Dual mechanisms for shedding of the cellular prion protein.

The cellular prion protein (PrP(C)) is essential for the pathogenesis and transmission of prion diseases. Whereas the majority of PrP(C) is bound to the cell membrane via a glycosylphosphatidylinositol (GPI) anchor, a secreted form of the protein has been identified. Here we show that PrP(C) can be shed into the medium of human neuroblastoma SH-SY5Y cells by both protease- and phospholipase-mediated mechanisms. The constitutive shedding of PrP(C) was inhibited by a range of hydroxamate-based zinc metalloprotease inhibitors in a manner identical to the alpha-secretase-mediated shedding of the amyloid precursor protein, indicating a proteolytic shedding mechanism. Like amyloid precursor protein, this zinc metalloprotease-mediated shedding of PrP(C) could be stimulated by phorbol myristate acetate and by copper ions. The lipid raft-disrupting agents filipin and methyl-beta-cyclodextrin promoted the shedding of PrP(C) via a distinct mechanism that was not inhibited by hydroxamate-based inhibitors. Filipin-mediated shedding of PrP(C) is likely to occur via phospholipase cleavage of the GPI anchor, since a transmembrane polypeptide-anchored PrP construct was not shed in response to filipin treatment. Collectively, our data indicate that shedding of PrP(C) can occur via both secretase-like proteolytic cleavage of the protein and phospholipase cleavage of the GPI anchor moiety.

The cellular prion protein (PrP C ) is essential for the pathogenesis and transmission of prion diseases. Whereas the majority of PrP C is bound to the cell membrane via a glycosylphosphatidylinositol (GPI) anchor, a secreted form of the protein has been identified. Here we show that PrP C can be shed into the medium of human neuroblastoma SH-SY5Y cells by both proteaseand phospholipase-mediated mechanisms. The constitutive shedding of PrP C was inhibited by a range of hydroxamate-based zinc metalloprotease inhibitors in a manner identical to the ␣-secretase-mediated shedding of the amyloid precursor protein, indicating a proteolytic shedding mechanism. Like amyloid precursor protein, this zinc metalloprotease-mediated shedding of PrP C could be stimulated by phorbol myristate acetate and by copper ions. The lipid raft-disrupting agents filipin and methyl-␤-cyclodextrin promoted the shedding of PrP C via a distinct mechanism that was not inhibited by hydroxamate-based inhibitors. Filipin-mediated shedding of PrP C is likely to occur via phospholipase cleavage of the GPI anchor, since a transmembrane polypeptide-anchored PrP construct was not shed in response to filipin treatment. Collectively, our data indicate that shedding of PrP C can occur via both secretase-like proteolytic cleavage of the protein and phospholipase cleavage of the GPI anchor moiety.
Transmissible spongiform encephalopathies encompass a group of neurodegenerative diseases including scrapie in sheep, bovine spongiform encephalopathy in cattle, and human pathologies such as Creutzfeldt-Jakob disease and Gerstmann-Straü ssler-Scheinker syndrome (1)(2)(3). The principal agent thought to be responsible for the transmission of these diseases is the infectious form of the prion protein (PrP) 1 (4). In prion diseases the normal cellular form of the prion protein (PrP C ) undergoes a conformational change to the scrapie isoform (PrP Sc ) that is partially resistant to proteases and accumulates in plaques (5).
PrP C is anchored to the extracellular surface of the plasma membrane by a glycosylphosphatidylinositol (GPI) anchor. Like most GPI-anchored proteins, PrP C is clustered into cholesterol-and sphingolipid-rich membrane microdomains called lipid rafts (6,7). At the plasma membrane, PrP C can interact with PrP Sc in the de novo formation of PrP Sc aggregates (8). A reduction in the level of cell surface PrP C through enhanced endocytosis or by release of the protein from the membrane may reduce PrP Sc production by limiting the amount of PrP C substrate available for conversion (9). Thus, the elucidation of the mechanisms involved in the cell surface release of PrP C may be of critical importance in the understanding of the pathogenesis of transmissible spongiform encephalopathies.
