Calpain and Other Cytosolic Proteases Can Contribute to the Degradation of Retro-translocated Prion Protein in the Cytosol*

PrP, a cell surface-localized N -linked glycoprotein, is required for the pathogenesis of prion diseases. Recent studies have revealed that prion protein (PrP) becomes neurotoxic and prone to aggregation when it is in the cytosol, suggesting that cytosolic PrP may play a role in the pathogenesis of prion disease. Retro-translocation of PrP from the endoplasmic reticulum to the cytosol for proteasome degradation offers a natural route for PrP to enter the cytosol, but whether PrP is subject to retro-translocation is controversial. In this study, we investi-gated the metabolism of endogenous wild-type PrP in several cell lines and in primary mouse cortical neurons. Our results suggest that a portion of the endogenous wild-type PrP is retro-translocated to the cytosol and degraded by the proteasome. Moreover, we also found that calpain and other cytosolic proteases could degrade PrP in the cytosol when the proteasome activity is compromised. These results provide the foundation for the hypothesis that cytosolic PrP may be involved in the pathogenesis of prion disease. Prion diseases are a unique group of neurodegenerative dis-orders that can be sporadic, dominantly inheritable, or infectious. The “protein-only” model postulates that PrP Sc , the misfolded 600 (cid:1) l of lysis buffer/well were used, and for all other cell types, 300 (cid:1) l of lysis buffer/well were used to lyse cells. The cell lysates were disrupted on ice by passing through 26-gauge needles and stored at (cid:2) 20 °C. To separate soluble and insoluble fractions, 50 (cid:1) l of PC3M cell lysates (100 (cid:1) l for other cells) were centrifuged at 16,300 (cid:1) g for 30 min at 4 °C. Supernatant was carefully transferred to a new tube, precipitated with 4 volumes of cold methanol ( (cid:2) 20 °C), and incubated at (cid:2) 20 °C for more than 30 min. Pellet fraction and precipitated supernatant proteins were sonicated in SDS-PAGE sample buffer containing 5% (w/v) SDS for 10 min. After boiling for 10 min, the samples were separated in 14% SDS-PAGE (Novex Tris-glycine gel; Invitrogen) and transferred to polyvinylidene difluoride membrane for immunoblot analyses. PrP was detected by the 3F4 monoclonal anti-PrP antibody (1:4000; Signet Laboratories) except experiments using mouse brain and primarily cultured mouse cortical neurons. In those experiments, the 7A12 monoclonal anti-PrP antibody (10) was used. The monoclonal 8B4 anti-PrP antibody (10) was used to determine the presence of the N terminus in PrP. All of the blots were developed by using a peroxidase-conjugated anti-mouse IgG secondary antibody (Bio-Rad) and ECL Plus reagent (Amersham Biosciences). Quantification was performed on a Storm phosphorus imaging system (Molecular Dynamics). To detect the signal peptide of PrP, a rabbit antibody raised against the signal peptide of PrP was used (1:1000), and the blot was developed by using a peroxidase-conjugated anti-rabbit IgG secondary antibody (Bio-Rad) and ECL Plus reagent (Amersham at 1:5000. RNase Protection Assay— RNase protection assay was performed as previously described (14). Human PrP cDNA was subcloned into the pBluescript SK (cid:3) vector. The PrP-pBluescript plasmid was digested with the restriction enzyme NcoI. After digestion, T7 RNA polymerase was used to generate antisense RNA probe. Human cyclophilin probe was a generous gift from Dr. Robert Costa at the University of Illinois at Chicago.

Prion diseases are a unique group of neurodegenerative disorders that can be sporadic, dominantly inheritable, or infectious. The "protein-only" model postulates that PrP Sc , the misfolded form of normal prion protein (PrP C ), is the infectious agent that can convert PrP C to PrP Sc in a self-templated reaction (1). Indeed, PrP Sc is the major component of the infectious agent and is found only in infected tissues (1,2). The central role of PrP in the transmission of prion diseases is supported by a variety of experimental evidence (1,3,4).
Aside from the infectious property of PrP, extensive efforts have focused on elucidating the mechanism of neurodegeneration in prion disease (5,6). Accumulated evidence revealed the contribution of normal host PrP C to the neurodegenerative process in prion disease. Using a neurografting technique, it has been shown that PrP Sc itself is not toxic to neurons that do not express PrP C (7). This conclusion is further supported by a recent study using conditional PrP knockout mice. Despite continuous accumulation of PrP Sc in these mouse brains, depletion of PrP C in neurons after prion infection not only stops the development of neurodegeneration but also reverses the neuropathology (8). Because PrP C is expressed abundantly in the brains of all mammals (9,10), the neurotoxic form of PrP has to be an abnormal form of host PrP C . Several forms of PrP have been implicated as the neurotoxic species in prion diseases, including transmembrane PrP (11,12), cross-linked PrP (13), and cytosolically localized PrP (14). Further investigation of these candidates is essential to identify which form of PrP is responsible for the neurodegeneration in prion disease or whether different neurotoxic PrP forms are responsible for different types of prion disease.
