Sequence-dependent Prion Protein Misfolding and Neurotoxicity*

Prion diseases are neurodegenerative disorders caused by misfolding of the normal prion protein (PrP) into a pathogenic “scrapie” conformation. To better understand the cellular and molecular mechanisms that govern the conformational changes (conversion) of PrP, we compared the dynamics of PrP from mammals susceptible (hamster and mouse) and resistant (rabbit) to prion diseases in transgenic flies. We recently showed that hamster PrP induces spongiform degeneration and accumulates into highly aggregated, scrapie-like conformers in transgenic flies. We show now that rabbit PrP does not induce spongiform degeneration and does not convert into scrapie-like conformers. Surprisingly, mouse PrP induces weak neurodegeneration and accumulates small amounts of scrapie-like conformers. Thus, the expression of three highly conserved mammalian prion proteins in transgenic flies uncovered prominent differences in their conformational dynamics. How these properties are encoded in the amino acid sequence remains to be elucidated.

Genetic and biochemical evidence indicates that the prion protein (PrP) 3 is the causative agent in the pathogenesis of prion diseases (1). The pioneers of prion biology successfully transmitted prions to many laboratory animals, including mice, rats, hamsters, and guinea pigs (2,3). However, rabbits proved resistant to different prion strains from humans and sheep and from mice-adapted scrapie sheep prions. One interpretation of these results is that the cellular environment of the rabbit may lack a conversion factor or express conversion inhibitors that prevent the acquisition of neurotoxic conformers. However, expression of recombinant PrP from sheep and rodents (mouse and bank vole) in rabbit RK13 epithelial cells followed by challenge with autologous prions led to persistent proteinase K-resistant PrP (PrP res or PrP Sc ) replication, indicating that rabbit cells provide a molecular environment consistent with PrP Sc conversion (4,5). Alternatively, key amino acid substitutions in the sequence of rabbit PrP may impose structural constraints that prevent its conversion. To answer these questions Priola and co-workers (6) expressed rabbit PrP (RaPrP) in prion-infected mouse neuroblastoma cells and showed that RaPrP could not be converted into PrP Sc , indicating that the amino acid sequence of RaPrP prevents its conformational conversion. Thus, understanding the conformational properties of RaPrP can contribute to unraveling the rules governing prion transmission and neuropathology.
PrP is a membrane-anchored glycoprotein highly enriched in the brain with the ability to undergo conformational changes. PrP Sc has been postulated as the causative agent in prion diseases, and it is composed of specific folding species of PrP that are replicated by autocatalytic mechanisms (7). However, the structure of PrP Sc has not yet been resolved. Structural studies performed with either purified or recombinant PrP C have confirmed the intrinsic ability of PrP to misfold into conformations that are transmissible and induce neurotoxicity (8,9). These in vitro analyses of PrP structure have mostly focused on understanding the unfolding/misfolding dynamics of the globular domain, while still accepting the potential contribution of the unstructured N-terminal domain (10). Recent studies have identified clear differences in the misfolding of purified hamster and mouse PrP, providing new insight into the structural basis of the species barrier phenomenon (11).
Here, we studied the structural properties and neurotoxic potential of wild type PrP from Syrian Golden hamster (HaPrP), mouse (MoPrP), and rabbit (RaPrP) in transgenic flies. This is the first time that these three proteins have been compared simultaneously in the same system, making their similitudes and differences highly relevant. Moreover, these experiments are carried out in transgenic animals expressing full-length PrP, providing a rich in vivo context to examine the properties of these proteins. Our studies confirmed that RaPrP does not convert into pathogenic conformations in vivo and does not induce neurotoxicity. More surprisingly, MoPrP produced mixed results, inducing early locomotor dysfunction but not spongiform degeneration or PrP aggregation, two characteristics of HaPrP. Thus, we have uncovered unsuspected conformational dynamics for MoPrP, indicating the sensitivity of the transgenic fly model to the subtle sequence differences between hamster and mouse PrP.
Generation of RaPrP Transgenic Flies-The open reading frame of the rabbit Prnp gene was isolated by PCR amplification from genomic DNA. EcoRI and NotI restriction sites were included in the primers (5Ј-GAATTCATCATGGCGCAC-CTCGGCTACTGG-3Ј and 5Ј-GCGGCCGCTCATCCCAC-GATCAGGAAG-3Ј) to facilitate cloning into the Drosophila pUAST vector (19). The resulting construct (UAS-RaPrP) was injected into yw embryos, and several single insertion lines were created by standard procedures (20).
Quantitative RT-PCR-To quantify the levels of PrP transcripts expressed from hamster, mouse, or rabbit PrP transgenes, we performed real time RT-PCR assays. Total RNA (TRIzol, Invitrogen) was isolated from five whole adult flies expressing PrP ubiquitously under the control of da-Gal4, and DNA traces were eliminated with turbo DNase (Ambion). Real time PCRs were done by amplifying each transcript with independent primers and using the same TaqMan probe for all the reactions, 6-carboxyfluorescein (FAM)-CGTGGTGGAGCA-GAT. For rabbit, an amplicon of 75 bp was detected with primers AGAACTTCACCGAGACCGACAT and CCTGCTGGT-ACTGCGTGATG. The 73-bp product from hamster was generated with primers TTCACGGAGACCGACATCAA and ACTCCTTCTGATACTGGGTGGTACA. The 80-bp mouse Prnp amplicon was generated with primers GAGACCGATG-TGAAGATGATGGA and TAATAGGCCTGGGACTCCTT-CTG. All PCR assays were run in the ABI PRISM 7000 sequence detection system (Applied Biosystems) in triplicate using standard conditions, and the relative amounts of mRNAs were calculated by amplifying RNA polymerase II mRNA in the same reactions. The HaPrP-M9 line was also used as a reference for a moderate line that induces very weak degeneration. Plotted values were obtained from three independent reactions and normalized against the strong HaPrP-M6 line.
