Functionally Relevant Domains of the Prion Protein Identified In Vivo

The prion consists essentially of PrPSc, a misfolded and aggregated conformer of the cellular protein PrPC. Whereas PrPC deficient mice are clinically healthy, expression of PrPC variants lacking its central domain (PrPΔCD), or of the PrP-related protein Dpl, induces lethal neurodegenerative syndromes which are repressed by full-length PrP. Here we tested the structural basis of these syndromes by grafting the amino terminus of PrPC (residues 1–134), or its central domain (residues 90–134), onto Dpl. Further, we constructed a soluble variant of the neurotoxic PrPΔCD mutant that lacks its glycosyl phosphatidyl inositol (GPI) membrane anchor. Each of these modifications abrogated the pathogenicity of Dpl and PrPΔCD in transgenic mice. The PrP-Dpl chimeric molecules, but not anchorless PrPΔCD, ameliorated the disease of mice expressing truncated PrP variants. We conclude that the amino proximal domain of PrP exerts a neurotrophic effect even when grafted onto a distantly related protein, and that GPI-linked membrane anchoring is necessary for both beneficial and deleterious effects of PrP and its variants.


Introduction
PrP Sc is the main constituent of prions [1], the infectious agents causing transmissible spongiform encephalopathies (TSE). PrP Sc is an aggregated and misfolded isoform of the cellular prion protein PrP C [2] which is expressed in a broad range of tissues of most vertebrates [3]. Nascent PrP C is exported to the lumen of the endoplasmic reticulum, deprived of its amino terminal signal sequence, glycosylated at two asparagine residues, and endowed with a GPI moiety which anchors it to the outer cell surface. Ablation of the Prnp gene, which encodes PrP C , abrogates prion replication [4] and toxicity [5]. Prnp o/o mice enjoy a normal life expectancy [6], but suffer from subtle neurological phenotypes [7] whose molecular basis has remained elusive [8].
Transgenic expression of amino proximally truncated PrP C mutants (PrP DCD , PrP DE and PrP DF , henceforth collectively termed DPrP) causes early-onset ataxia and white-matter degeneration (Fig. 1A). Toxicity appears to correlate with partial or complete deletions of the conserved PrP central domain (CD, residues 94-134) [9,10,11] which bridges the flexible amino proximal tail and the globular carboxy proximal domain [12].
Another neurotoxic phenotype was detected in compoundheterozygous Prnp o/ZHII mice and in homozygous Prnp ZHII/ZHII mice [13] whose Prnp ZHII allele leads to ectopic expression of the PrP Crelated protein Dpl [14,15,16,17]. Neuronal expression of Dpl in Tg(Dpl) or Tg(N-Dpl) mice induces ataxia within 40-60 days [18,19]. Despite 80% amino acid sequence dissimilarities [14], the overall 3D structure of Dpl is similar to that of PrP C (Fig. 1B) and includes an unstructured amino proximal tail, a globular three-helix domain [20], and a GPI anchor. However, Dpl is physiologically not expressed in the adult nervous system [21] and, importantly, lacks any sequences comparable to the CD. Therefore, Dpl resembles the neurotoxic DPrP mutants. What is more, the toxicity of both Dpl and DPrP is counteracted by co-expression of full-length PrP C [9,10,18,22,23], implying that it exploits common molecular pathways.
We reported previously that the removal of just the CD domain confers dramatic neurotoxicity to PrP. This suggests that the toxicity of Dpl may also result from the absence of a CD-like domain. Here, we tested this hypothesis by transgenic expression of two chimeric proteins, PrP_Dpl (residues 1-65 of Dpl replaced by residues 1-133 of PrP) and CD_Dpl (residues 90-133 of PrP inserted between residues 65 and 66 of Dpl). Transgenic mice expressing these proteins did not develop any clinical phenotypes. Additionally, coexpression of PrP_Dpl or of CD_Dpl ameliorated the clinical syndromes and prolonged the life expectancy of mice expressing neurotoxic DPrP mutants, in agreement with a previous report [24]. Since PrP is thought to be involved in signal transduction, we tested whether the toxicity of CD-deficient PrP mutants (PrP DCD ) may require localization to membrane lipid rafts. Indeed, removal of the GPI addition signal from PrP DCD prevents its neurotoxic effects.

