A mutant prion protein displays an aberrant membrane association when expressed in cultured cells.

Inherited forms of prion disease have been linked to mutations in the gene encoding PrP, a neuronal and glial protein that is attached to the plasma membrane by a glycosyl-phosphatidylinositol (GPI) anchor. One familial form of Creutzfeldt-Jakob disease is associated with a mutant PrP containing six additional octapeptide repeats. We report here our analysis of cultured Chinese hamster ovary cells expressing a murine homologue of this mutant PrP. We find that, like wild-type PrP, the mutant protein is glycosylated, GPI-anchored, and expressed on the cell surface. Surprisingly, however, cleavage of the GPI anchor using phosphatidylinositol-specific phospholipase C fails to release the mutant PrP from the surface of intact cells, suggesting that it has an additional mode of membrane attachment. The phospholipase-treated protein is hydrophobic, since it partitions into the detergent phase of Triton X-114 lysates; and it is tightly membrane-associated, since it is not extractable in carbonate buffer at pH 11.5. Whether membrane attachment of the mutant PrP involves integration of the polypeptide into the lipid bilayer, self-association, or binding to other membrane proteins remains to be determined. Our results suggest that alterations in the membrane association of PrP may be an important feature of prion diseases.

Inherited forms of prion disease have been linked to mutations in the gene encoding PrP, a neuronal and glial protein that is attached to the plasma membrane by a glycosyl-phosphatidylinositol (GPI) anchor. One familial form of Creutzfeldt-Jakob disease is associated with a mutant PrP containing six additional octapeptide repeats. We report here our analysis of cultured Chinese hamster ovary cells expressing a murine homologue of this mutant PrP. We find that, like wild-type PrP, the mutant protein is glycosylated, GPI-anchored, and expressed on the cell surface. Surprisingly, however, cleavage of the GPI anchor using phosphatidylinositolspecific phospholipase C fails to release the mutant PrP from the surface of intact cells, suggesting that it has an additional mode of membrane attachment. The phospholipase-treated protein is hydrophobic, since it partitions into the detergent phase of Triton X-114 lysates; and it is tightly membrane-associated, since it is not extractable in carbonate buffer at pH 11.5. Whether membrane attachment of the mutant PrP involves integration of the polypeptide into the lipid bilayer, selfassociation, or binding to other membrane proteins remains to be determined. Our results suggest that alterations in the membrane association of PrP may be an important feature of prion diseases.
The spongiform encephalopathies are a group of transmissible neurodegenerative disorders which are characterized by dementia, motor dysfunction, and in some cases by the presence of cerebral amyloid plaques (reviewed by Gajdusek, 1990;Prusiner and DeArmond, 1994). These fatal disorders include Creutzfeld-Jakob disease (CJD), 1 Gerstmann-Strä ussler syndrome, kuru, and fatal familial insomnia in human beings, and scrapie and bovine spongiform encephalopathy in animals. The infectious agent responsible for these diseases has been called a prion, to distinguish it from a conventional virus (Prusiner, 1982). The major, and possibly only, component of prion particles is a 35-45-kDa glycoprotein called PrP Sc (Prusiner et al., 1984;Bolton et al., 1987). PrP Sc is a post-translationally modified isoform of a glycosyl-phosphatidylinositol (GPI)-anchored, cell surface protein of the host called PrP C (Oesch et al., 1985;Chesebro et al., 1985). Prion replication is hypothesized to result from conversion of endogenous PrP C into infectious PrP Sc , a process that seems to involve species-specific molecular interactions between the two isoforms Bü eler et al., 1993;Kocisko et al., 1994). Recent evidence suggests that the isoforms differ in the conformation of their polypeptide chains, and it has been proposed that conversion of ␣-helices into ␤-sheets underlies the formation of PrP Sc Safar et al., 1993;Pan et al., 1993). The two forms of the protein also exhibit distinct biochemical properties. PrP Sc aggregates in non-denaturing detergents, and it is more resistant to protease digestion than PrP C , yielding an N terminally cleaved core fragment called PrP 27-30 (Oesch et al., 1985;Meyer et al., 1986).
Prion diseases are unique in having both a genetic as well as an infectious origin. Gerstmann-Strä ussler syndrome and fatal familial insomnia, and about 10% of the cases of CJD, are inherited in an autosomal dominant fashion (Prusiner and DeArmond, 1994). These inherited forms have been associated with specific mutations in the chromosomal gene that encodes PrP. Genetic linkage or association has been established for several amino acid substitutions in the C-terminal half of the molecule, as well as for insertions of between two and nine additional octapeptide repeats in the N-terminal half (reviewed by Brown et al., 1991;Prusiner and Hsiao, 1994). A genetic form of prion disease can also be produced in mice harboring a PrP transgene with a point mutation (Hsiao et al., 1990); mouse models of insertion mutations have not been reported.
It is presumed that mutant PrPs spontaneously assume the conformation of PrP Sc in the absence of exogenous prions (Prusiner and DeArmond, 1994). However, there is no experimental evidence for this point, and the cellular mechanisms that might be involved remain unknown. To address these issues, it would be desirable to analyze the processing and metabolism of mutant PrPs in cultured cells. We have previously characterized the trafficking of wild-type PrP C in cultured neuroblastoma cells and have defined the secretory and endocytic pathways followed by the cellular isoform Shyng et al., 1993Shyng et al., , 1994Shyng et al., , and 1995. We have now extended our studies to a murine form of PrP that carries an insertion of six additional octapeptide repeats. This mutation is homologous to one in human PrP that is strongly linked to CJD in a large family from southeast England. This family comprises seven generations, and its members are all descended from a single founder who was born in the late 18th century (Owen et al., 1990;Poulter et al., 1992;Collinge et al., 1992). Recently, two other families with PrP insertions encoding the same amino acids have been published (Nicholl et al., 1995;Oda et al., 1995). We report here that murine PrP containing six additional octapeptide repeats displays an abnormal association with the plasma membrane when expressed in cultured CHO cells.
