Disease-related Prion Protein Forms Aggresomes in Neuronal Cells Leading to Caspase Activation and Apoptosis*

The molecular basis for neuronal death in prion disease is not established, but putative pathogenic roles for both disease-related prion protein (PrPSc) and accumulated cytosolic PrPC have been proposed. Here we report that only prion-infected neuronal cells become apoptotic after mild inhibition of the proteasome, and this is strictly dependent upon sustained propagation of PrPSc. Whereas cells overexpressing PrPC developed cytosolic PrPC aggregates, this did not cause cell death. In contrast, only in prion-infected cells, mild proteasome impairment resulted in the formation of large cytosolic perinuclear aggresomes that contained PrPSc, heat shock chaperone 70, ubiquitin, proteasome subunits, and vimentin. Similar structures were found in the brains of prion-infected mice. PrPSc aggresome formation was directly associated with activation of caspase 3 and 8, resulting in apoptosis. These data suggest that neuronal propagation of prions invokes a neurotoxic mechanism involving intracellular formation of PrPSc aggresomes. This, in turn, triggers caspase-dependent apoptosis and further implicates proteasome dysfunction in the pathogenesis of prion diseases.

The molecular basis for neuronal death in prion disease is not established, but putative pathogenic roles for both disease-related prion protein (PrP Sc ) and accumulated cytosolic PrP C have been proposed. Here we report that only prion-infected neuronal cells become apoptotic after mild inhibition of the proteasome, and this is strictly dependent upon sustained propagation of PrP Sc . Whereas cells overexpressing PrP C developed cytosolic PrP C aggregates, this did not cause cell death. In contrast, only in prion-infected cells, mild proteasome impairment resulted in the formation of large cytosolic perinuclear aggresomes that contained PrP Sc , heat shock chaperone 70, ubiquitin, proteasome subunits, and vimentin. Similar structures were found in the brains of prion-infected mice. PrP Sc aggresome formation was directly associated with activation of caspase 3 and 8, resulting in apoptosis. These data suggest that neuronal propagation of prions invokes a neurotoxic mechanism involving intracellular formation of PrP Sc aggresomes. This, in turn, triggers caspase-dependent apoptosis and further implicates proteasome dysfunction in the pathogenesis of prion diseases.
Prion diseases are rare fatal neurodegenerative disorders, which include Creutzfeldt-Jakob disease in humans, bovine spongiform encephalopathy (BSE) 2 in cattle and scrapie in sheep. The molecular hallmark of these disorders is the accumulation of abnormal prion protein conformers (PrP Sc ) derived from normal cellular host prion protein (PrP C ) (1). The cause of neurodegeneration in these disorders is not well understood, and a major gap exists in the understanding of how the conversion of PrP C to PrP Sc ultimately kills neurons. Whereas PrP C is absolutely required for prion conversion and neurotoxicity (2), knockout of PrP C in adult brain (3) and in embryonic models (4,5) has no overt phenotypic effect, effectively excluding loss of PrP C function in neurons as a significant mechanism in prion neurodegeneration. However, there is much evidence that also argues against the direct neurotoxicity of PrP Sc or prions (whether or not they are identical). PrP C -null tissue remains healthy and free of pathology when exposed to PrP Sc (6,7), and there is no direct correlation between neuronal loss and PrP Sc plaques in Creutzfeldt-Jakob disease brains (8). Similarly, prion diseases in which PrP Sc is barely detectable have been described (9 -11), and subclinical infection where high levels of PrP Sc accumulate in the absence of clinical symptoms are also recognized (12). In such "subclinical disease states," the majority of the accumulated PrP Sc may be inert; alternatively, PrP Sc may not be the toxic entity, but instead a toxic oligomeric PrP intermediate species (PrP L for lethal) may be produced during prion conversion (12). Either this intermediate species or PrP Sc itself may then only elicit neurotoxic effects when present at sufficient concentrations in particular subcellular compartments.
Various mechanisms have been proposed to explain neuronal death in prion disease (reviewed in Ref. 13), which is thought to occur via an apoptotic mechanism (14 -16). In vitro studies have suggested that both full-length PrP Sc (16,17) and shorter peptide fragments (18) are toxic when applied to primary cultured neurons. Other mechanisms suggested relate to altered PrP C trafficking. It has been described that PrP C may assume a transmembrane topology ( Ctm PrP), the concentration of which has been suggested to correlate with neurotoxicity (19). More recently, it was proposed that prion-associated toxicity involves altered trafficking of PrP C , where inhibition of the ubiquitin-proteasome system (UPS) results in extensive PrP C accumulation in the cytoplasm and associated neuronal cell death (20). However, the data is conflicting, with evidence both for (21,22) and against (23)(24)(25) this cytoplasmic accumulation of PrP C having neurotoxic sequelae. One of the major drawbacks of many of these studies on cytosolic PrP C is the high levels of proteasome inhibition used, which may limit any physiological relevance to the situation in vivo (26).