Results
Transgenic mice expressing chimeric PrP-Dpl proteins and PrP DCDs All chimeric mutants of Dpl and PrP described here are based on the 'half-genomic' pPrPHG backbone [25] whose expression pattern has been recently studied in detail [26]. This construct contains a redacted murine Prnp gene which lacks intron #2 and is flanked by 6 and 2.2 kb of 59 and 39 genomic regions, respectively. Neuronal expression of Dpl leads to ataxia, neuronal loss and demyelinating neuropathy [17,18,19,22] while most of the toxicity of truncated PrP can be assigned to the lack of the central domain CD (residues 94-134) [10]. If the absence of a CD-like domain were responsible for its toxicity, addition of domains containing the CD region of PrP might detoxify Dpl.
We constructed CD_Dpl, a chimeric fusion protein consisting of codons 90-133 of mouse Prnp inserted between codons 65 and 66 of Prnd (Fig. 1A, F). This particular insertional position was chosen because hydrophobicity comparisons suggested that the resulting chimeric protein would resemble wild-type PrP (Fig. 1C-D). In a second construct termed PrP_Dpl, the amino terminus of PrP comprising codons 1-133 was fused to the carboxy terminus of Dpl comprising codons 66-179 ( Fig. 1A PrP and Dpl are tethered to the cell membrane by a C-terminal GPI anchor. PrP has been proposed to act as a signal transducer acting on various signaling pathways [9,10,27,28,29,30], and in this context it was speculated that PrP DCD toxicity may require membrane localization. To test this hypothesis, we introduced two point mutations at codons 232 and 233 (original mouse numbering) of the half-genomic construct PrP DCD [10], resulting in two in-frame stop codons. This prevents the translation of the carboxy terminal hydrophobic membrane anchoring domain of the precursor protein (see Fig. 1A), resulting in a secreted PrP mutant termed PrP DCDs . Because of the possible toxicity of the transgene, pronuclear injection was performed into hybrid B6D2F1 Prnp +/+ zygotes to generate PrP z=z DCDs (shorthand as above) transgenic mice. The latter mice were predicted to be viable due to the coexpression of wild-type PrP.
The levels of PrP DCDs in brains of both transgenic lines Tg40 and Tg42 was similar to that of Tg1046 PrP DCD [10] (Fig. 2B, C) and paralleled the measured amount of mRNA (Fig. S1). PrP DCDs showed a higher electrophoretic mobility than PrP DCD by 2-3 kDa, indicative of the missing GPI anchor.
Upon PNGase-F treatment, the complex banding pattern of PrP_Dpl, CD_Dpl, PrP C , PrP DCDs , and PrP DCD was reduced to one single band of lower molecular weight ( Fig. 2A-B), suggesting that these proteins were N-glycosylated. The strong reducing conditions prior to PNGase-F treatment prevented recognition of Dpl by anti-Dpl antibody (data not shown). suggesting that this antibody recognizes a discontinuous C-terminal epitope destroyed by reduction of the two disulfide bridges of Dpl. Milder pretreatment resulted in partial deglycosylation of Dpl ( Fig. 2A, arrowhead); under these conditions CD_Dpl extracts gave rise to two additional bands, which may indicate posttranslational cleavage ( Fig. 2A, arrowhead). PrP_Dpl extracts did not show this phenomenon.