Antisera-The antibody designated P45-66 was raised by immunization of rabbits with a synthetic peptide (CGGNRYPPQGGTWGQPH-GGGWGQ), encompassing moPrP residues 45-66, that was coupled via the additional N-terminal cysteine residue to maleimide-activated keyhole limpet hemocyanin (Pierce). The rabbit antiserum designated F35-96 was raised against a bacterial fusion protein encompassing chPrP residues 35-96 . Dr. Rick Kascsak (Institute for Basic Research, Staten Island, NY) kindly provided 3F4, a mouse monoclonal antibody that was raised against PrP 27-30 from infected hamster brain and that reacts with hamster but not mouse PrP (Kascsak et al., 1987). A mouse monoclonal antibody (TD.1) against the heavy chain of clathrin from bovine brain was a gift of Dr. Frances Brodsky (University of California at San Francisco).
PrP Constructs- Fig. 1 shows the structure of the murine PrP constructs used in this study. The coding region of wild-type moPrP (Prn-p a allele) was amplified by PCR from the plasmid N2TBX1 (ATCC 63010) using Vent polymerase (New England Biolabs) and the following primers: 5Ј-GACCAGAAGCTTATGGCGAACCTTGGCTACTGG-3Ј (primer 1) and 5Ј-GACCAGGGATCCTCATCCCACGATCAGGAAGAT-3Ј(primer 2). After cleavage with HindIII and BamHI, the PCR product was cloned into the expression vector pBC12/CMV that had been cleaved with the same two enzymes. This vector places inserted sequences under control of the human cytomegalovirus promoter/enhancer (Cullen, 1986).
To make PG11 moPrP, the following primers were first used to amplify the octapeptide repeat region of wild-type moPrP: 5Ј-GACCA-GAAGCTTATGGCGAACCTTGGCTACTGG-3Ј (primer 3) and 5Ј-AT-TATGGGTACCCCCTC-3Ј (primer 4). The amplified product was cleaved with BstXI, and a fragment of 72 base pairs comprising portions of the three central octapeptide repeats (R2-R3-R3 in Fig. 1) was isolated. This fragment was then ligated to itself, and after size fractionation of the ligation products on an agarose gel, a 216-base pair fragment was recovered. This fragment was then ligated to pBC12/CMV containing wild-type moPrP, after a 72-base pair fragment had been excised by digestion with BstXI.
To create moPrPs tagged with the 3F4 antibody epitope, the following PCR primers were employed: 5Ј-TGCCGCAGCCCCTGCCATAT-GCTTCATGTTGGTTTTTGGTTTGC-3Ј (primer 5), and 5Ј-GCAAAC-CAAAAACCAACATGAAGCATATGGCAGGGGCTGCGGCA-3Ј (primer 6). Untagged wild-type and PG11 cDNAs were first amplified with primer 1 and primer 5, and with primer 2 and primer 6. These two primary PCR products were then annealed with each other and amplified again with primer 1 and primer 2. After cleavage with HindIII and BamHI, the secondary PCR product was cloned into pBC12/CMV that had been cleaved with the same two enzymes.
The structure of each DNA construct was confirmed by sequencing. We note that moPrP contains a methionine residue at position 128, which is homologous to position 129 in human PrP where a valine/ methionine polymorphism has been described; the PG11 gene in humans is always found to contain a methionine residue .
Cell Lines-Chinese hamster ovary (CHO) cells were grown in MEM-␣ containing 7.5% fetal calf serum and penicillin/streptomycin in an atmosphere of 5% CO 2 , 95% air. CHO cells were transfected using Lipofectin (Life Technologies Inc.), according to the manufacturer's directions, with a 1:10 mixture of pRSVneo (Ulrich and Ley, 1990) and the pBC12/CMV expression plasmid encoding moPrP. Antibiotic-resistant clones were selected in 700 g/ml geneticin (G418), subcloned once, and maintained in 300 g/ml geneticin. The experiments described here were carried out on a single cell line expressing each construct.
Immunocytochemistry-Cells grown on glass coverslips were incubated with rabbit polyclonal antibody P45-66 for 1 h at 4°C. After rinsing, cells were fixed with methanol for 30 min at 4°C and then incubated with fluorescein-conjugated secondary antibody for 1 h at room temperature. Coverslips were mounted in glycerol and cells observed with a Zeiss Axioplan fluorescence microscope equipped with a Bio-Rad MRC1000 laser confocal scanning system.
Metabolic Labeling, Immunoprecipitation, and Biotinylation-Confluent cultures of CHO cells were labeled either in methionine-free MEM containing Tran 35 S-Label (250 -500 Ci/ml); in MEM plus 10% dialyzed fetal calf serum and non-essential amino acids, containing [ 3 H]ethanolamine (200 Ci/ml), or in DMEM plus 10 mg/ml fatty acidfree bovine serum albumin, containing a mixture of [ 3 H]stearic acid (250 Ci/ml) and [ 3 H]palmitic acid (1,250 Ci/ml). In some experiments, labeled cultures were chased in Opti-MEM (Life Technologies Inc.) for various times before harvesting. Cells were lysed in a buffer that contained 150 mM NaCl, 50 mM Tris, pH 7.5, 0.5% Triton X-100, 0.5% sodium deoxycholate and that was supplemented with protease inhibitors (1 g/ml pepstatin and leupeptin, 0.5 mM phenylmethylsulfonyl fluoride, 2 mM EDTA). Prior to lysis, cultures in some experiments were incubated with PIPLC (1 unit/ml in Opti-MEM) at 4°C for 2 h, or with 0.25% trypsin in PBS at 4°C for 10 min, followed by addition of fetal calf serum to 10% to terminate digestion.
