Glycosylation Deficiency at Either One of the Two Glycan Attachment Sites of Cellular Prion Protein Preserves Susceptibility to Bovine Spongiform Encephalopathy and Scrapie Infections*

The conversion into abnormally folded prion protein (PrP) plays a key role in prion diseases. PrPC carries two N-linked glycan chains at amino acid residues 180 and 196 (mouse). Previous in vitro data indicated that the conversion process may not require glycosylation of PrP. However, it is conceivable that these glycans function as intermolecular binding sites during the de novo infection of cells on susceptible organisms and/or play a role for the interaction of both PrP isoforms. Such receptor-like properties could contribute to the formation of specific prion strains. However, in earlier studies, mutations at the glycosylation sites of PrP led to intracellular trafficking abnormalities, which made it impossible to generate PrP glycosylation-deficient mice that were susceptible to bovine spongiform encephalopathy (BSE) or scrapie. We have now tested more than 25 different mutations at both consensus sites and found one nonglycosylated (T182N/T198A) and two monoglycosylated (T182N and T198A) mutants that rather retained authentic cellular trafficking properties. In vitro all three mutants were converted into PrPres. PrP mutant T182N/T198A also provoked a strong dominant-negative inhibition on the endogenous wild type PrP conversion reaction. By using the two monoglycosylated mutants, we generated transgenic mice overexpressing PrPC in their brains at levels of 2–4 times that of nontransgenic mice. Most interestingly, such mice proved readily susceptible to a challenge with either scrapie (Chandler and Me7) or with BSE. Incubation times were comparable or in some instances even significantly shorter than those of nontransgenic mice. These data indicate that diglycosylation of PrPC is not mandatory for prion infection in vivo.

The conversion into abnormally folded prion protein (PrP) plays a key role in prion diseases. PrP C carries two N-linked glycan chains at amino acid residues 180 and 196 (mouse). Previous in vitro data indicated that the conversion process may not require glycosylation of PrP. However, it is conceivable that these glycans function as intermolecular binding sites during the de novo infection of cells on susceptible organisms and/or play a role for the interaction of both PrP isoforms. Such receptor-like properties could contribute to the formation of specific prion strains. However, in earlier studies, mutations at the glycosylation sites of PrP led to intracellular trafficking abnormalities, which made it impossible to generate PrP glycosylation-deficient mice that were susceptible to bovine spongiform encephalopathy (BSE) or scrapie. We have now tested more than 25 different mutations at both consensus sites and found one nonglycosylated (T182N/T198A) and two monoglycosylated (T182N and T198A) mutants that rather retained authentic cellular trafficking properties. In vitro all three mutants were converted into PrP res . PrP mutant T182N/T198A also provoked a strong dominant-negative inhibition on the endogenous wild type PrP conversion reaction. By using the two monoglycosylated mutants, we generated transgenic mice overexpressing PrP C in their brains at levels of 2-4 times that of nontransgenic mice. Most interestingly, such mice proved readily susceptible to a challenge with either scrapie (Chandler and Me7) or with BSE. Incubation times were comparable or in some instances even significantly shorter than those of nontransgenic mice. These data indicate that diglycosylation of PrP C is not mandatory for prion infection in vivo.
Prion diseases include Creutzfeldt-Jacob disease, Gerstmann-Strä ussler-Scheinker syndrome, Kuru, and fatal familial insomnia in humans, scrapie in sheep and goats, bovine spongiform encephalopathy (BSE) 1 in cattle, transmissible mink encephalopathy in mink, feline spongiform encephalopathy in cats, and chronic wasting disease in cervids. Prion diseases are characterized by incubation times of up to several years or decades. The post-translational conversion of cellular proteasesensitive prion protein (PrP C syn. PrP sen ) into its proteaseresistant isoform (PrP res ) is the key feature in these infectious diseases (1). Cellular and pathological PrP are encoded by the same gene (Prnp) (2)(3)(4)(5), have the same covalent structure, and are both glycoproteins with a glycophosphatidylinositol anchor (6). The cellular isoform is bound to the external surface of the plasma membrane, whereas PrP res accumulates in late endosomes and lysosomes (7,8). Cellular PrP is rich in ␣-helices and is proteinase K-sensitive and soluble in detergents (9). In contrast, abnormal PrP res is rich in ␤-sheets (10,11), partially proteinase K-resistant, and is insoluble. Mouse PrP contains two consensus sites for N-linked glycosylation (12,13). The first is located at codon 180 and the second at codon 196. In the fully matured protein both N-glycosylation sites are occupied. Analysis of hamster-or mouse-derived PrP res revealed that the glycan chains represent a mixture of bi-, tri-, and tetra-antennary complex-type sugars with 70% of the galactose residues sialyzed (12,14,15).
A variety of experimental scrapie strains has been isolated and characterized in rodents, which show marked differences in the clinical phenotype, pathology, and in the biochemical features of PrP res . The existence of different strains still raises the question of how such a diversity can be encoded by a single protein composed of a defined amino acid sequence. Variations in the protein conformation and in the glycosylation of PrP have been postulated as modulators of the conversion and/or disease process and of prion strain properties (16 -19).
Prion strains can be distinguished (20) by PrP res glycosylation patterns (21). Although some scrapie strains such as Chandler produce predominantly heavily diglycosylated PrP res , in others (e.g. Me7) monoglycosylated PrP res is favored (22). Comparison of the glycoforms of normal and abnormal hamster PrP showed that both contain generally the same N-linked oligo-saccharides but at different ratios (23). Moreover, it has recently been shown that the PrP C glycosylation can significantly affect the PrP res formation, if both are derived from different species (24).
To elucidate the role of the glycans in prion infection and pathogenesis, we sought to develop transgenic mice that express mono-or nonglycosylated PrP C with retained authentic PrP trafficking. It is generally accepted that the surface expression of PrP C is necessary for a de novo prion infection of cells (25)(26)(27)(28)(29). Earlier work (30,31) showed that elimination of the first consensus site for N-glycosylation by substituting threonine by alanine at codon 182 leads to a PrP C retention in the cell with no expression on the plasma membrane. Transgenic mice carrying this mutation were resistant to a scrapie infection and showed only low expression levels of PrP C (19). On the other hand, the deletion of the second glycosylation site did not change the trafficking of the prion protein nor the fundamental susceptibility. However, expression levels that were achieved did not suffice to effectively support the infection cycle (19). Another approach was based on the direct mutation of the attachment sites (N180Q/N196Q) (32) in vitro. However, viable transgenic mice based on this concept have not been reported to date.