We then prepared detergent-resistant membranes (DRMs) from wild-type, PrP_Dpl and CD_Dpl, (Fig. 2D), PrP DCDs, anchorless PrP s , and PrP DCD brains (Fig. 2E) in the presence or absence of PrP C . The buoyancy of PrP_Dpl, CD_Dpl, and PrP DCD was similar to that of PrP C and flotillin ( Fig. 2D-E), suggesting that  they all reside in similar membrane microdomains. Therefore, most aspects of PrP_Dpl and CD_Dpl biogenesis appear to be similar to those of PrP C . In contrast, both PrP DCDs and PrP s displayed less buoyancy, suggesting no association with rafts in agreement with their biogenesis as soluble proteins. We then prepared DRMs from Tg42 PrP z=o DCDs mice coexpressing PrP C and PrP DCDs . Fractions were deglycosylated with PNGase F prior to western blotting. This experiment revealed that coexpression of wild-type PrP fails to recruit PrP DCDs to DRMs. Upon pretreatment with phosphatidylinositol-specific phospholipase C (PI-PLC) the buoyancy of the GPI-anchored PrP variants became similar to that of their anchorless counterparts (Fig. S2).
Finally, we determined the serum PrP concentration in PrP wt , PrP DCD , and PrP DCDs mice, as well as in GPI-Tg44 mice expressing anchorless full-length PrP s [32] (Fig. 2F). Despite similar PrP levels in brain homogenates, mice expressing anchorless versions of PrP (PrP s or PrP DCDs ) displayed up to 4fold higher serum levels. Therefore, PrP DCDs underwent normal maturation and glycosylation but was predominantly secreted, similarly to PrP s . , and monitored using a four-degree clinical score [10]. It has previously been shown that onset and development of disease correlate with expression levels of Dpl. Tg(Dpl)28272/ZrchI and (TgN-Dpl)32 mice, which express high amounts of Dpl, survived only 32 and 60 days respectively [18,19] whereas mice expressing lower Dpl levels, such as Prnp Ngsk/Ngsk mice [16], showed progressive symptoms of ataxia and were euthanized according to clinical scoring at $70 weeks of age. Instead, none of the PrP  (Fig. 3C) and wt mice (Fig. 3B). White matter pathology characterized by vacuolation and astrogliosis was seen in the cerebellum (arrows Fig. 3E) and in the corpus callosum of Prnp Ngsk/Ngsk mice (Fig. 3I). None of these changes were observed in wt, Tg1026 PrP   Table 2) as described [9]. Double transgenic TgF356Tg1071 PrP  (CGC) loss (Fig. 5H). Milder white-matter changes and much less severe CGC loss were observed in compound Tg10466Tg1071 PrP o=o DCD CD Dpl and Tg10466Tg1026 PrP o=o DCD PrP Dpl littermates euthanized at the same age (Fig. 5F-G and 5I-J). Western blot analysis of brain homogenates indicated that expression levels of the various transgenic proteins were unchanged in the compound transgenic mice independently of the respective combination. The steady-state levels of CD_Dpl exceeded those of PrP DCD PrP DF and PrP wt (Fig. 5K, M), whereas those of PrP_Dpl and PrP DCD were similar and much lower than those of PrP DF (Fig. 5L, N). Although expression of CD_Dpl was higher than that of PrP_Dpl, and compound PrP   [10]. We therefore conclude that removal of the lipid anchor from PrP DCD completely abolishes its neurotoxic properties.

Discussion
The results presented here confirm and extend a recent report that fusion of the complete amino-terminus of PrP detoxifies Dpl.  Tg(PrPN-Dpl) mice expressing a fusion protein consisting of amino acids 1-124 of PrP and amino acids 58-179 of Dpl failed to show Dpl typical neurological disorder and were able to prolong the onset of ataxia in mice with exogenous Dpl expression [24]. By generating chimeric proteins that contain either the entire aminoterminus of PrP linked to the carboxy-terminus of Dpl (PrP_Dpl) or the central domain of PrP alone (CD_Dpl), we found specific domains within the amino-terminus of PrP that are involved in the detoxification of Dpl in two distinct brain regions and cell types.
