A Transmembrane Form of the Prion Protein Is Localized in the Golgi Apparatus of Neurons*

CtmPrP is a transmembrane version of the prion protein that has been proposed to be a neurotoxic intermediate underlying prion-induced pathogenesis. In previous studies, we found that PrP molecules carrying mutations in the N-terminal signal peptide (L9R) and the transmembrane domain (3AV) were synthesized exclusively in the CtmPrP form in transfected cell lines. To characterize the properties of CtmPrP in a neuronal setting, we have utilized cerebellar granule neurons cultured from Tg(L9R–3AV) mice that developed a fatal neurodegenerative illness. We found that about half of the L9R-3AV PrP synthesized in these neurons represents CtmPrP, with the rest being SecPrP, the glycolipid anchored form that does not span the membrane. Both forms contained an uncleaved signal peptide, and they are differentially glycosylated. SecPrP was localized on the surface of neuronal processes. Most surprisingly, CtmPrP was concentrated in the Golgi apparatus, rather in the endoplasmic reticulum as it is in transfected cell lines. Our study is the first to analyze the properties of CtmPrP in a neuronal context, and our results suggest new hypotheses about how this form may exert its neurotoxic effects.

Prion diseases are fatal neurological disorders of humans and animals characterized by ataxia and neuronal degeneration (1). Unlike other neurodegenerative diseases, they can have an infectious as well as a sporadic or familial origin. Most cases are associated with the presence of PrP Sc , 1 a conformationally altered isoform of PrP C , a cell surface glycoprotein of uncertain function that is expressed primarily in neurons of the brain and spinal cord. PrP C is monomeric, protease-sensitive, and rich in ␣-helical structure. In contrast, PrP Sc is aggregated, protease-resistant, and rich in ␤-sheets. There is considerable evidence that PrP Sc is an infectious protein and that conversion of PrP C into PrP Sc is the central event in the propagation of prions, the infectious agents in these diseases (2,3).
Although it is clear that PrP Sc accumulates in the brain during most prion diseases, there is uncertainty about the mechanisms responsible for neuronal death. Several lines of evidence suggest that PrP Sc is not toxic when it is presented to neurons externally (4,5) and that conversion of PrP C to PrP Sc within neurons may generate toxic intermediates or by-products. However, there is debate about the identity of these neurotoxic species (6). Several alternative forms of PrP, distinct from both PrP C and PrP Sc , have been proposed as key pathogenic entities based on experiments in cell culture and transgenic mice. These include transmembrane PrP (7,8), cytosolic PrP (9), protease-sensitive PrP Sc (10), and PG14 spon PrP (11).
This report focuses on transmembrane PrP. PrP is unusual because it can exist in several different topological forms that are generated during synthesis in the endoplasmic reticulum (ER). Most molecules assume the form designated Sec PrP, in which the polypeptide chain has been fully translocated into the ER lumen with its C terminus attached to the lipid bilayer by a glycosylphosphatidylinositol (GPI) anchor (12,13). Under certain circumstances, however, the protein can adopt either of two opposite transmembrane orientations (designated Ctm PrP and Ntm PrP) in which a highly conserved stretch of hydrophobic amino acids in the middle of the sequence integrates into the membrane (7, 14 -16). In Ctm PrP, the C terminus of the protein lies in the ER lumen, whereas in Ntm PrP, the N terminus is luminal. Mutations within and adjacent to the transmembrane domain that enhance its hydrophobicity increase the proportion of Ctm PrP (from Ͻ10% to 20 -30% of the total PrP) (7,8,17). Some of these mutations (A117V and P105L) are associated with familial prion diseases, whereas others (3AV) are not seen in human patients.
Because the brains of patients with the A117V mutation do not contain conventional PrP27-30 (the protease-resistant core of PrP Sc ) (18), it was proposed that Ctm PrP, rather than PrP Sc , is the proximal cause of neurodegeneration in this these and possibly other cases of prion disease (7,8). Consistent with this hypothesis, transgenic mice expressing PrP with Ctm PrP-favoring mutations develop a spontaneous neurodegenerative illness that is similar to scrapie, but without detectable PrP Sc (7). Based on these and other results, it was suggested that Ctm PrP is a key neurotoxic intermediate in both familial and infectious prion diseases and that the amount of this form can be increased directly by pathogenic mutations, or indirectly by accumulation of PrP Sc (8). However, uncertainties remain about the role of Ctm PrP in prion diseases because of recent reports that Ctm PrP levels do not change significantly during scrapie infection (19) or as a result of most pathogenic mutations (17).
Our laboratory has been interested in investigating further the role of Ctm PrP in prion diseases. To this end, we have identified mutations that cause PrP to be synthesized exclu-sively with the Ctm PrP topology, thus facilitating analysis of this form in the absence of the other topological variants. We demonstrated that Ctm PrP has an uncleaved, N-terminal signal peptide (16) and that substitution of charged residues in the hydrophobic core of the signal sequence strongly favors synthesis of Ctm PrP (19). Combining one of these mutations (L9R) with a previously studied mutation in the transmembrane domain (3AV) resulted in a protein that was expressed entirely as Ctm PrP after in vitro translation and transfection of cultured cells (16). The presence of the L9R-3AV mutation caused a striking change in the subcellular distribution of PrP; the protein was no longer present on the plasma membrane like wildtype PrP but was instead concentrated in the ER of transfected CHO, BHK, and N2a cells (16). This result raised the possibility that the ER could be a potential site for the neurotoxic action of Ctm PrP.