Sequence Alignment-The alignment of hamster, mouse, and rabbit protein sequences was done using ClustalW2. Amino acid sequences for the three species were obtained from NCBI with the following accession numbers: B34759 (Syrian hamster), AAA39996 (mouse), and AAD01554 (rabbit). Alignment was done with the mature C-terminal sequences containing the globular domain of each protein, which spans amino acid positions 121-230 of MoPrP. The color-coded amino acids indicate properties relevant for protein structure (size and charge).
Histology-Flies expressing PrP throughout the brain under the control of the ubiquitous da-GAL4, along with control flies expressing LacZ, were collected at 1 and 30 days after eclosion. Plastic embedding was prepared as described previously (21), and semithin sections were cut at 1 m and stained with toluidine blue. Semithin sections were imaged in a Nikon 80i microscope equipped with a Zeiss Axiocam color camera, and the images were collected with the Axiovision software from Zeiss.
Immunofluorescence-For mushroom body analysis, we collected flies expressing PrP or LacZ under the control of OK107-Gal4 at days 1 and 40. The subcellular distribution of PrP was characterized by expressing the PrP constructs in motor neurons in the larval wall (BG380-Gal4) or by co-expression of PrP and GFP in the ventral ganglion of the larval brain (OK107-Gal4) for imaging of single neurons. Whole-mount immunohistochemistry of adult brains and larval tissues was conducted as described previously (21) using anti-HaPrP (3F4, 1:1,500, Millipore), anti-MoPrP/RaPrP (6H4, 1:2,000, Prionics), and anti-HRP (1:200, GenScript) primary antibodies and the antimouse-Cy3 (1:600, Molecular Probes) and anti-rabbit-FITC (1:600, Sigma) secondary antibodies. For membrane detection of PrP, we followed the same procedures, except that no detergent was used. Fluorescent images were collected in an LSM510 confocal microscope (Zeiss). Cell surface occupied by Kenyon cell clusters was measured by outlining the surface by hand in Photoshop from 10 to 20 independent confocal images from 1and 40-day-old flies. The data were exported to Excel for plotting and statistical analysis. The three-dimensional image of the mushroom body was reconstructed using Zeiss Zen software from a Z-stack of 35 images covering the whole brain with a Z-step of 3 m.
Locomotor Assays-PrP and control constructs were expressed in motor neurons using the BG380-Gal4 driver. The progeny was collected in 8-h intervals and then subjected to climbing assays (22). Briefly, 20 adult flies were placed in empty vials and forced to the bottom by firmly tapping against the surface. After 8 s, the number of flies above 5 cm was recorded. This was repeated eight times every 2 days for 30 days. Climbing ability was plotted as a function of age in Excel. For softwareassisted analysis of fly locomotion, we video recorded the climbing for 20 s, and the videos were downloaded in the computer and analyzed simultaneously with GroupScan (Cleveristics). Four experimental arenas were created to cover the surface of the vials (except the bottom and the stopper) and were analyzed simultaneously. Flies were considered "active" if they moved at least 0.5 mm/s. Speed per active fly was calculated every frame (1 s ϭ 30 frames) and was averaged for the 20 s of the video. At least 20 flies per genotype were analyzed in triplicate. Data were exported to Excel for t test statistical analysis.
Tissue Homogenates and Western Blot-Depending on the experiment, 1-20 whole flies expressing the PrP constructs ubiquitously under the control of da-Gal4 were homogenized in 20 -50 l of PBS containing 150 mM NaCl, 1% Triton X-100, 4 mM EDTA, and Complete protease inhibitors (Roche Applied Science) using a motorized pestle. Complete homogenates or subfractions from other experiments (see below) were resolved by SDS-PAGE under reducing conditions, electroblotted into nitrocellulose membranes, and probed against the 3F4 and 6H4 anti-PrP antibodies. The anti-␣-tubulin (1:200,000, Sigma) antibody was also used in all membranes as a loading control.
NaPTA Insolubility-We generated lysates from whole flies expressing each PrP construct ubiquitously (da-Gal4) at days 1 or 30. Homogenates were incubated with 10% Sarkosyl for 30 min at 4°C and then treated consecutively with 5 units of benzonase and 0.3% NaPTA for 30 min each at 37°C, followed by centrifugation at 16,000 ϫ g for 30 min as described previously (23,24). The presence of PrP in the soluble (supernatant) and insoluble (pellet) fractions, as well as an equivalent aliquot of the total fraction, were analyzed by Western blot.
Immunoprecipitation Assay-PrP Sc -specific conformations were detected in fly brain extracts using the 15B3 immunoprecipitation kit following the procedure provided by Prionics AG (Zurich, Switzerland) (25). Briefly, two PrP-expressing flies were homogenized in 40 l of 15B3 homogenization buffer and mixed with 450 l of 15B3 IP buffer and 10 l of rat anti-mouse IgM Dynabeads (Invitrogen) coated with mAb 15B3. After 2 h of incubation with gentle rotation at 25°C, Dynabeads were washed three times and boiled in loading buffer. Immunopre-cipitated proteins were detected by Western blot with a mixture of 3F4 and 6H4 antibodies against the same membrane.
Velocity Sedimentation in Sucrose Gradients-Tissue homogenates from 10 whole flies were incubated with an equal volume of 2% Sarkosyl for 30 min on ice. The samples were loaded atop 10 -60% step sucrose gradients and centrifuged at 200,000 ϫ g in the SW55 rotor (Beckman Coulter) for 1 h at 4°C as described previously (26). After centrifugation, 12 fractions were sequentially removed from the top. Aliquots of the 12 fractions were subjected to Western blot analysis with 6H4 and 6D11 antibodies against PrP.