Expression of PrP in the cytosol, instead of its normal plasma membrane-associated form, has been shown to cause rapid and extensive degeneration of cerebellar granular neurons in transgenic mice (14). This neuropathological phenotype is similar to that of transgenic mice expressing a disease-associated PrP mutant (15). Moreover, in cell culture experiments, a significant portion of PrP in the reducing cytosolic environment gains a conformation similar to PrP Sc , being highly aggregated with a signature proteinase K-resistant pattern (16,17). This observation is also consistent with results from biochemical studies of PrP folding, in which reducing conditions can fold recombinant PrP into an aggregated, proteinase K-resistant, amyloidogenic conformation (18 -20). Based on these findings, we have proposed that cytosolic PrP is the neurotoxic species in prion disease, and it may also play a role in the initial formation of the PrP Sc conformation in sporadic and inherited prion diseases.
As a glycosylphosphatidylinositol-anchored protein, PrP is co-translationally translocated into the lumen of endoplasmic reticulum (ER). 1 Thereafter, it travels within the secretory pathway and is not exposed to the reducing environment in the cytosol (21). Retro-translocation of misfolded PrP from the ER to the cytosol for proteasomal degradation has been proposed as a natural route for PrP to enter the cytosol. But whether PrP is subject to retro-translocation and proteasome degradation has been controversial (22,23). Accumulation of wild-type and mutant PrP during proteasome inhibition has been observed in various cell lines and in human primary neurons (24 -28). Because many of these studies used transfected cells expressing PrP under the control of cytomegalovirus promoter, a recent study argued that the accumulation of PrP results from an artifact of increased expression of the cytomegalovirus promoter under proteasome inhibition (29).
Given the central role of PrP retro-translocation to the cytosolic PrP hypothesis, we performed a detailed analysis of endogenous wild-type PrP in several cell lines and in primary mouse neurons. We found that a portion of PrP is retro-translocated to the cytosol for degradation by the proteasome. Moreover, we also found that other cytosolic proteases can contribute to the degradation of retro-translocated PrP when proteasome activity is compromised. These results support the model that cytosolic PrP may be involved in the pathogenesis of prion disease.

EXPERIMENTAL PROCEDURES
Cell Lines and Cell Culture-PC3M is a metastatic subline of PC3 human prostate cancer cell line (30). PC3M cells were cultured in RPMI 1640 with 10% fetal bovine serum, 10 mM HEPES, pH 7.2, 2 mM glutamine, and penicillin-streptomycin. DU145 cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, 10 mM HEPES, pH 7.2, 2 mM glutamine, and penicillin-streptomycin. NT2 cells were cultured in Opti-MEM with 5% fetal bovine serum, and penicillin-streptomycin. All of the cell culture media and reagents were purchased from Invitrogen.
Treatment of Cells and Immunoblot Analysis-All of the cells used for inhibitor analyses were cultured in 6-well tissue culture plates. The cells were cultured in a humidified incubator at 37°C with 5% CO 2 . After cells reached 90% confluence, the culture media were replaced with fresh media containing various inhibitors. The cultures were continued in the CO 2 cell culture incubator at 37°C for the time indicated. The inhibitors used in this study were: MG132 (Biomol Research Laboratory), epoxomicin (Biomol Research Laboratory), lactacystin (EMD Biosciences, Calbiochem), MDL28170 (Sigma), NH 4 Cl (Sigma), and cloroquine (Sigma).
After treatment with various inhibitors, cell culture media were removed, and the cells were washed once with phosphate-buffered saline and lysed on ice in a cell lysis buffer (150 mM NaCl, 50 mM Tris-HCl, pH 8.0, 2 mM EDTA, 0.5% Triton X-100, 0.5% sodium deoxycholate, 1ϫ complete protease inhibitor (Roche Applied Science)). For PC3M cells, 600 l of lysis buffer/well were used, and for all other cell types, 300 l of lysis buffer/well were used to lyse cells. The cell lysates were disrupted on ice by passing through 26-gauge needles and stored at Ϫ20°C. To separate soluble and insoluble fractions, 50 l of PC3M cell lysates (100 l for other cells) were centrifuged at 16,300 ϫ g for 30 min at 4°C. Supernatant was carefully transferred to a new tube, precipitated with 4 volumes of cold methanol (Ϫ20°C), and incubated at Ϫ20°C for more than 30 min. Pellet fraction and precipitated supernatant proteins were sonicated in SDS-PAGE sample buffer containing 5% (w/v) SDS for 10 min. After boiling for 10 min, the samples were separated in 14% SDS-PAGE (Novex Tris-glycine gel; Invitrogen) and transferred to polyvinylidene difluoride membrane for immunoblot analyses. PrP was detected by the 3F4 monoclonal anti-PrP antibody (1:4000; Signet Laboratories) except experiments using mouse brain and primarily cultured mouse cortical neurons. In those experiments, the 7A12 monoclonal anti-PrP antibody (10) was used. The monoclonal 8B4 anti-PrP antibody (10) was used to determine the presence of the N terminus in PrP. All of the blots were developed by using a peroxidaseconjugated anti-mouse IgG secondary antibody (Bio-Rad) and ECL Plus reagent (Amersham Biosciences). Quantification was performed on a Storm phosphorus imaging system (Molecular Dynamics).
