Most Pathogenic Mutations Do Not Alter the Membrane Topology of the Prion Protein*

The prion protein (PrP), a glycolipid-anchored membrane glycoprotein, contains a conserved hydrophobic sequence that can span the lipid bilayer in either direc-tion, resulting in two transmembrane forms designated Ntm PrP and Ctm PrP. Previous studies have shown that the proportion of Ctm PrP is increased by mutations in the membrane-spanning segment, and it has been hypothesized that Ctm PrP represents a key intermediate in the pathway of prion-induced neurodegeneration. To further test this idea, we have surveyed a number of mutations associated with familial prion diseases to determine whether they alter the proportions of Ntm PrP and Ctm PrP produced in vitro , in transfected cells, and in transgenic mice. For the in vitro experiments, PrP mRNA was translated in the presence of murine thymoma microsomes which, in contrast to the canine pancreatic microsomes used in previous studies, are capable of efficient glycolipidation. We confirmed that mutations within or near the transmembrane domain enhance the formation of Ctm PrP, and we demonstrate for the first time that this species contains a C-terminal glycolipid anchor, thus exhibiting an unusual, dual mode of membrane attachment. However, we find that pathogenic mutations in other regions of the molecule have no effect on the amounts of Ctm PrP and Ntm PrP, arguing against the proposition that transmembrane PrP plays an obligate role in the pathogenesis of prion diseases.

Prion diseases are neurodegenerative disorders characterized by spongiform destruction of brain tissue and the presence of cerebral amyloid plaques (1,2). These disorders include kuru and Creutzfeldt-Jakob disease in humans, "mad cow disease" in cattle, and scrapie in sheep. Prion diseases can have an infectious or genetic origin, or can arise spontaneously. The infectious agent is hypothesized to be PrP Sc , 1 a conformation-ally altered isoform of a normal cell-surface glycoprotein of unknown function called PrP C (3). PrP Sc from dietary or other infectious sources is thought to act as a catalyst or template to convert endogenous PrP C into more PrP Sc , which then accumulates, eventually causing disease. Familial forms of prion disease result from germline mutations in the PrP gene on chromosome 20, which are believed to favor conversion of the protein to the PrP Sc form (4). Sporadic cases may be due to rare, spontaneous conversion of wild-type PrP C to PrP Sc . No covalent modifications that distinguish PrP Sc from PrP C have been detected (5), but the conformations of the two isoforms are dramatically different, with PrP Sc having a much higher content of ␤-sheet (6,7). There are several biochemical properties that distinguish PrP Sc from PrP C , the most prominent being protease resistance. After treatment with proteinase K (PK), PrP Sc is cleaved near amino acid 90 to yield a protease-resistant core fragment known as PrP 27-30, while under the same conditions PrP C is completely degraded (8).
PrP Sc is found in the brain in most cases of infectious, familial, and sporadic prion disease (3). However, in some inherited prion diseases, for example cases of Gerstmann-Strä ussler syndrome due to an A117V mutation in PrP, PrP Sc has not been detected and brain material does not appear to be infectious when injected into rodents (9 -11). These exceptions raise the interesting possibility that PrP Sc , although it is the infectious form of the protein, may not be the proximate cause of neurodegeneration in at least some forms of prion disease.
Recently, it has been proposed that an alternate form of PrP that is distinct from PrP Sc may play an important role in prion pathogenesis. This form, designated Ctm PrP, has an unusual transmembrane topology (12). Most molecules of PrP C do not span the lipid bilayer and are attached to the cell surface exclusively by a glycosyl phosphatidylinositol (GPI) anchor appended to the C terminus of the polypeptide chain (13,14). In contrast, Ctm PrP is thought to span the membrane once, with its C terminus on the exofacial surface and a highly conserved, hydrophobic region in the center of the molecule (amino acids 111-134) serving as a transmembrane anchor. Another form of PrP, Ntm PrP, has also been described with the same transmembrane segment, but the reverse orientation (N terminus on the exofacial surface) (12). It has been proposed that the relative proportions of these three topological variants is influenced by as yet unidentified accessory proteins that interact with the translocation apparatus in the endoplasmic reticulum (ER) (15,16).
Recent studies have brought the potential biological relevance of transmembrane forms of PrP into sharper focus. These species were originally observed only after translation of PrP mRNA in vitro on rabbit reticulocyte or wheat germ ribosomes in the presence of canine pancreatic microsomes (17)(18)(19)(20)(21). In more recent investigations, however, Ctm PrP has been identified in brain membranes from transgenic mice that express PrP molecules carrying mutations within or near the transmembrane domain; in vitro translation experiments had indicated that these mutations increased the relative proportion of Ctm PrP (12,22). Transgenic mice expressing such Ctm PrP-favoring mutations at high levels develop a spontaneous neurodegenerative illness that bears some similarities to scrapie, but without the presence of PrP Sc (12,22). There is also indirect evidence that Ctm PrP accumulates in mice expressing wild-type PrP during the course of scrapie infection (22). Based on these results, it has been hypothesized that Ctm PrP represents a common intermediate in the pathogenesis of both infectious and genetic prion diseases (22). In this view, Ctm PrP is the ultimate cause of neurodegeneration, and PrP Sc acts indirectly by increasing the amount of Ctm PrP.