The concept that UPS inhibition may contribute to neurodegeneration is not new. Degradation of intracellular proteins via the UPS is a highly complex and tightly regulated process that plays a major role in a variety of cellular processes (27). Aberrations in this system have been implicated, either as a primary or secondary event, in the pathogenesis of a range of neurodegenerative diseases including Parkinson disease, Alzheimer disease, and Huntington disease (28). The degradative capacity of the UPS in the nervous system is known to become impaired in these neurodegenerative diseases as well as during the aging process itself (29) (reviewed in Ref. 48).
We therefore set out to examine further the nature of the PrP species responsible for neurotoxicity and to what extent low level or "physiological" UPS inhibition may be involved in prion disease pathogenesis. We chose to study these effects both in cell models of prion infection as well as examining overexpression of native PrP C . We found that whereas neuronal cells overexpressing PrP C developed cytosolic PrP C aggregates under conditions of mild proteasome inhibition, this did not cause cell death. However, under similar conditions, we found that neuronal propagation of prions invokes a neurotoxic mechanism involving intracellular formation of compartmentalized cytosolic PrP Sc aggresomes that triggers caspase-dependent apoptosis and implicates protea-some dysfunction in the pathogenesis of prion diseases. The aggresome has emerged as a key organelle in the clearance of toxic cytoplasmic misfolded protein aggregates (30). Interestingly, we found evidence for similar structures in vivo in brains of prion-infected mice.

MATERIALS AND METHODS
Cell Culture and Scrapie Infection-GT-1 and N2aPD88 mouse neuronal cells were infected with mouse-adapted RML scrapie prions or wild-type CD-1 mouse brain homogenate as described (31,32) and passaged to remove the initial brain inoculum. Cells were cultured in Opti-MEM (Invitrogen) with 10% fetal calf serum supplemented with 1% penicillin/streptomycin and maintained at 37°C in 5% CO 2 . Cultures were tested for the presence of newly generated PrP Sc by the scrapie cell assay (SCA) (32). The SCA was used to determine the percentage of infected cells (32). Cells were diluted in duplicate so that no more than 1000 cells were seeded into one well of a scrapie cell assay plate. To determine the number of infected cells, the assay was developed using the standard color reaction as described (32). To determine the total number of cells, the assay was developed using a 1:10 dilution of trypan blue as an indicator of cell viability. The percentage of infected cells was calculated as the number of infected cells expressed as a percentage of the total number of cells. Both ScN2aPD88 and ScGT-1 cells were cured of PrP Sc with 0.5 g/ml anti-PrP antibody ICSM18 (D-Gen Ltd., London, UK) for 14 days. Confirmation of clearance of PrP Sc was determined using the scrapie cell assay as described (32).
Generation of N2a moPrPC Cells-Exon 3 (bp 7-1316) encoding the full open reading frame of mouse prnp was cloned into the NotI-ClaI sites of the retroviral vector LNCX2 (Clontech). This vector was then packaged into the GP-E86 line using Fugene 6 (Roche Applied Science) and selected using G418. Viral supernatant was used to infect N2aPD88 and GT-1 cells with 4 g/ml polybrene (Sigma). At 24 h after retroviral transduction, stable exon 3 moprnp-expressing clones were selected using G418. Expression levels were quantified by immunoblotting as described below.
Cell Death and Apoptosis Assays-For cell death studies, dose-response curves were established using proteasome inhibitors, and concentrations causing ϳ20% cell death in wild-type cells were chosen. Analysis of cell death using the lactate dehydrogenase kit was as recommended by the manufacturer (Alexis, Nottingham, UK). Quantification of apoptosis using annexin V and propidium iodide staining and caspase 3 and 8 assays was performed according to the manufacturer's instructions (Oncogene). For caspase inhibition experiments, cells were preincubated for 2 h with benzyloxycarbonyl-DEVD-fmk or benzyloxycarbonyl-IETD-fmk before the addition of proteasome inhibitors. Annexin V-FITC binding was quantified using flow cytometry (fluorescenceactivated cell sorting). Cells were treated with 1 M lactacystin for 24 h and then harvested by trypsinization. A total of 10,000 cells/sample were analyzed for cell death by a FACSCalibur (BD Biosciences) with the Cell Quest software.
SDS-PAGE and Immunoblot Analysis-Cells were harvested, and brain tissue was homogenized on ice in PBS, freeze-thawed three times, and treated with benzonase (50 units/ml) to digest DNA. Protein concentration was determined by BCA assay (Pierce). The equivalent of 25 g of total protein was loaded onto 16% SDS-PAGE minigels (Novex, Paisley, UK) and analyzed by electrophoresis and immunoblotting as described (33). For proteinase K (Roche Applied Science) digestion, lysates were incubated with proteinase K at 1 g/mg protein (cells) or 5 g/mg protein (brain) at 37°C for 90 min. Equivalent protein loading was confirmed in non-proteinase K-treated lanes by stripping membranes and reprobing with anti-␤-actin antibody (Sigma).