To reveal PrP glycosylation-deficient mutants with retained authentic subcellular trafficking, we systematically studied more than 25 different mutations at the consensus sites. Among those we found only three mutants lacking either one or both glycosylation sites and with no detectable change in their expression topology in neuroblastoma, Chinese hamster ovary, and murine fibroblast cells. In scrapie-infected neuroblastoma cells, even mutants with hampered or blocked trafficking turned out to be convertible into their proteinase K-resistant isoform, but only mutants highly expressed on the surface promoted a de novo infection of cells. In addition, although being convertible itself, the unglycosylated mutant T182N/ T198A showed a remarkable dominant-negative inhibition of the endogenous murine PrP conversion process in ScN2a cells. Similar effects have been reported in the literature and were explained by conversion inefficiencies of PrP C across species barriers (amino acids 112-187) (33) and/or an interference with the "protein X"-binding site, a proposed chaperone-like factor facilitating murine PrP conversion (amino acids 167, 171, 214, and 218) (34 -37). Finally, we produced transgenic mice overexpressing these glycosylation-deficient PrP mutants and that were proven to be susceptible to challenge with BSE and scrapie prion strains.
Cloning of an N-terminal Deletion Mutant-This construct, missing the N-terminal signal sequence (amino acids , was produced by SOE-PCR (splicing by overlap extension) using the following primers: N-sig.seq, 5Ј-gaA TCA GTC ATC ATG AAG CGG CCA AAG CCT GGA GGG-3Ј; CMV.seq, 5Ј-cat atg cca agt acg ccc cct att gac g-3Ј; BsteII.rev, 5Ј-CAT CGG TCT CGG TGA AGT TCT CCC CCT TGG-3Ј; N-sig.rev, 5Ј-CGC TTC ATG ATG ACT GAT tct aga tct gc-3Ј. Three consecutive PCRs were performed as follows. The first used primers CMV.seq and N-sig.rev, and the second used primers N-sig.seq and BsteII.rev. A 3F4-tagged mouse ORF served as a template. Finally, the obtained PCR products were used as templates for the last PCR with primers CMV.seq and BsteII.seq. The fidelity was confirmed by sequencing.
Cell Lines-Murine neuroblastoma (N2a) and Chinese hamster ovary (CHO) cells were purchased from ATCC (Manassas, VA). Scrapieinfected neuroblastoma cells as well as murine fibroblast cell lines Pa 317 (41) and ⌿2 cells (42) (which were derived from NIH3T3 cells) (43) were generously supplied by B. Chesebro and S. Priola (Hamilton, MT). N2a cells were grown in MEM-NEAA, CHO cells in a 3:1 mixture Dulbecco's MEM and F-12, and murine fibroblast cells in RPMI 1640 and ScN2a cells in Opti-MEM or MEM-NEAA. Media were enriched with 10% FCS and contained penicillin/streptomycin. All cell culture reagents were purchased from Invitrogen. Cells were grown in an atmosphere of 5% CO 2 .
Transient or Stable Expression of Mutant PrP in Cell Cultures-N2a and CHO cells were transfected using LipofectAMINE (Invitrogen) and murine fibroblasts were transfected by CaPO 4 (Stratagene) according to the manufacturer's directions. For every construct at least two independent transfections were performed. The transfection efficiency was measured by FACS analysis using pEGFPn3 (Clontech, U57609). For N2a cells, transfection rates of about 70% were achieved. Antibiotic resistant clones of N2a and CHO cells were selected by using 500 g/ml Zeocin (Invitrogen), and when necessary the cells were subcloned and then maintained in 250 g/ml Zeocin. For every construct we performed at least two transfections. Calculations were made using the "Cellquest" program (BD Biosciences) on a Macintosh Quadra 650. Fluorescein (FITC)-and phycoerythrin (PE)-conjugated second antibodies were purchased from Jackson ImmunoResearch (West Grove, PA). To every FITC-labeled sample, 30 l of DNA-binding propidium iodide solution (1 mg/ml) (Sigma) was added prior to measurement in order to stain dead cells for an electronic exclusion. (The signal of FITC is measured at 530 nm, whereas propidium iodide has its emission at about 570 nm.) For every construct we performed at least two transient transfections. As a control, cells were FACS-analyzed by adding a conjugate only.
PrP Conversion in ScN2a Cells-PrP in vitro conversion assays were conducted by using a retroviral expression system for exogenous PrP (43). ScN2a cells (3 ϫ 10 5 ) were plated in medium and grown overnight. This medium was removed and replaced by another medium containing 4 g/ml Polybrene (Sigma). After 2 h a retroviral supernatant in different dilutions was added and incubated overnight. The cells were split on a T-25 flask (for PrP C detection), a T-75 flask (for PrP res detection), and a 24-well plate (for immunohistochemistry) and were grown to confluence and treated as follows.
For the detection of PrP C , cells from a confluent T25-flask were washed with PBS and resuspended with lysis buffer (0.32 M sucrose, 0.5% Nonidet P-40, 0.5% deoxycholate) to prepare a 10% cell homogenate, which could be then further analyzed by SDS-PAGE and immunoblotting.
For the detection of PrP res , cells from a confluent T-75 flask were harvested using 0.25% trypsin and pelleted at 800 rpm for 5 min. The cells were then resuspended in lysis buffer (10 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM EDTA, 0.5% Triton X-100, 0.5% deoxycholate), and debris was removed by mild centrifugation. The samples were digested with proteinase K (PK 25 g/ml, 37°C, 1 h), and PMSF was added to stop digestion followed by a final centrifugation at 265,000 ϫ g for 1 h at 4°C. After resuspension and heat denaturation in sample buffer (final concentration 4% SDS, 100 mM Tris-HCl, 175.2 mM sucrose, 1% ␤-mercaptoethanol, 0.02% bromphenol blue, pH 7.4), one-fourth from each sample was run on a 16% SDS-polyacrylamide gel followed by immunoblotting. Untransduced cells were loaded as a control, because initial experiments revealed no discrepancy between mock-and nontransduced cells (data not shown).