We have previously carried out extensive studies of the properties of mutant PrP molecules expressed in cultured cells and transgenic mice (23)(24)(25)(26)(27). The mutations analyzed in those investigations lie outside of the conserved hydrophobic region that serves as a transmembrane anchor in Ctm PrP and Ntm PrP. To test the hypothesis that transmembrane PrP is part of a general pathway for prion-related neurodegeneration, we undertook here to determine whether these mutations induce the formation of Ntm PrP and Ctm PrP in vitro, in cultured cells, and in transgenic mice. We confirm that mutations in the central, hydrophobic region enhance formation of transmembrane PrP, and we demonstrate for the first time that Ctm PrP contains a C-terminal GPI anchor as well as a transmembrane segment, thus exhibiting two modes of membrane attachment. However, we find that pathogenic mutations in other regions of the molecule have no effect on the amounts of Ctm PrP and Ntm PrP, arguing against the possibility that transmembrane PrP plays an obligate role in the pathogenesis of prion diseases.

EXPERIMENTAL PROCEDURES
Antisera-P45-66 antibody, raised against a synthetic peptide encompassing amino acids 45-66 of mouse PrP, has been described previously (14). Monoclonal antibody 3F4, which recognizes an epitope from hamster and human PrP encompassing residues 109 -112 (28), was a gift of Richard Kascsak (Institute for Basic Research, New York). R20 antibody, raised against a synthetic peptide comprising residues 218 -232 of mouse PrP (29), was a gift of Byron Caughey (Rocky Mountain Laboratories, Hamilton, MT). Anti-calnexin antibody was from Stressgen (Vancouver, British Columbia, Canada).
In Vitro Translation-Messenger RNAs were translated with [ 35 S]methionine in a final volume of 25 l containing 50% nucleasetreated, rabbit reticulocyte lysate (Promgea, Madison, WI) according to the manufacturer's instructions. Canine pancreatic microsomal membranes were purchased from Promega, or were prepared by the method of Walter and Blobel (30). Microsomes from BW5147.3 mouse thymoma cells were prepared as described (31), except that cells were lysed by Dounce homogenization and passage through a 27-gauge needle. Pancreatic and thymoma microsomes were used at 1.5 and 5 l per translation reaction, respectively, to equalize the amounts of ER marker proteins added. In some reactions, an oligosaccharide acceptor peptide (Bz-N-G-T-Ac; Bachem, Torrance, CA) was used at a final concentration of 0.5 mM to inhibit N-linked glycosylation. Translation reactions were incubated at 30°C for 60 min.
To detect protease-protected products, 5 l aliquots of translation reactions were incubated in a final volume of 50 l with 100 g/ml PK (Roche Molecular Biochemicals) in 50 mM Tris-HCl (pH 7.5) and 1 mM CaCl 2 for 60 min at 4°C, followed by addition of 5 mM phenylmethylsulfonyl fluoride to terminate digestion. Some digestion reactions also contained 0.5% Triton X-100 to solubilize membranes.
Samples were either analyzed directly by SDS-PAGE and autoradiography, or they were first subjected to immunoprecipitation or treatment with phosphatidylinositol-specific phospholipase C (PIPLC) as described below. Radioactive bands on gels were quantitated using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA).
Immunoprecipitation-Aliquots of the translation reactions were boiled in the presence of 1% SDS for 5 min to denature the proteins, and were then diluted with 10 volumes of radioimmune precipitation assay buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS) plus protease inhibitors (pepstatin A and leupeptin, 1 g/ml; phenylmethylsulfonyl fluoride, 5 mM). One l of the appropriate anti-PrP antibody was added, and samples incubated on ice for 60 min. Protein A-Sepharose beads were added, and samples were rotated at 4°C for 30 min. Beads were collected by low speed centrifugation and washed three times with radioimmune precipitation assay buffer, after which proteins were eluted with 1% SDS in 50 mM Tris-HCl (pH 7.5) and subjected to deglycosylation with peptide:Nglycosidase F (New England Biolabs, Beverly, MA) according to the manufacturer's directions. Following methanol precipitation, proteins were analyzed by SDS-PAGE and autoradiography.
PIPLC Treatment-Five l aliquots of translation reactions were diluted to 200 l with 1% Triton X-114 (precondensed as described (Ref. 32)) in phosphate-buffered saline (PBS). After incubation at 4°C for 20 min, samples were subjected to phase partitioning by incubation at 37°C for 10 min, followed by centrifugation at 16,000 ϫ g for 10 min to separate the phases. The aqueous phase was removed, and the detergent phase diluted to 200 l with PBS containing protease inhibitors and 100 g/ml bovine serum albumin. One unit of PIPLC from Bacillus thuringiensis (prepared as described in Ref. 33) was added to one half of the diluted detergent phase, and both halves were incubated at 4°C for 2 h. The phase separation was repeated, and proteins in the second set of aqueous and detergent phases either were precipitated with methanol and analyzed by SDS-PAGE or were subjected to immunoprecipitation with 3F4 antibody.