Size Exclusion Chromatography-A 1 ϫ 30-cm column of Superdex 200 HR beads (GE Healthcare) was used to determine the aggregated state of PrP molecules. Chromatography was performed in an FPLC system (GE Healthcare), as described previously (26), by injecting 200 l of fly homogenates in the column with a flow rate of 0.25 ml/min and collecting fractions of 0.25 ml each. Aliquots from every other sample were subjected to Western blot to detect PrP with 6H4 and 6D11 antibodies. The molecular weight of the PrP species recovered was evaluated according to a calibration curve generated with molecular mass markers (Sigma), including dextran blue (2,000 kDa), thyroglobulin (669 kDa), apoferritin (443 kDa), ␤-amylase (200 kDa), alcohol dehydrogenase (150 kDa), albumin (66 kDa), and carbonic anhydrase (29 kDa).

RESULTS
Conservation of Mammalian PrP-PrPs are highly conserved mammalian proteins. The overall sequence conservation of hamster, mouse, and RaPrP PrP is 85% identity and 96% similarity for the three proteins. The introduction of amino acid substitutions from RaPrP into mouse MoPrP demonstrated that there are critical positions for PrP conversion throughout the whole protein (6). However, recent structural studies indicate that the globular domain on the C terminus is the most critical for conformational conversion (10). The globular domain encompassing amino acids 122-232 in HaPrP shows very high conservation between hamster and mouse PrP (94% identity/98% similarity), although the conservation between RaPrP and the two rodents is slightly lower (Mo/Ra, 87% identity/98% similarity; Ha/Ra, 87% identity/95% similarity) (Fig.  1A). Some dramatic amino acid differences are known to have The ␤-sheet domains (blue) and ␣-helices (red) are indicated. Color coding of amino acids is as follows: red, small and/or hydrophobic; blue, acidic; magenta, basic; green, hydrophilic, charged. L42 indicates a RaPrP-specific epitope recognized by the L42 antibody that has no effect when introduced in MoPrP. The red box in MoV214 indicates a position key for MoPrP conversion with a conservative change in RaPrP and a radical change in HaPrP. B, detection of hamster, mouse, and rabbit PrP expressed in transgenic flies by Western blot. A mix of 3F4 and 6H4 antibodies was used for detection of the three proteins in the same membrane. HaPrP and MoPrP accumulate in two bands, whereas RaPrP accumulates in three, the top two being the more prominent. C, relative expression of PrP transcripts by quantitative RT-PCR. Two HaPrP lines, moderate (M9) and strong (M6), are shown as reference. The MoPrP-P1 line induces slightly lower (nonsignificant) levels than the strong HaPrP-M6 line. The RaPrP-F22 line induces 20% more PrP transcripts than HaPrP-M6. Values were normalized with RNA polymerase II transcript levels.
no effect when introduced in mouse, including the rabbit-specific epitope L42 in position RaPrPY146 (6). More conservative substitutions, like MoPrPV214I, a residue localized in the third ␣-helix of the globular domain, prevent MoPrP conversion. It is not clear, however, how this conservative Val to Ile replacement can have such a radical effect on PrP stability. On the other hand, HaPrP has a more dramatic substitution in the same position, a charged amino acid (HaPrPT215) that does not affect its ability to convert to PrP Sc . These observations suggest that we have no clear knowledge of the role of particular residues in determining the conformational stability of these proteins. These results suggest that the sequence/structure studies need to focus on global approaches that take into consideration the role of interactions among several residues in the three-dimensional conformation of the protein.
Expression of Hamster, Mouse, and Rabbit PrP in Transgenic Flies-To examine how the primary sequence of PrP determines its ability to convert into pathogenic conformers in vivo, we created transgenic flies expressing wild type PrP from hamster (HaPrP), mouse (MoPrP), and rabbit (RaPrP) under the control of the upstream-activating sequence (UAS) promoter sequences. Then we analyzed the expression levels of each strain to select mouse and rabbit PrP strains expressed at levels comparable with the strong HaPrP line previously characterized (12). For this, we induced ubiquitous expression of all our MoPrP (5 lines) and RaPrP (13 lines) strains plus two additional MoPrP lines provided by S. Supattapone (13), prepared protein homogenates from whole flies, and resolved them in a Western blot (data not shown). Because hamsters, mice, and rabbits have unique affinities for the anti-PrP antibodies, the relative expression levels of the three proteins cannot be determined by Western blot. However, these preliminary studies showed that the electrophoretic mobility of RaPrP differed dramatically from that of HaPrP and MoPrP (Fig. 1B). These distinct electrophoretic patterns could be due to differential post-translational modifications because the lower band corresponds to the size of the full-length mature protein. Each PrP contains two facultative N-glycosylation sites (NXT) perfectly conserved at positions 180 and 196 (MoPrP numbering) that can potentially yield di-, mono-, and unglycosylated fractions. Interestingly, hamster and mouse PrP accumulated mostly mono-and unglycosylated isoforms, whereas RaPrP produced a prominent diglycosylated band. Although the mechanisms and consequences of differential PrP glycosylation are unclear, they emphasize the unique properties of RaPrP.
To overcome the difficulty of comparing the relative expression levels of the three proteins in Western blot, we quantified Prnp transcript levels for hamster, mouse, and rabbit PrP strains by quantitative PCR. For this, we used a common Taq-Man probe that hybridized with all three transcripts taking advantage of the high sequence conservation of the Prnp genes. By accurately quantifying the expression of Prnp transcripts, we selected comparable hamster (HaPrP-M6) and mouse PrP (MoPrP-P1 from S. Suppatapone) strains, and a slightly stronger rabbit PrP line (RaPrP-F22) that was the closest match available (Fig. 1C). The rest of this work describes the properties of these three strains in histological, behavioral, and biochemical assays.