To detect the signal peptide of PrP, a rabbit antibody raised against the signal peptide of PrP was used (1:1000), and the blot was developed by using a peroxidase-conjugated anti-rabbit IgG secondary antibody (Bio-Rad) and ECL Plus reagent (Amersham Biosciences).
Deglycosylation-For deglycosylation reactions, total protein from cell culture lysates or mouse brain lysates was precipitated with 4 volumes of cold methanol (Ϫ20°C). The precipitated protein were sonicated in peptide:N-glycosidase F (PNGase F; New England Biolabs) digestion buffer or in endoglycosidase H (Endo H; New England Biolabs) digestion buffer. PNGase F or Endo H was added to the reaction mixture as indicated. Digestion was carried out at 37°C for 1 h, and all of the proteins were precipitated again by methanol after the digestion. To ensure that the digestion reaction does not interfere with PrP migration on SDS-PAGE, the pellet fraction of MG-132-treated PC3M cell lysates was sonicated into PNGase F buffer, incubated at 37°C for 1 h, and precipitated with methanol as well. All of the precipitated proteins were sonicated in SDS-PAGE sample buffer containing 5% (w/v) SDS for 10 min, boiled for 10 min, and separated in 14% SDS-PAGE. PrP was detected by using 3F4 or 7A12 anti-PrP antibody.
Pulse-Chase Analysis-PC3M cells were cultured in a 10-cm plate. When the cells reached 90% confluence, pulse-chase analysis was performed. The cells were incubated with labeling medium (Dulbecco's modified Eagle's medium without Met and Cys; Invitrogen) for 30 min at 37°C in a CO 2 tissue culture incubator. Afterward, the media were replaced with labeling media containing Tran 35 S-Label (200 Ci/plate; MP Biomedicals Inc.) with or without 50 M MG132. The pulse labeling was carried out at 37°C in a CO 2 tissue culture incubator for 6 h. The cells were washed three times with phosphate-buffered saline, and regular culture media were added in the presence or absence of 50 M MG132. After chasing for indicated time, the cells were washed once with phosphate-buffered saline and lysed in 1 ml of lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40, 2% SDS). To denature all proteins in the lysate, 1.5 ml of 8 M guanidinium HCl (Sigma) was added to 1 ml of lysate, and all of the proteins were precipitated with 4 volumes of cold methanol (Ϫ20°C). Precipitated proteins were sonicated into 1 ml of DLPC buffer (4.2 mg/ml L-␣phosphatidylcholine, 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2% sarkosyl). PrP was immunoprecipitated with 3F4 antibody and protein Aconjugated Sepharose 4B beads (Sigma). The beads were washed twice with buffer A (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, 0.5% Triton X-100, 0.5% deoxycholate, 0.5% sarkosyl), twice with buffer B (50 mM Tris-HCl, pH 7.5, 500 mM NaCl, 0.1% Triton X-100, 0.05% deoxycholate), and once with buffer C (50 mM Tris-HCl, pH 7.5, 0.1% Triton X-100, 0.05% deoxycholate). PrP was released from the beads by boiling in SDS sample buffer for 10 min and separated on a 12% SDS-PAGE. The signals were detected and quantified by a Storm phosphorus imaging system.
Primary Mouse Cortical Neuronal Culture-Frozen mouse cortical neurons were purchased from QBM Cell Science. These neurons were cultured exactly according to the protocol supplied by the manufacturer. Neurobasal medium, L-glutamine, penicillin-streptomycin, and B27 serum-free supplements were purchased from Invitrogen as suggested by the manufacturer. These neurons were brought up and cultured for 7 days before the experiments were performed.
Subcellular Fractionation Using Iodixanol Gradient-PC3M cells were cultured in 10-cm plates and treatments were performed when cells were 90% confluent. The culture media were replaced with fresh media with or without 50 M MG132, and the cells were incubated at 37°C in a CO 2 tissue culture incubator for 8 h. At the end of incubation, the cells were washed once in phosphate-buffered saline, scraped into 1 ml of membrane preparation buffer (8% sucrose, 20 mM Tris-HCl, pH 8.0, 1 mM EDTA), and homogenized on ice. The homogenates were centrifuged at 1,000 ϫ g for 5 min at 4°C to remove nuclei and unbroken cells. The post-nuclear supernatant (280 l/sample) was mixed with stock iodixanol solution (OptiPrep, Axis-Shield PoC, AS) to generate 1 ml 36% iodixanol solution, and the mixture was loaded at the bottom of an ultracentrifuge tube. A 1-ml 31% iodixanol solution and a 0.25-ml 5% iodixanol solution were loaded on top of the sample sequentially. The gradient was centrifuged at 186,000 ϫ g for 3 h at 4°C in a TLX Beckman bench top ultracentrifuge. 250-l fractions were collected from the top to bottom, and proteins in all fractions were precipitated by adding 250 l of 40% trichloroacetic acid. Precipitated proteins were sonicated in SDS-PAGE sample buffer and were used to perform immunoblot analysis. The antibodies used for the immunoblot analyses are: 3F4 anti-PrP antibody at 1:4000, anti-calnexin antibody (Stressgen) at 1:1000, and anti-Hsp70 antibody (Stressgen) at 1:5000.
