Altered Glycosylated PrP Proteins Can Have Different Neuronal Trafficking in Brain but Do Not Acquire Scrapie-like Properties*

N-Linked glycans have been shown to have an important role in the cell biology of a variety of cell surface glycoproteins, including PrP protein. It has been suggested that glycosylation of PrP can influence the susceptibility to transmissible spongiform encephalopathy and determine the characteristics of the many different strains observed in this particular type of disease. To understand the role of carbohydrates in influencing the PrP maturation, stability, and cell biology, we have produced and analyzed gene-targeted murine models expressing differentially glycosylated PrP. Transgenic mice carrying the PrP substitution threonine for asparagine 180 (G1) or threonine for asparagine 196 (G2) or both mutations combined (G3), which eliminate the first, second, and both glycosylation sites, respectively, have been generated by double replacement gene targeting. An in vivo analysis of altered PrP has been carried out in transgenic mouse brains, and our data show that the lack of glycans does not influence PrP maturation and stability. The presence of one chain of sugar is sufficient for the trafficking to the cell membrane, whereas the unglycosylated PrP localization is mainly intracellular. However, this altered cellular localization of PrP does not lead to any overt phenotype in the G3 transgenic mice. Most importantly, we found that, in vivo, unglycosylated PrP does not acquire the characteristics of the aberrant pathogenic form (PrPSc), as was previously reported using in vitro models.

Glycoproteins are subject to a number of post-translational modifications as they pass through the secretory pathway. During polypeptide chain synthesis, N-glycosylation is initiated by the transfer of core glycans to target asparagines. Processing of core glycans into the complex type is then achieved in the endoplasmic reticulum (ER) 4 and Golgi apparatus compartments (1). Protein-attached glycans have been shown to have a wide range of biological functions, most notably stabilization of protein structure and cellular trafficking (2).
PrP is a glycoprotein attached to the cell membrane by a glycosylphosphatidylinositol (GPI) anchor (3)(4)(5). Whereas its normal function has yet to be defined, expression of PrP is essential for the development of transmissible spongiform encephalopathy (TSE) or prion disease (6,7). The TSEs are a group of fatal neurodegenerative diseases that can be sporadic, inherited, or acquired by infection. TSE diseases include scrapie of sheep and goats, bovine spongiform encephalopathy in cattle, and a number of human forms of the disease such as Creutzfeldt-Jackob disease, variant Creutzfeldt-Jackob disease linked with bovine spongiform encephalopathy, Gerstmann-Straussler-Scheinker syndrome, Kuru, and fatal familial insomnia (FFI) (8 -10). A central event in all prion diseases appears to be a conformational modification of the normal cellular prion protein (PrP C ) from a soluble form with a predominant ␣-helical conformation to the pathogenic form (PrP Sc ) that is aggregated, rich in ␤-sheets, partially resistant to proteinase K digestion, and insoluble in nondenaturing detergents (11).
PrP contains two N-glycan attachment sequences (NXT) at amino acids 180 and 196 in mice. These sites are variably glycosylated in vivo such that un-, mono-, and diglycosylated glycotypes are observed (12,13). The biological significance of each of the glycotypes of PrP (un-, mono-, and diglycosylated) is unknown. Both N-glycosylation sites are conserved in the PrP gene (Prnp) from all species, suggesting that N-glycans play an important role in the protein function (14).
A number of reports have shown that the lack of sugars can induce the PrP C to PrP Sc transition in vitro, suggesting that perturbations in glycosylation may contribute to the development of disease, destabilizing PrP structure and allowing it to acquire spontaneously PrP-like properties (15)(16)(17)(18)(19). It has also been reported that alterations in glycosylation may alter the intracellular trafficking of PrP (16, 17, 20 -22).
To investigate the in vivo effect of glycosylation on PrP biochemical properties and its cellular biology, we have developed a gene-targeted transgenic model in which the host Prnp is replaced by a modified Prnp transgene in the correct genomic location (23). This model represents a valid tool to analyze the effect of mutations of the host Prnp in TSE susceptibility, since the Prnp gene expression is controlled by the normal regulatory elements of endogenous PrP (24 -28).
Three transgenic lines have been generated, each containing a point mutation in the Prnp gene eliminating the first, second, or both of the glycosylation sites: N180T (G1), N196T (G2), and N180T-N196T (G3). Using these mice, we have investigated whether the lack of glycans can alter the expression level of the PrP protein, its conformation and intracellular localization, and its ability to acquire the biochemical characteristics of the pathogenic form.
We report here that whereas glycans appear to control the cellular location of PrP, the presence of sugars does not dramatically change the biology of PrP, and there is no evidence of PrP Sc -like properties in either mono-or unglycosylated PrP.
The results reported here are important in determining the physiological function of PrP glycoforms and in understanding their role in the infectious and pathogenic process of TSEs.