Soluble PrP C is present in the medium of cultured cells (10 -15) and human cerebrospinal fluid and serum (16,17) and is released from human platelets (18). However, the mechanism(s) by which soluble PrP C is released from cells remains to be determined. Like other GPI-anchored proteins, PrP C can be released from the cell surface in vitro by the action of exogenous bacterial phosphatidylinositol-specific phospholipase C (PI-PLC) (19). In light of this fact, it has become widely assumed that soluble PrP C is produced via cleavage of its GPI anchor by an endogenous mammalian GPI-specific phospholipase (17). However, a growing number of cell surface proteins can also be proteolytically shed by the action of a group of zinc metalloproteases known as secretases or sheddases (reviewed in Ref. 20). This cleavage and secretion appears to be an important and widely used cellular post-translational regulatory process, because a variety of structurally and functionally unrelated cell surface proteins undergo such a process (reviewed in Ref. 21). Whereas most proteolytically shed proteins are derived from transmembrane polypeptide-anchored precursors, several GPI-anchored proteins, including the folate receptor (22), the ecto-ADP-ribosyltransferase ART2.2 (23), and a GPI-anchored construct of the angiotensin-converting enzyme (24) are shed by the action of a metalloprotease.
In the current study, we show for the first time that a significant proportion of PrP C is shed from the cell surface by the proteolytic action of a zinc metalloprotease with similarities to the ␣-secretase involved in the shedding of the amyloid precursor protein (APP). We also show that this protease-mediated shedding of PrP is distinct from the shedding induced by treatment with the lipid raft-disrupting agents, filipin or methyl-␤-cyclodextrin (M␤CD), which is most likely phospholipasemediated. Collectively, our data indicate dual mechanisms for the cell surface shedding of PrP.

EXPERIMENTAL PROCEDURES
Cell Culture-The human neuroblastoma cell line SH-SY5Y was cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, penicillin (50 units/ml), streptomycin (50 mg/ ml), and 2 mM glutamate (all from Invitrogen). Cells were maintained at 37°C in 5% CO 2 in air. For stable transfections, 30 g of DNA was introduced to cells by electroporation, and selection was performed in normal growth medium containing 500 g/ml neomycin selection antibiotic. SB 244000 and all other hydroxamate-based inhibitor compounds were synthesized at GlaxoSmithKline Pharmaceuticals (Harlow, UK) and were used at 20 M. Structural details of these compounds have been published previously (25). Phorbol myristate acetate (PMA), filipin, and M␤CD (Sigma) were used at 1 M, 10 nM, and 5 mM, respectively. Cells were incubated with copper concentrations of 0 -100 M achieved by the addition of a CuSO 4 stock diluted in fetal bovine serum. When the cells were confluent, the medium was changed to OptiMEM (Invitrogen), and the cells were incubated for 7 h or 30 min at 37°C with the indicated compounds (24 h in the case of the copper incubations). The medium was then harvested and concentrated. For analysis of cell-associated PrP, cells were washed with phosphatebuffered saline (20 mM Na 2 HPO 4 , 2 mM NaH 2 PO 4 , 0.15 M NaCl, pH 7.4) and scraped from the flasks into phosphate-buffered saline. Following centrifugation at 500 ϫ g for 5 min to pellet the cells, they were lysed in 0.1 M Tris/HCl, 0.15 M NaCl, 1% Triton X-100, 0.1% Nonidet P-40, pH 7.4.
Protein and Enzyme Assays-Protein was quantified using bicinchoninic acid (26) in a microtiter plate with bovine serum albumin as a standard. Alkaline phosphatase activity was assayed using p-nitrophenyl phosphate as substrate, and the product was quantified spectrophotometrically as described previously (27).
Peptide:N-Glycosidase F Deglycosylation-Enzymic deglycosylation was performed by adding 10 l of 5ϫ peptide:N-glycosidase F buffer (30 mM Na 2 HPO 4 /NaH 2 PO 4 , pH 7.2, 20 mM EDTA) to 35 l of concentrated conditioned medium or cell lysate along with 2.5 l of 10% (w/v) SDS and 2.5 l of ␤-mercaptoethanol. The samples were boiled for 5 min, and then 20% (v/v) Triton X-100 (2.5 l) was added along with 1 unit of peptide:N-glycosidase F. Samples were then deglycosylated for 16 h at 37°C.