RNase Protection Assay-RNase protection assay was performed as previously described (14). Human PrP cDNA was subcloned into the pBluescript SKϩ vector. The PrP-pBluescript plasmid was digested with the restriction enzyme NcoI. After digestion, T7 RNA polymerase was used to generate antisense RNA probe. Human cyclophilin probe was a generous gift from Dr. Robert Costa at the University of Illinois at Chicago.

Accumulation of Unglycosylated and Aggregated PrP upon
Proteasome Inhibition-To determine whether PrP is subject to retro-translocation to the cytosol for degradation by the proteasome, we analyzed endogenous wild-type PrP in a human prostate cancer cell line, PC3M. These cells express high levels of PrP that is fully glycosylated, soluble in mild detergents, and localized on the plasma membrane (Ref. 31 and data not shown). When the proteasome inhibitor MG132 was added to these cells, PrP started to accumulate. Notably, the majority of accumulated PrP is unglycosylated and insoluble with mild detergent extraction (Fig. 1A). After 8 h of proteasome inhibition, the level of soluble, fully glycosylated PrP remained about the same level as that of control cells (Fig. 1B). In contrast, the level of unglycosylated PrP increased to about 4-fold of that in control cells (Fig.  1B). After 24 h of proteasome inhibition, the level of soluble and fully glycosylated PrP was significantly decreased, presumably because of the effect of long term proteasome inhibition. How-ever, the accumulation of unglycosylated PrP was dramatically increased (Fig. 1B), with the total unglycosylated PrP about 6-fold of that in control cells, reflecting continuous accumulation of PrP destined for proteasome degradation.
Because many N-terminally truncated PrP fragments are present in normal cells (32), it is possible that the accumulated PrP could be a PrP fragment instead of full-length PrP. To differentiate between these two possibilities, we performed immunoblot analysis with 8B4 monoclonal antibody (Fig. 1C). This antibody recognizes an epitope at amino acid 37-44 of PrP, which is removed in almost all N-terminally truncated PrP fragments (32,33). In the supernatant fraction, the signal detected by 8B4 antibody was much weaker comparing to the signal detected by 3F4 antibody (comparing the supernatant fractions between panels A and C of Fig. 1), indicating that most of the soluble and fully glycosylated PrP in PC3M cells was N-terminally truncated PrP fragments. However, the PrP in the pellet fraction was clearly detected by the 8B4 antibody ( Fig. 1C, pellet). The presence of 8B4 epitope indicated that the PrP accumulated upon proteasome inhibition contains fulllength PrP.
Moderate Increase of Endogenous PrP mRNA Levels upon Proteasome Inhibition-To ensure that the accumulation of unglycosylated PrP is not due to an increase in PrP expression, we performed an RNase protection assay to measure the mRNA levels of PrP in PC3M cells treated with proteasome inhibitor. The probe against cyclophilin was included in the assay as an internal control for total RNA levels. Upon proteasome inhibition, we detected a small but consistent increase of PrP mRNA level. After 8 h of MG132 treatment, PrP mRNA level increased about 1.5-fold (Fig. 2). It is unlikely that this moderate increase of PrP mRNA levels could account for the accumulation of PrP upon proteasome inhibition. Increase of PrP expression should have lead to an increase of soluble and fully glycosylated PrP protein but not solely to the accumulation of unglycosylated PrP as we observed (Fig. 1). After 24 h of proteasome inhibition, the PrP mRNA level was 1.5-fold of that in untreated cells (Fig. 2), and the soluble fully glycosylated PrP protein level was much lower than that in control cells (Fig. 1B). In contrast, the accumulation of unglycosylated PrP was much higher, and the majority of the accumulated PrP was aggregated ( Fig. 1; a 30-fold increase of aggregated PrP after 24 h of proteasome inhibition). It is unlikely that a mere 0.5fold increase of PrP mRNA level can explain such a massive accumulation of unglycosylated and aggregated PrP in proteasome-inhibited cells.
