Prion Peptide Modulates the Aggregation of Cellular Prion Protein and Induces the Synthesis of Potentially Neurotoxic Transmembrane PrP*

Transcription

1 THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 277, No. 3, Issue of January 18, pp , by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A. Prion Peptide Modulates the Aggregation of Cellular Prion Protein and Induces the Synthesis of Potentially Neurotoxic Transmembrane PrP* Received for publication, May 14, 2001, and in revised form, October 5, 2001 Published, JBC Papers in Press, October 26, 2001, DOI /jbc.M Yaping Gu, Hisashi Fujioka, Ravi Shankar Mishra, Ruliang Li, and Neena Singh From the Institute of Pathology, Case Western Reserve University, Cleveland, Ohio In infectious and familial prion disorders, neurodegeneration is often seen without obvious deposits of the scrapie prion protein (PrP Sc ), the principal cause of neuronal death in prion disorders. In such cases, neurotoxicity must be mediated by alternative pathways of cell death. One such pathway is through a transmembrane form of PrP. We have investigated the relationship between intracellular accumulation of prion protein aggregates and the consequent up-regulation of transmembrane prion protein in a cell model. Here, we report that exposure of neuroblastoma cells to the prion peptide catalyzes the aggregation of cellular prion protein to a weakly proteinase K-resistant form and induces the synthesis of transmembrane prion protein, the proposed mediator of neurotoxicity in certain prion disorders. The N terminus of newly synthesized transmembrane prion protein is cleaved spontaneously on the cytosolic face of the endoplasmic reticulum, and the truncated C-terminal fragment accumulates on the cell surface. Our results suggest that neurotoxicity in prion disorders is mediated by a complex pathway involving transmembrane prion protein and not by deposits of aggregated and proteinase K-resistant PrP alone. Prion disorders manifest when the prion protein (PrP C ), 1 a normal cell surface glycoprotein, undergoes a conformational change from a predominantly -helical to a -sheet-rich structure that is pathogenic (PrP Sc ). This transformation is initiated by an exogenous source of PrP Sc in cases acquired by infection, triggered by mutation(s) in the prion protein gene in inherited forms, and is a random, spontaneous event in sporadic cases. Following the initial conversion, subsequent transformation of additional PrP C molecules progresses autocatalytically, resulting in deposits of PrP Sc in the brain parenchyma. Unlike PrP C, PrP Sc aggregates easily, is insoluble in nonionic detergents, * This work was supported by National Institutes of Health Grants NS35962 and NS39089 (to N. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. To whom correspondence should be addressed: Institute of Pathology, Case Western Reserve University, 2085 Adelbert Rd., Cleveland, OH Tel.: ; nxs2@po.cwru.edu. 1 The abbreviations used are: PrP C, normal cell-associated prion protein; PrP , PrP peptide including residues biotin-tagged at the N terminus; PrP Scr , PrP peptide with a scrambled sequence biotin-tagged at the N terminus; PrP Sc, conformationally transformed scrapie form of PrP; A, -amyloid peptide of amyloid precursor protein; PK, proteinase K; PI-PLC, phosphatidylinositol-specific phospholipase C; ER, endoplasmic reticulum; DAB, 3,3 -diaminobenzidine; GPI, glycosylphosphatidylinositol; PNGase-F, N-glycosidase F; Ctm PrP, transmembrane PrP; L, lysosomes; N, nucleus; E, endosomes. This paper is available on line at and is partially resistant to limited digestion by proteinase K. Deposits of PrP Sc in the brain parenchyma are believed to be the principal cause of neuronal toxicity in prion disorders (1 3). Although PrP Sc is believed to be responsible for both transmission and pathogenicity in all prion disorders, the molecular events leading to PrP Sc -induced transformation of additional PrP C molecules and the consequent neuronal toxicity are poorly understood. Because neurodegenerative changes typical of prion disorders are often seen without detectable PrP Sc, alternative mechanisms of neuronal death besides PrP Sc deposition have been suggested (4 6). One such mechanism is through the preferred synthesis of Ctm PrP, a transmembrane form of PrP that spans the endoplasmic reticulum (ER) membrane at residues with its N terminus in the cytosol, rather than the normal glycolipid-linked PrP C that is translocated co-translationally into the ER lumen. Mice carrying the mutant PrP transgene A117V that has an increased predilection for Ctm PrP synthesis show spontaneous neurodegeneration without detectable PrP Sc and, when challenged with infectious prions, show neurodegeneration earlier and with smaller amounts of accumulated PrP Sc than the corresponding animals with a deleted transmembrane domain. In fact, in these cases the extent of neurodegeneration correlates directly with the amount of Ctm PrP rather than PrP Sc load, indicating that Ctm - PrP, and not accumulated PrP Sc, is responsible for the observed neurodegeneration (7, 8). We have examined the initial events of PrP C aggregation in a cell model, and the correlation between intracellular accumulation of aggregated PrP C and Ctm PrP generation. To initiate the aggregation of endogenous PrP C, we have used an internal peptide of PrP comprising residues (PrP ) instead of the proteinase K (PK)-resistant core of PrP Sc that constitutes the infectious prion particle. PrP offers the advantage of being similar to PrP Sc in several respects and at the same time is more soluble and easy to manipulate for cell culture studies. For example, like PrP Sc, PrP is rich in -sheet structure, forms aggregates that are detergent-insoluble and PK-resistant, and is toxic to primary cultures of neurons (9 11). Furthermore, residues of PrP have been shown to be sufficient for initiating or blocking the transformation of PrP C, probably because these constitute the principal site of binding of PrP Sc to PrP C during the conversion process (12 15). This region can be further narrowed down to residues , which are considered particularly important for the binding and inhibitory effect (11, 14). In this study, we demonstrate that when PrP C -expressing neuroblastoma cells are exposed to nontoxic concentrations of PrP , micro-aggregates of PrP seed the aggregation of endogenous, cellular PrP C into thioflavin-binding deposits that accumulate in the lysosomes. Subsequently, there is increased synthesis of the potentially neurotoxic transmem-

2 2276 Induction of Ctm PrP by PrP Sc -like Forms in a Cell Model brane Ctm PrP in these cells. Our data provide the first direct evidence for nucleation-dependent transformation of PrP C into aggregated and weakly PK-resistant forms in a cell model. More importantly, our findings show a direct correlation between intracellular PrP C aggregation and Ctm PrP up-regulation, indicating that prion-related neuropathology is mediated by complex cellular pathway(s) involving Ctm PrP and not simply by deposits of PrP Sc. EXPERIMENTAL PROCEDURES Materials and Cell Culture Conditions All cell culture supplies were obtained from Invitrogen. Hygromycin B was from Calbiochem; Hoechst and Lysotracker dye were from Molecular Probes (Eugene, Oregon). Biotin-tagged PrP and biotin-tagged scrambled PrP were custom synthesized (Genemed Corp., San Francisco). Glutaraldehyde, osmium tetroxide, uranyl acetate, lead citrate, and epoxy resin were from Polysciences Inc. (Warrington, PA). Anti-PrP monoclonal antibody 3F4 (specific to PrP residues ) was from Richard Kascsak (New York State Institute for Basic Research in Developmental Disabilities). Anti-PrP monoclonal antibody 8H4, which binds to a site between residues of PrP, was raised in our facility (20). All other chemicals, including antibodies to the glial fibrillary acidic protein (anti-gfap) and neurofilament-specific (NF68) antibodies were purchased from Sigma. PrP C -expressing human neuroblastoma cells were generated as described previously (17) and cultured in the presence of 500 g/ml hygromycin B. Fetal brain cells were obtained from the Birth Defects Research Laboratory (University of Washington). Discarded brain tissue from human therapeutic abortions was washed with cold Hanks buffer containing gentamycin and dissociated by trituration through a fire-polished glass pipette. Cells were separated from large aggregates of connective tissue by a brief centrifugation at 200 g for 5 min. The supernatant containing suspended cells was recentrifuged, and the cells were cultured on Matrigel-coated dishes in minimum essential medium containing 10% fetal bovine serum, 1% penicillin/streptomycin, and 0.01% gentamycin for 24 h. The culture medium was then replaced with neurobasal medium containing B27 supplements. To enrich for neuronal cells, the mitotic inhibitor cytosine arabinoside (10 M) was added to the medium after 2 days of culture. For immunofluorescence analysis, the cells were plated on Matrigel-coated coverslips rather than in Petri dishes and subjected to the required experimental conditions. Treatment of Neuroblastoma and Mixed Brain Cells with PrP The following peptides were custom synthesized for the study: biotintagged PrP (PrP ), biotin- 106 KTNMKHMAGAAAAGAVVG- GLG 126 ; and biotin-tagged scrambled PrP (PrP Scr ), biotin- NGAMALMGGHGATKVKVGAAA. Biotinylation of PrP does not alter its biochemical properties (11, 16). Peptide stocks were dissolved in Me 2 SO and kept frozen until use. At the time of subculture, PrP or PrP Scr (5 M) was dissolved in culture medium and centrifuged at 1000 g for 10 min to eliminate large aggregates. The supernatant was used to culture PrP C -over-expressing neuroblastoma cells or primary brain cultures. Every 3 days, the neuroblastoma cells were trypsinized and subcultured with fresh peptide solution. Brain cells were cultured in the presence of peptide for 1 4 weeks. Immunofluorescence Staining and Confocal Microscopy For all immunostaining experiments, anti-prp antibody 8H4 was used to detect cellular PrP. This antibody is specific to residues of PrP and thus does not react with PrP The specificity of this antibody has been confirmed in several studies by independent laboratories (18 20). For detecting biotin-tagged PrP or biotin-tagged PrP Scr , streptavidin-conjugated-texas red was used because of the specific binding of streptavidin with biotin. For immunostaining cell surface proteins, permeabilization of the membrane with Triton X-100 was omitted. Neuroblastoma cells cultured in normal medium or in 5 M biotin-prp or biotin-prp Scr were fixed and stained with 8H4 (1:25)-anti-mouse-fluorescein isothiocyanate (1:25) (Southern Biotechnology Associates) followed by streptavidin-conjugated-texas red (1:40) (Pierce). For intracellular staining, fixed cells were permeabilized with 0.1% Triton X-100 prior to immunostaining as described above. Stained cells were incubated with Hoechst (1 g/ml) (Molecular Probes) for 5 min to detect apoptotic nuclei and mounted in Gel-mount (Biomeda Corp., Foster City, CA). All samples were observed with a laser scanning confocal microscope (Bio-Rad MRC 600). A single 5 m optical section and a composite of several sections were evaluated in each case. Electron Microscopy For immunoelectron microscopy, cells treated with biotin-prp Scr ( control ) or biotin-prp ( treated ) were fixed with 4% paraformaldehyde and 0.01% glutaraldehyde, and endogenous peroxidase activity was quenched by exposing fixed cells to 0.3% H 2 O 2. The cells were then immunostained with 8H4 (1:20) followed by peroxidase-conjugated anti-mouse (1:50) and mouse peroxidase-antiperoxidase (1:250), exposed to 3,3 -diaminobenzidine (DAB) containing 0.1 M imidazole to obtain an electron-dense reaction product and fixed again with 2.5% glutaraldehyde. Subsequently, the cells were post-fixed in 1% osmium tetroxide, dehydrated, embedded in Epon 812, and processed for electron microscopy. Post-staining with uranyl acetate and lead citrate was omitted to enhance the contrast for the PrP-specific DAB stain. For intracellular immunostaining, fixed cells were permeabilized with 0.1% Triton X-100 before processing as described above. For ultrastructural analysis, cells were processed by conventional methods and examined with a Zeiss electron microscope (Zeiss CEM 902; Carl Zeiss Inc., Thornwood, NY). SDS-PAGE and Western Blotting Cells treated with biotin- PrP Scr (control) or biotin-prp (treated) were processed for Western blotting as described (17, 37). Membranes containing transferred proteins were probed with either anti-prp antibody 8H4 (1:1000) or 3F4 (1:40,000) followed by anti-mouse-horseradish peroxidase (1: 4000) or streptavidin-conjugated horseradish peroxidase (1:40,000). Reactive bands were visualized on an autoradiographic film by ECL (Amersham Biosciences Inc.). Metabolic Labeling and Immunoprecipitation Cells treated with biotin-prp Scr (control) or biotin-prp (treated) were radiolabeled with 50 Ci/ml of Tran 35 S-label overnight in Dulbecco s modified Eagle s medium containing 5% dialyzed serum or with 75 Ci/ml of [ 3 H]ethanolamine in complete medium. Labeled cells were treated with PI-PLC (Oxford Glycosystems) in Opti-MEM for 1 h at 37 C. Cell lysates and PI-PLC-released proteins in the medium were subjected to immunoprecipitation with anti-prp antibody 3F4 or 8H4 as described (17, 37). Metabolic labeling at 15 C was performed in a refrigerated CO 2 incubator for 2 h, and lysates were immunoprecipitated with anti-prp antibody 2301 specific for C-terminal residues of PrP (17, 37, 38). Anti-PrP antibody 3F4 binds to residues 109 and 112 of PrP and therefore detects truncated Ctm PrP and the conventional PK-resistant 20-kDa fragment of PrP. Anti-PrP antibodies 8H4 and 2301 bind to both the 18- and 20-kDa fragments of PrP (Fig. 8). Assay of Detergent Insolubility and Proteinase K Resistance Cells treated with biotin-prp Scr (control) or biotin-prp (treated) were lysed in a buffer containing nonionic detergents Nonidet P-40 (1%) and deoxycholate (0.5%). Large aggregates and nuclei were pelleted by low speed centrifugation at 500 g, and the low speed supernatant was further ultracentrifuged at 100,000 g in a Beckman SW50 rotor for 1hat4 C. The low and high speed pellet and supernatant samples were analyzed by Western blotting. Duplicate sets of samples were electroblotted and probed with 8H4, 3F4, or streptavidin-horseradish peroxidase as described above. Only the low speed supernatant fraction is shown here because all of the aggregated PrP C and PrP were found in the low speed pellet and supernatant fractions rather than in the high speed pellet as observed for conventional aggregated PrP. For evaluation of proteinase K resistance, lysates of control (biotin- PrP Scr ) or treated (PrP ) cells were exposed to 3 g/ml PK for 0, 2, or 4 min at 37 C. The reaction was stopped with 1 mm phenylmethylsulfonyl fluoride, and proteins were electroblotted and probed with 3F4 as described (17, 37). Enzymatic Deglycosylation Deglycosylation with N-glycosidase F (PNGase-F) was performed essentially as described (17, 37). Cell Homogenization and PK Treatment Cells treated with biotin- PrP Scr (control) or PrP (treated) were washed and homogenized on ice in a buffer containing 10 mm HEPES (ph 7.9), 1.5 mm MgCl 2,10mM KCl, and 0.5 mm dithiothreitol by 20 strokes of a Kontes all-glass Dounce homogenizer. The homogenate was checked microscopically for cell breakage and centrifuged to pellet nuclei. The resulting supernatant was centrifuged at 20,000 g to pellet membrane vesicles. The pelleted vesicles were resuspended in 0.5 ml of transport buffer (25 mm HEPES, ph 7.4, 115 mm KOAc, 2.5 mm MgCl 2,10mM KCl, 2.5 mm CaCl 2,and1mM dithiothreitol) and treated with 20 g/ml proteinase K on ice for 30 min. After the addition of 5 mm phenylmethylsulfonyl fluoride to stop the reaction, membrane vesicles were pelleted again, solubilized in lysis buffer, and immunoblotted with 3F4. Staining with Thioflavin S Cells treated with biotin-prp Scr (control) or biotin-prp (treated) for 1 year were fixed and permeabilized as described above and immunostained with 8H4 (1:25 dilution)-anti-mouse-rhodamine isothiocyanate (1:40 dilution). Subsequently, the cells were incubated with a 1% aqueous solution of