Immunofluorescence and Confocal Analysis-Cells were fixed onto poly-L-lysine-coated glass coverslips using 4% paraformaldehyde for 20 min at room temperature, washed three times with PBS, and then permeabilized in methanol at Ϫ20°C for 15 min. Cells were then incubated in 10% normal goat serum for 30 min. Incubation with primary antibody was at 37°C for 1 h. After washing, cells were incubated for 45 min with the appropriate secondary antibody at 37°C, washed several times in PBS, and mounted in Antifade (Sigma) containing 1 g/ml 4Ј,6-diamidino-2-phenylindole (DAPI; Sigma). To remove PrP C and reveal PrP Sc , cells were exposed to 98% formic acid for 5 min after fixing and before permeabilization. For confocal analysis, images were obtained using an LSM510 confocal microscopy system (Zeiss). A ϫ63 oil immersion objective was used for all imaging. For some experiments, ICSM18 was conjugated to a fluorescent Alexa-488 FITC secondary antibody (Molecular Probes, Inc., Eugene, OR). Details of all antibodies are shown in supplemental Table S1.
Subcellular Fractionation and PrP Analysis-Analysis of PrP solubility and aggregation was performed with modifications to published procedures (20,34). Briefly, after 24 h of lactacystin treatment, cells were harvested, washed in ice-cold PBS, freeze-thawed three times, and then centrifuged at 1000 ϫ g for 10 min at 4°C to remove cellular debris. The supernatant was collected, adjusted with an equal volume of ϫ2 lysis buffer (100 mM Tris, pH 7.4, 300 mM NaCl, 4 mM EDTA, 1% Triton X-100, 1% deoxycholate), and incubated with benzonase (50 units/ml) for 20 min at 4°C before centrifugation at 100,000 ϫ g for 45 min. Proteins in the supernatant were precipitated at Ϫ20°C with methanol, air-dried, and then boiled in ϫ2 SDS-sample buffer, whereas proteins in the pellet fraction were boiled in ϫ2 SDS-sample buffer. Each fraction was analyzed by immunoblotting as described above.
Affinity Purification of PrP Sc Aggresomes-Magnetic tosyl-activated beads were coated with either mouse monoclonal antibody to vimentin or Brc126, an IgG isotype control antibody, according to the manufacturer's instructions (Dynal, Bromborough, UK). ScGT-1 cells (with or without proteasome inhibition) were harvested in ice-cold PBS, freezethawed three times before treatment with benzonase (50 units/ml) on ice for 15 min. Equivalent aliquots of sample (10 l containing 25 g of total cell protein) were incubated with vimentin antibody-coated beads, Brc126-coated beads, or magnetic beads alone (25-l bead bed volume) for 2 h at 37°C on an orbital shaker. Beads were concentrated in a magnetic particle concentrator (Dynal) and washed three times with PBS (3 ϫ 5 min, 500 l of PBS). Washed beads were resuspended in 20 l of PBS and analyzed after proteinase K digestion (final protease concentration of 1 g/mg protein, 90 min, 37°C) or in the absence of protease digestion. Beads were treated with 2ϫ sample buffer at 100°C for 10 min, and the supernatant was analyzed by immunoblotting with biotinylated anti-PrP monoclonal antibody ICSM35 (D-Gen). For analysis of brain, 10% brain homogenate was prepared in PBS and freezethawed three times before digestion with benzonase (50 units/ml on ice, 15 min) and centrifugation at 1000 ϫ g for 10 min to remove cellular debris. Aliquots of supernatant (25 l) were adjusted with an equal volume of 2ϫ lysis buffer (100 mM Tris, pH 7.4, 300 mM NaCl, 4 mM EDTA, 1% Triton-X-100, 1% deoxycholate) and incubated with either vimentin antibody-coated beads, Brc126 antibody-coated beads, or magnetic beads alone (25-l bead bed volume) for 2 h at 37°C on an orbital shaker. Beads were concentrated in a magnetic particle concentrator (Dynal) and washed three times with PBS (3 ϫ 5 min, 500 l of PBS). Washed beads were resuspended in 20 l of PBS and analyzed after proteinase K digestion (final protease concentration 5 g/mg protein, 90 min, 37°C) or in the absence of protease digestion. Beads were treated with 2ϫ sample buffer at 100°C for 10 min, and the supernatant was analyzed by immunoblotting with biotinylated anti-PrP monoclonal antibody ICSM35.