Transfected Cells-BHK cells were maintained in ␣-minimal essential medium supplemented with 10% fetal calf serum, nonessential amino acids, and penicillin/streptomycin. CHO cells were grown in ␣-minimal essential medium supplemented with 7.5% fetal calf serum and penicillin/streptomycin. Transfections were performed with Lipo-fectAMINE (Life Technologies, Inc.) according to the manufacturer's instructions. Cells were harvested 24 h after transfection by brief trypsinization or by mechanical detachment, rinsed twice with PBS, and resuspended in 250 l of 0.25 M sucrose, 10 mM HEPES (pH 7.4), 1 g/ml pepstatin A, and 1 g/ml leupeptin. After 5 min on ice, cells were lysed by 10 passages through silastic tubing (0.3 mm, inner diameter) connecting two syringes with 27-gauge needles (34). A post-nuclear supernatant was prepared by centrifugation at 5,000 ϫ g for 2 min. PK protection assays were performed by incubating post-nuclear supernatants in 50 mM Tris-HCl (pH 7.5), 250 g/ml PK, and in some cases 0.5% Triton X-100. After 60 min at 22°C, digestion was terminated by addition of 5 mM phenylmethylsulfonyl fluoride. Samples were deglycosylated with peptide:N-glycosidase F prior to analysis by Western blotting.
Brain Membranes-Fresh brain tissue (ϳ0.5 g wet weight) was homogenized using a Dounce apparatus in 5 ml of Buffer B (0.25 M sucrose, 10 mM HEPES (pH 7.5), 100 mM KOAc, 1 mM Mg(OAc) 2 , 1 g/ml pepstatin, 1 g/ml leupeptin), followed by passage (five times each) through 16-, 18-, 21-, 23-, and 27-gauge needles. The homogenate was centrifuged first at 12,000 ϫ g for 10 min in a microcentrifuge, and then at 541,000 ϫ g for 20 min in a Beckman TLA100.3 rotor, and the microsomal pellet was resuspended in 250 l of Buffer B. PK protection assays were carried out as described above for transfected cells, except that PK was used at 50 g/ml.

Microsomes from Canine Pancreas and Murine Thymoma
Cells Differ in Their Efficiency of GPI Anchoring-PrP is attached to the cell surface via a C-terminal GPI anchor (13,14). In carrying out studies of the membrane topology of PrP after in vitro translation, it would thus be desirable to utilize microsomal membranes that are capable of attaching the GPI anchor to newly synthesized polypeptides. Previous investigations have employed microsomes derived from canine pancreas, which do not effectively carry out GPI anchor addition (35). We therefore tested microsomes derived from BW5147.3 murine thymoma cells, which have been reported to be much more efficient in GPI anchoring (31,36).
We translated mouse PrP mRNA in vitro using rabbit reticulocyte lysate in the presence of pancreatic or thymoma microsomes, and then incubated Triton X-114 lysates of the membranes with or without PIPLC, a bacterial phospholipase that cleaves the GPI anchor. By partitioning the lysate into detergent and aqueous phases, we could then score the amount of PrP that had been rendered hydrophilic by removal of the anchor. As shown in Fig. 1, all translation reactions produced two groups of products of 32 and 25 kDa, representing, respectively, core-glycosylated and unglycosylated forms of PrP. The latter correspond to molecules that had not been translocated into the lumen of the microsomes, since they are susceptible to digestion with PK (see below). Less than 10% of the PrP chains translated in the presence of canine pancreatic microsomes shifted into the aqueous phase after PIPLC treatment, indicating that very few molecules carried a GPI anchor (lanes 1-4). In contrast, about 50 -60% of the glycosylated protein produced in the presence of thymoma microsomes shifted into the aqueous phase, indicating relatively efficient GPI anchoring (lanes 5-8). Untranslocated PrP chains, although they lack a GPI anchor, are retained in the detergent phase, presumably because they contain both N-and C-terminal signal peptides that are hydrophobic. Of note, PrP molecules shifted into the aqueous phase migrated with a slightly lower mobility than those that remained in the detergent phase (compare lanes 7 and 8), a phenomenon that is typical for polypeptides that have lost their GPI anchor (37). We believe that the actual efficiency of anchor addition by thymoma microsomes may be even higher than 60%, and that the persistence of some glycosylated PrP in the detergent phase after PIPLC treatment is likely to reflect inefficient PIPLC digestion and phase partitioning. In support of this idea, when samples were denatured in SDS prior to PIPLC treatment, to improve accessibility of proteins to the phospholipase, essentially all of the glycosylated PrP was converted to the more slowly migrating band (not shown). In contrast, PIPLC treatment after SDS denaturation did not alter the gel mobility of PrP synthesized with pancreatic microsomes.