Samples from metabolically labeled cells were immunoprecipitated as described previously , with the modification that immunocomplexes were collected using protein A-Sepharose. Immunoprecipitated proteins were analyzed by SDS-PAGE, and radioactive gels were quantitated using a PhosphorImager (Molecular Dynamics).
Surface biotinylation was carried out as described previously (Shyng et al., 1995). Rabbit anti-mouse IgG was included when biotinylated proteins were immunoprecipitated with 3F4 antibody.
Deglycosylation of PrP-Where indicated, cell lysates were treated prior to immunoprecipitation with 0.01 units/ml of N-glycosidase F for 16 h at 37°C. For the experiments shown in Fig. 4, immunoprecipitated PrP was eluted from protein A-Sepharose with 20 l of 50 mM sodium citrate, pH 5.5, containing 0.1% SDS, and then incubated for 16 h with either endoglycosidase H (1.6 milliunits) or neuraminidase (10 milliunits). After addition of 2 ϫ SDS sample buffer, samples were analyzed by SDS-PAGE and autoradiography.
Triton X-114 Phase Separation-Surface-biotinylated cells were solubilized at 4°C in 1% Triton X-114 in PBS containing protease inhibitors; the detergent was diluted from a 12% stock solution that had been precondensed according to Bordier (1981). After incubation at 37°C for 20 min, aqueous and detergent phases were separated by centrifugation. The detergent phase was diluted to the initial volume with PBS, incubated with or without PIPLC for 2 h at 4°C, and the phase separation repeated. PrP in each phase was immunoprecipitated and biotinylated proteins visualized as described (Shyng et al., 1995).
Alkaline Carbonate Extraction-Confluent CHO cells expressing PG11 moPrP protein were surface biotinylated as described above and incubated with PIPLC for 2 h at 4°C. Cells were then collected in PBS and passed nine times through 27-gauge needles connected by plastic tubing. The homogenate was centrifuged for 3 min at 300 ϫ g to remove nuclei and debris and then centrifuged for 5 min at 16,000 ϫ g to collect membranes. Membrane pellets were resuspended in either PBS or carbonate buffer (0.1 M, pH 11.5), and kept on ice for 10 min. After a further centrifugation for 5 min at 16,000 ϫ g, PrP was immunoprecipitated from pellets and supernatants; each fraction was also immunoblotted for clathrin heavy chain using antibody TD.1. It should be noted that the 5-min centrifugation at 16,000 ϫ g used to collect membranes is not sufficient to pellet PG11 moPrP protein from detergent lysates (not shown). Fig. 1 shows that wildtype mouse PrP (moPrP) contains five octapeptide repeats, each of which displays one of three related amino acid sequences: PQGGTWGQ (R1), PHGGGWGQ (R2), or PHGG-SWGQ (R3). The repeats are arranged in the order R1-R2-R3-R3-R2. We have constructed a mutant form of moPrP that contains 11 octapeptide repeats (designated PG11, since the repeats are rich in proline and glycine). The PG11 protein contains two additional copies of the sequence R3-R3-R2 (blocks labeled ϩ1 through ϩ6 in Fig. 1). To detect moPrP synthesized by transfected cells, we used a rabbit anti-peptide antibody designated P45-66 that is directed against moPrP residues 45-66 and that does not cross-react with hamster PrP.

CHO Cells Expressing Murine PrPs-
In some experiments, we used epitopically tagged versions of wild-type and PG11 moPrP, in which methionine residues were substituted for a leucine and a valine residue at two nearby positions ( Fig. 1). Methionine residues are found at the homologous positions in hamster and human PrP, and their introduction into moPrP allows the protein to be recognized by the species-specific monoclonal antibody 3F4 (Bolton et al., 1991). MoPrPs tagged with the 3F4 epitope have been used extensively in cell biological and transgenetic studies (Rogers et al., 1993;Scott et al., 1993).
We have expressed each of these moPrP molecules in stably transfected lines of CHO cells, which synthesize undetectable levels of endogenous hamster PrP (data not shown). A single cell line was analyzed for each construct. The unusual properties of PG11 moPrP that we describe here are not peculiar to a particular clone of cells, however, since both the untagged and 3F4-tagged versions of the mutant protein, expressed in separate lines, behaved similarly (see Fig. 6C). We noted that even though they had been subcloned, the lines expressing PG11 moPrP showed some decrease in their levels of PrP expression during prolonged passage. In most experiments, cells were used when the amounts of wild-type and mutant proteins were roughly equivalent (Figs. 2, 3, 4, 6, and 7). In two experiments (Figs. 5 and 8A), the amount of PG11 PrP was approximately 3-fold less than the amount of wild-type PrP. In none of our experiments, however, have we found that differences in expression levels had any effect on the results obtained.