Accumulation of Unglycosylated PrP Results from Inhibition of Its Degradation-An alternative explanation of unglycosylated and aggregated PrP accumulation would be that these PrP molecules were destined to the proteasome for degradation. To test this possibility, we performed pulse-chase analysis in PC3M cells with 3F4 anti-PrP antibody. Two pitfalls could have potentially interfered with pulse-chase analysis of retrotranslocated PrP. First, the pulse labeling generally is performed in a short time period, such as 30 min. Even with the PC3M cells that express a very high amount of endogenous wild-type PrP, only about 5% of total PrP accumulated after 4 h of continuous proteasome inhibition (Fig. 1), and only about 15% of total PrP accumulated after 8 h of proteasome inhibition. (Quantification was performed using a Storm phosphorus imaging system.) Therefore, short pulse labeling will not accumulate detectable amount of PrP molecules destined to proteosome degradation. Second, the majority of the accumulated PrP is insoluble in mild detergents (Fig. 1). Immunoprecipitation of aggregated proteins is generally inefficient, and these proteins tend to be removed during the preclear centrifugation step of immunoprecipitation. To overcome these two potential pitfalls, we pulse-labeled cell in the presence or absence of the proteasome inhibitor MG132 for 6 h and chased under the same condition for 12 or 18 h. After cell lysis, all of the proteins were denatured with 4.8 M guanidinium HCl and precipitated with methanol. A buffer used previously to solublize aggregated PrP Sc (34) was used to solublize methanol-precipitated proteins. In cells treated with MG132, a band migrated at the position of unglycosylated PrP was clearly detectable after pulse labeling (Fig. 3A, indicated by an arrow). This band did not appear in the control sample that was treated with MG132, pulse-labeled, and immunoprecipitated under the same procedure except without incubation with anti-PrP antibody (data not shown), suggesting that it represents the accumulated unglycosylated PrP after proteasome inhibition. After 12 h of chase, majority of the fully glycosylated PrP were degraded in control cells (Fig. 3, A and B), consistent with previous reports that PrP is a short-lived protein (34,35). In MG132-treated cells, the majority of the fully glycosylated PrP was also turned over after 12 h of chase (Fig. 3A, indicated by a bracket). In contrast, the portion of accumulated unglycosylated PrP in the same MG132-treated samples remained even after 18 h chase (Fig. 3A, bands indicated by a arrow). Taken together, these results suggest that the accumulation of unglycosylated PrP results from the inhibition of proteasome degradation rather than increased expression. Moreover, the normal turnover rate of fully glycosylated PrP in MG132-treated cells indicated that proteasome inhibition did not change the metabolism of PrP molecules going through the secretory pathway as previously reported (29).
Cytosolic Localization of Accumulated PrP upon Proteasome Inhibition-In mammalian cells, signal peptides are generally very efficient in directing the co-translational translocation of protein into the ER lumen. If the increase of PrP levels upon proteasome inhibition results from increasing PrP expression, accumulated PrP, even if aggregated, should be within the secretory pathway and associated with membrane. To determine the subcellular localization of accumulated PrP, we used a discontinuous iodixanol gradient to separate membrane fractions from cytosolic proteins. In this gradient, membrane fractions will flow to the top of the gradient, whereas cytosolic protein will remain in the sample-loading region. The successful separation of membrane and cytosolic fractions was confirmed by the enrichment of endogenous PrP and a membranebound chaperone, calnexin, in the top three fractions of the gradient (Fig 4A, Control). The cytosolic chaperone protein Hsp70 was used to identify fractions containing cytosolic proteins (the bottom five fractions). A very small amount of Hsp70 co-migrated with the membrane fraction (Fig. 4A, fraction 2), reflecting its association with membrane-bound co-chaperones (as reviewed in Ref. 36). When the homogenate from cells treated with MG132 was applied to the same gradient, the fully glycosylated and membrane-bound PrP still migrated to the top three fractions as well as calnexin (Fig. 4B). However, accumulated unglycosylated PrP migrated in the same fractions as cytosolic chaperone Hsp70, demonstrating that these PrP molecules are localized in the cytosol (Fig. 4iB). Cytosolic accumulation of unglycosylated PrP further supports that these PrP molecules are destined to the proteasome for degradation.
Accumulated PrP Has Been Proteolytically Processed in the ER-Cytosolic accumulated PrP could result from PrP molecules that has been in the lumen of ER, misfolded, and retrotranslocated. Alternatively, it may also come from PrP molecules that never enter the ER because of some aberrant effects of proteasome inhibition. To differentiate between these two possibilities, we took advantage of the fact that two proteolytic cleavages occur when PrP is in the lumen of ER, that is, cleavage of the N-terminal 22-amino acid signal peptide and cleavage of the C-terminal 22 amino acids for addition of the glycosylphosphatidylinositol anchor (21). Had accumulated PrP been in the lumen of ER, it should be through all these modifications and have the same molecular weight as that of deglycosylated wild-type PrP that has traveled through ER to the cell surface. To compare the size of accumulated PrP with deglycosylated PrP, we included a control N2A cell line (moPrP cell line, N2A cells stably transfected with mouse PrP carrying a 3F4 epitope) that has been used in previous studies (14,37). PrP from moPrP cells migrates with three distinct bands on SDS-PAGE, likely representing unglycosylated, mono-core-glycosylated, or di-core-glycosylated PrP (14). As we showed in previous study (14), PrP from moPrP cells was sensitive to both Endo H and PNGase F digestions (Fig. 5A, compare lanes 2,  4, and 6), typical of core-glycosylated N-linked oligosaccharides.