Antibodies
Mouse monoclonal antibody 8H4, epitope (residues 145-220) binding is independent of the N-linked glycosylation, because it reacts with both recombinant PrP and all native glycoforms (29). 7A12 (epitope 90 -140) is a mouse monoclonal anti-PrP antibody (30). FH11 is a mouse monoclonal antibody that binds the N-terminal region of PrP and is used extensively in enzyme-linked immunosorbent assays (31). AG4 is a mouse monoclonal antibody with epitope recognition between residues 31 and 51, with a further area of binding between amino acids 147 and 163. 1B3 and 1A8 are both rabbit polyclonal antibodies against PrP. Rat monoclonal anti-tubulin antibody (Abcam) has been used as loading control in Western blot experiments. Alexa Fluor 488 and Alexa Fluor 568 are IgG-labeled with fluorescent dye (Molecular Probes, Inc., Eugene, OR). Rabbit anti-cow glial fibrillary acidic protein (DAKO) is an antibody recognizing a specific astrocytic marker. The endoplasmic reticulum marker anti-ERp60 is raised in rabbit against porcine ERp60 peptide PIIQEEKPKKKKKAQEDL in the C terminus of the protein (32). The Golgi marker 23C rat monoclonal, IgG2c, clone 23c was raised in rats against recombinant mouse TCP-1␣, C-terminal half (33).

Generation of Targeting Vectors for the Prnp a180T and Prnp a196T Alleles
PrP codon 180 and 196 alterations were introduced into HM-1 embryonic stem (ES) cells. Briefly, a gene-targeting vector was constructed using isogenic 129/Ola Prnp a DNA from a HM-1 genomic library in DASH II (Stratagene). The PrP codon 180 and 196 alterations were introduced into a 1.1-kb XmaIII-EcoRI exon 3 fragment containing the open reading frame by the Kunkel method (34). This was ligated with the 5Ј and 3Ј homologous sequences derived from a 7.8-kb BamHI-EcoRV genomic clone spanning 129/OlaPrnp exon 3. A LoxPneomycin/thymidine kinase-selectable cassette (provided by Alan Clarke, University of Cardiff, UK) was ligated into a unique SalI site 1600 bases downstream of exon 3 in the pBluescript plasmid (Stratagene). The pBluescript vector previously had its SalI site removed, so this was a unique site in the targeting vector. The open reading frame encoding PrP in the targeting vector was sequenced at each step in the cloning procedure to confirm the presence of the alterations and the absence of any other cloning artifacts.

Embryonic Stem Cell Culture and Gene Targeting
Culture conditions for the ES cell line HM-1 have been described previously (35). HM-1 cells (5 ϫ 10 7 ) were electroporated using a gene pulser (Bio-Rad) at 800 V and 3 millifarads with 250 g of linearized targeting vector DNA in 0.8 ml of Hepes-phosphate-buffered saline, pH 7.05. Cells were rested for 15 min and plated at 5 ϫ 10 5 /10-cm plate. 24 -48 h after electroporation, G418 selection medium was added. Medium was changed every 2-3 days, and colonies were selected for PCR screening 15 days after electroporation. 10 7 targeted HM-1 cells in 0.8 ml of serum-free growth medium were electroporated with 25 g of the plasmid pCre2 (provided by Alan Clarke). Two pulses of 230 V, 500 microfarads were given. Cells were rested for 15 min and plated at 10 4 cell/10-cm plate. On day 6 after electroporation, 2 mM gancylovir was added to the growth medium. Colonies were picked and screened on day 15.

PCR Screening for Targeting Events
Half of the cells from surviving colonies were used to prepare DNA for all PCR analyses Detection of Homologous Recombination Events-A 1600-bp PCR product was synthesized between the neomycin/thymidine kinase cassette and a site outside the targeting vector. The reaction-specific oligonucleotides are LoxP, situated immediately upstream of the 3Ј loxP site (TCGATCGACTAGAGCTTGCGGA), and 3ЈMap1, located 200 bases 3Ј to the EcoRV site (CTAAGTGACCTAGGCACATGTC). The cycle conditions were 3 min at 94°C and then 35 cycles of 1 min at 94°C, 1 min at 60°C, and 2 min at 72°C and then 10 min at 72°C (GeneAmp 9700; PerkinElmer Life Sciences). Those positive for the selection cassette were then analyzed for the glycosylation mutation using the mismatch-specific PCR reaction described below for genotyping.
Removal of the Selectable Marker-The removal of the selectable marker left one LoxP site. This is screened for using oligonucleotides 5Ј and 3Ј to the PrP gene SalI site NLTVitro creA (AGAACAGGTCT-GACCACACTGGTT) and NLTVitro creB (AATGGTTAAACTT-TCGTTAAGGAT). Wild type PrP alleles will give a PCR product of 242 bp, whereas those containing a loxP site will be 342 bp. Sites containing an unexcised neomycin/thymidine kinase cassette would be over 5 kb. The cycle conditions were 3 min at 94°C and then 30 cycles of 45 s at 94°C, 45 s at 60°C, and 45 s at 72°C and then 10 min at 72°C.

Generation of Gene-targeted Mice
Targeted ES cells were used to generate chimeric mice as described previously (25) to obtain G1 and G2 heterozygous mice expressing mono-and unglycosylated PrP and G3 heterozygous mice expressing unglycosylated PrP. Heterozygous mice were bred to produce an inbred homozygous line. 129/Ola mice were used as wild type controls, since the transgenics had been generated on a 129/Ola background. NPU PrP Ϫ/Ϫ mice (7) were used as negative controls in all of the experiments performed.

PCR Genotyping of Mouse Tail DNA
G1 and G2 mutant alleles were detected using a mismatch PCR technique. An oligonucleotide mixture was used at 1 pmol that contained a forward oligonucleotide 9910 (AACCTCAAGCATGTGGCAGGGG-CTGCGGCAGCTGG), a reverse oligonucleotide 9912 (TCAGTGCC-AGGGGTATTAGCCTATGGGGGACACAG), and a mutant-specific or wild type oligonucleotide (also in the reverse orientation) in a ratio of 20:1:20.