Spongiform Degeneration in PrP Flies-Spongiform degeneration is the neuropathological hallmark of transmissible spongiform encephalopathies, and we have shown before that overexpression of wild type HaPrP induces prominent vacuolar degeneration of brain neurons in 30 days (12). To compare the ability of mouse and rabbit PrP to induce similar pathology, we expressed the three transgenes ubiquitously with the da-Gal4 driver and analyzed brain morphology in semithin sections in young and aged flies. Young flies expressing HaPrP were anatomically normal, indicating that HaPrP did not perturb brain development ( Fig. 2A). These flies showed well preserved architecture of the cortex, which contains the cell bodies of the brain neurons and the neuropile, the projections that occupy the center of the brain (Fig. 2B). In contrast, the brain of 30-day-old flies expressing HaPrP were smaller (reduced surface) and contained abundant vacuoles in the neuropile of the brain and optic lobes (Fig. 2C, arrows). At higher magnification, the cortex appeared very thin and contained numerous microvacuoles, suggesting significant neuronal loss (Fig. 2D, arrowheads). Next, we analyzed the brains of 30-day-old flies expressing MoPrP. These brains appeared bigger than the old HaPrP brains and showed fewer vacuoles that accumulated mostly in the optic lobes (Fig. 2E, white arrows). In these flies the cortex was larger and contained few microvacuoles (Fig. 2F, arrowheads). Finally, the brains of flies expressing RaPrP for 30 days were similar to the MoPrP brains and presented only a few vacuoles in the optic lobes and none in the brain neuropile (Fig.  2G, white arrow). The cortex of the RaPrP flies contained few microvacuoles and was similar to MoPrP brains (Fig. 2H, arrowhead). These results supported our expectations for hamster (strong degenerative changes) and rabbit PrP (no degeneration), but MoPrP produced unexpectedly weak degenerative phenotypes.
Mouse and Rabbit PrP Do Not Induce Axonal Degeneration-To further document PrP neurotoxicity in the transgenic fly models, we expressed the three PrP transgenes in a defined set of neurons involved in olfactory learning and memory that constitute the mushroom bodies under the control of OK107-Gal4 (27). The mushroom bodies are the dorsal and medial projections of Ϸ2,500 Kenyon cells, a group of tightly packed neurons in the posterior brain (Fig. 3A). The projections of the Kenyon cells extend forward and then split into one dorsal (␣) and two medial (␤ and ␥) branches to form the three mushroom body lobes on each side of the brain (Fig. 3, A and B). The mushroom body is a very robust structure that does not change during aging as shown by the thickness of the projection and distal arborizations in 40-day-old flies expressing a control transgene (lacZ) (Fig. 3B, arrow). The mushroom body also has a normal aspect in young flies expressing HaPrP (Fig. 3C), indicating that PrP does not affect mushroom body development. To compare the consequence of expressing hamster, mouse, and rabbit PrP in these neurons, we first analyzed the axonal projections in flies aged for 40 days in whole-mount brains. Expression of HaPrP in these neurons is highly neurotoxic and results in the progressive degeneration and retraction of the ␣ lobes (Fig. 3D) (12). The dynamic destruction of these projections can be seen by the active retraction of axonal terminals (blebbing) (Fig. 3D,  inset). However, older flies expressing MoPrP and RaPrP pre-sented completely normal morphology of ␣, ␤, and ␥ lobes (Fig.  3, E and F).
To support our observations on the degeneration of mushroom body projections, we also documented the effects of PrP on the Kenyon cells. Because these neurons form tightly packed clusters of 2,000ϩ neurons each, we measured the surface of the clusters as a means of estimating cell number. The size of the Kenyon cell clusters was comparable in young flies expressing either a control reporter gene (lacZ) or HaPrP (Fig. 3G), indicating that HaPrP did not interfere with the development of the Kenyon cells. Then we measured the clusters in flies aged for 40 days. Flies expressing LacZ showed an expansion of the clusters, likely due to reorganization of the cells within the clusters (Fig. 3, G and H). Older flies expressing HaPrP displayed a dramatic reduction in cluster size, indicating that these flies underwent significant cell loss (Fig. 3, G and I). Flies expressing MoPrP showed a modest but significant reduction in cluster size compared with controls, whereas flies expressing RaPrP showed a similar although not significant reduction in size (Fig. 3, G, J, and K). However, the Kenyon cell clusters in both mouse and rabbit PrP were significantly larger than in HaPrP. Overall, the ability of hamster, mouse, and rabbit PrP to induce degeneration of mushroom body neurons was consistent with our observations in whole brains.