Membrane Preparation and High Speed Centrifugation-SH-SY5Y cells were harvested as described above, and the cell pellet was resuspended in 30 ml of 50 mM Hepes, 20 mM CaCl 2 , pH 7.5. The cells were then disrupted by sonication (30% maximum power for 30 s using a , a copper-binding octapeptide repeat sequence (shaded box), and a 23-amino acid C-terminal GPI anchor addition sequence (dotted box). PrP-DA retains the C-terminal GPI addition sequence present in wt-PrP, but the signal peptide was replaced with the uncleaved signal sequence/transmembrane domain (checkered box) and stalk region (cross-hatched box) from murine aminopeptidase A (48). PrP-PG14 possesses an additional nine octapeptide repeats. B, expression levels of PrP constructs. Cell lysates were immunoblotted with antibody 3F4. C, detection of PrP in conditioned media. Cells were incubated for 7 h in the absence or presence of SB 244000 (20 M), and conditioned media were immunoblotted with antibody 3F4. D, phase separation of constitutively shed PrP. SH-SY5Y membranes and conditioned media were subjected to Triton X-114 phase separation, and the resultant detergent-rich (DR) and aqueous (AQ) phases were immunoblotted using antibody 3F4. Multiple immunoblots were quantified by densitometric analysis, and the results were expressed in terms of the percentage distribution of PrP between the two phases (means Ϯ S.D., n ϭ 3).
Branson Sonifier), and nuclei and cell debris were removed by centrifugation at 1000 ϫ g for 10 min. A total membrane fraction was then pelleted by centrifugation at 140,000 ϫ g for 90 min. High speed centrifugation of media to pellet particulate material was performed under the same conditions. PI-PLC Digestion and Triton X-114 Phase Separation-For cleavage of the GPI anchor with PI-PLC, samples were incubated for 1 h at 37°C with 1 unit/ml Bacillus thuringiensis PI-PLC. For Triton X-114 phase separation, samples (50 l) were mixed with 150 l of 10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2% (v/v) precondensed Triton X-114 and successively incubated for 10 min at 4°C and 3 min at 30°C. The sample was then layered over a 6% (w/v) sucrose cushion (300 l), and the aqueous and detergent phases were separated by centrifugation for 3 min at 3000 ϫ g. The volume of the detergent phase was made equal to that of the aqueous phase with 10 mM Tris-HCl pH 7.4, 150 mM NaCl.
SDS-PAGE and Immunoelectrophoretic Blot Analysis-Samples were mixed with an equal volume of reducing electrophoresis sample buffer and boiled for 3 min. Proteins were resolved by SDS-PAGE using 7-17% polyacrylamide gradient gels and transferred to Immobilon P polyvinylidene difluoride membranes as previously described (28). Monoclonal antibody 3F4 (Signet Laboratories, Inc., Dedham, MA) recognizes a 3F4 epitope tag (corresponding to amino acid residues 109 -112 of human PrP) at residues 108 -111 of the murine prion protein and was used at a concentration of 1:4000. Monoclonal antibody SAF-32 (Cayman Chemical, Ann Arbor, MI) recognizes the octapeptide repeat region located in the N-terminal region of PrP and was used at 1:3000. Monoclonal antibody R1 (a gift from Dr. A. Williamson (Department of Immunology, The Scripps Research Institute, La Jolla, CA) recognizes amino acid residues 220 -231 at the C terminus of PrP (29) and was used at 1:5000. Monoclonal antibody 6E10 (Signet Laboratories) recognizes amino acid residues 1-17 of the human amyloid beta sequence and was used at 1:2500. Bound antibody was detected using peroxidaseconjugated secondary antibodies in conjunction with the enhanced chemiluminescence (ECL®) detection method (Amersham Biosciences).

Metalloprotease-mediated Constitutive Shedding of PrP-In
order to elucidate the mechanism involved in constitutive PrP secretion, we compared the shedding of overexpressed wildtype PrP (wt-PrP) from human neuroblastoma SH-SY5Y cells with that of two mutant PrP constructs (Fig. 1A). In PrP-DA, the PrP signal peptide is replaced with the uncleaved signal sequence/transmembrane domain and stalk region from murine aminopeptidase A (30). PrP-PG14 is associated with an inherited prion disease (31) and possesses an additional nine octapeptide repeats. The expression levels of all three forms of PrP were essentially identical as determined by immunoblotting cell lysates with antibody 3F4 (Fig. 1B). Conditioned media obtained following incubation of the cells for 7 h in the absence or presence of the zinc metalloprotease hydroxamatebased inhibitor SB 244000 were also immunoblotted using antibody 3F4 (Fig. 1C). wt-PrP was effectively shed into the medium, and SB 244000 inhibited the shedding by 70%. In contrast, little or no PrP-DA or PrP-PG14 could be detected in conditioned media.