While PrP_Dpl showed no signs of cerebellar granule cell degeneration for at least 60 weeks, PrP o=o CD Dpl mice displayed mild astrogliosis within the CGC layer. This may point to some residual neurotoxicity of CD_Dpl. In contrast, white matter degeneration was observed in Dpl-expressing Ngsk mice yet was not seen in mice expressing either of the two transgenes, PrP_Dpl and CD_Dpl. Since leukoencephalopathy is the major life-shortening pathology associated with expression of truncated PrP and Dpl [10,33], both addition of the whole amino-terminus, or addition of the central domain alone resulted in a normal life expectancy in transgenic mice.
In addition to detoxifying Dpl, chimeric fusion proteins were able to partially antagonize the toxic effects of the PrP deletion mutants PrP DF and PrP DCD . While PrP_Dpl was able to antagonize cerebellar granule cell loss in PrP DF mice, CD_Dpl was not. Cerebellar white matter gliosis was milder in both PrP o=o DF CD Dpl and PrP o=o DF PrP Dpl mice. This lends further support to the conclusion that distinct domains within PrP exert neurotrophic functions in a variety of brain regions and cell types.
We have excluded that differences in expression level were responsible for the observed effects: Western blotting with antibody POM3 [31], which recognizes a domain common to both transgenes, showed a higher expression for CD_Dpl than for PrP_Dpl. All transgenic constructs were expressed using the same backbone, thereby reducing the likelihood of differential expression in distinct cell types. Thus the cell-specific effects of the different transgenes appear to be related to their structural features rather than to the levels or tissue-specific patterns of their expression.
Despite sequence homologies of ,20%, the carboxy terminal domains of Dpl and PrP have very similar folding patterns of the respective carboxy proximal regions, whereas their amino proximal portions are much less structured [20,34,35]. Hence the selective permutations of the less structured domains of the two proteins performed here are not very likely to alter the overall global fold of the resulting fusion proteins. We found that both PrP_Dpl and CD_Dpl underwent correct intracellular sorting and posttranslational processing ( Fig. 2A, C). Furthermore, in none of the transgenic mice (including the lines expressing the highest levels of transgene) did we detect any spontaneous formation of PK-resistant transgenic protein or PrP aggregates by Western blotting and histology (data not shown).
Further evidence for specific differences in the function of PrP C comes from the previous studies on transgenic mice expressing PrP C in a cell-type specific manner. While cerebellar granule cell loss in PrP DF mice was reversed by neuronal expression of PrP, white matter degeneration was rescued by myelin-specific expression of PrP [36].
Cell-specific requirements for distinct PrP domains might explain the discrepancies regarding the domains reported to be involved in cytotrophic functions. Several studies suggest that the octapeptide repeat region is crucially linked to the neuroprotective functions of PrP C [37,38,39]. On the other hand, a feature common to all the toxic PrP deletion mutants is the lack of the central domain (encompassing at least residues 105-125) within PrP C . This in turn points to a role of the central domain of PrP C .
The results presented here may help clarifying this controversy. The central domain (aa 94-134) appears to be crucial for myelin maintenance, while other domains within the amino terminus (aa 23-94) may be required for neuroprotection. Residues 23-94 consists of the amino-terminal charged cluster (aa 23-28) involved in endocytosis and of the octapeptide repeat region associated with neuroprotection via anti-oxidative function and copper binding (aa 50-90) [37,39,40]. It was initially reported that amino acids 23-88 are needed to fully suppress neurotoxicity on Purkinje cells [41], yet it was later shown that the octapeptide repeats are dispensable for this function. This suggests that the charge cluster may be more relevevant for the neuroprotection of Purkinje cells [24] that for other cell types. It is less likely that toxic domains within the amino-terminus of Dpl in CD_Dpl may be responsible for the observed residual neurotoxicity, since earlier studies showed that the proximate cause of cerebellar granule cell degeneration is not the amino terminus of Dpl, but rather its carboxy terminus [42].