Hamster and Mouse PrP Display Abnormal Cellular Distribution-Despite the small size of the Kenyon cells, hamster, mouse, and rabbit PrP appeared to accumulate with different cellular distribution (Fig. 3, I-K). HaPrP staining produced an uneven distribution with bright puncta (Fig. 3I), whereas MoPrP and RaPrP seemed more evenly distributed throughout the cell (Fig. 3, J  and K). At higher magnification, we confirmed the uneven distribution of HaPrP with low accumulation in the membrane and one bright spot per cell (Fig. 3L). In contrast, RaPrP showed perinuclear and cytoplasmic localization (Fig. 3N). MoPrP exhibited some puncta similar to HaPrP but also showed some diffuse cytoplasmic accumulation (Fig.  3M). Because these three proteins contain a signal peptide for secretion, they should be similarly processed and transported through the ER, Golgi, and the secretory pathway for final accumulation in the membrane through a GPI anchor. To investigate the subcellular distribution of these three proteins in more detail, we expressed the PrP constructs in larger and isolated motor neurons near the mandibular complex of the larva using the BG380-Gal4 driver. Whole-mount immunofluorescence showed HaPrP accumulation in multiple puncta throughout the cell body (Fig. 4A). Very little HaPrP was detected in the membrane, however. It is possible that the use of detergent (0.3% Triton) to permeabilize the membranes for immunofluorescence staining contributed to the extraction of membrane proteins, reducing the signal of membrane-bound PrP. To keep the membranes intact, we performed the immunofluorescence in the absence of detergents. In these conditions, HaPrP covered the membrane of the neuron and co-localized with the HRP epitope that labels the extracellular aspect of neuronal membranes (Fig. 4B). This result supports our previous detection of HaPrP in lipid rafts, a specialized compart- C, these brains are very thin and contain abundant vacuoles in the central brain (red arrow) and the optic lobes (white arrows). D, at higher magnification, the neuropile contains large vacuoles (red arrows), whereas the cortex is very thin and contains multiple microvacuoles (arrowheads). E and F, brain of 30-day-old flies expressing MoPrP. E, these brains display normal size and contain few vacuoles, mostly in the optic lobes (white arrows). F, at higher magnification, the cortex shows few microvacuoles and a larger cortex that contains more cells than the HaPrP flies. G and H, brain of 30-day-old flies expressing RaPrP. G, RaPrP brains are very similar to MoPrP; they are large and contain few vacuoles, typically in the optic lobes (white arrow). H, neuropile contains no vacuoles, whereas cortex is larger and shows few microvacuoles (arrowhead). ment of the plasma membrane relevant for PrP biology (12). Next, we detected MoPrP in the same motor neurons in permeabilized tissue and found some puncta similar to HaPrP, although less abundant, and some diffuse signals (Fig. 4C). In contrast, RaPrP showed perinuclear localization and diffuse distribution in the rest of the cytoplasm (Fig. 4D).
To pinpoint the localization of PrP to particular organelles, we co-expressed PrP constructs with reporter genes targeted to different cellular compartments in interneurons of the ventral nerve cord using OK107-Gal4. These experiments indicated the presence of HaPrP and MoPrP in the Golgi/secretory vesicle compartments by co-localization with GFP fusion proteins directed to the Golgi (galactosyltransferase-GFP, see Fig. 4E) and the late Golgi and secretory vesicles (Rab11-GFP, see Fig. 4,  F and H). Also, HaPrP did not co-localize with other organelles, like mitochondria (Fig. 4G), suggesting that HaPrP is confined to membranous structures involved in protein secretion. Additionally, MoPrP also accumulated more diffusely than HaPrP, supporting its presence in the ER (Fig. 4H, arrowhead). Finally, RaPrP displayed diffuse and homogeneous distribution through the ER and the membrane and co-localization with the ER marker KDEL (Fig. 4I).
Hamster and Mouse PrP, but Not Rabbit, Induce Locomotor Dysfunction-To investigate whether wild type PrP from hamster, mouse, and rabbit can cause neuronal dysfunction in flies, we expressed the PrP transgenes and control transgenes in motor neurons using the BG380-Gal4 driver. The motor coordination of males was assayed in climbing assays, which reflect the ability of flies to climb as a function of age (22). Control males expressing bacterial LacZ performed well over 25 days (measured at 50% climbing ability) and continued to move well for several more days (Fig. 5A). In contrast, hamster and mouse PrP induced severe locomotor dysfunction in 3-day-old flies (HaPrP) and 4-day-old flies (MoPrP) respectively (Fig. 5A). On the other hand, flies expressing RaPrP climbed well for 20 days, although not as well as control flies, and they completely lost the ability to climb by day 30 (Fig. 5A).
To further characterize the locomotor dysfunction of flies expressing PrP, we studied their movement with software-assisted video analysis. For this, we placed 20-day-old flies in empty vials and created custom arenas to match the vials (Fig.  5B). Then we tapped the vials to force the flies to the bottom and recorded the upward movement of the flies for 20 s. This procedure was repeated three times and later analyzed using the tracking software. As shown in Fig. 5B, hamster and mouse PrP flies were mostly found in the bottom of the vials at the end of the assay, whereas control and RaPrP flies occupied the top and middle of the vial, respectively. Because most of the hamster and mouse PrP flies remained in the bottom of the vial, we wanted to know how many were active, defined as the flies that move more than 5 mm/s (see "Experimental Procedures"). In the control group, 82% of the flies were active, leaving a small group of inactive flies (Fig. 5C). The HaPrP flies were by far the least active (35%), and most flies appeared paralyzed for long periods of time (Fig. 5C). Although the MoPrP flies performed almost as poorly as the HaPrP flies in the climbing assay, they showed a moderate but significant decrease in activity levels (59% active) (Fig. 5, A and C). The RaPrP flies showed a nonsignificant decrease in activity (70% active) compared with controls (Fig. 5C). Moreover, the HaPrP flies were significantly less active than flies expressing either mouse or rabbit PrP. Next, we measured the speed of the active flies. Control flies walked at a speed of 23 mm/s, reaching the top of the vial in 2-3 s. The HaPrP flies showed a dramatic reduction in average speed to 3 mm/s, whereas the flies expressing MoPrP demonstrated a more modest (still significant) decrease in their speed to 14 mm/s (Fig. 5D). In contrast, the RaPrP flies performed as well as the control flies with a speed of 22 mm/s (Fig. 5D). Furthermore, the performance of flies expressing either mouse or rabbit PrP was significantly better than those expressing HaPrP. Finally, we measured the cumulative distance traveled by the active flies in each vial. As expected, the HaPrP flies performed poorly, and the MoPrP showed a significant decrease in locomotor activity (Fig. 5D). However, the RaPrP flies traveled the same distance than control flies (Fig. 5E). The analysis of the movement of these flies demonstrated that HaPrP induced aggressive locomotor dysfunction, whereas RaPrP showed a mild dysfunction. Interestingly, MoPrP induced an intermediate locomotor phenotype that manifested more clearly in the climbing assay but did not register so strongly when other parameters were measured, like the speed of the flies.