When conditioned medium was subjected to phase separation in Triton X-114 (Fig. 1D), 85.0 Ϯ 8.7% of wt-PrP was detected in the aqueous phase, thereby confirming the hydrophilic nature of the secreted PrP. As a control, membranes isolated from SH-SY5Y cells were also subjected to phase separation (Fig. 1D). In the absence of prior bacterial PI-PLC treatment, 100% of the membrane-associated PrP was located in the detergent-rich phase, and following PI-PLC treatment, all of the PrP was located in the aqueous phase. Immunocharacterization of Constitutively Shed PrP-Lysates and conditioned media from mock-transfected and PrPtransfected SH-SY5Y cells were immunoblotted with the PrP antibodies 3F4, SAF-32, and R1 ( Fig. 2A). All three antibodies detected PrP as a diffuse band between 30 and 45 kDa in both cell lysates and conditioned media (Fig. 1, B-D). The amount of PrP in conditioned media was dramatically reduced in the presence of SB 244000. When deglycosylated samples were immunoblotted using antibody 3F4 (Fig. 2E), two major bands were detected at 27 and 21 kDa in cell lysates, and a 27-kDa band and a slightly less intense doublet at 20 -21 kDa were detected in the conditioned medium. Using antibody SAF-32 (Fig. 2F), a major band at 27 kDa and a much less prominent band at 21-22 kDa were detected in deglycosylated cell lysates. PrP was detected almost exclusively as a 27-kDa polypeptide in conditioned medium using SAF-32. Using antibody R1 (Fig.  2G), PrP was detected as multiple bands in the lysates of PrP-transfected cells. In conditioned media, PrP was detected primarily as 27-and 21-kDa bands, with a faint band also detectable at 24 kDa. Thus, it appears that both full-length PrP and a C-terminal fragment are shed from the cells.
Metalloprotease-mediated PrP Shedding Is Similar to That of the Amyloid Precursor Protein-Since PrP is GPI-anchored but appears to be shed by a metalloprotease, we compared its shedding mechanism with that of a transmembrane polypeptide-anchored protein known to be shed by the action of a metalloprotease, namely the APP. APP is expressed endogenously in SH-SY5Y cells; therefore, its shedding can be com-pared directly with that of PrP overexpressed in the same cells. Cells were incubated in the absence or presence of a range of zinc metalloprotease hydroxamate-based inhibitors (25). Conditioned media were immunoblotted with the APP monoclonal antibody 6E10 (Fig. 3A) and the PrP antibody 3F4 (Fig. 3B). Multiple blots were quantified by densitometric analysis, and the inhibition patterns of APP and PrP shedding were compared quantitatively (Fig. 3C). No significant differences could be detected between the inhibition profiles for APP and PrP shedding with a range of hydroxamate-based compounds.
APP shedding is known to be stimulated by activators of the protein kinase signaling cascade such as PMA (32). Therefore, we compared the PMA-regulated shedding of APP from SH-SY5Y cells with that of PrP. Cells were incubated in the absence or presence of PMA, and conditioned media were immunoblotted with the APP antibody 6E10 (Fig. 4A) or the PrP antibody 3F4 (Fig. 4B). Both APP and PrP shedding were strongly enhanced to a similar extent by PMA treatment, and this regulated shedding of the two proteins was effectively inhibited by SB 244000 (Fig. 4C).
Shedding of Endogenous GPI-anchored Alkaline Phosphatase Does Not Parallel That of PrP-Although hydroxamic acidbased compounds such as SB 244000 are active site-directed zinc metalloprotease inhibitors, we considered the possibility that they may have inhibited PrP shedding via interaction with a phospholipase rather than a protease. We have shown previously that SH-SY5Y cells possess endogenous alkaline phosphatase activity, which, like PrP (33, 34), is localized within lipid rafts (35). In the current study, we show that endogenous alkaline phosphatase is GPI-anchored, since the level of activity in conditioned media was increased 81-fold when cells were treated with exogenous bacterial PI-PLC, and SB 244000 had no effect on this release (Table I). In addition, the endogenous release of alkaline phosphatase was not inhibited by SB 244000 or stimulated by PMA (Table I), implying that these compounds were not exerting their effects on PrP via the inhibition or stimulation of a phospholipase activity.