Intrinsic Proteinase K Resistance of PrP C -Appropriate cell lysate equivalents were treated with one of the following procedures (at 37°C): 25 g/ml PK for 1 h, 1 g/ml PK for 30 min, 0.5 g/ml PK for 30 min, or 3.3 g/ml PK for 10 min. The digestions were stopped by heat denaturation at 95°C for 10 min. Samples were then pretreated with guanidinium HCl (final concentration 3 M) for 10 min at room temperature, and proteins were acetone-precipitated (final concentration 50%). The resulting pellets were resuspended in sample buffer, electrophoresed (16% gels), and immunoblotted using mAb 3F4 (44).
Deglycosylation of PrP C -Where indicated, cell lysates were treated with 10 units/ml N-glycosidase F or 0.26 -0.36 units/ml endoglycosidase H (both reagents were from Roche Applied Science) for 16 h at 37°C before the samples were electrophoresed followed by immunoblotting. To avoid incomplete digestion, higher endoglycosidase H concentrations were used as described in earlier protocols (31).
Biotinylation of Cell Surface Proteins-Cell membrane proteins were biotinylated as described (45) using Sulfobiotin-X-NHS (Calbiochem), which is not able to enter intact cells and was immunoprecipitated with the mAb 3F4 as described (46). Trypan blue staining (Serva, Heidelberg, Germany) was carried out to exclude that sulfobiotin-X-NHS was labeling intracellular proteins because of cytoplasmic leakages. As additional control, we performed an immunoprecipitation using mAb ␤-actin for detection of intracellular actin. N2a cells were biotinylated, and the reaction was stopped or not stopped and afterward immunoprecipitated by ␤-actin. Only data from experiments were used in which no signal was detectable in the stopped sample.
PI-PLC Release-N2a cells (80% confluent) from a T-75 flask were washed three times with MEM-NEAA and incubated with PI-PLC (0.1 units/ml in MEM-NEAA) (Roche Applied Science) at 37°C for 4 h. Following this treatment the supernatant was separated from floating cells by centrifugation (800 rpm, 5 min), and the proteins were then methanol-precipitated. The resulting pellet was resuspended in electrophoresis sample buffer. Cells were harvested and lysed (as described above for the detection of PrP C ). The cell lysate and PI-PLC released proteins were analyzed on 13% SDS-PAGE followed by immunoblotting. Identically treated tissue cultures without the addition of PI-PLC served as controls.
Solubility Assay-Cells from a T-75 flask were dissolved in 3 ml of solubility buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM EDTA, 0.5% Triton X-100, 0.5% deoxycholate), and debris was removed by mild centrifugation. PMSF (0.81 mM) was added, and the samples were pelleted by ultracentrifugation at 265,000 ϫ g, 4°C for 1 h. The pellets were resuspended in sample buffer, and supernatants were precipitated as described above.
Overlay Experiments-Brain tissue from mice infected with scrapie strains Chandler, Me7, and 79A were obtained from terminal stage C57Bl and CD1 mice. The brains were homogenized in Opti-MEM (10% w/v), subjected to ultrasonification, and heated for 10 min at 70°C. ScN2A cells were cloned twice and selected for PrP res negative cells. Such cells were then transfected with 1 ml of retroviral supernatant as described above and grown in T-75 flasks to a confluence of 70 -80%. Either 200 l of Me7, 400 l of 79A, or 400 l of Chandler homogenates (10%), which contained comparable PrP res amounts as determined by immunoblotting, were used for infection. The cells were washed three times with PBS and overlaid with 5 ml of Opti-MEM plus 10% FCS containing 400 or 200 l of brain homogenate. The cells were incubated for 4 h, and afterward 10 ml of medium was added per flask. The cells were grown for 4 days to confluence and treated as described for PrP res . For each construct every infection experiment was performed at least twice. Cells overlaid with 400 l of brain homogenate of a non-infected C57Bl mouse served as a control.
Generation and Characterization of Transgenic Mice-Suitable ORF mutations encoding non-or monoglycosylated PrP C were cloned into the vector phgPrP that comprised the murine prion gene but lacked intron 2 (39). Fertilized oocytes were taken from PrP 0/0 mice (40), microinjected with DNA of the relevant phgPrP mutant according to standard procedures, and reimplanted into NMRI mice. For the detection of transgenic mice, we applied the protocol of Bueler et al. (40). Vector phgPrP and PrP 0/0 mice were generously supplied by C. Weissman, London, UK.
Founder animals were mated with PrP 0/0 mice and the offspring examined for PrP C expression. 10% homogenates were prepared in lysis buffer by using an Ultra-Turrax (Janke and Kunkel, Germany) and were analyzed on SDS-PAGE followed by immunoblotting. The protein expression levels in the brains were determined by using a photo imager with the software "Tina" (Raytest, Straubenhard, Germany) and were compared with the normal expression level of nontransgenic mice. The Prnp copy number was estimated accordingly by comparing PCR amplifications to corresponding standard samples.
BSE and Scrapie Challenge of Transgenic Mice-Hemizygous transgenic mice were inoculated intracerebrally with 30 l of C57Bl mouse brain homogenates (in saline solution) containing 10% BSE or scrapie (Me7 and Chandler). Negative and positive controls included transgenic and conventional C57Bl mice inoculated either with non-infectious or with BSE or scrapie mouse brain homogenates. These mice were checked three times per week until the onset of clear clinical symptoms and were culled. Their brains were later sampled and analyzed by immunoblotting or immunohistochemistry for demonstrating the accumulation of PrP Sc .

Deletion of the First Glycosylation Site by Codon 182
Mutation-In earlier studies the deletion of the first glycosylation consensus site led to aberrant trafficking of PrP C in cells (19,30,31). However, these studies left the question unanswered as to whether these differences were because of the introduced point mutation at position 182 or because of the missing glycosylation of PrP itself. Because wild type PrP C synthesized in the presence of tunicamycin was readily present on the cell surface (31), the first explanation seemed to be more likely. Our initial experiments therefore intended to generate a mutant of the first glycosylation consensus site of PrP C exhibiting unaltered intracellular trafficking.