To extend our studies to an in vivo setting, we have recently created transgenic mice expressing L9R-3AV PrP. 2 We found that these mice develop a fatal, ataxic neurological disorder accompanied by extensive degeneration of cerebellar granule neurons and hippocampal pyramidal cells. To our surprise, we found that this phenotype was strongly dependent on coexpression of endogenous, wild-type PrP. Here we have taken advantage of the availability of Tg(L9R-3AV) mice to study the cell biology and metabolism of Ctm PrP in neurons. We find that, in contrast to transfected cell lines, cultured neurons expressing transgenically encoded L9R-3AV localize Ctm PrP to the Golgi apparatus rather than to the ER. This observation dramatically changes our view of the possible cellular pathways that may be responsible for Ctm PrP-induced neurotoxicity.

EXPERIMENTAL PROCEDURES
Transgenic Mice-Engineering of Tg(L9R-3AV) mice is described elsewhere. 2 The experiments reported here were carried out using mice from the B line, but similar results were obtained using mice from the C line (data not shown). Mice were hemizygous for the transgene array, and were maintained on both Prn-p ϩ/ϩ and Prn-p 0/0 backgrounds. Tg(WT-E1)/Prn-p 0/0 and Tg(PG14-A2)/Prn-p 0/0 mice have been described previously (20).
Cerebellar Granule Cell Culture-Primary cultures from 5-day-old pups were performed as described previously (21). Dissociated cells were resuspended in CGN medium (basal medium Eagle's, 10% dialyzed fetal bovine serum, 25 mM KCl, 2 mM glutamine, 50 g/ml gentamycin) and plated at a density of 500,000 cells/cm 2 in polylysinecoated plastic plates or 8-well glass chamber slides. Cells were used after 4 -5 days in culture. Based on staining with cell type-specific marker proteins, these cultures typically contained Ͼ95% granule neurons, with the remainder of cells being fibroblasts and astrocytes.
Metabolic Labeling of Cultured Cells and Immunoprecipitation of PrP-Cerebellar granule cells were labeled with 100 -500 Ci/ml of 35 S-Promix (Amersham Biosciences) in CGN medium lacking methionine, cysteine, and bovine serum and containing B27 vitamin supplement (Invitrogen). In some experiments, cells were chased in complete medium lacking radioactive methionine and cysteine, with or without PIPLC (1 unit/ml; purified from Bacillus thuringiensis as described by Shyng et al. (24)). Pulse-chase labeling of CHO cells was carried out as described previously (25).
Cells were lysed in 0.5% SDS, 50 mM Tris-HCl (pH 7.5), and immunoprecipitation of PrP was carried out as described previously (25). To treat PrP with glycosidases, protein was eluted from protein A-Sepharose beads by heating at 95°C for 5 min in 0.1% SDS, 50 mM Tris-HCl (pH 6.7). The eluate was incubated for 1 h at 37°C with endo H, neuraminidase, or PNGase F (all from New England Biolabs, Beverly, MA) according to the manufacturer's directions. Immunoprecipitated PrP was analyzed by SDS-PAGE and autoradiography.
PrP Membrane Topology Assay-Cerebellar granule cells were met-abolically labeled for 4 -6 h as described above. Cells were then scraped into PBS, spun at 2,000 ϫ g for 5 min, and resuspended in ice-cold homogenization buffer (250 mM sucrose, 5 mM KCl, 5 mM MgCl 2 , 50 mM Tris-HCl (pH 7.5)). Cells were lysed by 12 passages through silastic tubing (0.3-mm inner diameter) connecting two syringes with 27-gauge needles, and nuclei were removed by centrifugation at 5,000 ϫ g for 10 min. Aliquots of the postnuclear supernatant were diluted into 50 mM Tris-HCl (pH 7.5) and incubated for 60 min at 4°C with 250 g/ml PK in the presence or absence of 0.5% Triton X-100. Digestion was terminated by addition of phenylmethylsulfonyl fluoride (5 mM final concentration), and PrP was then immunoprecipitated and deglycosylated by treatment with PNGase F. Western Blotting-Brain tissue was homogenized using a Teflon pestle in 10 volumes of PBS containing protease inhibitors (phenylmethylsulfonyl fluoride, 20 g/ml; leupeptin and pepstatin, 10 g/ml). Homogenates were clarified by centrifugation at 2,000 ϫ g for 5 min. Cultured neurons were lysed in 0.5% SDS, 50 mM Tris-HCl (pH 7.5), and the lysates were heated at 95°C for 10 min. Protein was quantified using a BCA Assay (Pierce). Samples were analyzed by SDS-PAGE followed by immunoblotting with anti-PrP antibodies. In some cases, samples were treated with endo H or PNGase F, and proteins were recovered by methanol precipitation prior to SDS-PAGE.
Immunocytochemistry-Cerebellar granules cells were grown in 8-well chamber slides. For surface staining, cells were transferred to B27 medium (Dulbecco's modified Eagle's medium containing B27 vitamin supplement) and stained with anti-PrP antibodies (1:500 dilution) for 10 min at 37°C. After rinsing in PBS, cells were fixed for 10 min at room temperature in 4% paraformaldehyde, 5% sucrose in PBS, blocked for 10 min in PBS, 2% goat serum, and stained with Alexa 488-conjugated goat anti-mouse IgG (Molecular Probes, Eugene, OR). Cells were mounted in 50% glycerol/PBS. In some cases, cells were incubated at 37°C for 2 h with PIPLC (1 unit/ml) in B27 medium prior to fixation.