Southern Blot Analysis of Prnp
Genomic DNA was prepared using a Puregene Isolation kit (Gentra Systems). DNA (15 mg/reaction) was digested with restriction enzymes and then separated on a 1% agarose gel and blotted to Hybond-N nylon membrane (Amersham Biosciences). Hybridization was performed using ULTAhyb solution (Ambion) using a 700-bp EcoRV-BamHI fragment (3Ј probe) and an 884-bp PCR product as probes. Following strin-gent wash procedures (0.1ϫ SSC at 65°C), the blots were exposed to x-ray film for 2 days.

Northern Blot Analysis of Total mRNA from Brains of Transgenic Mice
Total RNA was isolated using RNAzol TM B (Biogenesis) based on the guanidinium thiocyanate/phenol/chloroform extraction method (36). A 20-g aliquot of total RNA was separated on a 1.0% agarose-formaldehyde denaturing gel, transferred to Hybond N (Amersham Biosciences), and probed with a 32 P-labeled 936-bp KpnI-EcoRI fragment from exon 3 of Prnp.

Western Blotting
Mice were killed by cervical dislocation, and brains were removed, flash frozen in liquid nitrogen, and then stored at Ϫ70°C until required. Half or whole brains were weighed and mechanically homogenized from frozen in nine volumes of ice-cold Nonidet P-40 lysis buffer (1% Nonidet 40, 0.5% sodium deoxycholate, 150 mM NaCl, 50 mM Tris, pH 7.5) with the addition of phenylmethylsulfonyl fluoride (PMSF) (final concentration 1 M; Sigma) to prevent protein degradation by endogenous proteases. The homogenate was centrifuged at 8000 rpm for 10 min to remove debris. Total protein was denatured in 1ϫ Novex Tris/ glycine SDS sample buffer (Invitrogen) and 1ϫ NuPage sample-reducing agent (Invitrogen) for 30 min at 95°C. Proteins were separated by electrophoresis at 125 V through a Novex precast Tris/glycine gel (12 or 14% acrylamide, Tris/glycine; Invitrogen). Proteins in the acrylamide gel were transferred to polyvinylidene difluoride membrane at 25 V (125 A/gel) using a semidry transfer blotter (Bio-Rad) in 1ϫ transfer solution (48 mM Tris, 39 mM glycine, 0.375% SDS, 20% methanol).

Proteinase K Digestion
Mouse brain homogenates (10%) were prepared in ice-cold Nonidet P-40 buffer. Each homogenate was then split into two aliquots, one treated with proteinase K (PK; Roche Applied Science) and one not. In order to assess the sensitivity to enzyme digestion, wild type, G1, G2, and G3 brain homogenates were each treated with varying concentrations of PK: 20, 10, and 5 g/ml at 37°C for 1 h. A milder treatment was also carried out, incubating the samples with PK (20 g/ml) at 4°C for 1 h. The samples were then analyzed by Western blotting using 8H4 or 7A12 monoclonal antibodies for PrP detection.
DELFIA enhancement solution was added to the samples to facilitate the formation of Eu-(2-NTA) 3 (TOPO) 2-3 . After 5 min of shaking at room temperature, Eu 3ϩ emission (615 nm) was calculated using a time-resolved technique. Between each step, the plate was washed in 1ϫ DELFIA wash concentrate (TBST; PerkinElmer Life Sciences) using the DELFIA automatic plate washer (Wallace). The program Work-Out was used to analyze absorbance from standard and samples and to produce the standard curve (based on a linear model of emission).

Detergent Solubility Assay
PrP C was extracted from brain homogenate (10 Ϫ1 tissue; Nonidet P-40 lysis buffer; 1 mM phenylmethylsulfonyl fluoride) by mechanical homogenization in 1 M guanidine hydrochloride (25 mM Tris, 1 M guanidine hydrochloride (Sigma); 0.5% Triton X-100 (Sigma)). This was then diluted in DELFIA assay buffer, leading to a final concentration equivalent to of 10 mg/ml of original tissue. Proteins insoluble in 1 M guanidine hydrochloride (PrP Sc ) were separated from those that were soluble (PrP C ) by centrifugation at 13,000 rpm for 10 min. The resultant pellet was resuspended in 6 M guanidine hydrochloride prior to dilution in DELFIA assay buffer, to a concentration equivalent to 10 mg/ml original tissue. Measurement of the sample concentration was then performed as described above (DELFIA analysis).

Phospholipase C Assay
Mouse brain homogenates (10%) were prepared by homogenizing in ice-cold PBS containing 10 mM phenylmethylsulfonyl fluoride. Each sample was centrifuged at 13,000 rpm for 5 min at room temperature. The supernatant was collected and centrifuged at 25,000 rpm for 10 min at 4°C. The pellets were resuspended in 500 l of cold PBS, and then each sample was split. One half was treated with phosphatidylinositol phospholipase C (PIPLC; 0.5 units/ml; Sigma), whereas the second remained untreated. The samples were incubated for 10 h at 4°C. Samples were then centrifuged at 13,000 rpm for 15 min at 4°C. The supernatant (membrane-released fraction) and the pellet (membrane-associated fraction) were analyzed by Western blotting using the monoclonal antibody 8H4.