PrP C was reported to inhibit the NR2D subunits of the NMDA receptor complex, and Prnp o/o hippocampal neurons display increased neuronal excitability and enhanced glutamate excitotoxicity [43]. It will be interesting to study whether chimeric PrP/ Dpl proteins exert PrP C like functional regulation of the NMDA receptor and whether central domain, octapeptide repeat region or amino-terminal charged cluster are involved in this function. It was suggested that homodimerization of PrP C mediates the transduction of extracellular signals [44,45,46]. The toxicity of truncated PrP and Dpl is counteracted by overexpression of fulllength PrP C [9,10,18,19] and exacerbated by removal of the endogenous Prnp gene, suggesting that PrP C and its variants compete for a common interacting molecule. The PrP/Dpl fusion proteins appear to partake in this competition as well, as both CD_Dpl and PrP_Dpl prolonged survival of PrP o=o DCD and PrP o=o DF mice. Perhaps the CD region is responsible for stringent proteinprotein interactions, whereas the structured carboxy termini of PrP and Dpl allow for more relaxed interactions and are therefore interchangeable. Such interactions might also include the formation of functionally relevant homodimers or homooligomers [47]. The residues 113-128 of PrP mediate interaction of PrP with stress inducible protein 1 (STI) [48] and heparan sulfate [49]. The incompleteness of the rescue in all tested paradigms of PrP In addition to the findings described above, we extended our analysis of functional domains within PrP to those determining the localization of the protein. Mice expressing anchorless PrP accumulate high titers of prions and protease-resistant PrP when challenged with scrapie [32,50], yet develop only subtle pathologies [51]. Here, anchorless PrP DCDs was expressed to high levels in transgenic mice, and was very efficiently secreted into the extracellular space of brain and in serum as a mature, fully glycosylated soluble form [52]. Although the deletion within PrP DCDs was identical to that of the neurotoxic membrane anchored PrP DCD , it did not induce any pathology in transgenic mice, irrespectively of the presence or absence of full-length PrP C . Since the total concentration of PrP DCDs in brain homogenates was as high as that of PrP DCD , and even higher than that of PrP DCD in the serum, lack of toxicity was unrelated to its expression level. Also, PrP DCDs failed to influence the survival of PrP DCD mice coexpressing PrP C , confirming that it exerts neither beneficial nor detrimental effects on the central nervous system.
PrP DCDs did not localize to detergent-resistant membrane (DRM) fractions, even when wild-type PrP C was coexpressed. This observation suggests that the genetic interaction between PrP C and its neurotoxic variants may physically necessitate membrane anchoring of all relevant partners. In contrast, soluble-dimeric prion protein (PrP-Fc 2 ) was found to translocate to the DRM compartment and to associate with PrP Sc upon prion infection of mice coexpressing PrP C and PrP-Fc 2 [51]. In this context, it may be of interest to study the localization of PrP DCDs in prion infected mice.
In conclusion, the above findings indicate that (1) the amino proximal domain of PrP contains minimal elements that are necessary and sufficient for PrP function, that (2) distinct domains within the amino-terminus of PrP exert site-and/or cell-specific functions, and that (3) GPI membrane anchoring is mandatory for exerting said function. The understanding of the physiological and pathophysiological functions of the prion protein will benefit from functional analyses of the proteinaceous [48] and non proteinaceous [49] constituents interacting with PrP and its variants. Finally, it will be of particular interest to explore whether the phenomena studied here share functional and molecular aspects with the neurotoxicity observed in prion diseases [53].

Ethics Statement
All mice were maintained under specific pathogen-free (SPF) conditions. Housing and experimental protocols were in accor-dance with the Swiss Animal Protection Law and in compliance with the regulations of the Veterinaeramt, Kanton Zurich.

Construction of the transgenes
The coding region of murine Prnp and Prnd gene were analyzed using DNAMAN software (Lynnon BioSoft, Canada), and hydrophobicity plots were generated using a window of 9 amino acid residues. The regions identified in these plots were used to define the CC, CD and HC domains. The chimeric fusion proteins of PrP and Dpl were designed such that their hydrophobicity characteristics would mimic that of wild-type PrP. Based on pPrPHG [25], a PmeI/NheI fragment was subcloned in the pMECA [54] backbone. To create the CD_Dpl cDNA, mouse genomic cDNA was used as template to obtain two PCR fragments with primer sets JP1 The two initial products were fused in a third PCR with the flanking primers JP1 and JP4. This product was digested with NsiI and ligated to the NsiI sites of the pMECA vector containing the pPrPHG subcloned PmeI/NheI sequence into which a second NsiI site had been engineered. After confirming insertion with the correct orientation, the insert was cloned back into the pPrPHG backbone using the PmeI/NheI sites.