3 Induction of Ctm PrP by PrP Sc -like Forms in a Cell Model 2277 thioflavin S for 20 min and washed three times with phosphate-buffered saline followed by one quick wash with 70% ethanol. Cells stained only with thioflavin S were prestained with the basic dye hematoxylin to reduce the background binding to nucleic acids, and processed as above. RESULTS PrP Induces the Aggregation and Intracellular Accumulation of PrP C In a previous report, we showed that exposure of neuroblastoma cells to nontoxic concentrations of PrP leads to intracellular accumulation of aggregated peptide, whereas scrambled PrP is degraded rapidly (16). To study the effect of long-term exposure of cells to PrP , neuroblastoma cells expressing high levels of transfected PrP C were cultured for 4 12 months in complete medium containing 5 M PrP or PrP Scr biotinylated at the N terminus. The cells were subcultured every 3 days, and the peptide was replenished. Cell viability was checked every day for the first month by staining with the nuclear dye Hoechst, and the rate of cell division was monitored by counting the number of cells every 3 days before subculturing. At this low concentration, the peptide dissolves easily in culture medium at ph 7.4 and is not toxic to neuroblastoma cells even after prolonged periods of constant exposure. Interaction of PrP and the corresponding scrambled peptide with PrP C -expressing cells was studied by immunofluorescence analysis. Monoclonal antibody 8H4 was used to detect cellular PrP C and Texas red-conjugated streptavidin to visualize biotin-tagged PrP or the corresponding scrambled peptide. Because the epitope for 8H4 lies between residues of PrP, it does not bind to PrP , and it provides a convenient method of differentiation between PrP C and PrP After 4 months of exposure to biotin-prp , cells were fixed with paraformaldehyde and immunostained for PrP C (green) orprp (red; Fig. 1). To visualize immunostained surface proteins, permeabilization with detergent was omitted. A composite confocal image was taken to detect staining at different depths, including invaginations of the plasma membrane. Small to large aggregates of PrP are evident at the cell surface (red), most of which co-localize with PrP C (yellow; Fig. 1A, panels 1 and 2). Some aggregates appear to be undergoing endocytosis (*, Fig. 1A, panels 1 and 2), whereas others are larger and more diffuse (Fig. 1A, panel 2). Control cells cultured with biotin-prp Scr show no staining for the peptide but a uniform surface staining for PrP C with 8H4 and 3F4 antibodies (Fig. 1A, panels 3 and 4). Nontransfected M17 cells show no reactivity with 8H4 (Fig. 1A, panel 5). These results were confirmed by immunoelectron microscopy. Cells treated with biotin-prp Scr or biotin- PrP were fixed and stained with 8H4 without prior permeabilization as described above. The bound antibody was reacted with the appropriate secondary antibody and DAB to obtain an electron-dense deposit as described under Experimental Procedures. Subsequent staining with uranyl acetate and lead citrate was omitted to obtain a better contrast with DAB staining. Cells cultured in the presence of scrambled peptide show a uniformly distributed, punctate staining of PrP C on the cell surface (PM: Fig. 1A, panel 6), whereas PrP treated cells show localized aggregation of PrP C at the plasma membrane Fig. 1A, panels 7 and 8). The aggregates increase in size gradually and are internalized in large, membrane-bound vesicles (Fig. 1A, panels 9 11). In some cases the aggregates are so large as to distort the nuclear membrane (Fig. 1A, panel 10). The intracellular localization of aggregated PrP and PrP C was assessed by immunofluorescence analysis of Triton X-100-permeabilized cells. Cells treated with scrambled peptide or PrP for 4 months were immunostained with 8H4 and Texas red-conjugated streptavidin as described above. Composite confocal images were taken to visualize the aggregates at different depths in the cell. Control cells treated with scrambled peptide show the expected intracellular immunoreactivity of PrP C in the Golgi area (Fig. 1B, panel 1), whereas PrP treated cells show aggregates of PrP (red) and PrP C (green) or complex aggregates containing PrP C and PrP (yellow) at the cell surface and in intracellular vesicles (Fig. 1B, panels 2 and 3). Immunostaining with an alternate anti-prp antibody 3F4 shows similar results. Scrambled peptide-treated cells show normal PrP immunoreactivity (Fig. 1B, panel 4), whereas PrP treated cells show large intracellular aggregates with immunoreactivity for PrP C (green) (Fig. 1B, panels 5 and 6). (Because 3F4 also binds to PrP (at residues 109 and 112), the aggregates with a yellow staining pattern could be combined aggregates of PrP PrP C or the aggregated peptide alone.) Immunoelectron microscopy shows a uniformly distributed punctate staining of PrP C on the plasma membrane of cells treated with scrambled peptide (Fig. 1B, panel 7). In contrast, PrP treated cells show large intracellular aggregates of PrP C enclosed by a membrane (Fig. 1B, panel 8), in some cases large enough to distort the nucleus. Nontransfected M17 cells treated with PrP and processed in parallel show no reactivity for PrP C (Fig. 1B, panel 9). The above described results show that exogenously added PrP induces the aggregation and internalization of cell surface PrP C in large membrane-bound vesicles. The aggregates degrade slowly and cause a reactive proliferation of lysosomes in cells treated with PrP (Fig. 1C, panel 2) as compared with controls exposed to PrP Scr (Fig. 1C, panel 1). Aggregates of PrP C Are Insoluble in Nonionic Detergents Insolubility in non-ionic detergents is one of the earliest changes observed during the transition of PrP C to PrP Sc. Conventionally, detergent insolubility is determined by preparing cell lysates in a buffer containing Nonidet P-40 (1%) and deoxycholate (0.5%), and subjecting the lysates clarified at low speed (500 g) to ultracentrifugation at 100,000 g. Aggregated PrP is usually detected in the pellet fraction of ultracentrifuged samples. To check whether aggregated PrP C is insoluble in non-ionic detergents, control cells cultured with scrambled peptide and cells exposed to PrP for 4 12 months were subjected to the above treatment, and the pellet and supernatant fractions from low and high speed centrifugation were immunoblotted with streptavidin or 8H4 to detect PrP or PrP C, respectively. Because most of the aggregates partitioned in the low speed pellet and supernatant fraction, the supernatant obtained after centrifugation at 500 g is shown in Fig. 2A. Immunoreaction with 8H4 and 3F4 antibodies shows normal PrP glycoforms for both control cells cultured with biotin-tagged PrP Scr and treated cells cultured in the presence of biotin-tagged PrP (Fig. 2A, lanes 1 4). In addition, protein bands that react with both 8H4 and 3F4 antibodies are detected in the stacking gel of PrP treated cells (Fig. 2A, lanes 2 and 4, overexposed lanes 9 and 11). Strikingly, the PrP-reactive bands cross-react with streptavidin (Fig. 2A, lane 7), indicating that the protein aggregates in PrP treated cells are too large to be resolved by the stacking or the separating gel and that these are complex aggregates of PrP C and PrP (Fig. 2A, lanes 2, 4, 7, 9, and 11, top two arrows). A significant amount of monomeric, detergent-soluble PrP is detected at 3 kda in treated lysates (Fig. 2A, lanes 7 and 11) that co-migrates with purified PrP (Fig. 2A, lane 5), confirming that PrP added to the culture medium is not pre-assembled into large aggre-