Transmembrane Forms of PrP Are Produced in Both Microsome Preparations, and Their Proportion Is Increased by Mutations in the Hydrophobic Region-To detect transmembrane forms of PrP, translation products were subjected to digestion by PK, which cleaves off portions of the polypeptide chain residing on the external side of the microsomal membrane and leaves transmembrane and lumenal domains intact. Samples were then immunoprecipitated with anti-PrP monoclonal antibody 3F4 to eliminate globin contamination contributed by the reticulocyte lysate, which obscures visualization of low molecular weight PrP fragments. Proteins were also subjected to enzymatic deglycosylation to reveal differences in the sizes of PrP species independent of glycosylation state. After PK digestion of translation reactions containing either canine pancreatic or thymoma microsomes, wild-type PrP was resolved into three forms of 25, 19, and 15 kDa (Fig. 2, A and B, lane 2). The 25-kDa form is the same size as PrP before protease treatment (lane 1) and therefore represents molecules that are completely protected from digestion, and which therefore have been fully translocated into the microsome lumen; for consistency with an earlier report (12), we refer to this form as Sec PrP. The two smaller fragments are derived from transmembrane species whose cytoplasmic domains have been digested by the protease, and whose lumenal and transmembrane domains have been protected by the microsome. No PrP is detected after PK treatment in the presence of a detergent that disrupts the microsomal membrane (lane 3), confirming that the 19-and 15-kDa fragments do not represent intrinsically protease-resistant portions of the molecule. As will be shown below, the 19-kDa fragment derives from Ctm PrP, a species whose C terminus resides in the ER lumen, and the 15-kDa fragment from Ntm PrP, a species whose N terminus is lumenal.
Consistent with previous studies (12,22), we find that two mutations in the central hydrophobic region of the PrP mole- cule increase the relative amount of Ctm PrP with little change in the amount of Ntm PrP (Fig. 2, A and B, lanes 4 -9; Table I). A116V is the mouse homologue of a mutation in human PrP (A117V) that is associated with Gerstmann-Strä ussler syndrome (38,39); 3AV is a triple alanine 3 valine mutation at residues 112, 114, and 117 that is not found in human PrP but that produces neurological disease in transgenic mice (12). The effect of the A116V mutation was weaker than that of the 3AV mutation.
We found that, for wild-type PrP, as well as for the two mutant PrPs, both Ntm PrP and Ctm PrP were produced in considerably smaller amounts in the presence of thymoma microsomes than in the presence of pancreatic microsomes (Fig. 2, compare A and B; Table I). This difference is not attributable to an inhibitory effect of GPI addition on the generation of transmembrane PrP in thymoma microsomes, since we have found that elimination of the C-terminal GPI addition signal does not alter the amounts of Ntm PrP and Ctm PrP produced in this system (data not shown). Presumably, there are other cell typeor species-related differences between the two microsome preparations that affect PrP topology.
Determination of the Topology of Ntm PrP and Ctm PrP-Two kinds of experiments were carried out. First, we investigated whether or not Ntm PrP and Ctm PrP were glycosylated by carrying out translations of 3AV PrP in the presence and absence of a peptide inhibitor of N-linked glycosylation (Bz-N-G-T-Ac). We observed that the protease-protected fragment of Ctm PrP shifted from 24 to 19 kDa upon inclusion of the inhibitor, while the migration of the Ntm PrP fragment was unchanged (data not shown). This result indicates that Ctm PrP but not Ntm PrP is glycosylated. Since the two consensus sites for N-glycosylation lie at positions 180 and 196, and the presumed transmembrane segment at positions 111-134, the data indicate that the C terminus of Ctm PrP and the N terminus of Ntm PrP lie in the lumen of the microsome.