For visualization of intracellular PrP, cells were first fixed for 10 min at 4°C in 4% paraformaldehyde, 5% sucrose in PBS and were then incubated for 10 min at 4°C, either in 0.2% Triton X-100 in PBS (to permeabilize all membranes) or in 10 g/ml digitonin in digitonin buffer (10% sucrose, 100 mM KOAc, 2.5 mM MgCl 2 , 1 mM EDTA, 10 mM HEPES-HCl, pH 7.0) (to selectively permeabilize the plasma membrane) (26 -28). After treatment with blocking solution, cells were stained with antibodies against PrP, giantin (Covance, Berkeley, CA), or PDI (StressGen, Victoria, British Columbia, Canada). Cells were then incubated with secondary antibodies (Alexa 488-conjugated goat anti-mouse IgG and Alexa 546-conjugated goat anti-rabbit IgG). In some experiments, cells were treated for 4 h with 10 g/ml brefeldin A (Sigma) prior to staining. Cells were viewed with a Zeiss LSM 510 confocal microscope equipped with an Axiovert 200 laser scanning system.

RESULTS
The L9R-3AV mutation is illustrated in Fig. 1. We have generated and characterized two lines of transgenic mice that express PrP carrying this mutation. 2 These Tg(L9R-3AV) mice FIG. 1. Schematic structure of mouse PrP. SP, signal peptide; TM, transmembrane; GPI, glycosylphosphatidylinositol anchor addition sequence. The numbers above the drawing are the amino acid positions that define the ends of each domain. The horizontal lines below the drawing indicate the locations of epitopes recognized by antibodies used in this study, with the amino acid positions of each epitope given in parentheses. The amino acid sequence surrounding the L9R and 3AV mutations are shown above the signal peptide and transmembrane domains, respectively. 3AV is the designation for the triple mutation A112V/A114V/A117V. The two underlined methionine residues at positions 108 and 111 were introduced to create an epitope for 3F4 antibody, which allows discrimination of transgenically encoded and endogenous PrP. develop a spontaneous neurological disorder characterized by ataxia and by loss of cerebellar granule cells and hippocampal pyramidal neurons. We utilized primary cultures of granule cells prepared from the cerebella of Tg(L9R-3AV) mice at postnatal day 5 to characterize the cell biology and metabolism of the mutant PrP in a neuronal setting.
Cultured Neurons from Tg(L9R-3AV) Mice Produce Both Ctm PrP and Sec PrP-We first used a protease protection assay to analyze the topology of L9R-3AV PrP in microsomes from granule neurons. Microsomes in a postnuclear supernatant prepared from [ 35 S]methionine-labeled neurons were subjected to digestion with proteinase K (PK), and PrP was then immunoprecipitated with 3F4 antibody after enzymatic deglycosylation (Fig. 2). We observed the following two protease-protected forms of PrP in approximately equal amounts (Fig. 2, lane 6): a 27-kDa band representing Sec PrP molecules that were fully protected from digestion because of their localization in the microsome lumen, and a 19-kDa fragment representing the luminal and transmembrane domains of Ctm PrP with the exposed cytoplasmic domain removed. We did not detect a 15-kDa fragment indicative of Ntm PrP. As expected, inclusion of Triton X-100 during protease treatment in order to disrupt microsomal membranes resulted in complete digestion of PrP (Fig. 2, lane 7). When the same analysis was carried out on microsomes from Tg(WT) neurons, only a fully protected 25-kDa band corresponding to Sec PrP was observed (Fig. 2, lane 2). Thus, granule neurons produce approximately equal proportions of Sec PrP and Ctm PrP from PrP carrying the L9R-3AV mutation. For both wild-type and mutant proteins, there was a decrease in the total amount of protected PrP after PK treatment, an effect that we attribute to the presence of some inside-out and damaged microsomes and not to the presence of cytoplasmic PrP.
Biosynthesis and Turnover of Mutant PrP-We used pulsechase labeling with [ 35 S]methionine to analyze the biosynthetic maturation and degradation of L9R-3AV PrP in cerebellar granule neurons (Fig. 3A). A portion of each cell lysate was incubated with endoglycosidase H (endo H) prior to immunoprecipitation of PrP in order to test whether the N-linked glycans had matured to an endo H-resistant state, a step that occurs as proteins transit the mid-Golgi apparatus. At the end of the 20-min pulse-labeling period, a single species of PrP with a molecular mass of 32 kDa was observed. This form was shifted to 27 kDa, the size of unglycosylated PrP, by treatment with endo H, indicating the presence of immature glycan chains. The 32-kDa species was converted into two glycoforms of 32 and 35 kDa over the next 20 min of chase. Both of these mature forms contained glycans that were endo H-resistant.
The 32-and 35-kDa glycoforms both decayed with a half-life of ϳ2.5 h, similar to the half-life of wild-type PrP in granule neurons (25). These data suggest the existence of a single metabolic pool of L9R-3AV PrP in neurons, implying that the Ctm PrP and Sec PrP in these cells do not have markedly different kinetic properties.