Both Wild-type and PG11 moPrPs Are Delivered to the Cell Surface-PrP C is a cell surface protein that is attached to the plasma membrane by a GPI anchor (Stahl et al., 1987). To determine whether PG11 moPrP was also present on the cell surface, we examined transfected CHO cells by immunofluorescence microscopy. Wild-type and PG11 moPrP were both detected on the surface of intact cells using confocal microscopy. In optical sections through the middle of the cell, staining appeared in small, bright patches around the cell periphery (Fig. 2, A and B). This pattern was quite distinct from the faint, nonspecific cytoplasmic fluorescence seen in untransfected CHO cells (Fig. 2C), or in transfected cells stained with anti-body that had been preincubated with the peptide immunogen (not shown). The punctate staining pattern seen for these mo-PrPs is similar to what we have described for chicken PrP expressed in neuroblastoma cells (Shyng et al., 1994). At this level of analysis, we could not discern any consistent differences between the staining patterns of the wild-type and mutant proteins.
These results suggested that the PG11 protein reached the cell surface after synthesis, in contrast to some other mutant proteins which remain trapped intracellularly as a result of aberrant folding of the polypeptide chain in the endoplasmic reticulum (Amara et al., 1992). To directly examine the kinetics with which the mutant protein was delivered to the cell surface, we performed a metabolic pulse-chase experiment (Fig. 3). We quantitated the amount of PrP transported to the cell surface by determining the proportion of labeled protein that was digested after incubation of intact cells with trypsin. In this experiment, we found that both wild-type and PG11 forms of moPrP were transported to the surface with similar kinetics (Fig. 3B). The two proteins both became accessible to external trypsin by 20 min of chase, and by 40 min approximately 60% was susceptible to digestion by the enzyme. Similar kinetics have been reported for delivery of pulse-labeled molecules of chicken PrP  and moPrP (Caughey et al., 1989;Borchelt et al., 1992) to the surface of mouse neuroblastoma cells.
Both Wild-type and PG11 moPrPs Are Glycosylated-We noted in the experiment shown in Fig. 3A that there was a marked shift in the molecular weight distribution of the labeled PrP molecules during the chase period. At the end of the pulse, PrP migrated as a series of three bands, ranging in size from 28 to 33 kDa for the wild-type protein and from 30 to 40 kDa for the PG11 protein. By 40 min of chase, the majority of the protein had shifted to 40 kDa for the wild-type form and to 40 -45 kDa for PG11. These time-dependent alterations in the molecular mass of each protein reflect changes in N-linked glycosylation, since N-glycosidase F digestion at each time point produces a single band at 29 kDa for wild-type moPrP and 33 kDa for PG11 (see Fig. 5A). These results indicate that the PG11 mutation does not block glycosylation of PrP. However, the pattern of glycosylation appears to be affected, since the mutant protein displays a greater proportion of lower molecular mass glycoforms than the wild-type protein (see Fig. 6A, lanes 1 and 6 and Fig. 6C, lanes 1 and 4).
To further explore the glycosylation of PG11 moPrP, we have tested the susceptibility of the protein to digestion by endoglycosidase H and neuraminidase after pulse-labeling. During the chase period, both wild-type and PG11 PrPs become resistant to digestion by endoglycosidase H (Fig. 4A) and sensitive to digestion by neuraminidase (Fig. 4B). These results indicate that both proteins traverse the mid-Golgi stack (site of acqui- Numbers beneath the schematics indicate amino acid positions. The numbers ϩ1 through ϩ6, above the PG11 schematic, indicate additional octapeptide repeats. The lollipop symbols indicate N-glycosylation sites, and GPI indicates the site of attachment of the glycolipid anchor. One set of constructs (untagged) had a leucine residue at wild-type position 108 (156 in PG11) and a valine residue at wild-type position 111 (159 in PG11). A second set of constructs (3F4-tagged) had methionine residues substituted at these positions, which allowed the proteins to be recognized by the monoclonal antibody 3F4. sition of endoglycosidase H resistance) and the trans-Golgi network (site of sialic acid addition) during their passage to the surface of CHO cells. Maturation of the oligosaccharide chains of moPrP to an endoglycosidase H-resistant form has also been observed in neuroblastoma cells (Caughey et al., 1989;Taraboulos et al., 1992). Taken together, our results indicate that the presence of additional octapeptide repeats does not markedly affect the kinetics or route of biosynthetic processing of PrP, but does affect the specific pattern of glycosylation.
MoPrP PG11 Is Not Released from the Cell Surface by PIPLC-Because of its GPI anchor, PrP C can be released from the surface of intact cells by digestion with the bacterial enzyme PIPLC, which removes the diacylglycerol 2 portion of the anchor (Englund, 1993). Since our previous results indicated that PG11 moPrP was present on the cell surface, we asked whether it was PIPLC-releasable like wild-type moPrP.
We first examined this question by performing a pulse-chase labeling experiment similar to that shown in Fig. 3, but using PIPLC rather than trypsin (Fig. 5). Cells were incubated with PIPLC at each time point and then lysed; the amount of PrP in cell lysates and PIPLC incubation media was determined by immunoprecipitation. During the chase period, wild-type mo-PrP gradually became accessible to digestion by PIPLC, so that by 30 min almost 40% of the protein was recovered in the PIPLC incubation medium (Fig. 5B). This increase reflects delivery of newly synthesized molecules to the cell surface, where they were susceptible to release by the phospholipase. The kinetics of surface delivery of wild-type moPrP measured in this way was quite similar to that measured using external trypsin (compare Figs. 3B and 5B).