Image Acquisition-Fluorescence images were obtained using a confocal microscope (Zeiss microscope LSM510 META) equipped with a "plan-Apochromat" ϫ63/1.40 oil differential interference contrast objective at room temperature and is controlled by Zeiss LSM software. Fluorescence was recorded at 488 nm using a 30-milliwatt argon laser for excitation or at 543 nm using a 1-milliwatt HeNE laser for excitation. Zeiss Immersol TM 518 F was used as imaging medium. Images not requiring confocal analysis were obtained using an Axioplan 2 MOT microscope (Zeiss) with filters for fluorescein isothiocyanate, rhodamine, and DAPI and Plan Neofluar ϫ10/0.30 Ph1 objective at room temperature. An AxioCam MRm (Zeiss) camera was used and was controlled using the Axiovision Control software (Zeiss).

PrP Sc Infection Sensitizes both GT-1 and N2aPD88
Cells to Mild Proteasome Inhibition-Mouse hypothalamic neuronal GT-1 (31) and highly prion-susceptible N2aPD88 cells (32) were infected with mouseadapted RML scrapie prions or wild-type CD-1 brain homogenate and passaged to remove the brain inoculum (31, 32) (Fig. 1A). Cells mockinfected with wild-type CD-1 brain homogenate were negative on the SCA, indicating that they were not prion-infected. 3 Cells were then treated with a range of doses of the irreversible proteasome inhibitor lactacystin (Fig. 1B, graph i). Significant differences in cell death were observed. At 1 M lactacystin, a highly significant difference (p Ͻ  4), wild type mouse neuroblastoma N2aPD88 cells (lanes 5 and 6), and RML scrapie-infected N2aPD88 cells (lanes 7 and 8) were incubated in the absence (Ϫ) or presence (ϩ) of proteinase K and immunoblotted using anti-PrP antibody ICSM35 to demonstrate PrP Sc infection in these cells. B, i, dose-response curves for proteasome inhibition by lactacystin (LAC) in GT-1 cells, ScGT-1 cells, or ScGT-1-18 cells. Cell death was determined after 24 h by measuring lactate dehydrogenase release (n ϭ 12). B, ii, the percentage of cell death after mild LAC treatment (1 M) is significantly different (p Ͻ 0.0001; n ϭ 12) in ScGT-1 cells compared with uninfected GT-1 cells or antibody-cured ScGT-1-18 cells. C, i, dose-response curves for proteasome inhibition by lactacystin in N2aPD88 cells, ScN2aPD88 cells, or antibody-cured ScN2aPD88-18 cells. Cell death was determined after 24 h by measuring lactate dehydrogenase release (n ϭ 12). C, ii, the percentage of cell death after mild LAC treatment (1 M) is significantly different (p Ͻ 0.0001; n ϭ 12) in ScN2aPD88 cells compared with uninfected N2aPD88 cells or antibody-cured ScN2aPD88-18 cells. Analysis by the SCA showed that ϳ52 and ϳ42% of the ScGT-1 and ScN2aPD88 cell populations were scrapie-infected, which is consistent with the percentages of cell death seen in B (ii) and C (ii). The prefix Sc denotes scrapie infection; the suffix -18 indicates the clonal line that has been cured of prion infection with anti-PrP antibody ICSM18. 0.0001) was observed in cell death in the ScGT-1 cells (52%) compared with uninfected GT-1 cells (17%) (Fig. 1B, graph ii). Mock-infected GT-1 cells resulted in levels of cell death similar to those of wild-type GT-1 cells, confirming the specificity of prion infection in sensitizing cells to mild proteasome inhibition (supplemental Fig. S1). This concentration of lactacystin was selected to mimic the degree of proteasome impairment that may occur in vivo (26,35). These results were reproduced with another specific proteasome inhibitor, epoxomicin (supple- mental Fig. S2). To ensure that the effects we observed were not confined to a subgroup of ScGT-1 cells known to have an apoptotic phenotype after scrapie infection (31), experiments were repeated using scrapie-infected clonal N2aPD88 cells (32). Again, RML-infected N2aPD88 cells were significantly more susceptible to induction of apoptosis after mild proteasome inhibition compared with uninfected N2aPD88 cells (p Ͻ 0.0001) (Fig. 1C, graphs i and ii). Analysis by the SCA (32) showed that ϳ52 and ϳ42% of the ScGT-1 and ScN2aPD88 cell populations were scrapie-infected, 3 which is consistent with the percentages of cell death seen in Fig. 1B (graph ii) and Fig. 1C (graph ii).
Curing Prion-infected Cells with Anti-PrP Monoclonal Antibodies Abrogates the Neurotoxic Effect of Proteasome Inhibition-To investigate whether the sensitivity of scrapie-infected cells was due to PrP Sc , cells were treated for 14 days with 0.5 g/ml anti-PrP monoclonal antibody ICSM18 (36) and confirmed to have undetectable levels of infectivity using the SCA. 3 Curing cells of prion infection abrogated the sensitivity to proteasome inhibition and resulted in the same degree of cell death as uninfected GT-1, mock-infected GT-1, and N2aPD88 cells (Fig. 1, B and C, and supplemental Fig. S1). Thus, prion propagation appears to sensitize these neuronal cells to mild proteasome inhibition.