Aging Experiment
A group of 10 homozygous G3 mice were monitored up to 850 days and compared with a group of wild type mice. The animals were ageand sex-matched. Animals were monitored constantly by a group of independent observers for any neurodegenerative signs. Between 800 and 900 days, mice were culled. Brains were retained; half of the brain was fixed for standard lesion profiling and plaque analysis, and the other half was flash frozen for biochemical analysis.

Lesion Profiles
Brain Sections were hematoxylin and eosin-stained and scored for vacuolar degeneration on a scale of 0 to 5 in nine standard gray matter areas and three standard white matter areas as described previously (38).

Immunocytochemical Analysis
Sections were immunostained using standard procedures. Briefly, sections were blocked with normal goat serum and probed overnight with polyclonal antibody anti-glial fibrillary acidic protein at a dilution of 1:400. A parallel panel of sections was also probed with normal mouse serum as a control. Antibody binding was detected with biotinylated goat anti-rabbit secondary antibody (Jackson) and the Vectastain Elite ABC Kit (Vector Laboratories). Reaction products were visualized with diaminobenzidine, and sections were lightly counterstained with hematoxylin. Pictures were taken using a Nikon Eclipse E800 microscope.

Confocal Analysis
Mouse brains were fixed in periodate/lysine/paraformaldehyde or in a paraformaldehyde/glutaraldehyde mix (4% paraformaldehyde, 0.025% glutaraldehyde) for 4 h. Brain sections were cut at a thickness of 70 m using a vibrating microtome (Leica). The sections were permeabilized for 1 h in PBS, 0.1% Triton at room temperature before blocking overnight at room temperature in PBS, 0.5% BSA in a humid chamber. Sections were blocked for a further 1 h at 37°C in Mouse On Mouse Ig blocking reagent/PBS when mouse monoclonal antibodies were used (Vector Laboratories). After blocking, the tissues were incubated at 37°C with primary antibody (diluted in PBS/protein concentrate when mouse monoclonals were used; Vector Laboratories) for 90 min (39). . After extensive washes in a calcium/magnesium-free PBS solution, the sections were incubated at 37°C with the secondary antibody goat anti-mouse Alexa Fluor 488 conjugate diluted 1:200 in PBS/BSA for 90 min. The samples were then washed with a calcium/magnesium-free PBS solution for 20 min and then stained with 4Ј,6-diamidino-2-phenylindole nuclear marker (1:10,000; Molecular Probes) for 30 min at room temperature. After extensive washes in ultrapure water, the sections were mounted for microscopic analysis. Co-localization experiments were carried out using the same basic method as above with some modifications. For ERp60/8H4 (1:400 and 1:500), the primary antibodies were mixed in PBS/BSA so that the tissue was incubated with them simultaneously. Similarly, the secondary antibodies, Alexa 488-conjugated goat anti-mouse antibodies and Alexa 568-conjugated goat anti-rabbit, were simultaneously incubated. Controls for cross-reactivity were used, and none was detected. For 23C/8H4, cross-reactivity was detected; to eliminate it, a sequential staining method was used. Sections were first incubated with 8H4 (1:500) diluted in PBS/protein concentrate (Vector Laboratories), washed in PBS, and then incubated with Alexa 488conjugated goat anti-mouse using a method identical to that for single 8H4 labeling. Sections were then washed 10 times in PBS before being blocked for 1 h in MOM Ig-blocking reagent (Vector Laboratories). Block was removed by washing for 10 min in PBS/protein concentrate before incubation of the section with 23C (1:50) in PBS/protein concentrate. The sections were then washed and incubated with Alexa 568conjugated goat anti-rat. The sections were washed, stained with 4Ј,6diamidino-2-phenylindole, and mounted for analysis. Sections were imaged with a Leica TCS SP2 laser-scanning confocal microscope.