Surprisingly, a very different result was obtained with CHO 2 We will refer to the portion of the anchor removed by PIPLC as diacylglycerol, although some anchors are known to contain alkylacylglycerol (Ferguson and Williams, 1988). MoPrP was immunoprecipitated from cell lysates using antibody P45-66, and was then either left untreated (Ϫ lanes) or was digested with neuraminidase (ϩ lanes) prior to SDS-PAGE. After 40 min of chase, both wild-type and PG11 moPrPs show an increase in electrophoretic mobility following treatment with neuraminidase, indicating acquisition of sialic acid residues. cells expressing the PG11 form of moPrP. Even after 30 min of chase, very little of the mutant protein was released by PIPLC, and most of it remained associated with the cell lysates (Fig.  5B). A similar result was obtained when cells were labeled continuously for 4 h with [ 35 S]methionine, eliminating the possibility that PG11 molecules become releasable by PIPLC at later times (Fig. 5C). The fact that PG11 moPrP cannot be released by the PIPLC contrasts with our observation that almost half of the mutant protein is susceptible to digestion by external trypsin 30 min after pulse labeling (Fig. 3B). Taken together, these data indicate that the PG11 protein is not releasable by PIPLC, even though substantial amounts of it are present on the cell surface.
To confirm this conclusion, we analyzed the effect of PIPLC on cell surface PrP that had been labeled by biotinylation of intact cells with the membrane-impermeant reagent sulfo-biotin-X-NHS. We found that ϳ90% of wild-type moPrP was released by PIPLC, in contrast to Ͻ5% of PG11 moPrP (Fig. 6A,  lanes 1, 2, 5, and 6). As a control to confirm that only PrP molecules on the cell surface had been biotinylated, we found that both forms of PrP were completely digested by externally applied trypsin (Fig. 6A, lanes 3, 4, 7, and 8). Lack of release of PG11 did not result from a slower action of PIPLC on the mutant protein, since extending incubation with the phospholipase to as long as 6 h failed to liberate any additional PrP (Fig. 6B). We obtained similar results after surface iodination of cells, indicating that the failure of PIPLC to release PG11 is not an artifact of biotinylation (data not shown).
To confirm that the unusual behavior of the mutant protein was not a peculiarity of the particular clone of CHO cells we were analyzing, we performed an identical biotinylation experiment on a separate pair of cell lines that expressed moPrPs tagged with the 3F4 antibody epitope (Fig. 6C). In agreement with our previous results, we found that Ͼ90% of wild-type moPrP was released by PIPLC, in contrast to ϳ5% of the PG11 mutant. We noted in this and other experiments that the small amount of the mutant protein released was usually the most heavily glycosylated isoform (Fig. 6C, lane 3). In preliminary FIG. 5. Pulse-labeled molecules of wild-type, but not PG11, moPrP are releasable by external PIPLC during the chase period. Panel A, CHO cells expressing untagged wild-type (WT) or PG11 moPrP were labeled with [ 35 S]methionine for 20 min and then chased in medium containing unlabeled methionine for 0, 15, or 30 min. At the end of the chase period, cells were treated with PIPLC for 2 h at 4°C before lysis. MoPrP in cell lysates (Cells) and PIPLC incubation medium (Medium) was immunoprecipitated with antibody P45-66 after treatment with N-glycosidase F to remove N-linked oligosaccharides. A single band of PrP is produced after deglycosylation (arrows). Exposure time was 48 h for both gels. Panel B, PrP bands in panel A were quantitated by PhosphorImager analysis, and the amount of PrP released by PIPLC was plotted as a percentage of the total amount of PrP present at each time point (medium ϩ cells). Panel C, CHO cells expressing untagged wild-type (WT) or PG11 moPrP were labeled with [ 35 S]methionine for 4 h and were then processed as described in panel A. PrP bands were quantitated by PhosphorImager analysis, and the amount of PrP released by PIPLC was plotted as a percentage of the total amount of PrP present (medium ϩ cells). Each bar represents the mean Ϯ S.D. of three separate experiments.
FIG. 6. Surface-biotinylated molecules of wild-type, but not PG11, moPrP are released by external PIPLC. Panel A, CHO cells expressing untagged wild-type (WT) or PG11 moPrP were biotinylated at 4°C with the membrane-impermeant reagent sulfo-biotin-X-NHS. One set of cultures was then treated with PIPLC for 2 h at 4°C prior to lysis, and moPrP in cell lysates (lanes 2 and 6) and PIPLC incubation medium (lanes 1 and 5) was analyzed by immunoprecipitation with antibody P45-66. Another set of cultures was treated with trypsin for 10 min at 4°C (lanes 4 and 8), and a third set was left undigested (lanes  3 and 7), after which cells were lysed and moPrP immunoprecipitated. SDS-polyacrylamide gels of the immunoprecipitates were electroblotted, and blots were developed with horseradish peroxidase-streptavidin and ECL to visualize biotinylated PrP. Panel B, CHO cells expressing untagged wild-type (WT) or PG11 moPrP were biotinylated at 4°C with sulfo-biotin-X-NHS and then treated with PIPLC at 4°C for the indicated times. MoPrP in cell lysates and PIPLC incubation medium was then analyzed as described in panel A. PrP bands on the ECL film were quantitated by densitometry, and the amount of PrP released by PIPLC was plotted as a percentage of the total amount of PrP (medium ϩ cell lysates). Each point represents the mean Ϯ S.D. of three separate experiments. Panel C, CHO cells expressing 3F4-tagged wildtype (WT) or PG11 moPrP were biotinylated at 4°C with sulfo-biotin-X-NHS and were then treated with PIPLC and processed as described in panel A, except that 3F4 was used as the primary antibody. experiments, we have also observed that PIPLC fails to release human PG11 PrP from the surface of biotinylated CHO cells (data not shown). We have also found that similar results are obtained whether CHO cells are stably or transiently transfected (not shown).