Cytosolic Accumulation of Detergent-insoluble PrP C Aggregates Is Not Neurotoxic after Mild Proteasome Inhibition-To further investigate whether our findings were due to PrP Sc , we generated N2aPD88 cells overexpressing ϳ3-fold full-length wild-type mouse PrP C (N2a moPrPC ) ( Fig. 2A). In non-lactacystin-treated N2a moPrPC cells, the PrP C was localized on the cell surface, in the lysosomal system, and in an ER-Golgi pattern with partial ER co-localization (63) (Fig. 2B). After low dose lactacystin treatment, the pattern of PrP C immunostaining changed in the N2a moPrPC cells, with the majority of PrP C deposition now in the cytoplasm co-localizing with the cytosolic protein Hsc70 (Fig. 2C).
However, despite the presence of cytosolic PrP C , there was no significant difference in cell death when compared with wild-type N2aPD88  NOVEMBER 18, 2005 • VOLUME 280 • NUMBER 46 cells (Fig. 2D, graph i), suggesting that at low level proteasome inhibition, cytoplasmic accumulation of PrP C is not neurotoxic. Notably, in our N2a moPrPC cells, there appeared to be a neuroprotective effect of PrP C overexpression (p Ͻ 0.001) after treatment with 10 M lactacystin compared with wild-type N2aPD88 cells (Fig. 2D, graph ii) as previously reported (24,25). At high doses of lactacystin (10 M), there were high levels of cell death in both the scrapie-infected and uninfected N2aPD88 (ϳ90% in ScN2aPD88 and ϳ75% in N2aPD88) (Fig. 1C, graph i) in agreement with previous studies (20). To assess whether the cytosolic PrP C observed in our lactacystin-treated cells had formed detergentinsoluble aggregates, we performed subcellular fractionation with analysis of detergent solubility and aggregation status of PrP C in cytosolic and membrane fractions (Fig. 2E). With no proteasome treatment, all of the PrP C expressed in our N2a moPrPC was fully detergent-soluble (Fig.  2E, lane 1); after low dose lactacystin treatment (1 M), there was an increase in detergent-insoluble aggregated PrP C isolated in the pellet fraction (Fig. 2E, lane 4). After high dose lactacystin treatment (10 M), all of the PrP C was aggregated and detergent-resistant (Fig. 2E, lane 6).

PrP Sc Forms Apoptosis-inducing Aggresomes
We then infected our N2a moPrPC cells with RML scrapie prions and observed correlation between the presence of PrP Sc and neurotoxic effect after 1 M lactacystin treatment (Fig. 2F). Thus, the presence of PrP Sc , rather than cytoplasmic aggregates of wild-type PrP C , appears to be associated with apoptosis after mild proteasome inhibition.

Prion Infection Induces Caspase 3-and 8-dependent Apoptosis, Which Is Abrogated by Specific Caspase
Inhibitors-To evaluate cell death after mild proteasome inhibition, nuclear DNA fragmentation analysis and annexin V and propidium iodide staining were performed in lactacystin-treated ScGT-1 cells, and apoptosis was quantified using fluorescence-activated cell sorting analysis. These results demonstrated that the ScGT-1 cells were dying by apoptosis (Figs. 3, A and B). Apoptosis may be initiated through a number of different pathways, and in vivo studies in prion disease have suggested that caspase 3-and 8-dependent pathways are activated (37,38). To study this process further, a time course study of caspase 3 and 8 activation was undertaken. From 1 h, there was a highly significant rise in caspase 3 and 8 activities in the ScGT-1 cells versus uninfected cells in a time-dependent manner (Fig. 3, C and D). At 24 h, there was a 120% increase in caspase 8 activation compared with uninfected GT-1 cells (Fig. 3D); this finding supports in vivo data suggesting that caspase 8-mediated apoptotic cell death plays a significant role in prion-mediated neuronal cell death (39). Apoptotic cell death was completely abrogated in the scrapie-infected cells using cell-permeable specific caspase 3 and 8 inhibitors (DEVD-fmk and IETD-fmk, respectively), supporting their pivotal role in scrapie-mediated neuronal cell death (Fig. 3E).