PrP_Dpl was created based on the plasmid pPrPHG [25]. A fragment (480 bp) was amplified using the primers pE2* (59-CAA CCG AGC TGA AGC ATT CTG CCT)/X2 (59-CCT GCT CAC GGC GCT CCC CAG CAT G) containing sequence information from Exon3 to codon 132/133 of the murine PrP. In a second PCR using genomic DNA as template and primers X3 (59-GGG AGC GCC GAC ATC GAC)/X4 (59-AAA GAA TTC CAC AAT TCT TAC TTC ACA ATG) a fragment (360 bp) containing codon 68 until polyadenylation site of Dpl was amplified. After purification both fragments were cut with HaeII mixed and directly ligated into the pCR-Blunt II-Topo vector. The transgene was then excised with AgeI/EcoRI and, after blunting the 39 EcoRI sites, ligated into the original AgeI/BbrPI site of pPrPHG. The presence of the new insert was confirmed by restriction analysis using SmaI.
PrP DCDs was generated using the pMECA PmeI/NheI subclone pPrPHG previously described [10]. The oligonucleotide primers dCDSol59 (59-CCT ATT ACG ACG GGA GAA GAT CCT GAT GAA CCG TGC TTT TCT CCT CC-39) dCDSol39 (59-GGA GGA GAA AAG CAC GGT TCA TCA GGA TCT TCT CCC GTC GTA ATA GG-39), each complementary to opposite strands of the vector, were extended during temperature cycling by PfuTurboH DNA polymerase. On incorporation of the oligonucleotide primers, a mutated plasmid containing staggered nicks was generated. After temperature cycling and treatment with DpnI to digest the parental DNA template and select for the desired DNA construct, the nicked vector DNA incorporating the mutations was transformed into E. coli. Clones were picked and sequenced. Finally the PmeI/NheI fragment containing the desired point mutation was religated into the pPrPHG vector as described before [10].

Generation, Identification, and Maintenance of Transgenic Mice
The pPrPHG plasmids containing the PrP or Dpl coding sequences were propagated in E. coli XL1 blue, the minigene excised with NotI and SalI, and processed as described [25]. Pronuclear injections into fertilized oocytes were carried out as described [55]. Transgenes on a Prnp o/o background were identified by PCR using the exon 2 primer pE2* (59-CAA CCG AGC TGA AGC ATT CTG CCT) and the exon 3 primer Ubl floxed Dpl (59-CTC GCT GGT GGA GCT TGC TAT C) resulting in a PCR product of 618 bp for CD_Dpl and 670 bp for PrP_Dpl or pE2* and exon 3 primer Mut217 (59-CCT GGG ACT CCT TCT GGT ACC GGG TGA CGC) resulting in a PCR product of 619 bp. PCR analysis in order to verify the outbreeding of the Prnp + allele was carried out using primers P10 (Prnp exon 3, 59-GTA CCC ATA ATC AGT GGA ACA AGC CCA GC), 39NC (non-coding region at 39 of exon 3, 59-CCC TCC CCC AGC CTA GAC CAC GA), and P3 (neoR gene, 59-ATT CGC AGC GCA TCG CCT TCT ATC GCC); P10 and 39NC gave an 560 bp signal for the Prnp + allele, and P3 and 39NC gave a 362 bp product for the Prnp 0 allele. Alternatively, to test for the presence or absence of the Prnp + allele an additional PCR was performed using primers P2 (Prnp int 2, 59-ATA CTG GGC ACT GAT ACC TTG TTC CTC AT) and P10rev (reverse complementary of P10 59-GCT GGG CTT GTT CCA CTG ATT ATG GGT AC) giving a product of 352 bp for the Prnp + allele. In order to distinguish between transgenic mice expressing PrP DCD and PrP DCDs , two separate PCR reactions were performed using primers pE2* and pdCDrev (59-GGA GGA GAA AAG CAC GGT GCT GCT) yielding a diagnostic amplicon of 666 bp, or using pE2* and pdCDsrev (59-GGA GGA GAA AAG CAC GGT TCA TCA) yielding a diagnostic amplicon of 666 bp.