PG11 moPrP Contains a GPI Anchor That Is PIPLC-cleavable-The PG11 protein may be resistant to release from the cell surface by PIPLC because it lacks a GPI anchor and is instead associated with the plasma membrane by another kind of attachment. To address this possibility, we incubated CHO cells with either [ 35 S]methionine to label the polypeptide chain, or with [ 3 H]ethanolamine, which is a constituent of the core structure of the GPI anchor (Englund, 1993). Both radioactive precursors were found to be incorporated into wild-type as well as PG11 molecules (Fig. 7A). The ratio (in arbitrary units) of 3 H to 35 S, calculated by densitometric scanning of the autoradiograms, was 1.3 for wild-type PrP and 1.1 for the mutant, indicating that both proteins carry a GPI anchor structure containing the same number of ethanolamine residues. PrP Sc purified from infected hamster brain has been reported to contain two ethanolamine residues in its GPI anchor (Stahl et al., 1992), although the anchor structure of PrP synthesized by cultured cells is unknown.
Our data thus far indicated that PG11 moPrP molecules are GPI-anchored, even though they cannot be released from the cell surface by digestion with PIPLC. A possible explanation for this fact might be that the mutant protein contains a modified GPI anchor that is intrinsically resistant to PIPLC cleavage, as has been reported for anchor structures in which the inositol ring is acylated (Rosenberry, 1991). Alternatively, aggregation of the protein on the cell surface, or interaction with other membrane components, might render the anchor physically inaccessible to the phospholipase. To investigate these possibilities, we incubated CHO cells with a mixture of [ 3 H]palmitate and [ 3 H]stearate in order to label the fatty acyl residues of the GPI anchor and then treated the intact cells with PIPLC. Since PIPLC cleaves the diacylglycerol moiety from the GPI anchor, the activity of this enzyme would be revealed by a loss of incorporated 3 H label after PIPLC treatment. We also labeled parallel cultures with [ 35 S]methionine to track the polypeptide chain. We found that both wild-type and PG11 moPrP molecules were labeled with 3 H-fatty acids, consistent with the presence of a GPI anchor on both proteins (Fig. 7B,  lanes 5 and 7). Importantly, PIPLC treatment of the cells removed the 3 H label from the PG11 protein almost completely (Fig. 7B, lane 8). Analysis of the [ 35 S]methionine-labeled dishes confirmed that after removal of the diacylglycerol moiety by PIPLC, PG11 molecules were retained on the cell surface (Fig.  7B, lane 4), while the wild-type molecules were released into the medium (Fig. 7B, lane 2). This result demonstrates that the GPI anchor of the PG11 mutant is cleavable by PIPLC, but that cleaved molecules are not released from the cell surface. Consistent with this conclusion, we also observed a small (1-2 kDa) decrease in SDS-PAGE mobility of the PG11 protein after PIPLC treatment (Fig. 7B, lanes 3 and 4). This shift in mobility is characteristic of many GPI-anchored proteins after digestion with PIPLC and is probably due to changes in SDS binding after removal of diacylglycerol (Ferguson and Williams, 1988;Rosenberry, 1991).
The PG11 moPrP Molecule Is Hydrophobic, and Carbonate Inextractable, Even after Cleavage of Its GPI Anchor-The fact that the PG11 protein is retained on the cell surface even after removal of the GPI anchor by PIPLC suggests that the polypeptide chain has an additional mode of attachment to the plasma membrane. One possibility is that the polypeptide chain interacts directly with the lipid bilayer. Alternatively, the protein may aggregate with itself or bind tightly to another membrane protein. To investigate these possibilities, we performed two tests that are commonly used to distinguish integral from peripheral membrane proteins.
First, we subjected moPrP to phase partitioning in the detergent Triton X-114. In this method, integral membrane and other hydrophobic proteins are found in the detergent phase, while peripheral membrane and soluble proteins are found in the aqueous phase (Bordier, 1981). We found that before treatment with PIPLC, both wild-type and PG11 moPrPs partitioned almost exclusively into the detergent phase, as would be expected for molecules carrying an intact GPI anchor (Fig. 8A,  lanes 1 and 2 and 5 and 6). PIPLC treatment shifted the wild-type protein almost completely into the aqueous phase (Fig. 8A, lanes 3 and 4 and Fig. 8B), a phenomenon that is observed for other GPI-anchored proteins (Englund, 1993), and which results from loss of the hydrophobic diacylglycerol por-  5-8). Cells then either remained untreated (Ϫ lanes) or were incubated for 2 h at 4°C with PIPLC (ϩ lanes) prior to lysis. MoPrP was then immunoprecipitated from cell lysates after enzymatic deglycosylation and analyzed by SDS-PAGE. Arrowheads indicate PrPspecific bands. A small (1-2 kDa) decrease in the electrophoretic mobility of PG11 PrP after cleavage with PIPLC can be seen by comparing lanes 3 and 4; this shift is more obvious in other experiments using longer gel runs (data not shown). tion of the anchor. In contrast, the mutant protein was only partially shifted, with over half remaining in the detergent phase after PIPLC treatment (Fig. 8A, lanes 7 and 8 and Fig.  8B). As a control, we analyzed the phase-partitioning of a transmembrane protein, a chPrP/mannose 6-phosphate receptor chimera (Gorodinsky et al., 1995), and found that it was almost completely retained in the detergent phase, whether or not it was treated with PIPLC (Fig. 8A, lanes 9 -12 and Fig.  8B). We thus conclude that, even after removal of the diacylglycerol portion of the GPI anchor, the PG11 molecule displays substantial hydrophobicity.