PrP Sc , but Not Aggregated PrP C , Forms Large Cytoplasmic Perinuclear Aggresomes, Which Appear Directly Neurotoxic-Aggresomes are located near the microtubule-organizing center at the centrosome (40), reflecting the fact that aggresome formation needs an intact microtubule network (40,41). They are also distinguished by a cage of the intermediate filament protein vimentin, which is an invariant feature of these structures (40,42). We used stringent formic acid pretreatment of cells to remove PrP C immunoreactivity and to reveal PrP Sc deposits in our scrapie-infected cells (Fig. 4A). Double-labeling immunostaining demonstrated that after mild proteasome inhibition, PrP Sc accumulates in ScGT-1 cells as large cytoplasmic perinuclear aggresomes and colocalizes with vimentin, Hsc70, 20 S proteasome, and ubiquitin (Fig. 4,  B-E). Similar structures were found in lactacystin-treated ScN2aPD88 cells (supplemental Fig. S3). Using confocal microscopy, cytoplasmic localization was confirmed by colocalization with the cytosolic chaper-one Hsc70 (Fig. 4D) and the absence of immunostaining with markers for the ER and nucleus. 3 To confirm that PrP Sc itself was a major constituent of these aggresomes, we performed an affinity purification of the ScGT-1 aggresomes using vimentin antibody-coated magnetic beads. Vimentin is a type-III intermediate filament protein that normally displays an extended cytoplasmic distribution. In aggresome-containing cells, vimentin is redistributed to form a cagelike structure wrapped around the exterior of the inclusion (41); it has been suggested that this contributes to the stability of the aggresome (42). Vimentin antibody-coated beads purified PrP Sc from lactacystin-treated ScGT-1 cells (Fig. 5A, lane 2), the specificity of this interaction was confirmed using isotype control antibody-coated beads that did not purify PrP Sc (Fig. 5A, lanes 3 and 4) and beads alone (Fig. 5A, lanes 5 and 6). Previous reports have suggested that cytosolic PrP C forms aggresomes after cyclosporin A treatment (43); we therefore performed immunostaining for aggresomes in our N2a moPrPC cells. Importantly, we found that cytoplasmic PrP C aggregates did not form aggresomes. 3 Mock-infected cells also did not form aggresomes and were indistinguishable from wild-type uninfected cells. 3 The role of aggresomes in cellular neurotoxicity is controversial; their formation in cells has been reported to be a neuroprotective mechanism to sequester toxic misfolded proteins (44), whereas others suggest that they are a toxic species (45,46). To investigate this further, we used agents that inhibit the formation of aggresomes by disrupting retrograde microtubule-mediated transport (47). Prior to treatment with colchicine, nocodazole, and cytochalasin D, we undertook dose-response curves in N2aPD88 and GT-1 cells to optimize treatment of cells with these agents (supplemental Fig. S4). Colchicine is an antimitotic agent that disrupts microtubule function. Treatment with colchicine (5 g/ml) prevented cell death in prion-infected cells after low proteasome inhibition (p Ͻ 0.0001) (Fig. 5B, graph i). To ensure that colchicine treatment had also prevented aggresome formation, we performed immunofluorescence, which confirmed that the prevention of PrP Sc aggresome formation had abrogated cell death (Fig. 5C, panels 1 and 2). We also examined the effect of nocodazole treatment on the formation of PrP Sc aggresomes; nocodazole is an agent that also disrupts microtubule dynamics (47). This supported the argument that the effect of colchicine treatment was via this mechanism, since cell death was also abrogated by 0.5 M nocodazole treatment (p Ͻ 0.0001) (Fig. 5B, graph ii) with prevention of aggresome formation (Fig. 5C, panel 3). To ensure that the abrogation of cell death induced by PrP Sc aggresomes by the microtubule-disrupting agents was specific, we treated our cells with 50 ng/ml cytochalasin D, which disrupts actin microfilaments that are not involved in aggresome formation (47). Treatment with cytochalasin D did not affect cell death (Fig. 5B, graph iii) or remove aggresomes (Fig.  5C, panel 4), confirming our data indicating that clearance of PrP Sc aggresomes selectively abrogates cell death. To ensure that colchicine and nocodazole were not exerting their antiapoptotic effect through clearance of PrP Sc , we treated ScGT-1 cells for 5 days with these drugs, which had no effect on PrP Sc levels in these cells assessed using the SCA. 3 Formation of PrP Sc Aggresomes Is Temporally Associated with Caspase 3 and 8 Activation-To examine whether aggresome formation is directly related to caspase activation, we measured caspase 3 and 8 activities after treatment with colchicine and nocodazole to prevent aggresome formation. This treatment also abrogated caspase activation in prion-infected cells (Fig. 6A), suggesting a direct relationship between the formation of PrP Sc aggresomes and caspase activation leading to apoptosis. We then performed a time course analysis of PrP Sc aggresome formation and caspase 3 and 8 activation in ScGT-1 cells after mild proteasome inhibition (Fig. 6, B-F, panels 1-3). At time 0, there are no PrP Sc aggresomes (Fig. 6B, panel 3) and no evidence of caspase 8 (Fig. 6B, panel 1) or caspase 3 activation (Fig. 6B, panel 2).