Q-PCR to determine genomic copy numbers
Total genomic DNA was prepared from mouse tails after PK digestion and purified according to standard procedures. Copy numbers were assessed by Taqman PCR using 2 ng of total genomic DNA and primer pairs CD Sonde59 (59-GGA GGG GGT ACC CAT AAT) and CD Sonde39 (59-GCG CTC CCC AGC ATG TAG) on C57Bl6, Tga20, Prnp o/o , Tg1025, Tg1026 and Tg1071 mice. For determination of copy numbers of Tg40, Tg42 primer pairs p60 (59-CGC TAC CCT AAC CAA GTG T) and p61 (59-GAT CTT CTC CCG TCG TAA T) were used. To standardize Taqman PCR on GAPDH using primers GAPDH up (59-CCA CCC CAG CAA GGA GAC T) and GAPDH down (59-GAA ATT GTG AGG GAG ATG CT) was done in parallel.

mRNA analysis
Total brain RNA was isolated in Trizol (Life Technologies), purified and DNase treated according to the manufacturer's manual (Roche). After reverse transcription (Geneamp; Roche) cDNA was used for Taqman PCR using primer pairs Dpl Taq59 (59-CTA CGC GGC TAA CTA TTG)/Dpl Taq39 (59-CGC CGG TTG GTC CAC) and PrP Taq59 (59-CAG TGG AAC AAG CCC AGC)/PrP Taq39 (59-CCC CAG CAT GTA GCC ACC). To standardize expression levels GAPDH using primers GAPDH up (59-CCA CCC CAG CAA GGA GAC T) and GAPDH down (59-GAA ATT GTG AGG GAG ATG CT) and 18S rRNA using primers 18S fw (59-GTA ACC CGT TGA ACC CCA TT) and 18S rc (59-CCA TCC AAT CGG TAG TAG CG) were used. Taqman PCR using SYBR-green (Roche) and determination of DDCT-values were done on a Applied Biosystems 7900 device. As control for possible DNA contamination, DNase-treated RNA from wt and tg mice that had not been reversely transcribed was used.

Flotation assays
Flotation of detergent insoluble complexes was performed as described [56]. Appropriate brain homogenates were extracted for 2 h on ice in cold lysis buffer (150 mM NaCl, 25 mM Tris-HCl, pH 7.5, 5 mM EDTA, 1% Triton X-100; total protein: 1 mg in 1.6 ml. Extracts were mixed with two volumes (3.2 ml) of 60% OptiprepH (Nycomed) to reach a final concentration of 40%. All lysates were loaded at the bottom of Beckman ultracentrifuge tubes. A 5-30% OptiprepH step gradient in TNE (150 mM NaCl, 25 mM Tris-HCl, pH 7.5, 5 mM EDTA) was then overlaid onto the lysate (8.4 ml of 30% OptiprepH and 3.6 ml of 5% OptiprepH). Tubes were centrifuged for 24 h at 4uC in a TLS55 Beckman rotor at 100,000 g. Fractions (1 ml) were collected from the top of the tube and processed for immunoblotting and visualization with anti-PrP antibody POM3 [31], anti-flotillin 1, and anti-GAPDH antibody (both BD Transduction Laboratories). In order to release GPI anchored proteins from membranes, brain homogenates were treated for 2 h at 37uC with 10 U/ml Phospholipase C (PI-PLC from Sigma) as described [57].