PrP from PC3M cells was sensitive to PNGase F digestion but resistant to Endo H digestion (Fig. 5A, compare lanes 1, 3, and 5), characteristic of N-linked oligosaccharides being modified in the Golgi complex. Smaller molecular mass bands (Fig. 5A, lane 5) represented the N-terminal truncated PrP fragments as showed in  To rule out the possibility that some cell culture-specific post-translational modifications may interfere with PrP migration, we also compared the size of accumulated PrP with deglycosylated PrP from wild-type mouse brain lysates. As expected, PrP from mouse brain was resistant to Endo H digestion but was sensitive to PNGase F digestion (Fig. 5B,  lanes 1-3). Importantly, the accumulated PrP (Fig. 5B, lane 4) migrated to the same position as deglycosylated PrP from mouse brain. The fact that accumulated PrP and deglycosylated PrP co-migrate at the exact same position on SDS-PAGE suggests that the PrP accumulated during proteasome inhibition has been processed in the ER.
Our second approach to investigate whether the accumulated PrP is retro-translocated from the ER to the cytosol utilized an antibody raised against the signal peptide region of PrP (29). This antibody has been shown to recognize the signal peptide of both mouse and human PrP (28,29). As a positive control for this anti-signal peptide antibody, we included a 3T3 cells that was stably transfected with wild-type mouse PrP with 3F4 epitope (PrP-3T3). As previously reported (29), a portion of accumulated PrP from proteasome inhibited PrP-3T3 cells contains the N-terminal signal peptide (data not shown). We compared the pellet fractions of detergent lysates of PC3M or PrP-3T3 cells treated with or without MG132 to determine whether the signal peptide of accumulated PrP in PC3M cells has been removed. To ensure the detection of signal peptide antibody was not interfered by the amount of PrP, we overloaded the samples from the PC3M cells while underloading samples from PrP-3T3 cells. Two blots with exactly the same samples were probed with either 3F4 anti-PrP antibody or the anti-signal peptide antibody. As expected, PrP accumulated in the pellet fraction in both PC3M cells and the stable PrP-3T3 cells upon proteasome inhibition (Fig. 6, anti-PrP antibody). However, accumulated PrP in MG132-treated PrP-3T3 cells migrated as two separate bands. The upper band was recognized by the anti-signal peptide antibody, suggesting this portion of accumulated PrP contained the signal peptide (Fig. 6,  open arrow labeled ssPrP). The accumulated PrP from MG132treated PC3M cells migrated as a single band to the position similar to the lower band of PrP from MG132-treated PrP-3T3 cells (Fig. 6, solid arrow labeled PrP). Importantly, the anti-signal peptide antibody did not recognize accumulated PrP from PC3M cells even after prolonged exposure (Fig. 6, antisignal peptide antibody). This result is in accordance with our previous finding that the accumulated PrP from PC3M cells has been proteolytically processed in the lumen of ER. In addition, the fact that the fast migrating species of accumulated PrP in PrP-3T3 cells did not contain the signal peptide (Fig. 6, solid arrow) suggests that this portion of PrP from this stably transfected cell line has also been processed in the lumen of ER. All together, these results suggest that the accumulated PrP has been translocated into the ER, processed, retro-translocated to the cytosol and accumulated in the cytosol when the proteasome is inhibited.
Accumulation of PrP in Different Cell Lines or Treated with Different Proteasome Inhibitors-To rule out the possibility that retro-translocation of PrP is a specific phenomenon of PC3M cell line, we tested two additional cell lines that express detectable levels of endogenous wild-type PrP. DU145 is a human prostate tumor cell line, and NT2 is a human teratocarcinoma cell line that has the characteristics of neuronal progenitor cells (38). The difference in the migration patterns of the soluble fully glycosylated PrP from these two cell lines represents the different modifications of N-linked oligosaccharides of PrP in these two cell lines (Fig. 7). Just as in PC3M cells, the accumulation of unglycosylated PrP in these two cell lines was detectable after 4 h of MG132 treatment and reached a higher level after 8 h of treatment (Fig. 7), and a significant portion of these accumulated PrP was aggregated in mild detergents.
To differentiate whether the accumulation of PrP is truly due to the effect of proteasome inhibition or to an aberrant effect of MG132, we tested several specific proteasome inhibitors. We treated PC3M cells with 10 M epoxomicin, 10 M lactacystin, or 50 M MG132 for 8 h, and accumulation of unglycosylated PrP in the pellet fraction was clearly detected in cells treated with all proteasome inhibitors but not in untreated cells (Fig.  8). Accumulation of PrP in different cell lines and with treatment of different proteasome inhibitors suggests that retrotranslocation of PrP to the cytosol for proteasome degradation is a general phenomenon and not an artifact of a particular cell line or a particular proteasome inhibitor.