4 2278 Induction of Ctm PrP by PrP Sc -like Forms in a Cell Model FIG.1.PrP induces the aggregation and internalization of cell surface PrP C. A, surface immunofluorescence analysis of cells treated with biotin-prp for 4 months with 8H4-fluorescein isothiocyanate to identify PrP C (green) followed by Texas red-streptavidin to detect biotin-prp (red) shows co-localization of PrP C and PrP aggregates on the plasma membrane (panels 1 and 2). Large aggregates of irregular shape that co-stain for PrP C and PrP (panel 2) and smaller aggregates that appear to be undergoing endocytosis are evident (panels 1 and 2, insets). Immunostaining of control cells treated with biotin-prp Scr for a similar period of time with 3F4 or 8H4 shows a normal pattern of PrP immunoreactivity on the cell surface and in the Golgi area (panels 3 and 4). Nontransfected M17 cells treated with biotin-prp and immunostained with 8H4 show no PrP immunoreactivity (panel 5). Bar, panels 1 and 2, 25 m; insets, 10 m. Immunoelectron microscopy of the above cells using 8H4 shows a uniform, punctate staining of PrP C on the surface of control cells treated with biotin-prp Scr (panel 6). In contrast, cells treated with biotin-prp show aggregates of PrP C on the cell surface (panels 7 and 8) that are gradually internalized in membrane-enclosed vesicles (panels 9 11). In some cells, the aggregates assume large proportions (panels 9 11), enough to distort the nucleus (panel 10). The aggregate in panel 10 is in a plasma membrane invagination, although it appears intracellular. PM, plasma membrane; N, nucleus. Bar, 1 m. B, immunofluorescence analysis of Triton-permeabilized cells treated with biotin-prp Scr (control) or biotin-prp with 8H4 shows normal staining pattern of PrP in control cells (green, panel 1). Cells treated with PrP show engorgement with PrP C (green), PrP (red), and combined (yellow) aggregates (panels 2 and 3). A similar analysis with 3F4 shows normal PrP staining pattern in control cells (green, panel 4), whereas PrP treated cells show intracellular aggregates of PrP C (green)orprp (red). (Because 3F4 cross-reacts with PrP , the yellow aggregates may represent only PrP mixed aggregates as in panels 2 and 3; immunoelectron microscopy of Triton-permeabilized cells shows similar results. A uniform, punctate reaction of PrP C can be detected on the plasma membrane of cells treated with biotin-prp Scr (panel 7). In contrast, biotin-prp treated cells show large intracellular aggregates of PrP C (panel 8). Nontransfected M17 cells treated with PrP show no PrP reactivity

5 Induction of Ctm PrP by PrP Sc -like Forms in a Cell Model 2279 FIG. 1 continued gates. Thus, the PrP C aggregates in treated cells are large and insoluble in non-ionic detergents and SDS. The aggregated PrP constitutes less than 10% of the total cellular PrP, which, as shown above, is detected in the detergent-soluble supernatant fraction. (A longer exposure of lanes 1 4 is shown in lanes 8 11 to show PrP /PrP C aggregates in the stacking gel.) To corroborate the above results, cells treated with biotin- PrP or biotin-prp Scr for 4 12 months were plated on glass coverslips and treated with Triton X-100 (0.1%) for 5 min on ice prior to fixation with paraformaldehyde. The resid- (panel 9). Bar, panels 1 6, 25 m; panels 7 9, 1 m. C, electron microscopy of cells treated with biotin-prp Scr (control) or biotin-prp for 1 year shows extensive proliferation of electron-dense lysosomes in PrP treated cells (panel 2) compared with the control sample (panel 1). There are no obvious signs of cellular toxicity, such as swollen mitochondria or condensation of nuclear chromatin in treated cells (panel 2). PM, plasma membrane; N, nucleus; L, lysosomes; M, mitochondria; G, Golgi apparatus. Bar, 0.7 m.

6 2280 Induction of Ctm PrP by PrP Sc -like Forms in a Cell Model FIG. 2.Aggregates of PrP C are insoluble in non-ionic detergents. A, lysates prepared from cells treated with biotin-prp Scr (Control) or biotin-prp (Treated) for 4 12 months were centrifuged at low speed (500 g), and the supernatant was immunoblotted with 8H4 (lanes 1 and 2), 3F4 (lanes 3 and 4), or streptavidin (lanes 6 and 7). Purified biotin-prp was added as a control in lane 5. Normal PrP glycoforms consisting of unglycosylated (27 kda), intermediate, and highly glycosylated forms are detected in 3F4 and 8H4 immunoblots (lanes 1 4). In addition, as expected, the 18- and 20-kDa fragments of PrP are evident in the 8H4 blot (lanes 1 and 2), whereas only the 20-kDa fragment is seen in the 3F4 blot (lanes 3 and 4). More importantly, treated samples show protein aggregates in the stacking gel and at the top of separating gel that immunoreact with 8H4, 3F4, and streptavidin (lanes 2, 4, and 7 or over-exposed lanes 9, 11, and 7). These aggregates are notably absent in control lysates (lanes 1, 3, and 6 or over-exposed lanes 8 and 10). A 3-kDa band of monomeric PrP can be detected by streptavidin in treated cells (lane 7) which, as expected, immunoreacts with 3F4 (lane 11) and co-migrates with purified PrP (lane 5). B, cells treated with biotin-prp for 4 months were plated on glass coverslips and treated with Triton X-100 on ice prior to fixation with paraformaldehyde. The residual membranes and insoluble cell debris were immunostained with 8H4-anti-mouse-fluorescein isothiocyanate to visualize PrP C (green) or with Texas redstreptavidin to detect PrP (red). Detergent-insoluble aggregates of PrP C that co-localize with PrP can be seen on residual membranes and as large aggregates (panels 1 and 2). Cells treated with biotinprp Scr and processed similarly show immunostaining of residual membranes, conspicuously lacking any aggregates of PrP C or PrP (panel 3). Bar, 10 m. ual insoluble material was immunostained with 8H4 and streptavidin. Detergent-insoluble aggregates of PrP C (green) and PrP (red) are detected in residual cell structures, some of which appear to be membrane ghosts (Fig. 2B, panel 1), whereas others bear the round shape of cellular organelles that have been solubilized by the detergent (Fig. 2B, panel 2). Cells treated with scrambled peptide show only occasional membrane ghosts, but conspicuously they lack large aggregates (Fig. 2B, panel 3). Aggregated PrP C Binds the Amyloid-specific Dye Thioflavin S To evaluate whether intracellular aggregates of biotin- PrP and PrP C bind the amyloid-specific dye thioflavin S, cells treated with biotin-prp for 1 year were fixed in paraformaldehyde, permeabilized with detergent, and stained with the basic dye hematoxylin followed by the amyloid-binding dye thioflavin S. Green fluorescence of thioflavin-stained aggregates can be seen in intracellular structures (Fig. 3, panels 1 4), similar to the large endocytic vesicles observed in Fig. 1, A and B. Co-immunostaining with 8H4 followed by rhodamine isothiocyanate-conjugated secondary antibody shows that most of the thioflavin-positive aggregates co-immunostain for PrP C (Fig. 3, panels 3 and 4, yellow). Some of the intensely thioflavin-positive aggregates do not show PrP C staining and are either adjacent to or surrounded by aggregated PrP C (Fig. 3, panels 2 and 3). These aggregates do not co-stain with the PrP marker streptavidin either (data not shown), suggesting that the epitopes for these indicators are not accessible because of aggregation or change in conformation of PrP C and PrP No thioflavin-positive staining was detected in untreated cells or in cells treated with scrambled peptide (data not shown). Immunoelectron microscopy of cells treated with PrP for 1 year shows a large intracellular aggregate that immunoreacts with 8H4, showing amyloid-like fibrils in the center of the aggregate (Fig. 3, panel 5, arrow). Thus, after prolonged exposure of cells to PrP , the internalized aggregates of both PrP and PrP C acquire some of the properties of amyloid.