Accumulation of Insoluble PrP in Aged Flies-Because hamster, mouse, and rabbit PrP show many differences in their pathological effects and subcellular distribution, we proceeded accumulates in a few puncta (arrows), and some protein labels diffusely the rest of the cell (arrowhead). D, RaPrP shows diffuse perinuclear accumulation, suggesting localization in the ER (arrowhead). E-I, high magnification confocal images of interneurons in the ventral cord expressing PrP under the control of OK107-Gal4. E, HaPrP (magenta) accumulates in distinct cytosolic puncta that co-localize with the Golgi marker galactosyltransferase (GalT-GFP, green, arrows); F, Golgi-secretory vesicle reporter Rab11-GFP (green, arrows); G, but not with mitochondria (mito-GFP, green). H, MoPrP (magenta) accumulates in distinct cytosolic puncta that co-localize with Rab11-GFP (green, arrows) and diffusely in the ER (arrowhead). I, RaPrP (magenta) accumulates with a diffuse distribution in the ER (arrowhead) and the membrane by co-localization with KDEL-GFP (green).
to characterize the biochemical and conformational properties of the three proteins. First, we determined the ability of sodium phosphotungstate (NaPTA) to precipitate PrP, which differentiates between native and misfolded conformations. For this, we prepared whole fly homogenates from young and old flies expressing PrP ubiquitously (da-Gal4) and incubated them in 10% Sarkosyl and 0.3% NaPTA to promote precipitation of misfolded PrP (23). Then we separated the soluble and insoluble fractions by centrifugation and resolved the fractions in Western blot. HaPrP was mostly Sarkosyl-soluble in young flies, but around 50% became insoluble in 30-day-old flies (Fig. 6A). MoPrP was mostly soluble in young flies but less than 50% was insoluble in 30-day-old flies (Fig. 6A). Finally, we determined the solubility of RaPrP and found that it was highly soluble in young flies and a small fraction was insoluble in older flies (Fig.  6A). Thus, the three proteins underwent progressive misfolding into Sarkosyl/NaPTA-insoluble isoforms.
Hamster and Mouse PrP Acquire PrP Sc -like Conformations-To further characterize the conformation of hamster, mouse, and rabbit PrP expressed in flies, we used an antibody that dis-criminates normal (PrP C ) and disease-specific (PrP Sc ) conformations. The 15B3 conformational antibody recognizes a common epitope in PrP present in the brain of several scrapieinfected animals, but it does not recognize PrP C from healthy brains (25). We have shown previously that HaPrP accumulates 15B3 epitopes in older flies (see Fig. 6B) and suggested that these conformers were responsible for the neurodegenerative changes in Drosophila (12). We next asked whether MoPrP and RaPrP could undergo the same conformational changes as HaPrP in transgenic flies. Young flies expressing MoPrP produced very weak reactivity to 15B3, but older flies accumulated more 15B3 epitopes (Fig. 6B). However, the amount of 15B3 immunoreactive MoPrP was approximately half the amount accumulated by HaPrP. Finally, neither young nor 30-day-old flies expressing RaPrP produced 15B3 immunoreactivity, supporting the hypothesis that RaPrP cannot convert into the pathogenic PrP Sc -like conformers spontaneously.
HaPrP Forms Large Aggregates-So far, RaPrP has shown a key biochemical difference with HaPrP and MoPrP, the complete absence of 15B3-positive conformers. This may explain the lack of toxicity of RaPrP in behavioral and histological analyses. If so, can we attribute the higher toxicity of HaPrP to the accumulation of more 15B3 epitopes in HaPrP flies or are there any other biochemical differences between hamster and mouse PrP? A consequence of the accumulation of misfolded conformers in animal models and human patients of prion diseases is the formation of large PrP aggregates organized into amyloid fibers. To examine the formation of high molecular weight aggregates of PrP, we performed sucrose gradients of tissue homogenates from young and old flies expressing hamster, mouse, or rabbit PrP. Sedimentation of PrP in sucrose step gradients has been used previously to determine the oligomeric state of PrP in normal and prion-infected brain tissues (26,28). After ultracentrifugation, the amount of PrP in each fraction was determined by Western blot analysis. In the samples from young flies (day 1) expressing hamster, mouse, and rabbit PrP, most PrP was recovered in fractions 1-3 (Fig. 6, C-E, top panels), representing the dominant soluble, monomeric PrP C . In the young flies, small amounts of HaPrP, but no mouse or rabbit PrP, were detected in fractions 10 -12 that contain PrP aggregates (Fig. 6C,  top panel). In the older flies from day 30, most PrP from the three constructs was still recovered in the top fractions (Fig. 6, C-E, lower  panels). Additionally, the older flies expressing HaPrP also accumulated large amounts of PrP in the high density fractions, indicating that HaPrP formed large aggregates over time (Fig. 6C, bottom panel). The levels of PrP aggregates in the elder flies expressing MoPrP were only moderate (Fig. 6D, bottom panel), whereas the elder flies expressing RaPrP revealed only traces of PrP aggregates (Fig. 6E, bottom panel).
Our findings on PrP aggregation with the sucrose step gradients were further confirmed by using gel filtration, which we used to determine the molecular size of PrP aggregates. For this, brain homogenates from young (day 1) or old (day 30) flies expressing hamster, mouse, or rabbit PrP were subjected to gel filtration, and 100 individual fractions were collected. The presence of PrP in these fractions was then determined by Western blot. Hamster, mouse, and rabbit PrP from the younger flies was recovered in fractions 56 -68 that elute molecules with an apparent molecular mass of 66 kDa or less, corresponding to small oligomeric or monomeric PrP molecules (Fig. 7, A-C, top panels).