For comparison, we also analyzed the biosynthesis of L9R-3AV PrP in transiently transfected CHO cells (Fig. 3B). In CHO cells, the mutant protein was initially synthesized as two species of 32 and 27 kDa. The 27-kDa form was not glycosylated, and the 32-kDa form was core-glycosylated, because it is sensitive to endo H. However, the 32-kDa form in CHO cells remained endo H-sensitive throughout the entire chase period, unlike the case for the glycosylated forms of L9R-3AV in granule neurons. These results indicate that, whereas the mutant protein is retained in a pre-Golgi compartment in CHO cells, it transits beyond the mid-Golgi in granule neurons. Despite these differences in glycosylation and cellular trafficking, L9R-3AV PrP in CHO cells decayed with a half-life (ϳ2.5 h) that was similar to the one observed for this mutant in neurons (see above) and for wild-type PrP in CHO cells (25).
L9R-3AV PrP Retains an Uncleaved Signal Peptide-We demonstrated previously that Ctm PrP synthesized by in vitro translation or by expression in CHO cells contains an uncleaved Nterminal signal peptide (16). To determine whether L9R-3AV PrP in cerebellar granule neurons also has the same feature, we immunoprecipitated PrP from [ 35 S]methionine-labeled cultures using anti-SP, an antibody that specifically recognizes PrP molecules containing an intact signal peptide (19). Parallel samples were immunoprecipitated with 3F4 or 8H4 antibodies, which recognize PrP molecules regardless of the presence of the signal peptide (see Fig. 1 for the location of antibody epitopes). As an additional way of discriminating molecules with and without the signal peptide, proteins were enzymatically deglycosylated to permit detection of the small size difference between the two species (27 and 25 kDa for signal peptide bearing, and signal peptide-cleaved forms, respectively).
When cultures from Tg(L9R-3AV)/Prn-p ϩ/ϩ mice were analyzed, two bands (25 and 27 kDa) of approximately equal intensity could be resolved after immunoprecipitation with 8H4 antibody (Fig. 4, lane 1). However, only the larger band was recognized by 3F4 and anti-SP antibodies (Fig. 4, lanes 2 and 3). This result demonstrates that the 27-kDa form represents transgenically encoded PrP that contains the 3F4 epitope and has an intact signal peptide, whereas the 25-kDa form represents endogenous mouse PrP that lacks the 3F4 epitope and has a cleaved signal peptide. Because Ͼ90% of the PrP recog-  4 and 8). The protease-protected forms of Sec PrP and Ctm PrP are indicated by arrows to the right of the gels. The protected bands appear as doublets in some samples due to nibbling of the polypeptide chain by PK. Molecular size markers are in kilodaltons. nized by 3F4 migrated at 27 kDa (lane 2), we infer that the majority of the L9R-3AV PrP in granule neurons, comprising both Sec PrP and Ctm PrP, retained an intact signal peptide. This conclusion was confirmed when we analyzed neurons from Tg(L9R-3AV)/Prn-p 0/0 mice. In this case, ϳ90% of the PrP immunoprecipitated by 3F4 migrated at 27 kDa, with the rest migrating at 25 kDa (Fig. 4, lane 5). A similar ratio of the two forms was observed with 8H4, the expected result because the neurons contain no endogenous PrP that would be recognized by this antibody (Fig. 4, lane 4). Again, only the 27-kDa form was immunoprecipitated by anti-SP (Fig. 4, lane 6). The amount of 25-kDa PrP in Tg(L9R-3AV) neurons varied somewhat in different experiments, possibly due to artifactual proteolysis after cell lysis, but it never exceeded ϳ20% of the total. As anticipated, Tg(WT) neurons produced only a 25-kDa form of PrP that was immunoprecipitated by 8H4 and 3F4 but not by anti-SP (Fig. 4, lanes 7-9).
Further evidence that Sec PrP contains an uncleaved signal peptide was provided by immunoprecipitation of PrP from PKtreated microsomes using anti-SP antibody (Fig. 2). The fully protected 27-kDa species in Tg(L9R-3AV) neurons reacted with anti-SP (Fig. 2, lane 8), whereas the corresponding 25-kDa band from Tg(WT) neurons did not (lane 4), implying that Sec PrP in the former cells contains an uncleaved signal peptide. As expected, the 19-kDa Ctm PrP fragment in Tg(L9R-3AV) neurons was not recognized by anti-SP (compare Fig. 2, lanes 6  and 8), because the N terminus of Ctm PrP lies on the cytoplasmic side of the membrane, and so its uncleaved signal peptide would be accessible to protease digestion.

Sec
PrP but Not Ctm PrP Is Present on the Neuronal Surface-In a previous study, we found by immunofluorescence staining that L9R-3AV PrP was absent from the surface of transfected CHO and BHK cells and was completely retained in the ER (16). This result was in accord with the observed endo H sensitivity of the protein in these cells. Our observation (Fig. 3) that L9R-3AV PrP in granule neurons matures to an endo H-resistant form suggested that localization of the mutant protein in these cells was likely to be different from CHO and BHK cells. We therefore analyzed the distribution of L9R-3AV PrP in granule neurons by using immunofluorescence microscopy.