Metalloprotease-mediated PrP Shedding Is Stimulated by Divalent Copper Ions-Although copper promotes the endocytosis of PrP from the cell surface (36), at lower concentrations we noticed that it promoted the shedding of PrP (Fig. 5). Whereas the expression of PrP was unaltered by copper treatment (Fig. 5A), the level of PrP in the conditioned media was increased 2-fold at a copper concentration of 1.5 M (Fig. 5, B and C). The amount of shed PrP continued to increase until a peak was reached at 5 M copper, after which levels began to decrease again but remained above control levels. Despite SB 244000 having no effect on the levels of PrP in cell lysates (Fig.  5D), copper-induced shedding into the medium was almost abolished by co-incubation with this compound (Fig. 5, E and F), indicating the involvement of a zinc metalloprotease in the copper-induced shedding of PrP.

Shedding of PrP Induced by Filipin and M␤CD Is Not Inhibited by Hydroxamate-based Compounds-
The lipid raft-disrupting agent filipin has previously been shown to stimulate PrP release from cells (9). In order to determine whether lipid raft-disrupting agents stimulate PrP release by increasing the metalloprotease-mediated shedding of the protein, we incubated PrP expressing SH-SY5Y cells with filipin or M␤CD in the absence or presence of SB 244000 (Fig. 6). After 30 min, the release of PrP from control cells was barely detectable on the immunoblots, whereas both filipin (Fig. 6, A and C) and M␤CD (Fig. 6, B and D) dramatically enhanced PrP secretion. However, this enhanced secretion was not inhibited by co-incubation with SB 244000. Thus, the shedding of PrP induced by lipid raft-disrupting agents was not mediated via the action of the zinc metalloprotease.
Filipin/M␤CD-mediated PrP Shedding Involves Cleavage of the GPI Anchor-Triton X-114 phase separation of conditioned media from filipin-or M␤CD-treated cells (Fig. 7A) showed that 89 Ϯ 2 and 88 Ϯ 2%, respectively, of PrP was located in the aqueous phase, indicating that the released PrP was hydrophilic in nature. In addition, PrP could not be pelleted from the media by high speed centrifugation (data not shown). To determine the possible role of the GPI anchor in filipin-mediated PrP shedding, we compared the shedding of wt-PrP with that of PrP-CTM. In PrP-CTM, the C-terminal GPI anchor addition sequence of PrP is replaced with the transmembrane and cytoplasmic domains of angiotensin-converting enzyme (30) (Fig.  7B). Both wt-PrP and PrP-CTM were expressed at identical levels in SH-SY5Y cell lysates (Fig. 7C). Since only the GPIanchored wt-PrP and not PrP-CTM was shed into the medium in response to filipin treatment (Fig. 7D), it would appear that cleavage of the GPI anchor is involved in filipin-mediated PrP shedding. Consequently, we examined whether endogenous GPI-anchored alkaline phosphatase was also shed into the medium in response to filipin treatment (Fig. 7E). The amount of alkaline phosphatase activity in conditioned media from filipin-treated cells was 4-fold higher than that in conditioned media from control cells, indicating that the raft-disrupting agents filipin and M␤CD promote the release of multiple GPIanchored proteins through cleavage of the GPI anchor. DISCUSSION In the present study, we have investigated the mechanisms behind the shedding of PrP C from human neuroblastoma SH-SY5Y cells. In order to determine whether PrP was constitutively shed by means of a specific enzymic mechanism and not through a nonspecific process such as membrane blebbing, we compared the shedding of wt-PrP to that of PrP-DA and PrP-  PG14. PrP-DA possesses an additional membrane anchoring domain at the N terminus (30), and PrP-PG14 is the mouse PrP homologue of a nine-octapeptide insertional mutation, which is associated with an inherited form of Creutzfeldt-Jakob disease in humans (37)(38)(39). PrP-PG14 is known to be resistant to PI-PLC cleavage (40), implying that the addition of the nine octapeptide repeats imparts an additional form of membrane interaction to the protein. In contrast to wt-PrP, neither PrP-DA nor PrP-PG14 was present in the conditioned medium to an appreciable level, indicating that wt-PrP was probably being cleaved from its membrane anchoring domain. Confirmation of this came from the observation that constitutively shed PrP was hydrophilic as assessed by temperature-induced phase separation in Triton X-114 (Fig. 1D) and could not be pelleted by high speed centrifugation (data not shown) indicating the removal of at least the fatty acyl moiety of the GPI anchor upon shedding.