RESULTS
Construction of Gene-targeted Mice with Altered N-Linked Glycosylation of PrP-Using the Cre-loxP recombination and gene-targeting approaches, three inbred lines of transgenic mice with alterations in the N-linked glycosylation consensus sites Asn-Xaa-Thr were generated. Gene targeting was used to alter the Asn residue to Thr at 180 (N180T), 196 (N196T), or both 180 and 196 (N180T/N196T) (Fig. 1) in the HM-1 ES cell line. The positive ES clones were subsequently confirmed by DNA sequencing and Southern blot analysis (data not shown).
The characterized targeted ES cells were microinjected into blastocysts of C57Bl6 mice to obtain chimeric mice. Chimeric mice were identified by coat color and mated with 129/Ola mice. ES cell-derived offspring were recognized by eye and coat color and genotyped to identify gene-targeted transgenic mice. The heterozygous offspring were then interbred to obtain homozygous inbred lines carrying the mutated Prnp gene.
mRNA Expression and Protein Levels of Mono-or Unglycosylated PrP Are Similar to Wild Type-The level of expression of the PrP gene from the different gene-targeted lines was assessed and compared with the wild-type gene. Northern blot analysis detected similar levels of PrP mRNA in mice with the mutant Prnp alleles (N180T, N196T, and N180T/N196T) and in wild-type mice ( Fig. 2A). Thus, the transgene expression levels in these new lines are the same as the wild type lines, showing that neither the point mutation nor the presence of the LoxP site downstream of Prnp interfere with gene transcription.
Western blot analysis of brain homogenates of the transgenic mice demonstrates that both G1 and G2 lines lack diglycosylated, but possess monoglycosylated and unglycosylated, PrP, whereas G3 mice only exhibit unglycosylated PrP, as confirmed by deglycosylation with peptide N-glycosidase F enzyme (Fig. 2B).
Western blot analysis using different monoclonal PrP antibodies indicated that the steady-state level of the PrP protein in glycosylationdeficient transgenic mice is apparently lower than that in wild-type mice (Fig. 2B). However, accurate quantification of the difference in amount of PrP between the lines of mice has proved difficult by Western blot analysis; thus, a more quantitative assay system has been utilized to address this question.
A DELFIA assay was set up in order to quantify the total amount of PrP in brains from the three transgenic lines and compare it with that in wild type. Levels of PrP in brains from G1 and G2 transgenic lines were similar to that of wild type brains in the DELFIA system. Unglycosylated PrP amount is slightly lower (10% less) by this analysis. Since only a slight reduction of PrP levels was observed in G3 mice, any biological effect is more likely to be due to lack of glycosylation than to this reduction in the amount of protein (Fig. 3). The presence of such a high level of the unglycosylated PrP was surprising, since with the absence of sugars, it was believed that this protein would not mature and be rapidly eliminated by the cell. This finding shows that the unglycosylated PrP is considered to be a normal protein by the cellular quality control system, since it accumulates to significant levels in vivo.
Altered Glycosylation: PrP Does Not Acquire PrP Sc -like Properties-It is possible to distinguish between PrP Sc and PrP C conformers on the basis of biochemical properties: (i) PK partial resistance; (ii) insolubility in detergents; (iii) resistance to the cleavage with PIPLC enzyme. Several studies performed in cell cultures (16,17,19) have shown unglycosy- . PrP quantification and solubility analysis in wild type and transgenic mouse brains. The total amount of PrP was tested using a DELFIAா immunoabsorbent assay. 96-well plates were coated using capture antibody FH11. Brain homogenates from wild type, G1, G2, G3, null, and ME7-infected mice were diluted in the appropriate buffer and added to the pretreated wells. PrP was detected using detector antibody 7A12 and shaking at room temperature for 1 h and based on europhium emission (615 nm) using a time-resolved technique. The total PrP amount in transgenic mice brains was comparable with wild type with a slight reduction in G3 brains. All PrP was recovered in the 1 M guanidine hydrochloride fraction (black bars), where PrP C is soluble. PrP was detected in the 6 M fraction (gray bars) only in the ME7-infected control brain. FIGURE 4. PK resistance Western blot analysis. Brain homogenates from wild type, transgenics, null, and ME7-infected mice were treated (ϩ) or not (Ϫ) with PK at different dilutions (20 g/ml (A) or 10 g/ml (B)). PrP was detected using monoclonal antibody 8H4. No bands were detected in wild type and in transgenic brains when PK-treated, whereas PK-resistant PrP was detected in the ME7-infected brain at both enzyme concentrations. The band between 64 and 50 kDa represents tubulin used as loading control.