7 Induction of Ctm PrP by PrP Sc -like Forms in a Cell Model 2281 FIG. 3.Intracellular PrP and PrP C aggregates bind thioflavin S. Cells treated with biotin-prp for 1 year were fixed, permeabilized with Triton X-100, and stained with thioflavin S to detect amyloid (green, panel 1) or immunostained with 8H4-anti-mouse-rhodamine isothiocyanate to detect PrP C (red) followed by staining with thioflavin S (panels 2 4). Green fluorescence of thioflavin S-positive aggregates of different sizes is seen in vesicular structures, some more intense than others (panel 1). Co-immunostaining for PrP C shows colocalization in some cells (panels 3 and 4, yellow), whereas in others the green fluorescence of amyloid is surrounded by PrP C (panel 2). Immunoelectron micrograph of PrP treated cells reacted with 8H4- peroxidase-dab shows an intracellular vesicle containing aggregated PrP with a fibrillar, amyloid-like appearance in the center (panel 5, arrow). Inset, 0.5 m. Bar, panels 1 4, 25 m; panel 5, 1 m. Primary Cultures of Human Neurons Show Similar Aggregation of PrP C Although the neuroblastoma cells used above tolerate relatively large intracellular aggregates of PrP C and PrP without significant toxicity (assessed by Hoechst staining), these are tumor cells that divide in culture and do not exactly replicate the metabolism of primary neuronal cells. To confirm whether similar aggregation of PrP C can be induced in primary neurons, mixed cultures of human brain cells prepared from discarded fetal tissue and enriched for neurons were treated with biotin-prp for 2 weeks as described above. The cells were washed thoroughly to remove precipitated peptide, fixed, permeabilized, and immunostained for PrP C and PrP as described above. Parallel cultures were immunostained with anti-glial fibrillary acidic protein (anti- GFAP) and anti-neurofilament (anti-nf68)-specific antibodies to identify specific cell types in the mixed brain cultures. Accumulation of PrP C and PrP can be seen in neuronal cell bodies, neuronal processes, and astrocytes (Fig. 4, panels 1, 2, and 3, respectively). Most of the intracellular deposits co-stain FIG. 4.Primary cultures of human neurons show similar aggregation of PrP C. Mixed cultures of human brain cells treated with biotin-prp for 2 weeks were immunostained for PrP C and PrP Aggregates of PrP C (green) and PrP (red) can be seen in neuronal cell bodies (panel 1), in axonal processes and nodal swellings (panel 2), and in astrocytes (panel 3). Most of the intracellular deposits co-stain for PrP C and PrP (yellow). Bar, 25 m. for PrP C (green) and PrP (red) (Fig. 4, panels 1 3, yellow), indicating the presence of mixed aggregates as seen above for neuroblastoma cells. Most of the cells in primary brain cultures show nuclear fragmentation after 4 weeks of PrP treatment, as opposed to cells cultured with PrP Scr , which are healthy for up to 8 weeks (data not shown). A C-terminal 20-kDa Fragment of PrP C Accumulates on the Surface of PrP treated Cells To investigate the metabolism of aggregated PrP C in cells treated with PrP , equal number of cells treated with PrP Scr (control) or PrP (treated) for 4 12 months were radiolabeled with [ 35 S]methionine overnight and treated with PI-PLC to cleave cell surface GPI-linked proteins. Both the lysate and PI-PLC-released samples were subjected to immunoprecipitation with anti-prp antibody 3F4 (specific to residues ). Analysis by SDS- PAGE fluorography shows significantly more PrP in the lysate sample of PrP treated cells compared with controls, indicating that some of the intracellular accumulated PrP C is soluble and immunoprecipitable before it finally aggregates (Fig. 5A, lanes 1 and 2). The amount of PrP C released from the cell surface is similar in control and treated cells, except for a significant increase in the 20-kDa fragment in treated cells (Fig. 5A, lanes 3 and 4). This fragment becomes more prominent after deglycosylation with PNGase-F (Fig. 5A, lanes 5 and 6), indicating the presence of glycosylated forms of 20 kda on the cell surface, which accumulate in treated cells. Radiolabeling with the GPI anchor component [ 3 H]ethanolamine and

8 2282 Induction of Ctm PrP by PrP Sc -like Forms in a Cell Model FIG. 5.A C-terminal 20-kDa fragment of PrP C accumulates on the surface of PrP treated cells. A, cells treated with biotin- PrP Scr (Control) or biotin-prp (Treated) for 4 12 months were radiolabeled overnight with [ 35 S]methionine and treated with PI-PLC to cleave surface-expressed PrP. Lysate and PI-PLC-released samples were immunoprecipitated with 3F4 and deglycosylated with PNGase-F. The amount of PrP C immunoprecipitated from treated cells is significantly more than with control cells (lane 1 versus 2). The 20-kDa fragment of PrP is detected only in treated cells (lane 2) but is significantly more in the PI-PLC-released sample compared with cellassociated proteins (lane 2 versus 4). The intensity of the 20-kDa fragment increases following PNGase-F treatment, indicating the presence of glycosylated forms that migrate at 20-kDa when glycans are removed (lane 6). Radiolabeling with the GPI anchor component [ 3 H]ethanolamine and immunoprecipitation with 3F4 show that the 20-kDa fragment is GPI-linked (lane 8). B, quantitative measurement of the 20-kDa fragment versus full-length PrP forms in control and treated samples was done by densitometric scanning of autoradiograms as described in A above. The data represent the percentage of 20-kDa fragment in comparison with total full-length PrP in each lane. The differences among mean values were tested by analysis-of-variance for repeated measurements. Statistical analysis within groups was carried out using Student s t test. Each bar represents the mean S.D. of three experiments. *, p immunoprecipitation with 3F4 confirms that the 20-kDa fragment is GPI-anchored and therefore C-terminal (Fig. 5, lanes 7 and 8). Quantitative estimation of the relative percentage of 20-kDa fragment in comparison with full-length PrP forms shows that in control cells, 20-kDa constitutes 2% of the total PrP, more than half of which is at the cell surface (Fig. 5B, lanes 1 and 3). After treatment with PrP , the total amount of 20 kda increases to 10%, more than half of which is at the cell surface (Fig. 5B, lanes 2 and 4). Treatment of PI- PLC-cleaved surface proteins with PNGase-F shows that the amount of 20-kDa fragment on the surface of treated cells is at least four-fold more compared to control cells (Fig. 5B, lanes 5 and 6). A similar ratio of 20-kDa versus full-length PrP is observed in ethanolamine-labeled samples (Fig. 5B, lanes 7 and 8). The C-terminal 20-kDa Fragment Is Truncated Ctm PrP Previous reports have documented that PrP C exists in different topological forms at the ER. Increased synthesis of one of the transmembrane forms, Ctm PrP, is believed to mediate the neurodegeneration observed in certain inherited and infectious prion disorders (7, 8). As opposed to PrP C, Ctm PrP is inserted in the ER membrane through its hydrophobic domain, including residues , with its N terminus in the cytosol and C terminus in the ER lumen. Thus, PK treatment of microsomes spares an 19-kDa C-terminal fragment, which is significantly increased in brain tissue obtained from diseased animals (7). Because this fragment