In a second experiment, we carried out epitope mapping to define the topology of Ntm PrP and Ctm PrP (Fig. 3). Translations of both wild-type and 3AV PrP were carried out in the presence of thymoma microsomes, and protease-protected products were immunoprecipitated with three different antibodies: P45-66, which recognizes the octapeptide repeats in the N terminus of PrP (14); 3F4, which reacts with an epitope encompassing residues 108 -111 (28) (this epitope, which is not normally present in mouse PrP, was introduced into all of our PrP constructs by changing residues 108 and 111 to methionine); and R20, which recognizes residues 218 -231 (29). We found that the Ntm PrP fragment was recognized by P45-66 and 3F4 but not by R20, indicating that the N terminus of Ntm PrP through residue 111 is protected by the microsome. In contrast, the Ctm PrP fragment was recognized by 3F4 and R20 but not by P45-66, indicating that the C terminus of Ctm PrP starting from residue 108 is protected. As expected, Sec PrP, which represents the full-length form of the protein, immunoprecipitates with all three antibodies. These mapping results, in conjunction with the observed sizes of the protected fragments, indicate that Ntm PrP and Ctm PrP span the membrane once, but in opposite directions, via the central hydrophobic segment (Fig. 3B). Similar results were obtained when pancreatic rather than thymoma microsomes were used. translation Messenger RNAs encoding wild-type (WT), 3AV, or A116V PrP were translated in the presence of microsomes from canine pancreas or murine thymoma cells. Samples were subjected to PK digestion, and PrP was immunoprecipitated, deglycosylated, and analyzed by SDS-PAGE as in Fig. 2. The percentage of each topological form was determined by phosphorImager analysis of SDS-PAGE gels. Each entry represents the mean of n experiments, with the S.D. given in parentheses below each entry. Values were corrected for the differences in methionine content among the three protease-protected products.  Ctm PrP Has a GPI Anchor-Since the C terminus of Ctm PrP is lumenal, this form of the protein has the potential to be GPI-anchored. To test this possibility, we carried out translations of wild-type and 3AV PrP in the presence of thymoma cell microsomes, which attach GPI anchors efficiently. As in Fig. 1, we utilized Triton X-114 phase partitioning after PIPLC treatment to assess the presence of the anchor. Translation reactions were treated with protease prior to detergent lysis to reveal the protected fragment derived from Ctm PrP, which migrates at 24 kDa because it has not been deglycosylated in this experiment. We found that this fragment, as well as fully protected Sec PrP (32 kDa), were shifted into the aqueous phase after PIPLC treatment, consistent with the presence of a GPI anchor on both species (Fig. 4, lanes 4 and 8). The change in partitioning of the 24-kDa fragment was especially easy to appreciate for 3AV PrP (lane 8), which produces larger amounts of Ctm PrP than wild-type PrP. Consistent with loss of the GPI anchor after phospholipase treatment, both Sec PrP and Ctm PrP in the aqueous phase displayed a slightly reduced gel mobility compared with the same species remaining in the detergent phase (compare lanes 7 and 8). About 40 -50% of the Ctm PrP fragment was shifted into the aqueous phase by PIPLC, about the same proportion seen for Sec PrP. As discussed above, we believe that the presence of residual PrP in the detergent phase after phospholipase treatment reflects inefficient PIPLC digestion and phase partitioning. These results indicate that Ctm PrP has a dual mechanism of membrane attachment, including both a transmembrane segment and a GPI anchor. From Fig. 4 it would appear that the transmembrane segment is not sufficiently hydrophobic to prevent segregation of the Ctm PrP fragment into the aqueous phase once the GPI anchor is removed by PIPLC.
Mutations Outside of the Central, Hydrophobic Domain Do Not Alter the Amount of Transmembrane PrP-It might be predicted that mutations such as 3AV and A116V would affect the proportions of Ctm PrP or Ntm PrP by altering the capacity of the central, hydrophobic region to serve as a transmembrane anchor. We wondered whether the same is true of mutations that lie outside of the hydrophobic region. We therefore carried out in vitro translations in the presence of thymoma microsomes of a number of PrP mutants whose human homologues are associated with familial prion diseases. The mutations analyzed are located both N-and C-terminal to the central, hydrophobic segment. As positive controls, we analyzed PrP containing the 3AV and A116V mutations, as well as another artificial mutation in the central region (K109I/H110I, referred to as KH-II) that has been reported to increase the amount of transmembrane PrP (12,22). We found that the proportion of Ctm PrP was not increased over wild-type levels by any of the mutations outside of the central, hydrophobic domain (Fig. 5, A  and B). As expected, the 3AV, A116V, and KH-II mutations resulted in significantly increased levels of Ctm PrP.
Since the amount of Ctm PrP associated with wild-type PrP in thymoma microsomes is small (ϳ10%) and close to the limits of detectability, we carried out similar experiments on mutant PrPs translated with canine microsomes, which generate larger amounts of Ctm PrP. In this system as well, the amount of Ctm PrP was not significantly altered by pathogenic mutations outside of the central region (Fig. 5B).
When samples were subjected to immunoprecipitation so that Ntm PrP could be visualized, the amount of this species produced in the presence of either thymoma or canine microsomes was not changed by any of the mutations (data not shown).
Little Ctm PrP Can Be Detected in Transfected Cells-To test whether transmembrane PrP can be detected in cultured cells, we carried out protease protection experiments on membranes in post-nuclear supernatants derived from transiently transfected BHK cells. As shown in Fig. 6, Ctm PrP was not detectable in cells expressing wild-type PrP, although small amounts of a 19-kDa band that is likely to represent the deglycosylated, protease-protected fragment of Ctm PrP were present in cells expressing 3AV or A116V PrPs. The identity of this band was confirmed by its reactivity with 3F4 and R20, but not with P45-66 (data not shown). Cells expressing PrP molecules with mutations outside of the hydrophobic region did not produce detectable amounts of Ctm PrP, indicating that either these mutations do not increase the amount of transmembrane PrP, or if they do, their effect is less pronounced than that of 3AV and A116V. No Ntm PrP was detected in any of the transfected cells. The integrity of the membrane vesicles in these experiments was confirmed by the production of a 70-kDa, proteaseprotected fragment of calnexin, which represents the transmembrane and lumenal domains of the protein (40). Results similar to those shown in Fig. 6 were also obtained in transiently transfected CHO cells (data not shown). We conclude from these experiments that amount of transmembrane PrP produced in cultured cells is considerably lower than after in vitro translation, with Ctm PrP representing Ͻ2% of the total PrP even for transmembrane-favoring mutations.