FIGURE 2. Expression of glycosylation-deficient PrP.
A, Northern blot analysis of PrP mRNA in the brain. Total brain RNA was probed with a 936-bp KpnI-EcoRI mouse PrP exon 3 DNA probe to demonstrate that PrP mRNA production levels in G1, G2, and G3 mice were comparable with wild type ones with no over-or underexpression phenomena. B, Western blot of brain homogenates from mice expressing wild type PrP and gene-targeted PrP. Samples were treated with peptide N-glycosidase F enzyme (ϩ) or not (Ϫ) and then resolved on SDS-PAGE and transferred to polyvinylidene difluoride membrane. PrP was detected using mouse monoclonal antibody 8H4. PrP Ϫ/Ϫ (null mice) were used as negative controls. Altered N-glycosylation has been successfully achieved, introducing the N180T or/and T196T point mutations in Prnp. The band between 64 and 50 kDa represents tubulin used as a loading control lated PrP spontaneously acquiring PrP Sc -like properties. We aimed to investigate whether these phenomena also occur in vivo with a normal level of PrP expression or if the effects observed in cell cultures might be due to transgene overexpression, ectopic expression, or some other anomaly of the in vitro system.
First we analyzed the sensitivity to proteolytic action of PK enzyme in brain homogenates from all three lines of transgenic mice. PrP Sc partial resistance to the proteolytic digestion of PK is detected by the presence of a 27-30-kDa fragment by Western blotting (11). Different enzyme concentrations (20, 10, and 5 g/ml) and reaction temperatures (4°C, 20 g/ml) did not reveal any difference in the PK resistance of PrP between wild type and transgenic forms, demonstrating that a lack of glycans does not increase the resistance of PrP to PK (Fig. 4, A and B, and data not shown).
To assess the solubility of mono-and unglycosylated PrP, a standard guanidine hydrochloride assay was employed. Brain homogenates were treated with guanidine hydrochloride at a low molarity (1 M) known to solubilize PrP; any material remaining insoluble at this concentration was then solubilized at a higher concentration (6 M) known to release PrP Sc into solution. Treated brain fractions were then assayed in a DELFIA assay. As in wild type animals, low guanidine hydrochlorideinsoluble PrP was not observed in any of the glycosylation transgenics, in contrast to the significant amount detected in the ME7 control (Fig. 3).
PIPLC bacteria-derived enzyme cleaves the eukaryotic GPI anchor, releasing bound proteins from the membrane. Different sensitivities between PrP C and PrP Sc to the action of the enzyme have been shown, with PrP C being sensitive and PrP Sc being resistant to the enzyme (40). To investigate this difference, brain homogenates were treated with PIPLC, and Western blot analysis was carried out. This assay revealed the same characteristics between wild type PrP and mono-or unglycosylated PrP. The ability of mono-and unglycosylated PrP to be cut by PIPLC enzyme is shown in Fig. 5A, where bands of all of the samples that were detected in the supernatant fraction (enzyme-released) migrated more slowly than the bands detected in the pellet (enzyme-resistant). No shifted bands were detected in the supernatant fraction of all trans-FIGURE 5. Membrane-attached PrP in wild type and transgenic mice. A, brain homogenates from wild type and transgenic mice were treated with PIPLC enzyme and incubated for 10 h at 4°C. The supernatant (S; membrane-released fraction) and the pellet (P; membrane-associated fraction) were separated by centrifugation at 13,000 rpm for 15 min at 4°C and then analyzed for the presence of PrP by Western blotting using the monoclonal antibody 8H4. All glycosylation-altered PrPs are associated with membranes as for wild type, since a slower migrating band was detected in the s lane in all samples following enzymatic digestion. B, when samples were not treated with PIPLC, no slower migrating bands were detected in the supernatant fractions. genics when brains had not been treated with the enzyme (Fig. 5B). This assay suggests that all three mutants are GPI-anchored proteins still sensitive to PIPLC enzymatic digestion. Notably, the presence of a GPI anchor in the unglycosylated PrP structure suggests that this protein is correctly processed, since the attachment of the anchor is the last event in the protein maturation process in the ER. The observation that this step occurs without the presence of any sugar in the PrP structure shows once again how the presence of unglycosylated PrP is tolerated within the secretory pathway as well as di-and monoglycosylated forms.
Aging Analysis of Transgenic Mice-All glycosylation-deficient transgenic animals bred and aged normally and did not display any overt phenotype.