The fact that constitutive PrP shedding was inhibited by SB 244000 suggested the involvement of a zinc metalloprotease in a proteolytic cleavage event. Consequently, we sought to determine the approximate site of cleavage within PrP by immunocharacterization of shed PrP (Fig. 2). Following deglycosylation, antibodies 3F4, SAF-32, and R1 all detected a 27-kDa band in the conditioned media that corresponds to either fulllength or only slightly truncated PrP. Antibodies 3F4 and R1 but not SAF-32 also detected a 21-kDa band in the conditioned media. Since this PrP fragment retained both the C-terminal and 3F4 epitopes but lacked the N-terminal epitope recognized by SAF-32, it most likely represents the C-terminal PrP-21 fragment that results from reactive oxygen species-mediated cleavage within the octapeptide repeat region of the protein (41). The fact that shedding of both the 27-and 21-kDa PrP species was inhibited by SB 244000 implied that PrP was cleaved by a zinc metalloprotease very close to the C terminus. Since antibody R1 recognizes an epitope within the C-terminal amino acid residues 220 -231 (29), the polypeptide chain of PrP must be cleaved very near the GPI anchor attachment site at Ser-231, thereby retaining sufficient amino acid residues on the shed protein to provide a minimum required epitope for the R1 antibody. An alternative explanation is that SB 244000 inhibited a phospholipase capable of cleaving PrP within its GPI anchor. However, this explanation is extremely unlikely for several reasons. First, neither the constitutive shedding of endogenous GPI-anchored alkaline phosphatase nor the shedding induced by exogenous bacterial PI-PLC treatment were inhibited by SB 244000 (Table I). Second, the inhibition profile for a range of structurally variant active site-directed hydroxamate-based zinc metalloprotease inhibitors (25) was exactly the same in relation to PrP shedding as it was to the shedding of the transmembrane polypeptide-anchored APP (Fig. 3). Fi- nally, PMA did not stimulate shedding of endogenous GPIanchored alkaline phosphatase (Table I) but did stimulate the shedding of PrP and APP in an identical manner that was sensitive to inhibition by SB 244000 (Fig. 4). Since inhibition by hydroxamate-based compounds and stimulation by PMA are characteristic features of zinc metalloprotease-mediated ectodomain shedding (21), we conclude that PrP is shed by such a mechanism.
Several studies have identified a secreted form of PrP and speculated that it was derived from GPI-anchored PrP by the action of a phospholipase (10 -12, 14, 16, 17). In contrast, Borchelt et al. (13) previously reported that PrP secreted from primary cultures of neonatal Syrian hamster brain exhibited no change in electrophoretic mobility upon treatment with aqueous hydrofluoric acid, indicative of loss of the entire GPI anchor modification. The authors suggested that PrP may be shed via the action of a protease, but no evidence was presented to support this theory. In addition, a proportion (15%) of PrP Sc extracted from hamster brain was found to terminate at Gly-228 just N-terminal to the site of GPI anchor addition at Ser-231 (42). Whether the zinc metalloprotease-mediated shedding of PrP C that we have reported here is responsible for the generation of truncated forms of PrP Sc awaits determination.
The amino-terminal half of PrP contains a series of histidineand glycine-rich octapeptide repeats that bind copper ions (43). Although at concentrations of 100 M or above, copper ions stimulate the endocytosis of PrP (36,44), at lower concentrations, we observed that they promoted the shedding of PrP and that this shedding was sensitive to SB 244000, implicating the involvement of a zinc metalloprotease (Fig. 5). The apparent decrease in the amount of PrP shed at the higher concentrations (100 M) of copper may be due to competition between the shedding and endocytosis of the protein. The enhanced shedding observed at low concentrations of copper may be due to the metal binding to PrP and inducing a conformational change in the protein that facilitates its cleavage by the zinc metalloprotease. Alternatively, the low concentrations of copper may directly stimulate the activity of the zinc metalloprotease. Interestingly, low micromolar concentrations of copper have been shown to stimulate the nonamyloidogenic shedding of APP, which is also a copper-binding protein (45). It remains to be determined whether the zinc metalloprotease-mediated ectodo- main shedding of other membrane proteins is also stimulated by copper ions.