Other Cytosolic Proteases Can Also Contribute to the Degradation of Retro-translocated PrP when Proteasome Activity Is
Compromised-We noticed that treatments with different proteasome inhibitors resulted in different amounts of PrP accumulation in the pellet fraction (Fig. 8). Because increasing epoxomicin concentration to as high as 100 M did not change the amount of accumulated PrP (data not shown), we concluded that different amounts of PrP accumulation reflect the pharmacological characteristics of different inhibitors rather than dosage differences of inhibitors. Compared with epoxomicin or lactacystin, MG132 is a rather nonspecific proteasome inhibitor. In addition to its strong proteasome inhibitory activity, it can also inhibit other proteases (39,40). To rule out the possibility that the PrP accumulation is due to the inhibition of other proteases, we treated PC3M cells for 8 h with lysosome inhibitor NH 4 Cl or chloroquine or with a calpain-specific inhibitor, MDL28170. Although unglycosylated PrP accumulated in the pellet fraction in cells treated with proteasome inhibitors, there was no PrP accumulation in cells treated with lysosome inhibitors or calpain inhibitor (Fig. 9A), suggesting that the proteasome is the primary degradation machinery for PrP in the cytosol.
An alternative explanation for the differences of various proteasome inhibitors could be that other cytosolic proteases will degrade PrP in the cytosol only if its primary degradation machinery, the proteasome, is inhibited. To test this hypothesis, we treated cells with epoxomicin or MG132 alone and treated them with epoxomicin plus NH 4 Cl, chloroquine or MDL28170 (Fig. 9B). Although epoxomicin plus lysosomal inhibitors (NH 4 Cl or chloroquine) had no effect on PrP accumulation, cells treated with epoxomicin plus MDL28170, a specific inhibitor against cytosolic calpains, showed significant increases in the amount of accumulated PrP in the pellet fraction (Fig. 9, B and C). This result demonstrates that other cytosolic proteases can also contribute to the degradation of PrP in the cytosol. The fact that cytosolic proteases instead of lysosome proteases contribute to the degradation of PrP further supports our results of cytosolic accumulation of these PrP molecules. Notably, even in cells treated with epoxomicin plus MDL28170, the level of PrP accumulation is not as high as that in cells treated with MG132 alone (Fig. 9, B and C), indicating that additional cytosolic proteases may contribute to cytosolic PrP degradation.
Accumulation of PrP in Primary Mouse Cortical Neurons upon Proteasome Inhibition-Because the main pathological changes during prion disease occur in neurons, retro-translocation of PrP has to occur in neurons for cytosolic PrP to be a valid candidate for the neurotoxic form in prion disease. To test whether PrP in neurons is subject to retro-translocation, we treated mouse cortical neuron primary cultures with or without MG132 or with MG132 plus brefeldin A (an ER stress-inducing agent) (41). Upon proteasome inhibition, there was a clear increase of unglycosylated PrP appearing in the pellet fraction (Fig. 10). Higher amounts of PrP accumulation were detected in cells treated with MG132 plus brefeldin A (Fig. 10), suggesting that brefeldin A increases PrP retro-translocation from the ER. Accumulation of PrP after proteasome inhibition in primary neurons suggests that retro-translocation of PrP occurs in neurons.

DISCUSSION
Retro-translocation of PrP from the ER to the cytosol is a natural route for membrane-associated PrP to enter the cytosol. Our results suggest that a portion of endogenous wildtype PrP is retro-translocated to the cytosol for degradation by the proteasome. Moreover, retro-translocated PrP is also subject to the digestion by other cytosolic proteases when the proteasome is inhibited. If escaped proteolysis, accumulated PrP in the cytosol will aggregate, and aggregated PrP molecules are apparently resistant to further degradation.
Whether the cytosolic PrP accumulation after proteasome inhibition results from retro-translocation of PrP from the ER or from the portion of PrP that never enters the ER has been controversial (29). Our previous results suggested that the accumulated PrP in the cytosol is from the PrP that retrotranslocated from the ER (27). Here, we showed that the accumulated PrP is without the N-terminal signal peptide (Fig. 6). Because the cleavage of the signal peptide occurs in the ER lumen, these PrP molecules must have entered the ER and retro-translocated to the cytosol. In addition, the accumulated PrP has exactly the same size as that of deglycosylated PrP from two different cell lines or from mouse brain (Fig. 5), suggesting these molecules have been through the same processing in the lumen of ER. Therefore, we conclude that the cytosolic PrP accumulation in these cells results from PrP molecules that have been in the lumen of ER.
An unexpected finding of this study is the increase of mRNA level of endogenous PrP upon proteasome inhibition. This small but very consistent increase of PrP mRNA levels is likely due to the fact that there are two heat shock transcription factor (HSF)-binding sites in the PrP promoter (42). Because PrP expression is up-regulated during heat shock treatment, these HSF binding sites are obviously functional inside cells (42,43). Proteasome inhibition is known to increase the halflives of various short-lived transcription factors, including HSF (44,45). Increased HSF protein levels may cause the PrP mRNA level increase during proteasome inhibition. But the small increase of PrP mRNA level cannot explain the massive accumulation of unglycosylated and aggregated PrP in the cytosol upon proteasome inhibition. The accumulation of unglycosylated and aggregated PrP results from the accumulation of retro-translocated PrP in the cytosol.