However, to rule out any PrP C /PrP Sc pathological transition due to the absence of carbohydrates over a long term period, an aging experiment of mice was set up. A group of uninfected G3 mice and wild type controls was monitored for up to 850 days. None of these mice developed clinical signs of TSE neurodegeneration. Moreover, to fully exclude the presence of PrP Sc , aged mice were culled at time points, and then brain homogenates were digested with PK and analyzed by Western blotting. As shown in Fig. 6, no PK-resistant PrP was detected in any of the animals, confirming that the in vivo lack of sugars, when altered PrP is expressed at natural levels, is not sufficient to destabilize PrP C structure and produce a pathogenic form. Microscopic analysis carried out in brain sections did not reveal any abnormal deposition of PrP in all of the aged mice, ruling out any possibility of aggregation of unglycosylated PrP (supplemental Fig. 1). Moreover, no gliosis was observed in aged wild type and transgenic mice brains (supplemental Fig. 2).

Unglycosylated PrP Localization Is Mainly Intracellular, whereas Monoglycosylated PrP Is Localized in the External Membrane-To
establish if the glycosylation is important in determining trafficking of PrP inside neurons, a confocal microscopy approach was developed using brain sections. Although highly expressed in the brain, PrP C is very difficult to detect. A number of monoclonal and polyclonal anti-PrP antibodies were tested: 1B3, 1A8, AG4, and 8H4 at different concentrations. Different types of tissue fixative methods were also considered (see "Experimental Procedures"). We observed the best results using 8H4 in 4% paraformaldehyde, 0.025% glutaraldehyde fixed tissues.
Labeling for PrP was seen throughout all of the brain sections, although it was less intense at the periphery of the brain. Analysis was focused on the hippocampus. In wild-type mice, PrP labeling was closely associated with the cell membrane, with some intracellular labeling also observed (Fig. 7A). This observation is in accord with several results previously obtained both in vitro and in vivo with different detection techniques (41)(42)(43)(44)(45)(46)(47)(48).
Clear membrane staining was also observed in G1 and G2 brains (Fig.  7, B and C), suggesting that the presence of just one sugar chain at either the first or the second site is sufficient for its trafficking through the secretory pathway to the external membrane. This differs from previous reports in transgenic mice overexpressing monoglycosylated hamster PrP or in transfected cell cultures, where altering the first glycosylation consensus site influenced the intracellular fate of PrP, blocking its trafficking to the cell membrane (16,21,49).
In G1 and G2 mice, although PrP was mainly on the cell membrane, there was a greater proportion of PrP protein located within the cell than in wild type mice. It is difficult to compare actual levels of labeling from one animal to the other, but it was clear that the ratio of membrane to cytoplasmic labeling was shifted toward the cytoplasm in the G1 and G2 mice. The more intense intracellular signal detected in these transgenic mice suggests that the intracellular trafficking rate may be altered when the first or second glycosylation site has been ablated with a slower migration of these proteins toward the outer membrane.
Furthermore, a clear and consistent increase in intracellular compared with membrane labeling for PrP was observed in G3 when compared with wild type, G1, and G2 brains (Fig. 7D). However, this intracellular labeling was of a pattern similar to that observed in the wild type, being tightly defined consistent with protein localization in subcellular compartments. Double labeling analysis using ER and Golgi markers revealed that mono-and unglycosylated PrP are not blocked in the ER. No co-localization was observed between PrP and ERp60 ER marker, suggesting that glycosylation-deficient proteins are able to leave this compartment and traffic toward the cytoplasm and the cell membrane (Fig. 7, A-D). This was further proven with a double staining using 8H4 and Golgi marker 23C. A significant proportion of the total intracellular anti-PrP was observed to be surrounded by anti-␤-cop labeling, indicating a localization of PrP in the Golgi. The ␤-cop labeling appeared to surround a significant proportion but not all of the intracellular PrP in both wild type controls and the glycosylation-deficient transgenics (Fig.  7, F-I). The non-␤-cop-associated intracellular fraction is most likely contained in endosomes, lysosomes, or similar endocytic compartments, since previous work has demonstrated significant localization of wild type PrP in these organelles (44,46). It appears that unglycosylated PrP therefore is able to enter in the secretory pathway but is then retained in the Golgi apparatus before it reaches the cell membrane. Importantly, we have demonstrated that unglycosylated PrP is a GPIanchored cytoplasmic protein, and it is therefore likely to be bound to intracellular membranes.
To establish the specificity of the fluorescent signal, PrP knockout mice brain sections have been treated with the same method. No signal was detected with both primary and secondary antibodies (Fig. 7, E and J) imaged with the same microscope settings as used for the wild type mice.
The microscopic analysis described here clearly shows how sugars can be important in determining the location of PrP protein in the cell.

