteins intersects that of proteins that are internalized via clathrin-coated pits. Thus, rafts could find their way to the late endocytic pathway. That lysosomes contain cholesterol was shown previously using filipin binding techniques.
In summary, the results presented in this report support a model in which the various aspects of PrP metabolism take place within cholesterol and glycolipid rafts. The exact subcellular compartments involved and their molecular components remain to be characterized.