PrP Sc aggresome formation initiates as small perinuclear structures at 6 h (Fig. 6C, panel 3), which directly correlates with the presence of de novo caspase 8 and 3 immunostaining in the cells at the same time point  NOVEMBER 18, 2005 • VOLUME 280 • NUMBER 46  2 and 3). Treatment with cytochalasin D did not clear aggresomes (panel 4), confirming our data indicating that clearance of PrP Sc aggresomes selectively abrogates cell death. PrP Sc -containing aggresomes were detected using anti-PrP antibody ICSM18 (green) and anti-vimentin antibodies (red). Nuclear staining was done with DAPI (blue). Scale bar, 20 m. (Fig. 6C, panels 1 and 2). After 12 h, the size of the PrP Sc aggresome increases (Fig. 6D, panel 3) with marked caspase 8 and 3 immunostaining in the cells containing PrP Sc aggresomes (Fig. 6D, panels 1 and  2). By 18 and 24 h, there is widespread intracellular caspase 8 and 3 immunostaining, and large perinuclear PrP Sc aggresomes are seen (Fig.  6, E and F, panels 1-3). These data are supported by the time course study of caspase 3 and 8 activation in ScGT-1 cells (Figs. 3, C and D), where from 6 h onward there was a significant increase in caspase 3 and 8 activity levels (p Ͻ 0.001; Figs. 3, C and D). These data support a direct relationship between the formation of PrP Sc aggresomes and caspase 8and 3-dependent apoptosis in neuronal cells after mild proteasome inhibition.

PrP Sc Forms Apoptosis-inducing Aggresomes
PrP Sc Is Also Associated with Aggresome-like Structures in Vivo-To assess whether PrP Sc aggresome structures occurred in vivo, we attempted to affinity-purify PrP Sc from terminal scrapie-infected CD-1 mice brains using vimentin antibody-coated magnetic beads (Fig. 7, A  and B). We demonstrate that there is a specific association between PrP Sc and the intracellular protein vimentin in vivo, since we were able to affinity-purify PrP Sc from scrapie-infected CD-1 mouse brain using vimentin antibody-coated beads (Fig. 7A, lane 3). Using densitometry and quantitative immunoblotting, the estimated proportion of total brain PrP Sc recovered by the vimentin beads is ϳ4%, which represents the intracellular PrP Sc associated with aggresomes in these terminal scrapie-infected mouse brains (Fig. 7B, lane 7). Isotype control antibody or beads alone did not isolate PrP Sc (Fig. 7A, lanes 1 and 4). There was no association between PrP C and vimentin in uninfected CD-1 brain, again  show that no PrP immunoreactivity is recovered from uninfected CD-1 mouse brain homogenate by co-precipitation with vimentin antibody-coated magnetic beads. B, quantification of PrP Sc isolated from RML-infected CD-1 mouse brain. Lanes 1-5 show the amount of PrP Sc present in 12.5, 6.25, 3.13, 1.56, and 0.78 l of 10% RMLinfected brain homogenate, respectively. Lane 7 shows the amount of PrP Sc recovered by vimentin antibody-coated beads from 25 l of 10% RML-infected brain homogenate. Densitometry indicates recovery of ϳ4% of total PrP Sc , representing intracellular PrP Sc bound to intracellular vimentin.
confirming the specificity of the interaction with PrP Sc and vimentin (Fig. 7A, lanes 5 and 6).

DISCUSSION
This study aimed to define further the cellular basis of neurotoxicity in prion-mediated neuronal death and the subcellular compartments in which toxicity may be generated. To investigate the role of the UPS in prion-mediated toxicity, we studied much milder levels of proteasome inhibition than reported in previous studies (20 -22, 34). Levels of proteasome impairment that we investigated are more compatible with the loss of proteasome activity associated with either senescence (45,48) or that may be seen in prion-infected brain (26,35).
We chose to study two separate mouse prion-propagating neuronal cell lines (N2aPD88 and GT-1) (31,32) to allow validation of the data in different prion-infected cell systems. There are very few cell lines able to stably propagate prions in vitro, and these two neuronal cell systems are well characterized and represent a valuable tool for analysis (49). We confirmed that both the N2aPD88 and GT-1 cells are able to propagate low levels of scrapie infectivity and suffer no obvious cytotoxic effects (31,32,50,51). This may be due to possible cell-specific properties of enhanced degradative capacity, where accumulation of PrP Sc does not occur to a level where it may become neurotoxic (52), which may account for the reason these neuronal cells are uniquely able to stably propagate low levels of mouse prions.
Here we show that prion-infected N2aPD88 and GT-1 neuronal cells were significantly more susceptible to cell death when treated with low dose proteasome inhibitors than uninfected or mock-infected cells. These cells underwent caspase 3-and 8-dependent apoptotic cell death that was abrogated by specific caspase 3 and 8 inhibitors. Curing these cells of prion infection with an anti-PrP antibody abrogated neurotoxicity; however, when the same cells were reinfected with prions, apoptosis occurred under conditions of mild proteasome inhibition. Therefore, neurotoxicity was dependent on continued PrP Sc propagation. To investigate whether apoptosis was due to a nonspecific cellular proteinopathy, we studied N2aPD88 cells overexpressing PrP C . Under the same low level of proteasome inhibition, these cells developed large cytoplasmic PrP C aggregates but did not undergo apoptosis. Neurotoxicity occurred only when these PrP C -overexpressing cells were infected with prions, arguing that prion infection was a prerequisite for apoptosis under these conditions. Under conditions of mild proteasome impairment, both prion-infected cell lines accumulated large cytoplasmic perinuclear aggresomes containing PrP Sc , heat shock protein 70, ubiquitin, proteasome subunits, and vimentin, characteristic of these structures. PrP Sc aggresome formation was temporally associated with caspase 3 and 8 activation and subsequent apoptosis. Inhibition of aggresome formation with different microtubule inhibitors abrogated both caspase activation and cell death, indicating that aggresome formation triggers apoptosis. Importantly, PrP Sc was associated with vimentin in RML prion-infected mouse brain, suggesting that similar PrP Sc aggresome structures may have relevance in vivo. Recently, granular deposits of disease-related prion protein have also been reported in the cell body of neurons, suggesting intraneuronal prion aggregates may play a role in Creutzfeldt-Jakob disease pathogenesis (53).