Additionally, a small amount of HaPrP was detected in fractions 30 -36 that elute molecules with an apparent molecular mass of 2,000 kDa or greater, corresponding to large PrP aggregates (Fig. 7A, top panel). This observation replicated the presence of small amounts of HaPrP in the heavier fractions of the sucrose gradient, suggesting that HaPrP starts forming aggregates very early. Moreover, in the older flies aged for 30 days, large amounts of HaPrP were eluted in fractions 30 -38 (Fig. 7A,  bottom panel), indicating the high propensity of HaPrP to form aggregates. In contrast, older flies expressing MoPrP accumulated in modest amounts in fractions 30 -38, and RaPrP did not accumulate in significant amounts in the same fractions, supporting the low rate of aggregation of these two molecules (Fig.  7, B and C, bottom panels).

DISCUSSION
We present here the first transgenic animal model expressing RaPrP in which we have examined its conformational dynamics and its ability to induce neurotoxicity in vivo. A previous report by Priola and co-workers (6) demonstrated that RaPrP could not be converted into proteinase K-resistant PrP (PrP Sc ) in prion-infected mouse neuroblastoma cells. By expressing RaPrP in Drosophila, a PrP-free animal model, we investigated the intrinsic conformational properties conferred by its amino acid sequence by focusing on spontaneous misfolding. We have identified fundamental biochemical and structural differences between RaPrP and HaPrP that support the hypothesis that RaPrP cannot form pathogenic conformers. Thus, key amino acid differences between rabbit and hamster PrP define the conformational properties of these proteins, providing higher structural stability to RaPrP (no conversion), while conferring higher instability to HaPrP (efficient conversion). However, figuring out which residues are responsible for these differences is not trivial for two main reasons. First, there are many differences between hamster and rabbit PrP (9 in the globular domain and 22 in the mature protein), obscuring the contribution of each residue to the structure of the protein. Amino acid substitutions from the RaPrP sequence into the N-terminal unstructured domain, the globular domain and the C-terminal region containing the GPI anchor, prevent the conversion of MoPrP into PrP Sc (6,29). Second, it is unclear which interactions between amino acids may contribute to the overall stability of the protein, and therefore, single amino acid substitutions may not reveal fundamental insight about the three-dimensional structure of PrP. An example of these complex interactions was uncovered by Priola and coworkers (6) when they replaced MoPrP residues with the equivalent positions in RaPrP. Whereas one substitution (MoL108M) was enough to inhibit MoPrP conversion, a second substitution (MoN107S/L108M) was easily converted to PrP Sc . Thus, unknown intramolecular interactions have an important role in determining the structural stability of PrP, hampering our ability to understand and predict how the amino acid sequence determines the conversion potential of different PrP molecules.
Our studies with hamster, mouse, and rabbit PrP have uncovered different spontaneously occurring PrP conformers with unique biochemical properties and pathogenic potential. Young flies expressing either protein accumulate a conformer consistent with PrP C that is soluble in NaPTA/Sarkosyl, negative for 15B3 epitopes and monomeric. However, each protein produced a different conformer in older flies (Fig. 8). Hamster, mouse, and rabbit PrP accumulated NaPTA-insoluble conformers consistent with progressive PrP misfolding. Because RaPrP is not neurotoxic, it was quite surprising to find NaPTAinsoluble RaPrP in the older flies, underscoring the instability of all prion proteins. Because the 15B3 conformational antibody does not recognize insoluble RaPrP, this may be a distinct con-  Transgenic flies expressing hamster, mouse, and rabbit PrP accumulate different conformers based on their biochemical and structural properties. The three proteins accumulate initially with properties typical of PrP C . However, older flies accumulate Sarkosyl-insoluble PrP, consistent with PrP misfolding. RaPrP may accumulate in a unique misfolded conformation that is not 15B3 immunoreactive and that we called PrP insol . Both HaPrP and MoPrP produce 15B3 epitopes, indicating that they accumulate PrP Sc -like or PrP* isoforms. However, only HaPrP accumulated large amounts of high molecular weight aggregates, suggesting that this is also a unique conformer (PrP aggre ). Some of these isoforms correlate with specific degenerative phenotypes, and PrP aggre is the only conformer associated with spongiform degeneration.
former that we called PrP insol . In addition to being NaPTAinsoluble, hamster and mouse PrP accumulated 15B3-positive (PrP Sc -like) conformers. We called these conformers PrP* in our previous publication based on the similarities with PrP* or PrP Lethal described by Harris and co-workers (30) and Collinge and Clarke (31). PrP* could represent a natural conformational progression from PrP insol , although RaPrP does not make this transition. Alternatively, PrP* may represent an independent folding pathway only permitted to hamster and mouse PrP. HaPrP undergoes a further transition toward the formation of high molecular weight aggregates (PrP aggre ), whereas MoPrP only accumulates modest amounts of these aggregates (Fig. 8). The simplest explanation for the formation of PrP aggre is that PrP* is a natural precursor of PrP aggre ; however, the different amounts of PrP aggre formed by hamster and mouse PrP suggest that only a subset PrP* of conformers are consistent with the formation of large aggregates.