For this purpose we substituted the codon for threonine at position 182 with codons for any other amino acid. We expressed these 19 mutants as well as a wild type PrP control in N2a cells, and we analyzed the effects by immunoblotting to reveal expression rates and glycosylation patterns (Fig. 1) and by FACS analysis to test for the presence of proteins on the cell surface (Fig. 2). All mutants were based on a template that encoded a 3F4 epitope-tagged mouse PrP. This allowed the detection of mutated PrP against the background of wild type mouse PrP. Mutations at the consensus site were introduced by using a set of wobble primers and PCR mutagenesis. All constructs were cloned into the expression plasmid pcDNA3.1Zeo(ϩ) and transfected into neuroblastoma (N2a) cells, and stably expressing cells were selected by Zeocin. For FACS analysis, we used living, non-permeabilized cells so that the antibodies only detected surface-expressed PrP C (see "Experimental Procedures"). Of the 19 different mutants at codon 182, we found only two that were still accessible extracellularly, T182S and T182N. However, only mutant T182N was monoglycosylated (Figs. 1 and 4a) because serine instead of threonine can also function as a glycosylation consensus site (Fig. 1). This mutant therefore served as a supplementary positive control in addition to mAb 3F4 epitope-tagged wild type PrP (data not shown).
Generation of Nonglycosylated, Surface-expressed PrP C -To generate a nonglycosylated PrP C mutant with retained intracellular trafficking, the T182N encoding substitution was combined with a T198A mutation by which the second N-linked glycosylation consensus site was also eliminated (choosing the same approach as for the first consensus site would have introduced an asparagine at position 198 which in combination with a threonine at position 200 would have generated a new consensus site ( 196 NFNET 200 )). This new construct T182N/ T198A was then transfected into N2a cells, and the subcloned cells were analyzed by immunoblot and FACS analysis. It was shown that this mutant PrP C was completely unglycosylated (Fig. 4b) while still being detectable on the cell surface (Fig. 2). This result demonstrated that the glycosylation is not mandatory for surface presentation of PrP C .

Comparative Analysis of Novel Mutants with the Earlier Described Glycosylation Site Mutants in Different Cell
Lines-To compare better the mutants described here (T182N

FIG. 1. Expression of PrPT182X mutants in neuroblastoma cells.
Nineteen different T182X mutants of PrP C were expressed in N2a cells using the pcDNA3.1zeo expression system. After Zeocin treatment stably mutant PrP-overexpressing cells were selected. Cell lysates were analyzed by immunoblot using mAb 3F4 as a detection antibody and visualized using a chemiluminescence system. Note that all mutations except for one abolish PrP glycosylation at the first consensus site. Only the threonine to serine mutation (different subclones are named f5, c4, and d3) still functions as a recognition site. Untransfected N2a cells (N2a), transfections using wild type 3F4 epitope-tagged PrP encoding constructs, and hamster PrP (brain lysate) served as controls. 0, 1, and ] ϭ non-, mono-, and diglycosylated PrPC. M ϭ molecular mass marker.

FIG. 2. Cell surface expression of
PrPC glycosylation mutants. T182N, T182A, and T198A single and double mutants of 3F4 epitope-tagged murine PrPs were expressed in N2a cells (pcDNA3.1zeo expression system). Stably transfected cells were selected using Zeocin, and mutant PrPCs overexpressing N2a cell clones were harvested after subcloning. FACS analysis was carried out on non-permeabilized cells using mAb 3F4 as a primary antibody and phycoerythrincoupled secondary antibodies. Dead cells were excluded by propidium iodide staining. Each analysis was performed with the addition of secondary antibodies (dark line) and without (light line) in order to discriminate between specific and nonspecific intrinsic fluorescence. By using this extremely sensitive technique, all single and double mutants were detected on the cell surface. Comparing the extent of shifting revealed differences in the levels of surface expression. A ⌬N signal sequence deletion mutant, which is incapable of reaching the cell membrane, was used as a negative control, whereas wild type 3F4-tagged PrP served as a positive control. and T182N/T198A) with those reported earlier in literature (T182A, T198A, and T182A/T198A) (19,30,31), expression plasmids were also cloned for T182A, T198A, and T182A/ T198A. Moreover, three different cell lines, N2a, CHO, and murine fibroblast cells, were transfected with each of these five mutants to determine specific effects on the glycosylation and trafficking. As it turned out, the choice of the cell line used for these studies had in general no impact on the results that were obtained (Table I).
Most surprisingly, as shown by FACS analysis and biotinylation studies, not only the newly generated mutants but also the earlier mutants were expressed on the cell surface, although at quite different levels. PrP mutants T182N, T198A, and T182N/T198A trafficked clearly onto the cell surface of subcloned N2a, CHO, and murine fibroblast cells, whereas only lower surface presentation levels were observed for PrP mutants T182A and T182A/T198A. Subcloned cells expressing 3F4-tagged PrP C lacking an N-terminal sequence were used as a additional negative controls in FACS analysis. As the Nterminal signal sequence is mandatory for the protein translocation into the endoplasmic reticulum, this mutant remained in the cytosol (Fig. 2). FACS results were reconfirmed by biotinylation of cell surface proteins followed by PrP C immunoprecipitation using mAb 3F4, immunoblotting, and detection of biotinylated proteins using streptavidin conjugated to peroxidase. Whereas mutants T182N, T198A, and T182N/T198A were comparable with those of 3F4 epitope-tagged wild type PrP, surface expression rates of mutants T182A and T182A/ T198A were considerably lower (Fig. 3). To exclude false-positive results because of dead and therefore permeable cells, we used DNA-binding propidium iodide (FACS analysis) or trypan blue staining (biotinylation assay) as markers for dead cells (see "Experimental Procedures"). When we tested the panel of glycosylation mutants in the different cell lines, no significantly diverging results were found in respect to surface expression and immunoblot profiles.
To determine whether these constructs were attached to the cytoplasmic membrane by a glycophosphatidylinositol anchor, transfected cells were treated with PI-PLC, which cleaves the phosphatidylinositol-diacylglycerol linkage specifically. With one exception, all cell surface-exposed constructs were released by PI-PLC. Only mutant T198A was resistant to PI-PLC-treatment, which was surprising as the double mutant T182N/ T198A was releasable (Table I). Therefore it must be assumed that the monoglycosylated protein is either anchored differently or, more likely, the cleavage site is sterically inaccessible to the enzyme.
To confirm that sugar residues were linked to asparagine, glycoproteins were treated with enzyme N-glycosidase F, which specifically cleaves off N-linked carbohydrate moieties. Independent from the cell line used, all three monoglycosylated mutants showed a shift in their molecular mass to that of nonglycosylated PrP C (Fig. 4, a and b).