Our data suggest a neurotoxic mechanism in prion disease where formation of intraneuronal cytosolic PrP Sc -containing aggresomes is associated with caspase 3-and 8-dependent apoptosis. They support a role for UPS dysfunction in the neuropathogenesis of prion disease but not a role for cytosolic aggregation of wild-type PrP C . Whereas we confirm in this study that high levels of proteasome inhibition (20) can result in accumulation of misfolded cytosolic PrP C and resultant neurotoxicity, such a degree of proteasome inhibition is unlikely to occur in vivo during prion pathogenesis (26,45,48).
It has been proposed that aggresome formation is a specific and active cellular response to cope with excessive levels of misfolded and aggregated proteins (40 -42). In support of the role for aggresomes in processing intracellular misfolded protein aggregates, proteasome components and molecular chaperones are actively recruited to aggresomes. Here we have identified for the first time the formation of cytosolic PrP Sc aggresomes and shown that their presence is deleterious to neuronal cells. Aggresomes contain ubiquitin, chaperones, and proteasome components, thereby lowering the degradative ubiquitin and proteasome-dependent proteolysis in the cell in a negative feedback loop, resulting in an autocatalytic chain leading to the induction of apoptosis (41,46) Accumulation of misfolded proteins at the centrosomes may also severely impair their function and therefore interfere with cell division (54). Recently, it has also been shown that UPS impairment by protein aggregates is global and that the capacity of the entire cellular UPS is compromised by the presence of aggregates that are restricted to either the cytoplasmic or nuclear compartments (55). In our cell system, prion aggresome formation appears directly related to caspase 8 activation, which then proteolytically activates downstream caspase 3 and induces neuronal apoptotic cell death. Similar intraneuronal caspase 8-mediated apoptosis in response to aggregated proteins has been described in Huntington disease (56) and Alzheimer disease (57). Whether PrP Sc aggresome structures occur in vivo is not known, but our data showing a specific association between PrP Sc and the aggresomeassociated intracellular protein vimentin in RML-infected CD-1 mouse brain suggest that this may be the case.
We also propose that it may specifically be the cytosolic accumulation of PrP Sc aggresomes that is particularly proapopotic. In support of this, inhibition of specific lysosomal cysteine proteases in GT-1 cells inhibited the degradation of PrP Sc and resulted in an accumulation of compartmentalized lysosomal PrP Sc (58,59); however, this was not cytotoxic (58). Whereas pathogenic prion protein mutants have been reported to form intracellular aggresomes in response to proteasome inhibition (43,60) and misfolded cytosolic PrP has been reported to form aggresomelike structures after cyclosporin A treatment (43), there have been no reports to date of PrP Sc aggresome formation or of the effects of aggresomes on cell viability in prion disease. How PrP Sc may enter the cytoplasm to form aggresomes has not been established, but little is known about the exact details of cellular PrP Sc trafficking (recently reviewed in Ref. 61). Possible sites of entry include retrotranslocation from the ER (62), as has been described for PrP C and some pathogenic prion mutants (34,(63)(64)(65), during its intracellular trafficking pathway or by intracellular trafficking from outside the cell. PrP Sc may then accumulate in the cytoplasm when the proteasome is inhibited as may occur in aging or during prion infection in vivo (35,48) and generate toxic aggresome structures as demonstrated in this study. A fundamental question raised by the present study is whether PrP Sc accumulation in aggresomes is accompanied by concomitant accumulation of a distinct neurotoxic PrP L species. Detailed physico-chemical characterization of PrP Sc aggresome formation will be required to pinpoint the neurotoxic PrP entity. The neurotoxic intracellular mechanism suggested by the present study is likely to be part of a multifactorial prion disease pathogenesis, which may also involve synaptic dysfunction and alterations to cellular membrane permeability (66). Given the critical role that has emerged for the UPS in protein misfolding disorders (28), combined with the age-dependent decrease in UPS activity, the design of drugs that improve UPS function in neurons may help provide effective inter-vention to slow or prevent these diseases in which toxic proteins are misfolded.