DISCUSSION
PrP glycosylation may represent the key factor in understanding not only PrP function but also TSE infectious process and the existence of a number of strains of agent in TSE disease (50,51).
We report here that un-or monoglycosylated PrP is expressed at physiological levels in the central nervous system. Unglycosylated PrP does not acquire PrP Sc characteristics, such as PK partial resistance or detergent insolubility. Thus, it is unlikely that a lack of glycans can FIGURE 6. PK resistance in aged wild type and transgenic mouse brains. Brain homogenates from wild type and transgenic mice aged for up to 850 days and culled at different time points were treated (ϩ) or not (Ϫ) with PK enzyme (20 g/ml) for 1 h at 37°C. Null and ME7-infected mouse brain homogenates were also analyzed as controls.
PrP detection was carried out by Western blot analysis using monoclonal antibody 8H4. No bands were detected in wild type and in transgenic brains when PK-treated, whereas PK-resistant PrP was detected in the ME7-infected brain. The band between 64 and 50 kDa represents tubulin used as loading control. Wt (1)  Colocalization of PrP and markers of the ER and Golgi in mouse brain sections using confocal analysis. PrP was detected using the mouse monoclonal 8H4 and an Alexa 488-conjugated anti-mouse (A-E) or Alexa 488-conjugated anti-mouse IgG1 secondary antibody (green) (F-J). The ratio of intracellular to extracellular PrP is higher in G3 brains (D and I) compared with wild type controls (A and F). No colocalization of PrP with endoplasmic reticulum-resident protein ERp60 was detected in either the wild type control (A) or the G1 (B), G2 (C), and G3 (D) transgenics, using a rabbit polyclonal antibody raised against an ERp60 C terminus peptide and an Alexa 568-conjugated anti-rabbit secondary antibody (red). Localization of PrP in the Golgi apparatus was determined using a rat monoclonal antibody that reacts with the ␤-cop subunit of coatamer and an Alexa 568-conjugated anti-rat secondary antibody (red) (F-J). In the wild type control, a proportion of intracellular PrP (green) colocalizes with the Golgi apparatus (F). A similar pattern of Golgi localization is seen in the G1 (G), G2 (H), and G3 (I) brain sections. The specificity of 8H4 staining is shown by the use of PrP null sections (E and J). destabilize the entire protein structure, facilitating the onset of a TSE phenotype.
To date, the role of carbohydrates in PrP cell biology had been addressed using in vitro transfection studies or overexpressing transgenic mouse models in vivo, which have produced contradictory results. PrP overexpression represents a major problem, because it is now clear that different results may be obtained when PrP expression levels are altered in both cell cultures and transgenic mice (23,52). Some experiments performed in cell culture models have shown that the lack of carbohydrates can in some way destabilize PrP C structure, thus allowing it to acquire all of the PrP hallmarks (16,17,19). However, recently, Neuendorf et al. (22) have shown that altered glycosylated PrPs display just some of the pathogenic protein characteristics, such as detergent insolubility, while maintaining the PK sensitivity of wild type PrP. Moreover, in the case of a monoglycosylated PrP, they also observed a resistance to PIPLC, whereas other mutants can be released by the enzyme. This discrepancy of results obtained may be the result of a combination of factors: different point mutations introduced in the PrP gene, different constructs, distinct cell lines, random integration of the transgene, and different levels of Prnp expression or in some cases use of drugs to prevent glycosylation that can cause intracellular stress. Moreover, experiments that introduced the T182A mutation to abolish the attachment of sugars at the first site (16,20,21) may be misleading, since this mutated PrP can cause familial TSE disease (49,53) in a glycosylationindependent manner (54).
A number of transgenic lines have been developed by our group using the gene targeting technique, whereby altered PrP has been introduced in the correct genomic location (23). This system is important not only for understanding the contribution to the disease by mutated endogenous PrP but also to study the biology of these proteins. Gene-targeted transgenic mice may indeed represent the best comparison for wild type animals, since any alteration observed can be directly related to the mutated PrP protein expressed. Importantly, this system allows a direct comparison not only between transgenic lines and wild-type mice but also between different transgenic lines.
All three glycosylation mutants we have generated retain the same biochemical characteristics of wild type PrP. Surprisingly, the amounts of unglycosylated protein in brain were comparable with wild type and monoglycosylated PrP levels. In theory, one of the functions of the sugars is to stabilize the glycoproteins and facilitate folding (1,2). If this does not happen, the protein can be considered nonmature, and it is eliminated by the ubiquitin/proteasome system (55). The lack of sugars has been previously reported to influence the degradation by the ubiquitin/proteasome system of PrP, decreasing its half-life in cell cultures, and it is generally believed that PrP without sugars is a nonmature protein (22,56). Here we report that unglycosylated PrP is apparently not considered by the ubiquitin/proteasome system to be nonmature and an unfolded protein, because its amount in the brain is similar to that of wild type. The additional observation that sugar-deficient PrP has a GPI anchor shows once again that during its synthesis and translocation in the ER, the protein is not blocked as would normally happen with unfolded proteins. Since the unglycosylated form of PrP exists in vivo, the cell does not need to eliminate it faster, suggesting that this unglycosylated protein is functional.
Using our models, we have also investigated the contribution of carbohydrates to directing PrP intracellular trafficking and localization inside neurons of the central nervous system. Several earlier cell culture studies investigated the possible role of sugars in PrP localization, and different results have been reported. In general, unglycosylated PrP is not detected at the cell surface, remaining trapped in the cytoplasm (16,20,21). However, Korth et al. (17) and later Neuendorf et al. (22) detected unglycosylated PrP on the cell surface of different cell lines when transfected with some but not all glycosylation mutants, suggesting that the mutation of an amino acid rather than the lack of sugars can influence the intracellular fate of PrP. However, it has been also observed that PrP without complex-type glycans after treatment with geldanamycin localizes to the cell surface despite any alteration in Prnp (19), suggesting that PrP can traffic independently of the presence of mature sugars or different amino acids. Our results show that the presence of just one sugar chain is sufficient for the protein to leave the intracellular compartments of the endoplasmic reticulum and Golgi apparatus and traffic to the cell surface. Contrastingly, unglycosylated PrP has a mainly intracellular localization. Using a double staining approach, we were able to establish that PrP without carbohydrates is not held in the ER, but its main localization is associated with Golgi apparatus. This protein is therefore considered mature by the cell and, leaving the ER, enters in the secretory pathway like di-and monoglycosylated forms. The lack of sugars causes PrP trafficking to stall and the protein to remain in the cytoplasm. Interestingly, this protein is still GPI-anchored, probably attached to intracellular membranes, where it may have a physiological role. Alternatively, di-and monoglycosylated proteins can leave the Golgi apparatus and traffic to the cell membrane with a small amount remaining in the cytoplasm. In our model, all PrP is expressed under the endogenous regulatory sequences at physiological levels and is subjected to the cellular control machinery without any possibility of artifacts arising from overexpression. Moreover, monoglycosylated PrP proteins produced here have shown biochemical characteristics and cellular localization similar to that of wild type PrP.
Whereas we cannot rule out the possibility that the differences in localization are due to the point mutations introduced rather than differences in glycosylation, we suggest that the results presented here point to differences in glycosylation being the most likely determinant of cellular localization.
Some reports have described cytoplasmic PrP in rodent brains (44,46). A more recent study suggested that intracellular localization of PrP is probably due to retrotranslocation of the protein from the cell membrane to the proteasome system (57). Here we propose that the presence of a certain amount of intracellular PrP is physiological for neurons and probably is due to the functional folded unglycosylated isoform.
The observed intracellular accumulation of PrP in G3 mice will also assist in understanding the potential neurodegenerative role of intracellular PrP. It has been shown that accumulation of unglycosylated PrP can be neurotoxic for transgenic mice that develop a severe ataxia, with cerebellar degeneration and gliosis (58). This is not the case in our transgenics, since the mice with accumulation of intracellular PrP did not develop any type of neurodegeneration even after a prolonged period. However, our intracellular PrP is GPI-anchored, whereas that described previously (58) is soluble, thus suggesting that accumulation of unanchored PrP may lead to neurodegeneration. If this is the case, the absence of the anchor rather then the lack of sugars may be the real signal for PrP to be considered an unfolded protein that should be retained in the ER.
In summary, using gene targeted animal models, we have reported that mono-and unglycosylated PrP retains the same biochemical characteristics of diglycosylated PrP, suggesting no influence of sugars in determining spontaneous PrP C to PrP Sc transition. However, glycosylation can influence PrP cellular localization in neurons of the central nervous system, balancing its presence between the cytoplasm and the cell membrane. These observations lead us to suggest that PrP may only be partially dependent on glycosylation.
We have also reported here that the presence of intracellular PrP is a normal event for the cell determined by the unglycosylated form attached to membranes within the cell, and its accumulation is not apparently toxic to the cell.
We now aim, therefore, to use these models to define the role of PrP in the infectious process.