Although the results of RaPrP confirmed our hypothesis that it would not induce neurotoxicity, we were particularly surprised by the results produced by MoPrP. Because mice are good laboratory animals for prion studies, we expected to find similar conformational dynamics for hamster and mouse PrP in transgenic flies. However, MoPrP exhibited many key differences with HaPrP, including lower neurotoxicity (brain sections and mushroom body integrity) and limited formation of 15B3 conformers and large PrP aggregates. Interestingly, the low neurotoxicity of MoPrP agrees with a previous study by Suppatapone and co-workers (13), in which expression of wild type MoPrP in transgenic flies failed to induce locomotor dysfunction and neuronal degeneration. In contrast, we reported previously (12) and confirmed here that MoPrP induces strong locomotor dysfunction using a stronger motor neuron Gal4 strain. These conflicting results suggest that MoPrP may acquire a conformation that causes aggressive motor neuron dysfunction but not cellular degeneration; thus, different PrP conformers may be responsible for these neurotoxic outputs. The early and aggressive toxicity described in locomotor assays makes oligomers good candidates as causative agents in both hamster and mouse PrP, which can be seen in gel filtration assays (Fig. 7). A different conformer that is only produced by HaPrP, possibly high molecular weight particles of PrP* (PrP aggre ), could be responsible for the spongiform degeneration of brain neurons that progresses over several weeks.
By expressing the full-length prion proteins from hamster, mouse, and rabbit in the same transgenic system and controlling for expression levels, we have produced highly comparable results between these three proteins for the first time. Moreover, the expression of these proteins in a transgenic model animal allowed us to collect data on multiple, relevant in vivo assays, including locomotor activity, neuronal degeneration, subcellular distribution, and a battery of biochemical/structural tests. Notably, our studies generally agree with observations in purified in vitro systems describing the different conformational stability of hamster, mouse, and rabbit PrP (11,29,32,33). These experiments are typically performed with the globular domain of the prion protein, so providing in vivo context with the full-length protein is critical. Although these studies do not deal with infectious PrP Sc , we believe that our direct comparison of three prion proteins in a transgenic model provides unique information about the structural/conformational properties of each protein, particularly their ability to undergo spontaneous conformational changes over time.
But are these findings just an oddity due to the heterologous expression of hamster and mouse PrP in flies or do they uncover any relevant structural differences between hamster and mouse prions? Interestingly, hamster-adapted prions show significantly shorter incubation times (263K Ϸ90 days) than similarly aggressive mouse prions (RML Ϸ150 days), suggesting that prion conversion progresses faster in hamsters. Also notable is the fact that there are many more prion strains in mice (14) than in hamsters (5; Ref. 34). Moreover, recombinant hamster and mouse PrP have shown important differences in the misfolding and conversion mechanisms as described recently by Robinson and Pinheiro (11). First, recombinant HaPrP misfolds at much lower concentrations of urea (3.4 M) than MoPrP (6.4 M), suggesting that the native conformation of HaPrP is more unstable and easily misfolded (11,32). However, this instability does not lead to more prion strains, uncovering an inverse correlation between conversion rates and folding variants (conformers). Second, whereas HaPrP seems to form intermediate precursors, MoPrP may transition directly to the misfolded conformation. It may be possible that the distinct mechanisms of misfolding proposed by Robinson and Pinheiro (11) can lead to structurally different prions, although this remains to be confirmed in vivo. These dramatic differences occur despite the fact that hamster and mouse PrP only differ in eight residues, and their three-dimensional structure is essentially identical. Thus, the expression of PrP in transgenic flies revealed key differences in the conformational dynamics of hamster and mouse PrP that may be applicable to other species.
In addition to the conformational differences indicated by the biochemical studies, hamster, mouse, and rabbit PrP also exhibit unique subcellular distributions and glycosylation patterns. Whereas hamster and mouse PrP accumulate prominently in secretory vesicles and lack diglycosylated protein, RaPrP accumulates diffusely in the ER and the secretory pathway and presents a predominant diglycosylated band. One way to explain this connection between PrP distribution and glycosylation is that hamster and mouse PrP may misfold rapidly in the secretory compartments following their synthesis in the ER, making the second glycosylation site inaccessible. If this is true, the lack of diglycosylated PrP may uncover clues about the early conformational changes of hamster and mouse PrP. However, the abnormal distribution does not explain the neurotoxicity by itself because PrP conversion can take place in different compartments, including the secretory pathway, although the lipid raft compartment of the membrane seems to play a prominent role in PrP biology (35). Alternatively, the oligosaccharide chains may modulate the conformational dynamics of PrP. Several pieces of evidence suggest that less glycan chains favor the formation of PrP Sc (36). First, an inherited prion disorder in a Brazilian family is associated with a mutation in PrP that eliminates one glycosylation site (T183A) (37). Second, PrP Sc is produced more rapidly when scrapie-infected cells are treated with tunicamycin, an inhibitor of N-glycosylation (38). Third, unglycosylated PrP seems to be a better substrate for in vitro con-version (39). Fourth, PrP mutants that lack one or two glycosylation sites efficiently convert into proteinase K-resistant PrP Sc in Chinese hamster ovary cells (36). However, conversion is not observed in transgenic mice expressing unglycosylated PrP, indicating important differences in the behavior of the protein in vivo. Although recent results demonstrate that prion transmission can take place regardless of PrP glycosylation (40), it seems to be important for the transmission of certain strains (41). Thus, mono-and unglycosylated PrP seem to be better substrates for conversion than diglycosylated PrP. Therefore, the high levels of the diglycosylated band in RaPrP may have a protective role. Our observations suggest that PrP misfolding and glycosylation are mutually regulated; early PrP misfolding (HaPrP, MoPrP) may hide the second glycosylation site, thus favoring further conformational changes during aging. On the other hand, the initial stability of PrP may favor the incorporation of more oligosaccharides, which in turn prevents PrP misfolding during aging. Overall, these studies demonstrate the benefits of Drosophila models expressing mammalian PrP to better understand the dynamics of sporadic PrP conversion in vivo.