Retro-translocation of protein from ER to the cytosol is a major cellular pathway to remove misfolded and misassembled proteins in the secretory pathway (46). Accumulated evidence suggests that, except for a few transmembrane proteins, most of these proteins are exported from ER through a protein channel (46 -48). To fit through the pore of the channel, retro- translocated proteins are believed to be at least partially unfolded (46,49). ER resident chaperones have been proposed to participate in recognizing and delivering the substrate to the channel for retro-translocation (49). As a lumenal glycosylphosphatidylinositol-anchored protein, PrP is likely to be retrotranslocated through the pore of the protein channel. To pass through this pore, the retro-translocated PrP has to remain unfolded. ER chaperones may play a role in maintaining the unfolded-status of PrP. Indeed, a disease-associated mutant PrP has been found to remain associated with ER chaperone BiP for a prolonged time and degraded by the proteasome (26).
The appearance of unfolded PrP molecules in the cytosol will lead to several consequences explaining our observations. First, it can be a substrate for proteasome degradation as other retro-translocated proteins. Second, when proteasome activity is compromised, unfolded PrP molecules will accumulate in the cytosol, an environment that is not natural for PrP. Therefore, these unfolded PrP molecules will be susceptible to digestion by various cytosolic proteases. Third, the conformational change of an unfolded PrP exposed to the reducing cytosolic environment is similar to the folding of a recombinant PrP under reducing conditions. It has been shown that recombinant PrP will fold into a ␤-sheet rich, proteinase K-resistant, aggregated conformation under the reducing conditions (18 -20). Thus, the majority accumulated PrP becomes aggregated in the reducing cytosolic environment (more than two-thirds of accumulated PrP was aggregated). Fourth, detection of aggregated PrP in cells treated only with very specific proteasome inhibitors (such as epoxomicin) indicates that some PrP molecules can escape other cytosolic proteases digestion. This is likely due to the inaccessibility of cytosolic proteases to these PrP molecules once they are aggregated.
We found that calpain is one of the cytosolic proteases contributing to the degradation of retro-translocated PrP in the cytosol when proteasome activity is compromised. The calpain proteases are a family of Ca 2ϩ -activated cysteine proteases localized in the cytosol (50). Interestingly, inhibition of calpaindependent endoproteolytic cleavage has been shown to inhibit the accumulation of PrP Sc in scrapie-infected cells and reduce infectivity associated with these cells (51). Whether cytosolic PrP plays any role in these calpain-mediated effects is unknown, but the contribution of calpain to the cytosolic PrP degradation could potentially explain these effects. Proteasome activity decreases during prion disease (52). Therefore, it is reasonable to postulate that scrapie-infected cells may also have a decreased proteasome activity. Under this circumstance, cytosolic proteases, like calpain, may play a role in removing cytosolic PrP. Inhibition of calpain activity could result in the accumulation of toxic cytosolic PrP molecules. In vitro, scrapie does not uniformly infect all cells; only some of the cells are infected. Upon calpain inhibition, infected cells would be gradually selected out because of the accumulation of the toxic cytosolic PrP. As a result, an increased number of uninfected cells will be populated in the culture, which will be reflected by reduced PrP Sc accumulation and infectivity.
In the course of our experiments, we observed increasing cell death (detached cells floating in the media) in primary neurons and neuronal like NT2 cells treated with proteasome inhibitor. Cell death was more obvious when proteasome inhibition lasted for a longer period of time. In contrast, no obvious cell death was observed in PC3M or DU145 cells even after 24 h of proteasome inhibition. These observations seem to correlate neuronal cell death with cytosolic accumulation of PrP. However, proteasome inhibition affects many aspects of cell metabolism, and events other than cytosolic accumulation of PrP could very well be the reason for neuronal cell death. Therefore, further investigation is required to determine whether cultured neurons can recapitulate the neurotoxicity caused by cytosolic PrP in transgenic mice (14).
Retro-translocation of PrP to the cytosol has been observed in human primary neurons by subcellular fractionation (28). Here, we observed accumulation of PrP in primarily cultured mouse cortical neurons with proteasome inhibition. The characteristics of in vitro cultured neurons may vary depending on what kind of neurons is selected, which procedure is used, and which culture condition is applied. However, observations of PrP retro-translocation by using two different assays in two different kinds of neurons suggest that this pathway exists in neurons (28). Notably, PrP in the cytosol has been detected in vivo in a subset of neurons in the hippocampus, neocortex, and thalamus of wild-type mouse brains by using high resolution cryoimmunogold electron microscopy (53). Retro-translocation of PrP in neurons could explain how PrP travels to the cytosol of these neurons. All together, the appearance of retro-translocated PrP in the cytosol when proteasome activity is compromised provides a foundation for the hypothesis that cytosolic PrP is involved in the pathogenesis of prion disease.