To determine processing deficits, monoglycosylated PrP C mutants were also analyzed by endoglycosidase H treatment. Carbohydrate moieties N-linked to PrP C are excessively trimmed and processed during their passage through the Golgi apparatus to complex-type oligosaccharides. Mannosidase II in the mid-Golgi compartment renders them resistant to endoglycosidase H cleavage. Therefore, no shift or only a partial shift of di-or monoglycosylated bands to the low molecular mass range of nonglycosylated PrP revealed disturbance in mid-Golgi processing. The following results were obtained: T182A showed low partial resistance, T182N high partial resistance, and T198A total resistance to endoglycosidase H (Fig. 4c). These results indicate that presumably only a few T182A mutants passed through the mid-Golgi apparatus, whereas the processing of T182N PrP C was only scarce and that of T198A PrP C was not disturbed at all.
Proteinase K Sensitivity and Solubility of Glycosylation Mutants-To exclude an intrinsic proteinase K resistance in the mono-or nonglycosylated PrP C molecules, mutant PrP expressed in N2a, CHO, or murine fibroblast cells was subjected  Fig. 2 for T182A/T198A. b ϩ(ϩ) indicates a signal as shown in Fig.3b for T182A/T198N. c ND indicates not done.

FIG. 3. Left and right, cell surface biotinylation of glycosylation-deficient PrPC mutants.
Mutants were stably expressed in N2a cells, and the outer membrane proteins were biotinylated using sulfobiotin-X-NHS which is not able to enter intact cells. The absence of ␤-actin biotinylation and trypan blue-permeable cells (Ͻ1%) was also monitored. Following cell lysis, PrP C was immunoprecipitated with mAb 3F4 and separated on 13% (left) or 16% (right) SDS-polyacrylamide gels and blotted onto Immobilon P membrane. Biotinylated PrP was detected by a streptavidin-horseradish peroxidase/chemiluminescence visualization system. Note that mutants T182N, T198A, and T182N/T198A were detected at abundant amounts on the cytoplasmic membrane, whereas mutants T182A and T182A/T198A were detected at much lower levels. Due to experimental variations between gel runs, the quantitation of the results shown on left and right is hardly possible. M ϭ molecular mass marker.
for different incubation times to digestions with various concentrations of proteinase K. In general, no particular partial or complete PK resistance was detected (data not shown).
It has been reported (31) that all glycosylation consensus site mutants based on the T/A substitution show detergent insolubility. To reveal whether this characteristic applies also to the newly generated non-and monoglycosylated mutants, a solubility assay was carried out. In accordance with earlier work, these new glycosylation mutants could also be detected after ultracentrifugation in the pellet as well as in the supernatant fractions, whereas wt3F4PrP could be detected in the supernatant only (Fig. 5). Therefore, it must be assumed that the depletion of the glycan side chains leads generally to an increased tendency of the non-and monoglycosylated PrP mutants to aggregate (49,50). In addition, we observed when developing blots with mAb SAF 70, which detects murine and 3F4-tagged PrP, that fully glycosylated murine PrP, which is also expressed in N2a cells, is not involved in this aggregation process (data not shown). Therefore, an interaction with resident cell PrP C could be excluded.
Conversion of PrP C Glycosylation Mutants into Their Proteinase K-resistant Isoforms-After the characterization of mutant PrP C , we wanted to know if these mutants were convertible into their proteinase K-resistant isoforms. Therefore, we performed conversion experiments in Chandler-infected ScN2a cells using a retroviral system for expression of mutant PrP C . For this purpose replication-defective murine retroviruses encoding mutant PrP molecules were cloned and generated using vector pSFF (38,43,52) and packaging cell lines (41,42). These retroviruses were then transduced into ScN2a cells. The advantage of this system is the high transfection efficiency (as revealed by immunocytochemistry), which leads to a reliable overexpression of exogenous mutant PrP in each cell (53,54). Cell lysates intended for PrP res detection were digested with proteinase K (25 g/ml, 1 h, 37°C), ultracentrifuged, and separated by SDS-PAGE. Those intended for PrP C detection were analyzed without pretreatment. The expression and conversion of mutated exogenous PrP in ScN2a cells were selectively detected by immunoblotting by using mAb 3F4, because mAb SAF70 or polyclonal antibodies Ra5/7 stain total (endogenous wild type plus exogenous) murine PrP.
All mutants were highly expressed in ScN2a cells (Fig. 6c), and no influence on the endogenous mouse PrP C production was detectable (Fig. 6d). Moreover, almost all mutants were efficiently converted into their proteinase K-resistant isoform except for mutant T182N/T198A which was only detected in small amounts (Fig. 6a). When looking at the total PrP res levels in the transduced ScN2a cells, with the exception of T182N/T198N,   FIG. 4. Processing and maturation of PrP C mutants. a and b, PNGaseF mutants of PrP C were expressed in N2a cells using the pcDNA3.1zeo expression system. After Zeocin treatment and subcloning, stably mutant PrP-overexpressing cells were selected. Cell lysates were analyzed by immunoblot using mAb 3F4 as detection antibody and a chemiluminescence visualization system. Asparagine linkage of PrP glycosylation was confirmed by enzymatic treatment of cell lysates using PNGaseF (ϩ). Transfections using wild type 3F4 epitope-tagged PrP-encoding constructs as well as hamster PrP (brain lysate) served as controls. c, Endo H. Stable mutant PrP-expressing fibroblast cells were lysed, and the lysates were treated with endoglycosidase H. Lysates were then immunoblotted, and PrP C signals were visualized by using mAb 3F4. Mid-Golgi matured PrP C is resistant to deglycosylation, and therefore PrP banding signals do not shift toward the non-glycosylated form. M ϭ molecular mass marker. 0, 1, and 2 ϭ non-, mono-, and diglycosylated PrP C , respectively.
FIG. 5. Solubility of PrP glycosylation mutants expressed in N2a cells. 3F4-tagged wild type PrP and corresponding mutants were expressed in N2a cells. Confluent cell layers were harvested and lysed using detergents, and the lysate was ultracentrifuged. Concentrated supernatant proteins (a) and the pellet fraction (b) were immunoblotted using mAb 3F4 to reveal possible insolubilities of PrP mutants. Note that wild type PrP C is only detectable in the supernatant fraction and hence mostly or rather fully soluble, whereas all glycosylation-deficient mutants were found in both fractions. It must therefore be concluded that the attached glycans contribute to the solubility of PrP C . M ϭ molecular mass marker.