In our first set of experiments, neurons from Tg(L9R-3AV)/ Prn-p 0/0 mice were stained in the living state without permeabilization to selectively recognize PrP on the cell surface. As shown in Fig. 5, A-C, we observed positive staining with antibodies directed against three different epitopes distributed along the length of the PrP molecule ( Fig. 1), including 8B4 (residues 34 -52), 3F4 (residues 108 -111), and 8H4 (residues 147-200). Staining was distributed along neuronal processes, which formed an extensive network throughout the culture. Because the epitopes for antibodies 8B4 and 3F4 are extracellular in Sec PrP but not in Ctm PrP, the accessibility of these epitopes to externally applied antibodies implied that at least some of the L9R-3AV PrP on the surface of the neurons must be in the Sec PrP form. Control experiments demonstrated that endogenous, wild-type PrP on neurons from Prn-p ϩ/0 mice stained with 8B4 and 8H4, but not with 3F4 (because endogenous PrP lacks the 3F4 epitope) (Fig. 5, D-F), and that none of the antibodies stained neurons from Prn-p 0/0 mice (not shown). In addition, neurons were not stained with an antibody directed against a cytoplasmic epitope of the Golgi protein, giantin, confirming the integrity of the surface membrane (not shown).
To further demonstrate the presence of Sec PrP on the cell surface, we treated living neurons with PIPLC, a bacterial enzyme which cleaves the C-terminal GPI anchor. PIPLC is predicted to release Sec PrP, but not Ctm PrP, from the cell surface, because the latter has a transmembrane segment that would maintain attachment to the plasma membrane even after the GPI anchor was cleaved. We found that PIPLC treatment eliminated virtually all surface staining for PrP, assayed by using 8H4 antibody, implying that most of the protein on the surface represented Sec PrP (Fig. 6, A and D). Any residual Ctm PrP should have reacted with 8H4, because the epitope recognized by this antibody is extracellular. As expected, wildtype PrP was also completely released by PIPLC from the surface of nontransgenic, Prn-p ϩ/ϩ neurons (Fig. 6, B and E).
As an additional control, we demonstrated that neurons from Tg(PG14) mice did display residual PrP staining after PIPLC treatment, consistent with the partial resistance of the PG14 protein to GPI anchor cleavage (20) (Fig. 6, C and F). This control rules out the possibility that protease contamination in the PIPLC preparation degraded surface PrP. Taken together, our results indicate that there is little Ctm PrP on the cell surface and that most of the protein in this location represents Sec PrP.
Ctm PrP Is Localized to the Golgi Apparatus-Because about half of the L9R-3AV PrP present in granule neurons represents Ctm PrP (Fig. 2), and because little of this form is present on the cell surface (Fig. 6), our results suggested that Ctm PrP was likely to be localized in an intracellular compartment. To visualize the intracellular distribution of the mutant protein, we stained neurons that had been permeabilized with Triton X-100. In control experiments with Prn-p ϩ/0 neurons, wild-type PrP was found to be localized primarily along neuronal processes (Fig. 7A), corresponding to the surface PrP visualized on these cells by staining without permeabilization (Fig. 5, D and  F). The distribution of PrP in neurons from Tg(L9R-3AV) mice, on both Prn-p 0/0 and Prn-p ϩ/ϩ backgrounds, was markedly different. The mutant protein was concentrated in discrete, perinuclear structures in the soma that colocalized with the Golgi marker protein, giantin (Fig. 7, D-F). Staining of neuronal processes was less prominent in these permeabilized neurons than in unpermeabilized ones (compare Figs. 5A and 7D), which reflects partial extraction of plasma membrane PrP as well as enhanced reactivity of cytoplasmic epitopes of Golgiresident PrP after Triton X-100 treatment. 3 To confirm the Golgi localization of L9R-3AV PrP, we treated neurons with brefeldin A, which causes fusion of the ER and Golgi compartments. This treatment resulted in a redistribution of both PrP and giantin to a more diffuse pattern, consistent with the conclusion that the two proteins reside in the same structures (Fig. 7, G-I). We conclude from these results, and from the results with unpermeabilized cells (Figs. 5 and 6), that L9R-3AV PrP in neurons is present in the Golgi apparatus as well as on the plasma membrane.
We performed a series of experiments to directly probe the membrane topology of L9R-3AV PrP in situ by immunofluorescence staining. Treatment of cells with digitonin, a cholesterolbinding detergent, permeabilizes only the plasma membrane but not internal membranes such as those of the Golgi and ER, because the latter have a lower content of cholesterol (26 -28). In contrast, Triton X-100 permeabilizes all membranes. We found that antibody 8B4, which is directed against an N-terminal epitope (Fig. 1, 34 -52), produced prominent Golgi staining in neurons permeabilized with either digitonin or Triton (Fig. 8, A and E). A similar result was seen with another N-terminally directed antibody, P45-66, which recognizes residues 45-66 within the octapeptide repeat region (not shown). In contrast, antibody 8H4, which reacts with a C-terminal region (Fig. 1, 147-200), stained PrP in the Golgi only after Triton treatment (Fig. 8, F-H); in digitonin-treated neurons, staining was visible exclusively on neuronal processes, representing cell surface PrP molecules (Fig. 8, B-D). The observation that N-terminal epitopes of L9R-3AV PrP are accessible without permeabilization of Golgi membranes, whereas C-ter-  minal epitopes are not, indicates that mutant protein in the Golgi has the Ctm PrP topology. When we performed the same experiment on neurons from nontransgenic Prn-p ϩ/0 mice, neurite staining was observed with both 8B4 and 8H4, regardless of the method of permeabilization (Fig. 8, I and K). This result reflects the fact that most wild-type PrP is found on the surface of neuronal processes, where it is present in the form of Sec PrP.