retains reactivity to 3F4, the proteolytic clip must spare residue 109. To determine whether the 20-kDa fragment observed in PrP treated cells is the proteolytically cleaved Ctm PrP, microsomes prepared from cells cultured in the presence of PrP Scr (control) or PrP (treated) for 4 12 months were treated with 20 g/ml PK on ice for 30 min as described in previous reports (7, 8). The reaction was stopped with phenylmethylsulfonyl fluoride, and the microsomes were solubilized in detergent and subjected to immunoblotting with 3F4 or anti-calnexin antibody. As expected, fulllength GPI-linked forms of PrP C are protected from the protease in both control and treated cells (Fig. 6A, lanes 2 and 4). In contrast, there is a small but significant increase in the 20-kDa fragment after protease digestion of treated cells (Fig. 6A, lane 3 versus 4), indicating that the 20-kDa fragment is generated by limited proteolysis of Ctm PrP and, more importantly, that the number of PrP molecules inserted in the Ctm PrP orientation is higher in treated cells than in controls. Cleavage of calnexin to produce a faster migrating species lacking the cytosolic C-terminal domain confirms the efficiency of PK treatment and the intactness of the microsomes (Fig. 6A, lanes 2 and 4). Because the number of PrP C molecules in the Ctm orientation constitute less than 2% of the total PrP C in Chinese hamster ovary and baby hamster kidney cells (34), the increase observed in peptide treated cells is noteworthy. A quantitative estimation of the percentage of 20-kDa fragment compared with full-length PrP forms shows a 4-fold increase in the 20-kDa fragment in treated cells as compared with the controls (Fig. 6C, lane 1 versus 3). More significantly, protease treatment of the control sample results in an increase of 2% in the amount of 20 kda, as compared with a 33% increase in the treated sample (Fig. 6C, lanes 1 versus 2 and 3 versus 4). (It is difficult to visualize the 20-kDa fragment in control cells by immunoblotting, but is detected by densitometry.) To confirm that the 20-kDa fragment indeed originates in the ER, control and treated cells were radiolabeled for 2hat15 C to block transport of secretory proteins from the ER. Lysates were subjected to immunoprecipitation with anti-prp antibody 2301 and analyzed by SDS-PAGE-fluorography. Because transport of PrP beyond the ER-cis-Golgi is blocked under these conditions, as expected, only PrP glycoforms with high mannose core glycans are detected. The highly glycosylated form and the 18-kDa C-terminal fragment that is a product of normal recycling of PrP from the plasma membrane are also absent (Fig. 6B, lanes 1 and 2). However, there is a clear accumulation of the 20-kDa fragment, which is significantly more in treated cells as compared with the control sample (Fig. 6B, lane 1 versus 2). Because 2301 is specific for the C-terminal residues of PrP, these results confirm that the 20 kda is indeed a C-terminal fragment of PrP, and it originates in the ER. Upon quantitative analysis, the 20-kDa fragment constitutes 0.8% of total full-length PrP in control cells, as compared with 10.5% in treated cells, an increase of 10-fold due to treatment of the cells with PrP (Fig. 6D, lanes 1 and 2). To exclude the possibility that 20 kda is a proteolytic product

9 Induction of Ctm PrP by PrP Sc -like Forms in a Cell Model 2283 FIG. 6.The C-terminal 20-kDa fragment represents truncated Ctm PrP. Cells treated with biotin-prp Scr (Control) or biotin- PrP (Treated) for 4 12 months were homogenized, and the resulting microsomes were treated with 20 g/ml PK for 30 min on ice. After inactivating PK, microsomes were solubilized with detergent and immunoblotted with 3F4 or anti-calnexin antibody. As expected, fulllength PrP C is protected from PK digestion in all samples (lanes 1 4). The amount of 20-kDa fragment is too low to be detected in control cells by immunoblotting either before or after PK treatment (lanes 1 and 2). In peptide-treated cells, on the other hand, the 20-kDa fragment increases in amount after PK treatment of microsomes (lanes 3 and 4). As expected, the C terminus of calnexin is cleaved by PK, increasing its migration by 10 kda (lanes 2 and 4). B, control and treated cells as above were radiolabeled with [ 35 S]methionine for 2hat15 C, lysed, and subjected to immunoprecipitation with anti-c-terminal PrP antibody 2301 to detect both the 18- and 20-kDa fragments. As expected, full-length PrP forms do not exit the ER, indicated by the absence of highly glycosylated forms of PrP in both cell lines (lanes 1 and 2). Strikingly, in comparison with control cells, there is a marked increase in the 20-kDa fragment in peptide-treated cells (lane 1 versus 2). A faint band of 18 kda is detected in peptide-treated cells (arrowhead) probably because of the exit of a small amount of PrP to the plasma membrane (*, lanes 1 and 2). C and D, quantitative analysis of the amount of 20-kDa versus full-length PrP forms in control and treated samples was carried out by densitometric scanning of autoradiograms as described in A and B above. The data represent the percentage of 20-kDa fragment in comparison with total full-length PrP. The differences among mean values were tested by analysis-of-variance. Statistical analysis within groups was carried out by Student s t test. Each bar represents the mean S.D. of three experiments. *, p of full-length PrP in the ER, PrP C cells were pulsed for 30 min at 37 C, and chased for 4hatthesame temperature in the presence of brefeldin-a to block transport of proteins from the ER. The 20-kDa fragment is detected soon after the pulse, but does not accumulate with chase, indicating that it does not arise from full-length PrP C (data not shown). If the brefeldin block is removed and radiolabeled proteins are allowed to exit the ER, the 20-kDa fragment, along with the full-length PrP C forms, can be recovered from the cell surface by PI-PLC treatment after1hofchase, and follows similar kinetics of turnover as PrP C (data not shown). Together, the above results show that the 20-kDa fragment is not a proteolytic product of full-length PrP C in the ER. Instead, it originates from an alternate form of PrP, probably Ctm PrP, which is up-regulated in PrP treated cells. The 20-kDa fragment is probably generated because of spontaneous cleavage of Ctm PrP at the cytosolic face of the ER, and is transported along the secretory path to the plasma membrane. Aggregated PrP C Is Resistant to Digestion by Proteinase K Resistance to limited digestion by PK is considered the hallmark of PrP Sc, and is one of the primary diagnostic tests available at this time for identifying prion-infected tissue. Although the amount of PK used for evaluating PK-resistant PrP in cell models is orders of magnitude lower, and the conditions less harsh than the ones used to detect PrP Sc in the brain, this test allows an evaluation of a change in conformation of PrP C, an important step toward the final transformation to PrP Sc.To evaluate if aggregated PrP C generated in our cell model is PK-resistant, lysates of control and treated cells were exposed to 3 g/ml of PK for 0, 2, or 4 min. The protease-resistant proteins remaining in the lysate were precipitated with cold methanol and analyzed by immunoblotting with 3F4. In