Ctm PrP Is Not Detectable in the Brains of Transgenic Mice Expressing a Mutant PrP-To determine whether transmembrane forms of PrP could be detected in brain tissue, we also carried out protease protection experiments on microsomal membranes prepared from Tg(WT) transgenic mice that express wild-type PrP, or from Tg(PG14) mice that express a mutant PrP carrying a nine-octapeptide insertion. Tg(PG14) mice spontaneously develop a neurological illness characterized by ataxia, granule cell apoptosis, synaptic PrP deposition, and accumulation of a weakly protease-resistant form of PrP (26,27). As was the case in cultured cells, neither Ctm PrP nor Ntm PrP were observable in the brains of mice expressing either wild-type PrP or PrP carrying the nine-octapeptide insertion (Fig. 7). Again, the integrity of the microsomes was confirmed by the presence of the 70-kDa protease-protected fragment of calnexin. Thus, in contrast to mutations in the hydrophobic region, which have been shown to produce detectable amounts of Ctm PrP in the brains of transgenic mice (12,22), a pathogenic mutation in the octapeptide repeat region does not increase the amount of transmembrane PrP above the level of detectability.
3AV PrP, but Not PG14 PrP, Can Be Released from the Cell   FIG. 4. Ctm PrP has a GPI anchor. Messenger RNA encoding WT or 3AV PrP was translated in reticulocyte lysate supplemented with murine thymoma microsomes. After treatment with PK, translation reactions were solubilized in 1% Triton X-114 at 4°C and were then phaseseparated by warming to 37°C. One half of the detergent phase was incubated with PIPLC at 4°C for 2 h (lanes 3, 4, 7, and 8), and the other half was incubated in the absence of PIPLC (lanes 1, 2, 5, and 6). The phase separation was repeated, and PrP in the detergent phase (D lanes) and aqueous phase (A lanes) was immunoprecipitated with 3F4 antibody and analyzed by SDS-PAGE and autoradiography. The positions of protected products derived from Sec PrP and Ctm PrP are indicated by the arrowheads to the right of the gel. The Ntm PrP product is present in only small amounts and is not seen on this gel.
Surface by PIPLC-Since Ctm PrP is attached to the lipid bilayer by both a transmembrane and a GPI anchor, it would not be released from the membrane after PIPLC cleavage of the anchor. To see if the increased amount of Ctm PrP associated with the 3AV mutation caused a reduction in the proportion of the mutant protein that could be released from the cell surface by PIPLC, we labeled transfected CHO cells with a membraneimpermeant biotinylation reagent (Fig. 8). We found that Ͼ90% of the surface biotinylated 3AV PrP was released by the phospholipase (lanes 3 and 4), a proportion similar to that for wild-type PrP (lanes 1 and 2). Thus, the 3AV mutation has no observable effect on the PIPLC releasability of surface PrP. In contrast, virtually no biotinylated PrP carrying the PG14 mutation could be released by PIPLC (lanes 5 and 6), even though this mutation does not affect the amount of Ctm PrP (Fig. 5). Similar results were obtained with transfected BHK cells (data not shown). DISCUSSION We show that PrP chains translated in vitro or expressed in cultured cells can adopt one of three possible topologies (Fig. 9). Molecules with the Sec PrP orientation have been fully translocated into the ER lumen and conjugated to a GPI anchor that serves as their sole means of attachment to the lipid bilayer. This is the topology that is presumably displayed by the majority of molecules present on the surface of cells. In contrast, Ntm PrP and Ctm PrP span the lipid bilayer once via the central hydrophobic region (residues 111-134), with either the N terminus or C terminus, respectively, in the ER lumen. Ctm PrP also acquires a C-terminal GPI anchor, and thus displays an unusual, dual mode of membrane attachment. We find that several mutations within or near the central, hydrophobic region increase the amount of Ctm PrP, but pathogenic mutations outside of this region have no effect. The results presented here confirm and significantly extend previous reports on the membrane topology of PrP (12,22).