In contrast to the constitutive PMA-and copper-induced shedding of PrP, the shedding induced by the lipid raft-disrupting agents filipin and M␤CD was not inhibited by SB 244000 (Fig. 6), indicating that the zinc metalloprotease was not involved. Nonetheless, the shed PrP was hydrophilic in nature as assessed by phase separation in Triton X-114 (Fig. 7A) and was not pelleted by high speed centrifugation (data not shown), suggesting partial or complete removal of the GPI anchor. The fact that the transmembrane polypeptide-anchored PrP-CTM construct was not released from cells in response to filipin treatment (Fig. 7D) strongly suggests that cleavage within the GPI anchor moiety, possibly by a phospholipase, is involved in this mechanism of PrP shedding. Further support for this cleavage mechanism comes from the observation that endogenous GPI-anchored alkaline phosphatase was also shed in response to filipin treatment. Filipin has previously been reported to promote the release of PrP C from the surface of neuroblastoma cells, but the mechanism of this release, whether shedding or membrane blebbing, was not investigated (9). Since filipin and M␤CD are structurally unrelated choles-terol-binding compounds but both stimulated PrP shedding, it is likely that their mechanism of action is through disrupting the structure of lipid rafts and the subsequent activation of a phospholipase. Both filipin and M␤CD have been used previously to remove GPI-anchored proteins from lipid rafts with the proteins then appearing to enter the detergent-soluble nonraft regions of the membrane (46,47). However, caution should be exercised in interpreting such results in light of the fact that these lipid raft-disrupting agents can actually induce the shedding of PrP and other GPI-anchored proteins from the cell surface.
In conclusion, we have shown for the first time that PrP can be shed from the surface of human neuroblastoma cells by two distinct mechanisms; one is a phorbol ester-stimulated process that involves the action of a zinc metalloprotease with a similar inhibition profile to the ␣-secretase that cleaves the amyloid precursor protein, and the other is stimulated by the lipid raft-disrupting agents filipin and M␤CD and involves cleavage within the GPI anchor possibly by a phospholipase. Since PrP C is located in lipid rafts at the cell surface (33,34), it is likely that the zinc metalloprotease-mediated shedding takes place within these structures. However, it was not possible to use FIG. 7. Characterization of filipin and M␤CD-induced PrP shedding. A, phase separation of conditioned media from filipin and M␤CD-treated SH-SY5Y cells. Conditioned media were subjected to Triton X-114 phase separation, and the resultant detergent-rich (DR) and aqueous (AQ) phases were immunoblotted using antibody 3F4. Multiple immunoblots were quantified by densitometric analysis, and the results were expressed in terms of the percentage distribution of PrP between the two phases (means Ϯ S.D., n ϭ 3). B, schematic of PrP-CTM. The C-terminal GPI signal sequence of PrP (residues 231-254) was replaced by the transmembrane (black box) and cytoplasmic (heavily shaded) domains of human angiotensin-converting enzyme (ACE). C, expression levels of PrP constructs. Cell lysates were immunoblotted with antibody 3F4. D, filipin-induced shedding of wt-PrP and PrP-CTM. Conditioned media from control and filipin-treated cells were immunoblotted using antibody 3F4. E, filipin-induced shedding of GPI-anchored alkaline phosphatase. Conditioned media from control and filipin-treated cells were assayed for alkaline phosphatase activity as described under "Experimental Procedures." Results are means Ϯ S.D. (n ϭ 3). lipid raft-disrupting agents to establish this, since such reagents increased the shedding of PrP through activation of another sheddase, probably a phospholipase. Since increased shedding of PrP C would reduce the amount of membranebound protein available for subsequent conversion to PrP Sc (9), these mechanisms of cell surface PrP C shedding may be of critical importance in the pathogenesis of transmissible spongiform encephalopathies.