As an additional test of the nature of the association between PG11 and the plasma membrane, we determined whether the protein was extractable by carbonate buffer at pH 11.5. Because many protein-protein interactions are disrupted at this pH, carbonate inextractability is commonly used as a criterion for stable integration of a polypeptide into the lipid bilayer (Fujiki et al., 1982). We prepared a membrane fraction from biotinylated CHO cells that had been treated with PIPLC and incubated it on ice with either PBS or carbonate buffer. We then collected the membranes by centrifugation and assayed the amount of PrP in the pellet and supernatant. We found that, even after PIPLC treatment, very little of the PG11 protein was extractable from the membranes by carbonate buffer (Fig. 8C, lanes 3 and 4), consistent with an alkaline-stable association. As a control for the effectiveness of the extraction conditions, we immunoblotted the supernatant and pellet fractions using an antibody to the heavy chain of clathrin and found that the majority of this peripheral membrane protein was solubilized by carbonate buffer, but not by PBS (Fig. 8C,  lanes 5-8). DISCUSSION The data reported here demonstrate that a mutant form of moPrP containing 11 octapeptide repeats displays an aberrant mechanism of membrane association when expressed in cultured CHO cells. These results are directly relevant to human prion diseases, since an insertional mutation in the PrP gene that results in 11 octapeptide repeats has been strongly linked to familial CJD in a large English family Collinge et al., 1992) and has also been described in two smaller families (Nicholl et al., 1995;Oda et al., 1995). This report presents the first description of an abnormality in the biochemical properties of a mutant PrP molecule, and it raises the possibility that an alteration in the membrane attachment of PrP is a critical feature of prion diseases.
We have presented several lines of evidence indicating that, while both wild-type and PG11 forms of PrP are glycosylated and delivered to the cell surface, they differ dramatically in their mode of membrane association. The surface localization of both proteins in intact cells has been shown by immunofluorescence staining, as well as by accessibility to externally applied biotinylation and iodination reagents, and to trypsin. Both wild-type and PG11 PrPs also contain a GPI anchor, as demonstrated by metabolic incorporation of [ 3 H]ethanolamine and 3 H-fatty acids, which are anchor constituents. The critical difference in membrane attachment between wild-type and PG11 PrP is revealed by digestion of intact cells with PIPLC, which is a bacterial enzyme that specifically cleaves the GPI anchor, removing the diacylglycerol portion that is embedded in the lipid bilayer. While the wild-type protein is released nearly quantitatively by PIPLC, the PG11 protein is almost completely retained on the cell surface after digestion. We have shown that the inability of PIPLC to release PG11 PrP is not due to a failure to cleave the GPI anchor, since the phospholipase almost completely removes 3 H-fatty acid label from the protein when applied to intact cells, and produces a character- FIG. 8. PIPLC-treated PG11 PrP is retained in the detergent phase after Triton X-114 phase partitioning and is not extractable with alkaline carbonate buffer. Panel A, CHO cells expressing untagged wild-type (WT) or PG11 moPrP, and N2a cells expressing a chPrP/M6PR transmembrane chimera, were surface-biotinylated, and lysed at 4°C in a buffer containing 1% Triton X-114. After the temperature was raised to 37°C, the detergent phase was recovered, diluted to the original volume, and split in half. Half was incubated with PIPLC at 4°C for 2 h (ϩ lanes), and the other half was left untreated (Ϫ lanes). Phase separation was then repeated, and moPrP in the detergent (D) and aqueous (A) phases was immunoprecipitated with antibody P45-66 (lanes 1-8); chPrP/M6PR was immunoprecipitated with antibody F35-96 (lanes 9 -12). SDS-polyacrylamide gels of the immunoprecipitates were electroblotted, and the blots developed with horseradish peroxidase-streptavidin and ECL to visualize biotinylated proteins. Panel B, PrP bands from the experiment shown in panel A, and from two additional experiments, were quantitated by densitometry, and the amount of PrP in the detergent phase after PIPLC treatment (lanes 3, 7, and 11) was plotted as a percentage of the total amount of PrP present (detergent ϩ aqueous phases). Each bar represents the mean Ϯ S.D. Panel C, CHO cells expressing PG11 moPrP were surface-biotinylated and incubated with PIPLC for 2 h at 4°C. Cells were then homogenized in PBS, and a postnuclear membrane fraction prepared. Membranes were resuspended either in PBS or in carbonate buffer and were then collected by centrifugation. MoPrP in the pellet and supernatant fractions was then analyzed by immunoprecipitation, followed by SDS-PAGE, electroblotting, and visualization with HRP-streptavidin and ECL (lanes 1-4; bracket indicates PrP-specific bands). Pellet and supernatant fractions were also immunoblotted to detect clathrin heavy chain (arrow), a marker peripheral membrane protein (lanes 5-8).
istic decrease in its electrophoretic mobility. These observations make it unlikely that aggregation of the mutant protein on the cell surface, or interaction with other membrane components, renders the GPI anchor inaccessible to enzymatic cleavage, at least for the majority of the molecules; however, our data do not rule out the possibility that the anchors of a small proportion of the molecules are resistant to cleavage.