FIG. 6. Conversion of PrP glycosylation mutants in ScN2a cells. 3F4-tagged PrP mutants were expressed in ScN2a cells by using recombinant replication-deficient retroviruses (pSFF system). For demonstrating PRP C or PrP res , cell lysates either pretreated without (c and d) or with proteinase K (25 g/ml, 37°C, 1 h followed by ultracentrifugation) (a and b) were separated by SDS-PAGE. Exogenous PrP was detected by probing the membrane with mAb 3F4 (a and c). For the detection of total (exogenous and endogenous) PrP, polyclonal antibodies Ra 5/7 were used. Nontransduced ScN2a-cells were used for controls. Every PRP C -mutant was expressed (c) without any inhibitory effect on the synthesis of endogenous PRP C (d). Note also that every mutant was efficiently converted into its PK-resistant isoform, except mutant T182N/T198A which was converted only in minimal amounts (a). The total PrP res accumulation in the infected neuroblastoma cells was not influenced by Wt3F4, T182A, T182N, T198A, and T182A/ T198A (b), again with exception of mutant T182N/T198A which depressed the accumulation of total PrP res dramatically (b). 0, 1, and 2 ϭ non-, mono-, and diglycosylated PrP C . M ϭ molecular mass marker. none of the glycosylation mutants lead to a stronger reduction of PrP res than the wild type 3F4PrP that was used as a control (Fig.  6b). However, in the case of an abundant PrP res expression of mutant T182N/T198A, the accumulation of endogenous murine PrP res was dramatically suppressed (Fig. 6b). To characterize further this phenomenon, ScN2a cells were transduced with different dilutions of retrovirus-containing supernatants. This experiment revealed that the unglycosylated mutant T182N/T198A shows a dominant-negative inhibition, whereas wt3F4-mouse PrP and mutant T182A/T198A does not. In other words, conversion was only detectable when this construct was expressed at low levels (Fig. 7, a and b). In the presence of high concentrations of mutated proteins, no conversion was found, and the detectable amounts of endogenous mouse-PrP were dramatically reduced. This effect has never been described before in context with mutations in this region of the protein. (Such an effect has been described for mutants carrying the 3F4 epitope alone (55,56), but under the experimental procedures presented here we saw no impact on wt3F4. During the course of the experiment, we mentioned that this effect provoked by the 3F4 epitope alone is dependent from the PrP res load of ScN2a cells. By applying ScN2a cells harboring much PrP res , no effect on wt3F4 could be seen, but in ScN2a cells expressing less PrP res , we saw this effect (data not shown).
The absence of the disulfide bridge can also hinder the conversion process of PrP C (57). Therefore, we analyzed whether mutant T182N/T198A, where the mutation is located in an area close to the disulfide bridge, still possessed such an in-tramolecular bond (Fig. 8). As such a linkage stabilizes the orientation of helices II and III, proteins where the disulfide bridge is disrupted are more extended and therefore migrate slower in SDS-PAGE. As a result, a slight bandshift can be observed when native proteins with an intact disulfide bridge are compared with dithiothreitol-denatured disulfide bridge proteins (Fig. 8). Our results for PrP C mutant T182N/T198A as well as for all other mutants expressed in N2a and murine fibroblasts clearly indicate the presence of disulfide bonds in these molecules (data not shown).
De Novo Conversion into PrP res Applying Overlay Experiments-Having clarified the convertibility of every mutant, we then wanted to simulate a natural scrapie infection, where the infectious agent first has to enter the cell. Therefore, transduced neuroblastoma cells expressing murine (wt3F4) PrP or T182N, T182A, T198A, T182A/T198A, and T182N/T198A PrP mutants were challenged with a scrapie mouse brain homogenate (Me7). Cell lysates were subjected to PK digestion, ultracentrifugation, and immunoblotting using either mAb 3F4 for the detection of de novo converted mono-and nonglycosylated mutants or antibody Ra5/7 for the detection of total PrP res . It   FIG. 7. Dominant-negative inhibition of mutant T182N/T198A tested in permanent scrapies-infected neuroblastoma cells. ScN2a cells were transduced using serial dilutions of recombinant virus containing supernatant (from 1 ϭ pure to 1:16 dilution). All lysates were treated with proteinase K. For wild type 3F4 and mutant T182A/T198A, there was no difference in the PrP res accumulation as a consequence of the resulting lower exogenous PRP C expression (a and c) and no effect on the endogenous PrP res deposition in the ScN2a cells (b and d). However, PrP res derived from mutant T182N/ T198A was detected only at lower exogenous PRP C expression rates (i.e. higher supernatant dilution rates). Hence the conversion of exogenous PrP was accompanied by an inhibitory effect on the endogenous PrP res accumulation All results shown were from one experiment, and the blot membranes were exposed together on one film. 0, 1, and 2 ϭ non-, mono-, and diglycosylated PrP C . M ϭ molecular mass marker.
FIG. 8. Verification of disulfide bridge. The existence of disulfide bonds was verified by mAb 3F4 immunoblotting of cell lysate proteins from mutant PrP-expressing N2a cells that had been electrophoresed under reducing (R) or nonreducing (N) conditions using dithiothreitol as a reducing agent. As can be seen, reduced PrP C lacking the disulfide bond and therefore being bulkier in shape is running slower in the acrylamide gel (12%) than oxidized PrP C . This slight bandshift between the nonreduced (N) an reduced (R) sample indicates the presence of a disulfide bridge. M ϭ molecular mass marker.

FIG. 9. De novo conversion into PrP res applying overlay experiments.