Two control experiments demonstrated the selectivity of the permeabilization procedures. Protein-disulfide isomerase (PDI), a luminal ER protein, was accessible to staining in Triton-treated but not digitonin-treated cultures, proving that digitonin did not permeabilize internal membranes. This result was most easily appreciated in the small number of fibroblasts present in the cultures, which have higher PDI levels than the neurons (Fig. 8, J and L). In contrast, an antibody to a cytoplasmic epitope of giantin stained the Golgi apparatus of neurons treated with either digitonin or Triton, demonstrating permeabilization of the plasma membrane by both detergents (Fig. 8, C and G).
Taken together, our immunofluorescence localization studies of intact and permeabilized neurons indicate that the Ctm PrP form L9R-3AV PrP is concentrated in the Golgi apparatus, whereas the Sec PrP form is present on the surface of neuronal processes like wild-type PrP.
Sec PrP and Ctm PrP Are Differentially Glycosylated-We wondered whether the 32-and 35-kDa glycoforms of L9R-3AV PrP seen in pulse-labeling experiments (Fig. 3) might correspond to the two different topological forms of the protein, one being Sec PrP and the other, Ctm PrP. To answer this question, we treated [ 35 S]methionine-labeled neurons with PIPLC in order to release Sec PrP from the cell surface. We found that only the 32-kDa band was released into the medium by PIPLC, implying that this glycoform represents Sec PrP (Fig. 9A, upper panels). At least some of the 35-kDa glycoform that remained cell-associated after PIPLC treatment must therefore represent Ctm PrP. However, our results do not rule out the possibil-ity that the 35-kDa band might also include molecules of Sec PrP that reside in intracellular compartments not accessible to externally applied PIPLC. In control experiments, wild-type PrP was completely released into the medium by PIPLC treatment, consistent with localization of virtually all of the protein on the cell surface (Fig. 9A, lower panels). In addition, both the 32-and 35-kDa glycoforms of L9R-3AV PrP underwent a characteristic decrease in gel mobility when detergent lysates were treated with PIPLC (Fig. 9B). Thus, lack of release of the 35-kDa form was due to its residence in an intracellular compartment, rather than to the presence of a phospholipaseresistant anchor structure.
To further analyze the glycoform profile of L9R-3AV in neurons, we tested the sensitivity of the oligosaccharide chains to neuraminidase, which cleaves sialic acid residues that are added in the trans cisterna of the Golgi. We found that both the 32-and 35-kDa forms of mutant PrP were shifted by neuraminidase treatment, although the shift was larger for the latter form (Fig. 9C, lanes 2 and 3). As observed previously, neither form was sensitive to endo H (Fig. 9C, lane 4). In contrast, both forms were shifted to 27 kDa, the size of unglycosylated PrP, by treatment with PNGase F, which cleaves all N-linked glycans regardless of structure (Fig. 9C, lane 5). These results indicate that both the 32-and 35-kDa forms of L9R-3AV PrP transit the trans cisterna of the Golgi in neurons but that the latter form is more heavily modified by sialic acid residues in that compartment. Consistent with an unusual oligosaccharide composition of the 35-kDa species is the fact that it has a higher M r than mature, doubly glycosylated, wild-type PrP from cultured neurons, which migrates at 32 kDa (Fig. 9C, compare lanes 1  and 2).
Because our previous experiments were carried out on isolated cerebellar granule neurons, we also analyzed the glycosylation pattern of L9R-3AV PrP in whole brain by Western blotting. We found that the mutant protein in brain, as in cultured neurons, was composed of 32-and 35-kDa glycoforms, both of which were endo H-resistant and PNGase-sensitive ( Fig. 9D, lanes 1-3). The proportion of the 35-kDa glycoform appeared to be lower in brain than in cultured neurons, but it was difficult to draw quantitative conclusions concerning the ratio of the two forms because of the weak reactivity of L9R-3AV PrP on Western blots. 2 In comparison, wild-type PrP in transgenic brain migrated as a major species of 32 kDa (Fig.  9D, lane 4 -6). These results suggest that the post-translational processing of L9R-3AV PrP by isolated cerebellar granule neurons may be qualitatively similar to its processing by neurons in situ within transgenic brain. DISCUSSION We report here our analysis of PrP synthesized in cerebellar granule neurons cultured from Tg(L9R-3AV) mice. The PrP molecules in these cells carry an L9R-3AV mutation that strongly favors synthesis of Ctm PrP (16). Tg(L9R-3AV) mice develop a fatal neurological illness accompanied by extensive degeneration of several populations of neurons in the brain, including cerebellar granule cells. 2 Our results demonstrate that, in cerebellar granule neurons, Ctm PrP is concentrated in the Golgi apparatus, rather than in the ER as it is in transfected cells. Our study is the first to analyze the synthesis and subcellular localization of Ctm PrP in a neuronal context, and our results suggest new hypotheses about the mechanism by which this form may exert its neurotoxic effects.