All previous studies of PrP membrane topology by in vitro translation utilized canine pancreatic microsomes (12,(17)(18)(19)(20)(21)(22), which we show here do not efficiently attach GPI anchors to newly synthesized polypeptide chains. GPI anchor addition occurs in the ER soon after completion of the polypeptide chain, and involves a transamidation reaction in which a hydrophobic, C-terminal sequence is cleaved and replaced by the pre-formed anchor structure (41). Presumably, the transamidase or other enzymes involved in GPI synthesis or addition are relatively inactive in microsomes derived from canine pancreas, or else endogenous phospholipases are more active. In our experiments, we have employed microsomes prepared from BW5147.3 mouse thymoma cells, which attach GPI anchors to the majority of PrP chains. Using this system, we find that both Sec PrP as well as Ctm PrP become GPI-anchored. The demonstration of a GPI anchor on Ctm PrP is a novel finding and B). The Ntm PrP product is not detectable because its migration is obscured by globin. The human homologues of the mutations are associated with the following familial prion diseases: Creutzfeldt-Jakob disease (PG11, PG14, V179I, E199K, V209I), Gerstmann-Strä ussler syndrome (A116V, F197S), and fatal familial insomnia (D177N). B, translations were carried out with either thymoma or pancreatic microsomes as in panel A, and the Ctm PrP and Sec PrP fragments produced after PK digestion were quantitated by PhosphorImager analysis of SDS-PAGE gels. The percentage of Ctm PrP was expressed as Ctm PrP/ ( Ctm PrP ϩ Sec PrP) ϫ 100. Ntm PrP was not included in this calculation because it cannot be visualized without first immunoprecipitating the samples to eliminate contaminating globin, which interferes with the migration of low M r species. The horizontal dashed lines indicate the amount of Ctm PrP observed for wild-type PrP. Each bar represents the mean Ϯ S.E. of 2-15 replicates. Values that are significantly different (p Ͻ 0.05) from wild-type PrP are indicated by an asterisk. indicates that this species is attached to the membrane by both a bilayer-spanning segment as well as a C-terminal GPI structure that is covalently linked to the outer leaflet of the bilayer. The existence of such a dual mechanism of membrane attachment is unusual, although several other proteins have been proposed to adopt such a topology (42)(43)(44).
We and others have previously observed that PrP molecules carrying pathogenic mutations are resistant to release from the cell surface by PIPLC (14,23,(45)(46)(47). The presence of a dual membrane anchor on the mutant proteins does not explain this phenomenon, since most mutations do not increase the amount of Ctm PrP. In fact, PrP molecules carrying the 3AV mutation, which enhances formation of Ctm PrP, are fully releasable by PIPLC, suggesting that Sec PrP is the major species on the cell surface. Presumably, this is because the amount of Ctm PrP that is present in cells is small in comparison to the amount of Sec PrP, or because Ctm PrP does not reach the cell surface. Our previous studies indicate that the most likely explanation for the PIPLC resistance of mutant PrPs is that the GPI anchors of these molecules are physically inaccessible to the phospholipase, possibly as a result of the conformational change that accompanies conversion to the PrP Sc state (45).
It had been previously reported that several mutations in or around the central, hydrophobic region increase the amount of Ctm PrP produced in translation reactions in vitro or in brain tissue (12,22), but the effect of mutations outside of this area had not been examined. The previously studied mutations included one (A116V) whose human homologue (A117V) is associated with Gerstmann-Strä ussler syndrome (38,39), and three (N107I, K109I/H110I, and 3AV) that are artificial and have not been described in humans. The data reported here confirm that these mutations increase the amount of Ctm PrP. A plausible explanation for this effect is that the amino acid changes alter the hydrophobicity, ␣-helical structure, or some other physical property of the central region in such a way as to make it a more suitable transmembrane segment. Two other artificial mutations in the central region of the molecule have been found to have the opposite effect, eliminating formation of Ntm PrP and Ctm PrP and causing exclusive production of Sec PrP (12,22). One of these (G122P) is likely to disrupt the formation of a transmembrane ␣-helix, and the other (⌬103-111) eliminates a segment of hydrophilic amino acids (designated "stop transfer effector") that strongly influences the topology of the adjacent FIG. 6. Only small amounts of Ctm PrP are produced in transfected cells. BHK cells were transiently transfected with plasmids encoding WT or mutant PrPs. Post-nuclear supernatants prepared from cells 24 h after transfection were incubated with (ϩ PK lanes) or without (Ϫ PK lanes) proteinase K in the presence (ϩ Det lanes) or absence (Ϫ Det lanes) of Triton X-100. Proteins were then solubilized in SDS, enzymatically deglycosylated, and subjected to Western blotting with 3F4 antibody. The Sec PrP band is indicated by an open arrowhead, and the Ctm PrP band (seen for 3AV and A116V) by the filled arrowhead. The bands indicated by the bracket for PG14 PrP represent nonspecific cleavage products produced by cellular proteases. In each experiment, samples were also Western-blotted with an antibody to calnexin to confirm the integrity of the ER membranes. A representative calnexin blot from an experiment on cells expressing wild-type PrP is shown in the lower right panel. Full-length calnexin migrates at 90 kDa (Ϫ PK/Ϫ Det lane). The 70-kDa band (ϩ PK/Ϫ Det lane) represents the transmembrane and lumenal domains of calnexin, which are protected from protease digestion. The 65-kDa band (ϩ PK/ϩ Det lane) is an intrinsically protease-resistant fragment of calnexin (40).  1 and 4) PK in the presence (lanes 3 and 6) or absence (lanes 1, 2, 4, and 5) of Triton X-100 (Det). Proteins were then solubilized in SDS, enzymatically deglycosylated, and subjected to Western blotting with 3F4 antibody. The Sec PrP band is indicated by an open arrowhead; no Ctm PrP fragment (19 kDa) is visible in lanes 2 and 5. The bands marked by a single asterisk represent PrP that has not been completely deglycosylated. Samples were also Westernblotted with an antibody to calnexin to confirm the integrity of the ER membranes. Microsomes from the Tg(PG14) mouse contain variable amounts of the 70-kDa protected fragment of calnexin even in the absence of PK treatment (lane 4, double asterisk), presumably due to endogenous protease activity. The Tg(PG14) mouse used for this experiment was terminally ill.