Neuroblastoma cells were first transduced with recombinant replication-deficient retrovirus encoding wild type and glycosylationdeficient PrP mutants and subsequently overlaid with a 10% mouse brain homogenate of a terminally diseased mouse infected with scrapie strain Me7 (m). Overlaying the cells with brain homogenate of a noninfected C57Bl mouse served as control (c). The amount of PrP res per flask is shown on the 3rd and 4th lanes of blot a (using Ra 5/7). Confluent cells were rinsed extensively and lysed, and the lysates were treated with proteinase K. Exogenous PrP res was revealed by immunoblotting using mAb 3F4 (b). For every construct the results of two independent experiments are shown. As can be seen, wild type 3F4 tagged PrP C , and mutants T182N, T198A, and T82N/T198A PrP were converted into PrP res , whereas mutants T182A and T182A/T198A PrP failed to be converted. Arrows indicate the position of non-and monoglycosylated PrP, respectively. M ϭ molecular mass marker. was shown that 3F4-tagged murine PrP C as well as mutants that showed wild type-like surface expression in tissue culture (T182N, T198A, and T182N/T198A) were effectively converted into PrP res (Fig. 9). It was therefore demonstrated that for a de novo conversion of mutated proteins in cells, the target protein (PrP C ) must be present in sufficient amounts on the cell surface.
Generation and BSE and Scrapie Challenge of PrP Glycosylation-deficient Transgenic Mice-PrP mutants with wild typelike characteristics (T182N, T198A, and T182N/T198A) were used for the generation of transgenic mice. For this reason, PrP 0/0 mice were microinjected with the corresponding constructs. The genome of the offspring was screened by PCR for the presence of the transgene and the transgene copy number (Table II). The protein expression was analyzed in a qualitative and quantitative manner by immunoblotting (Fig. 10). These transgenic mice overexpressed prion protein in their brains at levels of up to four times that of wild type mice. We then challenged these transgenic mice with mouse BSE or scrapie (Me7 and Chandler) strains. As expected, transgenic mice carrying mutations at either one of the glycosylation sites developed scrapie (Table III) and accumulated PrP Sc in their brains as detected by immunoblotting (Fig. 11) or immunohistochemistry. However, following BSE and Chandler scrapie challenge, the incubation times were substantially shorter in the first consensus site mutant (i.e. 160 -240 days) than in mice lacking the second glycosylation site (i.e. 305-410 days). This was true for both of the two challenged transgenic mouse lines of each construct. Most interestingly, compared with the other two strains, Me7 displayed a substantially prolonged incubation time in both transgenic lines carrying the mutation at the first glycosylation consensus site, whereas no such effect was observed with the mutation at the second site. Moreover, based on clinical observations, transgenic mice lacking the first or second PrP C glycosylation site developed distinct disease patterns, even when inoculated with the same infectious strain. Although the first site mutant mice died within hours after onset of clinical disease showing severe excitation and kyphosis, second site mutant mice became apathic and died eventually after a prolonged clinical phase. Unfortunately, no transmission results are as yet available for transgenic mice lacking both glycosylation sites (T182N/T198A).

Diglycosylation of PrP c Is Not Mandatory for the Prion Infection Cycle in Vitro and in
Vivo-In the experiments described here, we have studied the role of PrPC glycosylation in the molecular pathogenesis of prion infections. By using a systematic approach, we revealed that the PrP mutation T182N, which eliminates the first glycosylation consensus site, does not interfere with the transport of PrPC to the cell membrane. Mono-and even nonglycosylated (by additional T198A ablation of the second consensus site) PrPCs were expressed on the cell surface and were convertible in scrapie-infected neuroblastoma cells into PrPres. Moreover, the expression of these mono-and nonglycosylated PrPC mutants in neuroblastoma cells chal-lenged by scrapie strain (Me7) lead to a de novo accumulation of mono-and nonglycosylated PrPres. This is in contrast to cloned glycosylation-deficient mutants that have been described in literature (T182A and T182A/198A). They were not convertible in our overlay experiments. This in vitro result goes along with earlier published in vivo results (18); transgenic mice expressing mutant T182A or T182A/T198A were not susceptible to scrapie infection. (Therefore, these results give rise to the speculation that not only our monoglycosylated mice will be infectious.) However, these earlier described constructs did prove to be convertible in neuroblastoma cells already infected with scrapie. In these studies, the deletion of the first glycosylation consensus site by the single point mutation threonine/ alanine at codon 182 leads to aberrant trafficking and synthesis of the prion protein (19,30,31). Our data now support evidence reported previously that sufficient PrP C expression on the cytoplasmic membrane is a prerequisite for the susceptibility of cells (25,26,28,35). This is supposed to be the reason why transgenic mice carrying a PrP T182A mutation are completely or partially scrapies-resistant (19). Based on these findings we generated glycosylation-deficient transgenic mice which proved to be BSE0 and scrapies-susceptible. Our data underline that neither the PrP glycosylation at the first nor at the second site are essential for the trafficking process and full infection cycle of prions in vivo. The results of the ongoing infection experiments are needed to learn more about the situation in which neither of the glycosylation sites are occupied.
Synthesis and Processing of Consensus Site Mutated PrP C -A number of earlier studies have addressed questions on the synthesis, processing, cellular trafficking, and the metabolism of PrP C glycosylation mutants. Variable results regarding the transport of mono-and nonglycosylated PrP C have been reported in the literature (30 -32, 58). Whereas Rogers et al. (30) and Korth et al. (32) working with T182A mutations revealed no surface expression in transiently transfected CV1 and N2a cells, Lehmann et al. (31) observed this mutant on the cell surface of stable-transfected, subcloned CHO cells.
These discrepancies are presumably due to the different techniques that were used. During our experiments we found little surface expression in transfected Zeocin-treated, un- cloned cell populations, whereas surface expression of the same mutant was easily detectable in CHO and N2a subcloned cells and also in uncloned murine fibroblast cells. One explanation could be that due to the high overexpression rate of the subcloned and selected cell lines that were used, their ability to control protein sorting was exhausted, which resulted in an accumulation of glycosylation-deficient PrP C which would normally be eliminated by the proteasome. The second explanation could be the high sensitivity of our assay that was achieved by using phycoerythrin (PE)-labeled conjugates as a secondary antibody. This antibody label is 20 times more sensitive than FITC, which was used in the references cited above. For the sake of comparison we also performed FACS analysis by using FITC conjugates, but we did not detect any surface expression of the T182N constructs. Moreover, in order to exclude falsepositive results, an N-signal sequence deletion mutant (3F4tagged) PrP was used as a control. To reveal possible cell line-specific effects, we tested three different cell lines (N2a, CHO, and murine fibroblast cells) in parallel but found no substantial differences.