the control sample treated with scrambled peptide, PrP C is digested completely after 4 min of PK treatment, with only traces of the unglycosylated form remaining (Fig. 7, lane 1 versus 3). In contrast, a significant amount of PrP from cells treated with PrP is resistant to PK-digestion after 4 min at 37 C (Fig. 7, lane 5 versus 7). Deglycosylation of 2 min PK-treated samples with PNGase-F shows that the resistant species in treated cells comprises of full-length PrP C and a 20-kDa C- terminal fragment of PrP, both of which are at least 10-fold more in treated cells as compared with the control sample (Fig. 7, lane 4 versus 8). Because the PK-resistant fragment of PrP Sc has a ragged N terminus near residue 90 (21) and Ctm PrP is also cleaved at around residue 104 in intact microsomes (7), both fragments would have a similar molecular mass of 20 kda on SDS-PAGE, making it difficult to distinguish between the two. However, judging from the sensitivity of truncated Ctm PrP to PI-PLC cleavage and its solubility in non-ionic detergents (data not shown), it does not appear to be aggregated or PK-resistant. Thus, the PK-resistant 20-kDa fragment detected above represents the digested product of aggregated PrP C rather than truncated Ctm PrP. In addition, a considerable amount of full-length PrP is also resistant to mild protease digestion in the treated sample, probably because of its state of aggregation. DISCUSSION This report extends the current understanding of possible mechanisms of PrP Sc -induced aggregation of PrP C and consequent neurodegeneration in three distinct ways. First, we show that a -sheet-rich peptide of PrP catalyzes the aggregation of endogenous full-length PrP C in a cell model, confirming the nucleation hypothesis of PrP aggregation. Second, we demonstrate that intracellular accumulation of PrP aggregates leads to up-regulation of the synthesis of Ctm PrP, and finally, we show that majority of Ctm PrP accumulates as a C-terminal fragment on the cell surface. These results suggest that the active mediator of neurotoxicity in prion disorders is perhaps truncated Ctm PrP, which probably accentuates the neurotoxic

10 2284 Induction of Ctm PrP by PrP Sc -like Forms in a Cell Model FIG. 7.Aggregated PrP C is resistant to digestion by proteinase K. Lysates of cells treated with biotin-tagged scrambled PrP (Control) or biotin-tagged PrP (Treated) for 4 12 months were exposed to 3 g/ml PK for 0, 2, and 4 min at 37 C, and the 2-min sample was further subjected to deglycosylation with PNGase-F. Immunoblotting with anti- PrP antibody 3F4 shows almost complete digestion of PrP C in control lysates after 2 and 4 min of PK treatment (lane 1 versus lanes 2 and 3), whereas a significant amount of undigested PrP is present in peptide-treated lysates (lane 5 versus lanes 6 and 7). Deglycosylation of 2-min PK-digested samples shows an increased amount of 20-kDa fragment in peptidetreated cells (lane 4 versus 8). (3F4 is specific to PrP residues included in the PK-resistant fragment of PrP Sc.) effect of intracellular aggregated PrP by initiating the aggregation of additional PrP C molecules at the cell surface, or functions as a novel receptor for as yet unidentified factors that accelerate the neurodegenerative process. The Conformational Change of PrP C to PrP Sc This report is the first direct demonstration of a change in conformation of endogenous PrP C into an aggregated, partially PK-resistant form by an exogenously added -sheet rich peptide of PrP. Although the cellular and biochemical processes of PrP Sc propagation and neuronal toxicity remain contentious, substantial evidence indicates that transmission of PrP Sc occurs by inducing the conversion of host PrP C into the PrP Sc conformation. Two largely unsubstantiated hypotheses have been proposed to explain the mechanism of this conformational change: 1) template-assisted conversion or re-folding of endogenous PrP C to PrP Sc, and 2) the nucleation or seeding hypothesis (22, 23). We show that a -sheet-rich internal fragment of PrP, PrP seeds the aggregation of PrP C on the plasma membrane of cells, and catalyzes its accumulation in an exponential manner. We believe that PrP intercalates within the lipid bilayer, where the charged phospholipid environment induces its aggregation. Here, it acts as a seed for the deposition of additional PrP C molecules. Because -sheet rich peptides have an affinity for cholesterol (9, 24), aggregated PrP probably concentrates in cholesterol-rich lipid domains of the plasma membrane. Incidentally, this is also the preferred location for endogenous PrP C and the site for PrP C to PrP Sc conversion (25 27), thus providing an ideal environment for the interaction of PrP and PrP C, and subsequent aggregation of the latter. Following the initial aggregation of PrP C at the plasma membrane, the complex aggregates of PrP C and PrP are transported to the endosomal/lysosomal compartment, where low ph and the negatively charged membrane microenvironment probably plays a major role in promoting their transformation to thioflavin-s-binding, -sheet rich structures. Such an accumulation of virtually nondegradable PrP C and PrP aggregates explains the extensive proliferation of lysosomes observed in our cell model, and the presence of abundant PrP immunoreactive lysosomes in the neurons of scrapie-infected animals and new variant Creutzfeldt-Jakob disease patients (25, 28). A similar change in conformation from an -helical to a -sheet structure has been reported for recombinant PrP and for certain other peptides when exposed to an acidic ph (24, 29) and is consistent with the notion that lysosomotropic agents inhibit the accumulation and branched polyamines stimulate the degradation of PrP Sc in scrapie-infected cells (30, 31). The intracellular aggregates of PrP C thus generated display several of the biochemical characteristics typical of PrP Sc -detergent insolubility, partial resistance to digestion by proteinase K (PK), and amyloid-like characteristics as demonstrated by thioflavin-s binding. Some of the aggregated PrP C is insoluble even in SDS, and cannot be resolved by conventional SDS-PAGE. However, as opposed to conventional PrP Sc that can transmit disease, the apparently similar biochemical characteristics of PrP C aggregates observed in our cell model depict a particular conformational state of PrP C rather than infectivity per se. Whether this conformational state will eventually evolve into other PrP conformations leading to PrP Sc is unclear. It is interesting to note that clinical signs of prion disease have been produced in transgenic 196 mice injected with MoPrP peptide (89 143, P101L) without significant correlation with PK-resistant PrP Sc deposition (32), in part supporting the results obtained in our study. A similar aggregation of PrP C in primary human fetal brain cells exposed to PrP confirms our results in neuroblastoma cells, and suggests the possibility of a similar sequence of events during infection of animals and human beings exposed to exogenous PrP Sc infection. The aggregates of PrP and PrP C in brain cells are not only present in neuronal and glial cell bodies, but also in neuronal processes and axonal swellings. It is plausible that these aggregates travel along axons to neighboring cells by vesicular traffic, or, alternately, are extruded into the extracellular milieu and are subsequently internalized by other cells. Thus, intracellular PrP aggregates may be directly toxic to neurons, or alternately, the surrounding glial cells may release cytokines and other factors that participate in the observed neurotoxicity. Microglia cultured in the presence of A show similar