FIG. 8. 3AV PrP, but not PG14 PrP, can be released from the cell surface by PIPLC. CHO cells were transiently transfected with plasmids encoding WT, 3AV, or PG14 PrP. Twenty-four hrs after transfection, cells were surface-biotinylated and then treated with PIPLC for 2 h at 4°C. PrP was immunoprecipitated from PIPLC incubation media (M lanes) and cell lysates (C lanes) after enzymatic deglycosylation and was subjected to Western blotting with 3F4 antibody. Blots were developed with horseradish peroxidase-streptavidin to identify biotinylated PrP molecules.
transmembrane segment (19,20). Taken together, the available data indicate that the central region of the PrP molecule acts as a crucial determinant of membrane topology, presumably in conjunction with the N-terminal signal sequence, but that other sections of the molecule, including those where many pathogenic mutations lie, are topologically inert. Exactly how the topology-determining domains of PrP function at the level of the ER translocon, and how the final topology of the protein is regulated will require further investigation (15,16).
Our observation that most pathogenic mutations do not affect Ctm PrP production has important implications for the role of this transmembrane species in the pathogenesis of prion diseases. Hegde et al. (22) have hypothesized that the generation of Ctm PrP represents a final common pathway for neurodegeneration in both transmissible and familial forms of prion disease. This idea is based upon their observations that transgenic mice expressing Ctm PrP-favoring mutations at high levels develop a spontaneous neurodegenerative illness in the absence of PrP Sc , and that the amount of Ctm PrP increases during the course of scrapie infection of mice expressing wild-type PrP. However, our observation that seven different pathogenic mutations fail to alter the amount of Ctm PrP makes it unlikely that this species is an obligatory intermediate in all cases of prionassociated neurodegeneration. It has been speculated that some mutations like E200K indirectly enhance formation of Ctm PrP by first causing accumulation of PrP Sc (22). If this were the case, it might be argued that these mutations would not have an observable effect on Ctm PrP levels in our experiments if PrP Sc were not being produced. However, five of the pathogenic PrPs (PG11, PG14, D177N, F197S, and E199K) do in fact assume a protease-resistant form that resembles PrP Sc when expressed in transfected CHO and BHK cells (2, 23), 2 while three of mutants (A116V, V179I, and V209I) are proteasesensitive and display other properties characteristic of PrP C . 3 This comparison makes it clear that there is no correlation between the efficiency with which a mutation promotes conversion to a PrP Sc -like state in cell culture and its potency in inducing Ctm PrP. A similar pattern is seen in vivo, since the human homologues of some of the mutations we have examined are associated with infectious PrP Sc in the brains of patients while others are not (4). These results argue against the proposition that PrP Sc causes disease by enhancing formation of Ctm PrP. In cases where neither PrP Sc nor Ctm PrP are present, it seems likely that another form of PrP is the primary pathogenic entity.
Whether Ctm PrP is an irrelevant byproduct of PrP biogenesis, or whether it plays some more restricted role in the disease process remains to be determined. One possibility is that Ctm PrP is responsible for a subset of inherited prion diseases due to mutations within or near the transmembrane segment. Alternatively, when Ctm PrP is present, it may accelerate or enhance neurodegeneration triggered by PrP Sc . To evaluate these and other hypotheses, it will be crucial to understand more about the cell biology of transmembrane forms of PrP. We have observed that the amount of Ctm PrP produced in transfected cells is considerably less than after in vitro translation. This may reflect differences in the efficiency of transmembrane PrP synthesis between the in vitro and in vivo situations, or, alternatively, it may be due to more rapid degradation of transmembrane PrP in cells. It is also possible that neuronal cells are more efficient at synthesizing Ctm PrP than the BHK and CHO cells we have used here. It will be important now to analyze how transmembrane forms of PrP are generated during translation, where they are localized at the subcellular level, and how they are metabolized, to shed further light on the role of these unique species in the biology of prion diseases.