ELISA
PrP ELISA was performed as described in [58] 96-well plates (Nunc-Immuno Maxisorb; prod. no. 439454) were coated with 50 mL per well of POM1 (2 mg/ml, 1:5000 in 0.1 M sodium carbonate buffer pH 9.6 [1.58 g Na 2 CO 3 +2.94 g NaHCO 3 in 500 ml H 2 O]) over night at 4uC. All following incubation steps were made at room temperature. The plates were washed by immersing them 4-5 times in PBS with 0.1% Tween-20 (PBST). Plates were then incubated with 100 mL per well of blocking buffer (5% Top-Block in PBST) for two hours. A 1:3 dilution of recombinant murine PrP (rmPrP) (starting from 50 ng/ml) was used for a standard curve. Blood plasma from respective mice was diluted appropriately in sample buffer (1% Top-Block in PBST) and incubated for 1 h. Then, plates were washed 4-5 times in PBST and incubated with biotin-labeled POM2 (1 mg/ml, 1:5000 in sample buffer, 100 mL per well) for 1 h. Plates were washed 4-5 times and incubated with avidin-HRP (1 mg/ml, 1:1000 in sample buffer, 100 mL per well) for 1 h followed by another round of washing, 4-5 times in PBST and 2-3 times with PBS alone. Chromogenic substrate (Biosource, prod. no. SB02, 50 mL per well) was applied for up to 10 min. The reaction was stopped with 0.5 M H 2 SO 4 and absorbance was read at 450 nm.

Clinical scoring and observation
Mice were examined once weekly for clinical signs as described previously [10]. Mice were euthanized when they reached a score of 3.5 or higher. Statistical significance was assessed as indicated.

Morphological analyses
Brains, spinal cords and sciatic nerves were removed and fixed in 4% formaldehyde in PBS, pH 7.5, paraffin embedded, and cut into 2-4 mm sections. Sections were stained with hematoxylineosin (H&E), Luxol-Nissl (myelin and neurons), and commercial antibodies to GFAP (glial fibrillary acidic protein; activated astrocytes), MBP (myelin basic protein), NF200 (neurofilament 200), IBA1 (microglia) and SAF84 (PrP Sc aggregates). For semithin sections and electron microscopy mice were perfused with ice-cold 4% PFA/3.9% glutaraldehyde. Spinal cord tissues were removed, immersed in the same solutions, and kept in Phosphate buffer at 4uC until processing. Tissues were embedded in Epon, and semithin sections were stained with toluidine blue and paraphenylene diamine. Frozen sections for POM3 and Dpl staining were blocked with M.O.M Mouse IgG Blocking Reagent (Vector Laboratories) stained with anti Dpl GX-2D10-B1 (Dpl) or POM3 (soluble cellular PrP). Detection was achieved using both Goat anti Mouse AP and Donkey anti Goat AP (Jackson) with alkaline phosphatase fast red. Figure S1 Characterization of transgenic mice (A) Gene copy numbers per haploid genome in transgenic lines as determined by genomic Q-PCR. (B) relative mRNA level in brain extracts of transgenes compared to PrP mRNA in C57BL/6 mice (filled black columns and left y-axis) and compared Dpl mRNA in Prnp Ngsk/Ngsk mice (open columns and right y-axis) using either PrP or Dpl specific primer sets for Q-PCR. Each column represents the average of 3 mice. Found at: doi:10.1371/journal.pone.0006707.s001 (0.55 MB TIF) Figure S2 Characterization of membrane anchored and PI-PLC treated transgenic proteins. Density gradient DRM preparations of wild-type, PrP GPI anchorless (PrP s ), PrP DCD and PrP DCDs transgenic brains analyzed after PI-PLC treatment and deglycosylation with PNGase F with monoclonal antibody POM1. After PI-PLC treatment PrP and PrP DCD had similarly buoyancy like PrPs and PrP DCDs whereas flotillin a non GPI-anchored DRM associated protein still was found in fractions with higher buoyancy indicating the intactness of the DRMs. Found at: doi:10.1371/journal.pone.0006707.s002 (0.84 MB TIF)