11 Induction of Ctm PrP by PrP Sc -like Forms in a Cell Model 2285 results. Aggregated A accumulates in lysosomes and is ultimately released from the cell without degradation (33). Because most PrP Sc in diseased tissue is likely sequestered in plaques and therefore unable to interact with neighboring cells, the propagation of PrP Sc may occur through fragments like PrP that are relatively soluble, resistant to proteases, and have the potential to fold into -sheet upon contact with the membrane micro-environment. The occurrence of such a phenomenon would explain the spread of PrP Sc infection in the brain in an exponential manner. Ctm PrP As a Mediator of Neurotoxicity by PrP Sc This study shows, for the first time, that intracellular accumulation of aggregated PrP C up-regulates the synthesis of Ctm PrP, the proposed mediator of neurodegeneration in certain inherited and infectious prion disorders (7, 8). Several lines of evidence support our assertion that the 20-kDa C-terminal fragment of PrP that accumulates on the surface of PrP treated cells arises from Ctm PrP. 1) PK treatment of intact microsomes prepared from peptide-treated cells shows a small but significant increase in the expected 20-kDa C-terminal fragment, and a further increase in peptide treated cells, confirming its transmembrane orientation with N terminus in the cytosol. In PrPexpressing Chinese hamster ovary and baby hamster kidney cells, 2% PrP has been reported to exist in the transmembrane orientation (34), consistent with the small increase in the 20-kDa fragment observed in our cell model. 2) There is at least a 4-fold increase in the surface expression of 20-kDa fragment in PrP treated cells that is similar in molecular mass and immunoreactivity to the fragment of Ctm PrP obtained after PK treatment. 3) The amount of 20-kDa and not full-length PrP increases in the presence of a proteasomal inhibitor, 2 consistent with its origin from Ctm PrP, which is normally degraded by the proteasomes (35). 4) The 20-kDa fragment is GPI-linked (as reported for Ctm PrP) and is transported to the plasma membrane within 1 h of synthesis. 2 5) This fragment does not accumulate with chase under conditions in which transport of PrP from the ER is blocked, arguing against its origin from full-length PrP. 2 6) Finally, in a separate cell model expressing mutant PrP(S231T), which lacks the GPI anchor and aggregates in the ER, a similar 20-kDa fragment is generated that, remarkably, contains an intact GPI anchor. 3 Because PrPS231T is anchorless, the GPI-linked 20 kda probably arises from Ctm PrP, which, because of its transmembrane orientation, is probably in a suitable conformation to receive the GPI anchor despite the mutation at codon 231. Overall, the evidence presented above strongly favors the possibility that the 20-kDa fragment is derived from Ctm PrP and not from fully translocated full-length PrP in the ER. Although the involvement of Ctm PrP in mediating neurotoxicity has already been demonstrated by an indirect method in mice infected with PrP Sc (8), this study is the first direct demonstration of a correlation between intracellular accumulation of PrP aggregates and up-regulation of Ctm PrP synthesis. The fact that majority of Ctm PrP is cleaved spontaneously on the cytosolic face of the ER and the truncated C-terminal fragment is expressed on the cell surface provides novel and important information about the metabolism of Ctm PrP and possible mechanisms of neuronal toxicity. A recent study suggests that Ctm PrP perhaps exerts its neurotoxic effect through cell membrane perturbation (36). However, truncated Ctm PrP in our cell model is GPI-linked and does not appear to be inserted through the transmembrane domain because it is released into the medium by PI-PLC treatment. Perhaps cleavage of the N- 2 Y. Gu and N. Singh, unpublished observations. 3 Y. Gu, et al., manuscript in preparation. FIG. 8.Proposed model of PrP C aggregation and induction of Ctm PrP. Step 1, micro-aggregates of PrP bind to the plasma membrane and initiate the aggregation of PrP C. Step 2, the aggregated proteins are endocytosed in large vesicular structures and transported to lysosomes. Step 3, the intracellular PrP aggregates induce the synthesis of Ctm PrP through trans-activating factors. Step 4, through an as yet unknown mechanism, the N-terminal region of Ctm PrP is cleaved by cytosolic enzymes. Step 5, the C-terminal 20-kDa fragment of Ctm PrP is transported to the cell surface through the secretory path. Step 6, the truncated Ctm PrP is inserted in the outer leaflet of the plasma membrane through the GPI anchor. Step 7, during the normal re-cycling of PrP, full-length PrP C is cleaved at residue 111 or 112, resulting in a truncated 18-kDa C-terminal fragment that accumulates on the cell surface (38). PM, plasma membrane; E, endosomes; L, lysosomes; N, nucleus. terminal domain of Ctm PrP at the ER membrane destabilizes its association with the membrane at nearby residues of PrP, leaving GPI anchor as the sole membrane linkage of the 20-kDa truncated Ctm PrP. Moreover, because Ctm PrP has been reported to partition into the aqueous phase of Triton X-114 partitioning after cleavage of the GPI anchor (34), the transmembrane domain does not appear to be hydrophobic enough to maintain membrane association by itself. Thus, truncated Ctm PrP must use an alternative mechanism for neurotoxicity in our cell model. Because this fragment includes the amyloidogenic region of PrP, comprising residues , that is normally disrupted during recycling of PrP from the cell surface (38), the truncated Ctm PrP may initiate aggregation of additional PrP C molecules at the cell surface by acting as a nidus, or it may function as a novel surface receptor. Although the precise mechanism by which Ctm PrP mediates neurotoxicity is not clear from our data, this study demonstrates a direct correlation between aggregation of PrP C and induction of Ctm PrP and clarifies the downstream pathway of Ctm PrP transport and metabolism, thus laying the groundwork for future investigations on the cellular pathways of Ctm PrP-mediated neurodegeneration. Although truncated Ctm PrP appears similar in molecular mass to the mildly PK-resistant fragment generated from detergent-solubilized lysates of peptide-treated cells, the two fragments differ in important aspects. First, the limited PK resistance of PrP observed in the presence of non-ionic detergents reflects a change in the protease susceptibility of PrP C because of its conformational transition to PrP Sc. Truncated Ctm PrP, on the other hand, results from a proteolytic clip of transmembrane PrP in the ER and is unlikely to resist PK treatment in the presence of detergent. Second, protease-resistant PrP Sc accumulates in intracellular compartments, whereas truncated Ctm PrP is present on the cell surface and is releasable by PI-PLC treatment. Thus, the PK-resistant 20-