Many glycoproteins require an intact N-glycosylation for cell surface expression (59 -62). In accordance with Korth et al. (32), we could now show that PrP C is different in this respect and is rather similar to the canine histamine H2 receptor that retains its authentic trafficking after deletion of the N-glycosylation (63). Therefore, it seems that the peptide moiety plays an important role for the glycopeptide export and for directing it to a specific transport pathway (64). We believe that the distinguished surface expression of our T182N mutant is because of its retained stability as compared with the T182A mutant. In the past, spectroscopic and thermodynamic stability experiments revealed that the mutant T182A displays an extreme tendency to aggregate, resulting in a significantly reduced intrinsic stability and an extraordinary reduced cooperativity (65). These properties were considered to result from a destabilization of PrP C due to the loss of the two buried hydrogen bonds in the threonine side chain, to residues in the hydrophobic core (Tyr-161 and Cys-178), and to the loss of van der Waal's interactions of the absent side chain atoms within the protein. Asparagine and threonine are both neutral amino acids with two free carbon atoms in their side chain that are capable of building hydrogen bonds. Therefore, it is reasonable that only the T182N mutation produced a stable monoglycosylated PrP C molecule. Although the same would apply for a glutamine exchange mutant at this position, it must be due to the additional carbon atoms in the side chain giving the protein a bulkier shape, which leads to the destabilization of the analogous glycosylation mutant. Other than mutant T182N PrP, mutant T182A PrP res and PrP C were lacking the nonglycosylated band, which also indicates a reduced stability. Petersen et al. (45) demonstrated that the loss of glycosylation can lead to a decreased half-life time of PrP C . We found that the monoglycosylated band of T182N was running slower and broader in SDS-PAGE, indicating a more heterogeneous glycosylation pattern, which might also contribute to a higher stability of the molecule. Following synthesis, molecule aggregates are usually retained in the mid-Golgi compartment and are then eliminated by the cellular control machinery: either directly in the mid-Golgi compartment or after retrograde transport and final decomposition by the  11. Infection of transgenic mice overexpressing glycosylation-deficient murine PrPC with murine BSE and scrapie strain Chandler. Hemizygous transgenic mice T182N expressing mutated PrP C and lacking the first glycosylation consensus site were inoculated intracerebrally with 10% brain homogenates of either BSE or scrapie strain Chandler diseased mice. Following onset of clinical disease, mice were sacrificed and the PrP Sc accumulation in their brains verified by immunoblotting using mAb 3F4 as detection antibody and chemiluminescence detection. Where indicated, samples were treated with proteinase K (50 g/ml at 55°C for 1 h). Corresponding samples of hamster scrapie 263K and molecular mass markers were run as controls. M ϭ molecular mass marker. ϩ, after PK digestion; Ϫ, without PK digestion. proteasome (66,67). Due to the overexpression of the T182N mutant, some molecules must have escaped the control system and were transported onto the cell surface.
We were able to show that the destabilization of the T182A mutant was not due to the loss of the disulfide bridge. Comparing all possible mutants at codon 182 through the application of the Swiss model program (www.expasy.ch) underlined the importance of hydrogen bonds in the molecule. Similar to threonine, alanine, asparagine, and glutamine can build hydrogen bonds to cysteine 178, but only threonine and asparagine were able to build hydrogen bonds to tyrosine 161. Therefore, we postulate that the intramolecular binding between tyro-sine161 (placed in ␤-sheet 2) and the ␣-helix 2 is a key position for the biological behavior of PrP C .
Electrospray mass spectrometry on murine PrP res revealed that oligosaccharides attached to Asn-180 are largely composed of neutral and poorly sialylated bi-and triantennary structures within a mass range of 1.660 -2.340 Da (14). In contrast, the more sialyzed and acidic glycan components at Asn-196 mainly range between 2.000 and 3.020 Da and consist of tri-and tetraantennary structures. Corresponding immunoblot patterns were revealed for the glycosylation mutants described here; PrP res and PrP C missing the first glycosylation site (T182A and T182N) produced a broader monoglycosylated band of higher molecular weight in comparison to the construct missing the second glycosylation site (T198A). The monoglycosylated band of mutant T182N was broader and had a higher molecular weight than mutant T182A. As this effect was independent from the cell line used, we propose that this is not due to a postulated different enzymatic set of the individual cell line but to the polypeptide chain itself indicating that the mutant T182N may be more stable and therefore able to bind more sugar groups.
In Vitro Convertibility of PrP C Glycosylation-deficient Mutants-The PrP conversion has been proposed to involve two steps: initial PrP C binding to PrP res followed by the folding reaction itself. Each of our glycosylation mutants was convertible into its PK-resistant isoform in scrapie-infected ScN2a cells. However, for the mutant T182N/T198A, this process was rather inefficient. Although this mutant was convertible, it was also capable of inhibiting the endogenous conversion process. This effect is described in the literature as dominant-negative inhibition (33,35,37,55,56,68). Considering the results for the other mutants, this effect cannot be explained by the loss of glycosylation. Furthermore, this effect was not due to the single point mutation, because the corresponding single mutants T182N and T198A reacted like Wt3F4. One explanation for the observed inhibition of T182N/T198A PrP may be aberrant binding. Theoretically, there are three different binding partners: exogenous PrP C , endogenous PrP C , and endogenous PrP res . Aberrant binding could either mean autoaggregation or the interaction of endogenous with exogenous PrP C . However, autoaggregation would not explain inhibitory effects on endogenous PrP res because endogenous PrP C was not found in aggregates of exogenous PrP C in the performed solubility assays; therefore the trans-dominant inhibition must occur in the folding reaction itself. The most likely explanation is that the double mutant binds more tightly to endogenous PrP res than endogenous PrP C but requires binding to more than one PrP res molecule for conversion. Therefore, excessive exogenous PrP C molecules in the cell would outcompete PrP res binding and eventually hinder the conversion of either endogenous or exogenous PrP C .
Another less likely explanation is that T182N/T198A PrP C binds to PrP res , and due to being a rigid and stable molecule it reconverted endogenous PrP res to PrP* and PrP C . Another possible but also unlikely explanation is that T182N/T198A-PrP C is processed differently as wild type-PrP C , caused by a diverging translocation at the endoplasmic reticulum. Finally, depending on the expression level, different proportions of ctm, ntm, or sec forms are generated, some of which are possibly more or less resistant to conversion (51). This is the first description of the dominant-negative inhibition of mutant T182N/T198A, as the mutations were neither located in the forecasted region of the postulated protein X-binding site nor in the area assumed to be involved in species barrier effects (35,36).