Regulation of PrP Topology in Neurons-We found that granule neurons from Tg(L9R-3AV) mice express about 50% of the mutant protein as Ctm PrP and about 50% as Sec PrP. In contrast, we demonstrated previously that L9R-3AV PrP is synthesized almost exclusively with the Ctm PrP topology in transiently transfected CHO, BHK, and N2a cells (16). This difference between granule neurons and transfected cells may reflect the action of important regulatory mechanisms. The membrane topology of PrP is determined by sequence determinants in the polypeptide chain as well as by trans-acting factors that operate at the translocon during the translocation process (15, 29 -31). Ctm PrP is the default topology of PrP synthesized in translocation reactions reconstituted from minimal components (Sec61p complex and SRP receptor) (31). Inclusion of an additional protein complex known as TRAP (translocon-associated protein) allows synthesis of Sec PrP as well as Ctm PrP in these reactions (32). One plausible explanation for our results is that the amount or activity of the TRAP complex, or of some other accessory factors, is different in granule neurons compared with transformed cell lines. It is also possible that, in conjunction with decreased capability for synthesis of Ctm PrP, neurons possess mechanisms for selectively degrading this form. This latter process may be important if neurons are particularly susceptible to a toxic effect of Ctm PrP.
Ctm PrP and Sec PrP Are Differentially Localized-Our data demonstrate that Sec PrP is present on the plasma membrane, primarily on neuronal processes, whereas Ctm PrP is found in the Golgi apparatus in the cell body. This conclusion is based on immunofluorescence staining of cultured granule neurons and is consistent with the observed endo H resistance of L9R-3AV PrP in lysates of granule neurons and brain. We have also observed Golgi localization of L9R-3AV PrP in neurons by immunocytochemical staining of vibratome sections cut from the cerebellum, hippocampus, and cerebral cortex of Tg(L9R-3AV) mice, confirming the results reported here on granule neurons in culture. 2 Our data suggest the following model for the cellular trafficking of Ctm PrP and Sec PrP in neurons (Fig.  10). Both Ctm PrP and Sec PrP are initially synthesized in the ER and are then transported to the Golgi. Ctm PrP remains trapped in the Golgi, whereas Sec PrP continues its transit to the plasma membrane where it is eventually distributed along neuronal processes.
The explanation for the differential trafficking of Ctm PrP and Sec PrP remains to be determined. One possibility is that Ctm PrP contains a Golgi retention or retrieval signal that is absent in Sec PrP. The transmembrane domain of Ctm PrP is a likely candidate for such a retention signal, because a number of other Golgi resident proteins utilize membrane-embedded segments as retention signals (33). The L9R-3AV mutation itself does not seem to impair delivery of Sec PrP molecules to the cell surface, implying that this mutation does not result in gross misfolding of the protein.
The subcellular localization of L9R-3AV PrP in granule neurons is dramatically different from its localization in transiently transfected CHO, BHK, and N2a cells. In these transfected cells, the mutant protein is retained in the ER and remains endo H-sensitive throughout its metabolic lifetime (this paper and see Ref. 16). We have observed that L9R-3AV PrP is endo H-resistant in fibroblasts cultured from Tg(L9R-3AV) mice, 3 arguing that factors other than cell type determine the localization of the mutant protein. It is possible that the high expression levels characteristic of transiently transfected cells cause ER retention of Ctm PrP, whereas the more physiological levels present in cells from transgenic mice allow the protein to transit further along the secretory pathway.
Post-translational Processing of Ctm PrP and Sec PrP-Our pulse-chase labeling experiments indicate that L9R-3AV PrP is synthesized in the ER as an endo H-sensitive precursor that subsequently matures to two endo H-resistant glycoforms as the protein transits the Golgi. Our analysis indicates that these two glycoforms of 32 and 35 kDa correspond to Sec PrP and Ctm PrP, respectively. The 35-kDa form is larger than the mature, doubly glycosylated wild-type PrP (ϳ32 kDa), and it undergoes a larger shift in gel mobility after treatment with neuraminidase. These observations suggest that Ctm PrP is hyperglycosylated, possibly due to its protracted residence in the Golgi apparatus, where sialic acids are added in the trans cisterna. Because the two glycoforms of mutant PrP migrate with different mobilities even after neuraminidase treatment, Ctm PrP must be distinguished from Sec PrP by post-translational modifications in addition to sialic acid residues. These conclusions regarding the maturation and glycosylation of Ctm -PrP and Sec PrP have been incorporated into the model shown in Fig. 10.
Despite the difference in their subcellular localization and glycosylation, Ctm PrP and Sec PrP appear to decay with a similar metabolic half-life (ϳ2.5 h). The mechanisms responsible for degradation of L9R-3AV in neurons and other cells remain to be determined. In a previous report, we showed by Western blotting that L9R-3AV PrP accumulated in transiently transfected BHK cells treated for 16 h with a proteasome inhibitor (16). This observation suggested that Ctm PrP may be a substrate for proteasomal degradation following retrotranslocation from the ER. However, we subsequently discovered that long term treatment of cells with proteasome inhibitors causes an artifactual increase in PrP mRNA levels when expression is driven from a strong viral promoter (25). When we tested shorter inhibitor treatments in pulse-chase labeling experiments, we found that the inhibitors had no effect on the maturation or turnover of L9R-3AV PrP in either cerebellar granule neurons or transfected cells. 3 We thus conclude that the proteasome does not play a prominent role in the metabolism of L9R-3AV PrP in these cell types.