The results indicate that PG11 PrP must have a second mechanism of membrane attachment in addition to the GPI anchor. Several possible models can be envisioned. One is that the PG11 polypeptide chain is directly integrated into the lipid bilayer. In support of this proposal, we find that after cleavage of the GPI anchor with PIPLC, 65% of the mutant PrP still partitions into the detergent phase after Triton X-114 phase separation (Bordier, 1981), in comparison to 15% for wild-type PrP (Fig. 8B). These results indicate that the PG11 molecule is quite hydrophobic, even after removal of diacylglycerol from the GPI anchor, suggesting that a portion of the protein has the potential to interact with lipids. The fact that 85% of a bona fide transmembrane protein was retained in the detergent phase after partitioning in Triton X-114 (Fig. 8B) might indicate that PG11 possesses a shorter or less hydrophobic lipidassociated domain than the transmembrane protein. Direct association between the polypeptide chain and the lipid bilayer is also suggested by failure to release PIPLC-digested PG11 from isolated membranes using alkaline carbonate buffer, which disrupts many protein-protein interactions. We note that transmembrane forms of PrP have been observed when PrP mRNA is translated in cell-free systems (Hay et al., 1987;Yost et al., 1990;Lopez et al., 1990;De Fea et al., 1994) and that several other proteins are proposed to have both a GPI anchor and a membrane-integrated polypeptide chain (Köster and Strand, 1994;Hitt et al., 1994;Howell et al., 1994). Nevertheless, it will be necessary to employ additional techniques, such as labeling with lipid-soluble reagents and mapping the accessibility of sequence epitopes, to decide whether or not the PG11 polypeptide chain loops into or traverses the lipid bilayer. A second possible model is that PG11 moPrP binds tightly to other membrane-associated molecules, thereby preventing release of mutant PrP after cleavage of its PIPLC anchor. This hypothetical interaction would have to be stable in carbonate buffer at pH 11.5. One potentially relevant class of surface binding sites is proteoglycan molecules. PrP C binds to heparinagarose beads , and we have found that PrP-containing bacterial fusion proteins bind to glycosaminoglycan sites on the surface of cultured cells. 3 In this model, one would have to postulate that the conformation of the PG11 molecule is altered in such a way that it binds more tightly than the wild type molecule to membrane-associated sites.
A final possibility is that PG11 moPrP forms aggregates on the cell surface and that these remain attached to the membrane after PIPLC treatment by virtue of a small number of molecules whose GPI anchors have escaped, or are resistant to, cleavage. In fact, we have found the PG11 protein expressed in CHO cells forms detergent-insoluble aggregates that can be recovered by ultracentrifugation. 4 Arguing against an effect of aggregation, however, is the high efficiency of PIPLC cleavage (ϳ90%, as judged by loss of 3 H-fatty acid label; Fig. 7B, lanes 7  and 8); and the observation that no additional PG11 protein is released by prolonging the period of PIPLC digestion to 6 h (Fig. 6B).
We have noted that the glycosylation pattern of PG11 PrP differs from that of wild-type PrP, with the mutant protein displaying a larger proportion of less glycosylated forms (Fig.  6A). We have also observed that the small amount (1-5%) of PG11 PrP that can be released by PIPLC is usually the most heavily glycosylated form (Fig. 6C). These results suggest that there may be a relationship between the altered glycosylation state of mutant PrP and its aberrant membrane association, a proposal consistent with the observation that PrP Sc synthesized in infected neuroblastoma cells also displays a predominance of less glycosylated forms (Caughey and Raymond, 1991). We have recently found that PG11 moPrP produced in CHO cells displays two of the characteristic biochemical properties of PrP Sc , aggregation in non-ionic detergent, and resistance to protease digestion. 4 We have not yet tested whether the protein is infectious. It will be important now to determine how these scrapie-like properties are related to the abnormal membrane association of PG11, and whether the conformational changes that underlie generation of PrP Sc involve alterations in membrane attachment. Although there are no data available on the protease sensitivity of PG11 PrP from human patients, these individuals do not develop symptoms until adulthood , suggesting that accumulation of PrP Sc is a slow process that may require many years. Since we find that PG11 moPrP molecules in CHO cells acquire an aberrant membrane association, and become protease-resistant and detergent-insoluble, 4 within a matter of hours, the generation of PrP Sc may be more efficient in CHO cells than in neurons in vivo. There is one report that CHO cells cannot be infected by scrapie prions (Butler et al., 1988), but the cellular factors that influence prion replication are poorly understood and may be different from those that affect PrP Sc production from mutant molecules.
Although our study has focused on a mutant form of PrP, our results lead us to hypothesize that the PrP Sc produced during prion infection possesses an aberrant membrane association similar to that of PG11. This proposal is generally consistent with a large body of literature, some of it dating back more than 30 years, suggesting that infectious PrP Sc is tightly associated with cellular membranes, and is very hydrophobic (Hunter et al., 1964;Gibbons and Hunter, 1967;Prusiner et al., 1978). Of particular relevance, it has been reported that, although PrP Sc possesses a GPI anchor, it is not readily releasable from brain membranes by PIPLC (Stahl et al., 1990;Safar et al., 1991). Lack of release by PIPLC does not result from sequestration in a membrane-bound compartment, since PrP Sc is susceptible to biotinylation with membrane-impermeant reagents, and to cleavage by proteases; moreover, the GPI anchor of purified PrP Sc is PIPLC-sensitive, arguing that the anchor is not intrinsically uncleavable. These properties of PrP Sc from infected brain are strikingly reminiscent of those we have described for PG11 PrP in CHO cells. It is uncertain from the published studies whether the failure of PIPLC to release membranebound PrP Sc results from inaccessibility of the anchor structure, or rather, signifies a second, GPI-independent form of membrane attachment. Our results with PG11 PrP argue for the second possibility. It will be useful now to reexamine the membrane association of PrP Sc from scrapie-infected cells in culture, using experimental approaches similar to those we have described here.
Most of the known pathogenic mutations in human PrP are single amino acid substitutions in the C-terminal half of the molecule, in contrast to the 48 amino acid insertion in the N-terminal half that we have modeled here (Prusiner and Hsiao, 1994). We find that several moPrPs carrying point mutations homologous to those found in human PrP are not released from the surface of transfected CHO cells by PIPLC, implying an aberrant form of membrane association like that described for PG11. 5 These data support the idea that alterations in membrane attachment may be a general feature of both genetic and infectious forms of prion disease.