The signal sequence of Sec PrP, like that of other secreted glycoproteins, is normally cleaved by a signal peptidase that acts in the lumen of the ER. In contrast, Ctm PrP contains an uncleaved signal sequence (16), reflecting the fact that the N terminus of this form remains in the cytoplasm and does not enter the ER lumen. We report here that, in neurons, both the Ctm and Sec forms of L9R-3AV PrP contain an uncleaved signal sequence, even though the latter molecules have been completely translocated into the ER lumen. One explanation for this phenomenon is that the L9R mutation itself interferes with the action of signal peptidase. Alternatively, signal peptidase may be less active, or the topogenesis of Sec PrP may be different, in neurons compared with non-neuronal systems. Analysis of L9R PrP expressed in neurons would help resolve this issue.
The presence of an uncleaved signal peptide does not seem to interfere with trafficking of mutant Sec PrP to the neuronal cell surface. Moreover, Sec PrP could be released from the cell surface by PIPLC, suggesting that the hydrophobic signal peptide does not integrate into the lipid bilayer. In pulse-chase exper- iments (not shown), we found that L9R-3AV PrP molecules recognized by anti-SP antibody were present throughout the chase period and appeared to decay with a half-life similar to that of molecules recognized by 3F4 antibody. These results indicate that the mutant PrP in neurons is synthesized with an intact signal peptide and that the signal peptide is not selectively cleaved during maturation and turnover of the protein.
All of the experiments reported here have been carried out on cerebellar granule neurons cultured from Tg(L9R-3AV) mice. Of course, it is possible that other types of neurons may handle the mutant PrP differently. However, we find that Western blots of brain lysates from Tg(L9R-3AV) mice display the same endo H-resistant PrP glycoforms of 32 and 35 kDa that are seen in cultured granule neurons, albeit in different relative amounts. This result is consistent with the possibility that the post-translational processing and subcellular localization of mutant protein are qualitatively similar in many types of neurons in Tg(L9R-3AV) brain. This conclusion is also consistent with our observation that L9R-3AV PrP is concentrated in the Golgi of neurons from a number of brain regions, based on immunocytochemical staining of brain sections. 2 Comparison with Other Studies-Ours is the first published study to examine the localization and metabolism of Ctm PrP in cultured neurons. Hegde et al. (7) have reported that PrP carrying either of two Ctm PrP-favoring mutations (3AV or K109I/H110I) is endo H-resistant in brain lysates from transgenic mice. This result implies that Ctm PrP induced by these mutations has also transited the mid-Golgi, although immunocytochemical localization studies would be necessary to confirm its precise localization.
Singh and colleagues have reported increased surface expression of a C-terminal fragment of Ctm PrP in neuroblastoma cells that have been treated with the synthetic peptide PrP106 -126 (34), or that have been transfected to express another PrP mutant (P101L) (35). However, the fact that this fragment is releasable by treatment of cells with PIPLC calls into question its relationship to Ctm PrP.
Clues to the Neurotoxicity of Ctm PrP-Cerebellar granule neurons, as well as hippocampal pyramidal cells, undergo massive degeneration in the brains of Tg(L9R-3AV) mice. The localization of Ctm PrP in the Golgi apparatus of cerebellar granule neurons in culture raises the possibility that the toxic effects of Ctm PrP on this cell type in vivo may involve this organelle. Although apoptotic pathways are known to be triggered in the ER as a result of protein misfolding, the role of the Golgi in programmed cell death is less clear. The Golgi apparatus undergoes a dramatic disassembly process during apoptosis (36,37). In addition, there are several caspase substrates and at least one procaspase and a caspase inhibitor that reside in this organelle. Thus, it is possible that Ctm PrP in the Golgi directly initiates apoptotic signals or amplifies signals that originate elsewhere in the cell. On the cell surface, PrP C is known to be localized to lipid rafts that contain other GPIanchored proteins (38,39) and that have been implicated in several kinds of signaling pathways (40,41). Because Ctm PrP contains a GPI anchor (16,17), it is conceivable that this form is incorporated into lipid rafts that begin to assemble in the Golgi (40) and that this localization plays a role in neurotoxic signaling. Because the N-terminal half of the Ctm PrP molecule is cytoplasmic, interaction with pro-apoptotic proteins in the cytoplasm could play a role in the neurotoxic effects of Ctm PrP regardless of where along the secretory pathway this form is localized.
Most unexpectedly, we have found that the neurodegenerative phenotype in Tg(L9R-3AV) mice is strongly dependent on coexpression of endogenous, wild-type PrP. 2 For example, Tg(L9R-3AV-B ϩ/Ϫ )/Prn-p ϩ/ϩ mice develop neurological symptoms at ϳ170 days of age and die with extensive loss of cerebellar and hippocampal neurons at ϳ390 days of age. In contrast, mice from the same line on the Prn-p 0/0 background never develop symptoms and have histologically normal brains. These results imply that wild-type PrP C influences the transmission of a toxic signal from Ctm PrP. We have shown in this study that the Prn-p status of the mice from which granule neurons are cultured has no effect on the topology ratio, subcellular localization, or post-translational processing of L9R-3AV PrP. Thus, endogenous PrP C most likely influences the phenotype of the mice by altering the functional activity of Ctm PrP, rather than by changing its amount or distribution. We have postulated that wild-type Sec PrP normally mediates a neuroprotective signal that is converted to a toxic signal upon physical interaction with Ctm PrP. 2 This interaction may occur as Sec PrP transits the Golgi on its way to the cell surface. An important goal now will be the identification of the signaling pathways of which Sec PrP and Ctm PrP may be components. Neurons cultured from Tg(L9R-3AV) mice may prove crucial in these investigations.